Energy expenditure of adult green turtles (Chelonia mydas) at their foraging grounds and during simulated oceanic migration

Functional Ecology 2016, 30, 1810 1825 doi: 10.1111/1365-2435.12667 Energy expenditure of adult green turtles (Chelonia mydas) at their foraging grounds and during simulated oceanic migration Manfred R. Enstipp*,,1,2, Katia Ballorain 1,2,Stephane Ciccione 3, Tomoko Narazaki 4, Katsufumi Sato 4 and Jean-Yves Georges 1,2 1 IPHC, Universite de Strasbourg, 23 rue Becquerel, 67087 Strasbourg, France; 2 UMR 7178, CNRS, 67037 Strasbourg, France; 3 Kelonia, l observatoire des tortues marines, BP 40, 97436 Saint Leu, La Reunion, France; and 4 Atmosphere and Ocean Research Institute, University of Tokyo, 5-1-5 Kashiwanoha, Kashiwa, Chiba 277-8564, Japan Summary 1. Measuring the energy requirements of animals under natural conditions and determining how acquired energy is allocated to specific activities is a central theme in ecophysiology. 2. Turtle reproductive output is fundamentally linked with their energy balance so a detailed understanding of marine turtle energy requirements during the different phases of their life cycle at sea is essential for their conservation. 3. We used the non-invasive accelerometry technique to investigate the activity patterns and energy expenditure (EE) of adult green turtles (Chelonia mydas) foraging year-round at a seagrass meadow in Mayotte (n = 13) and during simulated oceanic migration (displacement from the nesting beach) off Moheli (n = 1), in the south-western Indian Ocean. 4. At the foraging site, turtles divided their days between foraging benthically on the shallow seagrass meadow during daylight hours and resting at greater depth on the inner side of the reef slope at night. Estimated oxygen consumption rates (s _V O2 ) and daily energy expenditures (DEE) at the foraging site were low (s _V O2 during the day was 16 and 19 times the respective resting rate at night during the austral summer and winter, respectively), which is consistent with the requirement to build up substantial energy reserves at the foraging site, to sustain the energy-demanding breeding migration and reproduction. 5. Dive duration (but not dive depth) at the foraging site shifted significantly with season (dive duration increased with declining water temperatures, T w ), while overall activity levels remained unchanged. In parallel with a significant seasonal decline in T w (from 289 01 C to 253 04 C), there was a moderate ( ~ 19%) but significant decline in DEE of turtles during the austral winter (901 111 kj day 1 ), when compared with the austral summer (1117 66 kj day 1 ). 6. By contrast, the turtle moved continuously during simulated oceanic migration, conducting short/shallow dives in the day, which (predominately at night) were interspersed with longer and deeper pelagic dives. Estimated oxygen consumption rates during a simulated migration (125 016 ml O 2 min 1 kg 083 ) were found to be significantly increased over the foraging condition, equal to ~ 3 times the resting rate at night (042 002 ml O 2 min 1 kg 083 ), and daily energy expenditure amounted to 2327 292 kj day 1, underlining the tremendous energetic effort associated with breeding migration. 7. Our study indicates that the accelerometry technique provides a new and promising opportunity to study marine turtle energy relations in great detail and under natural conditions. Key-words: accelerometry technique, diving, energetics, foraging, marine turtles, metabolic rate, migration, oxygen consumption, partial dynamic body acceleration *Corresponding author. E-mail: manfred.enstipp@iphc.cnrs.fr Present address. Institut Pluridisciplinaire Hubert Curien (IPHC), Departement Ecologie, Physiologie et Ethologie (DEPE), UMR 7178 CNRS-Universite de Strasbourg (UdS), 23 Rue Becquerel, 67087 Strasbourg Cedex 2, France. 2016 The Authors. Functional Ecology 2016 British Ecological Society

Green turtle energy expenditure 1811 Introduction The dynamic flow of energy within ecosystems is a central aspect to understanding how they function (Lindeman 1942). All living organisms must obtain energy from the environment. The rate at which they take up energy and materials, transform them and allocate them to survival, growth and reproduction represents their metabolic rate, which sets the pace of life (Brown et al. 2004). Metabolic rate plays a central role at various ecological scales. On the individual level, metabolic rate provides an index for the cost of living of an organism and allows important insights into how a variety of environmental, behavioural and physiological factors affect its energy requirements (Williard 2013). Beyond that, metabolic rate controls ecological processes at all levels of organization: from individual survival, growth and reproduction to population dynamics and ecosystem processes (Brown et al. 2004). Hence, measuring the energy requirements of animals under natural conditions and determining how the acquired energy is allocated to specific activities is a central theme in ecophysiology (Butler et al. 2004). Studying the energy requirements of marine top predators/megafauna (e.g. marine mammals, seabirds, marine turtles, large predatory fish) is not an easy task, since these animals spend most, if not all, of their life at sea, far away from shore. A variety of methods have been used to study the energy expenditure of animals in the field (Speakman 1997; Fort, Porter & Gremillet 2011; Green 2011). The pros and cons of two widely used methods, the doubly labelled water (DLW) method and the heart rate method, were discussed in detail by Butler and colleagues (Butler et al. 2004). Apart from issues concerning resolution/scale, one of the most important differences between these two methods is the period over which measurement is possible, which for the DLW method is in the order of days. By contrast, recent developments in solid-state electronics allow the heart rate method to measure energy expenditure over extended periods. For example, heart rate recordings have been used to estimate the energy expenditure of Antarctic fur seals (Arctocephalus gazella) at sea for periods of up to 3 weeks (Boyd et al. 1999). In seabirds, yearround recordings have been achieved: great cormorants (Phalacrocorax carbo carbo) in Greenland were implanted with heart rate data loggers to study their behaviour, physiology and energetics over the annual cycle (Gremillet et al. 2005; White et al. 2011), and similar studies were conducted with macaroni penguins (Eudyptes chrysolophus) (Green et al. 2005, 2009a) and eider ducks (Somateria mollissima) (Guillemette & Butler 2012). Marine turtles are large ectotherms with a low metabolic rate (nearly an order of magnitude lower than endotherms of equal size and body temperature; Bennet & Ruben 1979), which allows them to use their energy reserves sparingly. While there has been a tremendous increase in studies to track movements (e.g. Hays et al. 1999; Cheng 2000; Godley et al. 2003) and activity patterns (e.g. Hays et al. 2001; Rice & Balazs 2008; Blumenthal et al. 2010; Fossette et al. 2012) of marine turtles during different life-history stages, investigations concerning the energetics of marine turtles in their natural environment are less common. A detailed understanding of marine turtle energy requirements during the different phases of their life cycle is essential for their conservation, since a number of key demographic parameters are fundamentally linked with their energy balance (Bjorndal 1985; Hays 2000, 2008; Jones & Seminoff 2013). Accordingly, the rate of nutrient uptake (i.e. energy acquisition) and energy expenditure at their foraging site are major determinants of the interbreeding interval (the time between successive breeding migrations) and, hence, reproductive output (Bjorndal 1985; Hays 2008). Unless turtles reach a certain threshold body condition, they might not be able to breed (Hays 2000; Plot et al. 2013), with consequences for the development of the population in question. Hays (2008) identified completing energy budgets for species over ecologically relevant time-scales as one of the remaining key issues for our understanding of marine turtle biology and conservation needs. Green turtles (Chelonia mydas L. 1758) have a global tropical to subtropical distribution and a distinct life cycle, though recent studies have demonstrated a considerable behavioural plasticity (Hatase et al. 2006; Gonzales Carman et al. 2012; Cheng, Bentivegna & Hochscheid 2013). Upon hatching, green turtles typically take up a pelagic existence for 3 5 years as omnivores before they enter shallow coastal areas, switch to a herbivorous existence (mainly feeding on seagrass and/or algae) and grow (Reich, Bjorndal & Bolten 2007). Once they reach sexual maturity, adult turtles embark on regular breeding migrations. Mating typically occurs in the waters off the nesting beach, after which females will haul themselves on to the nesting beach to deposit their eggs, from which a new generation will hatch (Bolten 2003). Breeding migration, including mating and repeated nesting, is an enormous physical task. First, breeding migrations can cover substantial distances, depending on the population in question, during which turtles might travel hundreds of kilometres to and from their nesting sites (Mortimer & Carr 1987). Secondly, hauling out onto the beach to build a nest and deposit the eggs is physically demanding and oxygen consumption rates of nine to ten times the resting rate have been measured (Prange & Jackson 1976; Jackson & Prange 1979). Nesting has been compared with a human athlete running a marathon race and green turtles repeat this endeavour several times during their ~2-month breeding season (Jackson 2011). Lastly, green turtles from some populations are believed to fast from the time they leave their foraging grounds until they return several months later (Carr, Ross & Carr 1974; Hays et al. 2002b), though it has been suggested that turtles might minimize fasting periods by (i) feeding during the internesting period (Hochscheid et al. 1999; Hays et al. 2002b), (ii) combining oceanic crossing with extended coastal migration, permitting benthic foraging

1812 M. R. Enstipp et al. (Cheng 2000; Godley et al. 2002; Hays et al. 2002c; Blumenthal et al. 2006), and (iii) nocturnal feeding on upward-moving macroplankton during the oceanic phase of migration (Hatase et al. 2006; Rice & Balazs 2008; Seminoff et al. 2008; Cheng, Bentivegna & Hochscheid 2013). Nevertheless, green turtles need to build up substantial energy reserves before they can breed. In the extreme case of green turtles nesting on Ascension Island, Prange (1976) calculated that an adult turtle would need to build up a fat store of ~37 kg, equivalent to ~21% of its body mass, to complete the task of moving to and from the breeding grounds, not including the energy required at the breeding grounds. A more recent estimate by Hays & Scott (2013) arrived at a similar figure of ~31 kg fat for a 150-kg green turtle for the entire breeding endeavour. Faced with enormous logistic and methodological difficulties, energetic investigations into the metabolic rates of marine turtles in their natural environment have been scarce (Enstipp et al. 2011; Williard 2013). However, in recent years, accelerometry has emerged as a promising tool both to investigate animal behaviour (e.g. Fossette et al. 2012) and to estimate activity-specific metabolic rates of animals in the field with a high temporal resolution (Wilson et al. 2006; Halsey, Shepard & Wilson 2011). Required calibration studies that relate the metric derived from the recording of body acceleration [overall dynamic body acceleration (ODBA), or partial dynamic body acceleration, (PDBA)] to simultaneously measured metabolic rate (typically using respirometry) have been conducted for a wide range of animals, usually under controlled captive conditions (Halsey et al. 2009), but also in the wild (Elliot et al. 2013; using the DLW method). Significant relationships between body acceleration and metabolic rate have been found not only for terrestrial species, but also for a range of aquatic species (e.g. Fahlman et al. 2008; Payne et al. 2011), including marine turtles (Enstipp et al. 2011; Halsey et al. 2011a). The south-western Indian Ocean hosts important nesting and feeding areas for marine turtles, especially green turtles (e.g. Comoros Islands, Mayotte Island, Scattered Islands) (Bourjea et al. 2007, 2015; Lauret-Stepler et al. 2007; Ballorain et al. 2010). This setting provides a great opportunity to study the energy requirements of these animals during different phases of their life cycle (foraging site vs. breeding migration) in great detail and under natural conditions, using the non-invasive accelerometry technique. As turtles are present year-round, this setting also allows to investigate the effects that seasonal changes in environmental conditions (e.g. water temperature, light conditions, food availability) might have on behavioural patterns and energetics (Southwood, Darveau & Jones 2003a; Hochscheid, Bentivegna & Speakman 2004; Ballorain et al. 2013; Williard 2013). The aims of our study were to test the hypotheses that (i) green turtles at their foraging site will deploy a strategy of minimizing energy expenditure, thus maximizing energy gain, and (ii) energy expenditure of green turtles during breeding migration, when turtles might travel continuously, will be substantially increased over the foraging situation. To investigate these issues, we recorded dive patterns and body acceleration of adult green turtles (i) foraging year-round at a seagrass meadow in Mayotte and (ii) during simulated oceanic migration (displacement from the nesting beach) off Moheli. Materials and methods LOGGER DEPLOYMENTS Foraging site (Mayotte) In the Bay of N Gouja (Mayotte Island, Comoros Archipelago, south-western Indian Ocean, 12 58 0 S, 45 05 0 E), juvenile and adult green turtles exploit shallow seagrass meadows on a yearround basis (Taquet et al. 2006; Ballorain et al. 2010). Between April 2007 and September 2008, we equipped 13 adult green turtles (Chelonia mydas L. 1758; seven males, six females; mean curved carapace length (CCL): 984 46 cm; mean body mass (M b ): 1094 140 kg) with one of the following data loggers: W190L-PD2GT (length 114 mm, diameter 21 mm, mass 70 g) or M190L-D2GT (length 53 mm, diameter 15 mm, mass 18 g; both Little Leonardo, Tokyo, Japan). Both loggers were set to record dive depth (resolution: 005 010 m) and water temperature (resolution: 0018 005 C) at a frequency of 1 Hz, and biaxial acceleration (head to tail: surge, and ventral to dorsal: heave ; 12-bit resolution; recording range 3 g) at a frequency of 16 Hz. This provided a recording duration of up to 129 hours (mean: 107 27 h) after which the memory was filled. Turtles were initially captured by snorkelers in the shallow bay, restraining the turtles pectoral fins by hand. They were moved to the nearby beach and placed in a shaded wooden enclosure, where CCL and body mass were recorded, using measuring tape and an electronic spring balance (600 02 kg; Kern & Sohn, Balingen, Germany), respectively. A wire mesh was glued to the centre of the second vertebral scute of the turtle carapace using fast-setting epoxy. Before the turtle was released from the beach, the logger was fixed to the wire mesh in anterior posterior direction with cable ties. This allowed for easy retrieval and reattachment (multiple deployments) of the logger underwater, not requiring further capture of the animal (for details concerning turtle capture and attachment procedure, see Ballorain et al. 2013). Care was taken to attach data loggers in the same position/orientation during all deployments. Upon retrieval of the logger, data were downloaded onto a PC. Simulated oceanic migration (Moheli) In July 2010, a female green turtle (CCL: 110 cm; M b estimated after Hays et al. 2002a: 1545 kg) was captured at its nesting beach upon the completion of nesting (Itsamia Beach, Moheli Island, Comoros Archipelago, 12 19 0 S, 43 44 0 E) and equipped with a data logger (W1000-3MPD3GT; length 175 mm, diameter 26 mm, mass 140 g; Little Leonardo, Tokyo, Japan) attached to the centre of the fourth vertebral scute (see Data Analysis). The logger was programmed to record dive depth (range 0 to 1000 m, resolution 025 m) and water temperature (range 20 to 50 C, resolution 002 C) at a frequency of 1 Hz and triaxial acceleration (heave, surge and sway; range 490 ms 2, resolution 001 m s 2 ) at a frequency of 16 Hz. It also recorded speed and geomagnetic intensity as part of a different study, not reported

Green turtle energy expenditure 1813 here. The turtle was transported by boat to its oceanic release site, 150 km south-west of Itsamia Beach, and released immediately (water depth >1000 m). Post-nesting migrations of green turtles from the Moheli population cover a wide geographical area, from Madagascar in the east to the East coast of Africa in the west, including Mozambique, Tanzania, Kenya and Somalia (Goossens, A et al., unpublished data). The release site (~halfway between Moheli and the coast of East Africa) was along one of the previously tracked migration routes for turtles of this population (Goossens, A et al., unpublished data). Upon release, the turtle swam/dived continuously and returned to the nesting beach ~5 days later, when the logger was retrieved and data were downloaded. All experimental procedures during field work in Mayotte were approved by the Comite National pour la Protection de la Nature of the French Ministry in charge of Environmental Affairs and adhered to all institutional guidelines and legal requirements of the country in which the work was carried out. All work in Mayotte was carried out under Kelonia institutional licence (no. 20/DAF/ 06, Mayotte Prefecture). Field work in Moheli was conducted in accordance with institutional and national (French) guidelines and regulations (Permit number 34-100 to Simon Benhamou, approved by the Veterinary Services of the French Ministry of Agriculture) and authorized by the Parc marin de Moheli. DATA ANALYSIS At the foraging site, three of the 13 turtles were equipped for more than a single deployment period (two to three periods; deployment period was constrained by logger memory, which was filled after ~55 days). Since these deployment periods followed the first deployment immediately (filled logger was exchanged with a blank logger), data from multiple deployment periods were merged and treated as a single deployment (see Table 2 for deployment durations). Acceleration sensors malfunctioned during logger deployment in two turtles, while one logger failed completely in another turtle. Hence, at the foraging site, data from 12 turtles were included in the analysis of dive patterns, while data from 10 turtles were included in our analysis of accelerometry-derived energy expenditure (EE). Recorded data from each deployment were first visualized in LOGGER TOOLS (version 3.31 HS; Little Leonardo, Tokyo, Japan) and then analysed in detail as described below. Since logger deployment at the foraging site spanned both the austral summer and the austral winter, we investigated seasonal differences in our analysis (dive patterns and energy expenditure). Similarly, we also explored diel differences, as turtle behaviour, especially at the foraging site but also during the displacement study, differed markedly between day and night. In the Bay of N Gouja (Mayotte), turtles typically forage throughout the day at the shallow seagrass meadows and rest on the inner reef slope at night (Taquet et al. 2006; Ballorain et al. 2010, 2013). This pattern is very consistent and persists throughout the year and was clearly visible in our recorded depth and acceleration traces. To distinguish between periods of activity (day) and resting (night), we defined the beginning of night with the start of the first long resting dive, and the end of night with the end of the last long resting dive, which typically coincided with local times of sunset and sunrise, respectively. On average, turtles started to become active at 5: 13 h in the morning (i.e. ~45 min before sunrise) and became inactive at 18: 09 h in the evening (~sunset). For the displaced turtle, we used the local time for sunrise (6: 30 h) and sunset (18: 00 h) to investigate the diel patterns. Depth data were analysed using IGOR PRO (version 5.0; Wave- Metrics Inc., Portland, OR, USA). In accordance with the resolution of the pressure sensor, all dives with a maximum depth of less than 08 m were excluded from the analysis. For all other dives, we recorded dive duration, maximum depth reached and post-dive surface interval duration. Water temperature (T w ) for each recording was averaged over 5-min periods. From the recorded acceleration data, we calculated a trace of PDBA (using heave and surge axes) for each deployment following Wilson et al. (2006), as described in Enstipp et al. (2011). For each trace, mean PDBA was calculated every 5 min and converted into mass-specific rate of oxygen consumption (s _V O2, ml min 1 kg 083 ) using a calibration equation derived from adult green turtles resting and swimming in an aquarium (Enstipp et al. 2011). The scaling exponent of 083 has commonly been used in the turtle literature (Prange & Jackson 1976; Southwood, Darveau & Jones 2003a), and a similar exponent was also found during the aforementioned calibration study (Enstipp et al. 2011). We acknowledge that logger placement on the displaced turtle differed slightly from turtles at the foraging site and during the calibration study because the usual position (second vertebral scute) was taken by a satellite tag. Clearly, raw acceleration values will vary in relation to the placement of the logger. However, since the calculation of PDBA (or ODBA, if sway is also recorded using a triaxial acceleration logger) only considers the dynamic component of acceleration and removes its static component, this small variation in logger placement should have not affected PDBA values significantly. Halsey, Shepard & Wilson (2011) assessed the applicability of the acceleration technique to estimate energy expenditure for a wide range of animals. They argue that different fixed points on the body should return similar acceleration measures with a minimal variation, as long as logger attachment position is close to the animal s trunk. Furthermore, since PDBA/ODBA is intended to be a measure of acceleration around an animal s centre of mass, it is important that the acceleration logger is placed as close to it as possible (Halsey, Shepard & Wilson 2011). We believe that both positions (2nd and 4th vertebral scute) fulfil equally well this requirement, so that PDBA values calculated for foraging turtles and the displaced turtle are comparable. Rates of oxygen consumption (s _V O2 ) were converted into daily energy expenditure (DEE). Given the herbivorous diet of green turtles foraging at N Gouja (various seagrass species; Ballorain et al. 2010), we assumed a respiratory quotient of 093 (Nagy & Shoemaker 1984) and an energy equivalent of 2077 kj L 1 O 2 (Brody 1945) for these turtles. However, the displaced, breeding turtle was presumably post-absorptive, relying on fat metabolism. We therefore assumed a respiratory quotient of 071 and an energy equivalent of 1964 kj L 1 O 2 (Brody 1945) for this turtle. For all parameters investigated, we first calculated daily means [(i) 24-h periods, (ii) day periods and (iii) night periods], which were used to calculate overall mean values for each turtle. Grand means were calculated from individual turtle means. STATISTICAL ANALYSIS All statistical analyses were conducted using JMP (Pro 10; SAS Institute Inc., Cary, NC, USA). At the foraging site, differences in the daily (24-h) means, diel differences (day vs. night) and seasonal differences (summer vs. winter) of dive duration, surface duration, maximum dive depth, the number of dives, water temperature (T w ), hours of activity, PDBA, oxygen consumption rate (s _V O2 ) and DEE were tested using a linear mixed-effects model (standard least-squares regression fitted by REML). Diel phase/season was included as fixed effect, while turtle ID was included as a random effect. To test for differences in the various parameters between turtles at the foraging site and the displaced turtle during simulated oceanic migration, we conducted separate tests (linear mixed-effects model with standard least-squares regression fitted by REML), including status (foraging vs. migration) as fixed effect (and also diel phase and season, where appropriate), while turtle ID was included as a random effect. Significance for all statistical tests was accepted at P < 005. All mean values are presented with standard deviation (1 SD).

1814 M. R. Enstipp et al. Results FORAGING SITE (MAYOTTE) Diel patterns Turtles divided their days between foraging on the shallow seagrass meadow during daylight hours and resting at greater depth on the inner side of the reef slope at night. Figure 1a shows a typical 24-h period for one turtle ( Autonome ) during the austral winter: as indicated by the depth and PDBA trace, the turtle moved from the reef slope to the adjacent shallower seagrass meadow just before sunrise, foraged throughout the day, often at very shallow depths, and returned to the deeper reef slope just before sunset. Figure 1b illustrates the persistence of this activity pattern for the same turtle over the course of 5 days. Occasionally, turtles also rested for a few hours (1 3 h) at greater depth during the day. This behaviour typically coincided with very low tides and high solar radiation at the foraging site, which increased measured T w considerably. Throughout the year, turtles remained active for 130 10 h and rested for 109 10 h during each 24-h period. When foraging during the day, turtles conducted shallow and short dives (summer: maximum depth: 25 04 m, dive duration: 59 08 min; Table 1). At night, when turtles rested at the inner reef slope, both dive depth (summer: 115 19 m) and dive duration (summer: 436 57 min) were significantly greater than during the day (depth: F = 10730, P < 00001; dive duration: F = 9138, P < 00001; Table 1). In accordance with the observed activity patterns, oxygen consumption rates (s _V O2 ) of turtles during the day were approximately twice the rates at night (F = 11532, P < 00001) and this difference persisted throughout the year (Table 2). Seasonal patterns Turtles foraging at N Gouja were similarly active during both seasons, as indicated by the amount of time turtles were active (as opposed to resting) during each 24-h period (summer: 128 04 h day 1, winter: 131 12 h day 1 ; F = 019, P = 068; Table 2) and by mean daily PDBA, which did not differ between summer and winter (F = 017, P = 069). Mean daily T w during the austral summer was 289 01 C (range: 283 305 C) and was significantly lower during the austral winter (257 04 C, range: 248 273 C; F = 1678, P < 00001; Table 2). The diel differences in dive patterns persisted throughout the year. However, dive durations of both day and night dives were significantly greater during the austral winter, when compared with the austral summer (F = 63, P = 003; Table 1), while dive depth remained unchanged (F = 013, P = 073; Table 1). During the austral summer, mean daily s _V O2 of turtles was 074 006 ml O 2 min 1 kg 083 and was found to be significantly decreased during the austral winter (063 006 ml O 2 min 1 kg 083 ; F = 805, P = 002; Table 2). Accordingly, DEE of turtles during the austral winter was found to be significantly decreased by ~19%, when compared with the summer period (summer: 1117 66 kj day 1 ; winter: 901 111 kj day 1 ; F = 97, P = 002). This decline in DEE from the austral summer to winter occurred gradually (from 1132 36 kj day 1 in April to 905 80 kj day 1 in August) and was paralleled by a decline in mean daily T w (from 289 01 C in April to 253 04 C in August). SIMULATED OCEANIC MIGRATION (MOH EL I ) Over the course of ~35 days, following its release over deep water, the turtle moved continuously towards its nesting beach in Moheli and no rest phases were apparent in the recorded trace (Fig. 2). During this time, the turtle displayed two distinct dive types: during the day most dives were short and shallow, just below the surface (dive duration: 06 05 min; maximum depth: 16 04 m, Table 1), and might be indicative of travelling (Fig. 3a; H- type dives according to Hochscheid 2014). At night (predominantly), these dives were interspersed with deeper and longer pelagic dives (dive duration: 253 68 min; maximum depth: 204 23 m; Fig. 3b; F-type dives; see Hochscheid 2014). However, mean dive duration (07 02 min) and maximum dive depth (15 03 m) during the night were similar to day values (Table 1). After ~35 days, as the turtle neared Moheli, it started to conduct even deeper dives, which might be interpreted as orienting dives, potentially looking for the sea bottom (Fig. 2a). After ~4 days of constant movement, the first rest phase can be detected in form of a U-shaped dive (Fig. 3c; A-type dive; see Hochscheid 2014) and is clearly visible in the PDBA trace (Fig. 2a). Two further U-shaped resting dives occurred after ~45 days, just before the turtle hauled out onto the beach (Fig. 2a). Dive duration and maximum dive depth during these U-shaped dives were 408 46 min and 182 92 m, respectively. Oxygen consumption rates (s _V O2 ) of the displaced turtle during its Fig. 1. Typical recording from one male green turtle (Autonome) for one day (a) and (b) for entire recording duration at the foraging site in Mayotte during the austral winter, showing depth, T w and PDBA. Depth and T w were recorded at 1 Hz, while PDBA was averaged over 5-min periods from the recording of heave and surge at 16 Hz. The black bars on top indicate periods of darkness, while tick marks associated with a date indicate 00 : 00 h for that day. (a) The turtle ascends just before sunrise from its last resting dive, moves to the adjacent seagrass meadow, where it forages throughout the day at shallow depth. Just before sunset, the turtle moves back to the inner side of the reef slope and rests there for the night at greater depth. (b) illustrates the very consistent activity pattern of the turtle, foraging during the day and resting at night. Note the influence of tidal oscillations on the depth trace and the increase in T w, associated with foraging during low tide at very shallow depth. PDBA is elevated throughout the day, when the turtle is active, but approaches zero at night, when the turtle rests. PDBA spikes at night are associated with brief excursions to the surface for gas exchange.

Green turtle energy expenditure 1815 (a) 0 Depth (m) 5 10 15 28 T w ( C) 27 26 0 06 PDBA (g) 0 04 0 02 0 00 04:00 08:00 12:00 16:00 20:00 00:00 04:00 08:00 Time (h) (b) 0 Depth (m) 5 10 15 28 T w ( C) 27 26 0 06 PDBA (g) 0 04 0 02 0 00 04/07/2007 05/07/2007 06/07/2007 07/07/2007 08/07/2007 Date/time

1816 M. R. Enstipp et al. return trip to the nesting beach were significantly greater than rates of turtles at the foraging site (daily s _V O2 : F = 1094, P < 00001) and did not change appreciably with diel phase (Table 2). When compared with the s _V O2 of turtles resting at night in Mayotte at almost identical T w, the s _V O2 of the displaced turtle during the day was 32 times the resting rate (Table 2). Accordingly, mean DEE of the displaced turtle during its return trip was 2327 292 kj d 1 (Fig. 4). Discussion We used the non-invasive accelerometry technique to investigate the EE of adult green turtles during two important phases of their life cycle: at their foraging site and during (simulated) breeding migration. We found that s _V O2 and DEE at the foraging site were low (s _V O2 during the day was 16 and 19 times the respective resting rate at night during austral summer and winter, respectively; Table 2), especially when compared with the DEE of the displaced turtle during simulated oceanic migration. This is consistent with the requirement to cost-efficiently build up substantial energy reserves at the foraging site, which will be required to sustain the energy-demanding breeding migration. While EE during displacement was high (daily s _V O2 was ~3 times the resting rate), as the turtle moved for ~4 days without rest, it remained within the typically observed metabolic scope for sea turtles in water of four times standard metabolic rate (SMR) (Wallace & Jones 2008; Jones & Seminoff 2013), emphasizing the cost efficiency of underwater locomotion in marine turtles. FORAGING SITE (MAYOTTE) At the foraging site in Mayotte, turtles displayed a typical diel foraging/dive pattern that has been observed in numerous studies of green turtles foraging in neritic habitats (e.g. Ogden et al. 1983; Makowski, Seminoff & Salmon 2006; Taquet et al. 2006; Hazel, Lawler & Hamann 2009; Ballorain et al. 2010, 2013; Blumenthal et al. 2010). Turtles were active throughout the day feeding on the seagrass meadow (short and shallow dives) and rested at night on the inner reef slope (long and deep, U-shaped dives). Occasionally feeding was interrupted by a few resting dives in deeper water, typically coinciding with low tide levels that fully or partially exposed the seagrass meadow. How shallow water depth and strong solar radiation affected the temperature experienced by the turtles is evident in the T w trace in Fig. 1. We observed a maximum difference in T w experienced by the turtles over the daily cycle of 24 C and 30 C during the austral summer and winter, respectively. Shuttling to deeper and cooler water under these circumstances has been observed in immature and subadult green turtles foraging in shallow lagoons (Bjorndal 1980; Mendoncßa 1983) and might serve thermoregulatory purposes, preventing turtles from heat stress (Mrosovsky 1980; Spotila & Standora 1985). The observed activity patterns of green turtles at their foraging site are in agreement with a strategy of minimizing EE during the night, as indicated by their low oxygen consumption rates (Table 2). These values are even below the resting rates measured in captive adult green turtles at similar T w during the day (Enstipp et al. 2011), which might be explained by the much longer dive durations and, hence, resting periods in the current study (436 min vs. 92 min and 614 vs. 133 min for current study vs. Enstipp et al. 2011, for the austral summer and winter, respectively; Table 1). During the day, turtles followed a strategy of maximizing energy acquisition, as turtles foraged throughout daylight hours (Taquet et al. 2006; Ballorain et al. 2013), only occasionally interrupting feeding activity. By contrast, juvenile green turtles at their neritic foraging site typically show intermittent feeding behaviour, with peaks occurring during the morning and afternoon, interrupted by resting periods around midday (Bjorndal 1980; Mendoncßa 1983). This pattern of intermittent (rather than continuous) feeding might be explained by the lower digestive capacities of smaller/younger green turtles due to the differences in size and intestinal flora, when compared with larger/older individuals (Bjorndal 1980, 1997). Despite being active throughout daylight hours, oxygen consumption rate (s _V O2 ) during the day was low, similar or even below values measured in captive adult green turtles swimming at low pace (Enstipp et al. 2011) and equalled 16 and 19 times the respective resting rate (at night) during the austral summer and winter, respectively (Table 2). Few studies have investigated the metabolic rate of marine turtles at their foraging grounds. For green turtles, two studies estimated the metabolic rate of juvenile green turtles at their foraging grounds in Australia and Japan (Southwood et al. 2006; Okuyama et al. 2014), while studies for adults are lacking. Southwood et al. (2006) used the DLW method to estimate the field metabolic rates (FMR) of juvenile green turtles foraging in Australia and reported values (142 and 81 kj kg 1 day 1 during summer and winter, respectively) that were 85 to 10 times the metabolic rates of routinely active juvenile green turtles (post-absorptive) of slightly greater mass in captivity (Southwood, Darveau & Jones 2003a). These values are high, especially when considering that the metabolic scope for sea turtles during activity in water is usually up to four times SMR (Wallace & Jones 2008; Jones & Seminoff 2013) and that the metabolic scope for adult green turtles is considerably higher than that observed for juveniles of this species (Williard 2013). Differences in the feeding status of turtles in the studies concerned might be responsible for some of this, while methodological uncertainties with the derivation of metabolic rates from DLW studies might have led to an overestimation of turtle FMRs (Jones et al. 2009; Enstipp et al. 2011). Okuyama et al. (2014) used the accelerometry technique to estimate the metabolic rate of juvenile green turtles at their foraging site in Japan during summer (mean T w :28509 C). Using the calibration equation established for juvenile green turtles by Halsey et al. (2011a),

Green turtle energy expenditure 1817 Table 1. Dive patterns recorded from adult green turtles at the foraging site and during simulated oceanic migration Day Night Turtle Sex CCL (cm) Mb (kg) Dive duration (min) Surface duration (min) Max depth (m) No. of dives Dive duration (min) Surface duration (min) Max depth (m) No. of dives Foraging site Austral summer Soula M 98 1122 56 10 03 01 23 07 1243 306 349 55 19 02 118 10 192 26 Mireille F 103 1340 71 10 02 01 21 06 991 236 465 52 22 03 94 04 141 17 Chabou F 100 1144 49 04 02 01 25 03 1218 162 443 26 21 01 121 05 144 09 Gamee F 98 1084 61 07 03 01 31 09 1015 169 501 130 22 05 143 15 140 40 Leo M 92 970 59 08 04 03 27 01 1023 90 421 45 21 02 100 11 150 16 Grand mean 981 42 1132 134 59 08 a 03 01 25 04 1098 122 a 436 57 a,b 21 01 b 115 19 b 153 22 a,b Austral winter Lea F 104 1302 103 10 03 01 21 07 664 117 807 45 27 03 99 03 83 09 Mumu F 94 960 73 16 02 01 21 03 1030 105 510 91 18 03 77 05 123 39 Kantee F 99 1140 76 07 03 02 36 15 942 129 605 82 21 03 125 14 104 11 Ayoub M 91 868 80 13 03 01 21 03 893 96 476 112 19 05 107 18 145 39 Imas M 96 960 100 18 05 01 29 11 835 158 388 193 15 07 96 68 190 106 Autonome M 104 1090 87 14 03 01 18 02 630 172 855 34 32 03 150 07 85 10 Fantomasq M 103 1144 82 05 02 00 24 04 776 152 658 102 21 03 110 13 105 17 Grand mean 986 52 1066 147 86 12 03 01 24 06 824 145 614 172 b 22 06 b 109 23 b 119 38 b Displaced turtle F 110 1545 06 05 c 01 01 16 04 10592 4372 c 07 02 c 01 01 c 15 03 c 7482 2742 c Data at the foraging site were obtained during the austral summer (April 2007) and the austral winter (July 2007 and August September 2008), while data from simulated oceanic migration were obtained during the austral winter (July 2010). Values shown are means (SD), separated into day and night periods. Durations for day and night periods and Tw are indicated in Table 2. Grand means were established from individual turtle means. CCL, curved carapace length; M b, body mass. Max depth is the maximum depth reached during a dive, while No. of dives indicates the number of dives conducted during the respective period. a Significant difference from the austral winter. b Significant difference from day conditions. c Significant difference from foraging site.

1818 M. R. Enstipp et al. Table 2. Oxygen consumption rates of adult green turtles at the foraging site and during simulated oceanic migration Daily (24 h) Day Night Turtle M b (kg) Rec. duration (days) T w ( C) _VO2 (mlmin 1 ) s _VO2 (mlmin 1 kg 083 ) Duration (h) _VO2 (mlmin 1 ) s _VO2 (mlmin 1 kg 083 ) Duration (h) _VO2 (mlmin 1 ) s _VO2 (mlmin 1 kg 083 ) Foraging site Austral summer Soula 1122 101 289 01 3641 297 072 006 126 04 4424 554 088 011 115 04 2796 195 056 004 Mireille 1340 102 290 01 3989 245 068 004 127 07 4823 262 083 004 113 08 3051 202 052 003 Leo 970 47 288 01 3578 141 080 003 133 08 4398 059 099 001 108 08 2572 260 058 006 Grand mean 1144 186 83 32 289 01 a 3736 222 a 074 006 a 128 04 4548 238 a 090 008 a 112 04 b 2806 240 a,b 055 003 a,b Austral winter Lea 1302 90 254 03 3280 139 058 002 127 11 4114 148 072 003 113 11 2336 077 041 001 Mumu 960 38 253 02 2536 073 057 002 136 11 3054 082 069 002 103 09 1850 072 042 002 Kantee 1140 54 252 02 3005 145 059 003 129 03 3835 272 075 005 111 03 2038 163 040 003 Ayoub 868 41 254 01 2501 120 062 003 127 01 3194 186 079 005 113 01 1726 076 042 002 Imas 960 48 259 01 3185 096 072 002 155 15 3858 248 087 006 85 15 1978 127 045 003 Autonome 1090 52 261 01 3072 120 063 002 118 09 4201 145 086 003 122 09 1978 024 040 000 Fantomasq 1144 52 262 01 3501 082 068 002 127 15 4685 246 092 005 113 15 2187 204 043 004 Grand mean 1066 147 54 17 257 04 3011 372 063 006 131 12 3849 571 080 008 109 12 b 2013 203 b 042 002 b Displaced turtle 1545 47 259 02 8229 1033 c 125 016 c 115 8869 718 c 135 011 c 125 7823 954 c 119 015 c Data at the foraging site were obtained during the austral summer (April 2007) and the austral winter (July 2007 and August September 2008), while data from simulated oceanic migration were obtained during the austral winter (July 2010). Values shown are means (SD), separated into the daily cycle (24 h), day and night periods. Durations for day and night periods are indicated. Grand means were established from individual turtle means. Sex and curved carapace length (CCL) of turtles are given in Table 1. Mb, body mass; Rec. duration, recording duration; Tw, water temperature; _VO2, oxygen consumption rate; s _VO2, mass-specific oxygen consumption rate. a Significant difference from the austral winter. b Significant difference from day conditions. c Significant difference from foraging site.

Green turtle energy expenditure 1819 (a) 0 Depth (m) 10 20 30 40 50 26 5 T w ( C) 26 0 25 5 0 20 PDBA (g) 0 15 0 10 0 05 0 00 19/07/2010 20/07/2010 21/07/2010 22/07/2010 23/07/2010 Date/time (b) 0 Depth (m) 10 20 26 5 T w ( C) 26 0 25 5 0 16 PDBA (g) 0 12 0 08 0 04 0 00 05:00 09:00 13:00 17:00 21:00 01:00 05:00 Time (h) Fig. 2. (a) Entire recording for the displaced green turtle on its return trip to the nesting beach in Moheli, showing depth, T w and PDBA. The black bars on top indicate periods of darkness, while tick marks associated with a date indicate 00:00 h for that day. For the first ~4 days, the turtle moves continuously and does not rest. It conducts mainly short and shallow dives during the day, which, predominately at night, are interspersed with deeper pelagic dives. As it nears Moheli, the turtle conducts deeper orienting dives and, upon reaching the sea bottom, also conducts a few U-shaped resting dives, which are clearly detectable in the PDBA trace (values approaching zero). (b) Enlarged 24-h sequence during which the turtle conducted shallow dives throughout the day, indicative of travelling, while at night shallow dives are regularly interspersed with deeper pelagic dives. Note how PDBA declines during pelagic dives but never approaches zero, indicating that turtles remained active (i.e. stroking) during these dives.

1820 M. R. Enstipp et al. (a) Travelling dives (H-type) 0.0 Depth (m) 0 5 1 0 1 5 PDBA (g) 2 0 0 12 0 08 0 04 0 00. svo 2 = 1 39 ± 0 10 ml O 2 min 1 kg 0 83 12:23 12:28 12:33 12:38 12:43 12:48 Time (b) Pelagic dives (F-type) 0 Depth (m) 5 10 15 PDBA (g) 20 0 16 1 59 ± 0 28 0 12 0 08 0 96 ± 0 04 0 96 ± 0 03 0 04 0 00 04:35 05:05 05:35 06:05 Time (c) U-shaped resting dives (A-type) 0 Depth (m) 5 10 PDBA (g) 15 0 12 0 84 ± 0 23 0 08 0 04 0 39 ± 0 03 0 37 ± 0 00 0 00 20:45 21:15 21:45 22:15 Time Fig. 3. Dive types observed during the return trip of the displaced green turtle to its nesting beach in Moheli, showing depth and PDBA (averaged over 1-min intervals). Numbers in the PDBA traces refer to the associated oxygen consumption rates (means SD), s _V O2,in ml O 2 min 1 kg 083, for the indicated periods. (a) Shallow travelling dives (H-type). (b) Deeper, pelagic dives (F-type) that predominantly occurred at night and interspersed the shallow dive pattern. (c) U-shaped resting dives (A-type) that occurred towards the end of the trip, when the sea bottom was within reach, most likely indicating bottom resting [illustrated by partial dynamic body acceleration (PDBA) values approaching zero and low s _V O2 values]. Dive type letters follow the classification in Hochscheid (2014). they report oxygen consumption rates (s _V O2 )of069, 095 and 102 ml O 2 min 1 kg 1 during resting, foraging and other dives, which equates to 478, 659 and 707 kjkg 1 day 1 for the respective dive types (using a conversion factor of 2077 kj L 1 O 2 ). These values are ~half or less the values found in the DLW study, despite higher T w values in the study by Okuyama et al. (2014). Nevertheless, oxygen consumption rates (s _V O2 ) and DEE of adult green turtles foraging in Mayotte during both the austral summer and winter were considerably lower than the values reported for juveniles. Some of these differences may be explained by the greater amount of metabolically inert

Green turtle energy expenditure 1821 DEE (kj d 1 ) 2500 2000 1500 1000 500 0 Foraging Summer * Day Night Winter * Day Night Migrating green fat stored by adults when compared with juveniles/ immatures, resulting in lower mass-specific metabolic rates of adult turtles (Kwan 1994; Penick et al. 1996). In the current study, turtles remained active year-round at the foraging site and activity patterns did not differ between seasons (time spent active per 24-h cycle, daily PDBA; Table 2), despite the significant changes in T w. However, dive durations during both day and night were significantly greater during the austral winter when compared with the austral summer, while bathymetrically constrained dive depth did not change (Table 1). Oxygen consumption rate of turtles and their DEE declined gradually from austral summer to austral winter, in parallel with T w, so that mean daily s _V O2 of turtles was ~19% lower during the austral winter, when compared with the austral summer. Year-round activity and maintenance of foraging have been documented for both immature and adult green turtles at several tropical and subtropical foraging grounds (e.g. Bjorndal 1980; Mendoncßa 1983; Southwood et al. 2003b, 2006; Taquet et al. 2006; Ballorain et al. 2013). Southwood et al. (2003b) reported a significant seasonal shift in dive patterns (greatly increased dive durations during winter) and activity levels of juvenile green turtles at their foraging ground in Australia that were accompanied by a considerable change in FMRs (43% lower during winter when compared with summer; Southwood et al. 2006). This led the authors to conclude that multiple biotic and abiotic factors act in conjunction with temperature to elicit seasonal changes in the energetics of sea turtles and that temperature alone cannot fully explain seasonal shifts in metabolism (Southwood et al. 2006; Williard 2013). However, in the current study, turtles at their foraging site in Mayotte were similarly active, regardless of season. Accordingly, the moderate 19% reduction in oxygen Day Night Fig. 4. Daily energy expenditures (DEE) (kj day 1 ) of adult green turtles foraging at the seagrass meadow in the Bay of N Gouja (Mayotte) during the austral summer (n = 3 turtles, T w = 289 01 C) and winter (n = 7 turtles, T w = 257 04 C) and during simulated oceanic migration after displacement from the nesting beach in Moheli (n = 1 turtle, T w = 259 02 C). Energy expenditure (EE) contributions from day and night periods are indicated by the horizontal line in each bar. Values are means SD *Significantly different from displaced turtle. Significant difference between seasons. consumption rates (s _V O2 ) and DEE observed during the winter, when compared with the summer, is most likely largely explained by the lower water temperatures during winter, reducing turtle metabolism (Enstipp et al. 2011). SIMULATED OCEANIC MIGRATION (MOH EL I ) We used an experimental approach (translocation) to simulate the oceanic migration of a green turtle in order to investigate the activity patterns and associated energetic costs during breeding migration. Experimental displacement during the internesting period might have altered the turtle s natural behaviour due to the stress associated with capture/handling and the presumably strong drive to return to its nesting beach. Despite this, the dive patterns observed in the displaced turtle during simulated oceanic migration to its nesting beach in Moheli are consistent with previously reported patterns for green turtles during (natural) oceanic migration. For example, Hays et al. (1999) satellite-tracked green turtles during their postbreeding migration from Ascension Island to the coast of Brazil and found distinct differences in dive durations, with shorter submergences during the day and longer submergences at night. When attaching TDRs to green turtles displaced from Ascension Island, Hays et al. (2001) found that during their return trip to the island, turtles conducted mainly short (2- to 4-min) and shallow (09- to 15-m) dives, which were interspersed with long (20- to 30-min) and deep (10- to 20-m) dives that predominately occurred at night. Similar patterns were also observed in other studies concerning different populations (Godley et al. 2002 for the Mediterranean; Hatase et al. 2006 for Japan; Rice & Balazs 2008 for Hawaii). Short and shallow dives during the day are typically interpreted as travelling dives (Hays et al. 2001; Godley et al. 2002; Rice & Balazs 2008), and this fits well with the observation that travel speed in migrating turtles is faster during the day than at night (Luschi et al. 1996, 1998). Travelling between 1 2 m depth during the day may allow turtles to make use of the sun for navigational purposes (Rice & Balazs 2008), while also avoiding wave drag associated with swimming at or near the surface (Hertel 1966; Fish 2000) that would increase swimming costs considerably. A number of functions have been suggested for the deeper nocturnal dives observed in migrating turtles, including reduced activity or even rest (Hays et al. 1999, 2001), reduced predator exposure (Hays et al. 2001) and pelagic foraging on upward-migrating macroplankton (Hatase et al. 2006; Rice & Balazs 2008; Cheng, Bentivegna & Hochscheid 2013). However, our recording from the displaced turtle shows that while PDBA and associated s _V O2 are found to be significantly reduced during these predominately nocturnal pelagic dives, when compared with shallow travelling dives, the values are still considerably higher than during the observed U-shaped dives, when the turtle did not move, most likely resting at the sea bottom (Fig. 3). The reduction in PDBA and s _V O2 during these pelagic dives at