Do seasonal changes in metabolic rate facilitate changes in diving behaviour?

Diving animals frequently show flexibility or variability in their diving behaviour and performance. This variability can be observed within individuals(Beck et al., 2000), among individuals (Ropert-Coudert et al.,2003) or between populations(Tremblay and Cherel, 2003). The determinants of this variability are diverse. For example, variability in prey species may require a modification in dive profile(Takahashi et al., 2003), or,on a larger scale, temporal variability in the location of prey can lead to changes in foraging location and hence preferred diving depths(Charrassin and Bost, 2001). Some changes in diving behaviour have a physiological basis. Larger individual Magellanic penguins (Spheniscus magellanicus) tend to dive deeper and for longer than smaller birds (Walker and Boersma, 2003). The buoyancy, depth profile, stroking patterns and efficiency of dives by grey seals (Halichoerus grypus) and Weddell seals (Leptonychotes weddellii) are dependent partly on body condition (Beck et al., 2000; Sato et al., 2003). In both of these species, leaner animals tend to stroke less frequently and descend more rapidly than fatter, more buoyant animals and either dive for longer(Beck et al., 2000) or have a shorter recovery time at the surface between dives(Sato et al., 2003).

Green et al. (in press a)showed that the foraging and diving behaviour of macaroni penguins[Eudyptes chrysolophus (Brandt 1837)] varies substantially throughout their annual cycle. Specifically, during the winter months, the penguins tend to dive to greater depths and for greater durations than during the summer. However, in an earlier study of their diving energetics, Green et al.(2003) showed that female macaroni penguins may dive close to the limits of aerobic metabolism during the summer months. Calculated aerobic dive limit (cADL) is a quantity often used to assess the energetic cost of dives(Butler and Jones, 1997). cADL is the time at which an animal is estimated to have exhausted its usable oxygen stores while submerged, and is calculated from the total usable oxygen stores divided by a measure of the rate of oxygen consumption(V̇_O2) while submerged. In macaroni penguins, during the summer, cADL was estimated to be 138 s, and 95.3% of all dives were less than this threshold(Green et al., 2003). However,during the middle of winter, mean dive duration of foraging dives by female penguins was approximately 143 s(Green et al., in press a). This suggests either that the previous estimate of cADL was incorrect or that,during the winter months, the penguins were either subject to considerable ecological constraints that forced them to dive beyond their aerobic limits or else their physiological capability to dive increased.

In the present study, we compare diving behaviour and energetics during the summer/pre-moult foraging trip and winter/migratory period of macaroni penguins in order to determine what might cause the observed changes in diving behaviour. Specifically, we ask: (1) when do the observed large-scale seasonal changes in diving behaviour occur; (2) is variation in diving behaviour caused by variation in the energetic cost of diving, or vice versa; and (3)what physiological and/or ecological factors might cause variation in the behaviour and energetics of diving?

Study animals were captured and equipped with data loggers (DLs) at the British Antarctic Survey (BAS) base on Bird Island, South Georgia. All birds used in the experiments were breeding adults from the macaroni penguin colony at Fairy Point on the north side of Bird Island. Although the United Kingdom Animal (Scientific Procedures) Act 1986 does not apply to South Georgia, where this work was conducted, we followed its provisions, especially those set out by the Home Office in the Official Guidance on the operation of the Act. As our benchmark, we followed guidance to researchers using similar methods in the United Kingdom. Our procedures also conformed to the SCAR Code of Conduct for Use of Animals for Scientific Purposes in Antarctica.

Deployment of DLs took place during the austral summers of 2001/02 and 2002/03, and retrieval of DLs took place during the austral summers of 2002/03 and 2003/04. Study birds were identified and captured using the procedures described by Green et al.(2004), and DLs were surgically implanted into the abdominal cavity using previously established techniques (Stephenson et al.,1986; Green et al.,2003). Long-term implantation of these DLs has previously been shown to have no detectable adverse effects on the behaviour, breeding success and survival of this species (Green et al., 2004). In 2001/02, DLs were deployed during the chick rearing phase in January and February (N=39). In 2002/03, DLs were implanted during the incubation phase in November (N=19), chick rearing period in January and February (N=12) and moult phase in March(N=12). DLs were retrieved in the breeding season following implantation. In the 2002/03 breeding season, 34 of 39 penguins with DLs returned after the winter. In the 2003/04 breeding season, 43 of 43 penguins with DLs returned after the winter. In both years, return rates were not significantly different to control groups or previous data for this colony(Green et al., 2004). In both seasons, DLs that had failed during the winter migration were removed during the courtship phase, while DLs which had not failed during the winter were removed during the following moult phase. DLs were removed using the same procedures as during implantation and, after recovery, the penguins resumed their normal activities.

The DLs used in the present study were designed by one of the authors(A.J.W.) and were the same instruments (Mk 3a, Mk 3b) used to study year-round diving behaviour of macaroni penguins(Green et al., in press a). Mk 3a instruments were used for all deployments during the 2001/02 season. Mk 3a instruments were also used for 13 of the deployments in the 2002/03 season,with Mk 3b used for the remainder. Both instruments had dimensions of 36×28 ×11 mm and weighed 18 g before and 21 g after encapsulation in paraffin wax to provide waterproofing and a silicone coating for biocompatibility. Both instruments could record hydrostatic pressure for conversion to dive depth, heart rate (fh), body attitude(upright or prone) and abdominal temperature (T_ab). Mk 3a DLs had a 32 Mb memory capacity and were programmed to record dive depth every 2 s, fh and body attitude every 10 s and T_ab every 30 s for 453 days. Mk 3b DLs had a larger memory capacity (64 Mb) and were programmed to record dive depth every 1 s, fh and body attitude every 10 s and T_ab every 15 s for 542 days. Mk 3a DLs had a depth resolution of approximately 0.3 m, while a technical problem with the Mk 3b DLs meant that, although they had a depth resolution of approximately 0.09 m,they failed to record further depth changes deeper than approximately 25 m(Green et al., in press a). All DLs were individually calibrated and had a temperature resolution of approximately 0.06°C. The time that each logger was started, implanted,removed and stopped was carefully noted as GMT. After retrieval, data from all DLs were downloaded onto a computer before being transferred to a UNIX workstation or PC for further analysis.

While evaluating dive records, dives with maximum depths of <2.4 m were ignored during analyses since, between the surface and this depth, wave action, recorder noise and the interaction between temperature and pressure degraded depth accuracy making it impossible to accurately characterise dives this shallow. No distinction was made between foraging and non-foraging(travelling or searching) dives, and hence all dives with maximum depth of×2.4 m were used in all analyses. In total, 1 616 403 dives were analysed and, for each dive, maximum depth, dive duration and subsequent surface duration were extracted.

Data were prepared and analysed using purpose-written computer programs within the SAS statistical package (version 8.2; SAS Institute, Cary, NC,USA). As in previous work (Green et al.,in press a), careful observation of the breeding behaviour of the individual penguins and preliminary analysis of the depth and body attitude data allowed each day of the deployment for each penguin to be assigned to a phase of the annual cycle and a day since the start of that phase for that bird (phaseday). These phases were: (1) incubation trip, (2) brood, (3)crèche, (4) pre-moult trip, (5) fail (foraging behaviour during the breeding season following the failure of a breeding attempt), (6) winter. In the current project, only data from the pre-moult trip (N=54 birds)and winter (N=46 birds) phases were used for analysis and are referred to as summer and winter, respectively. The data were further filtered to reduce variability. The duration of the pre-moult trip varied slightly between individuals, so only the first 16 days of the pre-moult trip were used, as this was the minimum duration of this phase in all birds. For winter,the first 13 days of winter were removed as, until this time, some birds still returned to the colony regularly and had not engaged in their full migration,despite the completion of moult. The duration of the winter phase was limited to 190 days, as this was the minimum duration of this phase. For analyses involving dive depth, only data from the 2001/02 season were used(N=31 for summer, N=26 for winter).

The body mass (M_b) of both male and female macaroni penguins fluctuates during the breeding season, by a factor of up to two(Croxall, 1984). M_b at the end of winter is substantially greater than at the beginning (Croxall, 1984)but, other than this, M_b is unknown during the winter period. To understand the relative changes in metabolism, it is beneficial to calculate the rate of oxygen consumption(V̇_O2),independently of the potentially confounding effects of fluctuating M_b and body size. In recent years, there has been much attention on the correct way to account for the effects of M_b in physiological investigations (e.g. Packard and Boardman, 1999). Unusually, in macaroni penguins, the mass exponent of V̇_O2 is one,whereas there is no relationship between heart rate (fh)and M_b (Green et al.,2001; J. A. Green, C. R. White and P. J. Butler, unpublished data). We found that using mass-specific V̇_O2(sV̇_O2) accounts for all of the variation between individuals in V̇_O2 at a given fh in this species and allows the construction of predictive relationships between fh and sV̇_O2(Green et al., 2001). Therefore, in the current study, the use of mass-independent sV̇_O2 not only accounts for variation in body size and mass between individuals but also allows us to account for the confounding effects of variable body mass within individuals at different times of year. sV̇_O2 (in ml min^-1 kg^-1) was estimated from fhusing the equations of Green et al.(2005), and the standard error of the estimate (s.e.e.) was calculated following the method of Green et al. (2001).

In an effort to evaluate potential changes in minimum or maintenance metabolic rate (MMR), a running average of sV̇_O2 was computed for each 12-min period throughout each day. The period with the minimum value was assigned as the basal rate of oxygen consumption for that day(sV̇_O2min). Twelve minutes was selected as the interval to evaluate sV̇_O2min, as it was the inflection point in a plot of running average size against minimum fh (Withers,2001). In an effort to correct for the possible effects of anapyrexia (a regulated decrease in body temperature) on MMR,sV̇_O2min was normalised to a temperature of 39°C(sV̇_O2minC). sV̇_O2minC was calculated according to van't Hoff principles, assuming an apparent Q₁₀ of 3 (Heldmaier and Ruf,1992).

Mean abdominal temperature was calculated for each day. To gain a better understanding of the many physiological processes leading to changes in body temperature, a running average of T_ab was calculated every 12 min for each day for each penguin. The minimum and maximum value for each day for each animal were then extracted (T_ab,min and T_ab,max, respectively). T_ab,max was used to represent the normal, core body temperature as it was independent of the effects of circulatory changes, regional hypothermia or metabolic suppression associated with diving(Ponganis et al., 2003; Butler, 2004). These factors can conspire together to reduce T_ab, and the magnitude and duration of this decrease are dependent on the duration of diving bouts(Green et al., 2003).

sV̇_O2 while submerged cannot be measured during dives(Costa, 1988), but sV̇_O2DC, the sV̇_O2 of a complete dive cycle (dive plus the subsequent surface period), can be estimated from mean fh of that dive cycle(Fedak, 1986; Bevan and Butler, 1992; Butler, 1993). Calculated aerobic dive limit (cADL) is usually calculated by dividing useable oxygen stores by an estimate of V̇_O2 while submerged (Butler and Jones,1997). Green et al.(2003) used sV̇_O2DC as a measure and indicator of V̇_O2 while submerged and showed that cADL is not necessarily a fixed quantity. In macaroni penguins, cADL increased with increasing dive duration, because sV̇_O2DC decreased with increasing dive duration. As a result of this, in the present study, cADL was calculated, at each observed dive duration, and cADL described in the text is the threshold at which dive duration exceeded the cADL calculated at that dive duration. cADL was calculated using oxygen stores of 58 ml kg^-1, as in a previous study(Green et al., 2003). 95%confidence intervals of cADL were calculated by repeating this process but substituting sV̇_O2DC with the upper and lower 95% confidence intervals of sV̇_O2DC at each dive duration.

Estimates of sV̇_O2 are quoted in the text ± s.e.e., and comparisons of these data were made using Woolf's test for differences, which is most appropriate for the analysis of data derived from a predictive relationship (R. L. Holder,personal communication). All other data were analysed using general linear model (GLM), and means are quoted as ± s.e.m. Results were considered significant at P<0.05, and the significance of statistical tests is quoted in the text.

Both sexes showed clear differences in their diving behaviour between summer and winter (Figs 1, 2). Both mean and maximum dive durations were greater during winter than summer. In the winter, mean and maximum dive durations started at a level around 10–15% higher than that during summer. Both quantities then increased steadily until around day 50 of winter, when they reached a stable level approximately 35% greater than that during summer. These stable levels were maintained until around day 100 of winter, whereupon they declined over the next 20 days to a second stable level, slightly lower than before but still 20–25% greater than those during summer (Fig. 1). Although there was considerably greater variability, maximum dive depth showed a similar pattern of change (Fig. 2A). Mean dive depth also showed a similar pattern of change but was up to 50% greater during the middle of winter than during summer. During late-winter, mean dive depth appeared to be closer to mean dive depth during summer (Fig. 2B).

To investigate these differences in more detail, mean and maximum dive duration and depth were derived for each bird for the summer (276 641 dives)and for the middle of the two stable periods during mid- and late-winter(phasedays 69–85, 121 681 dives; phasedays 149–165, 113 174 dives,respectively). These periods are illustrated in Fig. 1. There was no difference between the sexes in mean and maximum dive duration and depth. Mean dive duration was significantly different between each of these periods (GLM with Tukey post-hoc tests; F_2,113=95.1, P<0.0001; Table 1). Mean maximum dive duration was also significantly different between each of these periods (GLM with Tukey post-hoc tests; F_2,113=154.7, P<0.0001; Table 1). Mean dive depth was significantly greater during mid-winter than during summer and late-winter(GLM with Tukey post-hoc tests; F_2,113=12.5, P<0.0001; Table 1). Mean maximum dive depth was not significantly different between mid- and late-winter, but both were significantly greater than mean maximum dive depth during summer (GLM with Tukey post-hoc tests; F_2,113=13.8, P<0.0001; Table 1).

There was some variation between individuals in the pattern of change of daily abdominal temperature (T_ab,D), but Fig. 3A shows the mean for all animals. In female penguins, T_ab,D decreased by approximately 1°C during early winter from the summer level then remained at this level for the rest of the winter. In male penguins, T_ab,D declined by approximately 0.8°C at the beginning of winter from the mean summer level but later increased to approximately the same as during summer. Changes in minimum abdominal temperature(T_ab,min) also varied between individuals but a general pattern was identified (Fig. 3B). In both male and female penguins, T_ab,minwas variable during winter but was up to 2°C higher than during summer. Significant changes in maximum abdominal temperature(T_ab,max) were far more consistent between individuals. T_ab,max declined steadily over the first 100 days of winter to a level around 1.5°C lower than that during summer. T_ab,max then remained at this level for the rest of the winter in female penguins but increased slightly, while remaining below summer levels, in male penguins (Fig. 3C). GLM was used to further investigate the changes in temperature (T_ab,D, T_ab,min or T_ab,max) for summer, mid-winter and late-winter as described above. In each analysis, temperature (T_ab,D, T_ab,min or T_ab,max) was the dependent variable, with sex and season as factors(Table 2). Mean T_ab,D was significantly greater in males than females but not significantly different between the seasons. Mean T_ab,min was not significantly different between the seasons or sexes. Mean T_ab,max was significantly greater in males than females and was significantly lower during mid-winter and late-winter than during the summer.

Minimum sV̇_O2showed an inverse pattern to that of diving behaviour(Fig. 4A). sV̇_O2min was lower during winter than summer. During winter,sV̇_O2min decreased initially until approximately day 50 and then remained relatively constant at around 50% of the summer value until the end of winter. Normalising sV̇_O2min for the effects of body temperature had little effect on this pattern(Fig. 4B). Mean sV̇_O2min during summer was significantly greater than sV̇_O2min during mid-winter (days 69–85) whether normalised or not (Woolf's test for differences; Table 3). The magnitude of the decrease in sV̇_O2min from summer to winter only changed from 51% to 49% with normalisation.

Changes in maximum dive duration and depth as winter progressed were mirrored by changes in the energetic cost of diving. For any given dive duration, mean sV̇_O2DC decreased as winter progressed, whereas during summer,sV̇_O2DC was unchanged (Fig. 5). At the start of winter,sV̇_O2DC was slightly less than that during the summer and decreased over the first part of the winter migration before reaching a stable level after approximately 50 days. This stable level was then maintained for most of the winter before increasing slightly at the end. As with mean dive duration, mean sV̇_O2DC was derived for each bird for the summer and for mid- and late-winter (phasedays 69–85 and phasedays 149–165, respectively) at three ranges of dive duration (41–50 s, 91–100 s,141–150 s). With the exception of 141–150 s dives in females,sV̇_O2DC was significantly different among these periods at each range of dive duration in both sexes (Woolf's test for differences; Table 4). Post-hoc Z-tests with Bonferroni corrections showed that, where there was a significant difference,sV̇_O2DC was the same during both winter periods but both of these were significantly lower than sV̇_O2DCduring summer.

The net cost of diving above MMR or maintenance metabolism(sV̇_O2net) was estimated by subtracting mean sV̇_O2min from mean sV̇_O2DC for each day (Fig. 6). sV̇_O2net was derived for each bird for the summer and for mid- and late-winter (phasedays 69–85 and phasedays 149–165, respectively) at three ranges of dive duration (41–50 s, 91–100 s, 141–150 s). With the exception of 141–150 s dives in females,sV̇_O2net was not significantly different among these periods at each range of dive duration in both sexes (Woolf's test for differences; Table 4). Post-hoc Z-tests with Bonferroni corrections showed that in dives of 141–150 s duration in females,sV̇_O2net was the same during both winter periods but both of these were significantly higher than sV̇_O2DCduring summer.

If sV̇_O2DC at a given dive duration decreases, then if oxygen stores remain unchanged, cADL should increase. cADL and the proportion of dives exceeding cADL were calculated for the three sampling periods used previously(Fig. 7). cADL is derived from several estimated quantities. It is therefore difficult to make statistical comparisons, other than to examine the extent of overlap in the ranges of 95%confidence intervals. Fig. 7indicates that in both males and female penguins, cADL was up to 45% greater during the mid- and late-winter than during summer. In female penguins, the proportion of dives within the cADL was the same during all three phases whereas in male penguins the proportion of dives within the cADL was the same during summer and mid-winter but was greater during late-winter.

In our earlier study of the diving physiology of macaroni penguins(Green et al., 2003), our measurements of fh were approximately 30% lower than those in the present study. After conversion, our estimates of sV̇_O2DC were approximately 45% lower in the previous study. We therefore derived an estimate of cADL during the summer of the present study that was 36% less than that in the previous study and which, because of the shape of the distribution of dive durations, resulted in 58% fewer dives being within the cADL. Some of the difference in sV̇_O2DC may be due to natural variability in metabolic rate and/or the demands of diving in different seasons, but re-examination of both data sets has unearthed other differences. We conclude that rates of oxygen consumption are higher in the present study because we have improved our methods of filtering out periods when the heart rate measurements are unreliable. The percentage of dives within the cADL in the present study now matches that found in other studies of diving birds (Butler, 2001)and suggests that V̇_O2 while submerged is substantially less than sV̇_O2DC. While cADL remains a useful comparative measure, both within and between species,this result once again highlights the importance of obtaining improved estimates of both usable oxygen stores and V̇_O2 while submerged if cADL is to truly reflect the duration of dives at which usable oxygen stores are exhausted (Green et al.,in press b).

In both sexes, diving behaviour during the pre-moult trip was relatively consistent, with little change in mean and maximum dive duration and depth. During the winter migration, however, the picture was rather different and two distinct stable phases can be identified. At the start of winter, both mean and maximum dive durations and depth were greater than those during the summer and proceeded to increase further until around day 50. These higher levels were maintained until around day 100. During this mid-winter period, mean and maximum depth and duration were significantly greater than those during the summer. There then followed a period of change where mean depth, mean duration and maximum duration decreased to other lower stable levels. During this late-winter stable period, maximum depth was unchanged but mean depth and mean and maximum duration were significantly lower. Indeed, mean depth was equal to that during summer.

The driving force behind these changes in behaviour is unclear but is likely to depend on the distribution of prey species with respect to depth. Antarctic krill (Euphausia superba) is the dominant species in the zooplankton assemblage of the Southern Ocean(Everson, 2000). Macaroni penguins feed predominantly on Antarctic krill, at least in the summer(Croxall et al., 1997), but in decreasing amounts in recent years (Barlow et al., 2002). Antarctic krill vary annually, seasonally,diurnally and geographically in their location in the water column(Godlewska, 1996) but usually undertake diurnal vertical migration; spending the night close to the surface and the day at deeper depths. Despite this, during the day, krill will nearly always be located within the diving depth range of macaroni penguins. The mean depth of the krill varies substantially between locations in the Atlantic sector of the Southern Ocean, Scotia Sea and waters around South Georgia(Everson, 1984; Godlewska, 1996). The mean depth in a review of studies varied from 28 to 156 m and the amplitude of migration from 2.5 to 59 m, depending on the timing and location of the studies. Indeed, some data suggest that close to South Georgia, pressure from predatory fish feeding on the shelf bottom causes diurnal vertical migration to be reversed, with krill found relatively close to the surface (around 50 m)during the day and dispersed throughout the water column at night(Everson, 1984).

There is little information on the foraging location of macaroni penguins,especially during winter. During the chick rearing season, the penguins undertake short foraging trips and remain close to South Georgia over the continental shelf (Barlow and Croxall,2002). During the longer incubation foraging trip, the penguins range further afield to far deeper waters in the Polar Frontal Zone(Barlow and Croxall, 2002). The incubation foraging trip is of a similar duration and for a similar purpose(rapidly to replenish or increase body reserves) to the pre-moult foraging trip, so it seems reasonable to assume for the time being that the penguins forage in a similar location. The location of macaroni penguins during the winter is currently unknown. However, it seems likely that during the winter they migrate to the open ocean away from the vicinity of South Georgia. A difference in the depth of krill swarms in the open ocean is therefore a likely explanation for the difference in mean diving depth between summer and winter. Figs 1 and 2 suggest that there was a further change in the mean depth of the krill from mid- to late-winter, due either to a difference in the location of the penguins or to the behaviour of the krill. In late-winter, mean depth, mean duration and maximum duration declined while maximum depth did not change. The decrease in maximum duration was only slight (6.5%; Table 1)and there were no changes at this time in the physiological parameters measured (Figs 3, 4, 5, 6), implying that the diving capacity of the penguins did not change. As a result, they were able to dive more comfortably within their limits. This is reflected in an increase in the proportion of dives within the cADL between mid-winter and late-winter(significantly so in male penguins), despite no significant change in the cADL(Fig. 7).

The increases in mean and maximum dive depth and duration from summer to winter, and the continued progressive increases through the first part of winter, were inversely matched precisely by changes in the energetic cost of diving. As mean and maximum dive duration increased, mean sV̇_O2DC for a given dive duration decreased (Fig. 5). It appears therefore that increased dive duration and depth were facilitated by a decrease in the energetic cost of diving. A progressive improvement in diving ability due to a reduction in energetic costs has not previously been observed in a mature diving animal. Juvenile diving animals commonly show an improvement in their ability to dive during development(Burns, 1999; Ponganis et al., 1999; Noren et al., 2001). Other studies have shown that oxygen stores can vary seasonally in adult animals(MacArthur, 1990) but that any apparent advantage of this is cancelled out by an increase in V̇_O2 while diving(MacArthur et al., 2000). Increased oxygen availability during diving was induced in tufted ducks(Aythya fuligula) by training them to dive for longer(Stephenson et al., 1989). However, a similar increase induced by training in muskrats (Ondatra zibethicus) was again cancelled out by an increase in V̇_O2 during diving(MacArthur et al., 2003). In the current study, we do not know how, or even whether, oxygen stores varied. However, the proportion of dives in excess of the cADL did not vary between summer and mid-winter (Fig. 7),suggesting that whether or not the estimate of oxygen stores we used is accurate, it did not vary. The penguins must have been under pressure to dive deeper and for longer during winter, and apparently this increase was accommodated solely by the decreased energetic cost of diving. This decrease must be achievable only in the time scale and under the conditions experienced in winter, or else the penguins would modify their energetic costs in the same way during the summer and we would see more dives within the cADL at this time.

Although seasonal variation in sV̇_O2 while diving has not been demonstrated previously, some authors have speculated on its existence (Bennett et al.,2001). In the current study, we show that, in macaroni penguins,variation in sV̇_O2while diving is due almost entirely to variation in sV̇_O2min. It is not possible to measure basal metabolic rate (BMR) in an active animal, but sV̇_O2min provides an approximation of the maintenance requirements of the penguins while at sea,excluding diving behaviour. sV̇_O2min showed a pattern of change which matched that of sV̇_O2DC(Fig. 4) and there are good linear relationships between mean daily sV̇_O2min and sV̇_O2DC(r²=0.73 and 0.83 for dives of 100 s duration in females and males, respectively). It seems clear that the decrease in sV̇_O2DC and increase in diving capacity are facilitated by this drop in minimum or basal metabolic rate.

Seasonal variation in BMR has been demonstrated in many species. Among diving animals, captive female grey seals(Boily, 1996; Boily and Lavigne, 1997),harbour seals (Phoca vitulina) and harp seals (Phoca groenlandica; Renouf and Gales,1994) show a marked underlying seasonal variability in resting metabolic rate, although the pattern of this variability varies between species.

The maintenance of a high core body temperature in endotherms carries a high metabolic cost (Bennett and Ruben,1979). In water, which has a thermal conductance 25 ×greater than air, these metabolic costs are likely to be even higher, even for an animal as well adapted to an aquatic lifestyle as a penguin. In several species of penguins, resting V̇_O2 is approximately twice as high when they are in relatively cold water than when they are in air (Stahel and Nicol,1982; Culik et al.,1991; Bevan et al.,1995). Other studies have shown that penguins have no thermoneutral zone in water, and metabolic rate increases with decreasing water temperature (Stahel and Nicol,1982; Barré and Roussel,1986). In little penguins (Eudyptula minor), this trend occurs until a critical temperature, beyond which metabolic rate increases sharply (Stahel and Nicol,1982).

Resting sV̇_O2during summer of macaroni penguins in water at 6.8°C, recorded using respirometry, was 27.0 and 24.5 ml O₂ min^-1kg^-1 for females and males, respectively(Green et al., 2005). These values are similar to sV̇_O2min during summer in the present study. However, by mid-winter,sV̇_O2min was closer to values obtained from the same penguins while resting in air (10.7 and 9.7 ml O₂ min^-1 kg^-1 for females and males, respectively; Green et al.,2005). In fasted king penguins (Aptenodytes patagonicus),resting V̇_O2 in water was approximately 2 × that in air. However, in well-fed penguins,resting V̇_O2 in water was substantially lower and was the same as resting V̇_O2 in air(Fahlman et al., in press). It is suggested by these authors that a complex interaction between nutritional state, vasoconstriction, fat deposition and fat mobilisation causes a decrease in thermal conductance (the inverse of insulation) of around 25% and therefore a substantial reduction in metabolic rate.

Perhaps macaroni penguins are also able to reduce the cost of thermoregulation during the winter. Thermoregulatory costs can be reduced by either decreasing the temperature gradient (by reducing body temperature or increasing external temperature) or by decreasing the rate of heat transfer or thermal conductance. Temporarily decreasing core temperature (anapyrexia)while inactive on a daily or seasonal basis is used by some endotherms to conserve energy when the temperature is very low and/or food is scarce(Nedergard and Cannon, 1990). Even in a period of high activity, barnacle geese (Branta leucopsis)were found to save considerable amounts of energy for their spring migration through having a reduced T_ab(Butler and Woakes, 2001). However, in the present study, mean T_ab,D did not change from the summer to winter (Table 2) and the modest decline in T_ab,max (a proxy for body temperature) of approximately 1.5°C(Table 2) was not nearly sufficient to explain the decrease in sV̇_O2min(Fig. 4). Perhaps more likely is that, by migrating northwards, macaroni penguins are able to forage in warmer waters during winter than they are during the summer. Factors controlling the sea surface temperature (SST) are beyond the scope of this article but, as an example, at the longitude of Bird Island (approx. 38°W) during July, SST increases by approximately 0.92°C for each degree of latitude travelled north from 53° S (the latitude of Bird Island) to 38° S and beyond.

Thermal conductance has been studied several times in penguins (summarised in Luna-Jorquera et al.,1997), although frequently in air rather than water. Thermal conductance was lower in air than water in little penguins(Stahel and Nicol, 1982). In water, erection of feathers is not possible as they must form a waterproof layer. This will trap less air, reduce the effective plumage depth, and hence the contribution to insulation will decrease. When the penguins dive, even more air will be removed from the plumage and the remaining air will be compressed, both of which will further increase conductance(De Vries and Van Eerden,1995). It is estimated that penguin plumage comprises 73–87%of total insulation in air (Le Maho et al., 1976; Stahel and Nicol,1982; Barré,1984), but this declines to 8 and 10% while fully immersed in water for macaroni and king penguins, respectively(Barré and Roussel,1986), with internal insulation accounting for the remainder. Internal insulation will therefore play a major role in thermoregulation while the macaroni penguins are engaged in long foraging trips at sea. During the winter, it may be supposed that the body condition of the penguins is better than that during the pre-moult period, as the penguins are less constrained in diet, foraging area and behaviour. This may enable them to have a more efficient and evenly distributed insulative fat layer than during the summer. In other animal models, the abdominal fat is the first resource to become exhausted during fasting and the first restored during re-feeding, while subcutaneous tissues are the last to be restored(Blem, 1990). Fasting,re-feeding and the large variation observed in body mass during the breeding season in macaroni penguins (Croxall,1984) may be the result of changes in abdominal fat stores, while during the winter the more insulative subcutaneous tissues are restored. Furthermore, although the insulative contribution of the feather layer decreases, its insulative effect is bound to be greater during the winter when the feathers are new, rather than immediately before the moult when they will be worn and undoubtedly less effective.

Interactions between T_ab, water temperature and insulation can be summarised in a simple model of thermal conductance changes. Conductance (C in W deg.^-1 m^-2) can be calculated simply as C=Ṁ_O2(T_b–T_a)^-1A^-1,where T_b and T_a are core body and ambient body temperature, respectively (T_ab and SST in this case), and A is the surface area of the bird in m². Ṁ_O2 (metabolic rate) is calculated from sV̇_O2min using an equivalence of 18.889 J ml^-1 O₂(Green et al., 2002). Surface area is given by Meeh's formula, with a constant of 0.77(Barré and Roussel,1986). While there is much debate on the correct constant for use in penguins (Luna-Jorquera et al.,1997), the value has no effect in our model, which looks at relative effects of insulation and T_a in isomorphic individuals. Exact mass is unknown but it is assumed to be the same for summer and winter at 4 kg. Varying mass has little effect on the conclusions of the model. We do not know how temperature changes with depth where the penguins are located, so T_a is assumed to be equal to SST. During the pre-moult trip, if we assume that the penguins forage close to the Maurice Ewing Bank (longitude –41.5°, latitude –51.5°) then SST during March 2002 was 5.8°C. Conductance can therefore be calculated for the pre-moult trip. If we assume that C is equal in the winter or decreases by 10, 20, 30 or 40% (i.e. insulation increases) then we can calculate what percentage of the decrease that we observe in sV̇_O2min might be explained by increasing water temperatures, i.e. reducing ΔT,the temperature difference between T_a and T_b (Fig. 8).

Modelling thermal conductance in this way assumes that animals are passive bodies, and studies have shown that such an assumption can oversimplify the complexity of changes in conductance in animals(Hind and Gurney, 1997). Furthermore, heat generated during locomotion may be used in thermoregulation,while movement itself will alter conductance(De Vries and Van Eerden,1995; Hind and Gurney,1997). However, penguins do not dive to forage at night(Wilson et al., 1993; Green et al., in press a) and therefore at this time probably remain inactive at the water surface. Indeed,in the present study,sV̇_O2min was nearly always recorded during the hours of darkness. Therefore, while changes in conductance are unlikely to provide the full explanation for the decrease in MMR during winter, Fig. 8suggests that this will be an important component. This simple model shows us that small increases in insulation, particularly internal insulation, and/or water temperature can have a large effect on the reduction of metabolic rate. The progressive decrease in sV̇_O2min at the start of winter and the increase towards the end of winter could be explained by the penguins moving from cooler to warmer waters and vice versa or a progressive improvement in body condition and insulation. In great cormorants (Phalacrocorax carbo carbo), water temperature, body temperature and body fat thickness were found to be major contributors to diving energetics (Grémillet et al., 1998). Further work on the winter location, body condition and mass of penguins will assist us in assessing the relative contribution of the effect of a change in thermal conductance and/or ΔT in macaroni penguins.

List of symbols

 
• fh

  heart rate

 
• C

  conductance

 
• cADL

  calculated aerobic dive limit

 
• M_b

  body mass

 
• MMR

  minimum metabolic rate

 
• Ṁ _O2

  metabolic rate

 
• sV̇_O2

  mass-specific rate of oxygen consumption

 
• sV̇_O2DC

  mass-specific rate of oxygen consumption during dive cycles

 
• sV̇_O2min

  minimum mass-specific rate of oxygen consumption

 
• sV̇_O2minC

  minimum mass-specific rate of oxygen consumption normalised to 39°C

 
• sV̇_O2net

  net cost of diving

 
• T_a

  ambient temperature

 
• T_ab

  abdominal temperature

 
• T_ab,D

  mean daily abdominal temperature

 
• T_ab,max

  daily maximum abdominal temperature

 
• T_ab,min

  daily minimum abdominal temperature

 
• T_b

  body temperature

 
• V̇ _O2

  rate of oxygen consumption

The authors would like to thank Nick Warren, who assisted in the retrieval of the data loggers, and the rest of the science team at Bird Island,especially Jane Tanton. The authors would also like to thank Dr Craig White for helpful discussions and comments on the manuscript. We are grateful to the British Atmospheric Data Centre, which provided us with access to the Met Office Sea Surface Temperature Data. This work was funded by NERC, under their Antarctic Funding Initiative (AFI), with logistical support provided by the British Antarctic Survey.