Subsurface swimming and stationary diving are metabolically cheap in adult Pacific walruses (Odobenus rosmarus divergens)

The Pacific walrus (Odobenus rosmarus divergens) inhabits a region of the Arctic that is experiencing some of the most rapid loss of summer sea ice (Laidre et al., 2015; Markus et al., 2009). The sea ice is a critical habitat for this species as it provides a location for resting between foraging bouts and immediate access to their benthic offshore foraging habitat (Fay, 1982). The sea ice platform is also used for breeding, and females and young walruses usually haul out onto sea ice throughout the year, while adult males haul out on sea ice primarily during winter and spring (Fay, 1982). In recent years, sea ice has retreated beyond the continental shelf in the eastern Chukchi Sea in the autumn (Markus et al., 2009). In response, female and young walruses are hauling out in large numbers on land during the summer where they either forage for nearshore prey, or travel farther to more productive benthic habitats in offshore areas (Jay et al., 2012). Walruses with only access to land in the summer spend more time in water while not foraging, and less time hauled out resting as compared with walruses with access to sea ice habitats (Jay et al., 2017). In addition, walruses hauling out on land on the northwest coast of Alaska have been documented to migrate to the coast of northern Chukotka in September and October when sea ice is sometimes preserved (Jay et al., 2012). This migration occurs with minimal foraging activity but provides the opportunity to maintain offshore locations (Jay et al., 2012). Migration and extended trips to maintain access to feeding habitat may increase energetic costs while decreasing energy intake, which could have implications for individual health, survival and breeding success (MacCracken, 2012). For instance, female walruses increase body size and lipid deposition during pregnancy to support early lactation, but only 27% of caloric requirements during lactation are met by body reserves (Noren et al., 2014). Thus, foraging during lactation is critical for calf survival and walrus population dynamics are known to be sensitive to successful calf survival (Udevitz et al., 2013). Furthermore, projections of female walrus behaviour with continued sea ice loss suggest that these activity patterns (less time invested on resting and foraging) will become more common in the future (Udevitz et al., 2017).

Understanding the implications of changing behavioural patterns for walrus energetic balance requires information on the energetic demands or metabolic rates associated with individual behaviours. Satellite radio tags deployed on walruses can record location, pressure and conductivity, which together identify whether a walrus is likely in water (foraging or not foraging) or hauled out (Fischbach and Jay, 2016; Jay et al., 2006). This information was recently used in combination with estimated metabolic rates of different behaviours to develop a bioenergetic model for female walruses (Noren et al., 2012; Udevitz et al., 2017). This bioenergetic model allowed prediction of how changes in behaviour in response to sea ice loss or other factors might affect walrus energetic costs. However, model outcomes were most sensitive to estimates of the cost of activity in water (Noren et al., 2012; Udevitz et al., 2017), which has not been directly measured in adult walruses. Data on metabolic rates of walruses are currently limited to a study that reported field metabolic rate (FMR) in two free-ranging adult males (Acquarone et al., 2006) and a study that measured metabolic rates of juveniles while resting in water or during horizontal subsurface swimming (Rosen, 2020). FMR in the two free-ranging adult male Atlantic walruses was measured using doubly labelled water, which provided estimates of the total energetic costs, but not the energetic costs associated with individual behaviours (Acquarone et al., 2006). More recently, Rosen (2020) reported the first data of activity-associated energetic costs in two trained juvenile walruses. Although these first studies have provided important information to help understand the energetic requirements in this species, we still lack direct measurements of the metabolic cost of specific behaviours for adult walruses.

To increase our understanding about the energetic expenditure in the walrus, the present study aimed to measure the metabolic rate during three different in-water behaviours using flow-through respirometry. Three adult female Pacific walruses were trained to perform the experimental procedures for measuring the O₂ consumption rate (V̇_O2, l O₂ min⁻¹): (1) while floating inactive at the water surface (surface); (2) during shallow stationary dives (stationary); and (3) during horizontal subsurface swimming (swim). We hypothesized that the V̇_O2 would be lower for stationary dives and subsurface swimming compared with periods while floating at the surface as previously reported for other pinniped species (Castellini et al., 1992; Fahlman et al., 2013; Hurley and Costa, 2001; Reed et al., 1994; Sparling and Fedak, 2004). In addition, we also hypothesized that the V̇_O2 would be higher during subsurface swimming than that measured during stationary dives.

Three adult (not spayed) female Pacific walruses (Odobenus rosmarus divergens Illiger 1815) born in 2003 and housed at the Oceanogràfic (Valencia, Spain) participated in the present study. Prior to initiating data collection, the animals underwent 7 months of desensitization to the respirometry equipment and were trained to perform the different experiments using operant conditioning. This allowed for data collection in a relaxed physiological state, where the animals were free to decide their participation in each experimental trial and could withdraw at any point. The health of the walruses was assessed daily, and all experiments were approved by the Animal Care and Welfare Committee of Fundación Oceanogràfic de la Comunitat Valenciana (Animal Care number: OCE-1-18), and the Animal Care and Use Committee of the U.S. Geological Survey Alaska Science Centre (Review code 2020-01).

Data collection took place between August 2018 and November 2019. No data were collected during the reproductive season (from around February to March in the study subjects) as during this time the animals showed little interest in participation. Body mass (M_b, kg) was measured before the first feeding in the morning the same week of the metabolic trial (±0–4 days from data collection) and ranged from 682 to 1035 kg (n=3, mean±s.d.=842±116 kg) during the overall experimental period. Ambient pressure, temperature and humidity were measured before commencing each experimental trial (thermometer and hygrometer OH513 Oh Haus & Co), and averaged, respectively, 102.0±0.6 kPa (range=100.9–103.3 kPa,), 19.1±2.8°C (range=13.0–25.0°C) and 69±7% (range=53–84%), during the data collection period. The animals were post-absorptive during metabolic measurements and had not had a meal for at least 15 h prior to initiating the experimental trials, but some food reinforcement was provided during the last minute of most respirometry trials to help reinforce the required experimental behaviour. The regular diet included a mixture of fish (capelin and herring) and molluscs (squid and mussels) supplemented with vitamins. The estimated energy intake (kJ_est, kJ) during each metabolic trial was calculated using the measured amount of consumed food (kg) and the measured energetic content for each type of food (kJ kg⁻¹). This energetic content was estimated in-house through bomb calorimetry of random samples of the food items that arrived at the aquarium.

All metabolic measurements were conducted in a 3 m deep seawater pool with a total water volume of 267 m³. The water temperature at the facility was measured daily and ranged from 12.9 to 18.6°C (mean±s.d.=15.9±1.0°C) during the experimental period. Data collection was performed using an open flow respirometry system described below in this section. The animals were trained to surface into a respirometry dome and keep breathing inside the respirometer while floating in a relaxed and inactive state. The walruses showed some corporal movement while maintaining the required position inside the respirometer. Consequently, the animals were not quiescent throughout the whole duration of the experimental trials. The measured changes in O₂ and CO₂ content of the air exiting the respirometry dome relative to atmospheric air were used to estimate the V̇_O2 and the CO₂ production rate (V̇_CO2, l CO₂ min⁻¹) for each experimental trial. The estimated V̇_O2 was used to define the metabolic rate for the three in-water behaviours: (1) while floating at the water surface [metabolic rate at the surface (MR_Surface, l O₂ min⁻¹)]; (2) during shallow stationary dives [diving metabolic rate during stationary dives (DMR_Stationary, l O₂ min⁻¹)]; or (3) during subsurface swimming dives [diving metabolic rate while swimming (DMR_Swim, l O₂ min⁻¹)].

The MR_Surface was estimated from at least 5 min of data collection while the walrus remained calmly floating inside the respirometer. For DMR_Stationary measurements, the walrus was directed to submerge to a fixed point at the bottom of the pool and to remain stationary while holding its breath for at least 3 min. For the DMR_Swim estimation, the animal swam back and forth between two underwater target poles that were positioned approximately ∼7 m apart, until reaching ∼90 m of horizontal subsurface swimming. The walruses swam at a depth of ∼1.5 m, and all trials were timed to calculate the swimming velocity. These methods are similar to the swimming trials conducted with juvenile walruses by Rosen (2020). To determine either DMR_Stationary or DMR_Swim, the walrus was guided to surface inside the respirometry dome at completion of the dives to measure the gas exchange during the post-dive recovery period. The post-dive recovery period ended when the V̇_O2 and V̇_CO2 had returned to similar values recorded during the last portion of MR_Surface measurements, and remained stable for at least 1 min. Exploratory trials prior to data collection showed that the V̇_O2 and V̇_CO2 returned to steady values within the range (mean±s.d.) of those measured during MR_Surface trials between ∼4 and ∼6 min for the stationary dives and subsurface swimming, respectively (see an example in Fig. 1). Therefore, to ensure the inclusion of the entire physiological recovery for all trials, the post-dive recovery period was extended, respectively, to 5 and 7 min for the stationary dives and subsurface swimming.

The open-flow respirometry system (Fig. 2) consisted of a vacuum pump (FlowKit Mass Flow Generator, FK-500-1, Sable Systems International, Las Vegas, NV, USA) pulling air through a floating transparent Plexiglas dome of 120 cm internal diameter (i.d.) via an 800 cm length and 4.5 cm i.d. plastic corrugated tube. The dome was made buoyant by attaching polyethylene foam to the base. A flow-through rate of 500 l min⁻¹ of air assured that the O₂ and CO₂ were maintained >19% and <2%, respectively. The O₂ and CO₂ content were measured using a fast-response gas analyser (Gemini Respiratory Gas Analyzer, part no. 14-1000, CWE Inc., Allentown, PA, USA), which pulled a subsample of the outlet air from the corrugated tube at a flow rate of 200 ml min⁻¹ via a 310 cm length and 2 mm i.d. firm-walled, flexible tubing. This flexible tubing was attached to a hydrophobic filter (13 mm i.d.), followed by a 60 cm length and 1.5 mm i.d. Nafion^© sample line connected to the gas analyser. The gas analyser was routinely calibrated using ambient air and a commercial gas mixture (5% O₂, 5% CO₂ and 90% N₂; UN1956 Air Liquide, USA). Both, respiratory gas concentrations and air flow rates were captured at 400 Hz using a data acquisition system (Powerlab 8/35, ADInstruments, Colorado Springs, CO, USA), and displayed on a laptop computer running Labchart (v. 8.1, ADInstruments). A simultaneous CO₂ and N₂ dilution test was conducted to evaluate the system for possible leaks, and to the assess the accuracy in measured O₂ and CO₂ concentrations (Fahlman et al., 2008; Fedak et al., 1981), which were within 6% of estimated values. The effective volume of the system was 465 l, which resulted in a time constant of 0.93 min (Bartholomew et al., 1981). The time required to reach a 95% fractional transformation to a new steady state was 167 s (2.79 min) or 3 times the time constant (Bartholomew et al., 1981; Fahlman et al., 2007).

The gas analyser baseline drift during an experimental trial was corrected by measuring the ambient air concentrations at the end of the trial. The mass flow generator automatically corrected flow rates to standard temperature and pressure conditions (STP), and measured flow was corrected for humidity inside the respirometry dome to standard temperature and pressure dry (STPD; Quanjer et al., 1993). Corrected flow rates and measured gas concentrations were then multiplied to calculate the instantaneous V̇_O2 and V̇_CO2, which were corrected for variation in the respiratory exchange ratio (Withers, 1977).

The following criteria were used to include a metabolic trial in the analysis. For all trials, the walrus had to remain calm with all breaths taken inside the respirometer. For experiments while floating at the surface, the walrus had to remain for at least 5 min inside the respirometer. For diving trials, the walrus had to: (1) complete the pre-determined dive duration of 3 min for the stationary dives or the horizontal swimming distance of ∼90 m for the subsurface swimming, and (2) remain inside the respirometer during the post-dive recovery for at least 5 min for stationary dives or 7 min for subsurface swimming.

To determine the total volume of O₂ consumed during each experimental trial, the instantaneous V̇_O2 was integrated over the period from when the walrus surfaced into the respirometer, until the O₂ had returned to ambient levels after leaving the respirometer (Fig. 1). The measured volume of O₂ consumed was then divided by the measurement period for the MR_Surface trials, and by the total dive cycle or the dive time plus the post-dive recovery time for the DMR_Stationary and DMR_Swim trials (see an example in Fig. 1).

The Kleiber ratio was calculated for each behaviour by dividing the measured V̇_O2 by the predicted basal metabolic rate (BMR_est, l O₂ min⁻¹) for a similarly sized terrestrial mammal (BMR_est=0.0093M_b^0.75 l O₂ min⁻¹; Kleiber, 1975). The V̇_CO2 for each behaviour was computed as described for V̇_O2 and used to determine the respiratory exchange ratio (RER, V̇_CO2 divided by V̇_O2). The mass-specific metabolic rate was calculated for each experimental procedure (floating at the surface: sMR_Surface; stationary dives: sDMR_Stationary; subsurface swimming: sDMR_Swim; ml O₂ kg⁻¹ min⁻¹) by dividing the metabolic rate by M_b and assuming isometry. Measured metabolic rates and mass-specific conversions for the different experimental procedures were converted into daily energetic requirements as: metabolic rate (kJ day⁻¹)=(16.218+4.716×RER)×V̇_O2×60×24. This formula is based on the assumptions that protein catabolism is negligible, while carbohydrates and lipids use 20.93 and 19.55 kJ per 1 litre of consumed O₂, respectively (see table 17 in Jungas et al., 1992). The respiratory frequency (f_R, breaths min⁻¹) was computed for each behaviour as the number of observed breaths divided by the total time of the measurement period. The calculated aerobic dive limit (cADL, min) was estimated by dividing the estimated total body O₂ stores by the obtained DMR_Swim (Butler, 2006). The calculations to yield the total accessible body O₂ stores during diving are detailed in the supplementary Materials and Methods and were performed by summing the O₂ capacity of different tissues (blood, muscle and lungs), and by estimating the level of O₂ utilization from each tissue during the dive (Kooyman, 1989).

We initially analysed the relationship between measured MR_Surface, DMR_Stationary, DMR_Swim and the experimental covariates (M_b, and water temperature as continuous variables) using linear models (lm function in R; https://www.r-project.org/). We used a backward stepping procedure implemented by the R-function ‘step’ to identify whether M_b and water temperature were important in affecting metabolic rates, based on the combination of variables that had the lowest Akaike information criterion (AIC). The importance of the covariates in the final models were verified by performing the log-likelihood ratio test to confirm that the full model was better than a model containing just the intercept. Measured MR_Surface, DMR_Station and DMR_Swim were then compared using an ANCOVA F-test that included the experiment as a fixed factor, and the covariates identified from the backward stepping procedure as independent variables. The animal identification (ID) was initially included as a random effect to account for the correlation between repeated measurements on the same individual (Littell et al., 1998), but was removed from the models due to its strong correlation with M_b. A Tukey HSD post hoc test was used to determine whether DMR_Station and DMR_Swim were different from MR_Surface. For all analyses M_b and V̇_O2 were transformed using the base 10 logarithm. Homoscedasticity and normality were assessed through evaluation of residual plots. In the present study, P-values ≤0.05 were considered significant, and data are presented as means±s.d.

The three female Pacific walruses participated in a total of 108 respirometry experimental trials including training and sampling sessions (2018: n=20 trials, 2019: n=88 trials). From the total experimental trials, 5 trials for each animal and behaviour (n=45 for all animals and behaviours) passed the selection criteria to be included in the analysis. For selected trials, the measured M_b ranged from 688 to 1030 kg (835±112 kg; Table 1) and the kJ_est ranged from 96 to 6238 kJ (n=44) which represented 0.1–5.5% of the established total daily energy intake of the animals for the dates of data collection (Table 1). For trials included in the analysis, the measured humidity outside and inside the dome was 71±7% (54–82%) and 69±7% (53–81%), respectively, while the ambient pressure and air temperature at the facility housing the animals were 101.8±0.4 kPa (100.9–102.6 kPa) and 19.2±2.1°C (15.4–23.1°C), respectively. The water temperature for selected trials averaged 15.4±1.0°C (Table 2). The backward stepping procedure showed that the elimination of the covariate ‘water temperature’ from the model increased the AIC for MR_Surface trials (ΔAIC=0.313) and decreased AIC for both underwater behaviours by less than 2 units (DMR_Stationary: ΔAIC=1.28; DMR_Swim: ΔAIC=1.74). For all experiments, the coefficient for water temperature was not significant (Wald's t-test, P-value: MR_Surface=0.18; DMR_Stationary=0.45; DMR_Swim=0.65). Because water temperature either did not improve model fit or had only a minor, non-significant effect on metabolic rate, we did not include it in models comparing metabolic rates between behaviours. Alternatively, the log-likelihood ratio test showed that inclusion of M_b improved the fit of the model compared with the intercept-only model (P<0.05 for all experiments; Table 3). Therefore, we included only M_b in the ANCOVA F-test comparing metabolic rates between behaviours. For each studied behaviour, measured M_b, V̇_CO2, f_R and RER are reported in Table S1, while measured V̇_O2 for each participating walrus and behaviour is reported in terms of daily energetic requirements in Table S2.

The average duration for MR_Surface trials was 5.7±0.6 min (Table 2). The average MR_Surface (4.64±1.04 l O₂ min⁻¹; Table 4) was higher than the predicted BMR_est for a similarly sized terrestrial mammal (1.55±0.17 l O₂ min⁻¹; Fig. 3), and the Kleiber ratio was >1 for all animals (3.0±0.5; range: 2.3–3.9). The MR_Surface increased with M_b with a mass-exponent close to 1 (Table 3 and Fig. 3).

The average breath-hold duration for stationary dives and subsurface swimming was 3.2±0.1 and 1.6±0.2 min, respectively (Table 2), with an average swimming velocity of 1.0±0.1 m s⁻¹ (range: 0.9–1.3 m s⁻¹) during the subsurface swimming. The average post-dive recovery duration for DMR_Stationary trials was 5.2±0.1 min, and 7.2±0.1 min for DMR_Swim trials (Table 2). The MR_Surface differed from those measured during the underwater behaviours (one-way ANCOVA F-test, F_2,41=16.48, P<0.05; Fig. 3), where the DMR_Stationary (3.82±0.56 l O₂ min⁻¹; Table 4) was significantly lower as compared with MR_Surface (Tukey HSD test, mean difference=−0.071, 95% CI [−0.117, −0.026], P<0.05), while DMR_Swim (4.91±0.77 l O₂ min⁻¹, Table 4) and MR_Surface did not differ (Tukey HSD test, mean difference=0.034, 95% CI [−0.011, 0.080], P<0.05). The Kleiber ratio was higher than 1 for all animals and underwater behaviours (Fig. 3) and averaged 2.5±0.3 (range=1.9–2.8) for stationary dives, and 3.2±0.3 (range=2.6–3.5) for subsurface swimming. For an 835 kg adult female walrus the cADL was 7.7 min.

In this study, we report the first metabolic rate estimates from adult individuals of the largest pinniped species yet studied while floating at the water surface and during stationary dives and subsurface swimming in a controlled environment. Our results from the three adult female Pacific walruses are similar to those previously reported for post-absorptive adult pinnipeds, where measured resting metabolic rate (RMR) was 2–3 times greater as compared with BMR_est (Dassis et al., 2012; Fahlman et al., 2008, 2013; Hurley and Costa, 2001; Sparling and Fedak, 2004). The range for sMR_Surface for the adult female walruses in our study overlapped and exceeded the upper end of sMR_Surface measured for juvenile walruses (Rosen, 2020). We hypothesized that the DMR_Stationary and DMR_Swim would be lower than the MR_Surface, and that subsurface swimming would be more energetically costly than stationary dives. However, the results indicated that DMR_Stationary was lower than MR_Surface and DMR_Swim, while DMR_Swim was similar to MR_Surface. These results agree with previous studies in pinnipeds that showed a lower diving metabolic rate (DMR) for inactive dives compared with RMR at the surface (Castellini et al., 1992; Hurley and Costa, 2001), and other studies reporting a DMR_Swim lower or similar to RMR (Castellini et al., 1992; Fahlman et al., 2008, 2013; Sparling and Fedak, 2004). Our results suggest that both stationary diving and subsurface swimming are not energetically costly for walruses. The value obtained for sDMR_Swim in the present study was lower than that measured in juvenile walruses performing similar subsurface swimming trials (Rosen, 2020) and also lower than the mass-specific FMR for free-ranging adult male walruses (Acquarone et al., 2006). The estimated cADL was within the range of dive durations observed in free-ranging walruses (Acquarone et al., 2006; Born et al., 2005; Gjertz et al., 2001; Wiig et al., 1993).

Measured MR_Surface in the present study was similar to results from previous studies reporting basal metabolic rate (BMR) or RMR from post-absorptive and desensitized adult pinnipeds, where measured metabolic rates were 2–3 times greater than that predicted by Kleiber's equation for terrestrial mammals (Dassis et al., 2012; Fahlman and Madigan, 2016; Fahlman et al., 2020, 2008, 2013; Hurley and Costa, 2001; Rosen, 2020; Sparling and Fedak, 2004). While the reasons for the higher measured BMR or RMR in marine mammals as compared with the predictions from Kleiber's remain unclear, it has been suggested that different methodologies and physiological states of the animals during metabolic measurements may be possible reasons for these differences (Lavigne et al., 1986). Kleiber (1975) defined the standard conditions that a study subject should accomplish to obtain comparable metabolic measurements under basal conditions or BMR: (1) reproductively mature individuals, under (2) post-absorptive state of digestion, (3) thermoneutral conditions, and (4) minimal activity level while awake. In the present study, the procedures fulfilled most of these established criteria. The three female walruses were adults, and the experimental procedures took place following an overnight fast. Although a small amount of food was given during the last minute of the trial, this is unlikely to have significantly altered measured metabolic rates considering that it takes approximately 30 min to detect a metabolic increase after food ingestion (Rosen and Trites, 1997). The thermoneutral zone for walruses in water is still unknown, and the water temperatures in their natural environment at the Chukchi Sea (temperature range at the surface: from 0 to 10°C, and at 30 m depth: from −1 to 6°C; Luchin and Panteleev, 2014) are lower than those measured during selected trials in the present study (12.9–17.1°C). However, measured metabolic rates were not related to the range of water temperatures during the metabolic trials in the current study. Therefore, owing to the lack of information about the thermoneutral range in this species, we do not propose that the measured MR_Surface in the current study are estimates of BMR for the walrus.

In addition, the animals were trained to maintain a relaxed and inactive state during metabolic measurements to allow for measuring RMR, which is defined as the metabolic rate measured at rest when other conditions for BMR are not met. However, it is worth mentioning that the measured MR_Surface may have been higher than actual RMR because the walruses had to move to maintain their upright position inside the respirometry dome. In fact, juvenile walruses that remained motionless while in a respirometry dome (Rosen, 2020) had slightly lower sMR_Surface (73.1–166.1 kJ kg⁻¹ day⁻¹) than the adult females in our study (126.3–203.9 kJ kg⁻¹ day⁻¹; Table S2), despite higher sMR_Suface being expected for juvenile individuals because of their higher mass-specific metabolic rates and investment in growth and development (Costa, 1993). Other studies have similarly noted difficulty maintaining the animals motionless when in water (Dassis et al., 2012; Fahlman et al., 2008). For example, in Patagonia sea lions (Otaria flavescens), measured RMR on land was 1.4 times lower compared with MR_Surface, which was attributed to movement inside the respirometry dome (Dassis et al., 2012). Still, the movement shown by the participating walruses at the water surface in this investigation may be consistent with the actual behaviour in their natural environment, where it could be difficult to maintain a motionless position at sea surface conditions. Therefore, although the obtained Kleiber ratio for measured MR_Surface in the present study was similar to BMR or RMR previously reported for pinnipeds, we suggest that measured MR_Surface in the current study may be elevated as compared with RMR because of observed movement while in the respirometer, but may represent a more appropriate approximation of the actual MR_Surface for those animals in the wild. In addition, metabolic rates have been reported to scale exponentially with M_b with a mass-exponent of 0.75 (Kleiber, 1975). While in the present study obtained MR_Surface closely followed a linear relationship with M_b (see Table 3 and Fig. 3), the limited M_b range may not allow an appropriate analysis to determine whether the relationship is linear or exponential.

On average, DMR_Stationary of the adult female walruses was lower as compared with MR_Surface and DMR_Swim, which is consistent with previous studies investigating stationary or inactive dives in pinnipeds (Castellini et al., 1992; Hurley and Costa, 2001). For example, in trained California sea lions (Zalophus californianus), Hurley and Costa (2001) reported lower DMR_Stationary than MR_Surface for shallow dives (see fig. 3 in Hurley and Costa, 2001). In freely diving Weddell seals (Leptonychotes weddellii), Castellini et al. (1992) reported metabolic rates during short sleep or inactive apneas (<14 min) below MR_Surface (see fig. 3 in Castellini et al., 1992). The same authors also reported that the overall DMR_Swim during free diving was 1.5 times greater than that observed for sleep apneas of similar duration (Castellini et al., 1992), which is similar to the results from our study in which DMR_Swim was 1.3 times greater than DMR_Station.

Alternatively, DMR_Swim did not differ from MR_Surface which agrees with previous studies in pinnipeds that reported a DMR_Swim close to or below RMR in freely diving animals (Castellini et al., 1992; Sparling and Fedak, 2004) or in subjects trained to swim vertically (Fahlman et al., 2008, 2013). In addition, these studies also reported a decreased DMR_Swim with increased dive duration, suggesting that pinnipeds can decrease their energetic investment depending on the dive requirements. This extended diving hypometabolism was first described by Scholander (1940), and was suggested to be an ability to reduce metabolism by decreasing the activity of different metabolic processes (e.g. digestion, liver and kidney function, etc.), that would increase the dive duration. Although changes in DMR with increased dive duration were not assessed in the present study, our results suggest an efficient O₂ use during underwater behaviours in the walrus. This physiological capacity could explain the reduced DMR_Stationary and how measured DMR_Swim was similar to MR_Surface despite the additional energetic investment during subsurface swimming.

Despite the similarity of the procedures during subsurface swimming trials in two juvenile walruses trained to swim horizontally for 110 m, at a depth of ∼1.2 m, and at constant velocities (male: 1.6±0.1 m s⁻¹; female: 2.1±0.1 m s⁻¹), the obtained average sDMR_Swim (male: 242.9±60.0 J kg⁻¹ min⁻¹; female: 336.6±84.4 J kg⁻¹ min⁻¹; Rosen, 2020) were up to 2.8 times greater than that measured in the present study (121.3±7.0 J kg⁻¹ min⁻¹). However, Rosen (2020) noticed that the estimated cost of transport during subsurface swimming in these juvenile walruses was elevated over to that predicted for a similarly sized marine mammal, which was suggested to be related with non-optimal swimming speed and with the higher underlying maintenance costs of young individuals (Rosen, 2020). Thus, the decreased sDMR_Swim reported in the adult female walruses may be partly related with the lower swim speeds (1.0±0.1 m s⁻¹) and distance (90 m) for subsurface swimming trials in the present study, but also to the decrease in mass-specific metabolic rates with age (Kleiber, 1975). In addition, obtained RER ranges for all experimental behaviours (0.81–1.06; see Table S1) were within expected values. While the RER immediately following a dive tends to change dynamically as the blood and tissues are recovering (Fahlman et al., 2008; Reed et al., 1994), the average for RER during trials at the water surface (0.97±0.05) were similar to those obtained during the post-dive recovery period for both underwater experiments (stationary dives: 0.94±0.07; subsurface swimming: 0.94±0.06), and agreed with data reported for resting periods at the surface and for conditioned dives in Steller sea lions (see table 1 in Fahlman et al., 2008).

When translating the results obtained in this study to free-ranging walruses, we found similar characteristics of subsurface swimming behaviour with those previously reported in free-ranging animals that were associated with travelling or exploratory behaviours (Gjertz et al., 2001; Jay et al., 2001). These dives are reported to be shallow (<4 m), with most of them lasting <3 min, and occurring most frequently at the beginning and ending of foraging trips (Gjertz et al., 2001; Jay et al., 2001). The associated ascent and descent swimming velocity reported for these short dives is ∼0.5 m s⁻¹, but exceeds 1 m s⁻¹ for longer and deeper dives (Gjertz et al., 2001; Jay et al., 2001). Thus, considering that the animals in the present study performed the subsurface swimming trials at their individual comfort, and the similarities with described underwater behaviour of wild walruses, we suggest that measured DMR_Swim could be considered as a suitable approximation of the energetic swimming requirements for walruses when travelling to and from foraging areas or over long-distance migrations. However, the experimental procedures at facilities housing animals are not usually representative of the wild, and differences in underwater behaviours (e.g. social interactions, exploratory behaviours, etc.), or environmental conditions (e.g. water current, water temperature, etc.) experienced by free-ranging walruses could result in different energetic requirements during diving and swimming. This could explain why the observed sDMR_Swim in the present study (see Table S2) is ∼2–3 times lower than the mass-specific FMR measured in two male adult Atlantic walruses, as that approximation accounted for behaviours in the wild (Acquarone et al., 2006). Despite this inherent limitation, these novel metabolic measurements on adult female walruses in water, could help improve energetic requirements estimations during these travelling subsurface dives, while helping interpret measurements of FMR associated with behavioural information.

The estimated cADL in the present study (7.7 min) was longer than the dive duration for both underwater behaviours (stationary dives: 3.2±0.1 min; subsurface swimming: 1.6±0.2 min) and the duration of travelling and foraging dives observed in free-ranging male Pacific walruses (median dive duration = 7.2 min; Jay et al., 2001) that were suggested to occur within the aerobic dive limit (ADL). Therefore, the dives by females in our study were well within the estimated cADL and likely supported by aerobic metabolism. In addition, the cADL of the walruses in this study was also longer than the average durations for most frequent dives (5–6 min) observed in adult male walruses foraging over continental shelf habitats at depths from 10 to 70 m (Acquarone et al., 2006; Gjertz et al., 2001; Wiig et al., 1993). However, walruses also take deeper and longer dives, exceeding 100 m depth (Acquarone et al., 2006) with a maximum reported dive depth of 234 m (Born et al., 2005) and a dive duration of 24 min (Gjertz et al., 2001). While, reported cADL in the present study could be overestimated as a result of lower energetic demands derived in laboratory conditions, if hypometabolism is related to dive duration in the walrus as in other pinnipeds, cADL may in fact longer during extended dives. Thus, further investigations of the metabolic requirements for dives of different duration in this species would help clarify whether these animals are able to achieve the reported prolonged dive durations (>10 min; Acquarone et al., 2006; Gjertz et al., 2001; Wiig et al., 1993) while avoiding the blood lactate accumulation.

Walruses are currently facing changes to their environment that have altered the availability of their sea ice habitat, which have resulted in changes in their travelling and foraging behaviour, and geographical distribution (Jay et al., 2012; Udevitz et al., 2017). This has led to an increased concern about the future of the walrus populations, raising interest in studies aimed to obtain information about basal, reproductive, or diving and swimming energetic demands (Noren et al., 2012, 2014; Rosen, 2020; Udevitz et al., 2017). Studies measuring the energetic costs for different behaviours and life stages will help improve the accuracy of previous energetic models to predict the potential impacts of environmental change (Udevitz et al., 2017). In the present study, we report the first direct measurements of metabolic rates in female adult Pacific walruses while floating in water and during short and shallow stationary dives and subsurface swimming. The results are consistent with previous work reporting a low energetic cost of diving in marine mammals and confirms that the walruses are able to reduce their metabolic rates during both stationary dives and subsurface swimming. These metabolic estimates would help quantify energetic consequences of reported increased in-water behaviours in the walrus (Jay et al., 2012). In addition, as a benthic feeding marine mammal, walruses spend 70–93% of their time in water, of which much of this time is spent diving to the ocean floor to feed on benthic invertebrates (Jay et al., 2017; Udevitz et al., 2009). Therefore, while the data reported in this study would help improve bioenergetic models, further investigations measuring the metabolic rate for different types of dives (e.g. longer durations or higher swimming requirements), and/or environmental parameters (e.g. water temperature or wind and water current), will help quantify the effect of these parameters on metabolic rates and further improve these models. Similarly, measurements of the metabolic demands while resting on land would add relevant information that will also improve theoretical models aimed at forecasting the conservation scenario of this species.

The authors thank the personnel at the Oceanogràfic (Valencia, Spain) who helped with the initial management of the experimental set up, particularly to Roberto Tejero and Daniel García-Párraga. We are especially grateful to all the animal care professionals at the Artico facility, whose perseverance, positive attitude and enthusiasm for science and conservation made possible the accomplishment of this project (Noni, Carlos, Jose, Lara, Raúl y Guaci). A special thanks to Andrés Jabois for their constant assistance, advice and support, and to Chad Jay at the U.S. Geological Survey who initiated and advised on this study. Lastly, thanks to several interns from Artico and Research departments that gently helped with the equipment and procedures when necessary. Any use of trade, firm, or product names is for descriptive purposes only and does not imply endorsement by the U.S. Government. We are grateful for the constructive comments provided by two anonymous referees that helped improve the final version of this article.