Modelled larval dispersal and measured gene flow: seascape genetics of the common cockle Cerastoderma edule in the southern Irish Sea
- 939 Downloads
The role of marine currents in shaping population connectivity in the common cockle Cerastoderma edule was investigated in the southern Irish Sea. C. edule is one of the most valuable and exploited shellfish species in the area, yet very little is known about its population dynamics. In the present study, coupled hydrodynamic and particle tracking models are used in conjunction with genetic data collected at twelve microsatellite loci to estimate the influence of the Celtic Sea front on larval transport between the coasts of Britain and Ireland. Genetic analysis highlights the presence of at least three genetic clusters partitioned within locations, suggesting a contact zone between separate subpopulations. Samples collected from the Irish coast are most similar to each other. On the British coast, the Burry Inlet appears genetically isolated while samples collected from the coast of Pembrokeshire show evidence of connectivity between Britain and Ireland. These results agree with the model’s predictions: away from the coastal zone, residual baroclinic currents develop along tidal mixing fronts and act as conduit systems, transporting larvae great distances. Larvae spawned in south Wales are capable of travelling west towards Ireland due to the Celtic Sea front residual current, confirming the action of the Celtic Sea front on larval transport. Sheltered, flood-dominant estuaries such as the Burry Inlet promote self-recruitment. The validation of the model using genetic data represents progress towards a sustainable future for the common cockle, and paves the way for a more effective approach to management of all Irish Sea shellfisheries.
KeywordsParticle tracking model Larval dispersal Common cockle Cerastoderma edule Celtic sea front Burry inlet
It has long been recognized that landscape features can have an important influence on the diversity of natural populations (Fisher and Ford 1947). However, methodologies for studying landscape genetics have been developed only during the last decade, by combining landscape ecology with population genetics (Manel et al. 2003; Sork and Waits 2010; Storfer et al. 2010). These are now beginning to be applied to the aquatic environment (Gilg and Hilbish 2003; Galindo et al. 2006; Selkoe et al. 2006, 2010; Schunter et al. 2011; Selkoe and Toonen 2011).
In the marine realm, the early stages of life are considered crucial in shaping diversity. The vast majority of marine species begin their life cycles with a pelagic larval stage (Pechenik 1999). When adults are characterized by restricted mobility, the larval stage is likely to bear principal responsibility for population connectivity (Cowen et al. 2006). Larvae can spend anywhere between hours to several weeks in the plankton depending on species, with oceanographic processes affecting their transport, and ultimately their distribution (Palumbi 2004; White et al. 2010), by acting as barriers to dispersal (Galarza et al. 2009; Sala-Bozano et al. 2009), or conveying larvae for tens of kilometers (Shanks et al. 2003; Galindo et al. 2010) and potentially homogenizing populations (Bortolotto et al. 2011; Watson et al. 2011).
In order to study the physical and biological factors that influence larval dispersal and species diversity, we investigated a system in the southern Irish Sea, a region known for complex hydrodynamic processes (Dabrowski et al. 2010). The circulation in the Irish Sea is strongly influenced by tides (Robinson 1979) and also baroclinic currents formed by seasonal heating during the summer months (Hill et al. 1994; Brown et al. 2003). Large tidal ranges in the east (e.g. the Bristol Channel) produce strong barotropic velocities. However, tidal flows are oscillatory and it is the residual, baroclinic, currents that become more important for larval transport. In the southern Irish Sea, baroclinic currents develop along the Celtic Sea tidal mixing front, during summer (Simpson 1976). The front might be responsible for the diversity in species composition (Gosling et al. 2008) and population differentiation (Mariani et al. 2012) observed between the south (Atlantic-wave exposed), and the east (Irish Sea) of Ireland.
The common cockle Cerastoderma edule L. is a long-lived (6–10 years), widespread bivalve occurring in intertidal soft-sediment locations along the coast of North-western Europe (Beukema and Dekker 2009; Ponsero et al. 2009; Freire et al. 2010; Malham et al. 2012). The species is a broadcast spawner, with external fertilization (André et al. 1999). Maturity is reached at about 18 months of age, and spawning times vary between populations (Seed and Brown 1977), triggered by environmental cues such as temperature (Ducrotoy et al. 1991). C. edule has a prolonged spawning season that can last from April (Guillou et al. 1990) to October (Cardoso et al. 2009). In the UK, spawning generally occurs between May and August (Seed and Brown 1977), with occasional peaks in September (Hancock 1967). C. edule produce a planktotrophic larva (veliger) that can live in the water column for up to 5 weeks and could be affected by local currents (Jonsson et al. 1991). Following primary settlement, cockle larvae can disperse again through a ‘bisso-pelagic’ dispersal, if environmental conditions are unsuitable (deMontaudouin 1997; Bouma et al. 2001; Huxham and Richards 2003). Information on vertical migration characteristics and other biological traits of C. edule is sporadic in the literature, and dissimilar between case studies: in the Wadden Sea, for example, cockles spawn in June–September and show diel migratory rhythms, yet, as postlarvae, they can settle by passive deposition (Baggerman 1953; Armonies 1992, 1994; Armonies and Hartke 1994).
Cerastoderma edule forms some of the most commercially valuable fisheries in the UK and Ireland (Dare et al. 2004; Hervas et al. 2008). However, the population dynamics of C. edule in the UK and Ireland are poorly understood, with little systematic research conducted to date (Seed and Brown 1977, 1978; Dare 1992; Dare and Walker 1993; Dare et al. 2004; Cesar and Frid 2009) and minimal stock management. Population genetics of cockles in the UK were investigated by Beaumont and Pether (1996) using allozyme markers. However, the resolution of this study was insufficient to reveal statistically significant genetic structure between cockle populations. The need for evidence-based management of cockle populations has been emphasized by annual mass mortalities in the Burry Inlet (Malham et al. 2008; Elliott et al. 2012) and southern Ireland over the last decade. Although management plans have been put in place to promote the recovery of the Burry Inlet cockle fishery, these are hindered by a lack of concrete information regarding the dynamics and genetics of Burry Inlet cockle populations.
The main aim of this study is to investigate the roles of mesoscale tidal and residual (baroclinic + wind) currents in the southern Irish Sea shelf region in determining the population structure of highly valuable marine species. In particular, the research focuses on the cockle C. edule and the influence that residual currents associated with the Celtic Sea front has on self-recruitment within their natal population and connectivity with other oceanographically distinct populations. Biophysical modelling has been performed to predict larval transport in the southern Irish and Celtic Seas, and analysis of genetic population structure will be used to qualitatively validate the model.
Materials and methods
Sampling and genotyping
Sampling locations, population codes and genetic diversity parameters inferred from microsatellites
All individuals were genotyped using twelve species-specific markers (Martínez et al. 2009) arranged in two multiplex reactions as follows: CeATC1-5(NED), CeATC2-12(VIC), CeATC2-11(6FAM), CeATC1-52(NED), CeATC2-44(PET), CeATC2-46(PET), CeATC1-34(VIC) in multiplex1, and CeATC2-51(VIC), CeATC2-4(6FAM), CeATC1-54(NED), CeATC1-36(6FAM), CeATC1-22(PET) in multiplex2. Amplification was carried out using Qiagen Multiplex PCR Kit (Qiagen, CA-USA) in a final volume of 10 μl, containing 5 μl of Multiplex Kit Buffer 2× and 2.5 μg of genomic DNA. Primer concentrations were different according to optimized conditions, and the PCR cycle was the same for both reactions: 95 °C for 15 min, followed by 35 cycles of 45 s at 94 °C, 45 s at 59 °C, 45 s at 72 °C, and a final extension step at 72 °C for 45 min. Products were then run alongside a GS500LIZ size standard in an ABI3730xl Genetic Analyzer (Applied Biosystems) and alleles were scored using Genemapper4.0 (Applied Biosystems).
Analysis of genetic diversity and population structure
Frequencies of null alleles were estimated using FreeNA (Chapuis and Estoup 2007). Genetic variation was assessed estimating expected (H e) and observed (H o) heterozygosities using GenAlEx v6 (Peakall and Smouse 2006). Linkage disequilibrium was estimated using Genepop v4 (Raymond and Rousset 1995; Rousset 2008) and allelic richness (A R), F IS and F ST, [θ estimator (Weir and Cockerham 1984)] were calculated with Fstat 2.9.3 (Goudet 1995). Finally, a mantel test (Mantel 1967; Smouse et al. 1986) was performed in R (R Development Core Team 2011) using the package Ecodist (Goslee and Urban 2007) to test for the presence of isolation by distance (IBD).
In order to confirm the neutrality of the markers employed in this study, a neutrality test as implemented in Lositan (Antao et al. 2008) was performed. The test is based on the FST-outlier method, originally described by Beaumont and Nichols (1996). The ‘outlier’ loci will show increased levels of population differentiation if they are under selection or linked to a locus that is (‘genetic hitch-hiking’). The program was run for each population pair under the default settings.
Analysis of molecular variance (AMOVA) (Excoffier et al. 1992) as implemented in Arlequin 3.5 (Excoffier and Lischer 2010) was performed to estimate genetic diversity within and among populations, and significance was tested after 1,000 permutations. Several group designs were tested, and those combinations that maximized FCT values (among-group variance) and minimized FSC values (among populations, within-group variance) were assumed to indicate the most probable subdivisions.
Population structure was investigated using two statistical approaches. Firstly, the Bayesian clustering method implemented in Structure 2.3 (Pritchard et al. 2000; Falush et al. 2003, 2007), allowing for admixture and correlated allele frequencies, and using 500,000 iterations following a burn-in period of 100,000. The number of clusters k was calculated by averaging the mean posterior probability of the data L(k) over 10 independent runs as well as by calculating the second order rate of change in probability between successive k values as described in Evanno et al. (2005) and implemented in Structure Harvester (Earl and vonHoldt. 2012). Independent runs for each k were averaged using Clumpp 1.1 (Jakobsson and Rosenberg 2007) and visualized in Distruct (Rosenberg 2004).
The second method used is the discriminant analysis of principal components (DAPC) implemented in Adegenet (Jombart 2008; Jombart et al. 2010) for R (R Development Core Team 2012). Both methods assign individuals to a number of clusters (k) without making any assumptions about their origin, but while Structure groups them in genetic clusters by minimizing Hardy–Weinberg and linkage disequilibria, DAPC simply maximizes genetic separation among groups, while minimizing variation within groups (Jombart et al. 2010). The optimal k was selected based on the associated Bayesian information criterion (BIC) value calculated, after 107 iterations: the lowest BIC value represents the optimal k.
The modelling methodology was based on simulations of larval transport in the southern Irish Sea. Firstly, we used a 3-dimensional (3D) hydrodynamic model to reproduce the observed barotropic and baroclinic circulation and, in particular, residual currents associated with the Celtic Sea front. Secondly, we used Lagrangian particle tracking models (PTMs) to predict the proportion of larvae contributing to self-recruitment at each sample site (Fig. 1), and connectivity between these populations. Larvae were released (spawned) from April to September and the pelagic larval duration (PLD) was 28 days (d), based on data reported in the literature (Siegel et al. 2003; Hartnett et al. 2007; Banas et al. 2009; Tian et al. 2009) and in relation to sea surface temperature (O’Connor et al. 2007). Three vertical larval migration strategies were tested: passive, tidal-stream transport, and diel transport.
The model domain is shown in Fig. 1, extending from the northern Celtic Sea to beyond the North Channel. In some respects, the domain was typical of semi-enclosed basins around the world (Taylor 1919) and, hence, analogous to other marine ecosystems. However, tidal ranges are extremely large on the Welsh and English coasts, especially in the Bristol Channel (>11 m) (Robinson 1979), generating large tidal flows (Neill et al. 2009). In contrast, there is an amphidromic point (zero tidal range) off Ireland. The model experiments focus on larval transport on the shelf; transport within estuaries may depend on local conditions and must be addressed separately (e.g. Robins et al. 2012). We therefore assume that larvae are already located off-shore and will be transported to either their natal habitat (self-recruitment) or to a habitat elsewhere (connectivity).
A 3D free-sea-surface, Princeton Ocean Model (POM, Blumberg and Mellor 1987; Mellor and Yamada 1982) was applied as our model system for the Irish Sea. It was assumed that the weight of the fluid balances pressure [hydrostatic assumption (Mellor et al. 1994)], and also that density differences are neglected unless influenced by gravitational forces (Boussinesq approximation). The prognostic variables (e.g. elevation, velocity, temperature) were solved using finite-difference discretization on an orthogonal (Arakawa-C) numerical model grid in the horizontal plane and terrain-following (sigma) layers in the vertical plane. The model horizontal cell size was 1/30° (longitude) by 1/60° (latitude), giving a resolution of approximately 1.85 km. In the vertical, 20 equally segmented sigma-layers gave minimum resolution at mean sea level of approximately 10 m.
In order to validate the model, a mean year in the recent past was chosen (namely 1990), based on a decadal simulation (1989–1998) of bed shear stress, wave heights and temperatures, from a model of the Northwest European Shelf (Neill et al. 2010), and was then used to provide hydrodynamic input for the PTM. The Irish Sea model was forced at the open boundaries with the 6 principal tidal constituents. The European Centre for Medium-Range Weather Forecasts ECMWF-ERA-Interim reanalysis (Simmons et al. 2006) was the source of synoptic meteorological input fields, available at a resolution of 1.5°, 3-hourly.
Particle tracking model
Lagrangian PTMs were used in conjunction with the hydrodynamic model to simulate individual particle displacement in space and time based on advection, sub-grid-scale turbulent mixing, and individual particle behaviour. Velocity and diffusivity output from the hydrodynamic model were used, off-line, in PTMs, so that the hydrodynamics could be used to simulate a large number of cluster-release scenarios. Velocities were tri-linearly interpolated to the position of each particle, and the time interval of hydrodynamic model (15 min) linearly interpolated to 5 min for the PTM. Each particle was then iteratively advected in space and time.
Summary of hydrodynamic (POM) and particle tracking model (PTM) simulations
PTM release schedule
Apr. 1st–Sep. 30th, 1990
PTM-2: Tidal @ 3 mm s−1
PTM-3: Diel @ 3 mm s−1
• Patches of 10,000 particles released from each sample site (see Fig. 1).
• 6 separate releases: (once per month), each simulating a 28 days PLD
For each PTM scenario, cohorts of 10,000 larvae were released at the sample locations (particle release locations were randomly chosen within 1 km of each sample site): Burry Inlet (both north and south sites have been combined for PTM releases), Dale, BB, TRA and FS. Test simulations with decreasing numbers of particles (starting with 1 × 107 particles) showed that 10,000 particles were sufficient to statistically represent dispersal (i.e. the proportion of larvae in each model cell varied by less than 10 %, from the initial simulation). Larval cohorts were released 6 times with start dates chosen each month (April–September). The simulations were repeated at different dates in order to assess the variance in dispersal. Although these results are not shown, in each case, the dispersal patterns were similar to the results presented hereafter, which are therefore representative of all the simulations. Since this study focused on potential dispersal, no mortality was assigned to the larvae. If larvae were advected onto a land model cell, they were reflected back to their position within the domain at the previous time-step (a similar method was used by North et al. 2008). Upon completion of each simulation, larvae within 10 km of a sample site would settle there (Cowen et al. 2006), whereas the others were considered ‘unsuccessful’, and were not accounted for in estimation of connectivity and self-recruitment among the considered sample sites. This distance is comparable to the distance larvae travel in mean currents of 0.5 m s−1 (Robins et al. unpublished), over a flood or ebb tidal phase.
Estimates of self-recruitment and connectivity, C ij , were calculated based on the proportion of larvae released in population i that settle in population j. A value of C ij close to unity indicates high connectivity whereas a value of C ii close to unity indicates high self-recruitment.
Genetic diversity and population structure
Evidence of null alleles was detected in four loci, with high frequency (i.e. r ≥ 0.20): CeATC1-5 in all populations, CeATC2-12 in seven populations (excluding NBA, for which r = 0.18), CeATC2-46 (in three populations: NBA, SBA and NBJ) and CeATC2-51 (in four populations: FS, NBA, SBA and NBJ). Null alleles have been described previously at these loci by Martínez et al. (2009) and, in general, heterozygote deficiency due to null alleles has been extensively measured across marine molluscs (Foltz 1986; Hoare and Beaumont 1995; Beaumont and Pether 1996; Reece et al. 2004; Mariani et al. 2012). It has been demonstrated that the inclusion of loci with null alleles in population studies can lead to an overestimation of genetic differentiation (F ST) (Chapuis and Estoup 2007; Carlsson 2008), while it can improve the levels of assignment when testing for population structure using clustering programs (Carlsson 2008; Griffiths et al. 2010; Schunter et al. 2011). Hence, genetic diversity indices were calculated after removing the four loci with a strong signal of null alleles (Table 1), whereas overall and pairwise FST (Weir 1996) were calculated after ENA correction performed in FreeNA retaining all loci (Table 1, lower).
Expected heterozygosity (H e) ranged from 0.708 (NBJ) to 0.741 (NBA), and was higher than the observed (H o). Even after removing four loci, F IS was positive and significant in all populations, ranging between 0.088 (BB) and 0.202 (SBA). Of the remaining eight loci, three seem to be responsible for producing high values of F IS (data not shown): CeATC1-52, CeATC1-54 and CeATC2-4. Their null allele frequencies averaged across populations were ‘intermediate’, being respectively 0.098, 0.118 and 0.086. Inbreeding coefficients (F IS), positive and significant, and the heterozygosity deficiency (H o < H e) detected at each sampled population were likely results of the Wahlund effect, due to the coexistence of genetically distinct cohorts within each sampling location, as confirmed by the assignment tests presented below (Fig. 2S).
A very weak but significant IBD was found (R2 = 0.19, Mantel r = 0.44, 95 % CI: 0.33–0.65) across all populations.
Overall, F ST before ENA correction was 0.0073 (95 % CI: 0.0046-0.0105), and was 0.0081 (95 % CI: 0.0047–0.0125), after ENA correction. Based on F ST pairwise comparisons (Table 1), the Irish populations are similar to one another, apart from FS and TRA, with a low but significant F ST (0.005, 95 % CI: 0.001–0.011). BB, TRA, FS and DA are then all significantly different from the Burry Inlet populations, with significant F ST values ranging from 0.02 (BB/NBJ) to 0.007 (FS/NBJ). Within the Burry Inlet, only two samples showed significant differentiation: the sample from the South Burry collected in April, before the mortality event (SBA), was significantly different from those collected in the same month in the North Burry (NBA, F ST = 0.006, 95 % CI: 0.0006–0.015) and from the same location after the mortality (SBJ, F ST = 0.008, 95 % CI: 0.002–0.016).
Lositan confirmed the neutrality of the set of markers, and no significant linkage disequilibrium was detected.
AMOVA results highlighted the lack of differentiation amongst Irish populations and amongst Welsh ones. Even within the Burry, no diversity was detected between North and South samples as well as between April and July samples. The only significant results were those separating Ireland and the Burry Inlet samples, either pooling DA with the Burry (F SC = 0.005, P = 0.006; F CT = 0.006, P = 0.014) or with the Irish populations (F SC = 0.004, P = 0.006; F CT = 0.006, P = 0.03) (see Table 1 Supplementary).
The three genetic groups assumed were not distributed following any specific patterns, with an overlapping distribution across locations (Fig. 2; Fig 2S).
Modelling larval transport
Larval dispersal was predicted to be greater in high-velocity environments (e.g. Dale (DA), Fig. 5a), and also in regions influenced by residual flow pathways, e.g. Irish populations transported within the Celtic Sea front residual (Fig. 5b). In the Burry Inlet, a more sheltered estuary, self-recruitment was predicted to be high and connectivity with other populations low. In addition, models predicted a strong effect of vertical migration behaviour. Larvae drifting passively would disperse off-shore from their natal habitats, whereas tidal or diel transport promoted retention or led to transportation to other coastal areas, but in most cases, active larvae did not remain off-shore. For example, Fig. 6 shows predictions that a group of larvae from Dale performing diel transport would be retained in South Wales while a second group would cross the Irish Sea and reach the Irish coast.
Hydrodynamic controls on larval dispersal
Connectivity matrices for (a) passive transport, (b) tidally-stream transport and (c) diel transport
The bathymetry and shape of the coastline surrounding the release zone affects larval transport and dispersal. Relatively long estuaries, such as the Burry Inlet (NB/SB), which is approximately 15 km in length, tend to promote flood-dominance (Friedrichs and Aubrey 1988), where net transport over a tidal cycle is directed up-estuary. Hence, larvae released in the Burry were predicted to show high self-recruitment, whereas larvae released from more exposed coastlines or smaller embayments (e.g. Irish populations), were predicted to show reduced self-recruitment and greater dispersal (Figs. 6, 7). For instance, larval retention levels of 50 % for passive transport and almost 100 % for tidal and diel transport were simulated from releases in the Burry Inlet (Fig. 7; Table 3). In contrast, simulated larvae released from TRA (which is a smaller bay), were quickly exposed to strong off-shore currents and travelled away from the parent population (self-recruitment levels were less than 1 % for passive transport and approximately 70 % for tidal/diel transport (Fig. 7; Table 3).
Modelled depth-averaged residual currents for the period 1st June–31st September are shown in Fig. 6b. This period was chosen to include maximum summer stratification. Wind effects also contribute to the residuals, although these are small (and confined to the surface) compared with the baroclinic component. In the southern Irish Sea, an important residual current (of magnitude 0.05–0.1 m s−1) flows westward from the Bristol Channel towards southern Ireland, across St George’s Channel. The net current continues westward (at 0.1–0.2 m s−1) along the south coast of Ireland towards the Celtic Sea. Another significant residual current that has been simulated is the western Irish Sea gyre, which flows in an anti-cyclonic direction, in between North Wales and Ireland. The gyre generates strong residual currents (>0.2 m s−1) flowing south along the west coast of Ireland, which meet the Celtic Sea front residual in the southern Irish Sea.
In terms of larval transport, these residual currents generate strong directionality in connectivity patterns; the fate of larvae that are entrained in these flows is ultimately controlled by them, until they reach land where coastal currents become important for larval transport (Robins et al. unpublished). An illustration of this process is presented in Fig. 6, where passive larvae generally travelled west, entrained in the Celtic Sea front residual current. Vertical migration strategies were shown to advance or retard the advection of larvae, but not notably alter their trajectories. As a consequence, there was connectivity of South Wales populations (DA) with Irish populations (BB, TRA and FS), although connectivity was generally less than 1 % in all cases (Table 3). However, there was no reciprocal connectivity of Irish larval populations with Welsh populations (Figs. 6, 7). Connectivity of Irish populations with neighboring populations to the west was as high as 20 % (e.g. connectivity of BB passive populations with TRA, Table 3a).
Role of vertical migration strategy on larval dispersal
For the majority of larvae simulated, self-recruitment and connectivity of passive larvae were lower than for larvae with tidal or diel transport, because more passive larvae remained off-shore in deeper water (i.e. high ‘larval wastage’). Self-recruitment of passive larvae was less than 50 %, whereas for the other strategies, self-recruitment was greater than 70 % (Table 3). As well as promoting self-recruitment, tidal and diel strategies also increased levels of connectivity with nearby populations (Table 3), because the larvae were exposed to on-shore surface currents. As a result, tidal and diel strategies generated low larval wastage. One exception is closure of connectivity of Burry populations with Dale, for tidal/diel transport (Fig. 7). In these cases, Burry populations do not emigrate out of the estuary. The effect of larval strategy is contingent to the local circulation patterns: Irish populations, for example, show considerably more dispersal of passive than active larvae, because the westward (and on-shore) residual currents in the region (Fig. 5b) promote retention of active larvae. Transport patterns of tidal and diel larvae were similar, although diel transport resulted in higher connectivity’s at some sites (Table 3; Fig. 6).
In this study, the role of the marine currents in determining population connectivity of C. edule in the southern Irish Sea shelf region was investigated. Microsatellite markers were employed to assess population structure, and results were qualitatively compared with estimates of population connectivity through larval dispersal calculated using a biophysical modelling approach. These comparisons represent a first step towards the validation of the model as a predictor of population connectivity and hence of dispersal and gene flow.
In general terms, the model predicts connectivity between C. edule populations on the Welsh and Irish coasts due to the action of the Celtic Sea front which forms during the cockles’ spawning season. This prediction is supported by genetic data which indicates that allele frequencies within Irish and Welsh populations of C. edule are similar, despite the long-distance separating the two stretches of coast and the restricted mobility of adult individuals.
More specifically, the oceanographic and PTMs show that residual currents, caused mainly by atmospheric events, are important for shelf-scale larval transport. Both tidal velocities (governed by tidal range and local bathymetry) and residual currents (controlled by atmospheric heating and mixing, and also wind) contribute to larval dispersal (Robins et al. unpublished). Tidal flows, however, are oscillatory and result in small net dispersal over a tidal cycle. Residual flows, on the other hand, advect larvae much further (up to 400 km over a period of 4 weeks (Robins et al. unpublished). Shelf-scale density-driven, or baroclinic, currents develop along oceanographic fronts where stratified regions meet well mixed waters and, during summer months, contribute significantly to the circulation of the Irish Sea. Such currents act as conduits, funneling larvae in one direction. In the southern Irish Sea, residual currents direct larvae westwards across the Celtic Sea front. The baroclinic flow can be attributed to the observed position of the Celtic Sea tidal mixing front in the southern Irish Sea (Simpson and Hunter 1974). Simulated larvae released from Dale in South Wales, for example, were more likely to infiltrate Irish populations (~200 km to the west) than settle in the Burry Inlet (~50 km to the east), over the 28 day period of larval dispersal. The Celtic Sea front also restricts eastward transport of Irish larval populations; it would be unlikely that cockles spawned from southern Ireland would settle in England or Wales within a pelagic larval dispersal (PLD) of around 1 month.
In coastal areas, topographic features become important for larval transport. For instance, the Burry Inlet is flood-dominant which leads to high self-recruitment of larvae and low connectivity with other populations. Indeed, our model’s estimates of larval retention and connectivity (Table 3) confirm that the Burry Inlet is characterized by low larval exchange, compared to the other sample sites. Vertical migration significantly affects estimates of larval dispersal modelled here. When tidal or diel dispersal were implemented in the model, retention at each site increased, especially at the Burry Inlet, where almost ‘full-retention’ of the population was predicted. This effect shows the importance of including species-specific larval behaviour, or accounting for it in the most comprehensive way, when empirical data is not available, such as in the case of the common cockle. The similarity of tidal and diel transport is because, over the 28 days period (which is a multiple of the spring-neap cycle), the different strategies cause larvae to spend equal proportions of time in the surface and near-bed currents. However, the fact that diel swimmers spend half as much time migrating upwards and downwards than tidal swimmers means that they travel slightly further/faster.
Results obtained from the microsatellite analysis revealed the presence of at least three genetic clusters (Figs. 2, 4). These groups are partitioned between sampling sites in similar proportions (Fig. 2S). Despite this, the pairwise F ST (Table 1, lower diagonal), the F ST-based AMOVA analysis and the DAPC (Fig. 3), reveal the presence of a weak but significant geographical pattern, with the Irish populations not significantly different from each other or from Dale, but separated to some extent from the Burry Inlet samples, in accordance with the model’s predictions. The genetic isolation of the Burry Inlet was already detected by Beaumont and Pether (1996) and, as stated above, can be explained by the sheltered nature of the estuary, where larval retention is promoted (Fig. 7). Within the estuary, despite the significant F ST values, both AMOVA and assignment tests fail to detect any substructure, either temporally (April and July samples), and spatially (North and South samples).
Considering the genetic results in the light of biophysical tracking models allows the formulation of several scenarios that could have led to the observed population sub-structuring in the region, although the identification of specific causes is outside of the general aims of this study. Firstly, the Irish Sea could represent a suture zone, where different stocks of C. edule from different geographical areas, favoured by the mixing action of the residual currents, converge and overlap. Separate haplogroups of common cockle have been described recently by Krakau et al. (2012) along the Atlantic European coast, as a result of recolonisation of the area from northern periglacial refugia, located in the North-east Atlantic. Based on mitochondrial DNA (mtDNA) analysis, at least two main groups have been described: a ‘southern’ haplogroup, distributed from Spain to the English Channel, and a ‘northern’ haplogroup, including the North Sea and Scandinavia. The existence of a third ‘Arctic’ haplogroup is also hypothesized. Secondary contact after this ancient separation could be responsible for the genetic structure detected by microsatellites, which are more suitable than mtDNA for studying present demographic phenomena.
Alternatively, the genetic structure described in C. edule could be maintained by its irregular spawning behaviour (Ducrotoy et al. 1991; Genelt-Yanovskiy et al. 2010). In the Irish Sea, C. edule can undergo multiple spawning events within a single year (Seed and Brown 1977; Morgan et al. 2012) and these could explain the co-existence of three genetic clusters as well as the heterozygote deficiency detected at each sample even after correcting for null alleles. Departure from Hardy–Weinberg expectations of heterozygosity is a well-documented characteristic of marine bivalves (Zouros and Foltz 1984) and populations of C. edule (Martínez et al. 2012), that does not seem to lead to erosion of the genetic diversity.
The strong directionality of the Celtic Sea front, from the Welsh and English coast towards Ireland, however, implies that the genetic composition of the Irish populations result mainly from the Welsh spawning, while larvae released in Ireland are likely to be transported westward, into the Celtic Sea. As well as subsequent spawning events, seasonal fluctuations in the number of spawners could also produce, and/or maintain, genetic diversity by genetic drift. It has been observed that in some cases recruitment is more successful after a harsh winter or in general where there is a low density of adults (Ducrotoy et al. 1991).
In addition, it is also possible to speculate that, based on the genetic and PTM results, cockle larvae in the southern Irish Sea seem to be mainly passively transported across great distances. Obviously, further experimental work needs to be carried out in order to confirm this hypothesis. Nevertheless, if larvae were influenced mainly by tidal and/or diel migration, the resulting population structure would be likely very different, with populations from different estuaries being genetically more distinct, due the higher self-recruitment rates (Fig. 7; Table 3). The general agreement between the two approaches is promising, and suggests a series of further steps to be taken. An immediate benefit would be the identification of key source populations, which, through long-distance dispersal of larvae are responsible for recruitment over a wide area.
Future work should aim to fully validate the model by quantitatively comparing simulations and empirical data, by expanding the geographical area covered and the number of species involved. The integration of empirical and simulated estimates of migration represents a difficult task (see Galindo et al. 2006; Hohenlohe 2004). The possibility of reliably estimating dispersal rates from populations departing from Hardy–Weinberg assumptions will aid the statistical validation of the PTM.
Model resolution in coastal areas should also be increased so that local circulations near natal populations are well resolved. Temporal sampling could be employed to confirm whether the population structure detected here is stable over time. The inclusion of further shellfish species with a similar pelagic larval phase will allow this approach to be applied to wider issues of shellfisheries management in the Irish Sea and beyond. For example, it might be used to inform the designation of marine protected areas (MPAs) or special areas of conservation (SACs) to preserve shellfish stocks. Furthermore, it has been shown that baroclinic currents can have a significant influence on the transport of pelagic larvae, and preliminary results indicate that these currents are affected by increasing sea surface temperatures (Robins et al. unpublished data). Hence, great attention should be given to the incorporation of climate change predictions in the hydrodynamic model.
Once validated, the implementation of this approach in management plans could aid in ensuring the future sustainability of shellfisheries resources in the southern Irish Sea and other fisheries threatened by overexploitation or environmental change.
SUSFISH is a 3 year project funded by the European Union Regional Development Fund (ERDF) under the Ireland Wales Programme 2007–2013—Interreg 4A, Project No. 042. The authors would like to thank Matt Longshaw (CEFAS, UK) and Emer Morgan (UCC, Ireland) for sharing samples and John Hickeys (BIM, Ireland) for assisting with sampling in Ireland.
- Antao T, Lopes A, Lopes RJ, Beja-Pereira A, Luikart G (2008) LOSITAN: a workbench to detect molecular adaptation based on a FST-outlier method. BMC Bioinformatics, 9Google Scholar
- Armonies W, Hartke, D (1994) Floating of mud snails (Hydrobia ulvae) in tidal waters of the Wadden Sea, and its implications on distribution patterns. Helgolhnder Meeresunters 49:529–538Google Scholar
- Baggerman B (1953) Spatfall and transport of Cardium edule L. Neth J Zool 10:316–342Google Scholar
- Bortolotto E, Bucklin A, Mezzavilla M, Zane L, Patarnello T (2011) Gone with the currents: lack of genetic differentiation at the circum-continental scale in the Antarctic krill Euphausia superba. BMC Genet 12Google Scholar
- Dare PJ (1992) A 95-year record of cockles and mussels spatfalls in the Wash: the use of qualitative information recorded in Sea Fishery Committee annual reports. First Yerseke Workshop, YersekeGoogle Scholar
- Dare PJ, Walker P (1993) Spatfalls of cockles and mussels in the Wash in relation to preceding winters temperatures and possible effects of Spring wind regimes upon larval dispersal: a preliminary analysis. Second Yerseke Workshop, YersekeGoogle Scholar
- Dare PJ, Bell MC, Walker P, Bannister RCA (2004) Historical and current status of cockle and mussel stocks in the wash. Lowestoft, pp 85Google Scholar
- Ducrotoy JP, Rybarczyk H, Souprayen J, Bachelet G, Beukema JJ, Desprez M, Dorjes J, Essink K, Guillou J, Michaelis H, Wilson JG, Elkaim B, Ibanez F (1991) A comparison of population dynamics of the cockle (Cerastoderma edule) in North-Western Europe. In: Elliot M, Ducrotoy JP (eds) Estuaries and coasts: spatial and temporal intercomparisons. Olsen and Olsen, Fredenborg, pp 173–183Google Scholar
- Elliott M, Burdon DR, Callaway R, Franco A, Hutchinson T, Longshaw M, Malham S, Mazik K, Otto Z, Palmer D, Firmin C, Smith T, Wither A (2012) Burry inlet cockle mortalities investigation 2009–2011. Technical Report to the Environment Agency Wales. Reference No: YBB140-Tech Final-2012. P363Google Scholar
- Excoffier L, Lischer HEL (2010) Arlequin suite Ver 3.5: a new series of programs to perform population genetics analyses under Linux and Windows. Mole Ecol Res 10:564–567Google Scholar
- Freire R, Arias A, Méndez J, Insua A (2010) Identification of European commercial cockles (Cerastoderma edule and C. glaucum) by species-specific PCR amplification of the ribosomal DNA ITS region. Eur Food Res Technol 1–4Google Scholar
- Goff LJ, Moon DA (1993) PCR amplification of nuclear and plastid genes from algal herbarium specimens and algal spores. J Phycol 7Google Scholar
- Goslee SC, Urban DL (2007) The ecodist package for dissimilarity-based analysis of ecological data. J Stat Softw 22(7):1–19Google Scholar
- Goudet J (1995) FSTAT (Version 1.2): a computer program to calculate F-statistics. J Hered 86:485–486Google Scholar
- Griffiths AM, Machado-Schiaffino G, Dillane E, Coughlan J, Horreo JL, Bowkett AE, Minting P, Toms S, Roche WK, Gargan P, McGinnity P, Cross T, Bright D, Garcia-Vasquez E, Stevens JR (2010) Genetic stock identification of Atlantic salmon (Salmo salar) populations in the southern part of the European range. BMC Genet 11:31PubMedCrossRefGoogle Scholar
- Hervas A, Tully O, Hickey J, O’Keefe E, Kelly E (2008) Assessment, monitoring and management of the Dundalk Bay and Waterford estuary cockle (Cerastoderma edule) fisheries in 2007. In: Fisheries resource series, p 38. BIM, DublinGoogle Scholar
- Jakobsson M, Rosenberg NA (2007) CLUMPP: a cluster matching and permutation program for dealing with label switching and multimodality in analysis of population structure. Bioinformatics 23(14):1801–1806Google Scholar
- Malham SK, Woolmer AP, Jones G, Cock D, Callaway R (2008) A report into the causes of mortality of the edible cockle (Cerastoderma edule) in the burry inlet and three rivers, South Wales, p 73Google Scholar
- R Development Core Team (2011) R: a language and environment for statistical computing. R Foundation for Statistical Computing, ViennaGoogle Scholar
- R Development Core Team (2012) R: a language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. http://www.R-project.org/
- Raymond M, Rousset F (1995) GENEPOP (version 1.2): population genetics software for exact tests and ecumenicism. J Hered 86:248–249Google Scholar
- Selkoe KA, Watson JR, White C, Ben Horin T, Iacchei M, Mitarai S, Siegel DA, Gaines SD, Toonen RJ (2010) Taking the chaos out of genetic patchiness: seascape genetics reveals ecological and oceanographic drivers of genetic patterns in three temperate reef species. Mol Ecol 19:3708–3726PubMedCrossRefGoogle Scholar
- Shanks AL (1995) Mechanisms of cross-shelf dispersal in larval invertebrates and fish. In: McEdward L (ed) Ecology of marine invertebrate larvae. CRC Press, Boca Raton, pp 324–367Google Scholar
- Simmons A, Uppala S, Dee D, Kobayashi S (2006) ERA interim: new ECMWF reanalysis products from 1989 onwards. ECMWF Newsl 110:25–35Google Scholar
- Weir BS (1996) Genetic data analysis II. Sinuaer Associates, Sunderland (MA)Google Scholar
- Zouros E, Foltz DW (1984) Possible explanations of heterozygote deficiency in bivalve molluscs. Malacologia 25:583–591Google Scholar