How fast does a seal swim? Variations in swimming behaviour under differing foraging conditions

Breath-holding divers must meet several challenges if they are to forage effectively in the marine environment. Firstly, they must balance the time spent exploiting underwater food resources with time spent unloading CO₂ and loading oxygen at the surface. Secondly, they have to operate in a three-dimensional environment where food resources may be scattered in aggregated patches, sometimes at great depths. This means that at the end of a surface period, they have to either relocate the previously visited patch or search for another more profitable patch(Ydenberg and Clark, 1989). Most optimal foraging models assume that the number of prey encountered during a feeding event is a linear function of the time spent foraging, i.e. foragers should maximise time spent feeding in the prey patch. To do this, they should reduce both their transit time to the feeding area and their recovery time between two successive feeding events (e.g. Ydenberg and Clark, 1989). When marine divers feed at greater depths, the time spent travelling to depth and hence the energy cost increase. Consequently, the expected net foraging gain of a dive decreases with increasing depth, and the time–energy trade-off becomes more constrained, thereby limiting the behavioural options available to the diver (Charnov,1976; Kramer,1988; Houston and Carbone,1992; Beauchamp et al.,1992).

The rate of oxygen consumption during a dive is directly proportional to the rate of energy expenditure, which is in turn a function of the swimming speed (Davis et al., 1985; Feldkamp, 1987; Fedak et al., 1988; Thompson et al., 1993; Stelle et al., 2000; Rosen and Trites, 2002). The swimming speeds employed during a dive will have a major impact on the rate of depletion of limited oxygen reserves (Davis et al., 1985; Williams et al.,1991; Thompson et al.,1993; Wilson et al.,2002). Effectively, in order to maximise prey ingestion while minimising the cost of transport, breath-holding divers are expected to modulate their swim speed, body angle and swimming pattern(Dunstone and O'Connor, 1979; Sato et al., 2003). Thompson et al. modelled how the optimal foraging tactics of seals may change as a function of the interactions between physiological constraints (cost of swimming) and constraints of prey availability(Thompson et al., 1993). Their optimality model suggested that seals should swim at the minimum cost of transport (MCT) speed in deep dives, but in shallower dives they could increase the proportion of time spent at the foraging area by swimming faster between the surface and the prey patch.

Animals swimming in the water column are affected by buoyancy, which has been reported to significantly affect the diving behaviour of seabirds and marine mammals (Lovvorn and Jones,1991; Webb et al.,1998; Skrovan et al.,1999; Beck et al.,2000; Williams et al.,2000; Biuw et al.,2003; Watanuki et al.,2003; Sato et al.,2003; Miller et al.,2004; Sato et al.,2007). In diving air-breathing animals that do not trap air in thick fur or feathers, or that regularly dive below the depth at which lung collapse occurs, the net buoyancy is largely determined by the relative amounts of low-density lipid and high-density lean tissues in the body. Individuals that are positively or negatively buoyant expend more energy to maintain a position in the water column than individuals of the same species that are neutrally buoyant (Lovvorn and Jones, 1991), and net buoyancy might be expected to directly influence the speeds and swimming modes during descent and ascent swimming(e.g. Webb et al., 1998; Beck et al., 2000). However, in studies on free-diving animals, it is often difficult to distinguish between the effect of buoyancy and effects of drag and motivational state, and results from these studies are therefore often inconclusive on this point.

This paper describes the swimming behaviour of grey seals in relation to food resource availability (prey density and patch distance) in an experimental set-up that provided us with a unique opportunity to remove the effect of buoyancy on dive behaviour. We describe the swimming patterns employed during foraging dives. We use observed behaviours to examine whether divers decrease their swim speed as patch distance increases, as predicted by the swimming behaviour model of Thompson et al.(Thompson et al., 1993). As all dives are horizontal rather than vertical and therefore not affected by buoyancy, we hypothesise that the descent and ascent swimming mode and speed will be similar in any particular dive.

Seven female grey seals Halichoerus grypus Fabricius 1791 (five juveniles <1 year of age and two adults) were used in this study. All seals were caught in the wild at Abertay Sands (Fife, Scotland, UK) and kept in outdoor seawater pools in the captive facility at the Sea Mammal Research Unit(SMRU, St Andrews, UK). Seals were fed daily on herring Clupea harrengus and sandeels Ammodytes marinus. They were regularly weighed (to ±0.1 kg) and body composition was determined periodically using isotopically labelled water(Speakman, 2001; Sparling et al., 2006). Seals were released back into the wild after a maximum of 10 months in captivity. All captures and handling procedures occurred between 2002 and 2004 and conformed to the Animals (Scientific Procedures) Act 1986 under project licence number 60/2589.

The SMRU experimental facility comprises a large seawater pool (40 m×6 m×2.5 m) that can be covered by aluminium meshed panels to restrict animals to surface only in a breathing box linked to an open circuit respirometry system. The pool could also be longitudinally partitioned with nets into lanes so that the distance between the breathing box and a simulated foraging patch (i.e. travel distance) could be varied between 40 and 120 m(Fig. 1). As phocid seals have no air in their fur and routinely exhale before diving, they have few compressible spaces, so that change in depth has only a small effect on buoyancy and no discernible effect on shape or drag. Net buoyancy effects due to high or low body density will approximately cancel out during the descent and ascent phases in vertical dives, so horizontal, underwater swimming to and from the feeder will be energetically similar to vertical diving to an equivalent depth. Hereafter, travel distance from the breathing box(equivalent to the surface) to the feeding patch (equivalent to the bottom)will be referred as `descent' and `ascent'.

The foraging patch consisted of a purpose built device that delivered food on a conveyor belt to a feeding window 2 m below the surface (see Sparling et al., 2007). An important aspect of the design was that seals were free to dive at will and select their own foraging behaviour.

The experimenter controlled the prey encounter rate (PER) by varying the spacing of prey items on the conveyor belt. PER was held constant within a given dive, but changed randomly between dives. PER varied between 0 and 13.8 fish min^–1. The upper limit of PER corresponds to the highest PER recorded in the wild with video cameras attached to freely diving harbour seals Phoca vitulina feeding on sandeel(Bowen et al., 2002).

Foraging behaviour was investigated in relation to both prey density and the distance between the prey patch and the breathing box. Trials were conducted with the feeding station positioned 40 m, 80 m and 120 m from the breathing box. Five of the seven seals were tested at all three distances while two of the juvenile seals were only tested at 80 m. Animals were fasted for >15 h before each feeding trial.

The start and the end time of each dive and times of arrival at and departure from the feeder were recorded from direct observation at the breathing box and an underwater video system at the feeder. These recorded observations represent the visual data. Swim speed was estimated as distance between breathing box and feeder divided by time between observed departure and arrival. Time–depth recorders (TDR; MK 8, Wildlife Computers,Redmond, WA, USA), attached to the seals' heads, provided an independent measure of swimming speeds and of durations of travelling and surface periods. The MCT speed was estimated from the metabolic rate–swim speed relationship for grey seals swimming in a flume tank(Fedak et al., 1988; Thompson et al., 1993). The measured MCT speed was approximately 1.3 m s^–1 and unrelated to body mass.

PER and mass as continuous variables and patch distance as a factor (40 m,80 m, 120 m) were included as fixed effects and seal ID and PER were included as random effects. This allows the model to fit separate slopes and intercepts for the relationship between index and PER for each seals. We used swim speed data obtained visually from five seals travelling to 40, 80 and 120 m to carry out this analysis. Deletion tests were used to assess the significance of each parameter in the models.

Swimming behaviour was investigated using Mk8 TDRs that incorporated a turbine swim speed sensor. The TDR was set to log velocity every second. Each TDR velocity meter was calibrated for both pups and adults by recording the times taken to swim fixed distances along lanes in the pool. Swim speed profiles were plotted and examined visually for each dive for each animal(Fig. 2).

Swim speed profiles were characterised by a `descent' phase (active swimming from the breathing box to the feeder), a phase of foraging at the feeder, and an `ascent' phase (active swimming from the feeder to the breathing box). The start and end of the dive were determined both by the wet/dry sensor of the TDR and by a sudden change in swim speed between 0 and 1.5–2 m s^–1 within 2 s. The foraging phase was characterised by stationary periods or relatively slow swimming in the vicinity of the feeder with occasional bursts of activity. The `ascent' phase was similar to the `descent' phase with a rapid increase in swim speed as seals left the feeder and a decrease upon arrival to the breathing box. For each dive, we calculated the mean speeds and the number of acceleration/deceleration phases (resulting from so-called `burst-and-glide'swimming) during the `descent' and the `ascent' phases. Speeds are presented as means ± 1 standard deviation (s.d.).

Because seals were swimming horizontally, buoyancy cannot directly explain potential differences between `descent' and `ascent' speeds in our study. However, it is possible that seals may employ a different level of swimming effort during `descent' and `ascent' as a conditioned response to their actual body condition and buoyancy. To test this hypothesis, we estimated the buoyancy for each seal through the year where body composition measures were available.

A total of 3220 dives obtained from visual data were analysed for seven female grey seals (two adults: L and Q, and five pups: K, N, R, W and X). Dive characteristics recorded during these experiments were similar to those recorded in the wild (for details, see Sparling et al., 2007).

A summary of the swim speed characteristics is shown in Table 1. Overall mean `descent'and `ascent' speeds were 1.71±0.41 m s^–1 and 1.37±0.36 m s^–1, respectively. Mean `descent' swim speeds were faster than the estimated MCT speed (1.3 m s<sup>–1</sup>)(Thompson et al., 1993) for all seals except for the adult Q. Mean `ascent' swimming speeds were faster than the calculated MCT speed for all pups (range 1.53–2.00 m s^–1), while for adults, ascent swimming speeds were close to or less than MCT speed (Table 1).

Of the 3220 dives for which swim speed was calculated from visual data,independent swim speeds were also obtained from the TDR data for 1289 dives. There was a strong positive correlation between the visual and TDR data(correlation= 0.786, Z=53.51, P<0.001), but the speed estimates obtained from visual data were significantly higher than those from TDR records (paired t-test, T=25.17, P<0.001)(Fig. 3). The variance in the relationship is due to measurement error in recording the start and end of active swimming and seals slowing or stopping when out of view during transit to and from the surface. Visually recorded swim speed therefore provides a noisier index of true swim speed, while TDR-derived data may be less noisy but may slightly underestimate the true swim speed. While this will reduce the statistical power of any comparisons and increase the probability of type II error it does ensure that any observed relationships are likely to be real,i.e. there is little chance of type I error.

A total of 1672 dives obtained from visual data for five female grey seals(L, Q, K, N and R) were used to create linear mixed effects models. X and W only swam to 80 m, therefore their data were not use in this analysis. A summary of the parameters obtained is presented in Table 2.

The full mixed effects model including data from all five seals suggested that body mass had a significant negative effect on swim speed. However, it was clear that this was being driven mainly by the mass change of one large pregnant adult female `Q', whose swim speed index decreased throughout the study (Fig. 4). This was also apparent when comparing swim speeds to different distances. During the early pregnancy of seal Q, swim speed index to 80 m was 0.5 and in late pregnancy the index fell to –0.6 (Fig. 5). Seal Q was therefore considered separately and removed from the overall analysis.

Four seals (three pups, N, R, K, and one adult, L) were therefore included in the final model. Results are summarised in Table 2.

There was significant variation between individual seals, but the model indicated that seals did not adjust their transit swim speeds in response to changes in either PER (Fig. 6)or body mass (Fig. 4). However,they did appear to adjust their swim speed in response to patch distance(Table 2, Fig. 7). Transit swim speeds decreased significantly with increasing patch distance. This decrease in swim speed was most pronounced between 40 m and 80 m, with little change between 80 m and 120 m.

In order to describe the swimming patterns of our animals, swim speed profiles obtained from TDR records of 1289 dives were examined from all seven seals (L, Q, K, N at 40 m and 80 m and R, W, X at 80 m).

The mean `ascent' speed was always significantly slower than the mean`descent' speed for all seals, at all distances (paired t-test, P<0.01). Interestingly, all seals swam between 10% and 20% slower on their way back to the breathing box despite not having to work against buoyancy in either direction (Fig. 8). We calculated the buoyancy through the year for each seal where body composition was available. The mean body mass of the adult grey seal was 129.7 kg and 34 kg for the pups, with adipose tissue accounting for between 7.1% and 40.8% of body mass (Table 3). There was no relationship between estimated buoyancy and the relative swim speed during `descent' and `ascent'. `Ascent' swimming was always slower than `descent' swimming whereas buoyancy varied widely from +16 N to –20 N.

All seals used burst and glide swimming during both `descent' and `ascent'. Repeated acceleration (stroke) and deceleration (glide) phases were apparent in swim speed records in all dives (Fig. 2). The number of burst and glide phases was generally significantly higher on the `ascent', except for N and R at 80 m (paired t-test, P<0.05, Fig. 9).

This study examines the swimming behaviour of grey seals in relation to resource availability, i.e. prey density and patch distance. The experimental set-up allowed us to test theoretical models of swimming speeds in breath-holding divers. In addition, we investigated `descent' and `ascent'swimming in the absence of buoyancy effects. The study was based on a sample of seven female grey seals, a relatively large sample compared to previous captive marine mammal studies. Although the sample size restricts our ability to investigate size-related effects, the behaviour of the non-pregnant adult was qualitatively similar to that of the pups throughout.

The observed mean swimming speeds for both adults and pups were similar to those reported for adult grey seals foraging in UK waters(Thompson et al., 1993) but were higher than those of adult grey seals foraging around Sable Island(Beck et al., 2000). However,Beck et al. reported `descent' and `ascent' rates that are only analogous to swim speed if the seals were diving vertically to the bottom(Beck et al., 2000). At any other angle, such rates would underestimate swim speed.

Marine mammals and penguins should swim at or near the MCT speed when swimming to and from the surface in order to maximise the amount of oxygen available during the foraging phases of dives(Davis et al., 1985; Feldkamp, 1987; Ponganis et al., 1990; Ponganis et al., 1992; Fish, 1993; Thompson et al., 1993; Williams et al., 1993; Ropert-Coudert et al., 2002). The MCT speed for grey seals swimming in a flume tank was approximately 1.3 m s^–1 (Fedak et al.,1988; Thompson et al.,1993), which was similar to those recorded for harbour seals in similar conditions (between 0.85 and 1.4 m s^–1)(Davis et al., 1985). On average, our seals swam approximately 20% faster than the estimated MCT speed when returning to the surface (Table 1), and over 40% faster than the expected MCT speed when going to the feeder. Estimates of MCT speed based on animals swimming in flume tanks may be underestimates. Seals had to swim actively against the flow while breathing at the surface. Drag is greatly enhanced at or close to the surface(Hertel, 1966) so seals in flume tanks would experience higher drag than during submerged swimming at similar speeds in our set-up or in the wild. At higher speeds this effect is exacerbated by seals spending proportionately longer breathing(Fedak et al., 1988). Our seals may therefore have been swimming at or close to MCT speeds during`ascent'. If so, however, they must have been swimming faster than MCT speed when travelling to the feeder.

Our largest animal was approximately six times heavier than our smallest so we would expect its MCT to be only around 5% higher. This is consistent with previous studies on animals in the wild showing that swim speed is relatively constant at around 1–2 m s^–1and not correlated to body mass over a wide range, from 30 tonne sperm whales to 0.5 kg seabirds(Ponganis et al., 1990; Sato et al., 2007).

Although dive duration is ultimately limited by oxygen stores, it has been suggested that seals may alter their diving behaviour in response to their perception of both the quality and depth of a prey patch(Thompson and Fedak, 2001). Grey seals do alter their dive durations in response to changes in patch quality, by ending their dives early at low prey densities(Sparling et al., 2007). The results presented here suggest that transit swim speeds are not simply related to prey density. This is consistent with the assumption that seals will maximise prey acquisition by maximising the rate of delivery of oxygen to the foraging patch irrespective of patch quality.

Swim speeds in deep dives should approach, but never go below, MCT speed,whereas in shallow dives higher swim speeds would allow animals to maximise the proportion of time spent foraging at the bottom(Thompson et al., 1993). All seals tested in this study did alter their swimming behaviour in response to changes in patch distance, swimming faster to 40 m compared to 80 m or 120 m.

Sparling et al. found that dives to 40 m were generally much shorter than the estimated aerobic dive limit (ADL; equivalent to the oxygen stores divided by the metabolic rate) but in dives to 120 m, seals were approaching ADL at the highest prey densities (Sparling et al., 2007). Reducing swim speeds and therefore metabolic rates during transit in deep dives would allow seals to spend longer at the feeding patch without exceeding their estimated ADL. In contrast, studies on Northern fur seals, New Zealand sea lions and Brunnich's guillemots found that swimming speed during descent increased significantly in deeper dives(Ponganis et al., 1992; Crocker et al., 2001; Lovvorn et al., 2004). Unlike phocid seals, these species store air in their lungs, fur or plumage and therefore have to work hard against buoyancy at the start of the dive. In shallow dives a higher proportion of the descent is spent working against buoyancy so that the apparent drag forces experienced by the animal will be greater in shallow compared to deep dives.

Differences between `descent' and `ascent' swim speeds are usually explained in terms of changes in buoyancy forces(Webb et al., 1998; Williams et al., 2000; Beck et al., 2000; Sato et al., 2003). However,despite the fact that all swimming between the surface and the foraging site was horizontal in our study, seals nevertheless swam slower on `ascent' than on `descent' despite not having to work against negative buoyancy(Table 1 and Fig. 7). This difference between `descent' and `ascent' speeds was maintained with increasing patch distance and there was no indication that the slower `ascent' swimming was a conditioned response to perceived buoyancy. Seals were choosing to swim faster to their feeding patch. Buoyancy cannot therefore completely explain why negatively buoyant seals swim more slowly during `ascent'. Motivational state may have a direct effect on swim speeds; seals may swim faster to the feeding patch in anticipation of finding food while they might save energy for the next dive by swimming slower on their way back.

All our seals used burst and glide swimming during both `descent' and`ascent' (Fig. 2). This is possibly an energy-efficient way of travelling for marine mammals(Lovvorn et al., 1999; Williams et al., 2000; Lovvorn et al., 2001). Data from TDRs in this study indicate that there were fewer acceleration and deceleration phases during the faster `descent' compared to the slower`ascent' (Fig. 8). Seals might have increased the frequency and/or the amplitude of their strokes to increase their speed on the `descent', but our TDR records were not sensitive enough to detect individual swimming strokes. Several studies have suggested that speed and acceleration are mediated via changes in stroke amplitude rather than stroke frequency (Lovvorn et al.,1999; Wilson and Liebsch,2003; Lovvorn et al.,2004). In order to determine in detail the swimming tactics used by the seals in our set-up, however, we would need to use more precise accelerometer devices.

In summary, the results of the present study indicate that swim speeds in grey seals are closely related to resource accessibility, i.e. distance, but not to the patch quality. Seals adjusted their swim speeds in relation to dive distance allowing them to increase their time spent foraging underwater. In addition, our unexpected discovery that seals swim slower on their way back to the surface in the absence of buoyancy effects suggests that the swimming behaviour exhibited by foraging grey seals during vertical swimming is primarily dependent on behavioural choices rather than a result of buoyancy effects (Fedak and Thompson,1993; Thompson and Fedak,2001; Sparling et al.,2007).

List of abbreviations

 
• ADL

  aerobic dive limit

 
• COT

  cost of transport

 
• DLV

  diving lung volume

 
• ID

  identification

 
• MCT

  minimum cost of transport

 
• PER

  prey encounter rate

 
• TBF

  total body fat

 
• TDR

  time–depth recorder

 
• TLC

  total lung capacity

We are grateful to Simon Moss for his expertise in animal care and handling. Special thanks to Stuart Middlemas for insightful comments on various drafts of this manuscript. We also wish to thank the two anonymous reviewers for their helpful comments. This work was funded by a Natural Environmental Research Council small grant to D.T. and M.A.F.(NER/B/S/20002/00300), and a Marie Curie Post-Doctoral Fellowship(HPMF-CT-2000-00816) to J.Y.G.