Influence of hunting strategy on foraging efficiency in Galapagos sea lions

Study site and data collection

This study was undertaken as part of the GSL population-monitoring program conducted by the Galapagos National Park Directorate (GNPD) and the Universidad San Francisco de Quito (USFQ) under research permits PC-19-12, and PC-61-13. The methods described here were reviewed and approved by the USFQ and GNPD committee responsible for assessing animal welfare in research activities.

This study was conducted at the El Malecon colony, San Cristóbal Island, Galapagos Islands (0.9019°S, 89.6142°W; Fig. 1) from November–December 2012 (i.e., during a La Niña phase (NOAA, 2020)). The El Malecon colony is the largest colony of GSL and San Cristóbal Island hosts approximately 13% of the GSL population (Riofrío-Lazo, Arreguín-Sánchez & Páez-Rosas, 2017). Adult females nursing pups 1–3 months old (Riofrío-Lazo, Arreguín-Sánchez & Páez-Rosas, 2017) were selected opportunistically according to ease of capture access at the site. Individuals were captured using a modified hoop net before being tranquillised with a 0.8-1 mL intra-muscular injection of Telazol (teletamine/zolazepam HCl) at a 100mg/ml concentration. Once tranquillised, individuals were weighed using a sling, tripod and a digital scale (±0.1 kg). Standard length and axillary girth (Mammals, 1967) were recorded using a tape measure (±0.5 cm).

Individuals were instrumented with an ARGOS satellite telemetry transmitter or Fastloc GPS data logger combined with a dive behaviour data logger (MK9-AF and MK10-AF, Wildlife Computers, Richmond, WA, USA). The GPS devices were programmed to record spatial locations every 15 min with an estimated accuracy of 0.5–11 km (Costa et al., 2010). Dive behavior loggers were set to sample depth (± 0.5 m) at 2 s intervals. Instruments were attached to a neoprene layer and mesh netting and glued with 5-min epoxy (Loctite) to the dorsal midline between the scapula. A VHF radio transmitter (Sirtrack, Havelock North, New Zealand) was also attached to the dorsal midline posterior to the data loggers to assist in relocation for recapture. A tri-axial accelerometer data logger (X6, Gulf Coast Data Concepts, Waveland, MS, USA) was attached using the same method to the top of the head between the ears and programmed to sample at 25 Hz. Animal handling lasted <30 min before individuals were released to resume normal behaviours. Individuals were recaptured after 1–7 foraging trips to sea and the devices removed by cutting through the neoprene layer underneath them. The residual epoxy mounts fall off within a few months during the animals’ annual moult.

Dive behaviour and prey encounter events

Dive behaviour data were initially zero-offset corrected (ZOC) to account for pressure sensor drift and summary statistics for each dive were extracted using the diveMove package (Luque, 2007) within R statistical environment version 3.5.2 (R Core Team, 2018). Only submergences exceeding a depth >5 m and durations >12 s were considered foraging dives and included in analyses (Jeglinski et al., 2012; Villegas-Amtmann et al., 2013). The remaining 10261 foraging dives were visually classified based on depth and dive profile as: (a) epipelagic (<100 m; EP); or (b) mesopelagic (>100 m; MP) if with V-shaped or U-shaped profiles; and (c) shallow benthic (<100 m; SB) or (d) deep benthic (>100 m; DB) if with square or flat-bottom dives.

The GPS positions were decoded and filtered using a speed filter of 6 km h⁻¹ to remove erroneous locations (Villegas-Amtmann et al., 2013). Satellite data (ARGOS) were pre-processed using a conservative forward/backward speed filter of 10 m s⁻¹ to remove erroneous locations (Patterson et al., 2010). The ARGOS data were additionally processed using a State Space Model (SSM) in order to obtain positions estimates that included errors related to each ARGOS location class and the dynamics of the movement process (Jonsen, Flemming & Myers, 2005; McConnell et al., 2010; Patterson et al., 2010). Both the GPS and ARGOS location data were then interpolated at 10-s intervals and merged with dive summaries. Female GSL are known to haul-out in multiple locations within a foraging trip (Villegas-Amtmann et al., 2008; Montero-Serra et al., 2014). Consequently, at-sea tracks were visually analysed to determine when an individual left and returned to the colony to define the start and end times of a foraging trip in the dive record. In addition, periods when individuals were hauled-out were subtracted from the duration away from the natal colony to obtain the effective time at sea. As GSL enter the water near the colony to thermoregulate, only departures from the colony >2 h were considered foraging trips.

The diving data from the individual foraging trips were merged with the tri-axial accelerometer data in the software package IgorPro (WaveMetrics, Lake Oswego, OR, USA). Potential prey captures (PPC) were estimated from peaks in the Y-axis (sway) acceleration data using the Ethographer package (Sakamoto et al., 2009) in IgorPro following the methods of Volpov et al. (2015) and Foo et al. (2016). Briefly, a 3-Hz high-pass filter was applied to the sway acceleration to isolate head movements from movements associated with the body (Viviant et al., 2010). A 1.5-s moving time window was then applied to the filtered data to determine the standard deviation (SD) and a threshold was calculated above which peaks were classified as a PPC. As peaks occurring within quick succession could reflect prey handling rather than additional prey encounters, a minimum time interval was used between two successive peaks to separate PPC. This method of identifying PPC has been validated using animal-borne video cameras on free-ranging female Australian Fur Seals (Arctocephalus pusillus doriferus) (Volpov et al., 2015).

To identify the optimal threshold on the SD of the sway acceleration and minimum time interval for detecting PPC in GSL, a range of threshold values from 0.1 to 0.6 g were tested for each individual. Thresholds were determined by: (1) assessing the dominant SD of the sway acceleration values within a frequency histogram; and (2) examining the raw acceleration data (Foo et al., 2016). The threshold was then chosen based on the highest rate of change. Minimum time intervals increasing by 0.5 s from 1–7 s were used to calculate PPC and the rate of change in PPC detection for each individual. The optimal minimum time interval was selected using the same methods used for the optimal threshold. A threshold on the SD of the sway acceleration of 0.4 and minimum time interval of 4 s was selected to determine PPCs for all GSL individuals except for one. For this individual the SD of a sway acceleration threshold of 0.2 and a minimum time interval of 5 s were selected due to a reduced detection rate. This was likely due to the placement of the accelerometer causing inaccurate sampling (Preston, Baltzer & Trost, 2012).

Energy expenditure and foraging efficiency

Previous studies have shown a correlation between vectorial dynamic body acceleration (VeDBA), calculated from the tri-axial accelerometer data, and energy expenditure in a range of species (Qasem et al., 2012; Jeanniard-du Dot et al., 2017a; Grémillet et al., 2018; Sutton et al., 2020). Correspondingly, static acceleration (S; due to individual’s position relative to gravity) was calculated for each axis (x, y and z) using a 2-s running mean and subtracted from the acceleration (A) to provide dynamic acceleration due to movement (Shepard et al., 2008; Qasem et al., 2012). The VeDBA throughout complete foraging trips was calculated using the equation: ${\displaystyle VeDBA=\sqrt{{\left(A_x-S_x\right)}^2+{\left(A_y-S_y\right)}^2+{\left(A_z-S_z\right)}^2}.}$

The resulting VeDBA values were used to estimate energy expenditure (MJ kg⁻¹) from conversion equations previously determined in the Antarctic (Arctocephalus gazella) and northern (Callorhinus ursinus) fur seals (Jeanniard-du Dot et al., 2017a) during diving and while at the surface:

${\displaystyle E{E}_{\text{dive}}\left(\text{MJ}{\text{kg}}^{-1}\right)\sim \left(0.10\pm 0.10\right)+\left(91.99\pm 4.42\right)\times {VeDBA}_{\text{dive}}\left({\text{ms}}^{-2}{\text{kg}}^{-1}\ast \text{day}\right)}$ ${\displaystyle E{E}_{\text{surface}}\left(\text{MJ}{\text{kg}}^{-1}\right)\sim \left(0.06\pm 0.07\right)+\left(23.40\pm 1.48\right)\times {VeDBA}_{\text{surface}}\left({\text{ms}}^{-2}{\text{kg}}^{-1}\ast \text{day}\right)}$ ${\displaystyle \text{Total}EE\sim E{E}_{\text{dive}}+E{E}_{\text{surface}}.}$

Estimates of the energy content (EC) gained by PPC were obtained from information on the prey items (species and size) reported in a concurrent diet study on GSL at the same location (Páez-Rosas, Villegas-Amtmann & Costa, 2017) and energy content values for these species derived from the literature (Table S1). Prey species were classified according to their habitat to obtain an average prey energy content for the different dive types identified (Lee, Gelwick & Davis, 2010). Estimated EC gained by individual GSL was then calculated based on the PPC and the dive type in which they occurred.

Foraging efficiency (FE; kJ s⁻¹) was calculated from the estimated gross energy expenditure and the total energy gained while foraging using the equation: ${\displaystyle \mathrm{FE}=\frac{{\mathrm{GE}}_{\text{consumed}}-G{E}_{\text{expended}}}{{\text{time}}_{\left(\text{s}\right)}}}$

where GE_consumed is the number of prey items multiplied by the estimated EC of potential prey items based on Dive Type for individuals’ dives or summed across a foraging trip, and GE_expended is the energy expended during individual dives or the entire foraging trip.

Linear mixed-effects models (LMEs) were used to evaluate the influence of Dive Type on the response variables of PPC, energy consumed, energy expended and foraging efficiency estimates using the lme4 (Bates et al., 2015) package in R. The individual was categorised as a random factor to account for repeated measurements (dives) within individuals. Parameters describing individual components of dives such as dive depth, duration, and bottom time were not considered part of the modelling process due to their use, and thus collinearity, in the classification of dive types. Model residuals were tested for skewness by visually assessing boxplots, and no transformations were necessary. Differences between Dive Types were evaluated using an ANOVA run on the model outputs, which were followed by Tukey’s multiple comparisons using the multcomp package (Hothorn, Bretz & Westfall, 2008).

As individuals displayed multiple Dive Types within and between foraging trips, to investigate factors influencing potential prey capture rates and FE at the trip level a multivariate hierarchical classification analysis was conducted. The classification using Euclidean distance and the ward D2 linking method (Murtagh & Legendre, 2014) within the pvclust package (Suzuki & Shimodaira, 2006) and groupings were based on a 95% confidence interval. Dive parameters used for cluster analysis included: total time at sea (h); total number of haul-outs; total time at haul-out (h); vertical dive rate (m s⁻¹); percentage of trip spent diving (%); percentage of dives using DB, SB, EP or MP Dive Types; mean bottom time (s); mean dive duration (s); mean maximum depth (m); mean post-dive duration (s); and ascent and descent rates (m s⁻¹). Where means were used, standard deviations and skewness values were included to account for variation within the data. The resultant groupings were considered as Foraging Trip Types.

To assess the influence of Foraging Trip Type on PPC, gross energy consumed and expended, and foraging efficiency, LMEs were created with Foraging Trip Type as a fixed effect and individual as a random effect. LMEs were analysed using the same process outlined above. Parameters describing individual components of Foraging Trip Types that were used in cluster analysis were not considered part of the modelling process due to their use, and thus collinearity, in the classification of Foraging Trip Types. As morphological features have been linked to foraging behaviours in GSL (Jeglinski et al., 2013), their influence on Foraging Trip Type was investigated with linear mixed effect models (Bates et al., 2015). To avoid pseudo-replication issues, these analyses were not conducted at the Dive Type level. Other factors such as pup size and sex were not included in analysis as the data were not available.

All statistical analyses were conducted in the R statistical environment (version 1.1.463). Unless otherwise specified, all further data are reported as mean ± SE, and results considered significant at P < 0.05. Bathymetry for maps were reproduced from the GEBCO Compilation Group (2019) website (https://www.gebco.net/) and land data were sourced from the CIAT-CSI SRTM website (http://srtm.csi.cgiar.org).

Potential prey captures, dive behaviour and foraging efficiency

In the present study, PPC were estimated utilising accelerometer data following methods previously validated with animal-borne video data loggers (88.3% precision) in another otariid species (Volpov et al., 2015; Foo et al., 2016). While it is possible that the number of PPC recorded in the present may have over- or underestimated prey captures, the average PPC rate of 0.97 per dive is comparable to that observed in Australian fur seals (Foo et al., 2016), Antarctic fur seals (Jeanniard-du Dot et al., 2017a), and northern fur seals (Jeanniard-du Dot et al., 2018).

The rate of PPC in the present study was strongly influenced by the Dive Type, with Benthic dives having a lower PPC rate than pelagic dives within the same depth range. Previous studies have suggested that individuals will have fewer prey encounters when feeding on larger, more energy-dense prey items, a trend observed in benthic foragers (Costa, 1991; Naito et al., 2013). However, in the present study a higher PPC rate was observed in both the deeper MP and DB dives. There is a potentially greater energetic cost associated with increased transit times in deep diving individuals (Rosen, Winship & Hoopes, 2007), which is possibly outweighed by a greater prey abundance and/or prey energetic content at those depths.

Sustainable dive strategies should have a favourable ratio of energy gain to cost (Emlen, 1966; MacArthur & Pianka, 1966). The GSL benthic dives had a higher energy expenditure rate than pelagic ones occurring to similar depths. This is consistent with recent studies of New Zealand sea lions (Phocarctos hookeri) where benthic foraging strategies appeared to be more energetically expensive than the other strategies observed in that species (Chilvers & Wilkinson, 2009). Benthic divers tend to spend a greater amount of time (and energy) foraging at depth and, therefore, must increase energy consumption by targeting larger prey or higher energy content items to maintain foraging optimality (Costa, 1991; Costa et al., 2004). Pelagic divers in the current study conducted dives of shorter duration than benthic divers when foraging at similar depths as has been observed in Californian sea lions (Zalophus californianus) (McHuron et al., 2016). Pelagic foragers also tend to consume smaller prey items, therefore, in order to gain adequate energy an individual must consume higher quantities of prey (Costa, 1991).

Mesopelagic dives were the exception to the positive ratio of energy consumed to energy expended in the present study. Since energy consumption was partially based on estimates of energy content of prey from previous studies, there may be biases in the estimations of prey energy content due to prey fluctuations between years or sampling bias. Nonetheless, MP dives are likely to have high energetic costs due to the high energetic cost of both transit and thermoregulation associated with achieving deeper dives. Páez-Rosas, Villegas-Amtmann & Costa (2017) reported the predominant mesopelagic species in the GSL diet to be Lanternfish (Myctophidae) and Panama lightfish (Vinciguerria lucetia). Both are small, deep pelagic fish of high energy content. Lanternfish, in particular have high levels of lipids including high levels of triglyceride amongst up to 14 other fatty acids in some species (Chai et al., 2012), potentially providing critical nutritional components to the GSL diet.

Previous studies have revealed the GSL diet is often dominated by 4–8 species of importance from distinct habitats that are complemented by a range of other species that make up a small proportion of their diet (Páez-Rosas & Aurioles-Gamboa, 2010; Páez-Rosas & Aurioles-Gamboa, 2014). Pelagic prey items such as Big-eye scad (Selar crumenophthalmucs) or Lanternfish are small vertically migrating fish with high energy density. In comparison, benthic prey items such as rock cod or groupers (Serranidae) and scorpion fish (Scorpaenidae) tend to be larger and have lower energy density (Eder & Lewis, 2005; Kumar et al., 2014; Páez-Rosas, Villegas-Amtmann & Costa, 2017). Although pelagic divers may have to invest more effort in terms of the number of PPC per dive, the energy density of their prey is generally higher, with the opposite being true for benthic divers.

Influence of foraging trip behaviour on foraging efficiency

GSL are considered generalist foragers, although recent studies have observed individuals from various colonies specialising their diet and foraging methods (Villegas-Amtmann et al., 2008; Villegas-Amtmann et al., 2013; Páez-Rosas & Aurioles-Gamboa, 2010). This trend has also been observed in other generalist populations including pinnipeds such as the Californian sea lion (Z. californianus) (McHuron et al., 2016), New Zealand sea lions (Chilvers & Wilkinson, 2009), and northern fur seals (Jeanniard-du Dot et al., 2018), as well as in sea otters (Lee, Gelwick & Davis, 2010), and seabirds (Camprasse et al., 2017). The variation of foraging behaviours or niche separation of individuals within a population should result in reduced competition pressures and increase foraging efficiency of individuals (Van Valen, 1965).

In previous studies of niche segregation in GSL, age and mass were determined to have some influence on foraging behaviour (Jeglinski et al., 2013). Similar finding have also been observed in California sea lions (Weise, Harvey & Costa, 2010) and Gentoo penguins (Pygoscelis papua) (Camprasse et al., 2017). In contrast, no significant differences in morphometrics or body condition indices were observed between the different Foraging Trip Types in the present study. While this could be due to the low sample size and variation in the individuals sampled, similar findings have been reported for New Zealand sea lions (Chilvers & Wilkinson, 2009).

Three Foraging Trip Types were observed in the present study. Two of the foraging strategies were comprised of individuals that predominately utilised either shallow benthic or epipelagic dive strategies throughout the trip. These individuals tended to complete shorter trips and had lower PPC rates and ratios between energy consumed to energy expended. Individuals that utilised the remaining strategy had predominantly longer foraging trips and spent a greater amount of time diving. Consequently, these individuals expended significantly more energy within a foraging trip. This could be related to distance travelled and the energetic costs of thermoregulation due to greater time at sea (Montero-Serra et al., 2014). Similar studies have found that increased field metabolic rate and, therefore, energy expenditure, are higher in individuals that complete extended foraging periods (Villegas-Amtmann et al., 2017).

Despite variation in energy consumption, energy expenditure, and dive parameters between groups, foraging efficiency was not significantly different between groups. The mean foraging efficiency of Foraging Trip Types ranged from 0.4 to 1.3 kJ s⁻¹, values consistent with estimates for Antarctic fur seals (2.02 kJ s⁻¹) (Jeanniard-Du-dot et al., 2017b) and northern fur seals (∼0.15 kJ s⁻¹) (Jeanniard-du Dot et al., 2018). Similar to GSL, the variation in foraging efficiency in these species are linked to the foraging behaviours. In northern fur seals, the individuals predominately using shallow benthic foraging were foraging closer to shore. They had lower foraging efficiency and PPC rates than individuals that foraged further from the colony and utilised predominantly pelagic foraging.

In 2012, the Galapagos region experienced a La Niña event, which is usually associated with cooler waters and greater productivity (NOAA, 2020). However, the 2012 La Niña was unusually warm (NOAA, 2020) and may have contributed the variation in GSL foraging trip strategies observed in the present study. Nonetheless, the variation in observed foraging trip strategies was not associated with differences in foraging efficiency. This suggests that the various currents, habitats and bathymetry surrounding the Galapagos Islands have enabled GSL to adapt various successful foraging behaviours (Drago et al., 2016; Páez-Rosas, Villegas-Amtmann & Costa, 2017). However, with the increasing frequency of major El Niño events that influence changes in climate, ecological interactions, and food availability (Cai et al., 2014), the flexibility of foraging behaviours in GSL and the ability to adapt to these stresses may be impacted (Trillmich et al., 2014). Given the current study covered only 1–7 foraging trips over a 1 month period and it is not known whether the strategies observed reflect individual specialisation (Kernaléguen et al., 2016). Further longitudinal studies are required to determine whether the observed differences in Foraging Trip Type reflect individual foraging specialisation and if they confer life-history advantages.