The role of diving respiratory air volume (DRAV) in deep-diving phocid seals remains poorly understood, largely because of the lack of methods for measuring DRAV in free-ranging divers that exhale before diving. We developed a method to estimate DRAV using a hydrodynamic glide model applied to descent glides recorded using multi-sensor data loggers. We estimated dive-by-dive DRAV for six negatively buoyant female southern elephant seals (Mirounga leonina). During shallow descent glides, rapid compression of DRAV influenced net buoyancy and gliding speed, making this phase suitable for estimating DRAV. Our results revealed dive-by-dive variation in DRAV, which was positively correlated with root mean square (RMS) sway acceleration (a proxy for per-stroke effort) and the depth at which gliding began during the initial descent. DRAV increased with both tissue density and maximum dive depth, suggesting that seals adjusted their DRAV to stay closer to neutral buoyancy through their dives. However, the observed level of adjustment did not result in neutral buoyancy at half of the maximum dive depth, as predicted to minimise round-trip locomotion costs. Instead, the seals typically adjusted DRAV to reach neutral buoyancy at∼30 m depth, <10% of their mean maximum dive depth. This indicates that strong negative tissue density imposes transit costs that cannot be fully compensated for by DRAV adjustment alone. Future work should explore whether other breath-hold divers show similar patterns of DRAV adjustment and quantify the associated physiological and ecological benefits.

Field site and study animals

Animal-attached movement data were collected from six, post-breeding female southern elephant seals (Jouma’a et al., 2017; Goulet et al., 2020). The seals were tagged at two locations: the Kerguelen Islands and Península Valdés, Argentina. Tagging occurred in October of 2017 (seals ml17_301 and ml17_280), in October of 2018 (ml18_292, ml18_294a, ml18_294b), and in October of 2019 (ml19_295a). Each seal was anaethetised and then equipped with a head-mounted DTAG-4 sound and movement tag (97×55×33 mm, 200 g in air) as well as a neck-mounted Argos tag (SPOT-293, Wildlife Computers, 72×54×24 mm, 119 g in air). The tags were retrieved when the seals returned to moult in December and January. Full details on the animal handling and the tagging process can be found in Jouma’a et al. (2017) and Goulet et al. (2020), respectively. Ethical approval was provided by the French Committee for Polar Environment and the University of St Andrews Animal Welfare and Ethics Committee.

Dive analysis

Tag-data analysis was conducted in Igor Pro 8 (WaveMetrics, Inc., Lake Oswego, OR, USA) and MATLAB (MathWorks, Inc., Natick, Massachusetts, USA). The tag sampled high-resolution depth (25 Hz) and triaxial acceleration (200 Hz) for over 30 days (Goulet et al., 2020). To capture long-term, independent changes in buoyancy (i.e., tissue density), each 24-hour record was spaced at least six days apart to ensure independent tissue density estimates.

For the analysis, the depth and triaxial acceleration data were downsampled to 1 Hz and 25 Hz, respectively. Speed was calculated as the change in depth per unit time (seconds) divided by sin(pitch) for pitch angles greater than |30| (Miller et al., 2004).

A dive was defined as when the animal dived below 5 m depth (> 5 m), and a glide was defined as the passive movement of an animal through a medium. Only glides below 5 m were used in this study to remove errors in the sway dynamic acceleration due to surface-flipper interactions. Each dive was segregated into four distinct phases: surface, descent, bottom, and ascent phases. The surface phase is when the seal was shallower than 5 m. As defined by Miller et al. (2004), the start of the descent phase was recorded when the animal left the surface and dived deeper than 5 m depth. The bottom phase started when the pitch angle of the seal first exceeded 0, and the ascent phase started at the last point in time when the pitch angle was negative. Since all the elephant seals were negatively buoyant, this study focused on using the descent phase, particularly shallow descent glides, to estimate DRAV at the start of each dive.

Stroke analysis

Flipper movements, or strokes, were defined as side-to-side oscillations, and one stroke cycle refers to the completion of one complete flipper oscillation. The stroking pattern was measured using triaxial specific acceleration in the body frame, which includes surge (anterior-posterior axis), sway (left-to-right axis) and heave (dorsal-ventral axis) accelerations. Given that phocid seals fluke on the left-to-right axis of the body frame, sway acceleration was used to detect strokes. The accelerometer recorded both gravity-based acceleration (reflecting body orientation) and dynamic acceleration (reflecting stroking) (Sato et al., 2003). Gravity-based acceleration is the overall movement of the body axis, while dynamic acceleration captures stroking. A low-pass filter was applied to the measured acceleration of each axis to separate the gravity-based acceleration from the dynamic acceleration. The filter frequency was determined from the power spectral density of the sway acceleration. Following Adachi et al. (2023), gravity-based acceleration was used to calculate pitch. Dynamic acceleration was filtered out by subtracting gravity-based acceleration from the total measured acceleration (Sato et al., 2004). The resulting high-pass filtered acceleration was used to detect strokes and glides, with stroke detection thresholds specific to each individual. Stroke intensity was quantified using the root mean square (RMS) of dynamic acceleration over a half-stroke (the period between two consecutive positive zero-crossings). The RMS of strokes across depths between 5 and 30 m with at least three complete half-strokes during the descent phase were analysed to assess the correlation with DRAV estimates.