1. Top of page
  2. Abstract
  3. Methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References

When the northern fulmar expanded its northeast Atlantic breeding range from the two known colonies, Grimsey in northern Iceland and St Kilda in the Outer Hebrides, Scotland, about 350 yr ago, the geographical pattern of colonisation – initially the Faroes, then Scotland, followed by Ireland and southern Britain – led James Fisher to propose a sole Icelandic source for the colonists. However, previously-analysed mitochondrial DNA from contemporary samples indicated a St Kildan origin for at least some colonists. If Fisher's hypothesis is correct and Iceland and not St Kilda was the source population for all of the new colonies, the Icelandic signal should be stronger in museum samples collected 100 yr ago when St Kilda was populated by people who harvested large numbers of fulmars. Patterns of genetic, specifically, nucleotide, diversity suggest an Icelandic origin for the pre-1940 samples. St Kilda birds contained a number of closely related haplotypes whereas Grimsey, Iceland, the other putative source population, contained diverse haplotypes. These two patterns are indicative of a younger and older population, respectively. When both nuclear aldolase and mitochondrial control region sequence data from historical samples collected on the newly colonized islands were examined, they contained highly divergent haplotypes characteristic of Grimsey, not St Kilda. Comparison of mitochondrial data from samples collected in the early and late 20th century showed an interesting pattern of haplotype turnover on St Kilda. Prior to 1940 the haplotypes present on St Kilda were genetically similar to one another, yet haplotype sampling in the 1990s showed highly divergent haplotypes on the island. We propose that these new haplotypes are not the result of mutation, but immigration from other colonies in the North Atlantic.

The northern fulmar Fulmarus glacialis, a medium-sized seabird with a distinctive tube-nose and opportunistic feeding habits (Hatch and Nettleship 1998), is a pelagic petrel that moults and over-winters in the open ocean and roams widely to forage during the breeding season. Today the species is ubiquitous across the North Atlantic, its range having expanded rapidly between the mid 18th century and the late 20th century (Fisher 1952a). Fisher (1952b) hypothesized the range expansion occurred in response to the increasing availability of scraps from the whaling industry and then of offal from expanding commercial fisheries. Alternatively, Salomonsen (1965) and Brown (1970) thought it was due to changes in oceanographic conditions.

The first record of an Atlantic breeding colony dates to the 1640s when nesting was recorded on Grimsey, an island north of Iceland (Fisher 1952a). The only other documented colony in the 17th century was on St Kilda in the Outer Hebrides, Scotland (Fisher and Waterston 1941). The large northern fulmar breeding population on St Kilda, which local residents harvested annually until 1930, remained stable throughout the initial range expansion and until about 1970 (Fisher 1952a, Lloyd et al. 1991). It then grew over the next 20 yr but has changed little since 1990 (Lloyd et al. 1991, Mitchell et al. 2004).

The historic spread of the northern fulmar to new colonies on Iceland and the eastern Atlantic has been described in meticulous detail (Fig. 1, Fisher 1952a). Following colonisation of the Faroe Islands between 1816 and 1839, the fulmar's range expanded to the British Isles and then to northwestern Europe. The path of expansion through the British Isles has been well documented, starting on Foula, Shetland Islands, followed by other sites in the Northern Isles, and then, more or less contemporaneously, sites bordering the Irish Sea to the west and the North Sea to the east. This geographical pattern led Fisher (1952a) to propose that, at least in the earlier stages, the expansion was driven by birds emigrating from Iceland colonizing islands as they moved in a southward direction.


Figure 1. Samples analyzed in this study and Burg et al. (2003) were obtained at the sites indicated by squares, while the two putative source populations of St Kilda and Grimsey are indicated by stars. Sites providing only contemporary samples (Burg et al. 2003) are indicated as white squares, sites providing both historical and contemporary samples are shown as grey squares, and sites providing only historical samples are shown as black squares. Sampling sites for the historic fulmars include St Kilda (StK), Fair Isle (FI), Orkney (Ork), Western Isles (WI), Coquet (Coq), Shetland Islands (Shet), Faroes (Far) and three sites in Iceland (Ice-N, Ice-S, Ice-G) including Grimsey. Contemporary only sampling sites in the North Atlantic include Ailsa (Ail), Ireland (Ire), southern Norway (Nor), Svalbard (Sval) and one site in Iceland (Ice-SW) and three sites in the Canadian Arctic: Prince Leopold (PL), Devin Island (DI) and Baffin Island (BI). Nor, Sval, PL, DI and BI sites from Kerr and Dove (2013) are not shown on the map. Dates beside Grimsey and St Kilda indicate the year since which occupation has been continuous. The number beside other sampling sites indicates the approximate date of colonisation (Fisher 1952b). Note that, although the Icelandic sampling site of Hvalfjordur (Ice-SW) was colonised in 1930, the Icelandic population expansion and colonisation of nearby sites started no later than 1753. Modified from Burg et al. (2003).

Download figure to PowerPoint

To date, two studies have examined phylogeography of northern fulmars (Burg et al. 2003, Kerr and Dove 2013) and both showed little genetic differentiation among contemporary populations in the North Atlantic. Burg et al. (2003) used mitochondrial DNA (mtDNA) data from modern birds sampled from Iceland, Faroe Islands, Ireland and the British Isles in the late 1990s in an attempt to determine if birds from both Iceland and St Kilda or from Iceland alone contributed to the range expansion. The results from highly variable control region sequences indicated both St Kilda and Iceland contributed to the expansion, in contrast to Fisher's (1952a) scenario. However, pairwise comparisons within Burg et al.'s study did support one aspect of Fisher's hypothesis; the colonisation of the British Isles occurred in a “stepping stone” pattern, proceeding from one island to the next (Wright 1951).

The failure of modern samples to convincingly support Fisher's ‘out-of-Iceland’ hypothesis raised the possibility that any Icelandic signal might have been obscured by emigration of birds from St Kilda in the second half of the 20th century. If so, then samples obtained in the late 19th and early 20th century might reveal a different picture. At this time islanders still harvested about half the chicks, around 6000–10 000, reared on St Kilda each year (Lloyd et al. 1991). To test this idea we sampled fulmars in museum collections from Iceland, Faroes and Great Britain that had been collected between 1868 and 1939, approximately 30–100 yr before the period of population expansion on St Kilda (Lloyd et al. 1991). We hypothesised that these samples would show a genetic signature consistent with the out-of-Iceland hypothesis.


  1. Top of page
  2. Abstract
  3. Methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References

DNA samples

Toe pads collected between 1868 and 1939 were obtained from 73 specimens held in the Natural History Museum (Tring), Edinburgh Museum, Norwich Museum, Manchester Museum, National Museum (Liverpool) and the Copenhagen Museum. These samples represent most of those that are available during that time period. Samples were from ten sites (defined as an island or archipelago) (Fig. 1), four of which were also sampled by Burg et al. (2003). The samples were collected during the breeding season: May–September. One sampling site is the island of Grimsey, the original Icelandic site, whose birds were not sampled as part of Burg et al.‘s (2003) study of modern fulmars.

All molecular work, including the DNA extractions, was performed in a lab where no previous fulmar work had been done. DNA was extracted using a QIAGEN DNeasy kit following the manufacturer's protocols.

DNA amplification

Duplicate copies of the control region have been reported in several seabird species (Abbott et al. 2005, Eda et al. 2010, Morris-Pocock et al. 2010, Rains et al. 2011). We designed primers specific to each of the two copies (F1: complete control region and F2: partial duplication) to amplify a single copy. A 551 base pair (bp) fragment of the control region was amplified using 100 μM dNTP, 2 mM MgCl2, 0.4 μM GluF6 and H505 primers (Supplementary Material Appendix 1, Table A1), and 1 U GoTaq Flexi DNA Polymerase (Promega) in a 25 μl reaction. Amplification consisted of one cycle at 94°C for 120 s, 45°C for 30 s and 72°C for 20 s; 31 cycles of 94°C for 45 s, 45°C for 30 s and 72°C for 20 s; and one final cycle of 94°C for 45 s, 45°C for 30 s and 72°C for 5 min. Samples were sequenced using ABI chemistry on a ABI3130xl and the resulting sequences were aligned by eye in MEGA 4.0 (Tamura et al. 2007).

We were able to sequence portions of the F1 and F2 mtDNA control region similar to those reported in other seabirds, where the F1 and F2 copies produce similar trees and evolve in concert (Dover 1982, Abbott et al. 2005, Eda et al. 2010, Morris-Pocock et al. 2010). The F1 and F2 copies diverged for 65 bp at the start of the control region, but were otherwise similar (Supplementary material Appendix 1, Table A2). The divergent portion, starting at position 300 in Burg et al. (2003), contains 26 substitutions including two indels, at positions 305 and 348, and 10 transversions. The divergent portion was not included in Burg et al. (2003); however, alignment with Coquet2 from Burg et al. (2003) is shown for reference. While we were able to align the fulmar sequences to the albatross, Thalassarche species, from Abbott et al. (2005), it was not possible to determine which of the two copies in fulmars corresponded to the F1 and F2 copies in albatross. As such the F1 and F2 in fulmars were arbitrarily named. A series of primers was designed to amplify a single region of the mitochondrial control region (Supplementary material Appendix 1, Table A1) based on the F1/F2 control region data.

Due to the degraded nature of the DNA, semi-nested polymerase chain reactions (PCR) were performed to amplify one large fragment or two smaller, overlapping fragments. DNA was amplified in a 25 μl reaction using the same conditions as the GluF6/H505 PCR above with 1.5 mM MgCl2 and an annealing temperature of 50°C. The first round of PCR used FgF2L and FgH450 primers (Supplementary material Appendix 1, Table A1) and the semi-nested PCR used diluted PCR product from round one as template and primers FgF2L and FgH395. Individuals which failed to amplify as one large PCR product were amplified as two smaller fragments using the same protocol with the exception of primer pairs. Either Fg2L or H505 were used along with one of the internal primers and two smaller, overlapping fragments were amplified (Supplementary material Appendix 1, Table A1).

The aldolase B, fructose-bisphosphate locus on the Z chromosome was amplified using AldB.6F and AldB.8R primers (Hackett et al. 2008). We designed fulmar specific primers to amplify the aldolase gene using a semi-nested PCR (Supplementary material Appendix 1, Table A1) and resulting sequence identity confirmed against seabird sequences in GenBank (Thalassarche albatross and Sula booby). DNA was amplified in a 25 μl reaction using the same conditions as the GluF6/H505 PCR above with an annealing temperature of 52°C.

PCR products were sequenced on an ABI 3130xl and sequences were visually aligned using MEGA 4.0 (Tamura et al. 2007). All variable sites were confirmed by visual inspection. A minimum of nine individuals was sequenced two or more times across the entire fragment to assess the accuracy of the sequences and another 55 were sequenced in two fragments with more than 100 base pair overlap; identical results were obtained in both instances.

Genetic data analyses

The 73 historical control region sequences from 10 sites (Coquet, Fair Isle, Faroes, Grimsey, northwestern Iceland, southern Iceland, Orkneys, Shetlands, St Kilda and Western Isles) were analyzed both as the long fragment (346 base pairs) and were then shortened to 299 base pairs to allow direct comparison with the 189 contemporary sequences from 12 populations (Atlantic: Ailsa, Coquet, Faroes, southwestern Iceland, Ireland, Norway, Shetlands, Svalbard and St Kilda; Canadian Arctic: Baffin Island, Devon Island, and Prince Leopold Island) analyzed in Burg et al. (2003) and Kerr and Dove (2013). Four of the sampling sites (Coquet, Faroes, Shetlands and St Kilda) were the same in the current study and the 2003 study while the location of samples from Iceland differed slightly between the two studies. For the nuclear data, only historical samples were sequenced. Each locus was analyzed separately.

As the aldolase gene is sex-linked, we reconstructed the alleles for the nuclear locus using default settings (100 iterations, thinning interval of 1 and 100 burn-in) of PHASE in DnaSP 5.10.01 (Librado and Rozas 2009) for all known males. Data on the sex of the birds was obtained from the museums who provided the samples. A statistical parsimony network was constructed using TCS 1.21 (Clement et al. 2000) for the two datasets: phased nuclear dataset and the mtDNA data. Allelic diversity (ad) was calculated following Nei (1987). Nucleotide diversity (π) and pairwise population divergence estimates, namely FST, were calculated in Arlequin 3.11 (Excoffier et al. 2005) and significance tested using 10 000 permutations. Geographic distance (shortest distance over water) and genetic distance were compared in all-against-all isolation by distance (IBD) analyses and as distance from a putative source population (e.g. St Kilda, Ice-G).

Relative migration rate estimates were calculated with the aldolase data using the isolation with migration model as implemented in IMa ver. 3.5 (Hey and Nielsen 2007). We grouped samples into Iceland (Ice-N and Ice-G), St Kilda and all remaining sites. Multiple initial runs were performed to identify upper parameter bounds and an ideal heating scheme. Three final runs were performed for each comparison using the HKY model, an inheritance scalar of 0.75, 1 000 000 burn-in and 5 000 00 to 10 000 000 post burn-in MCMC steps, geometric heating, and 30 chains with 30 chain swaps between steps.

Fu's FS and Fu and Li's F* and D* were calculated to test for deviations from neutral evolution resulting from selective sweeps, background selection or population expansion (Fu 1997). Neutrality tests were performed in Arlequin ver. 3.11 (Fu's FS, 10 000 permutations) and DnaSP ver. 5.10.01 (Fu and Li's F* and D*) (Excoffier et al. 2005, Librado and Rozas 2009). Fu's FS has more statistical power to detect expansion and selective sweeps (Fu 1997) whereas Fu and Li's F* and D* are better at detecting background selection. When FS is significant and F* and D* are not, the deviation from neutrality is the result of expansion and not background selection.


  1. Top of page
  2. Abstract
  3. Methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References


Control region sequences from the museum samples contained 21 variable sites defining 25 haplotypes (Table 2, Supplementary material Appendix 1, Table A3). Among the museum samples, 13 haplotypes were represented once, and 12 were found in two or more individuals (Table 2).

Table 2. Geographic distribution of mtDNA haplotypes, sample size (n) and number of haplotypes (hn) among museum fulmars. Data from contemporary samples are shown in brackets.
Unique(3)(3)(6) (5)122 (2)(4)12 (2)(2)(5)(2) (3)8 (37)
A(3)(1)(2) (3)134 (7)(6)22 4 (7)(1) (4)(3)(4)16 (41)
B (1)  (1)  3 (2)(2)11 (2) (1)(3) (1)5 (13)
C       (2)(1)   (3) (1)(1)  (8)
D   1   (2)     (1)    1 (3)
E       (2)    2 (2)(1)(1)   2 (6)
F       (1)(1)     (1)   (3)
G    (1)  (1)     (1) (2)  (5)
H (1)     (2)1       (1) 1 (4)
I       (1)      (2)   (3)
J   1 2 1 (2)(1) 3  (1) (3)  7 (7)
K       (2)  1    (1)  1 (3)
L    (1)  (5)(3)   (1)(1)  (3) (14)
M    (1)   (2)         (3)
N(1)      (2)    2     2 (3)
O  (3) (2)113 (2)4   2(1)(1) (1)(5)11 (15)
P              (2)(1)  (3)
Q   1(1) 1       (1)   2 (2)
R     11(1)      (1)(1)(1) 2 (4)
S              (2)   (2)
T            2    (1)2 (1)
U        1      (1)(1)(3)1 (5)
V   1        2     3
X   1   1   12 1   6
Y    (1) 1   1       2 (1)
Z    (1) 1           1 (1)
AA  (1)    (1)          (2)
n(7)(6)(12)5(17)61014 (37)6 (20)410116 (17)(9)1 (18)(19)(10)(17)73 (189)
hn(5)(6)(9)5(14)577 (18)3 (11)3717 (7)(9)1 (15)(11)(6)(8)25 (61)

To explore geographical structure in the museum samples, pairwise FST values were calculated. Five were significant (Table 1); however, three of the significant differences may be due to the small sample size from Shetland (n = 6). Similarly, differences may go undetected due to small sample sizes and high levels of variation (e.g. Ice-N and Ice-S). The absence of geographical clustering is also apparent in the statistical parsimony network of the pre-1940 samples (Fig. 2). As with the contemporary samples, haplotypes A (22%) and O (15%) were the most common (Table 2).

Table 1. Pairwise FST values for historical samples collected during the breeding season for mtDNA (a) and aldolase (b). FST (lower left) and p values (upper right). Only populations with n > 5 were analyzed (see Fig. 1 for site abbreviations). Significant values after FDR (false discovery rate) correction are in bold and values no longer significant after FDR are italicized.
(a) mtDNA
Ice-N 0.5870.7400.9030.4440.3480.381
Ice-S−0.389 0.4550.3240.4680.8580.117
Ice-G−0.0510.000 0.264 0.032 0.1640.425
Far−0.0880.0180.019 0.224 0.044 0.194
Shet0.0180.015 0.138 0.035  0.028 0.023
Ork0.012−0.0640.056 0.102 0.151   0.003
StK0.0030.6780.0010.027 0.166 0.174  
(b) aldolase
Ice-N  0.028 0.3390.1420.1390.2720.202
Ice-G 0.164   0.008 0.039 0.008 0.033 0.001
Far0.013 0.123  0.4350.3160.0730.107
FI0.100 0.115 0.001 0.2790.1200.087
Ork0.081 0.189 0.0160.028 0.0690.240
Shet0.133 0.155 0.1170.1730.190 0.062
StK0.040 0.196 0.0280.0540.0160.129 

Figure 2. Statistical parsimony network using 299 bp mtDNA control region sequence data from historical samples only (n = 73). Individuals are represented by single squares and haplotypes numbered following Table 2 and Supplementary material Appendix 1, Table A3. Haplotype in the inset are combined into three main groups: Iceland, St Kilda or new colonies. The size of the circle is proportional to the number of individuals.

Download figure to PowerPoint

St Kilda samples pre-1900 and 1900–1939 were compared, to check whether the older samples lacked haplotypes that might have been introduced by immigrants in the early 20th century when the expansion was fully underway. Among the six pre-1900 samples, four haplotypes were present (A, O, T and X), of which two, haplotypes O and T, were not found on St Kilda after 1900, though haplotype O was present in birds from other sites. One of the other six pre-1900 St Kilda birds captured outside the breeding season also had haplotype T (data not shown). Haplotype T was not found at any other historic sampling site. Samples collected at St Kilda from 1900–1939 (n = 10) represent five shared haplotypes (A, E, N, V and X). Thus three haplotypes, E, N and V, were recorded for the first time on St Kilda in the first half of the 20th century, but two of these, N and V, are not represented in the contemporary samples from St Kilda.

Combining the pre-1940 samples with the previously published sequences (Burg et al. 2013, Kerr and Dove 2013) and including eight additional modern Coquet sequences revealed 45 unique haplotypes, 26 shared haplotypes and 53 variable sites (Fig. 3, Table 2, Supplementary material Appendix 1, Fig. A1, Table A3). Some of the unique haplotypes from Burg et al. (2003) and Kerr and Dove (2013) are now shared haplotypes. The main differences in haplotype frequencies concern haplotype O which was more common prior to 1940 (15 vs 8%) and haplotype L which was absent in the museum samples (0 vs 7%). Two shared haplotypes (V and X) were present only in the pre-1940 samples. Together they constitute 4/16 (25%) of St Kilda samples and are found mainly in St Kilda birds (67 and 33% respectively). Haplotype C, previously found in three (18%) modern St Kilda birds and other sites (Coquet, Faroe, Ireland) was absent in the pre-1940 samples. The TCS network (Fig. 3) depicts the relationship amongst all sampled mitochondrial haplotypes revealing few inferred/unsampled haplotypes despite high diversity.


Figure 3. Statistical parsimony network using 299 bp mtDNA control region sequence data from historical (red outline) and contemporary (black outline) samples (n = 262). Most contemporary samples, see text, are from Burg et al. (2003) and Kerr and Dove (2013). The size of the circle is proportional to the number of individuals. Haplotypes are numbered following Table 2 and Supplementary material Appendix 1, Table A3. Network in Supplementary material Appendix 1, Fig. A1 contains geographic locations for each sample.

Download figure to PowerPoint

MtDNA data for all sampling sites and from St Kilda only showed no evidence of IBD (p = 0.85 and 0.92, respectively) while IBD from Iceland showed a strong, but not significant, positive correlation (p = 0.19). Historical samples from Iceland have a significant FS and corresponding non-significant F* and D* indicating population growth, not background selection (Table 3). Orkney, St Kilda and Faroe populations have near-significant (p = 0.06–0.10) and strongly negative FS values. Seven of the contemporary populations show the same pattern of population growth with FS, F* and D*.

Table 3. Tests for neutral evolution, mismatch distribution, sample size (n) and mtDNA haplotype diversity. Significant values are in bold. Values for Fu and Li's F* or D* were not significant (values reported as > 0.1, no actual p-value given). Historical samples were also obtained from Coquet (n = 1), Fair Isle (n = 4) and Western Isles (n = 1), however, these were not included in the tests for neutral evolution or mismatch due to the small sample size.
Fu's FS −3.29 −1.90 −1.33 −1.790.30−2.02−2.29
p-value 0.02 0.05 0.01 0.100.490.070.06
Fu and Li's F*−0.402−0.461−0.174−1.012−1.396−1.1420.980
Fu and Li's D*−0.232−0.440−0.138−0.895−1.325−1.1381.272
Fu's FS−1.77 −3.71 −5.03 −9.87 −8.88 −4.56 −0.33 −7.39 −8.09 −2.64−1.99−1.81
p-value0.12 0.024 0.006 < 0.001 < 0.001 0.01 0.45 < 0.001 < 0.001
Fu and Li's F*−0.449−1.071−2.259−1.937−1.042−2.3280.081−0.667−1.131−0.146−0.436−1.134
Fu and Li's D*−0.370−0.995−2.404−1.735−0.844−2.1340.190−0.520−1.0200.0007−0.368−1.058


The aldolase sequence is perhaps the better of the two markers in terms of containing moderate levels of variation. A total of 37 fulmars were sequenced from seven sites and 26 alleles were present among the phased data (n = 65 Z chromosomes). Allelic diversity was highest on Ice-G (0.929), one of the original colonies, and ranged from 0.644–0.909 at other colonies (Table 4). Little geographic clustering of haplotypes is evident in the TCS network (Fig. 4). Two alleles (alleles a and b) were shared among 58% of the sampled individuals and were present at more than one site (Fig. 4). Nucleotide diversity was lower in the aldolase dataset (0.002–0.010) than in the mtDNA. Prior to FDR (false discovery rate) corrections, six pairwise FST values were significant and another three (Shet-Ork, Shet-StK, Shet-Far) were marginally significant (p = 0.062–0.070). Three comparisons remained significant following FDR correction and all involved comparisons with Ice-G and other populations (Table 1).

Table 4. Geographic distribution of shared aldolase alleles in the historical samples. Summary statistics: number of Z chromosomes sampled (nz), number of alleles detected (an), allelic diversity (ad), and nucleotide diversity (π) and results from Fu's and Fu and Li's statistics.
a2 4346322
c      22
d   2   2
e 2     2
unique246 14421
Fu's FS−0.56 −2.20 −3.10 1.99−0.05 −2.74 −2.50  
p-value0.227 0.034 0.018 0.8310.383 0.008 0.035  
Fu and Li's F*1.4890.0470.5951.622−0.177−0.312−1.466 
Fu and Li's D*1.4670.0710.7211.551−0.28−0.135−1.246 

Figure 4. Statistical parsimony network using 356 bp aldolase sequence data from 37 fulmars collected prior between 1868 and 1939. Data were phased using DnaSP and the 65 alleles were used to create the network. Individuals are represented by single squares and alleles numbered following Table 4 and Supplementary material Appendix 1, Table A4.

Download figure to PowerPoint

None of the IBD analyses were significant (p = 0.34 (all), 0.25 (St Kilda), 0.33 (Iceland) (graphs not shown)). Marginal distribution curves for IMa were unimodal and approached 0 on each side. In each case ESS estimates were > 100 (often > 10 000). Average parameter estimates were similar in both directions for the comparison between Iceland and the other populations (m1 = 0.01 and m2 = 0.02), but showed increased gene flow into St Kilda (Ice vs St Kilda: m1 = 0.037 and m2 = 0.197; the rest vs St Kilda: m1 = 0.01 and m2 = 0.083). Grimsey (Ice-G), Orkney, St Kilda and Faroe Islands show significant deviation from neutral evolution. Fu's FS and F* and D* show these deviations are consistent with population growth (Table 4).


  1. Top of page
  2. Abstract
  3. Methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References

We studied the patterns of genetic diversity in fulmar samples collected from breeding sites during the 1868–1939 breeding seasons in an attempt to determine if there was a genetic signature consistent with a single initial range expansion from Iceland. The pattern is not clear-cut and the data indicate three possibilities: 1) the observed high levels of mtDNA variation necessitate larger sample sizes to detect a convincing Icelandic signal; 2) St Kilda and Iceland both contributed to the initial expansion; or 3) Icelandic birds dispersed to St Kilda and other North Atlantic colonies prior to the main period of expansion. The three options are not mutually exclusive.

While the number of historical samples is limited, the fact that many of the haplotypes are found in several individuals, within the same time period and between periods, suggests that sample size is adequate to assess levels of variation present at the time. If it was not, then many more haplotypes would have been represented by single individuals. However, rare haplotypes are likely to be underrepresented due to the sample size. Sample size is more of a concern with the mtDNA data where higher levels of variation are present and this is supported by rarefaction curves where populations with less than seven samples and Grimsey (n = 10) have not levelled off (data not shown). For the less variable aldolase gene, the curves plateau much sooner, with the exception of Grimsey.

We now address the issue of gene flow between Iceland and St Kilda. Determining patterns of gene flow is difficult given the high diversity and complex haplotype network, particularly for the mtDNA data (Fig. 3). Historically St Kilda had unique mitochondrial haplotypes which are absent in contemporary samples and haplotypes that are now found at other sites (e.g. haplotype E), supporting the idea that gene flow was occurring between St Kilda and other colonies after the 1930s. The two main aldolase alleles (a and b) are also present on both St Kilda and Iceland; however, the fact that none of the remaining 24 alleles are present at multiple sites limits our conclusions. That alleles from historical samples are shared between Iceland and St Kilda for both markers suggest population connectivity between the two sites, but in what direction and when? Aldolase data show evidence of asymmetrical migration between Iceland and St Kilda. The levels of immigration to St Kilda are higher than emigration rates answering the question of directionality, but not the timing.

Historically haplotype diversity on St Kilda was comparable to levels found at some of the old colonies (e.g. Grimsey and Faroes); however, both mitochondrial and nuclear data show reduced nucleotide diversity on St Kilda relative to other colonies. Control region haplotypes on St Kilda are all separated by a single nucleotide substitution from haplotypes A and O (Fig. 2), which themselves are separated by a single mutation; and all but one of the seven aldolase alleles (Fig. 4) are a single step from allele a. This historical pattern is unique to St Kilda. In contrast, contemporary levels of genetic diversity on St Kilda are different: mtDNA haplotype diversity is much lower having decreased from 0.84 to 0.76 and nucleotide diversity increased (0.0056 to 0.0102). The divergent mtDNA haplotypes could not have been the result of mutation in such a short period of time and, as historical and contemporary sample sizes are similar (16 and 17 birds, respectively), sampling artefacts are unlikely to explain the difference in either diversity measure. One possible explanation is a selective sweep, yet this is not supported by the current data (Table 3). Alternatively if the population declined, haplotypes would have been lost (i.e. haplotype diversity reduced) through genetic drift and the loss would have been random. However, this still leaves the unanswered question of where did the divergent mtDNA haplotypes come from? As mentioned earlier, the historical and contemporary mtDNA haplotypes on St Kilda show some differences. Of the 12 mitochondrial haplotypes detected at any time on St Kilda, only two appear in both time series (haplotypes A and E). While some turnover in haplotypes occurred on the Faroe Islands, the only other site with comparable samples sizes in the two sampling periods, temporal differences in haplotypes on the Faroes are mainly due to the presence of eleven new haplotypes in the contemporary samples. In contrast, only two of the historical haplotypes sampled on St Kilda remain in the more contemporary samples. New, more divergent haplotypes (e.g. B and C) have appeared on St Kilda suggesting immigration to St Kilda. Immigration to St Kilda is supported by IMa results for the aldolase data.

The absence of long term isolation and presence of gene flow between St Kilda and other populations prior to the 1940s is evident in pairwise FST (Table 1) and the statistical parsimony network (Fig. 2). Prior to the 1940s, only two populations showed significant differences to St Kilda (Grimsey and Orkney) and each for a single locus.

Among the historical samples, a signal for population growth was detected in Iceland, Faroes, Orkney and St Kilda (Table 3, 4). Differences in marker variation and sample size may explain some of the differences between the aldolase and control region datasets. However, with the exception of northern Iceland, the two datasets show similar trends, with older colonies (Faroes, Iceland and St Kilda) exhibiting stronger and newer colonies weaker signals of population growth. These differences in the signal for historical population growth may reflect the amount of time since the colony was founded. A northern gannet Morus bassanus colony took almost 50 yr after founding for the colony to start growing (Siorat and Rocamora 1995). Fisher and Waterston's (1941) detailed study of the fulmar expansion shows some colonies took 15–30 yr before experiencing exponential growth while others remained small for much longer periods. Part of this lag time between founding and rapid increase in growth can be attributed to colony size and breeding success with larger colonies being more successful (Fisher and Waterston 1941, Fisher 1952b). Thus, while the long-established Icelandic colonies were growing, the more recently established colonies on the Northern Isles had perhaps not yet embarked on the phase of rapid growth. At the time the historical samples were collected, Faroe (1879–1928), Shetland (1905–1939) and Orkney (1906–1934) had been colonised for 40–112, 27–61 and 6–34 yr respectively.

Moving forward to the contemporary samples, the majority of the sites show population growth (Table 3), the exceptions being Ireland, a newer colony, and St Kilda. The fact that the genetic data do not show population growth at St Kilda at a time when independent colony counts indicate growth (1970–1990) suggests that the genetic tests are conservative. One interesting finding concerns changes in haplotypes. Haplotype diversity increased at both Faroe and Shetland. If migration rates fluctuated over time or birds were dispersing from islands with large colonies (e.g. Iceland) after 1940, this could explain these patterns.

The historical samples have shed light on the history of the fulmar expansion, but have not provided a conclusive answer to the out-of-Iceland hypothesis. In support, strong signals of population expansion and high haplotype/allelic and nucleotide diversity are evident in the historical samples from Iceland and indicative of an older population. In contrast, nucleotide diversity in historical samples from St Kilda is much lower, a pattern indicative of a relatively young, but isolated population. This is not to suggest St Kilda was colonized as recently as the other north Atlantic sites. Rather the colony may not be as old as the population on Grimsey, and the diverse haplotypes found on the newer colonies likely did not originate from St Kilda. Two other pieces of evidence support the out-of-Iceland hypothesis. First, the change in genetic diversity at St Kilda, where more divergent haplotypes are now present, suggests immigration to St Kilda. And second, haplotypes found at newer colonies are highly divergent and not characteristic of historical haplotypes present on St Kilda, but are typical of those from Iceland. However, we cannot rule out the possibility of emigration from St Kilda to the surrounding colonies.


  1. Top of page
  2. Abstract
  3. Methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References

The authors would like to thank Curators of Natural History Museum (Tring), National Museums Scotland (Edinburgh), Norwich Museum, Manchester Museum, National Museum (Liverpool) and the Copenhagen Museum for access to their specimens (Supplementary material Appendix 1, Table A5). We also thank Júlio Neto for constructive comments.


  1. Top of page
  2. Abstract
  3. Methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References
  • Abbott C. L. and Double M. C . 2003 . Phylogeography of shy and white-capped albatrosses in ferred from mitochondrial DNA sequences: implications for population history and taxonomy . – Mol. Ecol. 12 : 27472758 .
  • Abbott C. L. , Double M. C. , Trueman J. W. H. , Robinson A. and Cockburn A . 2005 . An unusual source of apparent mitochondrial heteroplasmy: duplicate mitochondrial regions in Thalassarche albatrosses . – Mol. Ecol. 14 : 36053613 .
  • Brown R. G. B . 1970 . Fulmar distribution, a Canadian perspective . – Ibis 112 : 4451 .
  • Burg T. M. , Lomax J. , Almond R. , Brooke M. de L. and Amos W . 2003 . Unraveling dispersal patterns in an expanding population of a highly mobile seabird, the northern fulmar (Fulmarus glacialis) . – Proc. R. Soc. B 270 : 979984 .
  • Clement M. , Posada D. and Crandall K. A . 2000 . TCS: a computer program to estimate gene genealogies . – Mol. Ecol. 9 : 16571659 .
  • Dover G . 1982 . Molecular drive: a cohesive mode of species evolution . – Nature 299 : 111117 .
  • Eda M. , Kuro-O M. , Higuchi H. , Hasegawa H. and Koike H . 2010 . Mosaic gene conversion after a tandem duplication of mtDNA sequence in Diomedeidae (albatrosses) . – Genes Genet. Syst. 85 : 129139 .
  • Excoffier L. , Laval G. and Schneider C . 2005 . Arlequin ver. 3.0: an integrated software package for population genetics data analysis . – Evol. Bioinform. Online 1 : 4750 .
  • Fisher J . 1952a . The fulmar . – Collins .
  • Fisher J . 1952b . A history of the fulmar Fulmarus and its population problems . – Ibis 94 : 334354 .
  • Fisher J. and Waterston G . 1941 . The breeding distribution, history and population of the fulmar (Fulmarus glacialis) in the British Isles . – J. Anim. Ecol. 10 : 204272 .
  • Fu Y.-X. 1997 . Statistical tests of neutrality of mutations . – Genetics 133 : 693709 .
  • Hackett S. J. , Kimball R. T. , Reddy S. , Bowie R. C. K. , Braun E. L. , Braun M. J. , Chojnowski J. L. , Cox W. A. , Han K. L. , Harshman J. , Huddleston C. J. , Marks B. D. , Miglia K. J. , Moore W. S. , Sheldon F. H. , Steadman D. W. , Witt C. C. and Yuri T . 2008 . A phylogenomic study of birds reveals their evolutionary history . – Science 320 : 17631768 .
  • Hatch S. and Nettleship D. (eds) 1998 . Northern fulmar (Fulmarus glacialis) . – The birds of North America Inc, Philadelphia .
  • Hey J. and Nielsen R . 2007 . Integration within the Felsenstein equation for improved Markov chain Monte Carlo methods in population genetics . – Proc. Natl Acad. Sci. USA 104 : 27852790 .
  • Kerr K. C. R. and Dove C. J . 2013 . Delimiting shades of gray: phylogeography of the northern fulmar, Fulmarus glacialis . – Ecol. Evol. 3 : 19151930 .
  • Librado R. and Rozas J . 2009 . DnaSP v5: a software for comprehensive analysis of DNA polymorphism data . – Bioinformatics 25 : 14511452 .
  • Lloyd C. , Tasker M. L. and Partridge K . 1991 . The status of seabirds in Britain and Ireland . – T and AD Poyser .
  • Mitchell P. I. , Newton S. F. , Ratcliffe N. and Dunn T. E . 2004 . Seabird populations of Britain and Ireland . – T and AD Poyser .
  • Morris-Pocock J. A. , Taylor S. A. , Birt T. P. and Friesen V. L . 2010 . Concerted evolution of duplicated mitochondrial control regions in three related seabird species . – BMC Evol. Biol. 10 : 14 .
  • Nei M . 1987 . Molecular evolutionary genetics . – Columbia Univ. Press .
  • Rains D. , Weimerskirch H. and Burg T. M . 2011 . Piecing together the global population puzzle of wandering albatrosses: genetic analysis of the Amsterdam albatross Diomedea amsterdamensis . – J. Avian Biol. 42 : 6979 .
  • Salomonsen F . 1965 . The geographical variation of the fulmar (Fulmarus glacialis) and the zones of the marine environment in the North Atlantic . – Auk 82 : 327355 .
  • Siorat F. and Rocamora G . 1995 . Changes in numbers and distribution of the northern gannet (Morus bassanus) on Rouzic Island, (reserve naturelle des Sept-Iles, Bretagne), France 1939–1994 . – Colonial Waterbirds 18 : 172178 .
  • Tamura T. , Dudley J. , Nei M. and Kumar S . 2007 . MEGA4: molecular evolutionary genetics analysis (MEGA) software ver. 4.0 . – Mol. Biol. Evol. 24 : 15961599 .
  • Wright S . 1951 . The genetical structure of populations . – Ann. Eugenics 15 : 323354 .

Supplementary material (Appendix JAV-00262 at < >). Appendix 1.