Plunge-diving into dynamic body acceleration and energy expenditure in the Peruvian booby

Energy balance in wild animals determines important outcomes in their life history, such as migration patterns (Guillemette et al., 2012), breeding strategies (Dunn et al., 2018; Viera et al., 2011), growth rates (Navarro et al., 2015) and ultimately survival (Daunt et al., 2007; Elliott et al., 2014). These linkages occur because energy balance affects how an animal decides which activities to perform, such as foraging, commuting or resting (Nagy, 1987; Nagy et al., 1999). These different activities have different energetics costs and those costs will determine how an individual allocates resources (Speakman, 1997). Flapping flight is energetically expensive compared with other forms of locomotion, and this is especially consequential for seabirds that often need to travel great distances to find food at great energetic cost in very patchy environments (Ballance et al., 1997; Weimerskirch, 2007). Energy expenditure becomes especially important during high demand periods, such as breeding at a central place, or during food shortages associated with environmental change (Ashmole, 1963; Elliott et al., 2014, 2009; Tremblay et al., 2022; Welcker et al., 2010).

The estimation of energy expenditure in the wild (‘field metabolic rate’) has been a difficult task in ecological research (Elliott, 2016; Speakman, 1997). Several methods have been used for this purpose such as mass loss, heart rate, accelerometry, respirometry in semi-wild conditions, thermal imagery and doubly labelled water (DLW), although they all pose certain limitations or biases (Elliott, 2016; Green, 2011; Speakman, 1997). DLW is widely used to provide accurate estimates of field metabolic rate in animals over 24–48 h (Butler et al., 2004; Shaffer, 2011; Speakman, 1997). DLW enables calculation of activity-specific field metabolic rate using a time budget approach, which can then be applied over shorter or longer time scales (Wilson and Culik, 1993). The development of small accelerometers has facilitated more accurate estimates of activity budgets improving the time budget approach (Elliott et al., 2013; Hicks et al., 2017).

Dynamic body acceleration (DBA) is the most commonly used metric in accelerometry to estimate energy expenditure directly as opposed to via time budgets (Wilson et al., 2006). DBA is calculated using tri-axial acceleration (heave, surge and sway) allowing for the measurement of fine-scale activity-specific energy costs (Chivers et al., 2015; Kokubun et al., 2015; Stothart et al., 2016). DBA is the area under the accelerometer–time curve after removing the static component associated with posture (Wilson et al., 2006). Specifically, work is equal to the integral of force (F) over distance (x) as given by ∫Fdx=mv∫adt=mvDBA, where m is mass, and therefore mass-specific energy expenditure at a constant speed (v) is proportional to DBA, provided all work is in the direction of travel (Gómez Laich et al., 2009; Stothart et al., 2016). Because locomotion makes up a substantial proportion of an animal's energy budget, DBA and the rate of oxygen consumption (V̇_O2) are often highly correlated, but the strength of the correlation (r) can vary from 0.6 to 0.99 (Elliott et al., 2013; Gómez Laich et al., 2009; Stothart et al., 2016). All validations of the relationship between daily energy expenditure (DEE) calculated from DLW and DBA (DEE_DLW–DBA) in birds to date have occurred below their thermoneutral zone, often in polar or subpolar regions, and the variation in r may be due to thermoregulation decoupling the DEE_DLW–DBA relationship (Elliott et al., 2013; Stothart et al., 2016; Hicks et al., 2017; Sutton et al., 2021). That is, individuals foraging in colder water may have additional energetic costs not accounted for via movement and DBA, potentially confounding the relationship. There is therefore a need to examine the DEE_DLW–DBA relationship in tropical seabirds, foraging in environments within their thermoneutral zone.

Energy balance can be an important metric for understanding the impact of extraction pressure on wildlife (Masello et al., 2021; Yin et al., 2024). The Peruvian Humboldt Current system (PHCS) is one the most productive marine upwelling systems on the planet, once supporting 10 million tons of seabird guano prior to the collapse of the anchovy fishery and subsequent major drop in the seabird population of the system in the 1950s (Barbraud et al., 2018; Chavez et al., 2008; Jahncke et al., 2004). There is a need to understand energy limitations for these guano birds to fully understand the factors hampering their recovery (Chavez et al., 2008; Jahncke et al., 2004; Weimerskirch et al., 2012; Zavalaga et al., 2010). Among the guano birds, Peruvian boobies (Sula variegata), are interesting energetically because they forage exclusively during the day and are plunge-divers with high aspect ratios that do not make deep dives or significantly use their wings to propel themselves underwater (Brewer and Hertel, 2007; Gaston, 2004; Nelson, 1978; Ropert-Coudert et al., 2009, 2004; Van Oordt et al., 2018). Peruvian boobies exclusively plunge-dive, potentially reducing both flying and diving energy expenditure at the cost of reduced mobility underwater (Nelson, 1978), contrasting with gannets, which plunge-dive and also flap their wings under the water (Ropert-Coudert et al., 2009; Sutton et al., 2023). Nonetheless, DEE in gannets is higher than in many other seabird despite low flight costs (Green et al., 2009), possibly because plunging is an energetically demanding activity, especially when flapping underwater, or because they nest in dense aggregations with many aggressive interactions (Adams et al., 1991; Côté, 2000; Ropert-Coudert et al., 2009; Shaffer, 2011; Siegel-Causey and Hunt, 1981). There is a need to understand the relative costs of flying and plunge-diving in boobies to better understand the relative role of anchovy density, distance to anchovy schools or depth of anchovies in limiting the net energy gain of boobies and thus limiting their reproductive success and population growth.

To examine the DEE_DLW–DBA relationship in a warm water, tropical species, we developed biologger validations for plunge-diving Peruvian boobies for both time budget and DBA approaches using accelerometers. Specifically, we measured energy expenditure in Peruvian boobies using DLW and accelerometry. We used the accelerometers, time–depth recorders (TDR) and GPS to classify activities into flying, foraging, at the colony and resting (away from the colony). We predicted that: (1) total daily DBA would highly correlate with total DLW-estimated mass-specific DEE given the lack of a confounding effect of temperature, and (2) DBA would better predict DLW-estimated mass-specific DEE compared with time budget approaches because DBA incorporates variation in work within activities with variable activity rates, such as plunge-diving. To understand how DEE is influenced by activity budgets, we also report time in each activity and activity-specific metabolic rates.

We performed our study between 11 and 17 November 2019 at Guañape Norte Island (08°32′41″S, 078°57′49″W) within the Reserva Nacional Sistema de Islas Islotes y Puntas Guaneras, Peru, when sea surface temperature was roughly 20°C. Thirty-one chick-rearing Peruvian boobies (hereafter boobies), Sula variegata (Tschudi 1843), were captured on their nests by approaching them during the early hours of the day and using a noose pole (>3 m long). We equipped 21 of these birds with biologgers and handling time was less than 5 min, both at deployment and at recapture. Sex was determined in the field based on size and vocalizations (Zavalaga et al., 2009). All boobies were wrapped tightly in a fabric bag and measured on an electronic scale (±1 g). We attached Technosmart (Rome, Italy) axy-trek accelerometers (18 g; sampling rate: 1 GPS fix per minute, 1 Hz for pressure; 50 Hz for triaxial acceleration) to the four central tail rectrices using two zip-ties, super glue and Tesa tape. Devices, including attachment materials, weighed 30 g, ranging from 1.6% to 2.9% of the body mass of our birds, which we assumed produced minimal effects on the birds (Bodey et al., 2018; Geen et al., 2019). No abandonment was detected as a result of biologging. We used tail-mounted devices to reduce the loss of devices due to the bird's plunging behaviour, as devices detach more easily from the back (Vandenabeele et al., 2014).

This study was approved by the McGill University Animal Care Committee.

Using the same birds for equilibrium and final blood samples can alter measurements of energy expenditure because of altered behaviours due to handling stress (Schultner et al., 2010). We used the single-sample method to minimize stress of the deployed birds (handling time <5 min for non-equilibrium birds), as described by Stothart et al. (2016) and well documented by Speakman (1997). Briefly, we divided our sample of 31 birds in two sets: one set for estimation of equilibrium of isotopic concentrations, and a second set on which we deployed accelerometers (as described above) and on which we calculated activity budgets.

The first set of birds consisted of 10 birds used to determine the relationship between isotopic dilution and body mass known as the equilibrium rate. We extracted blood in three capillary tubes (∼150 µl) from the brachial vein of all birds as soon as they were captured for isotopic background value readings. We injected 1 ml of DLW (50% H₂¹⁸O and 50% D₂O; see Elliott et al., 2013, for details) in the pectoralis muscle and kept the birds in a dark box for 1 h before collecting three new capillary tubes (∼150 µl) of blood from the other wing's brachial vein. All capillary tubes obtained in the field were flame sealed for later analysis. We obtained a dilution relationship equation per isotope (deuterium and oxygen-18) with high correlation values (Pearson's r>0.7; Fig. S1), so we were confident that using the single-sample technique for the second set of birds would provide accurate estimates of energy expenditure.

The second set of birds consisted of 21 individuals that were captured and injected with 1 ml of DLW (same mixture as described above) and released with GPS-accelerometers within 5 min of injection. All birds were recaptured ∼48 h after deployment (range 46.1–61.3 h, s.d.=3.7), except for one bird, which was recaptured ∼100 h after and was not included in the energy expenditure validation analysis because the biologger turned off before recapture. Upon recapture, we obtained a blood sample from the brachial vein in two to three capillary tubes (∼150 µl). Biologging devices were carefully removed to avoid the loss of feathers.

All blood samples were distilled in the lab through evaporation after 24 h on a hotplate while in flame-sealed glass Pasteur pipettes. DLW analysis was performed using a Los Gatos Liquid Water Isotopic EP Benchtop Analyzer (model GLA430 LWIA-912, Los Gatos Research, San Jose, CA, USA; hereafter LWIA). Samples were run later in batches of corresponding enrichment using DLW standards for each batch (high enrichment for initial samples and very low enrichment for background value samples, both from the first set of birds, and low enrichment for final samples from the second set of birds). Each run consisted of five preparation injections and five measured injections, which were averaged to obtain the final value for the H²/H¹ and O¹⁸/O¹⁶ ratios for all posterior calculations. We used a low and a high standard at the beginning and end of the run, as well as one of each standard between every 2–4 samples to correct for any residual effect during the run. In the case of some unusual readings, such as low density (due to bubbles in the vials) or pressure errors, samples were rerun or redistilled when needed.

All calculations followed Speakman (1997) as described in the Appendix.

We used Hidden Markov Models (HMMs) with the momentuHMM package (McClintock and Michelot, 2018; Patterson et al., 2019) to classify four behavioural states of boobies: colony, commuting (exclusively flying), foraging (plunging, underwater time, lift-off; may include short-distance flying before and after the plunge), and resting (away from the colony) (see Fig. 1 for example behavioural data and a mapped track). Because this method requires a consistent sampling frequency and step length (distance travelled between GPS locations), we interpolated all tracks to 1 min sampling rates using the crawlWrap function from momentuHMM (to account for potential gaps in GPS locations). We estimated behavioural states using four data streams (presence at the colony, wingbeat frequency, diving, step length) according to the parameters show in Table 1. Briefly, we calculated metrics for all data streams at the original accelerometer rate (50 Hz), interpolating non-accelerometer data when needed (GPS locations and depth). Then we obtained a mean value by extracting the 50th quantile per minute, to thin the data for the HMM models. ‘Colony’ state was based on the distance from the colony (distances ≤0.5 km from the colony=1, distances >1 km from the colony=0); wingbeat frequency, calculated as the peak frequency of the z-axis over a 30 s window using a Fast Fourier Transform; ‘diving’ as the proportion of time underwater during a 1 min window according to the depth sensor (e.g. not diving during 1 min window=0, or diving for 30 s=0.5). Probability distributions for each data stream and starting values for each behavioural state used in the HMM are provided in Table 1.

All statistical analyses were performed using R (http://www.R-project.org/). We first tested differences between sexes in total mass-specific DEE_DLW using linear models, as boobies are sexually dimorphic in size (body mass). No deviations from normality in the model were significant. Although there were significant differences between sexes in mass-specific DEE in boobies, we grouped all individuals after doing this comparison because of small sample sizes per sex, considering that males and females represented a continuum of body mass along the x-axis. Sex-specific models are presented in Table S1.

We then compared all 10 time budget and DBA models and two null models (an intercept model and a total DBA model) using the corrected Akaike's information criterion (AICc) approach (see Table 3). The AICc ranks models, penalizing them if they have an increased number of parameters without an improvement in fit, and it applies a sample size correction which additionally increases penalization. The best models were then plotted with marginal effects using the ggpredict() function from the ggeffects package (https://CRAN.R-project.org/package=ggeffects) with 95% confidence intervals.

To additionally assess the predictive capacity (Stothart et al., 2016; Sutton et al., 2023) of the best ranked models by the AICc, we investigated the correlation strength between predicted values and the actual DEE values using linear regression and reporting the correlation coefficients (r). We also present the mass-specific DEE (hereafter _MSDEE_DLW) estimation values for each activity for both approaches (time budgets and DBA) to demonstrate activity-specific metabolic rates and plot flying costs using DBA activity-specific estimates in comparison to other flying–gliding birds. As our DBA metrics were normalized to a daily basis, we avoided Halsey's time trap (Halsey, 2017). We compared time budgets between sexes using linear models (ANOVA), because boobies are highly dimorphic and our DEE estimates are mass specific, and DBA relationships with time with linear regressions (to further determine whether relationships were driven by time or DBA; Halsey, 2017). For comparative purposes we also calculated BMR using the updated allometric formula presented in McKechnie and Wolf (2004) and also compared our results with those obtained using the Seabird FMR Calculator (Dunn et al., 2018).

_MSDEE_DLW in Peruvian boobies averaged 1.12±0.29 kJ day⁻¹ g⁻¹ (mean±s.d.), and was higher in males (t_1,18=4.21, P<0.001) (Table 2).

Three models predicted _MSDEE_DLW with ΔAICc <2. The best model to predict _MSDEE_DLW was daily DBA (Table 3, r=0.55; Fig. S2, Table S2). The other two models had only two predictor variables: (1) DBA at colony with resting away from the colony and DBA commuting with foraging, and (2) DBA at colony and DBA away from colony (commuting, foraging and resting) (Table 3; Fig. S2, Table S2). Both two-parameter models had strong predictive capacity (r=0.61, P≈0.004) (Fig. 2). Also, activity-specific daily DBA while resting was lower compared with that of all other activities (Table S3). Total DBA at the colony was weakly correlated with time spent at the colony (r=0.46), which is the opposite to what we would expect due to the ‘Halsey trap’. In contrast, time spent away from the colony and DBA away from the colony were strongly correlated (r=0.83), suggesting that DEE in the colony is independent of time spent, and associated with specific behaviours (e.g. territoriality, thermoregulation, etc.).

Peruvian boobies averaged more than 19 h of the day in the colony and ∼3.8 h in commuting flight, with a maximum of 6 h spent commuting among the sampled birds (Fig. 3; Table S3). In both estimation approaches (time budgets and DBA), total energy spent at the colony was highest, followed by commuting and finally resting – but only because much more total time was spent at the colony (Table S3). On an hourly basis, energy costs were 2.7 times higher in commuting flight than at the colony, and 4.6 time higher when foraging (Table 4). Conversely, costs of resting away from the colony were 2.6 times higher than at the colony. Additionally, hourly DBA at the colony during the day was significantly higher than at night (F_1,40=45.88, P<0.0001).

DBA predicted DEE in a plunge-diving, flap-gliding seabird, the Peruvian booby. Interestingly, whereas most other studies have found that DBA in flight and diving partitioned separately from other activities (Ste-Marie et al., 2022; Stothart et al., 2016), in our study, the most parsimonious model predicted DEE from DBA alone, or perhaps with colony partitioned separately (Table 3). This means that DEE can be estimated in wild plunge-diving boobies from a single metric, DBA, across many different locomotory modes. Whereas Sutton et al. (2023) found that the predictive capacity in gannets increased with the incorporation of other movement metrics (e.g. total distance travelled), our DBA-only models achieved a predictive capacity within the range of their best models (from 0.6 to 0.8). Whereas all previous studies occurred in waters below the thermoneutral zone of birds, either in polar regions or in southern Australia (Elliott et al., 2013; Stothart et al., 2016; Ste-Marie et al., 2022; Hicks et al., 2020; Sutton et al., 2021, 2023), with a DEE_DLW–DBA r>0.6, our study occurred in waters with sea surface temperatures of 20°C. Thus, the unexplained variance in previous studies does not appear to be directly related to thermoregulation in cold water (which would increase DEE for a given DBA). Perhaps overheating in flight or at the colony, leading to higher DEE_DLW but not higher DBA, is a greater source of error (Guillemette et al., 2016).

We present the first estimate of energy expenditure in Peruvian boobies using DLW. The need to calculate DEE of seabirds in the highly productive and over-exploited PHCS has been noted since the late 1960s. Our calculated energy expenditure values are similar to those of other sulids, such as red-footed boobies (∼1 kJ day⁻¹ g⁻¹, from Balance, 1995) and northern gannets (∼1–1.5 kJ day⁻¹ g⁻¹, from Birt-Friesen et al., 1989; Pelletier et al., 2020), but are slightly higher than those presented by Laugksch and Duffy (1984), who estimated energy requirements in a larger ecosystem-level analysis. Conversely, Boyd et al. (2014) modelled energy expenditure of Peruvian boobies by means of allometric equations, and found a DEE of ∼0.8 kJ day⁻¹ g⁻¹, much lower than our DLW direct estimation (even lower than for males alone, see Table 1). Similarly, the Seabird FMR Calculator developed by Dunn et al. (2018) estimated the Peruvian booby's field metabolic rate (FMR) for a colony at the same latitude as our study site as ∼0.9 kJ day⁻¹ g⁻¹, still lower than our DEE_DLW direct measurements, demonstrating the need for in situ estimates of FMRs.

During chick rearing, Peruvian boobies spent about ∼80% of their time in the colony, mainly attending their nest. Boobies spent most of their time away from the colony commuting, and individual foraging bouts were short because of their plunge diving habits (Fig. 1; see Appendix). As expected in diurnal, chick-rearing birds, time budgets showed a high bias toward time in the colony (Collins et al., 2016). Unlike in other sulids (e.g. northern gannets), foraging trips in boobies are fairly short (Carboneras, 1992; Nelson, 1978), hence the large amount of time spent at the colony, but energy expenditure values are fairly close to those of other sulid species that are present at the colony only 50% of the day (Sutton et al., 2023), implying that activity costs at the colony were higher at our study site.

The correlation between DBA and DEE despite the high time spent at the colony implies that DBA predicts DEE even when ‘resting’ at the colony. Densely populated seabird colonies like those of Peruvian boobies are prone to having intense aggressive interactions that imply higher than expected energy costs (Viera et al., 2011). Peruvian boobies, especially males, show highly aggressive behaviour toward other individuals in transit to their nests or scouting for females. This behaviour may occur in Peruvian boobies, as in other sulids, such as Nazca boobies, that commonly perform aggressive non-parental adult visits to chicks (Müller et al., 2011). This could explain why DBA models, rather than time budget models, and more especially the model with DBA in the colony as a separate parameter from other activities, ranked better at predicting _MSDEE_DLW. Using heart rate loggers coupled with accelerometers to further explore the periods of time at the colony when DBA and metabolic rate are high would help us to understand this relationship better.

As observed in other plunge-diving seabirds such as gannets, Peruvian boobies' flight (while commuting) metabolic rate estimated from daily time budgets was higher than those of pure gliders relative to body mass (Fig. 4). However, boobies' flight costs seemed to be much closer to the glider line than those of gannets, implying that they are more efficient flyers, perhaps because they do not flap underwater, allowing them to specialize in flying (Fig. 4). BMR of Peruvian boobies, estimated by the allometric formula from McKechnie and Wolf (2004), was 4.39 W (compared with a value of 0.78 kJ day⁻¹ g⁻¹ or 12.6 W for our birds ‘at the colony’). Our estimation of commuting flight costs was 7.8×BMR, whereas foraging was 13×BMR, which is consistent with the plunge-diving behaviour of this species, when repeated lifting from the water should be energetically demanding.

Our behavioural classification appeared to reliably partition the daily activity patterns in Peruvian boobies in the PHCS using accelerometry. We restricted our classification to four behaviours thought to be the most important energetically. Potentially, a more refined classification could be achieved by increasing behaviours (searching, resting on water, resting on land, aggressive behaviour at colony, resting at colony, etc.). Nonetheless, our best model included either two sets of two (commuting+foraging and colony+resting) or three grouped behavioural parameters (commuting + foraging + resting away from the colony) of our four initial independent set, which may indicate that modelling DEE may not improve with a greater set of parameters, as noted by Stothart et al. (2016).

Although Peruvian boobies are numerous with a stable population, their high energy requirements – being one of the booby species with the largest clutch sizes, with no signs of siblicide, and one with the highest chick rearing rates, with 1.75 chicks per breeding attempt (Carboneras, 1992) – and more importantly their high foraging costs would explain the slow recovery of their population prior to their collapse in the 1970s in the face of competition with fisheries and lower food availability as a result of climate change. Total time spent away from the colony in boobies is 1.5 times higher during years with harsh conditions (van Oordt, 2024) with an estimated DEE of 0.476 kJ day⁻¹ g⁻¹ whereas in a normal year it is 0.39 kJ day⁻¹ g⁻¹. This results in ∼20% energetic increase for Peruvian boobies. Future population models and projections should include accurate energetic requirements for boobies or any species of seabird, as well as the energetic demands associated with climate change and fisheries, which pose a threat to their population stability, recovery and management.

We would like to thank ‘guardaislas’ Moises Tomario and Andres Flores for their support during the fieldwork at Isla Guañape Norte. Special thanks to Rodger Titman for his help in the field handling and banding birds. We would like to thank SERNANP for research authorizations and SERFOR for the import permit for DLW blood samples. Thanks to Miguel Martinez for his logistic help shipping the blood samples from Peru to Canada. Thanks to Don Powers for training on lab DLW analysis methods. Thanks to Fred Tremblay for help, guidance and advice for running the LWIA machine. We thank Lauren Jackson for digitizing the drawing of boobies used in the manuscript. Thanks to Christina Petalas for her support with graphing in R and Katelyn Depot for revising the manuscript. We thank an anonymous reviewer and Jon Green for their very useful comments on the manuscript.