Turtle

We compared the stroke frequencies and swimming speeds of a range of animals in relation to their body sizes. Owing to morphological differences among species, body mass was used as an index of body size. Morphological measurements were used to estimate mass for adult Weddell seals (Sato et al. 2003 (link)), leatherback turtles and sperm whales (Miller et al. 2004 (link)). For the killer whale, we used typical sex-specific masses reported in the literature (Williams 1999 ; Rohr & Fish 2004 (link)). Mass of the other species was measured directly using balances. We recorded the behavioural context of each species during the period when data were collected (table 1). We specified migrating contexts, including short-distance translocations, and breath-hold diving for foraging as periods when animals are expected to swim efficiently. Seabirds were classified into one of two groups: those with a specialized swimming organ (e.g. flippers of penguins and feet of shags); and those that use the same organ for both flight and swimming (wings of auklets, guillemots and razorbills).
Field experiments using accelerometers were conducted from tropical to Antarctic regions. Detailed protocols of the field experiments were already published for the sperm whale (Amano & Yoshioka 2003 ; Miller et al. 2004 (link)), Weddell seal (Sato et al. 2003 (link)), Baikal seal (Watanabe et al. 2004 ), finless porpoise (Akamatsu et al. 2002 ), emperor penguin (Sato et al. 2005 (link)), king penguin (Sato et al. 2002 (link)), Adélie penguin (Sato et al. 2002 (link)), macaroni penguin (Sato et al. 2004 (link)), little penguin (Watanuki et al. 2006 (link)), Brünnich's guillemot (Watanuki et al. 2003 ), European shag (Watanuki et al. 2005 (link)), common guillemot (Watanuki et al. 2006 (link)), razorbill (Watanuki et al. 2006 (link)), rhinoceros auklet (Watanuki et al. 2006 (link)), chum salmon (Tanaka et al. 2001 (link)) and Japanese flounder (Kawabe et al. 2004 ). The location and time of the studies for other animals were as follows: killer whales (Tysfjord, Norway, November 2005; and SE Alaska, USA, July 2006); chinstrap penguins (Signy Island, South Atlantic, January 2001); gentoo penguins, black-browed albatrosses and South Georgian shags (Bird Island, South Georgia, January 2005); southern elephant seal (Kerguelen Islands, South Indian Ocean, December 2002); northern elephant seal (California, USA, March 2003); streaked shearwaters (Sangan Island, Japan, September 2004); and leatherback turtles (French Guiana, South America, May 2001–2004). Study protocols followed those of the above-mentioned published studies.
We used acceleration data loggers (D2GT and PD2GT, Little Leonardo Ltd, Tokyo; Dtag, the Woods Hole Oceanographic Institution; Johnson & Tyack 2003 ) to detect the stroking movement and the swim speed of animals. The D2GT was 15 mm in diameter, 53 mm in length, with a mass of 16 g in air, and recorded depth, two-dimensional acceleration and temperature. The D2GT was deployed on smaller species of penguins (macaroni and little) and flying seabirds. The Dtag (150 g in air) was used to study killer whales. Both the PD2GT and the Dtag were used for the sperm whales (two whales by PD2GT and nine whales by Dtag). The swim speed was calculated from the pitch angle from longitudinal acceleration and vertical velocity from depth data (Watanuki et al. 2003 ; Miller et al. 2004 (link)). There are two types of PD2GT depending on the memory size. They were used for the other animals and are 27 or 22 mm in diameter, 128 or 122 mm in length, with masses of 73 or 101 g in air, and recorded swim speed, depth, two-dimensional acceleration and temperature. The swim speed was converted from the rotation of an external propeller using a calibration line that was estimated for each animal (Sato et al. 2002 (link), 2003 (link)). According to calibration experiments of the PD2GTs using a water circulation tank (Akamatsu et al. 2002 ; Kawabe et al. 2004 ) or swimming pool (Tanaka et al. 2001 (link)), linear relationships between rotation number of propeller and water flow speed were obtained with a high coefficient of determination larger than 0.9, which enabled us to compare swim speeds among species. The mean swim speeds were calculated during propulsive swimming. For example, the speed during the ascent phase was excluded from analyses for penguins and seabirds because they glided up to the surface using buoyant force (Sato et al. 2002 (link); Watanuki et al. 2003 ). The accelerometers can measure both dynamic acceleration (such as propulsive activities) and static acceleration (such as gravity). Low-frequency components of the longitudinal acceleration, along the long axis of the body, were used to calculate the pitch angle of the animals (Sato et al. 2003 (link)).
We could detect the duration of each stroke cycle from the time-series data, but our goal was to determine the dominant stroke cycle frequency for each animal. The periodic properties of the acceleration signal allowed us to apply a Fourier Transform to determine the dominant frequency. Power spectral density (PSD) was calculated from the entire acceleration dataset of each animal, or a subsample during identified foraging or migration behaviour to determine the dominant stroke cycle frequency using a Fast Fourier Transformation with a computer program package, Igor Pro (WaveMetric, Inc., Lake Oswego, OR, USA). For the sperm whale, the bottom phase of the dive was not used as it is typified by body rotations, which can occur at similar rates to the fluking action.

Since this study began in 1981, age in years upon initial capture of each turtle was estimated as the number of winters post-hatching. Age estimation is the same every year and, thus, our methods for age determination apply to all data considered herein (population and immune subsample). Because only a single scute annulus is produced each year in this population, age of immature turtles usually equals the total number of annuli present. However, during extremely harsh years, there may not be any growth, and two annuli may appear as one. Hence, counts of annuli are minimum ages, and might be underestimated if annuli are not counted and evaluated in the context of general shell growth patterns in the whole population. By comparing the pattern of the increments in scute growth of an unaged juvenile turtle with the general pattern of scute growth in the population, age-at-first-capture could be reliably determined up to at least 12 years. We counted the minimum number of annuli on adult turtles captured in the early years of the study, and age was estimated for all of those turtles if the number of annuli was less than 20. However, most turtles were initially marked individually and aged before their sixth winter, and aged subsequently based on the actual recapture interval. Adult females were significantly older (and smaller) than males in our samples (Table S3; unpaired t-test, P = 0.0003, See Fig. S2 and Fig. 5 in [33 ]). Therefore age was z-transformed within each sex to a mean of 0 with unit variance. This variable “zAge” was used in all statistical analyses. We excluded data from 26 unsexed juveniles.

We conducted radio-telemetry surveys from 2014 to 2019. During the turtle active season, from May to mid-September, we tracked turtles two to three times weekly, whereas in the non-active season, we tracked once monthly. Each year, a subset of headstarted turtles were outfitted with radio transmitters (Advanced Telemetry Systems R1600) and tracked for at least one year. To determine which turtles would receive a radio transmitter, headstarted turtles from each cohort were separated into two groups based on presumed sex (i.e., incubation temperature). Each turtle was assigned a unique number and using the random number function (= RAND) in Excel, a turtle was selected to be outfitted with a radio transmitter. Each year, more female turtles were selected for radio transmitter attachment. The combined weight of the transmitter and the epoxy was approximately 10 g, which was less than 10% of the average turtle body mass. Headstarted turtles were tracked using an R410 or R4000 Receiver (Advanced Telemetry Systems, Inc., MN). We tracked 10 turtles from the 2014 release cohort, 21 turtles from 2015, 24 from 2016, 16 from 2017, 22 from 2018, and 23 from the 2019 release cohort. We excluded the 10 individuals from the 2014 release cohort in our analysis because of missing data. Most turtles were only tracked for a year, but a few headstarted turtles remained in the study to be radio-tracked for 2–5 additional years (Table 1). The turtles that remained in the study were affixed with a new radio transmitter each year.