The study area comprised the whole of the UK, Ireland and Isle of Man. This area was divided into regions for habitat preference modelling ( Figure 1 ). Regional boundaries were broadly based on the Seal Monitoring Units (SMUs) used in UK seal conservation and management (SCOS, 2020), but with some alterations to incorporate Ireland and the Isle of Man, and to maximise the predictive power of the models, accounting for the availability of tracking data ( Figure 2 ), as well as spatial differences in movement patterns and the heterogeneity of habitat, and ensuring that clusters of haulout sites were not split by regional boundaries. Each region was assigned a sequential number, used here throughout ( Figure 1 ).

Grey (n = 114: 45 M, 69 F) and harbour seals (n = 239: 107 M, 132 F) were tagged at 26 sites in the UK and Ireland between 2005 and 2019 ( Supplementary Material S1 ). All seals were estimated to be >1 year old based on mass and body length. Seals were captured on, or close to, haulout sites using seine, pop-up, tangle, or hand nets. Seals were instrumented with Fastloc^® GPS phone tags or dual tags (GPS-GSM-ARGOS) (SMRU Instrumentation, UK), glued to fur at the base of the skull following the procedure outlined by Sharples et al. (2012). For analysis in habitat preference models, data were cleaned following Russell et al. (2016) to remove erroneous location estimates and data during the breeding and moulting seasons (grey seals: September – April, harbour seals: June – August), regularised to a constant 2 h time step and partitioned into trips at sea. Trips were then assigned to one of the ten regional designations for habitat preference models (see Figure 1 ) based on the location of the haulout sites used at the start and end of each trip. At-sea locations were only included in the analysis if they belonged to trips that originated and terminated at haulout sites in the same region (grey seals: 93% of locations, harbour seals: 99.5% of locations). For detailed data cleaning and processing protocol, see Supplementary Material S2 . The resulting dataset for grey seals comprised 3,276 trips (61,296 temporally regularised at-sea locations) throughout the summer foraging season (May – August) ( Figure 2A ). The dataset for harbour seals comprised 8,579 trips (99,159 temporally regularised at-sea locations) throughout the autumn-winter-spring foraging season (September – May) ( Figure 2B ).

Under a “use-availability” study design (Johnson, 1980; Aarts et al., 2008; Matthiopoulos et al., 2020), accessibility polygons were generated for each haulout location in the tracking datasets, with a radius based on the maximum geodesic distance (shortest path at sea without crossing land) from a haulout by any seal in the cleaned tracking dataset (Russell et al., 2016) (grey seals: 448 km, harbour seals: 273 km), thus representing the species’ maximum foraging range (Wakefield et al., 2009). For each regularised seal location, a random sample of control points was generated within the corresponding accessibility polygon, representing the available habitat that is accessible to the tagged seal (Aarts et al., 2008; Beyer et al., 2010). The ratio of control points to presences was set using model coefficient stability analysis (Beyer et al., 2010; Ventura et al., 2019), and varied between the species and among regions from 15:1 to 25:1 (see Supplementary Material S3 for more details). In this use-availability framework, areas where seals go (presences; seal location data) are modelled alongside areas where they could go (control points) (Matthiopoulos, 2003). Preference for a particular habitat type is inferred where its use is disproportionate to its availability (Johnson, 1980; Matthiopoulos et al., 2020).

A range of static and seasonally dynamic environmental variables were extracted for each seal location and control point and included as candidate explanatory covariates in a maximal model for each species in each region ( Figure 3 ). These covariates included: distance to haulout site, water depth, substrate type, mean winter sea surface temperature (SST) lagged by one year, water column vertical stratification, Δ stratification (spatial variation in stratification values, analogous to frontal intensity), seabed slope, rugosity and proximity to the continental shelf break (see Supplementary Material S4 for more detail). Covariates were chosen because of known or potential biological relevance to seals and/or their prey, or to account for the effects of accessibility on habitat selection.

For each species-region combination, a binomial response term (0/1; control/presence) was modelled as a function of environmental covariates in a generalized additive mixed model (GAMM) using the package “mgcv” (Wood, 2020) in R (R Core Team, 2020). Individual seal ID was included as a random intercept. This meant that the relationships in the models were estimated across individuals rather than data points, ensuring that data-rich individuals did not have a disproportionate influence on the overall trends. Control points were weighted in the models according to the ratio of control points to presences, such that each set contributed the same as one presence. I.e., if the ratio of control points to presences was 20:1, then each control point would be weighted at 0.05 (1/20). Each continuous covariate was fitted as a smoothed term with shrinkage, such that uninformative terms can be penalised to zero, effectively making them linear (Wood, 2020). To ensure smooth functions reflected plausible biological relationships, the number of knots (k) was determined for each smooth by trialling different values (2≤k ≤ 10) and selecting that which minimised the model Akaike Information Criterion (AIC) score whilst still returning a relationship that was biologically interpretable. Maximal models were simplified to a minimal adequate model using a two-stage approach comprising: (i) backwards model selection by AIC score and (ii) k-fold cross validation ( Supplementary Material S5 ).

Haulout count data were used to weight model spatial predictions ( Figure 3 ; see below). These data comprised counts of grey and harbour seals on terrestrial haulouts from aerial and ground survey platforms, conducted during August ( Supplementary Material S6 ) (Morris and Duck, 2019; SCOS, 2020). In brief, Scotland (SCOS, 2020), Northern Ireland (SCOS, 2020), eastern England (with the exception of the Tees Estuary) (Cox et al., 2020; SCOS, 2020), and Republic of Ireland (Morris and Duck, 2019) were surveyed aerially. Data for the remaining areas (rest of England, Isle of Man and Wales) were comprised of ground and boat counts from multiple organisations (see Supplementary Material S6 for full list of sources). Counts were aggregated to 5 km x 5 km grid cells (hereafter ‘haulout cells’). Survey effort varied spatially; for this study the most recent available counts were used for each haulout cell. The vast majority of counts (94.4%) were recorded from 2015 - 2018 ( Figure S6.1 ).

A prediction grid was generated on a 5 km x 5 km cell resolution encompassing the at-sea area accessible to seals from all haulout cells in the study area (UK, Isle of Man and Ireland). Environmental data corresponding to the modelled covariates were extracted for each cell (see Supplementary Material S7.1 ). Data for non-static covariates were extracted for 2018. Spatial predictions of relative at-sea density (mean and associated lower and upper 95% confidence intervals) were generated per species using the corresponding region-specific habitat preference model ( Figure 3 ). Predictions were made emanating from each haulout cell with a non-zero count in the most recent survey ( Figure S6.2 ). Raw predictions (on the logit scale) were exponentiated, then normalised (Manly et al., 2002; Matthiopoulos et al., 2020). Haulout-specific prediction surfaces were then weighted by the number of individuals counted on the most recent survey. These weighted prediction surfaces were then summed into a multi-region surface and normalised. The mean distribution estimates therefore represent the percentage of the at-sea population for the UK and Ireland (i.e. excluding hauled-out animals) estimated to be present in each cell at any one time (see Supplementary Material S7.2 for more detail). Thus, cells in the mean relative density distribution map sum to 100%. Cell-wise lower and upper 95% Bayesian credible intervals (hereafter confidence intervals) were generated using a posterior simulation approach (Wood, 2006; Augustin et al., 2013) ( Supplementary Material S7.3 ). For some applications, estimates of absolute density (i.e. number of seals) are favourable to relative density. Here, we present distribution maps as relative density (see “Technical Considerations” in Discussion for rationale). However, the mean prediction can be converted to absolute density using population scalars derived from telemetry data, and this process is explained in Supplementary Material S7.4 . In brief, haulout counts are scaled to total population size for UK and Ireland using the mean estimated proportion of the population hauled-out during the survey window (and thus available to count). Total population size is then scaled to at-sea population size using the mean estimated proportion of time seals spend at-sea during the months that the maps represent (i.e. excluding breeding and moulting). Relative density estimates (percentage of the at-sea population) can then be converted to absolute density using the estimated at-sea population size. However, an important consideration with these absolute density estimates is that uncertainty in the scalars is not propagated through to the confidence intervals around the mean, which only reflect uncertainty in the habitat preference relationships.

Using the spatial prediction approach described above, distribution estimates can be made emanating from specific haulout cells, allowing users to apportion at-sea abundance to specific areas, such as designated sites. To investigate the at-sea distribution of seals that haul-out in SACs, at-sea density estimates (mean and associated lower and upper 95% confidence intervals) were generated, for both species, for haulout cells within each of the SACs in the UK and Ireland for which the species is a primary or qualifying feature for designation using the associated regional habitat preference model. As above, estimates were normalised to represent the percentage of the at-sea population for that SAC present in each grid cell (i.e. relative density), thus cells in the mean distribution map sum to 100%, representing at-sea distribution of all individuals hauling-out in the SAC during the main foraging season. To demonstrate how these estimates may be used in an applied sense to apportion at-sea seal density to an SAC, grey seal estimates for both the Isle of May SAC and the Berwickshire and North Northumberland Coast SAC were converted to absolute density using the most recent counts (SAC totals) and population scalars (Russell et al., 2015; Russell and Carter, 2021) ( Supplementary Material S7.4 ). The number of seals estimated to be present within a windfarm footprint (Neart na Gaoithe; currently under construction 15 km off the east coast of mainland Scotland) was estimated for (a) the UK and Ireland population, (b) seals hauling-out at Isle of May SAC, and (c) seals hauling-out at the Berwickshire and North Northumberland Coast SAC. Cell values in the absolute density layers were weighted by multiplying the number of seals by the proportion of the cell that overlapped the windfarm polygon (i.e., a cell with an estimated ten seals present, but with only a tenth of its area overlapping the polygon would be counted as one seal). Area-based (i.e., windfarm footprint) 95% confidence intervals were estimated by adapting the posterior simulation process (see Supplementary Material S7.3 ).

Grey seals aggregate in large numbers to breed, and SACs are often designated for this species based on breeding numbers, but there may be temporal variation in abundance at these sites. To investigate the relative importance of these sites for grey seals in both the breeding and foraging seasons, the most recent available counts (2019) during the breeding season (pup production estimates; SCOS (2021)) for frequently monitored sites (Scotland and eastern England) were compared to counts of grey seals hauled-out during the harbour seal moult surveys in August (representing the main foraging season for grey seals). Counts for SACs were calculated as percentage of the area-wide totals (i.e., including monitored sites outside of SACs). To update the knowledge base on seasonal movements of grey seals presented by Russell et al. (2013), tracking data from females tagged in this study were examined for evidence of breeding. This analysis focussed on females that were tagged in and/or likely bred in SACs. Following the protocol developed by Russell et al. (2013), a female was assumed to have bred if it was recorded as hauled-out for the majority of an 18-day period during the breeding season and spent <10% of this time diving. Breeding locations were then compared to tagging locations (individuals were tagged during the summer foraging season) to examine seasonal variation in SAC use.

Habitat preference modelling revealed regional and inter-specific differences in environmental drivers of distribution. Below we outline some key results with the covariates retained in each regional model shown in Table 1 , but full results and associated plots of modelled relationships from each region are given in Supplementary Material S8 (grey seals) and S9 (harbour seals).

For grey seals, distance to haulout site was the primary driver of distribution in all regions, with predicted density declining with increasing distance. In addition to distance to haulout, the following covariates were retained in the models: substrate type (Regions 1, 2, 3, 4, 6, 7, 8), water depth (Regions 2, 4, 5, 6, 8, 10), stratification (Regions 2, 3, 4, 6, 8), mean winter SST lagged 1 yr (Regions 1, 5, 6, 7, 10), Δ stratification (Regions 1, 5), shelf edge (Regions 5, 10) and rugosity (Region 6) ( Table 1 ). The best model for grey seals in Region 1 (Southeast England) according to model selection criteria revealed a positive influence of Δ stratification, with seals preferentially selecting habitat in areas with high frontal intensity ( Figure S8.1 ). Mean winter SST (lagged 1 year) was also retained in the model, and preference peaked at values between 6-8°C. The model also revealed an effect of substrate type, with sandy mud having a negative effect on preference. Grey seals in Region 4 (North Coast and Northern Isles) demonstrated a more coastal distribution than in other regions, having a negative association with areas >100 km from the haulout ( Figure S8.4 ). The best model for grey seals in Region 5 (Western Isles) revealed a positive association with shelf edge; areas at the shelf edge had a stronger influence on habitat selection than areas on-shelf (proportionate to their availability), although confidence intervals overlapped ( Figure S8.5 ). Indeed, nine (60%) of the tagged individuals in this region performed repeated trips to the shelf edge. A similar relationship with shelf was found for grey seals in Region 10 (West Ireland; Figure S8.9 ).

As with grey seals, distance to haulout site was also the primary driver of distribution for harbour seals in all regions. The following covariates were also retained in the models for harbour seals: Δ stratification (Regions 1, 3, 6), mean winter SST lagged 1 yr (Regions 1, 2, 4), substrate type (Regions 5, 6) and rugosity (Regions 1, 7) ( Table 1 ). The best model for harbour seals in Region 1 revealed a similar but stronger relationship to that of grey seals for mean winter SST, with preference peaking at values between 5-7°C. However, in contrast to grey seals the best model also revealed a negative influence of frontal intensity ( Figure S9.1 ). Water depth was also an important driver of distribution for harbour seals, with predicted density declining with increasing depth in Regions 1, 2, 3, 5, 7 and 9. A negative effect of distance to haulout >130 km was predicted for harbour seals in Region 1, but other regional models revealed a much more coastal distribution, with a stronger negative effect for lower values of distance to haulout. For example, in Regions 2 (East Coast; Figure S9.2 ) and 4 ( Figure S9.4 ), the best model revealed a strong negative association with areas >50 km from the haulout. In Regions 5 and 10 where grey seals showed positive association with the shelf edge, harbour seals had a tight coastal distribution; the best model for Region 5 revealed a strong negative association with both distance to haulout and depth ( Figure S9.5 ), and the best model for Region 10 revealed a strong negative association with distance for areas >20 km from the haulout ( Figure S9.9 ).

Figures 4 , 5 show the mean and associated cell-wise 95% confidence intervals of the at-sea distribution of grey and harbour seals hauling-out in the UK and Ireland according to the combined predictions from the best model for each region. Large areas of relatively high at-sea density are apparent for grey seals adjacent to centres of abundance in Orkney, Northeast England, Southeast England and the Western Isles ( Figure 4 ). The maps also reveal hotspots of grey seal density offshore from Southeast Scotland and Northeast England, along the southern fringes of the Dogger Bank, and at the shelf edge west of Scotland ( Figure 4 ). High at-sea density areas are revealed for harbour seals adjacent to relatively large centres of abundance in Shetland, The Wash (Southeast England), and West Scotland ( Figure 5 ).

At-sea distribution estimates for grey seals hauling-out in SACs ( Supplementary Material S10.1 ) revealed that hotspots of density occur offshore, often >150 km from the SAC itself. This is particularly apparent for SACs in regions where seals travel further offshore, for example the Humber Estuary SAC in Southeast England, and the Monach Islands SAC in the Western Isles ( Figure 6 ). In general, at-sea distribution estimates of harbour seals ( Supplementary Material S10.2 ) were much more tightly concentrated in waters surrounding the SAC, with hotspots of density extending outwards to ~50 km from the SAC boundaries.

SACs represent 42.8% of pup production in Scotland and eastern England, and host 30.4% of individuals during the foraging season, with all but one SAC (Humber Estuary) having a higher proportion of pup production than summer count ( Table 2 ). Some SACs host large breeding numbers but only a small number in the foraging season (e.g. Isle of May: 2.9% of the pup production but only 0.1% of the summer haulout count).

Analysis of absolute density emanating from both the Isle of May and the Berwickshire and North Northumberland Coast SACs revealed large differences in the number of animals estimated to be within the footprint of the Neart na Gaoithe windfarm at any one time. Of the 148 (81-204 95% CIs) seals estimated to be within the footprint, 68.2% (101: 48-116 95% CIs) can be apportioned to the Berwickshire and North Northumberland Coast SAC (~50 km from the windfarm), and 1.4% (2: 1-2 95% CIs) can be apportioned to the Isle of May SAC (~15 km from the windfarm) ( Figure 7 ).

Of the 69 grey seal females tagged in this study, 12 transmitted data through to the breeding season, showed evidence of breeding activity and were tagged in and/or likely pupped in SACs. Of the 12, ten were tagged and likely pupped in an SAC, one was tagged in an SAC but appeared to pup outside of an SAC, and one was tagged outside of an SAC but appeared to pup in one. Half of the ten females that were tagged and likely pupped in SACs appeared to pup in a different SAC to the one in which they were tagged. Three of these movements constituted a transition between different Seal Monitoring Units (SMUs) ( Table 3 ). For two of the females that had different pupping and tagging sites, the tags continued to transmit after breeding and the tracks revealed the seals returning to the tagging site ( Figure 8 ).

This study presents the first habitat-based at-sea distribution estimates for the entire UK and Ireland populations of grey and harbour seals. Critically, region-specific habitat preference models were used to allow for spatial variation in the species-environment relationship around the coast. UK and Ireland-wide predicted distributions revealed large areas of relatively high at-sea density of grey seals adjacent to large haulout clusters in Southeast England, Northeast England, Orkney and the Western Isles ( Figure 4 ). Hotspots of density were also predicted offshore along the continental shelf edge west of Scotland, off the east coast of Scotland and off the east coast of England extending out to the southern fringes of the Dogger Bank in the central North Sea ( Figure 4 ). The regional habitat preference model for grey seals in Southeast England revealed a positive association with areas of strong heterogeneity in water column stratification values ( Figure S8.1 ), analogous to areas of high frontal intensity. The Flamborough tidal mixing front is a persistent frontal system that occurs from late spring to early autumn off the east coast of England, extending offshore from Flamborough Head and encircling the Dogger Bank (Hill et al., 1993). Our results suggest that this feature is an important driver of distribution for grey seals hauling-out in Southeast England. This finding supports the results of Wyles et al. (2022) who showed that the fringes of the Dogger Bank represent an important foraging area for grey seals hauling-out in Southeast England. Being primarily a bottom feature with weak surface signature (Hill et al., 1993), this frontal system may be particularly attractive to grey seals, which predominantly target benthic and demersal prey (Gosch et al., 2019; Wilson and Hammond, 2019). Tidal mixing fronts can alter the density and behaviour of prey and thus may represent a temporally and spatially predictable area of high foraging success for predators (Cox et al., 2018). Similar tidally driven processes may occur at the continental shelf break, where upwellings from deeper water may provide predictable foraging opportunities (Cox et al., 2018). While persistent surface fronts, detectable with satellite imagery, have been shown to correlate with the foraging behaviour of seabirds (Scales et al., 2014; Grecian et al., 2018), our results demonstrate the importance of also considering vertical water column structure as a potential predictor of habitat use for aquatic predators.

Harbour seal distribution was predominantly driven by a negative association with increasing distance from haulout and water depth. As such, predicted density is more tightly concentrated in coastal and inshore waters compared to grey seals, but with fine-scale structuring of density around the coast ( Figure 5 ). The exception to this is in Southeast England, where the negative associations with distance and depth were weaker. High at-sea density areas of seals hauling-out in this region extended further offshore than elsewhere around the coast ( Figure 5 ). In general, the influence of dynamic covariates (SST, stratification and Δ stratification) was variable among the regions. We found relatively high densities of grey seals predicted in coastal waters ( Figure 4 ) in areas where harbour seal populations have suffered the most rapid declines: including Orkney and the east coast of Scotland which have exhibited declines of over 80% between 2000 and 2016 (Thompson et al., 2019). Competition with grey seals for prey resources is one of several putative drivers of regional harbour seal declines (Thompson et al., 2019). Although comparing inter-specific differences in habitat associations and space use is an important step in quantifying the potential for competition, inference of inter-specific competition for prey resources is limited by a number of factors. For example, high levels of overlap in distribution may equally be indicative of exploitative competition, or sufficient prey resources to support both species. Similarly, low levels of overlap may result from competitive exclusion or spatial partitioning of foraging resources between the species. Moreover, it is impossible to disentangle spatial variation in prey preference or prey distribution from competition without examining changes in foraging habitat preference and distribution through time. This study considered overall habitat preference (i.e. not discriminating by behaviour) in order to capture the overall distribution of seals, since all forms of habitat use are of relevance for marine spatial planning. A priority for future work is to identify foraging behaviour using robust statistical models (such as hidden Markov models; Carter et al., 2016), which will facilitate a comprehensive study of foraging habitat preference and may provide more insight regarding the dynamics of competition between the two species.

The habitat preference approach used here provides some key improvements over previous large scale distribution estimates for grey and harbour seals (i.e. usage maps (Matthiopoulos et al., 2004; Jones et al., 2015)). For example, where no tracked seal visited a specific cluster of haulout sites, predictions emanating from the haulout sites were based on null usage; a simple study-wide relationship between distance from haulout site and observed usage (Jones et al., 2013; Russell et al., 2017). In our approach, such predictions are based on a region-specific species-environment relationship, which includes distance to haulout and any other covariates deemed to be good predictors of habitat preference by the model selection process. As such, areas of suitable habitat not visited by the tagged seals may still be detected, whereas they may be overlooked or underrepresented in the usage maps. Nevertheless, in data-rich areas (e.g. harbour seals in Southeast England), distribution patterns resulting from the track smoothing approach may be favourable to habitat-based predictions since they are able to describe features of the distribution which may not be captured by modelled habitat preference relationships averaged across individuals and dependent on relatively simple descriptors of habitat. Future work should seek to combine these two modelling frameworks to maximise robustness and ecological insights; previous work has been restricted to a small, discrete spatial extent (Jones et al., 2017). The usage maps have been an important resource for management applications, and they have also been used in ecological studies (Sadykova et al., 2017; Sadykova et al., 2020). For example, Sadykova et al. (2017) inferred habitat preference relationships for grey and harbour seals from the usage map distribution estimates presented by Jones et al. (2013). However, such inferred relationships, and resulting ecological conclusions, may be subject to distortion due to the reliance on null usage and the assumption that habitat preference does not vary among regions. Moving forwards, the maps presented here offer an updated resource for marine spatial planning, with increased potential for ecological insights on both regional and population-wide scales.

While the distribution estimates presented here represent a valuable resource for both ecology and management, as with any predicted species distribution, there are considerations which should be acknowledged depending on their application. Below we outline four important such considerations. (i) As with the “usage maps” (Jones et al., 2013; Jones et al., 2015; Russell et al., 2017), the upper and lower 95% confidence intervals presented in this study are generated on a cell-wise basis, meaning that they should not be summed across an area (e.g., windfarm footprint), as doing so would lead to inflated uncertainty estimates. Specific area-based confidence intervals can be generated on a case-by-case basis, as demonstrated in the Neart na Gaoithe windfarm example presented here. (ii) The latest count for each haulout cell was used to weight the distribution estimates, and these are taken from different days, and in some cases different years around the coast ( Figure A6.1 in Supplementary Material ). There may be spatial as well as day-to-day variation in the proportion of the population hauled-out during the surveys, and such variation is not considered here. (iii) For some applications, such as identifying areas of ecosystem-level importance, it should be noted that the estimates presented here represent the at-sea distribution of seals hauling-out in the UK and Ireland and are not inclusive of seals from haulouts in continental Europe. (iv) There are additional considerations when using absolute density estimates, and thus relative density estimates should be used where possible (e.g. for comparison of density between areas). Uncertainty in total abundance (population size) for the study area could not be propagated. Furthermore, the proportion of time spent hauled out, and thus the at-sea population size, varies across the focal months (Russell et al., 2015) but was assumed constant (averaged) to generate maps of absolute abundance.

The distribution estimates presented here provide an opportunity to link changes in abundance detected at land-based monitoring sites with areas at sea where such changes are likely to be mediated. The ability to draw such links is particularly pertinent for SACs, which provide monitored conservation areas with management efforts focussed on maintenance of (or recovery to) Favourable Conservation Status (FCS). By extension, knowledge of the at-sea distribution of individuals hauling out in an SAC is critical to predicting and managing the potential impact of offshore activities on the integrity of the SAC. Furthermore, in instances of declining SAC abundance, particularly for SACs no longer in FCS, these maps give guidance on where such change may be mediated. The results show that, unless the SAC encapsulates the majority of the abundance within a region (e.g. Monach Islands SAC), UK and Ireland-wide distribution maps may not be representative of SAC-specific distributions. For example, the Firth of Tay and Eden Estuary SAC has seen a steep decline in harbour seal abundance over the last two decades (Thompson et al., 2019). As such, overall distribution patterns in this area are largely driven by individuals hauling-out elsewhere within the region, and hotspots of density are not representative of the distribution of individuals using the SAC ( Figure S10.2 ). Thus, the locations of processes occurring at sea most likely to impact the SAC population cannot be easily identified using the UK and Ireland-wide maps. However, an important consideration is connectivity between haul out sites; unlike true central place foragers (e.g. breeding seabirds) which always return to the same site between foraging trips, seal foraging trips often start and end at different haulout sites. Indeed, only 35% of grey seal trips and 44% of harbour seal trips analysed here started and ended at the same 5 km x 5 km grid cell. These differences in central place strategies mean that seals may utilise multiple designated and non-designated sites throughout the foraging season. Inter-regional movements within the foraging season are more limited (Russell et al., 2013) particularly for harbour seals (Carroll et al., 2020), thus we advise that stakeholders use the SAC-specific maps in conjunction with the overall prediction for the area to identify both primary and potentially secondary (through use of multiple haulouts) areas of linkage with SACs.

In addition to the management implications of movements within the foraging season (discussed above), there are particular complexities for grey seals associated with the relationship between breeding and foraging distributions on land. Grey seals are classic capital breeders and aggregate in large numbers during the breeding season (Bonner, 1972). SACs are generally designated based on breeding numbers, and population monitoring is focussed on trends in pup production (Russell et al., 2019). However, grey seal use of sites on land varies between the foraging and breeding season on two spatial scales. Indeed, breeding sites do not always represent a central place for foraging ( Table 2 ). For example, the Isle of May SAC contributes 2.9% of pup production for Scotland and eastern England with >1,800 pups born annually (almost a quarter of the total pups born in SACs in this area), but hosts just 0.1% of seals hauled-out in this area during the foraging season (most recent count = 40) ( Table 2 ). In such cases, hotspots of density in the area are unlikely to be attributable to individuals hauling-out at the SAC despite the large reported population size (based on breeding numbers). Our analysis reveals that only 1.4% (n=2) of seals estimated to be present within the footprint of the Neart na Gaoithe windfarm at any one time can be apportioned to individuals hauling-out in the Isle of May SAC (~15 km away), while the majority (68.2%; 101) can be apportioned to the Farne Islands within the Berwickshire and North Northumberland Coast SAC ~75 km away ( Figure 7 ). Any decline in the Isle of May breeding population would therefore not be easily attributed to processes occurring at sea, since the non-breeding distribution of these seals is largely unknown.

Grey seals also exhibit seasonal redistribution at broader spatial scales. Our analysis reveals that, of the 12 females that transmitted data through to the breeding season and were tagged in and/or likely bred in SACs, six (50%) bred in different SACs to where they were tagged, with four (33%) breeding at colonies in different Seal Monitoring Units (SMUs) ( Table 3 ). Although these results come from a relatively small sample size, they support the findings of Russell et al. (2013) that 21-58% of breeding females use different regions for breeding and foraging. This highlights the level of uncertainty associated with linking abundance trends observed on grey seal breeding colonies to processes occurring at offshore foraging areas. Partial migration, where a population is comprised of migratory and resident individuals, is a widespread phenomenon in the animal kingdom (Chapman et al., 2011), but has not been widely identified in seals. Evidence from other terrestrial-breeding marine predators has demonstrated that migratory strategy may influence fitness; Grist et al. (2017) showed that breeding success was higher for non-migrating than migrating pairs of European shags (Phalacrocorax aristotelis) breeding on the Isle of May. In grey seals, the drivers of this behaviour are likely related to an inclination to natal philopatry and fidelity to breeding sites following recruitment (Pomeroy et al., 2000). Indeed, the rate of increase in pup production in Southeast England over recent decades lagged behind the rate of increase in seals counted during the foraging season (Russell et al., 2019) likely due to seals foraging in the south, where trends in abundance indicate conditions are favourable compared to further north where a proportion returned to breed. Further work is required to determine the implications of partial migration in grey seals for breeding success, population dynamics and conservation management.