Diving behaviour of southern elephant seals: new models of behavioural and ecophysiological adjustments of oxygen store management Free

In living organisms, maintaining energy balance is key to survival. It depends on the ability of individuals to acquire resources at rates sufficient for sustaining their overall metabolism. Successful reproduction requires energy gains that outweigh the expenditure corresponding to maintenance requirements and growth. Individuals that maximise their energy intake per unit of time and mass have more energy to allocate to reproduction and thus should increase their fitness (Boggs, 1992). In this way, natural selection favours animals that balance their overall energy expenditure with the energy they extract from their environment (Kleiber, 1975). Among mammals, pinnipeds have a bimodal cycle with periods at sea and time on land or ice (Harrison and Kooyman, 1968). On land, during breeding, pinnipeds fast, and their body condition deteriorates as all their metabolism is exclusively fuelled by their lipid, protein and mineral body stores (Hindell et al., 1994). These reserves need to be restored during their at-sea foraging trips.

As obligate air breathers at sea, pinnipeds forage at depth and return to the surface to breathe between dives. As such, they have been classified as ‘ocean-surface central-place’ foragers. To forage effectively, these predators must modify their diving behaviour to minimize the energy costs involved in searching and capturing prey while maximizing energy acquisition. Optimal diving theory (ODT), predicts that animals should maximise their time foraging underwater relative to their oxygen capacity. Individuals should end their dives when their oxygen reserves are depleted (Carbone and Houston, 1994; Houston and Carbone, 1992; Kramer, 1988; Thompson et al., 1993). However, other factors such as abundance, depth, energy content of the prey or individual body density impact the decision to continue or stop the dive (Adachi et al., 2014; Jouma'a et al., 2016; Thompson and Fedak, 2001). Pinnipeds have developed morphological and physiological adaptations to simultaneously decrease the cost of exercising underwater (i.e. diving metabolic rate) and increase the magnitude of body oxygen stores. Individuals may also continue their dives once their oxygen reserves have been exhausted. In this case, they switch from aerobic metabolism to anaerobic metabolism at the cost of a disproportionate increase in surface recovery time (Kooyman et al., 1983). In phocids, oxygen is almost exclusively stored in blood and muscles (Hassrick et al., 2010) and the blood volume relative to the seals' mass is twice that of humans (Irving et al., 1941; Ponganis, 2015; Ponganis et al., 2011). Not only is their blood volume large but the haematocrit value, and consequently the haemoglobin content of phocid blood, is also very high (Blix, 2018; Lenfant et al., 1970). The oxygen-carrying capacity of blood and hence the blood oxygen store may therefore be as high as 64 ml O₂ kg⁻¹ in northern elephant seals (Ponganis, 2015; 11 ml O₂ kg⁻¹ in humans). In addition, seal muscles are conspicuously dark because of their very high myoglobin concentration (Scholander, 1940). In elephant seals, muscles can bind and store 26 ml O₂ kg⁻¹ (Ponganis, 2015; 3 times more than human muscle), whereas only 4 ml O₂ kg⁻¹ are stored in the respiratory system. The total oxygen stores of an elephant seal correspond to a staggering 94 ml O₂ kg⁻¹ with 68% in the blood, 28% in the muscle and 4% in the respiratory system (Davis, 2014). In comparison, humans have total oxygen stores of approximately 25 ml O₂ kg⁻¹, of which ∼30% is contained in the lungs (Blix, 2018; Jouma'a et al., 2016; Richard et al., 2014). Further optimising oxygen stores in pinnipeds, the cardiovascular dive response is a decrease in heart rate (i.e. bradycardia) and a global vasoconstriction while maintaining arterial pressure, which results in reduced blood flow in non-essential tissues in order to keep O₂ for vital organs like the brain (Costa, 2007; Irving et al., 1941). Elephant seals display large variability in the degree of diving bradycardia, with extreme values as low as 2 beats min⁻¹ (Hindell and Lea, 1998).

To minimise the cost of transport, reduce oxygen depletion and therefore maximise foraging time, seals rely on a streamlined body and adjust their swimming effort (i.e. hind flipper stroke frequency) relative to their buoyancy (Sato et al., 2003). Body composition (i.e. the proportion of lipids, protein, water and ash) influences the animal's buoyancy, which rises with increasing lipid content, with direct consequences on its cost of transport and therefore its swimming effort (Adachi et al., 2014; Jouma'a et al., 2016; Richard et al., 2014; Sato et al., 2003; Webb et al., 1998). With cost of transport being minimal at neutral buoyancy when the animal's body density equals the density of the fluid in which it is immersed, any deviation from neutral buoyancy will result in increased swimming effort and, therefore, energy expenditure (Adachi et al., 2014; Miller et al., 2012). Richard et al. (2014) found that in negatively buoyant southern elephant seals (SESs), a 1% decrease in buoyancy resulted in a 20% increase in swimming effort during the ascent phase of the dive. Therefore, for a given amount of oxygen stored, buoyancy is a critical parameter in determining the swimming effort and thus oxygen consumption during a dive, with direct consequences on the overall dive duration (Beck et al., 2000; Crocker et al., 1997; Sato et al., 2003). These adaptations associated with long dive durations, suggest that elephant seals are not very active swimmers and that they adopt an ‘inexpensive’ foraging dive behaviour compared with other marine mammals (Costa, 1991b; Costa and Maresh, 2017; Naito et al., 2013).

Physiological data are essential for understanding diving mammal metabolism (Butler and Jones, 1997) but they remain difficult to collect in free-ranging animals. Kooyman and colleagues (1967, 1971, 1973), who recorded physiological data and calculated metabolic rate for the first time on free-ranging Weddell seals (Leptonychotes weddellii), did so by using a respiratory chamber on a man-made ice hole. However, transferring this method to ocean-ranging species is not possible. Recently, the development of bio-logging has allowed researchers to access some physiological data such as heart and breathing rates (Day et al., 2017; Fletcher et al., 1996; Génin et al., 2015; le Boeuf et al., 2000). High-frequency accelerometers incorporated into time–depth loggers provide reliable behavioural information on variations in energy expenditure (Jeanniard-du-Dot et al., 2017; Maresh et al., 2014; 2015) relative to diving depth and duration as well as prey-encounter rates (Gallon et al., 2013; Naito et al., 2010; Suzuki et al., 2009; Viviant et al., 2010). Some physiological data such as the behavioural aerobic dive limit (ADL_B) are still not directly measurable but they can be estimated from diving parameters: the threshold for a dive duration beyond which surface intervals begin to disproportionately increase in relation to the increase in dive duration (Burns and Castellini, 1996; Hindle et al., 2011; Kooyman et al., 1983).

Elephant seals are the most extreme divers among pinnipeds as they perform the longest [24.5±6.7 min (mean±s.d.); up to 2 h] and deepest dives (433±166 m; up to 2000 m) (Slip et al., 1994). Diving parameters suggest that despite such extreme diving behaviour, the vast majority of dives performed by free-ranging elephant seals remains within the ADL_B (Le Boeuf et al., 1988, 1996; Meir et al., 2009). Despite these large variations in dive duration, we expect elephant seals to adjust some physiological parameters such as the overall amount of oxygen stored, depending on dive duration and oxygen expenditure, to remain within the ADL_B (Day et al., 2017; Génin et al., 2015). However, it is unclear how seals manage and balance oxygen stores and expenditure while swimming (Maresh et al., 2015).

The main objective of our present study is to investigate how SESs modify their oxygen consumption through behavioural and ecophysiological adjustment, according to dive duration, to remain within their aerobic dive limits even for the extreme dive durations observed. We hypothesise that with increasing dive duration, SESs reduce their overall swimming effort to a greater extent than previously estimated and/or undertake some ecophysiological adjustments, such as a decrease in body temperature during extended dives, to further decrease their metabolic rate to remain within the ADL_B. Furthermore, SES females continue to grow with age (Bell et al., 2005). As oxygen stores are related to body mass/size, we expect phenotypic/age differences in size to strongly influence the diving behaviour and aptitudes of SES females.

Data were collected on 66 post breeding female southern elephant seals [Mirounga leonina (Linnaeus 1758)] from the Kerguelen archipelago (49°20′S and 70°20′E) between 2010 and 2018. Females were sedated with an intravenous injection of Tiletamine–Zolazepam (Zoletil 100; dosage 0.6–0.7 ml per 100 kg) (McMahon et al., 2000) and equipped with different data loggers (Table 1) glued on their fur using fast-setting epoxy glue (Araldite AW 2101, Ciba). Female body length (standard length), referred to as ‘size’ in this paper, was measured to the nearest centimetre and female body mass was measured to the nearest kilogram. Data loggers were glued to the animal's back or head depending on the other tags it carried (Table 1). Loggers recorded pressure, surrounding temperature, and the animal's location (via the GPS and ARGOS system) over time. These were coupled to a 3-axis accelerometer recording at a frequency of 16 Hz (MK10-X, Wildlife Computers, Redmond, USA; 56 mm×38 mm×42 mm; 165 g for individuals in 2010–2016) or 12 Hz (CTD, SMRU Instrumentation, University of St Andrews, UK; 10.5 mm×7 mm×4 mm; 545 g for individuals in 2016–2019) with a resolution of 0.05 m s⁻². Once equipped, females were released. When they returned onshore to moult (i.e. during the austral summer), females were recaptured, anaesthetised using the same protocol and data loggers were recovered.

In 2017, an additional subcutaneous data logger (DST milli-HRT heart rate and temperature logger for animals; Star Oddi; 13 mm×39.5 mm; 11.8 g) was deployed on one female to measure under-blubber temperature. The DST milli-HRT logger sampled temperature (±0.2°C) every 5 min. To attach the logger, the female was captured according to the above procedure. The logger was then implanted, using the surgical method of Chaise et al. (2017), under the blubber layer and against the thoracic muscle at the apex beat projection. The device was removed with the other loggers when the female came back to moult. Capture and surgical procedures were the same as the methods used by McMahon et al. (2000) and Chaise et al. (2017).

To construct a relationship between mass and length, we used data from all female SESs from Kerguelen archipelago that were weighed and measured since 2008 (n=253) using the same anaesthesia protocol previously described.

The ‘Comité d'Ethique ComEth Anses/ENVA/UPEC’ validated all scientific procedures performed on elephant seals under projects 16-078-2016100709156983; 19-040 #21375: Physiological and energy expenditure adaptations of elephant seals to environmental constraints during their life cycle as part of the IPEV programme 1201-CycleEleph. All animals in this study were handled and cared for in total accordance with its guidelines and recommendations.

Depending on the time of the cycle (fasting, moulting, period at sea), the mass of the animals varies owing to the loss or gain of primarily fat tissue and, to a lesser extent, protein tissues. The first step was constructing an allometric relationship between mass and length to compensate for these variations.

Diving data were processed using MATLAB (The MathWorks, Natick, MA, USA) and a custom-written script described in Vacquié-Garcia et al. (2015). A dive cycle was defined as a dive with three periods (descent, bottom and ascent) followed by a recovery period at the surface. We consider that animals were diving when the depth was greater than 15 m. At less than 15 m depth, SESs were considered as being at the surface. This value was chosen to eliminate seal subsurface movements. The three periods of each dive were extracted using a vertical speed criterion (Vacquié-Garcia et al., 2015). The descent and ascent phases were identified as the periods before or after the surface where the vertical speed was greater than 0.75 m s⁻¹. Bottom phases were defined between the descent and ascent phases, where the vertical speed remained below 0.75 m s⁻¹. For each dive, the maximum depth (m), dive duration (s), descent, ascent and bottom duration (s), and water temperature (°C) were computed.

Acceleration data were processed according to Viviant et al. (2010) and Gallon et al. (2013), using accelerometers to measure static and dynamic acceleration. Static acceleration relates to the position of the animal's centre of gravity relative to the gravity vector and corresponds to low-frequency signals. Dynamic acceleration describes the acceleration induced by the animal's body and head movements, and corresponds to high-frequency signals (Richard et al., 2014; Vacquié-Garcia et al., 2015; Viviant et al., 2010). The three axes were filtered using a high-pass filter at 0.02 Hz to remove noise from the signal due to static acceleration.

Prey-capture attempts were estimated from the accelerometer data (Gallon et al., 2013; Guinet et al., 2014; Viviant et al., 2010). The dynamic acceleration was extracted and prey-capture attempts were calculated using a MATLAB custom-made script according to the method described in Jouma'a et al. (2016).

Two models were built to study SES oxygen consumption during dives: one as a function of dive duration and the second as a function of body density. The first model estimates oxygen consumption for each dive duration and compares this consumption to the behavioural aerobic dive limit. The second model estimates oxygen consumption according to the body density, still at the dive scale. As body density of SESs decreases during foraging trips (Richard et al., 2014), body density is used as a proxy to represent the time spent at sea.

In SESs and most marine mammals, oxygen reserves are located in the blood and muscles. Our model did not consider air in the lungs as this oxygen store is minimal in elephant seals, given that they exhale before diving. The lungs collapse from a certain depth (between 30 and 50 m), with residual air being chased toward the upper trachea where it cannot be used (Falke et al., 1985; Reed et al., 1994). When the lungs are collapsed, there is very little exchange of O₂, if any, between the lungs and blood (Davis and Weihs, 2007). Elephant seals have a blood volume of about 250 ml kg⁻¹, of which 33% is arterial and 67% is venous circulation (Kooyman, 1989). The haemoglobin concentration is 250 g l⁻¹, and its oxygen-binding capacity is 1.34 ml O₂ g⁻¹ haemoglobin (Davis, 2014; Meir et al., 2009). At the beginning of a dive, arterial blood is considered to be oxygen saturated and venous blood is considered to be 85% oxygen saturated (Davis and Weihs, 2007; Kooyman et al., 1980; Meir et al., 2013; Ponganis et al., 1993). Skeletal muscle accounts for ∼28% of the animal's total mass (Davis, 2014). The myoglobin concentration in muscle is 65 g kg⁻¹ and its oxygen-binding capacity is 1.34 ml O₂ g⁻¹ myoglobin (Kooyman, 1989). At the beginning of the dive, all the animal's myoglobin is considered to be saturated with oxygen (Davis and Weihs, 2007; Hassrick et al., 2010). This corresponds to the maximum oxygen capacity that the animal can have in reserve. During diving, Meir et al. (2009) showed that only 91% of arterial oxygen but 100% of venous and skeletal muscle oxygen can be used by the animal. In addition, Tift et al. (2014) demonstrated that elephant seals have high carboxyhaemoglobin concentrations. The maximum carboxyhaemoglobin value they measured resulted in a 7% reduction in the total oxygen reserves of the individuals. To take account of the presence of this high carboxyhaemoglobin concentration, the calculated oxygen reserves were therefore reduced by 7%.

We retained locomotion cost, basal metabolism and digestion cost [the heat increment of feeding (HIF)] as factors influencing energy expenditure during dives. (1) Locomotion cost was estimated by the energy expenditure of a flipper stroke during the short trip between breeding and moulting using the 0.24±0.04 J kg⁻¹ stroke⁻¹ value found for northern elephant seals (Maresh et al., 2015). Maresh et al. (2015) calculated a mean energy expenditure over the travel period, which we corrected by plotting the histogram of mean stroke amplitude, thereby hypothesising that the most representative intensity corresponded to an expenditure of 0.24 J. Then, a simple proportionality rule was applied to correct this energy expenditure per flipper stroke according to mean stroke amplitude. During a dive, locomotion cost depends on flipper strokes and, therefore, stroke rate and amplitude. As these parameters vary according to the dive duration, we estimated a mean locomotion cost for each dive duration. Energy expenditure costs were converted to oxygen consumption using conversion factors of 0.239 cal J⁻¹ and 0.208 ml O₂ cal⁻¹ (Schmidt-Nielsen, 1979). (2) Basal metabolic rate (BMR) was estimated using the Kleiber predictions. Although BMR has never been measured in adult elephant seals, previous experiments on phocids have demonstrated that Kleiber's prediction of BMR approach empirically determined BMR for adult phocids (Castellini et al., 1992; Hurley and Costa, 2001; Sparling et al., 2007b; Sparling and Fedak, 2004; Webb et al., 1998; Williams et al., 2004). To take into account a cooling of the animal's body from the outside during the dives, a correction factor related to the sub-blubber temperature was applied to the BMR. On the basis of various studies on the decrease in body temperature in Weddell seals during dives, we established that a decrease of 3°C of subcutaneous temperature leads to a 15% decrease in BMR O₂ consumption (Blix, 2018; Castellini et al., 1992; Hill et al., 1987). The change in the sub-blubber temperature on the SES during the dives was determined thanks to the individual equipped with the DST milli-HRT logger (see Table 1). This hypothesis will be discussed in the Discussion. (3) To account for digestion cost (HIF), we used a mean value calculated in the study of Maresh et al. (2015). This cost is estimated at 11.6% of metabolizable energy in juvenile elephant seals (Barbour, 1993). Taking the mean values of travel time and energy ingested during the short trip (Maresh et al., 2015), we estimated the digestion cost at 483 MJ. This value was converted to a cost per unit of time during the trip (in s).

For this second model, the same factors previously described were used (locomotion cost, BMR and HIF). The changes in dive duration, surface duration, mean stroke amplitude and the number of flipper strokes as a function of seal density were modelled from our recovered SES data. Then, assuming that the cost of transport is minimal at neutral buoyancy (i.e. 1027 kg m⁻³; Miller et al., 2012), and thus, that parameters are optimal for this density, we modelled the changes to all the parameters in percentages. Finally, we calculated the change in oxygen consumption as above (as the sum of locomotion cost, BMR and HIF).

Statistical analyses were made using R-Studio, v. 3.6.1 (https://www.r-project.org/). Since our study concerns SES behaviour during foraging dives only, a filter was applied to all the data to exclude transit dives (dives between the archipelago and the hunting areas). Dives recorded close to the Kerguelen archipelago (between 47°17'S–50°40'S and 71°34'E–67°84'E), mainly corresponding to the offshore or inshore transit of females, were removed. For each SES, dives that lasted longer than 4000 s and those with a calculated stroke rate greater than 1.5 were considered as outliers and removed from the analyses owing to their infrequency (Aoki et al., 2011).

To study whether different diving strategies exist, clustering methods (k-means and dendrograms) and principal component analyses were performed using mean dive duration, mean stroke amplitude and dive depth as parameters for each individual. The most representative number of clusters was chosen using the Calinski–Harabasz index.

To create our models and correct the factors influencing oxygen consumption (see previously), the energetic parameters (mean stroke amplitude, stroke rate, number of flipper strokes, surface duration) were studied separately as a function of dive duration and then seal density, using the mixed linear model function ‘lmer’ (R package lme4: https://CRAN.R-project.org/package=lme4). For each parameter, the most appropriate model (between a model with random intercept and slope, random intercept, or no random effect) was selected according to the Akaike Information Criteria (AIC). Individuals were used as a random factor consider that each animal performed a different number of dives, and assumptions of the mixed random models were tested. The R² was obtained using the R function ‘rsquaredGLMM’ (R package MuMIn: https://CRAN.R-project.org/package=MuMIn).

To check how the sub-blubber temperature changed during diving, a standard linear model was performed as a function of the dive duration for the equipped individual. Assumptions of the linear models were tested. This relationship was calculated for a single individual and a single body location.

Only surface intervals between 0 and 600 s were used as we focused on the surface duration directly related to the previous dive. Longer surface duration can be related to other behaviours, such as socializing or digestion (Sparling et al., 2007a). We calculated the mean surface recovery time over a 50 s time-series window for each dive duration. The breakpoint was then found using the ‘segmented’ R function (https://CRAN.R-project.org/package=segmented). Then, this ADL_b was compared with the theoretical ADL (ADL_t) calculated with our first model.

For the accelerometery analysis, 63 of the initial 66 individuals were included. One individual was discarded for lack of data as it did not travel far enough from the archipelago, and two others were considered as outliers for marginal behaviour (longer dives). In total, 157,091 dives were analysed. Dives without corresponding accelerometery data (2901 dives), dives longer than 4000 s (11 dives, not representative and marginal behaviour) and dives with a calculated stroke rate above 1.5 (213 dives considered as outliers) were discarded.

To construct this relationship, data from 253 adult females were used. Female mass ranged between 237 and 547 kg, while standard length ranged from 210 to 284 cm. As size increased, mass increased significantly according to a power function relationship (P<0.01; Fig. 1). Consequently, animal mass was estimated based on this relationship in all of our analyses.

A dendrogram and a k-mean clustering of individuals were performed based on their diving behaviours (mean dive duration, depth and mean stroke amplitude). Both show that individuals can be divided into two groups: the deep- and shallow-diving groups (referred to as ‘deep group’ and ‘shallow group’ hereafter). A striking contrast exists between individuals (n=31) performing long (20.13±0.67 min) and deep dives (528±23 m) with low mean stroke amplitude (0.87±0.04 m s⁻² stroke⁻¹) and individuals (n=32) performing short (16.41±0.53 min) and shallow dives (380±25 m) with a significant higher mean stroke amplitude (1.01±0.06 m s⁻² stroke⁻¹) (paired Wilcoxon tests, P<0.01). Individuals diving deeper and for longer also had a notably larger body size than those performing shallower and shorter dives (Table 2).

The change in stroke rate with dive duration was identical in both groups, as was the mean stroke rate. Neither of the diving modes seemed to confer any advantage in body condition recovery, as the change in ascent mean stroke amplitude (i.e. ascent swimming effort) with time was not statistically different between the two groups (Table 2). However, a similar rate of change in density will result in a larger absolute mass gain in the longer female group compared with the shorter ones, as we discussed. In terms of energetic reserves, since the deep group's individuals were larger (longer and heavier) than the others, they theoretically had comparatively more oxygen reserves. Thus, their ADL_t was higher (Table 2).

Then, we investigated whether these differences were due to age, as elephant seals exhibit continuous growth. Dive duration increased significantly with the body length of the individuals (F_1,61=3.851, P<0.01; Fig. 2A), and mean stroke amplitude tended to decrease as size increased (F_1,61=3.205, P=0.08; Fig. 2B). However, one large individual fell into the shallow group as it performed short and high-intensity dives (Fig. 2A,B).

We then assessed the variation in mean stroke amplitude as an index of energy expenditure through the dive duration to integrate these changes in the models to correct locomotion cost and BMR. To limit convergence problems in the linear models, we removed individuals with less than 100 dives recorded (2 individuals in the deep group and 6 in the shallow group).

Mean stroke amplitude varied negatively with dive duration (t=−6.33, P<0.01; Fig. 3A) and decreased at the same rate for both groups (t=−1.11, P>0.05). However, the shallow group recorded a higher mean stroke amplitude than the deep group (t=2.31, P<0.05; Fig. 3A). The relationship between mean stroke amplitude and dive duration is expressed for both groups in Table 3.

No differences in stroke rate were found between the shallow and deep groups (t=−1.69, P>0.05). Regardless of the group, the stroke rate varied negatively with increasing dive duration (t=−6.62, P<0.01; Table 3 and Fig. 3B).

To limit convergence problems in the linear models, we removed individuals for which less than 5 body densities could be estimated from their accelerometer data (7 in the deep group and 11 in the shallow group). Therefore, this analysis was performed on 37 SES females.

Seal density and mean stroke amplitude were positively correlated (deep group: t=2.94, P<0.01; shallow group: t=86.80, P<0.001), but the relationship differed between groups. In the deep group, a polynomial relationship of 2 degrees provided the best fit (Table 3). In contrast, a linear relationship was applied to the shallow group (Table 3).

The number of flipper strokes was different between the shallow and deep groups (t=20.54, P<0.01): stroke number varied negatively with seal density for the deep group (t=−26.19, P<0.01), whereas a positive relationship was found for shallow group (t=2.19, P<0.05), as expressed in Table 3.

Dive duration varied according to seal density and the seal diving group (t=−19.45, P<0.01). For the deep group, dive duration was negatively correlated with seal density (t=−28.58, P<0.01; Table 3), whereas for the shallow group, dive duration did not vary according to seal density (t=−1.74, P>0.05). The time spent at the bottom of the dive decreased with seal density (t=−42.7, P<0.01), regardless of the group (t=−1.54, P>0.01).

Finally, using the linear mixed model, the variation in surface duration according to seal density, dive duration and mean stroke amplitude was found to differ between the diving groups (t=7.98, P<0.01). Surface duration decreased with increasing seal density for the deep group (t=−8.17, P<0.01; Table 3). In contrast, it remained constant for the shallow group (t=−1.84, P>0.05).

For both groups, a box plot was constructed to represent surface duration for each dive duration (Fig. 4). For the deep group, surface duration initially increased slowly before increasing drastically beyond 2065±63 s of diving (Fig. 4A). This 2065±63 s breakpoint represents the ADL_b. Only 1% of the dives exceeded the ADL_b for the deep group. For the shallow group, surface duration increased with dive duration, but with no breakpoint characterising the ADL_b was identifiable (Fig. 4B).

Since stroke rate, mean stroke amplitude and sub-blubber temperature varied with dive duration, these probably influenced oxygen consumption during the dive. An oxygen-consumption model was developed for both the shallow and deep groups to assess the influence of these factors.

As animal density varies over time with an effect on diving parameters, we modelled the change in oxygen consumption according to SES density. Since the energetic parameters differed between the two groups, a model was built for each group.

For the shallow group, only mean stroke amplitude varied while dive and surface duration remained constant. Mean stroke amplitude decreased by 70±6% with seal density decreasing from 1060 to 1027 kg m⁻³.

In the oxygen-depletion model according to body density, among the deep group individuals, dive duration increased as their body condition increased (Fig. 6). Assuming that these individuals dived close to their ADL and that only muscle oxygen reserves increased over time, we estimated their gain in muscle mass. We calculated that to remain at their ADL_b, their muscle mass should have increased by about 17±3% during the trip.

For the deep group, a box plot was constructed to represent prey-capture rate as a function of duration according to depth. For these individuals, the number of prey-capture attempts exhibited a trend to decrease with depth (Fig. 7).

The oxygen management of SESs during their post-breeding foraging trip was investigated to understand how SESs adjust their diving energy expenditure according to dive duration and body density to maximise their foraging time. The use of time–depth recorders and accelerometers to monitor diving behaviour and estimate oxygen consumption in female SESs during their post-breeding foraging trip provides insight into the adjustment of energy expenditure during foraging dives. Changes in seal density that affect the seal's mean stroke amplitude directly affect oxygen consumption throughout the dive. Seal density, related to the body composition (i.e. the relative proportion of proteins, lipids, ash and water), influences SES diving behaviour, with a decrease in body density resulting in a decrease in swimming effort and an increase in dive duration (Jouma'a et al., 2016; Miller et al., 2012; Richard et al., 2014). During the trip, SES density decreases as seals improve their body condition and mass through an increase in the relative proportion of lipids over proteins (Richard et al., 2014). Indeed, SES females were estimated to gain 0.9±0.4 kg (range: 0.2–2 kg) per day spent at sea (Guinet et al., 2014).

SES females from Kerguelen Island could be divided into two groups according to their diving behaviour. Some individuals swam actively at shallower depths and for shorter periods. In contrast, others swam less actively but performed deeper and longer dives. Costa (1991a) predicted this relationship. These differences could be age related as females in the deep-diving group were, on average, longer, larger and possibly older than females in the shallow-diving group. In addition, the tendency of changes in dive duration and mean stroke amplitude with size to be gradual increases the likelihood that such changes are related to age-related size but also possible phenotypic differences (i.e. larger individuals for a given age are more likely to dive longer). Females grow as they age (Bell et al., 2005), and their muscle mass and blood volume increase. Muscles and blood are the main oxygen-storage compartments for SESs (Hassrick et al., 2010) and therefore, their oxygen reserves increase. It is likely that while absolute metabolic rate increases with female mass, the mass-specific metabolic rate could also tend to decrease with mass as observed at inter-specific levels in mammals (Brody and Lardy, 1946; Kleiber, 1932; Rea and Costa, 1992) but this remains to be verified in SESs. An increase in oxygen reserves associated with lower mean stroke amplitude could explain the gradual transition from shallow and short dives to longer and deeper dives. However, one larger individual performed active shallow diving, so other factors, such as inter-individual ontogenic and/or phenotypic behavioural differences, might be in action. An alternative hypothesis is that senescence could contribute to this change, with older and therefore longer, individuals returning to shorter and shallower dives (Hassrick et al., 2010; Hindle et al., 2011). To summarise, the differences in individual diving strategies appear to be mainly driven by physiological constraints such as oxygen storage. Still, to adequately address this question, working with individuals of known age and/or longitudinal monitoring of SES females over years of diving behaviour is necessary.

In parallel, no differences in the speed at which individuals improved their body condition were found between the shallow and deep groups as the change in ascent mean stroke amplitude with time was similar between the two groups (Table 2). Therefore, no particular diving strategy tends to confer an advantage on the rate of improvement of the females' body condition (i.e. buoyancy). However, a similar rate of change in buoyancy will result in different absolute mass gains, with larger females belonging mainly to the deep group gaining, over the whole foraging trip, a greater absolute amount of body mass compared with smaller females mainly in the shallow group. However, large inter-individual differences in buoyancy changes were observed within each diving group, revealing differences in foraging performances between individuals according to the foraging location, year and/or foraging efficiency.

The number of prey-capture attempts during the time spent at the bottom of the ocean tend to decrease with increasing diving depth for the deep group. This result is consistent with other studies performed in our group relying on prey sizes estimated from an acoustic micro-sonar deployed on SES females (Goulet et al., 2019), which showed that fewer but larger prey tend to be captured at greater depth (our unpublished data) and therefore that it might be beneficial to dive deeper.

Both the mean stroke amplitude and stroke rate exhibited significant and concomitant changes with dive duration. This finding is consistent with the optimal diving theory (Kramer, 1988), which defines the best strategies for animals to maximise their time foraging underwater. By reducing their locomotion costs, individuals can dive longer. Our results clearly emphasize this, with individuals diving deeper and longer having a lower mean stroke amplitude than shorter and shallower diving ones.

According to several studies (Adachi et al., 2014; Jouma'a et al., 2016; Richard et al., 2014), seal density is an important factor to be considered in energy-expenditure calculations. SES density impacts swimming effort during dives (Maresh et al., 2014; Miller et al., 2012; Richard et al., 2014), with swimming effort decreasing along with the increase of buoyancy (i.e. a decrease of density) toward neutral buoyancy, for which the cost of transport is minimal (Miller et al., 2012). This allows the SESs to reduce their swimming effort and, therefore, oxygen consumption rate, enabling them to dive longer. However, no resulting dive duration increment was observed in the shallow group. In the shallow group, the better the body condition (i.e. the higher buoyancy is) of individuals, the less effort was required to perform the same dive. The individual's stroke rate and, therefore, the number of flipper strokes decreased. Thus, a decrease in swimming effort does not necessarily change the dive duration.

Individuals belonging to the deep group dove longer and thus got closer to their ADL. Their decrease in oxygen consumption allowed them to increase their dive durations, particularly the amount of time spent at the bottom of their dive. The bottom phase is the sequence where SES foraging activity is the most important (Guinet et al., 2014; Jouma'a et al., 2016; Thompson and Fedak, 2001). Increasing the length of time spent there allows SES to improve their foraging efficiency and, therefore, the number of prey-capture attempts performed. These results support the notion that larger SESs are less constrained in their diving and foraging ability (Horning, 2012). Our study emphasises the critical ecological importance of buoyancy change through the foraging trip to SES foraging efficiency. Access time (descent and ascent) of the deep group SES increased as their buoyancy improved. This could mean that these individuals tended to dive deeper. In parallel, for these individuals, the number of prey-capture attempts seemed to decrease with depth. Thus, as no differences in how quickly individuals improved their body condition were found between the two groups, the deeper dives performed by individuals in the deep group may allow them to access larger prey than those accessed by the shallow group (Naito et al., 2017). This hypothesis will need to be confirmed in future studies.

Our models were constructed and the energy expenditure estimated based on several hypotheses. First, we used a simple proportionality relationship between mean stroke amplitude and the energy cost of a flipper stroke. Since no studies have been conducted on SESs to measure energy expenditure in relation to exercise intensity, our hypothesis relied on studies conducted on other species, such as dolphins and goldfish (Smit et al., 1971; Williams et al., 1993). Oxygen consumption increases linearly with the force of flipper strokes and curvilinearly with swimming speed in dolphins (Williams et al., 1993). In our study, we measured the mean stroke amplitude corresponding to the mean thrust generated by each flipper stroke. Therefore, based on Williams et al. (1993), we applied a linear relationship to estimate the energy cost of a stroke as a function of mean stroke amplitude. Future studies on the possible variation of energy expenditure associated with a stroke in relation to seal/hind flipper size are needed to confirm or invalidate this hypothesis.

Regarding the effect of temperature on the BMR, few studies have been conducted on body-temperature change and its effect on seal metabolism during dives. Scholander (1940) showed that oxygen consumption during diving was reduced by 70% in marine mammals. Similarly, Castellini et al. (1992) measured an approximately 20% decrease in metabolic rate in Weddell seals during long dives. Cooling of the body due to peripheral vasoconstriction (Irving et al., 1941; McKnight et al., 2019), probably contributes to this hypometabolism (Blix, 2018; Blix et al., 2010; Hill et al., 1987; Scholander et al., 1942). Although studies on specific organs have demonstrated and quantified the effect of temperature (Kuhn and Costa, 2006), no general relationship at the individual level has been established. Indeed, at the brain level, in harp and hooded seals, the temperature can be reduced by up to 3°C. Thus, cerebral oxygen requirement decreases by ∼15–20% or even 25% in some cases (Blix et al., 2010; Odden et al., 1999). However, in elephant seals, Meir and Ponganis (2010) showed that core body temperature was significantly but weakly related to dive duration. Nevertheless, no temperature coefficient (Q₁₀) effect on metabolism was demonstrated. More recent sub-blubber temperature data (our unpublished data), from two SESs with a continuous recording temperature logger, showed a clear 24 h cycle. However, the factors explaining this remain to be understood but appear to be unrelated to the water temperature (our unpublished data). According to our data, the sub-blubber temperature decreased during the dive by about 3°C for a 30 min dive and this is consistent with the data recorded for northern elephant seals (Favilla et al., 2022; Meir and Ponganis, 2010). This temperature change is likely to reflect a combination of several effects: a 24 h diurnal temperature cycle causing a change in core temperature, combined with cooling from colder waters during the dive, possibly over a multi-dive time scale, rather than simply changes in the core temperature (our unpublished data). However, our logger was implanted under the blubber layer, an area whose temperature is more dependent on internal than external temperature (Favilla et al., 2022). In addition, these results were obtained from a single individual. However, to take this phenomenon into account in the BMR calculation, we chose, based on the various studies cited above, to consider a drop in the sub-blubber temperature of 3°C as leading to a 15% decrease in BMR. Changes in the under-blubber and core temperatures and their impact in the metabolism should nevertheless be appropriately investigated in future studies in free-ranging seals (Favilla et al., 2022).

When comparing ADL_t and ADL_b, models integrating a BMR correction, considering the decrease in under-blubber temperature through the dive, estimate the ADL more accurately than models without this correction. This result suggests that temperature might be a significant factor when estimating SES energy expenditure, as in king penguins (Handrich et al., 1997). Compared with ADL_t, the closest estimate of ADL_b was found with the model not integrating HIF, which slightly overestimates the observed ADL. HIF enables the expenditure due to digestion to be included, as it is a significant factor in energy expenditure (Maresh et al., 2015). However, several studies indicate that elephant seals may dissociate foraging from digestion so that digestion is performed during drift dives (Crocker et al., 1997; Meir et al., 2013; Rosen et al., 2007; Sparling et al., 2007a). By isolating foraging and digestion in different dives, the animal can then optimise its use of oxygen reserves. Meir et al. (2013) showed that these different functions would have the same metabolic impact on northern elephant seals.

At the foraging scale, our model suggests that surface time increases as the individual's body condition increases (i.e. seal density decreases) for large deep-diving individuals. We related this to the dive duration, which also increases, as well as the bottom duration. This change could be related to a decrease in mean stroke amplitude in relation to the improvement of buoyancy, combined with increased oxygen body store related to an increase in muscle mass during the foraging trip. Consequently, replenishing this larger oxygen store may require more breaths at the surface and, therefore, a longer surface interval between dives (Génin et al., 2015). Thus, the increase in surface time over the foraging trip may provide a way to evaluate this increase in oxygen stores, muscle-mass gain and a possible increase in blood volume. Although the current study hypothesises that blood volume remains constant, this assumption needs to be verified in the future. Body composition and blood volume studies should allow testing of the hypothesis that the increase in dive and surface durations provides an accurate indicator of oxygen store changes throughout the foraging trip.

Accelerometers provide a unique means to assess seal density and stroke amplitude. Our results support the hypothesis that diving behaviour and oxygen consumption change according to SES size and buoyancy. These changes are characterised by a decrease in estimated diving energy expenditure, allowing individuals to maximise their time at the bottom of the ocean, where SES foraging activity is the most important (Guinet et al., 2014; Jouma'a et al., 2016; Thompson and Fedak, 2001). Combined with measurements of the body composition of individuals at departure and arrival, future studies should evaluate changes in body composition and body mass. An increase in buoyancy, related to a change in the proportion of fat versus lean tissue, can only be achieved by an overall increase in the body mass of seals through the at-sea foraging trip. Therefore, monitoring the change in dive duration, controlled by seal buoyancy and stroke amplitude, may provide useful information for assessing changes in SES body mass and body composition. Finally, new methods for monitoring tissue oxygenation, for example, in the brain, have been developed and are now starting to be used (McKnight et al., 2019). These will allow more detailed monitoring of the oxygenation of different tissues over time to estimate new oxygen-saving mechanisms or detail those known in SESs.

SES data were gathered as part of the ‘Système National d'Observation: Mammifères Echantillonneurs du Milieu Océanique’ (SNO-MEMO, PI C. Guinet) and in the framework of the ANR HYPO2. Fieldwork in Kerguelen was supported by the French Polar Institute (Institut Polaire Français Paul Emile Victor) as part of the CyclEleph programme (no. 1201, PI C. Gilbert) and with the help of the Ornitho-eco programme (no. 109, PI C. Barbraud). Data acquisition was also supported by CNES-TOSCA (Centre National d’Études Spatiales). The authors wish to thank all individuals who, over the years, have contributed to the fieldwork of deploying and recovering tags at Kerguelen Island.