Past Colony Connectivity of a Declining Seabird Derived From Host–Parasite Genetic Data
C. P. Cargill, K. D. McCoy, B. E. Scott, E. A. Masden, J. Miller, L. Ruffino, A. Payo‐Payo
TL;DR
This study uses genetic data from seabirds and their parasites to understand how black-legged kittiwake colonies are connected across the North Atlantic.
Contribution
The study introduces a novel use of host–parasite genetic data and Bayesian analysis to infer past colony connectivity in a wide-ranging seabird.
Findings
Kittiwake dispersal and summer breeding movements show a longitudinal east-to-west trend.
Connectivity among colonies is less likely across the Atlantic Ocean.
Geographic distance weakly constrains connectivity, with other factors like colony status playing a role.
Abstract
The black‐legged kittiwake ( Rissa tridactyla , hereafter ‘kittiwake’, conservation status ‘Vulnerable’) is a long‐lived, highly motile and wide‐ranging seabird. Breeding kittiwake colonies are abundant across the northern hemisphere. The kittiwake's life history and the spatial scale of its breeding distribution make understanding colony connectivity a challenge; current species management models kittiwake colonies as closed units. Here, we explored the use of Bayesian analysis of multilocus microsatellite genotypes in the program BayesAss (BA3) to infer dispersal and seasonal summer breeding movements (information‐gathering behaviour; prospecting) (collectively ‘connectivity’) of kittiwakes around the North Atlantic. This approach uses the concept of inheritance by descent (IBD) (the formulation of genotypes within a population mediated by inheritance) and Markov‐chain Monte Carlo…
Genes, proteins, chemicals, diseases, species, mutations and cell lines named across the full text — each resolved to its canonical identifier and authoritative record.
Click any figure to enlarge with its caption.
FIGURE 1
FIGURE 2
FIGURE 3
FIGURE 4
FIGURE 5
FIGURE 6
FIGURE 7| Maritime region | Country | Region | Population (map ID) | Latitude | Longitude | Year |
|
|
|---|---|---|---|---|---|---|---|---|
| Baffin Bay | Canada | Qikiqtaaluk, Nunavut | Prince Leopold Island (PL) | 74.038 | −90.111 | 1993 | 11 | 0 |
| Greenland | Avannaata | Hakluyt Island (HI) | 77.432 | −72.708 | 1998 | 10 | 0 | |
| Labrador Sea | Canada | Newfoundland and Labrador | Lester Point (LP)* | 54.201 | −58.359 | 1992 | 8 | 0 |
| Atlantic Ocean | Canada | Newfoundland | Burgeo (BU)* | 47.612 | −57.617 | 1992 | 8 | 0 |
| Cape St Mary (CM) | 46.822 | −54.197 | 1997 | 0 | 36 | |||
| Gull Island (GU)** | 47.952 | −53.038 | 1997 | 0 | 21 | |||
| Baccalieu Island (BA)** | 48.137 | −52.798 | 1997 | 0 | 25 | |||
| Denmark Strait | Iceland | Breiðafjörður, Vestfirðir, Northwest | Flatey Island (FT) | 65.374 | −22.916 | 2001 | 24 | 0 |
| 1998 | 0 | 14 | ||||||
| Sea of the Hebrides | Scotland | Inner Hebridean Islands | Isle of Colonsay (CO) | 56.076 | −6.234 | 2001 | 30 | 0 |
| Irish Sea | Ireland | County Dublin | Lambay Island (LB)† | 53.491 | −6.018 | 2001 | 29 | 0 |
| Rockabill (RB)† | 53.598 | −6.004 | 2001 | 32 | 0 | |||
| Atlantic Ocean | France | Brittany | Cap Sizun (CP) | 48.049 | −4.695 | 1997 | 0 | 25 |
| North Sea | Scotland | Orkney Islands | Birsay (OK) | 59.117 | −3.319 | 2001 | 18 | 25 |
| Firth of Forth | Isle of May (IM) | 56.186 | −2.556 | 1999 | 0 | 33 | ||
| Peterhead, Aberdeenshire | Whinnyfold (WF) | 57.388 | −1.867 | 2001 | 28 | 35 | ||
| Shetland Islands | Foula Island (FO) | 60.138 | −2.076 | 2001 | 30 | 18 | ||
| Fair Isle (FI) | 59.534 | −1.629 | 2001 | 29 | 43 | |||
| Sumburgh Head (SH) | 59.854 | −1.275 | 2001 | 30 | 25 | |||
| Greenland Sea | Svalbard | Kongsfjorden | Krykkejefjellet (KR) and Ossiansarsfjellet (OS) | 78.928 | 12.482 | 1999, 2001 | 88 | 0 |
| Norwegian Sea | Norway | Vesterålen, Nordland | Nykvåg (NK) | 68.777 | 14.473 | 2000 | 20 | 24 |
| Harstad, Troms | Bjarkøya (BJ) | 68.996 | 16.505 | 2000 | 16 | 0 | ||
| Barents Sea | Norway | Nordkapp, Finnmark | Gjesvær (GJ) | 71.100 | 25.371 | 2000 | 29 | 9 |
| Båtsfjord, Finnmark | Syltefjord (SF) | 70.550 | 30.077 | 2000 | 22 | 24 | ||
| Vadsø, Finnmark | Ekkerøya (EK) | 70.075 | 30.103 | 2000 | 30 | 23 | ||
| Vardø, Finnmark | Rein Island/Reinøya (RN)†† | 70.393 | 31.133 | 2000 | 30 | 22 | ||
| Hornøya (HN)†† | 70.387 | 31.159 | 2000 | 39 | 0 | |||
| 1998 | 0 | 86 | ||||||
| Total | 561 | 488 | ||||||
- —UKRI NERC Scottish Universities Partnership for Environmental Research (SUPER)
- —University of Aberdeen10.13039/501100000882
Peer Reviews
No public reviews on file for this paper yet. If you reviewed it on a platform where reviews are public (OpenReview, ICLR, NeurIPS, ICML), you can paste yours below so the community can read it here.
Videos
No videos yet. Explain this paper in a talk, walkthrough, or lecture? Add one.
Taxonomy
TopicsAvian ecology and behavior · Genetic diversity and population structure · Animal Behavior and Reproduction
Introduction
1
Dispersal is a fundamental part of the biology and ecology of a species, facilitating persistence and adaptation by introducing genetic diversity across spatially and temporally heterogeneous or unpredictable environments (Hedrick and Gilpin 1997; e.g., Dayton and Szczys 2021; Herman et al. 2024). For long‐lived and wide ranging taxa, such as seabirds, dispersal facilitates the exploitation of novel and/or less competitive habitat (e.g., Munilla et al. 2016; Christensen‐Dalsgaard et al. 2020) when natal philopatry (returning to the location of birth/fledging) or breeding philopatry (inter‐annual use of a same breeding site) come at a cost to individual fitness (e.g., Fayet et al. 2021). For example, dispersal among seabird colonies tends to occur in relation to low individual and/or colony‐level breeding success (e.g., in Laridae: Cam et al. 2004; Boulinier et al. 2008; Palestis and Hines 2015; Oro et al. 2021). For a number of seabirds of conservation concern, dispersal likely underlies some of the observed spatio‐temporal variation in long‐term colony trends (e.g., European herring gull: Kentie et al. 2022; black‐legged kittiwake: Kildaw et al. 2008; Ponchon, Garnier, et al. 2014; Horswill et al. 2022) through spatial bias in breeding recruitment and/or the formation of ‘source‐sink’ metapopulation dynamics (defined in Levins 1969). However, spatially discrete populations (defined in Berryman 2002) of highly motile species which lack sufficient data to estimate dispersal are often defined as geographic closed units. When population‐level analyses adopt a precautionary approach (e.g., Cooney 2004), the species assessments which underpin environmental policy may lack biological realism. This limits their capacity to support decision‐makers and to reduce uncertainty within species management and conservation (see Camus and Lima 2002).
The black‐legged kittiwake, Rissa tridactyla (Linnaeus 1758), hereafter referred to as ‘kittiwake’, is a long‐lived (Golet et al. 2004) migratory pelagic gull of the northern hemisphere with a widespread breeding distribution in the High Arctic, Atlantic Ocean, Pacific Ocean and adjacent straits and shelf seas (BirdLife International 2019). The kittiwake breeds in colonies in predominantly coastal habitat on sheer cliffs and rocky stacks, often among large numbers of heterospecifics. The global conservation status of the kittiwake is ‘Vulnerable’ due to marked declines in the number of breeding pairs (BirdLife International 2019; OSPAR 2023) and declines in the average rate of breeding success at monitored colonies (CAFF 2020; Frederiksen et al. 2022); the overall global rate of decline is estimated to be in the range of 30% to 49% over three generations (39 years) (BirdLife International 2019).
The dispersal of kittiwakes is well‐studied relative to other northern hemisphere gulls, including the little gull ( Larus minutus ), black‐headed gull ( Chroicocephalus ridibundus ), great black‐backed gull ( L. marinus ) and lesser black‐backed gull ( L. fuscus ) (Horswill and Robinson 2015). Frequent movements among neighbouring kittiwake colonies have been observed during the summer months when both pre‐breeders (juveniles and sub‐adults; Cadiou et al. 1994; Cam et al. 2002) and adults engage in information‐gathering (‘prospecting’) prior to breeding (Danchin and Cam 2002; Ponchon, Gremillet, et al. 2014). ‘Adults’ refers to experienced breeders and includes three sub‐demographics: sabbaticals (adults that skip a breeding attempt; as in Aubry et al. 2009), failed breeders and successful breeders, the latter excluding active breeders as individuals engaged in incubation or chick‐rearing do not exhibit dispersive behaviour (Ponchon et al. 2015; Krali et al. 2023). The recruitment of first‐time breeders away from the natal colony (‘natal dispersal’) has been estimated at an annual mean rate of 0.890 (Porter and Coulson 1987; Horswill and Robinson 2015) and breeding dispersal (the relocation of adult kittiwakes) has been estimated in increasing and decreasing populations at annual mean rates of 0.012 ± 0.018 (standard deviation) and 0.062 ± 0.045, respectively (Danchin and Monnat 1992; Horswill and Robinson 2015). Natal dispersal is thought to occur most frequently among colonies less than 100 km (km) apart (Porter and Coulson 1987; Coulson and Nève De Mévergnies 1992; Danchin and Monnat 1992; Coulson and Coulson 2008), with a second pulse in dispersal posited at larger spatial scales—between 400 km and 900 km (Coulson and Nève De Mévergnies 1992). Studies of genetic structuring among kittiwake colonies indicate that dispersal is weakly associated with geographic distance (McCoy, Boulinier, and Tirard 2005; Sauve et al. 2019) and follows either a panmictic island model (all colonies have an equal likelihood of being connected by random dispersal) or frequent stepping‐stone model (geographic neighbours receive more dispersers) (McCoy, Boulinier, and Tirard 2005).
Study Rationale, Aims and Objectives
1.1
Despite their high estimated potential for dispersal, distinct breeding colonies of kittiwakes remain treated as closed units for the purposes of species management, conservation actions and related policy (Black and Ruffino 2019; Ruffino et al. 2020). The traditional capture‐mark‐recapture based approach to inferring population vital rates of kittiwakes—colour‐ringing (e.g., McKnight et al. 2019)—has been tested as a potential approach for improving estimates of dispersal among kittiwake colonies. However, the expected number of recaptures across space and time is too low given the realistic spatial scales of kittiwake colony distribution and the availability of resources (see O'Hanlon et al. 2021). Direct observational data of kittiwake dispersal are spatially constrained to local movements in France (Danchin and Monnat 1992) and northeast England (Porter and Coulson 1987; Coulson and Nève De Mévergnies 1992; Coulson and Coulson 2008). Anecdotal evidence from local ringing projects and large‐scale ring recovery networks (i.e., Suryan and Irons 2001; Poisblaud and Dromzée 2022) or biologging (France: Ponchon, Aulert, et al. 2017; Norway: Ponchon et al. 2015; Ponchon, Iliszko, et al. 2017; Russia: Ezhov et al. 2021) suggests that large‐scale dispersal occurs but provides no clear estimates on their frequency.
The spatio‐temporal resolving power of genetic indices of population connectivity is inherently coarse in long‐lived species and where there are few (i.e., < 10) suitable genetic markers available, a common problem when using nuclear microsatellite loci (Zimmerman et al. 2020; e.g., Dufresnes et al. 2023). To improve the temporal resolution of genetic outputs, previous studies have employed additional sources of genetic data derived from biological systems operating at finer temporal scales, i.e., short‐term seasonal dynamics. Of particular interest are host–parasite symbioses (Nieberding and Olivieri 2007; e.g., Sweet et al. 2020). In parasites, population structure and dispersal is often associated with the movements of the host (Combes 2001). Ixodes uriae , also known as the common seabird tick (hereafter referred to as ‘tick’), is a hard tick of the family Ixodidae. This tick parasitizes a wide range of seabird species in both the northern and southern hemispheres (Dietrich, Kempf, et al. 2014) and has been observed to form specific host‐associated genetic races within mixed‐species seabird colonies (McCoy et al. 2001; McCoy, Boulinier, and Tirard 2005; Kempf et al. 2009; Dietrich et al. 2012). The redistribution of the tick across geographic space occurs within the temporal bounds of the seabird breeding season when the end of the tick's annual blood meal coincides with the sequential information‐gathering behaviour of the seabird host (‘prospecting’ and ‘squatting’, see Cadiou et al. 1994; Cam et al. 2002; Cadiou 2008; Boulinier et al. 2016). Therefore, ticks can be used as a tool to improve inference of the inter‐colony movements of seabirds within the summer months (e.g., McCoy, Boulinier, and Tirard 2005).
Challenges to improving genetic inference of dispersal among kittiwake colonies include disentangling contemporary gene flow and post‐glacial range expansion (see Sauve et al. 2019) and capturing the directionality and relative magnitude of dispersal among genetically related colonies (see McCoy, Boulinier, and Tirard 2005). These elements are particularly vital for species assessments and developing subsequent strategic monitoring approaches for conservation and management. The aims of this study were to explore an alternative approach to improving inference of kittiwake dispersal using genetic data, and to build on previous genetic work reported in McCoy et al. (2003); McCoy, Boulinier, and Tirard (2005). The objectives were, (i) to use Bayesian inferential multilocus analysis to derive directional estimates of dispersal and seasonal breeding movements among kittiwake colonies (collectively ‘connectivity’), and (ii) to model colony connectivity as a function of geographic position and spatial isolation. We tested the null hypotheses, ‘connectivity among kittiwake colonies is heterogenous across temporal and spatial scales’ and ‘connectivity among kittiwake colonies is negatively correlated with geographic distance’. We used a past host–parasite (kittiwake–tick) microsatellite dataset from kittiwake colonies sampled across the Atlantic Ocean basin between the years of 1992 and 2001. Outputs are average per‐generational estimates of colony connectivity covering zero to three non‐overlapping generations prior to the time of sampling (Wilson and Rannala 2003).
Materials and Methods
2
Study Species
2.1
Kittiwake
2.1.1
Kittiwakes usually breed for the first time between 3 and 5 years of age for both sexes (Wooller and Coulson 1977; Porter and Coulson 1987; Cam et al. 2002) but breeding can be delayed for up to 8 years (Wooller and Coulson 1977): the average age at first breeding is 4 years (Horswill and Robinson 2015). Breeding occurs annually during the summer months (April through August) and is generally sustained throughout the adult life, with reproductive senescence observed after approximately 12 breeding years and/or immediately prior to mortality (Coulson and Thomas 1985; Coulson and Fairweather 2001). The introduction of genetic material among kittiwake colonies can occur via the relocation and recruitment of first‐time breeders away from their natal colony (‘natal dispersal’), and the recruitment of experienced breeders (‘breeding dispersal’).
Tick
2.1.2
The seabird tick I. uriae has a reproductive lifecycle of between two and 4 years, typically three (Barton et al. 1996). In general, this tick takes one blood meal per year, and it is only during this period that they parasitize the seabird host. Attachment duration while feeding ranges from three to 12 days depending on the life stage (larval, nymphal, adult) and host suitability (Eveleigh and Threlfall 1974; McCoy et al. 2002; Dietrich, Lobato, et al. 2014). The majority of ticks, regardless of lifecycle stage, feed during the incubation and chick rearing seasonal phases of the seabird host (e.g., kittiwakes breeding in Scotland: May through early June), before dropping off the host and overwintering in the substrate of the nesting area until the next opportunity to feed (Danchin 1992; Barton et al. 1996).
Sampling Distribution
2.2
Past microsatellite data were available for kittiwakes and ticks sampled from kittiwake colonies in the High Arctic (northeast Canada; northwest Greenland; eastern Svalbard; Norway) and distributed around the Atlantic Ocean (eastern Canada; eastern Iceland; northwest France; western Scotland), in addition to the western North Sea (northeast and eastern Scotland) and Irish Sea (western Ireland) (Figure 1). The total number of kittiwakes sampled was 561 and the total number of ticks sampled from kittiwakes was 488 (Table 1). Data were collected during six breeding seasons between the years of 1992 and 2001, inclusive, falling within a single kittiwake generation (Table 1). Colonies less than 50 km apart were pooled during analyses because movements are very frequent at this scale (e.g., Ponchon, Gremillet, et al. 2014); it is likely that this movement strongly correlates with dispersal, meaning that little genetic differentiation is expected (see e.g., McCoy, Boulinier, and Tirard 2005). These colonies were in the western Atlantic Ocean: Gull Island and Baccalieu Island, eastern Canada (c. 25 km apart); in the Irish Sea: Lambay Island and Rockabill Island, western Ireland (c. 10 km); in the High Arctic: Krykkejefjellet and Ossiansarsfjellet, eastern Svalbard (c. 10 km) and in Norway: Reinøya and Hornøya (c. 1 km). Due to small sample sizes, the samples for kittiwakes from Lester Point (Newfoundland and Labrador) and Burgeo (Newfoundland), approximately 750 km apart, were also pooled to give one overall sample for kittiwakes representing the western Atlantic Ocean. Inferential analyses of kittiwake dispersal used kittiwake microsatellite data and inferential analyses of kittiwake seasonal breeding movements used tick microsatellite data.
Sampling distribution of kittiwakes and ticks during the breeding seasons of 1992 through 2001 from, Burgeo (BU); Cape St Mary (CM); Baccalieu Island (BA); Gull Island (GU); Lester Point (LP); Prince Leopold Island (PL); Hakluyt Island (HI); Flatey Island (FT); Krykkejefjellet (KR); Ossiansarsfjellet (OS); Cap Sizun (CP); Lambay Island (LB); Rockabill Island (RB); Whinnyfold (WF); Sumburgh Head (SH); Foula Island (FO); Fair Isle (FI); Birsay (OK); Isle of Colonsay (CO); Isle of May (IM); Nykvåg (NK); Gjesvær (GJ); Hornøya (HN); Ekkerøy (EK); Rein Island/Reinøya (RN); Syltefjord (SF); Bjarkøya (BJ). Abbreviations are as follows, Bar. Sea, Barents Sea; Den. Str, Denmark Strait; Gre. Sea, Greenland Sea; I. Sea, Irish Sea; Lab. Sea, Labrador Sea; N. Sea, North Sea; Nor. Sea, Norwegian Sea; S.o.t.H, Sea of the Hebrides; Shetland Is, Shetland Isles. Map made in R (R Core Team 2023) using the following packages, ‘basemaps’ (Schwalb‐Willmann 2024); ‘mapedit’ (Appelhans et al. 2020), ‘ggplot2’ (Wickham 2016); ‘raster’ (Hijmans 2024a); ‘sf’ (Pebesma 2018; Pebesma and Bivand 2023); ‘ggspatial’ (Dunnington 2023). Two hundred Nautical Mile (NM) Exclusive Economic Zone (EEZ) boundaries from Flanders Marine Institute (2023). Southern extent of the Arctic Circle from Natural Earth. Map CRS: EPSG 9809/Oblique Stereographic (PROJ Contributors 2023).
Multilocus Estimation of Colony Connectivity
2.3
BayesAss (BA3)
2.3.1
Multilocus identity by descent (IBD) genotyping of individuals (the formulation of genotypes within a population mediated by inheritance: Thompson 2013) facilitates the derivation of quantitative estimates of dispersal among populations. For example, the Markov‐chain Monte Carlo‐based program BayesAss (BA3) (Wilson and Rannala 2003) assumes that the genotype of the first filial generation within any given ‘receiving’ population, denoted i, is produced in part by the input of genotypes from a ‘donor’ population, j. Proportional estimates of immigrants within population i originating from population j are derived from the posterior probability distributions of recent individual ancestries, themselves derived from Markov‐chain Monte Carlo resampling of multilocus genotypes.
The relatedness of a set of sampled populations is therefore defined as patterns of spatial aggregations of alleles; states of genetic similarity or dissimilarity between sampling locations expressed as the mean per‐generation proportion of individuals, m, within receiving population i derived from donor population j (mi,j) (mathematical notation follows Wilson and Rannala 2003). Here, BA3 ‘MSAT’ v3.0.5.7 for Windows 64 was used to estimate the mean values and plausible ranges of values (95% credible intervals, CIs) for per‐generation connectivity among kittiwake colonies. It is important to note that inferential estimates derived using BA3 are not interpreted as a population vital rate (i.e., percentage proportions of individuals moving between populations) and are not necessarily comparable between species, rather they provide a quantitative measure of relative connectivity among a sampled set of populations (see Wilson and Rannala 2003). Approximate CIs around the mean mi,j were calculated as shown in Equation (1), where μ (mu) refers to the population mean and σ (sigma) the standard deviation.
Input Data
2.4
The full microsatellite dataset comprised diploid nucleotide base‐pair lengths for seven highly polymorphic loci isolated from the kittiwake (K6, K31, K32, K67, K71, UL12 and HC6) and the tick (T22, T38, T39, T44, T47, T5 and T35) (kittiwake microsatellites are published in Tirard et al. 2002; locus UL12 is referred to as ‘ulo12a12’ in Ibarguchi et al. 1999; tick microsatellites are published in McCoy and Tirard 2003). Allele data in GENEPOP format were converted to IMMANC/BAYESASS format using the data conversion software PGDSpider v2.1.1.5 (Lischer and Excoffier 2012). McCoy, Boulinier, and Tirard (2005) previously assessed the quality of the data and met assumptions of no linkage disequilibrium among loci and Hardy–Weinberg equilibrium within populations, although the latter was not a requirement for this study (see Wilson and Rannala 2003). Measurable genetic differentiation among colonies is highly dependent on the allelic variation present within each colony sample (see Hedrick 1999); here, we anticipated that the low information content of few microsatellite loci with a high degree of allelic polymorphism would reduce population inbreeding coefficients and limit statistical power. The observed allelic polymorphisms at each locus and corresponding estimates of population inbreeding coefficients are provided in Data S1.
Markov‐Chain Monte Carlo Conditions
2.5
The Bayesian priors implemented within BA3 are uniform in the interval [0,1] and constrained such that at least two‐thirds of each colony sampled for both the kittiwake and the tick were assumed to be philopatric (see Wilson and Rannala 2003). The percentage of complete genotypes available was 97.86% for the kittiwake dataset and 87.80% for the tick dataset—missing alleles are dealt with by BA3 by replacement with alleles drawn at random (Wilson and Rannala 2003). Delta values for the BA3 MCMC parameters migration rate (‘migration’ is common parlance in genetic studies for dispersal), allele frequency and inbreeding coefficients were adjusted to as close as possible meet the recommended acceptance rates for proposed parameter changes of between 40% and 60% (Wilson and Rannala 2003). Each MCMC run comprised a total of 50,000,000 iterations, with the first 10,000,000 iterations discarded as burn‐in and samples taken every 2000 iterations: the total number of samples used to inform the posterior probability distribution was 20,000. For both kittiwake and tick, five independent MCMC test runs were first initiated with different random number seeds. Bayesian deviances (see Meirmans 2014; Appendix S1) were calculated in the statistical environment R v4.3.2 (R Core Team 2023) for each test run and used as an objective optimality criterion to identify the single test run that best fit the observed data. A test run was only considered for acceptance where the Bayesian deviance between subsequent runs was at least negative two. Additionally, log‐probability plots for each MCMC test run were visually assessed to evaluate the convergence and mixing of MCMC chains and consistency of estimates between runs. As an objective measure, R‐hat (R^) convergence diagnostics (Gelman and Rubin 1992; Vehtari et al. 2021; Stan Development Team 2024) were calculated for each run; an R^ value of less than 1.05 is indicative of adequate mixing of MCMC chains. The best MCMC runs for kittiwake and tick provided the estimates of colony connectivity reported in this study. All subsequent data management, statistical analyses and graphical outputs were conducted in R and all values are reported rounded to four significant figures.
Exploring Colony Connectivity as a Function of Geographic Position and Spatial Scale
2.6
Large‐scale spatial trends in colony connectivity (east/west and north/south) were identified by calculating the net connectivity specific to each colony, ∑i=1nmi,j, where n = 17 for the kittiwake and n = 14 for the tick, and fitting a linear regression between ∑i=1nmi,j and colony longitude and latitude (units: decimal degrees). Pair‐wise shortest geographic distances around land barriers among colonies, hereafter referred to as the explanatory variable distance (units: km), were then calculated as follows: High‐resolution coastline data (CRS: WGS84/EPSG: 4326) of all relevant land masses including islands were downloaded as shapefiles from the publicly available Global Administrative Areas (GADM) vector dataset series v4.1 (GADM 2018– 2022, URL: https://gadm.org) and rasterized in R using the packages ‘sf’ (Pebesma 2018; Pebesma and Bivand 2023) and ‘terra’ (Hijmans 2024b) with a final xy grid cell resolution of 0.36° × 0.04° (units: decimal degrees). The distance between grid cells representing each sampling location was calculated using the function ‘gridDist’ from the R package ‘terra’ (Hijmans 2024b) by using the rasterized land shapefiles as a mask. Great‐circle (Haversine straight‐line) distances between sampling locations were also considered; the relative values did not differ substantially from the shortest distance around land barriers and were therefore not used within subsequent analyses. The relationship between distance and the strength of connectivity among colonies was modelled using Gaussian Generalised Additive Models (GAMs) with an identity link function using the package ‘mgcv’ (Wood 2011) with method ‘GCV.Cp’ as shown in Equation (2), where s() denotes a non‐parametric smooth; both linear and non‐linear functions of distance were included to allow for the detection of multiple peaks in colony connectivity at any given value of distance (such as reported for kittiwakes in Coulson and Nève de Mévergnies 1992), as well as any overall negative or positive relationship between distance and colony connectivity. To reduce over or under smoothing, the basis dimension, k, was optimised to 15 for kittiwake and 5 for tick (k values of 15, 10, 7, 5 and 3 were tested sequentially and the lowest k value with k‐index above a value of 1 chosen Wood 2011). Statistical significance was assumed when p values were less than 0.05. To evaluate how well the fitted GAMs described the observed data, plots of model deviance residuals were examined for normality and homoscedasticity and percentages of deviance explained derived from model summaries.
Results
3
Kittiwake Microsatellite Analysis: Dispersal
3.1
The kittiwake colony at Krykkejefjellet/Ossiansarsfjellet (Svalbard) was the largest net ‘donor’ colony of those sampled, and Prince Leopold Island and Hakluyt Island (both in Baffin Bay), the largest net ‘receivers’ (Figure 2). The highest estimate of mean per‐generational dispersal between any one colony pair was from Krykkejefjellet/Ossiansarsfjellet (Svalbard) to Hornøya/Reinøya (Norway); mHN/RN,KR/OS = 0.048 ± 0.066. Overall, kittiwakes were less likely to disperse across the Atlantic Ocean compared to around the east Atlantic shelf seas. However, the Atlantic Ocean was not a definitive geographic barrier to dispersal—dispersal was more likely to be from east to west across the Atlantic Ocean compared to the reverse (Figure 2). No latitudinal trend of dispersal among kittiwake colonies was observed. The multimodal behaviour of the negative relationship between pair‐wise dispersal and geographic distance (Figure 3) is indicative of dispersal bias among kittiwake colonies.
Overall dispersal among kittiwake colonies was longitudinal from east to west (Linear regression line ±95% confidence interval (dashed); degrees of freedom = 16, t = −3.04, p = 0.0079). Reference lines at ∑i=n17mi,j = 0, longitude = 0° are shown in grey. BJ, Bjarkøya; BU, Burgeo/LP, Lester Point; CO, Isle of Colonsay; EK, Ekkerøy; FI, Fair Isle; FO, Foula Island; FT, Flatey Island; GJ, Gjesvær; HI, Hakluyt Island; HN, Hornøya/RN, Reinøya; KR, Krykkejefjellet/OS, Ossiansarsfjellet; LB, Lambay Island/RB, Rockabill Island; NK, Nykvåg; OK, Birsay; PL, Prince Leopold Island; SF, Syltefjord; SH, Sumburgh Head; WF, Whinnyfold.
The impact of increasing geographic distance on the strength of pair‐wise dispersal (mi,j+mj,i) (open circles) among kittiwake colonies was negligible at spatial scales up to 3000 km but became increasingly negative above this approximate threshold, where the curve crosses zero (Generalised Additive Model smooth ± standard error lines (dashed); deviance explained by model: 44.1%; non‐linear term: Effective degrees of freedom = 12.8, F = 47.07, p < 2e‐16; linear term: T = 2.89, p = 0.004).
Confidence in kittiwake dispersal estimates was limited by low statistical power: in all cases the ranges of plausible values (credible intervals) around the means encompassed zero and overlapped for opposing colony pairs (Figure 4). Minor hills and valleys were observed in all MCMC test traces following burn‐in and R^ exceeded 1.05 (R^ = 1.051). This indicates some problems with convergence and mixing of MCMC chains. The final delta values (Δ) and acceptance rates for the BA3 parameters were, migration rate: Δ0.50 and 0.65; allele frequency: Δ0.50 and 0.39; inbreeding coefficient: Δ0.50 and 0.58. The full datasets generated are provided online in Data S2 (Table S1) and Data S3 (Table S1).
Estimates of mean per generational dispersal (mi,j) among Atlantic black‐legged kittiwake colonies derived from Bayesian inferential analysis of multilocus microsatellite genotyping, shown with 95% credible intervals in grey. BJ, Bjarkøya; BU, Burgeo/LP, Lester Point; CO, Isle of Colonsay; EK, Ekkerøy; FI, Fair Isle; FO, Foula Island; FT, Flatey Island; GJ, Gjesvær; HI, Hakluyt Island; HN, Hornøya/RN, Reinøya; KR, Krykkejefjellet/OS, Ossiansarsfjellet; LB, Lambay Island/RB, Rockabill Island; NK, Nykvåg; OK, Birsay; PL, Prince Leopold Island; SF, Syltefjord; SH, Sumburgh Head; WF, Whinnyfold. Receiver colonies are plotted in the same order left to right as donor colonies.
Tick Microsatellite Analysis: Kittiwake Seasonal Breeding Movements
3.2
Tick microsatellite analysis highlighted a number of predominant ‘donor’ kittiwake colonies of ticks during the summer breeding season (Figure 5) which linked colonies in the following maritime regions (Figure 6): the western Atlantic Ocean (Figure 5a Cape St Mary's); North Sea, Denmark Strait and Norwegian Sea (Figure 5b Isle of May); North Sea and eastern Atlantic Ocean (Figure 5c Whinnyfold); Barents Sea (Figure 5d Hornøya/Reinøya). The highest estimate of seasonal movements between any one colony pair was from Cape St Mary's to Baccalieu Island/Gull Island, both in eastern Canada: mBA/GU,CM = 0.214 ± 0.054. Whinnyfold (Scotland) was a net ‘donor’ of ticks: ∑j=113mWh,j = 0.167, ∑i=113mi,Wh = 1.19, net = −1.024. The significant negative relationship between tick dispersal and geographic distance (Figure 7) suggests more localised movements of kittiwakes during the summer breeding season; the dispersal of ticks was highest among kittiwake colonies less than 500 km apart. There was no effect of either latitude or longitude on tick dispersal among kittiwake colonies. The full datasets generated are provided in Data S2 (Table S2) and Data S3 (Table S2).
Estimates of mean per generational dispersal (m<sub>i,j</sub>) for ticks sampled from black‐legged kittiwake colonies in Newfoundland (Cape St Mary's), Scotland (Isle of May, Whinnyfold) and Norway (Hornøya/Reinøya) derived from Bayesian inferential analysis of multilocus microsatellite markers; 95% credible intervals are shown with dashed horizontal lines at mi,j=0. BA, Baccalieu Island; BU, Burgeo; CM, Cape St Mary; CP, Cap Sizun; EK, Ekkerøy; FI, Fair Isle; FO, Foula Island; FT, Flatey Island; GJ, Gjesvær; HN, Hornøya; IM, Isle of May; NK, Nykvåg; OK, Birsay; RN, Rein Island/Reinøya; SF, Syltefjord; SH, Sumburgh Head; WF, Whinnyfold.
Kittiwake colonies grouped by tick dispersal (indicative of kittiwake seasonal breeding movements); closely linked colonies graphically represented within the same grey circles. Map made in R (R Core Team 2023). Southern extent of the Arctic Circle from Natural Earth. Map CRS: EPSG 9809/Oblique Stereographic (PROJ Contributors 2023).
There was a negative impact of increasing distance on the strength of pair‐wise colony connectivity by tick dispersal (mi,j+mj,i) (open circles), particularly above an approximate threshold of 2000 km, where the curve crosses zero (Generalised Additive Model smooth ± dashed standard error lines; deviance explained by model: 19.9%; non‐linear term: Effective degrees of freedom = 1.88, F = 16.74, p = 1.78e‐07, linear term: T = 2.81, p = 0.0059).
In contrast to the kittiwake microsatellite analysis, multilocus genotype analysis of the ticks was more powerful; all MCMC test chains showed adequate convergence and mixing in log‐probability plots and by using objective criteria (maximum R^ = 1.011). Final delta values and acceptance rates for BA3 MCMC parameters were, migration rate: Δ0.35 and 0.68; allele frequency: Δ0.35 and 0.64; inbreeding coefficient: Δ0.70 and 0.74.
Discussion
4
In this study, we explored the application of Bayesian inferential analysis to a past host–parasite microsatellite dataset to estimate directional connectivity (dispersal and seasonal breeding movements) among Atlantic kittiwake colonies. Analysis of kittiwake microsatellites suggested an overall longitudinal trend in dispersal among kittiwake colonies, from colonies sampled in the east (Barents Sea, Greenland Sea) to those sampled in the west (Baffin Bay, Labrador Sea). Analysis of tick microsatellites, used here as a proxy for kittiwake seasonal breeding movements (i.e., prospecting), showed clear movements among colonies in the Denmark Strait (Iceland), Norwegian Sea (Norway), North Sea (Scotland) and the eastern Atlantic Ocean (northwest France), colonies in the east Atlantic Ocean (Newfoundland) and colonies in the Barents Sea (Norway). A non‐linear negative relationship between colony connectivity and geographic distance indicated that although isolation by distance may play some role in constraining colony connectivity, it is unlikely to be the sole determinant—note, for example, the weakly negative relationship in Figure 3 and the data points around 2000 km and exceeding 3000 km in Figure 7. There was a lower propensity for both dispersal and seasonal breeding movements to occur across the Atlantic Ocean. Overall, these results concur with contemporary anecdotal evidence from colour‐ringing studies for of dispersal among kittiwake colonies operating across large spatial scales. For example, an ongoing study at the industrial‐urban kittiwake colony in Boulogne‐sur‐Mer (northern France) has resighted kittiwakes ringed as chicks in Wales (Anglesey), east Scotland (Inchkeith), England (from north to south: Howick, Gateshead, Hartlepool and Lowestoft), Sweden (Nidinghen) and western France (Cap‐Sizun), across spatial scales of approximately 190 km to 1000 km (Poisblaud and Dromzée 2022). Of the 28 kittiwakes resighted, 15 have successfully paired, mated and raised at least one chick in Boulogne‐Sur‐Mer (Poisblaud and Dromzée 2022). Further, given the longitudinal migratory seasonal movements of the kittiwake between breeding and overwintering areas (Frederiksen et al. 2011; Amélineau et al. 2023) and the strong attraction of kittiwakes to conspecifics (Danchin et al. 1998), this study supports the theory that the spatio‐temporal associations of kittiwakes outside the breeding season could contribute to dispersal and recruitment dynamics (e.g., Bogdanova et al. 2011, 2017; Swindells 2019).
Positive skew in the deviance residuals and the presence of systematic change in the spread of the deviance residuals (heteroscedasticity) of the GAMs fitted indicated that geographic distance alone failed to explain some significant patterns in the microsatellite dataset. Instead of stepping‐stone or panmictic island‐type models of dispersal where colony connectivity is predominantly described by geographic distance, the dispersal and seasonal breeding movements of this highly motile and wide‐ranging seabird are more likely underpinned by distance‐cost trade‐offs (see Bonte et al. 2011) operating in combination with a number of factors associated with the breeding grounds. For example, information on colony breeding status/habitat quality and conspecific attraction (see Kildaw et al. 2008). It is also plausible that the ability of non‐breeding kittiwakes to prospect more widely in space and time (Kotzerka et al. 2010; Goutte et al. 2014) relative to active breeders (Chivers et al. 2012; Schlener et al. 2024), and the spatio‐temporal overlaps in at‐sea habitat use by kittiwakes from different colonies (Redfern and Bevan 2014; Paredes et al. 2014; Ponchon, Aulert, et al. 2017; Petalas et al. 2021), may facilitate prospecting and recruitment decoupled from geographic distance.
Case Study: Kittiwakes Breeding in Scotland
4.1
In the UK, kittiwake breeding and count data are available for most Special Protection Area (SPA) populations through the Seabird Monitoring Programme (SMP, URL: https://app.bto.org/seabirds/public/index.jsp) from the late 1980s to present day. The dispersal estimates in this study represent average per‐generational values for colony connectivity broadly covering zero to three non‐overlapping generations prior to the time of sampling (Wilson and Rannala 2003). Kittiwakes recruit at an average age of 4 years (Horswill and Robinson 2015), and the tick has a reproductive life span of 2 to 4 years (Barton et al. 1996). Therefore, it was possible to contextualise the overall genetic signals of connectivity among Scottish kittiwake colonies using spatially explicit population trend data (productivity, counts of breeding pairs) for the decade prior to genetic sampling. SMP data are provided online in Data S4.
Under a hypothesis of performance‐based conspecific attraction (Danchin et al. 1998; Suryan and Irons 2001), recruiting kittiwakes preferentially avoid colonies experiencing high levels of breeding failure. Data from Birsay (Marwick Head SPA) and Fair Isle (Fair Isle SPA) kittiwake colonies support this assumption: Productivity at Birsay generally exceeded productivity at Foula Island throughout the 1990s. Productivity at Fair Isle was lower than Foula Island at the start of the decade; however, the decadal trend was to exceed that of Foula Island from 1993 onwards. Here, we found genetic signals of dispersal from Foula Island to both Birsay and Fair Isle. Overall, population declines at Foula Island (Foula Island SPA) support the genetic signal of kittiwakes leaving Foula Island (this study).
Connectivity ecology among kittiwake colonies is likely more complex. A declining population offers available nesting sites to prospective breeders, and larger colonies also confer additional benefits on new recruits, such as a higher availability of mating opportunities and a lower risk of predation (see Gochfeld 1980), and may also be more attractive to prospecting individuals due to differences in acoustic cues (Mulard et al. 2009) and/or olfactory cues (for kittiwake, see LeClaire et al. 2009, 2011; Pineaux et al. 2023, also in, e.g., Procellariiformes: Buxton and Jones 2012). Productivity at Whinnyfold (Banff and Buchan SPA) and Sumburgh Head (Sumburgh Head SPA) was lower than Foula Island on average across the decade, and in particular in the second half of the 1990s. Yet we found genetic signals of dispersal into these two populations, from Foula Island into Whinnyfold and from Whinnyfold and Foula Island into Sumburgh Head. The Whinnyfold breeding population was an order of magnitude larger than Foula Island and in a state of decline (from 24,957 breeding pairs in 1995 to 13,330 in 2001, a loss of over half (53%) of the breeding population over a 6 year period). In comparison, the decline of the colony at Sumburgh Head occurred at a lower rate, despite being substantially smaller than Foula Island and impacted by reduced breeding effort following the ‘Braer’ tanker oil‐spill in the northwest North Sea in 1993 (see Walton et al. 1997). Placed, albeit to a limited extent, in the wider ecological context of the time, Sumburgh Head may have been functioning as an ecological ‘sink’ for a short period of time; sustained to a degree by immigration from a wider metapopulation rather than by natal philopatry (Levins 1969).
In general, patterns of connectivity among kittiwake colonies are likely to be highly dynamic over space and time because the underlying drivers of prospecting and recruitment are associated with spatio‐temporal dynamics in biological, ecological and anthropogenic pressures. For example, since these microsatellite data were collected, the continued decline of the Foula Island SPA kittiwake colony—to a low of 259 breeding pairs in 2019—has been linked to the collapse of the Shetland sandeel stock (Furness and Tasker 2000; McGregor et al. 2022) and predation by great skua ( Stercorarius skua ) (Oro and Furness 2002). An interesting area of research would be to integrate time series data on colony status to define Bayesian priors for inferential estimation of dispersal using genetic data. This would allow for a direct examination of the additional informative value of applying species‐specific ecological hypotheses, such as performance‐based conspecific attraction, to reduce uncertainty around derived estimates of recent colony connectivity, and capture some of the latent biological and anthropogenic variables underlying kittiwake prospecting and recruitment.
Study Limitations
4.2
Analyses of DNA polymorphisms are well established as a tool for defining population structure in avian conservation (e.g., Plaza et al. 2023; Kubacka et al. 2023), including estimating dispersal between ocean basins (e.g., tropicbirds (Phaethontidae): Humeau et al. 2020) and within ocean basins (e.g., black‐legged kittiwake: McCoy, Chapuis, et al. 2005; roseate tern ( Sterna dougallii ): Byerly et al. 2023; tufted puffin ( Fratercula cirrhata ): Graham et al. 2023). The BayesAss software (BA3) has been previously used to derive estimates of population connectivity for a number of seabirds, including Galapagos Nazca ( Sula granti ) and red‐footed ( S. sula ) boobies (Levin and Parker 2012; Morris‐Pocock et al. 2016), great ( Fregata minor ) and magnificent ( F. magnificens ) frigatebirds (Levin and Parker 2012; Nuss et al. 2016) and several species of tern (Sterninae) (Faria et al. 2010; Boutilier et al. 2014; Szczys et al. 2017; Gu et al. 2021; Byerly et al. 2023). However, uncertainty around derived estimates of kittiwake dispersal indicates that future research is likely to benefit from using more information‐rich single nucleotide polymorphism (SNP) data (see Mussmann et al. 2019), or exploration of an alternative approach (e.g., epigenetics: Oleksiak and Rajora 2019; Norouzitallab et al. 2019). Additionally, it is unknown whether spatial sampling bias within colonies in this past genetic dataset disproportionately favoured individuals with high or low dispersal probability: intra‐colony spatial hierarchies in nesting habitat quality creates intra‐colony spatial heterogeneity in dispersal probability associated with individual state (breeding experience, breeding status and performance) (Acker et al. 2017). Although both central and peripheral colony areas were known to be accessed during sampling, the inferential methodology used in this study is highly sensitive to sampling bias within populations (Wilson and Rannala 2003) and a comparison of the impact of sampling strategy (including a truly randomised or structured design) on outputs is highly desirable.
Finally, note that the seasonal breeding movements of kittiwakes inferred from tick microsatellites reflect only a single spatio‐temporal ‘snapshot’ of the information‐gathering behaviour of the kittiwake. Prospecting kittiwakes sequentially visit colonies in the same breeding season (Ponchon, Gremillet, et al. 2014; Acker et al. 2017) and the redistribution of ticks among kittiwake colonies depends on the suitability of the habitat for both tick and kittiwake and on the attachment duration of the tick. The latter is highly variable, ranging from three up to 12 days, and it is known that engorged ticks are able to delay detachment until the host is present in a suitable environment for the tick (White et al. 2012). Kittiwake seasonal breeding movements inferred from tick analyses are therefore predisposed towards identifying ‘receiver’ kittiwake colonies visited by kittiwakes at the end of the tick's annual blood meal, and ‘donor’ kittiwake colonies with relatively extensive tick infestations. For example, the kittiwake colony at Whinnyfold is over 60 years old, with a documented history of tick infestation (Boulinier and Danchin 1996). In this study, tick microsatellites identified the seasonal breeding movements of kittiwakes from Whinnyfold northwards into the Orkney Islands (Birsay) and the Shetland Islands (Foula Island, Fair Isle and Sumburgh Head). Assuming a hypothesis of local breeding failure prompting wide‐ranging prospecting movements in kittiwakes (see Ponchon, Garnier, et al. 2014; Ponchon, Gremillet, et al. 2014; Ponchon et al. 2015; Ponchon, Iliszko, et al. 2017; Coulson and Nève De Mévergnies 1992; Danchin and Monnat 1992), the carriage of ticks from Whinnyfold into regional colonies is expected given the long‐term declines in kittiwake productivity at this colony throughout the 1990s (SMP data, see Data S4).
Management Implications and Recommendations
4.3
The governance of important habitat patches (i.e., SPAs) and habitat patch networks (e.g., the European Natura 2000 network and the United Kingdom National Site Network) (European Commission: Directorate‐General for Environment and Sundseth 2015; JNCC 2024) for the kittiwake is underpinned by Population Viability Analysis (PVA, Gilpin and Soulé 1986). Important contemporary and developing policy for kittiwakes informed by PVA includes ornithological compensation for marine renewables developments (e.g., GoBe Consultants Ltd. 2022; Searle et al. 2022) and environmental appraisals preceding the decommissioning of oil and gas rigs colonised by breeding seabirds (e.g., McLean et al. 2022). Whilst noting some key analytical and inferential limitations, this study has provided further empirical evidence of heterogenous seasonal breeding movements among kittiwake colonies and recruitment bias across large spatial scales. As such, the current precautionary ‘closed unit’ definition of kittiwake colonies within PVA is likely to be lacking in biological realism by precluding the modelling of spatial bias in ‘donor’ and ‘receiving’ colonies. Geographic distance may be useful as a coarse explanatory variable to define conservation units for kittiwakes within the bounds of the breeding season (this study) but appears to be less suited to defining relationships among colonies in PVA‐based approaches (i.e., metapopulation viability analysis (mPVA), see simulation study in Miller et al. 2019). However, in lieu of more robust dispersal estimates, the qualitative correlations observed between genetic signals of connectivity among Scottish kittiwake colonies and time series (SMP) population data indicate that the integration of geographic information with spatially explicit population trend data remains a worthwhile approach. Uncertainty around derived predictions of colony connectivity could be further reduced by using information on habitat use and spatio‐temporal associations of conspecifics outwith the breeding season. Finally, given the large spatial scales across which kittiwake colony connectivity apparently occurs (this study), the interpretation and applicability of future genetic outputs will likely be improved by sampling, for example, the wider North Sea coastline and accessible offshore colonies, the latter likely to comprise an important subset of national kittiwake populations (i.e., Christensen‐Dalsgaard et al. 2020; Delahay et al. 2025).
Author Contributions
C. P. Cargill: conceptualization (lead), data curation (supporting), formal analysis (lead), investigation (lead), methodology (lead), project administration (lead), visualization (lead), writing – original draft (lead), writing – review and editing (lead). K. D. McCoy: data curation (lead), resources (lead), supervision (supporting), writing – review and editing (equal). B. E. Scott: conceptualization (supporting), funding acquisition (equal), project administration (supporting), supervision (equal), writing – review and editing (equal). E. A. Masden: conceptualization (supporting), funding acquisition (equal), supervision (equal), writing – review and editing (equal). J. Miller: conceptualization (supporting), funding acquisition (equal), supervision (equal), writing – review and editing (equal). L. Ruffino: conceptualization (supporting), funding acquisition (equal), supervision (equal), writing – review and editing (equal). A. Payo‐Payo: conceptualization (supporting), funding acquisition (lead), project administration (supporting), supervision (equal), writing – review and editing (equal).
Funding
This work was supported by UKRI NERC Scottish Universities Partnership for Environmental Research (SUPER), NE/S007342/1.
Conflicts of Interest
The authors declare no conflicts of interest.
Supporting information
Data S1: ece373204‐sup‐0001‐DataS1.xlsx.
The reference list from the paper itself. Each links out to its DOI / PubMed record.
- 1Acker, P. , A. Besnard , J.‐Y. Monnat , and E. Cam . 2017. “Breeding Habitat Selection Across Spatial Scales: Is Grass Always Greener on the Other Side.” Ecology 98, no. 10: 2684–2697. 10.1002/ecy.1962.28746975 · doi ↗ · pubmed ↗
- 2Amélineau, F. , A. Tarroux , S. Lacombe , et al. 2023. “Multi‐Colony Tracking of Two Pelagic Seabirds With Contrasting Flight Capability Illustrates How Windscapes Shape Migratory Movements at an Ocean‐Basin Scale.” Ecography 2024, no. 2: e 06496. 10.1111/ecog.06496. · doi ↗
- 3Appelhans, T. , K. Russell , and L. Busetto . 2020. “mapedit: Interactive Editing of Spatial Data in R, R Package Version 0.6.1.9002.” https://github.com/r‐spatial/mapedit.
- 4Aubry, L. M. , D. N. Koons , J.‐Y. Monnat , and E. Cam . 2009. “Consequences of Recruitment Decisions and Heterogeneity on Age‐Specific Breeding Success in a Long‐Lived Seabird.” Ecology 90, no. 9: 2491–2502. 10.1890/08-1475.1.19769127 · doi ↗ · pubmed ↗
- 5Barton, T. R. , M. P. Harris , S. Wanless , and D. A. Elston . 1996. “The Activity Periods and Life‐Cycle of the Tick Ixodes uriae (Acari: Ixodidae) in Relation to Host Breeding Strategies.” Parasitology 112, no. 6: 571–580. 10.1017/S 0031182000066154.8684831 · doi ↗ · pubmed ↗
- 6Berryman, A. A. 2002. “Population: A Central Concept for Ecology?” Oikos 97: 439–442.
- 7Bird Life International . 2019. Rissa tridactyla (Amended Version of 2019 Assessment). The IUCN Red List of Threatened Species 2019: e.T 22694497 A 155617539. Bird Life International. 10.2305/IUCN.UK.2018-2.RLTS.T 22694497 A 155617539.en. · doi ↗
- 8Black, J. , and L. Ruffino . 2019. Better Linking Effects of Offshore Windfarms on Black‐Legged Kittiwakes to Populations. JNCC Report No. 645. Joint Nature Conservation Committee. https://hub.jncc.gov.uk/assets/bdd 7e 6f 0‐b 0fe‐431d‐b 403‐85a 456aacdc 9.