The annual cycle of many animals is characterized by the need to satisfy different life history priorities, often requiring seasonal movements. For such species, investigating carry-over effects (such as the year-long drivers of breeding success) and managing protected areas effectively, relies on quantifying these movements. Here, we model the seasonal movements of the UK population of grey seals Halichoerus grypus and show how insights from the model can improve its management.
We fit a hidden process model to two types of information – regional population redistribution and individual movements – to estimate the seasonal transition probabilities of breeding female grey seals among four regions around the UK.
We found that between 21% and 58% of females used different regions for breeding and foraging.
For our study period, we detected an increase in the breeding performance of animals that foraged in the Hebrides and South-East Coast.
Grey seal Special Areas of Conservation (SACs) were designed to encompass a significant proportion of the UK breeding population: ˜40% of the breeding females in our study area. Of the females breeding on SACs, only 15% breed in Northern Scotland, but up to 50% forage there. Our results indicate that, by only considering the breeding distribution of females that breed in SACs, the impact of anthropogenic activities on nearby SACs may be overestimated, whereas impacts on remote SACs may be underestimated.
Synthesis and applications. By quantifying the link between the foraging and breeding distributions of grey seals, management of breeding populations can be focused on the foraging regions where the resources necessary for reproduction are acquired. The construction of marine developments is dependent on demonstrating that they will not have an adverse effect on the integrity of Special Areas of Conservation (SACs), and we have shown that this requires consideration of the seasonal transition probabilities estimated in this study. Our specific results provide support for management strategies that jointly consider SACs and Marine Protected Areas (MPAs). More generally, we prescribe combinations of data on population size, breeding performance and individual movement that can enable our framework to be applied to seasonally migrating species.
The annual cycle of many animals must satisfy different life history priorities, and this often requires seasonal migrations. Unravelling the relationships between these priorities is critical for understanding population dynamics (Harrison et al. 2011) and therefore, for appropriate population management (Webster et al. 2002). For example, the seasonal movements of mobile species must be considered when defining population management units (Mauritzen et al. 2002), managing protected areas (Thirgood et al. 2004) and examining carry-over effects (the influence of conditions in one season on performance in the next – Bearhop et al. 2005). Understanding such movements is becoming increasingly possible with individually referenced data such as those obtained via telemetry (Bogdanova et al. 2011), mark–recapture (Studds, Kyser & Marra 2008), fatty-acid sampling (Walton, Henderson & Pomeroy 2000) and stable-isotope analysis (Bearhop et al. 2004). Because such data are costly and laborious to collect, they are usually derived from only small numbers of animals. Here, we show how population-level transition probabilities can be estimated by combining sparse individual-based data with population data, allowing fundamental applied questions to be addressed. Our objectives are to (i) develop a conceptual framework for quantifying seasonal movement between distinct geographical regions, (ii) apply this framework to a nontrivial case study, the metapopulation of grey seals Halichoerus grypus in the UK, (iii) illustrate how applied inferences can be drawn from such a model, and (iv) prescribe the type of data that must be collected to inform such models.
The relationship between sites where animals spend the nonbreeding and breeding seasons is particularly important in capital breeders because poor maternal condition cannot be offset by successful foraging during the breeding season (Boyd 2000). Grey seals are capital breeders and, in the UK, the links between their foraging and breeding areas are obscure (Thompson et al. 1996). For the majority of the year, they alternate foraging at sea with visits to land, to haul-out. In the breeding season (September to December), females produce a pup that is weaned after ~18 days (Twiss, Duck & Pomeroy 2003). Approximately 36% of the world's grey seal population breed in the UK (Duck 2010), and it is therefore essential to manage their potential interactions with human activities, such as fisheries (Matthiopoulos et al. 2008) and marine developments (Madsen et al. 2006). Quantifying the seasonal links between foraging and breeding areas will help address two key management issues. The first relates to the spatial drivers of inter-annual variation in pup production. Regional breeding performance can be the result of foraging performance in a number of different regions. Understanding the spatial drivers of breeding performance will allow an assessment of the importance of factors affecting pup production such as food availability, interactions with fisheries and other competitors, disease and at-sea disturbance. Most importantly, it will allow management to be focused where resources are acquired for reproduction. The second question deals with the potential effects of human activities on grey seal breeding populations within designated areas, specifically the Special Areas of Conservation (SACs) that have been established for grey seals in Europe to support the maintenance of Favourable Conservation Status (FCS; Council of the European Communities 1992). SACs for grey seals were selected to include large breeding colonies while covering the geographic breeding range in the UK. Before offshore developments can be licensed, it is necessary to show that they ‘will not adversely affect the integrity of the site [SAC] concerned’ (Council of the European Communities 1992). Pup production in SACs will be driven by the breeding condition of females established during the preceding foraging season (Boyd 2000). Thus, to appropriately determine the potential impact of such developments, the at-sea foraging distribution of females that breed in SACs needs to be quantified.
We develop a Bayesian hidden process model (Buckland et al. 2007) to estimate the transition probabilities of breeding female seals among four UK regions (Fig. 1) used for foraging and breeding at different times of the year. We fit the model to regional population estimates from the 2008 foraging season (Lonergan et al. 2011), regional pup production estimates from the two adjacent breeding seasons (2007 and 2008; Duck & Mackey 2008; Duck 2009), and to telemetry data on the transitions of individual breeding females. The model allows us to obtain point and interval estimates of the transition probabilities between foraging and breeding regions and to address our applied objectives, the stated management issues.
Table 1. The seven Special Areas of Conservation (SAC) in our study showing the estimated pup production and coefficient of variation (CV) for 2008. The 95% credible intervals of the number of animals which foraged in each region, estimated from the model, are shown
These colonies were assigned a CV of 0.01 on the basis that they were ground-counted colonies (see text).
The study area (Fig. 1) was split into four regions: Hebrides; Northern Scotland; East Coast and South-East Coast. The regions were selected to ensure main breeding, and haul-out sites were not adjacent to regional boundaries and to correspond as closely as possible to the seal management areas used by the Scottish Government (Scottish Government 2011). The regions included in this study encompass the majority of seals that breed in Britain and Ireland (93%; Duck & Thompson 2007; Ó Cadhla et al. 2007). We define the breeding season for each female as the time from pup birth to weaning, with the rest of the year defined as the foraging season. We only consider the movements of breeding females for three reasons: (i) grey seals are polygynous and thus pup production is unlikely to be limited by male abundance, (ii) the number of breeding females (but not adult males) can be estimated using pup production estimates, and (iii) males do not provision pups.
The foraging population of grey seals in each region in 2008 was estimated from aerial survey counts of haul-out sites conducted during the end of July and beginning of August. Counts in 2008 covered the entire study area, with the exception of Shetland (Northern Scotland) and parts of the west coast of Scotland (Hebrides) which were counted in 2007 or 2009. All counts were singular and did not distinguish between sexes or age classes. They were scaled to provide an estimate of population size and associated uncertainty using the estimates of the proportion of time seals spent hauled out during the survey period from telemetry data (Lonergan et al. 2011). These regional population estimates reflect the number of animals that use haul-out sites and are therefore only a proxy for regional foraging populations (see 'Discussion').
Grey seal pup production at a colony can be equated to the number of breeding females because, for the purposes of this study, we define breeding females as those that will produce a pup in the next breeding season. Pup production was estimated using data from survey counts. Most counts were conducted using aerial surveys, with the main colonies surveyed four or more times during the breeding season. Pup production in 2007 and 2008 and their coefficients of variation (CVs) were estimated for each colony (Duck & Mackey 2008; Duck 2009). In contrast to estimates from aerial surveys, ground-based counts did not have associated uncertainties. However, they are considered to be more precise. Thus, we assigned the lowest estimated CV obtained for the colonies which were surveyed by air (0.01) to the ground-counted colonies. A few colonies were not surveyed during 2007 and 2008, and so their population size in these years was extrapolated from counts outside the survey years. The regional CVs for each year were only based on counts in the year concerned. Nevertheless, over 95% of pup production was estimated from surveys conducted in the year of inference in both years and all regions. In 2008, c. 40% of pup production in the study area took place in the seven SAC colonies (Fig. 1, Table 1). The northern boundary of the Berwickshire and North Northumberland Coast SAC which contains the Farnes colony intersects an additional colony, Fast Castle. As the majority of this colony is outside the boundaries, it was not included as an SAC colony.
Telemetry data were used to identify transition events between the region in which a seal spent the majority of the year (foraging region) and the region in which it bred. The Sea Mammal Research Unit has deployed 109 telemetry tags on female grey seals, aged one or older. These transmit data on location and activity, including time spent hauled out and diving. Data from tags that were attached to a seal during both the foraging and breeding season were investigated. Breeding was identified and assigned to a region if a female was recorded as hauled out for the majority of an 18-day period during the breeding season and spent <10% of the time diving. This is a conservative definition of breeding because in some colonies, lactating grey seal females can spend <40% of their time on land (Caudron, Joiris & Ruwet 2001). Fifteen females, tagged between 1992 and 2008, met this criterion and each was assigned to a foraging region (see Appendix S1, Supporting Information). The duration of the pre-breeding tagging data for these animals varied between 58 and 207 days (median: 142). Seals that bred in a different region from the one in which they had been foraging did not leave their haul-out in the foraging region until seven to nine days before they were first located at their breeding colony. Therefore, we were able to include all 15 seals for which we had pre-breeding data confident that their pre-breeding region was a foraging region. For the two seals that had locations from more than one region, their assigned foraging region contained the majority (>95%) of locations.
Hidden process models allow both process (environmental, demographic) and observational uncertainty to be reflected in the parameter estimates (Buckland et al. 2007). We estimated scaling factors describing the proportion of the regional foraging populations in 2008 that produced pups in 2007 and 2008 and hence quantified the spatial and temporal variation in the breeding performance of foraging populations. We also estimated the regional foraging distribution of females which bred in SACs and compared this with their breeding distribution.
The number of breeding females (gi,t) in the ith foraging region in year t (t =2007, 2008) was estimated as a proportion (si,t; population scalar) of the total regional foraging population (χi; eqn 3) in 2008.
Breeding females from the ith foraging region have a probability (pj,i) of breeding in the jth breeding region. We first sought to inform these transition probabilities from the population data using the mismatch between number of breeding females in a region during the foraging (gi,t) and breeding season (mj,t; pup production). We modelled population-level movement as a deterministic process because, given the large sizes of the regional populations, a random process would not introduce noticeable demographic stochasticity. The number of pups (mj,t) produced in each breeding region is
Independently obtained regional foraging population estimates (Pi) reflect the population sizes (χi) in 2008 with an associated Gaussian estimation uncertainty () derived from Lonergan et al. (2011)
Similarly, the independently estimated pup production (Ψj,t) in the jth breeding region is a reflection of the number of pups produced (mj,t) accounting for Gaussian estimation error ().
From the telemetry data, the likelihood of the transitions can be obtained as follows: given a number of tagged females that are foraging in the ith region, a realization (Yi) of the transition process can be obtained from
We fitted the model simultaneously to the regional population estimates and the telemetry data using a joint likelihood (see Appendix S2). The final likelihood function (eqn 6) has five components: two Gaussian distributions; a uniform prior; a multinomial distribution and a nonstandard normalized beta prior:
Of the total number of females breeding in SACs in 2008 (mSAC), the proportion breeding in each region is denoted by αj and the regional proportion of breeding females that breed in an SAC is denoted by dj:
where mSAC,j is regional pup production in SACs calculated using observed pup production and uncertainty as in eqn 4. We then used the transition probabilities to estimate the proportion of females breeding in SACs that foraged in each region in 2008:
The model was fit, using Markov Chain Monte Carlo (MCMC) methods, in the software package WinBUGS (Lunn et al. 2000). Two chains were used with mixing and convergence assessed by visual inspection of the MCMC chains and the Gelman-Rubin diagnostic (Brooks & Roberts 1998). After an initial burn-in of 2 × 105 iterations, 107 iterations were generated.
Priors were required for the transition probabilities (pj,i) and population scalars (si,t). We had no a-priori knowledge of these transition probabilities, and we therefore needed to design diffuse joint priors conforming to a unit-sum constraint (see Appendix S3), which led to two normalized beta priors. The sensitivity of parameter estimates to these priors was investigated in simulations (see Appendix S3). The two priors produced similar results, and the Beta (α = 0.5, β = 2) was selected for the final analyses.
Some prior knowledge was available about the range of values for population scalars in 2007 and 2008. The mean value required to convert the total foraging population estimate across all regions in the summer of 2008 into the estimated number of pups produced (breeding females) in 2008 was 0.51 (95% CI: 0.42, 0.60). Fecundity estimates (Boyd 1985; Smout, King & Pomeroy 2010, 2011) were used to calculate the minimum and maximum values for the scalar in 2008, where fecundity is defined as the proportion of mature females that produce a pup. The resulting scalar range is larger than required to reflect variation in fecundity alone (see Appendix S3) and thus, although we do not explicitly account for the age/sex structure of the population, variation in these structures between the regions could also be incorporated within the range. The total population size is likely to have changed little between the summer of 2008 and the breeding season in 2008 (c. 2 months). However, this is unlikely to have been the case between the 2007 breeding season and the summer of 2008 (c. 10 months). Thus, for the 2007 scalar, the limits of the prior for the 2008 scalar were extended to allow for a change of upto 25% in the population size of any foraging region. An increase of 25% reflects the observed increase in the foraging population of the South-East Coast between 2007 and 2008, the region in which population size has increased most rapidly (SMRU, unpublished data). There are no data on the maximum decrease, so an allowance was made for a decrease between 2007 and 2008 of up to 25%. Within these limits, we had no knowledge of the likely values of the scalars, so we chose to use uniform priors of U(0.186, 0.883) and U(0.246, 0.698) for 2007 and 2008, respectively.
We identified the foraging and breeding regions for 15 tagged females (Table 2). Together with the data on foraging population size and pup production, the telemetry data allowed us to estimate transition probabilities between regions (Fig. 2). We found that although many seals bred in the same region in which they foraged, between 21% and 58% used a different region, with the degree of fidelity varying among regions (Fig. 2). The transition probabilities can be reversed to estimate the probability that an individual that bred in region j foraged in region i (Fig. 3).
Table 2. Seasonal transitions of breeding female grey seals between their foraging and breeding regions as indicated by telemetry data. The year in which the telemetry devices were attached are shown. The region assigned as the foraging region was the one from which the majority (>95%) of locations during the foraging season
2003 (n = 2)
2004 (n = 2)
1992 (n = 1)
1998 (n = 2)
1998 (n = 1)
2004 (n = 1)
2008 (n = 1)
2008 (n = 2)
2005 (n = 3)
We estimated population scalars, the proportion of the regional foraging populations in 2008 that were breeding females in 2007 and 2008, for each foraging region. In 2008, the foraging populations of the Hebrides and Northern Scotland had the highest proportion of breeding females whereas the South-East Coast had the lowest (Fig. 4). Changes in the population scalars between 2007 and 2008 reflect changes in the number of females in each foraging region that bred in 2007 and 2008 (Fig. 5). As such, they represent changes in the contribution of these foraging regions to the breeding population. There was a significant (at the 10% level) change in the population scalars between 2007 and 2008 in both the Hebrides and South-East Coast, with a median increase of 10% and 8%, respectively. We found that the proportion of females breeding in SACs that bred in each region differed from the proportion that had foraged there (Fig. 6). For example, 60–61% of females that bred in SACs bred in the Hebrides but only 41% (95% CI: 23, 57) foraged there. Incorporating the uncertainty surrounding both pup production estimates and the transition probabilities, we quantified the foraging distribution of females breeding in each SAC (Table 1).
We combined comprehensive population data and sparse telemetry data to quantify the regional movements of UK breeding female grey seals between the foraging and breeding season. We also, for the first time, described the movements of breeding female grey seals showing that if they bred in a different region from the one in which they foraged, they only leave their haul-out in the foraging region seven to nine days before they were first located at their breeding colony. The estimated transition probabilities can be used to address two important management issues: 1) the identification of spatial drivers of pup production and how these may vary over time and 2) the assessment of the potential effects of human activities on SACs. The wide confidence intervals surrounding our parameter estimates are a consequence of the small sample size of animals whose tags were operational in both the foraging and breeding seasons. When provided with similar, limited data (spatial variation in population scalars and transition probabilities; scenario Cii in Appendix S4), exhaustive simulations illustrated successful recovery of parameters providing confidence in the findings of our investigation with the seal data.
Our conclusions depend on four assumptions: (i) female grey seals spend the majority of the foraging season in one region, (ii) regional population estimates from haul-out counts are a reliable proxy for the regional foraging population, (iii) the transition probabilities remained constant between 2007 and 2008, and (iv) the observed transitions from telemetry data were a representative sample for 2007 and 2008. Within the foraging season, most trips (88%) return to the same haul-out (McConnell et al. 1999). Using telemetry data during the foraging season from 53 females (see Appendix S5), we found that more than 90% of the locations obtained from 48 (91%) females were within their primary region (the region that encompassed the majority of their locations). Furthermore, 43 (81%) females were only recorded in their primary region showing that females have high regional fidelity during the foraging season. We also found that haul-outs occurred in all regions which encompassed more than 10% of locations indicating that at the regional scale, haul-out population estimates are a reliable proxy for foraging population size. Constancy of transition probabilities over a 2-year period will be affected by the fidelity of individual animals to their foraging and breeding regions. Breeding female grey seals show strong site fidelity even when pupping is unsuccessful in the previous year (Pomeroy, Twiss & Duck 2000). We found high between-year fidelity to foraging regions: for the 11 study seals for which we had pre- and post-breeding data, all the pre- and post-breeding regions were the same. Furthermore, Vincent et al. (2005) found high inter-annual fidelity to foraging season haul-outs (females: 70–95%), with regional fidelity likely to be higher still.
Representativeness of the transition data (assumption iv) cannot be tested directly. However, even when grouped by their transitions, the extent of the regional movements of the 15 study seals was qualitatively similar and thus representative of the extent of the usage by all tagged seals (see Appendix S5). Thus, it is unlikely that the transitions of the study animals were not representative of the regional populations due to spatial variation within regions in transition probabilities. Whether the observed transitions were a representative sample of the population will also depend on how much the transition probabilities have changed during the 17 years of telemetry data. Although, as discussed earlier, transition probabilities of individuals are likely to remain constant, recruiting females may have different transition probabilities to those already in the breeding population. In the absence of population change, changes in transition probabilities caused by turnover of breeding females, through recruitment and death, would be gradual because females start to breed at about age 5 and can live for over 30 years. However, the population increased by 14% between 1998 and 2008, mostly as a result of increased pup production in the North Sea (Thomas 2011), which must have been a consequence of an increase in the recruitment of breeding females. Fortunately, most of the telemetry data (11 of 15 individuals) were collected within 5 years of 2008 and all but one seal were tagged between 1998 and 2008. Furthermore, there is evidence (Table 2) that transitions were constant through time in tagged animals which foraged in the Hebrides (2004 and 2005) and Northern Scotland (1992 and 1998). It is possible that due to rapid population change, the South-East Coast transition data are not representative of the population in 2007 and 2008. In this region, there is a mismatch between the increase in the foraging population (137%) and pup production (53%) between the year for which we have telemetry data (2005) and our year of interest (2008; SMRU, unpublished data). This indicates that some females that foraged in the South-East Coast must have bred elsewhere in 2008, but all three animals tagged in 2005 remained in the South-East Coast to breed. Nevertheless, the foraging population of the South-East Coast comprises only 11% of the total population, and thus, any temporal changes in transition probabilities would have little effect on transition probabilities for the other regions. The current lack of knowledge about movements of breeding females in the South-East Coast is reflected in the uncertainty estimates surrounding the transition probabilities.
Our approach can be applied to many other species that perform seasonal migrations. To facilitate the use of available data to address particular research questions, we prescribe five combinations of data requirements that could make use of our framework (Table 3). These are mapped to both estimable parameters and applied questions, with examples of species for which such data appear to exist. The types of data most practical to collect will vary with species. For example, in some species, population estimates may not be available, but with sufficient individually referenced data, the probability of animals moving to another region given their starting region can be estimated (conditional probabilities). If these telemetry data could be combined with spatially resolved population estimates from one season, population-level transition probabilities and thus spatially resolved population estimates in the second season could be estimated. Similarly, the applied questions that require addressing for effective management will vary with species.
Table 3. Data requirements for using this framework to answer both pure and applied research questions. The applied questions referred to here are (i) determination of the spatial drivers of breeding success, and (ii) management of protected areas. Complete movement data refer to a large sample of data originating from all study areas
Harp seals (Haug et al. 1994; Folkow, Nordøy & Blix 2004; Nordøy et al. 2008)
Population-level transition probabilities
This study (although we use amalgamated population-level data on breeding success)
Implications for management
In the UK, grey seals are primarily managed on a regional basis with sub-populations defined by discrete geographical groupings of breeding sites (Fig. 1; Scottish Government 2011). Population trends within management regions are currently monitored through pup production (Thomas 2011). We have illustrated patterns in seasonal movements which show that resources for reproduction may be being acquired in other management regions. These movement patterns therefore need to be considered in developing management strategies for grey seals. For example, up to half of the females breeding in Northern Scotland acquired their resources in the East Coast region and thus management of the Northern Scotland breeding population needs to take into account conditions in both regions. Such consideration of foraging areas is of particular importance in the North Sea where the Convention for the Protection of the Marine Environment of the North-East Atlantic have an Ecological Quality Objective that there will not be a decrease in pup production of ≥10% in any sub regions (OSPAR 2009), which include Orkney (Northern Scotland), Firth of Forth (East Coast), Farnes Islands (East Coast) and Donna Nook (South-East Coast). In addition, we quantified the spatial (resulting from variation in population structure or fecundity) and temporal (resulting from changes on population size or fecundity) variation in the breeding performance of foraging regions. Although the demonstrated temporal changes were only between 2007 and 2008, these are likely to be representative of recent trends as pup production levels in these years were in keeping with the regional trends (Duck 2009). Between 2007 and 2008, the proportion of the foraging population in 2008 that were breeding females increased in the Hebrides and South-East Coast by 10% and 8%, respectively, indicating that no change in management is required to maintain the status of these populations.
Before offshore developments can be licensed, it must be shown that they will not affect the integrity of SAC sites, no matter how far away these sites are from the proposed development (Council of the European Communities 1992). The competent authority primarily uses estimates of pup production in SACs near to the proposed development to assess the scale of any potential impact of marine developments on grey seal SACs. In addition, seal telemetry data from the foraging season are used to investigate whether seals that haul-out in an SAC use the area of potential marine development (connectivity with the SAC). However, as we have shown (Table S6, Supporting information), females are faithful to regions within a season and thus although useful, in most cases such data will underestimate the potential impact of developments on the breeding populations in SACs outside that region and over-estimate the potential impact within a region. Given that any effect on pup production is likely to be mediated through female condition, the foraging distribution of females that breed in SACs, also needs to be considered. For example, solely considering pup production within SACs in the Hebrides leads to an overestimation of the potential impact of marine developments in the Hebrides on the females which breed in SACs because even though about 60% of SAC pup production occurred there, only 41% (95% CI: 23, 57) of females that bred in SACs foraged there. The opposite was true in Northern Scotland where the proportion of females breeding in SACs that bred and foraged in Northern Scotland was 15% and 30% (95% CI: 12, 50), respectively. Thus, developments in Northern Scotland will have potential impacts on SAC populations outside that region. Specifically, although marine developments on the East Coast have the potential to impact the two SACs in that region, they will also have a potential impact on between 9 (c. 242) and 49% (c. 1374) of the females that breed on Faray and Holm of Faray SAC. The potential impact on this SAC must be considered during the consenting process for marine renewable developments in the East Coast region. Ultimately, the results of this study illustrate the difficulties in using terrestrial breeding sites, which we have shown, even at the regional level, do not equate to foraging sites, to assist in the maintenance of FCS. Thus, we argue that the designation of marine protected areas to protect the foraging hotspots of this species is also needed to assist the maintenance of FCS. Future work should target telemetry tagging of females prior to the breeding season, especially in the South-East Coast, to allow us to estimate transition probabilities more accurately and precisely. Although the telemetry sample size meant it was necessary to define large regions for this study, with foraging populations defined on the basis of haul-outs, once regions of interest are identified, at-sea usage emanating from these haul-outs based on telemetry (Matthiopoulos et al. 2004) can be examined. In combination with our results here, this may uncover more spatially resolved links between foraging regions and breeding performance and allow quantification of the overlap between marine developments and foraging space-use by animals that breed in SACs.
D.J.F.R is funded by the UK Department of Energy and Climate Change (DECC) as part of their Offshore Energy Strategic Environmental Assessment programme. For fieldwork and helpful discussions, we thank P.Hammond, G.Hastie, E.Jones, M.Lonergan, S.Moss, W.Patterson, P.Pomeroy, S.Smout, and L.Thomas, University of St Andrews, and B.Chapman, Lund University. We the thank M.Mangel, two other referees and the journal's editorial team for their constructive comments. Data from one of the tags were supplied by P.Thompson, University of Aberdeen. The telemetry tags were funded by DECC, the Natural Environment Research Council and Scottish Government.