Study site

All birds used in the current study were mature adults from the breeding colonies at Wedge Light and Pope's Eye, in the Pope's Eye Marine Reserve (38°16′35″S, 144°41′54″E), located off Queenscliff near the entrance to Port Phillip Bay, Australia. We assumed that there were no sex differences in physiology in this monomorphic species and therefore sex was not considered as a factor in any part of this study. All experiments were carried out with the approval of the La Trobe University animal ethics committee (AEC 04/37L) and appropriate wildlife permits were issued by Parks Victoria.

Laboratory measurements

Nine gannets, Morus serrator Gray 1843, (mean mass ± s.e.m., 2.31±0.02 kg) were brought into the laboratory in January and February 2006. All birds had either failed in their breeding attempt that year or had already fledged a chick. Birds were collected in groups of two or three and transported in ventilated pet carriers by boat to Queenscliff and onwards to La Trobe University in an air-conditioned car (~2 h total journey time). Once at La Trobe, the birds were kept in an outside aviary (7.5×5 m) which had a sand base, perches, protection from rain and sun, and a rectangular pool (2×1.5 m) for bathing and drinking. Birds were kept captive for no more than 7 days, after which point they were returned to their colonies. During captivity, the birds were hand-fed 300 g of pilchards twice daily, supplemented with a daily multi-vitamin tablet. Birds were used once each in two sets of experiments: to examine metabolic changes associated with diel rhythms and exercise, and define a calibration relationship between heart rate (f_H) and rate of oxygen consumption () (present study), and to examine heat stress and thermoregulation at high temperatures (J.A.G., E.J.A.-S. and P.B.F., unpublished data).

We measured and the rate of carbon dioxide production () using an open-circuit respirometry system. Animals were placed in a clear Perspex respirometer (45×70×62 cm), which was located in a constant temperature (CT) room within which the ambient temperature could be controlled. Air was drawn from the respirometer using a compressor pump (Millipore, Billerica, MA, USA) at either 12 or 24 l min⁻¹ and flow rate was monitored using an electronic mass-flow meter (Sierra Instruments Inc., Monterey, CA, USA). A subsample of air was drawn from this main flow using a small aquarium air pump (Rolf C. Hagen, Castleford, West Yorkshire, UK) and passed through a drying column (Drierite, Hammond, Xenia, OH, USA), then analysed for the fractional content of O₂ and CO₂ by a combined gas analyser (ADInstruments, Sydney, Australia). A solenoid valve (Bürkert, Ingelfingen, Germany) was used to switch between the respirometer and ambient air samples. Heart rate was monitored using a customised heart rate transmitter and a receiver unit (Polar Electro Oy, Kempele, Finland). The transmitter was attached dorsally to the feathers using adhesive tape (Tesa, Hamburg, Germany) and custom-made brass electrodes were inserted under the skin at the bottom of the neck and base of the spine. A thermocouple was inserted ~8 cm into the rectum of the birds to monitor body temperature (T_b). The heart rate receiver output, outputs from the gas analyser, flow meter and thermocouples (ambient temperature, T_a, was also recorded continuously) were collected at 100 Hz (ADInstruments) and displayed on a computer using Chart software (ADInstruments). was determined from the rate of airflow from the respirometer and the difference in the fractional concentration of O₂ between ambient and outflowing air. Instantaneous corrections of the gas concentrations were calculated dry at standard temperature (273 K) and pressure (101.3 kPa) using a previously described method (Frappell et al., 1989), assuming a first-order linear system (chamber volume, 195 l; flow, 12 or 24 l min⁻¹; time constant τ, 12.8 or 7.8, determined from a semi-logarithmic plot of concentration against time following a perturbation; r²=0.97, 0.99, respectively). was calculated with consideration of respiratory quotient (RQ)-related errors (Frappell et al., 1992). Whole-system accuracy was determined to be within 6% by bleeding CO₂ into the respirometer at known flow-rate and back-calculating this rate.

Prior to experiments, food (not water) was withheld for 12 h, sufficient to ensure that the birds were post-absorptive (Laugksch and Duffy, 1986). Birds were equipped with the heart rate transmitter and thermocouple and introduced into the respirometer at ~11:00 h with the CT room set to 23°C. The birds were allowed to settle for ~1 h until and f_H had stabilised. They were then used in the thermoregulation experiment for ~3 h (J.A.G., E.J.A.-S. and P.B.F., unpublished data). After being allowed to settle again for 1–2 h, they were then walked at a series of speeds from 0.1 to 1.4 km h⁻¹. Birds walked for 3–15 min at each speed and had frequent rest periods during which f_H declined to resting levels. When compared with other species where a similar approach has been used (e.g. Green et al., 2009b), the gannets were relatively reluctant to walk on the treadmill and not all birds would walk at the highest speeds. One bird would not walk at all, and data from this animal were not included in analyses. However, instantaneous corrections allowed and to be calculated over relatively short exercise periods of 3 min, which were interspersed with rest periods where f_H and returned to pre-exercise levels. This entire ‘activity’ protocol therefore took less than 2 h in the late afternoon in each case. After exercise had finished, birds were allowed to rest in the respirometer for ~36 h to record values during ‘inactivity’. During this time the birds were free to rest, preen, sleep or investigate the respirometer at will. The CT room was maintained at 22°C in a photoperiod of 14.5 h L:9.5 h D (lights on 05:00 h–19:30 h) to match conditions in the colony at this time of year. Throughout this period f_H and were monitored continuously, with the solenoid automated to switch the gas analyser from excurrent to incurrent (baseline) measurements for 5 min every 2 h. Observations of the bird could be made without disturbance via a closed circuit camera system.

Laboratory data analysis

Data from the laboratory were initially processed using Chart software and further analysed using Excel (Microsoft, Redmond, WA, USA) and Minitab (Minitab Inc., State College, PA, USA). There was no relationship between body mass and resting in our data set. As a result, mass corrections were not applied to any data in the present article and all data are presented as ‘whole-animal’ . Similarly, there was no effect of body mass on either the slope or the intercept of individual relationships between f_H and . We assume therefore that body mass does not have any influence on the relationship between f_H and (Halsey et al., 2007). Should larger changes in body mass occur during the annual cycle, outside the range measured during calibration, leading to changes in resting , we assume that these would be accompanied by other physiological adjustments, causing appropriate changes in f_H in line with the relationship between f_H and . This approach is in line with similar long-term energetic studies of seabirds (Guillemette and Butler, 2012) and is consistent with studies examining year-round changes in the relationship between f_H and in species with highly variable body mass (Green et al., 2005b; Portugal et al., 2009). A full discussion of this issue is presented in a recent review (Green, 2011).

Mean values of f_H, and T_b were calculated every 3 min throughout the experiments. From these values we calculated grand means for both light and dark values for the 24 h period commencing at midnight at the end of the first day. To further examine the diel rhythm, mean values of each quantity were also calculated for each of these 24 h. Resting f_H and were defined as the lowest of these hourly values for each animal, and resting was assumed to be equivalent to BMR.

When constructing a calibration relationship between f_H and , it is essential that, where possible, the collection and analysis of data are used to create a relationship that is analogous to conditions encountered by free-ranging experimental animals (Green, 2011). In the present study, while the birds were free to move in the respirometer during the inactivity period, their large size meant that movement was constrained when compared with other studies that have used these data alone to generate calibration data (e.g. Steiger et al., 2009). As a result, the amount of data recorded while the animals were inactive constituted 97% of the calibration data recorded, which is clearly unrepresentative of the behaviour of free-ranging Australasian gannets, which spend over 40% of their time engaged in foraging activity (Bunce, 2001). To account for this, we could have selected data from the inactive set to combine with data while the birds were active on the treadmill. However, rather than discard valuable data, we instead combined the active and inactive data sets and calculated mean f_H and in heart rate bins of 10 beats min⁻¹ for each bird (Storch et al., 1999). Relationships between f_H and were constructed using general linear model (GLM) with f_H as a covariate and gannet identity as a random factor. From initial visual inspection it was clear that the data would not be best described by a linear relationship; thus, data were log_e-transformed. Within each bird, some bins had more data points than others (range 1–310) so this variability was accounted for by weighting data in the GLM by the square root of this value. This entire approach allowed us to adopt a ‘one-model’ method so that the behaviour of animals in the field would not need to be known (Green et al., 2009b; White et al., 2011). This was particularly important in the present study for two reasons. Firstly, it meant that it was not necessary to define an unambiguous continuous time-budget for the gannets as this has previously been shown to be impossible with this data set (see Green et al., 2009c). Secondly, it allowed us to account for the fact that calibration data could not be obtained from flying or swimming birds. We assumed the curvilinear relationship generated by log_e-transforming the data was analogous to the relationships previously obtained from walking and swimming great cormorants (Phalacrocorax carbo) (White et al., 2011) and walking and flying geese (Ward et al., 2002). To account for the possibility that flying gannets had a drastically different relationship between f_H and we also conducted a sensitivity analysis (see Appendix). This allowed us to look at how different trajectories of the relationship that could conceivably occur during flight might affect our findings.

Field measurements

As described in more detail previously (Green et al., 2009c), six mature breeding adult gannets (mean mass, 2.58±0.05 kg) were surgically implanted with a custom-built data logger (DL), which recorded f_H, depth and abdominal temperature (T_ab; not analysed in the present study). The gannets were selected randomly in September of the 2004–2005 breeding period while incubating their eggs, and all six apparently bred normally (Green et al., 2009c), suggesting that, as in previous studies, the birds were not negatively affected by the presence of the DL (White et al., 2013). We have previously used this data set to describe activity-specific f_H and diving behaviour and physiology of this species. In the current study we focused on larger scale changes in behaviour and energy expenditure. Five of the six data loggers had recordings of f_H for the full 5 month breeding season (mid-September to mid-February), whereas one unit only recorded until mid-October. Regular observations of the colony allowed us to establish laying dates, hatching dates and fledging dates for each breeding attempt by each study animal. Each day of the deployment was therefore categorized as being in one of the following phases: pre-breeding, incubation, chick rearing, winter or failed.

Field data analysis

Periods where the gannets were in flight were identified from data on heart rate for all six gannets, following the procedure outlined previously (Green et al., 2009c). In brief, 5 min running means of f_H were calculated for each second of each day. Flight was considered to have occurred when this 5 min mean f_H was greater than a ‘flight threshold’ value for at least 20 s. Flight f_H was then calculated as the mean f_H during the >20 s period when the running mean was greater than the flight threshold. To select the flight threshold value, a range of threshold f_H values between 160 and 360 beats min⁻¹ was tested for each individual bird. We then plotted daily flight time as a function of threshold f_H. Daily flight time decreased with increasing threshold, but in each case there was a point of inflection where this decrease decelerated, indicating that the appropriate threshold (where f_H rapidly increased as a result of flight) had been identified. For example, for bird 112, below 240 beats min⁻¹, a 20 beats min⁻¹ change in threshold f_H resulted in a 2 h change in total daily flight time, whereas the same change in threshold above 240 beats min⁻¹ resulted in a 0.5 h change in total daily flight time. The flight threshold was either 220 beats min⁻¹ (birds 110, 135, 535) or 240 beats min⁻¹ (birds 112, 244, 292). The time spent in flight was calculated for each gannet for each day. The mean of these individual values gave the daily time in flight (DTF) for each day of the breeding season. Diving data were available from four individuals (birds 112, 135, 292, 535) as previously described (Green et al., 2009c). In the current study we calculated the total time submerged and the number of dives for each gannet for each day. The mean of these individual values gave the daily time submerged (DTS) and daily number of dives (DND) for each day of the breeding season.

To estimate metabolic rate, mean f_H was calculated every 3 min, to match the recording interval in the laboratory calibration procedure, for each animal in the field. Estimates of for these 3 min periods were then made using the calibration relationship. The standard error of the estimate (s.e.e.) for these estimates was calculated using a procedure outlined previously (Green et al., 2001), which accounts for all of the error inherent in the calibration process. For some analyses, we converted to metabolic rate using a factor of 18.7 kJ l⁻¹ O₂, assuming that the gannets had a diet composed predominantly of fish of the same nutritional composition as sardines (Frankel and Smith, 1998; Green et al., 2006). The timing of sunrise and sunset at Pope's Eye was calculated using data from Geoscience Australia (http://www.ga.gov.au/geodesy/astro/sunrise.jsp). Mean f_H and hence estimated and FMR were calculated for each day of the summer breeding period, the entire summer breeding period and each phase of the breeding season. In the case of the different phases there were only sufficient data to evaluate the two main phases, incubation and chick rearing. Estimates of or metabolic rate cannot be treated as parametric data as this would ignore the residual error associated with the calibration process (Green, 2011). Thus estimates of or metabolic rate (±s.e.e.) were compared using the proximate normal test, paired where appropriate.

As well as calculating FMR for each day, we also calculated MMR for each day of the breeding season (see Guillemette and Butler, 2012; Guillemette et al., 2007). To do this we derived a 15 min running mean of f_H and used the lowest value of this quantity each day to estimate and metabolic rate. NMR (total metabolic costs above resting) was calculated by subtracting MMR from FMR for each day (e.g. Guillemette et al., 2007; Ricklefs et al., 1996; White et al., 2011). In doing so we assumed that MMR and NMR are additive components of FMR, because of strong support for a positive Pearson's correlation in our data set between MMR and FMR (P<0.001) and the lack of support for a correlation between MMR and NMR (P=0.06). This agrees with findings from the great cormorant (White et al., 2011) and supports Ricklefs' ‘partitioned pathways’ model of energy metabolism (Ricklefs et al., 1996).

Changes in metabolic measurements (FMR, MMR and NMR) and behavioural measurements (DTF, DTS and DND) over time during the breeding season were checked for normality and analysed using regression analysis, weighted by the inverse square root of s.e.e. or s.e.m., where days since 1 September 2004 was the independent variable and measurement (FMR, MMR, NMR, DTF, DTS and DND) was the dependent variable. Relationships between date and FMR and MMR were compared using GLM with measurement (FMR or MMR) as a fixed factor and days since 1 September 2004 as a covariate. For all these analyses, mean FMR, MMR, NMR, DTF, DTS and DND was calculated for 14 day intervals to control for the effects of serial autocorrelation in time-series data. The value of 14 days was determined by correlating residual values for successive days until the correlation was no longer significant, which occurred at an interval of 14 days. In all statistical testing we followed the advice of Sterne and Smith (Sterne and Smith, 2001) in the interpretation of P-values.