Study areas

The data were collated from a number of separate studies carried out between 1999 and 2010 and from across nine sea areas (Fig. 1; Table 1): the southern North Sea (ICES area IVc); the central North Sea (ICES area IVb); north-east Scotland (including the Moray Firth and Shetland Isles (ICES area IVa); the Viking Bank (ICES area IVa), the Scottish west coast (ICES area VIa); the Irish Sea (ICES area VIIa); the Celtic Sea (ICES areas VIIb and VIIf); and the western and eastern English Channel (ICES areas VIId and VIIe, respectively). These nine areas covered most of the distribution of cod around the British Isles, although the number of cod tagged and returned in each area varied resulting in some areas being better represented, for example the eastern English Channel, than others, for example the west coast of Scotland. Tagging was carried out mainly during the spawning season (classed here as from 1 January to 30 April).

Data storage tag (DST) and implantation

Data storage tags used in this study were either the Centi or Milli (Star-Oddi Marine Device Manufacturing Ltd, Gardabaer, Iceland), the LTD1200 or LTD1110 (Lotek Marine Technologies Inc., Ontario, Canada), or G5 (Cefas Technologies Ltd, Lowestoft, UK). Each DST was programmed to record depth and temperature every 5, 10 or 20 min depending on the memory capacity and battery life of the tag. Cod were captured using a modified commercial bottom trawl or by line in inshore areas where the seabed was too rough to trawl. Cod were anaesthetized in sea water containing MS222 or phenoxyethanol (Sigma-Aldrich Company Ltd, Gillingham, UK). For internal implantation of tags, a small incision was made in the belly along the mid-ventral line and a sterilized DST inserted into the body cavity. A plastic filament (~7 cm long) from the DST remained external and secured with a rubber bead. The incision was then sutured with two stitches using braided nylon (Ethicon Ltd, Johnson & Johnson International). A second conventional fish tag with return details (Floytag Mfg Ltd, Seattle, WA, USA) was sewn through the dorsal musculature. External tags were attached by threading nylon or monofilament line from the attachment points of the DST through the muscles anterior to the first dorsal fin and fastening the ends securely on the other side (Righton et al. 2010). The tagged cod were allowed to recover from the procedure in a tank and, provided they showed normal buoyancy control, were released overboard and the position recorded from the ship's GPS. A reward for return was offered to fishermen. In the United Kingdom, this research was carried out by Home Office licensed personnel under a number of Animals (Scientific Procedures) Act project licenses issued by the UK Home Office during the course and across the range of the study.

Estimation of geographic position

Hydrostatic (tidal) data, derived from the sinusoidal pressure cycle recorded in the depth data when a fish is at rest on the seafloor, were used to enable the geographical reconstruction of an individual cod's movements. This method of geolocation is referred to as the tidal location method (Metcalfe & Arnold 1997). We used a novel Fokker–Planck-based method that combines the tidal location method with a hidden Markov model to estimate, for each day at liberty, the nonparametric probability distribution of geographic position (Pedersen et al. 2008). The method was slightly modified for cod in the north-east Scotland, west coast of Scotland and Viking Bank areas to include an exponential probability density function that decayed as the height above the seabed increased. This was necessary to prevent the model returning geolocations at unrealistic heights above the seabed when the cod were in proximity to deep waters off the continental slope. The uncertainty associated with each daily geolocation mainly varied according to the quality of the tidal signal in the data and the geographical proximity to tidal amphidromic nodes (Pedersen et al. 2008).

Data analysis

Spatial analysis of the geolocation data was undertaken using ArcGIS 10.0. The data were subset and analysed separately for each release area and for the spawning period (1 January to 30 April) and out with the spawning period (1 May to 31 December). A bounding polygon was created encompassing all geolocation points within each data subset. Directional distribution ellipses were then created for each data subset. Centred on the mean position, these ellipses enabled identification of geolocations falling within one standard deviation from that mean, which were then used to create a second, inner, bounding polygon. For each individual, a measure of the straight-line distance travelled between successive geolocations (used to calculate average monthly movement rates), and between each geolocation and the release location (with the maximum used as a measure of migration distance) was calculated using the great circle equation. Individual movement paths were inspected and visually classified as home ranging or migratory (following the classification of Bunnefeld et al. 2011). Since most data sets were <1 year in duration and therefore did not encompass a full migratory cycle, migrations were further classified into feeding migrations (movement >100 km away from release area followed by residency) or spawning migrations (movement away from release area before return or movement to another area). In those cases where data sets were more than 250 days long, and therefore encompassed close to a yearly behavioural cycle, migrations were classified into homing migrations (return to release area at spawning time) or straying migrations (no return to release area at spawning time).

The extent of exchange of cod between areas was assessed by calculating the number of individual fish that were within a rectangle of 1° of longitude by 0·5° of latitude during the spawning period (defined as from 1 January to 30 April) and out with the spawning period (1 May to 31 December). Each cod was classed as being present or absent from a square for each day that it was at liberty. The proportion of cod from each area in each square during the spawning period and out with the spawning period was then calculated. Where only a single individual was recorded in a square, that square was excluded.

The temperature and depth data from each tag were summarized to give daily means and standard deviations for each fish. Mean monthly depth, mean standard deviation of depth and mean monthly temperature were then determined for each area. For those areas for which temperature data were recorded in each calendar month, the average monthly temperature per stock was extrapolated to a full year, to provide an estimate of degree days.

Southern north sea

Thirty-three of 59 cod from this area showed migratory behaviour (Table 1). The maximum migration distance travelled from the release position was 578 km, with the average migration distance at 177 km (±145 km). Of the migratory cod, nine moved westward into the English Channel, three moved eastward across the North Sea towards the Danish coast (for example Fig. S1, Supporting information), and ten moved northward into the central North Sea (for example Fig. S2, Supporting information) where they probably mixed with cod from that area (Fig. 3a,b). The 26 resident individuals remained within the southern North Sea, many within their immediate release area (37·3 ± 15·6 km). The extent of migrations gave the stock a latitudinal range of 5°. Home range was the greatest of all the stocks and was larger out with the spawning season (Table 2). The contraction in range at spawning time appeared to be even, and so the centres of gravity of spatial distribution were similar for both seasons. There was no evidence to suggest that southern North Sea cod migrated far enough north to mix with cod from the northern North Sea or far enough west to mix with those from the western English Channel or beyond.

Central north sea

Only five cod out of 22 released in this area showed migratory behaviour. One individual moved towards the Danish coast during the spawning season, and several cod moved into the nearby southern North Sea once the spawning period was over. This brought the centre of gravity for the feeding season to the south and east of the spawning location. The movement of this group of cod, together with the northward movement of the southern North Sea stock, resulted in a large overlap in home range between cod from these two areas (Fig. 3a and b). Northward movements extended to 57°N but not as far as the Viking Bank area. There was no evidence of movement into the English Channel. This stock had the narrowest latitudinal range of all stocks (0·8°). Spawning home range was marginally smaller than during the feeding season.

Viking bank

Five of 16 cod from the Viking Bank area showed migratory behaviour. Two cod made return north–south movements of over 300 km following the shelf break of the Norwegian trench. One individual crossed the Norwegian trench (Fig. S3, Supporting information) and did not return. It spent its second year at liberty on the Norwegian shelf break having travelled nearly 1000 km from its release site. This group had a latitudinal range of 4°. Two cod moved 150 km west towards the Shetland Isles where they could have mixed with cod tagged and released in that area.

North-east Scotland

Cod were released mainly in the waters around the Shetland Isles. Eleven out of 27 cod in this area showed migratory behaviour, three of which returned to the release location to spawn a year later. The cod moved up to 150 km west during the summer. A few immature cod tagged on the east side of Shetland moved offshore after the first year and into areas where cod from the Viking Bank area were present. The latitudinal range was approximately 3°, and the average migration distance for cod in the area was amongst the lowest (Table 2).

Scottish West Coast

There were too few cod tagged from the Scottish west coast to infer firm conclusions. Two cod of the four cod showed migratory behaviour, and there was evidence that cod from the Clyde Sea area may mix with cod from the Irish Sea (outside of the spawning season).

Irish sea

Twenty-five of 33 cod from the Irish Sea showed migratory behaviour. Five individuals moved north, which could have brought them into contact with cod from the southern part of the Scottish west coast and the firth of Clyde. Four individuals moved sufficiently far south to bring them into contact with cod from the Celtic Sea, and one individual migrated into the western section of English Channel during the spawning season. Latitudinal range was 5·3°, and the migratory distance of this group of cod was amongst the highest (Table 2). Relative to cod from other areas, the home range area of Irish Sea cod during the spawning season or feeding season was quite restricted.

Celtic sea

Twenty-three of 33 cod from the Celtic Sea showed migratory behaviour. No cod went north into the Irish Sea, although five individuals moved towards the western section of the English Channel. Individuals showed extensive and complex migrations, turning back on themselves numerous times, but returning to roughly the same area during spawning time (Fig. S4, Supporting information). Latitudinal range was approximately 3°, and the home range of these cod was large compared to cod from the neighbouring Irish Sea (Table 2). They overlapped in range with cod from the western and eastern channel and on occasion with cod from Irish Sea (Fig. 3a,b).

Western English channel

Twelve out of 15 cod from the western English Channel were migratory and had a relatively large home range. Those migratory individuals made extensive movements mainly to the west. Five individuals moved into the Celtic Sea area, one of which continued north to the Irish Sea. Others moved east to the Eastern Channel, but none as far as the southern North Sea. The overlap between cod tagged in the western English Channel and eastern English Channel was large and occurred mainly during the feeding season.

Eastern English channel

Twenty-five out of 43 cod tagged in the eastern English Channel were migratory and moved more than 100 km from release. They moved either west into the western Channel or Celtic Sea during the summer, or east into the southern North Sea. Most returned to near the release location or further east into the North Sea at spawning time. In consequence, this stock had the largest latitudinal range (nearly 7°). This stock had the second largest home range. In one case, a cod spent the summer in the Celtic Sea, before migrating to the southern North Sea in time for spawning (Fig. S5, Supporting information).

Environmental experience

Cod occupied average monthly depths of between 15 and 165 m. The average monthly depth of each area broadly reflected the local variability in bathymetry within the home range (Table 3, Fig. S6, Supporting information) and was thus generally <100 m in the shallow southern and central North Sea, but frequently greater than 100 m in the deeper northern North Sea and western areas. Cod experienced average monthly temperatures between 6 and 17 °C that clearly differed between areas (Table 3, Fig. S7, Supporting information). Cod from the Irish Sea, English Channel, southern North Sea and central North Sea all experienced strong seasonal cycles in temperature whereas those from the Celtic Sea and Viking Bank areas showed less seasonal variability. At the extremes of the thermal range (Table 3), cod from coldest area (the Viking Bank) experienced three-quarters of the annual number of degree days than cod from the warmest area (the central North Sea). Interestingly, however, there was no simple latitudinal gradient in temperature; cod from the Celtic Sea experienced an annual number of degree days lower than cod from the more northerly Irish Sea and central North Sea (Table 3).

Stock structure

Three largely discrete groups of cod were evident: (i) cod from the Viking Bank; (ii) cod from the central and southern North Sea; and (iii) cod from the western areas (Irish and Celtic Sea). While cod from the offshore Viking Bank area were isolated from southerly and western groups in the study, there was some overlap in range with cod from the adjacent Shetland Isles. This overlap may arise if the Shetland Isles represent a nursery area used by both local inshore cod and the offshore cod. Interestingly, one cod tagged on the Viking Bank moved into Norwegian coastal water, which raises the possibility that the Viking Bank cod could mix with, or even be the southerly extent of, a northerly population that extends to the Norwegian coast. That the Viking Bank area is home to a distinct population of cod suggests that it should be assessed and managed separately. This population grows more slowly and matures at a larger size and older age than other cod populations (Wright, Millar & Gibb 2011), and so, even if it is currently abundant (Holmes, Wright & Fryer 2008), it may actually be less resilient to fishing than other populations. Furthermore, it cannot be expected to repopulate the depleted southern or inshore areas, because Viking Bank cod do not range there, are genetically distinct and may even be adapted to a colder, more stable thermal environment.

The other major separation appears to lie between the western areas (Clyde, Irish and Celtic Seas) and the North Sea areas. This is again consistent with the finding of significant genetic distinction between these western groups and the North Sea groups (Heath et al. 2013). There were too few data in our study to assess a potential northerly exchange via the west coast of Scotland to the North Sea, and further work is needed to address connectivity of cod found to the north-west of Scotland and those in the North Sea.

Movement of individuals between areas was high among neighbouring areas in coastal areas. Only two individuals (from the western and eastern English Channel) were observed to move beyond a neighbouring area during the spawning period. These individuals, however, were tagged during the winter when they may have been returning to their natal spawning site. Within each of the main population units identified in this study, a small number of individuals made long-distance migrations, but the majority of individuals did not. This suggests an even finer ecological structuring may persist, perhaps similar to the ‘bay-scale’ population structure described for coastal Atlantic cod off Canada (Ruzzante et al. 2000). This type of population structure conforms to the concept of a meta-population (Kritzer & Sale 2004). For species that have a meta-population structure, the assumptions of what constitutes a stock from an assessment and management perspective become questionable. A ‘stock’ is assumed to be a discrete group of fish that show little mixing with adjacent groups and that have the same growth and mortality parameters across a particular geographical area (Gulland 1983). The present study suggests that the spatial scale at which this assumption remains valid may actually be finer than what even the genetic analyses suggest and further questions whether it is appropriate to apply the concept of a stock to species that exhibit a meta-population structure.

Thermal environment

The north-west European shelf represents one of the warmest areas for cod (Brander & Mohn 2004; Righton et al. 2010), and there was considerable thermal variation between study areas. This variation reflected depth, latitude and oceanic influence and may act to prevent a uniform cod stock from establishing across all areas. Although cod have been shown to be tolerant and flexible in response to thermal conditions, they may be thermally sensitive at spawning time (Righton et al. 2010). It may be that acclimation or adaptation to local thermal conditions constrains ranging and favours a resident strategy, especially if spawning and feeding grounds can be found in close proximity to one another (Neat et al. 2006). This may be crucial to the interpretation of reports of habitat shifting in temperate marine fish species (Dulvy et al. 2008). Although cod in the southern North Sea experienced summer temperatures considered super-optimal for the species (Neat & Righton 2007), they showed strong site fidelity to spawning grounds. Species like cod that are behaviourally constrained in their range shifting may turn out to be those that are most vulnerable to climate change (Heath et al. 2012).

Applied implications and conclusion

Failure to incorporate information on population structuring of marine fishes into fishery management risks sequentially depleting local population units and eventual stock collapse (Hilborn et al. 2003; Kritzer & Sale 2004; Heath et al. 2008). This is likely to have happened to cod around the British Isles over the last century where inshore stocks were first depleted and the fleet then moved progressively further offshore (Greenstreet et al. 2009). This study suggests that a finer scale of population structure exists in cod around the British Isles than is currently recognized in management and that each population unit experiences different thermal conditions that will affect physiology and growth. The applied implication of this is that local adjustment of exploitation strategies, such as the setting of maximum sustainable yield for each population unit, may be required to ensure sustainable harvesting (Holmes et al. 2014). Future management of fisheries should account for this current understanding of population structure and explicitly allow for local population recovery.