Trade-offs between foraging reward and mortality risk drive sex-specific foraging strategies in sexually dimorphic northern elephant seals

Sex-specific phenotypic differences are widespread throughout the animal kingdom. Reproductive advantages provided by trait differences come at a cost. Here, we link sex-specific foraging strategies to trade-offs between foraging reward and mortality risk in sexually dimorphic northern elephant seals (Mirounga angustirostris). We analyse a decadal dataset on movement patterns, dive behaviour, foraging success and mortality rates. Females are deep-diving predators in open ocean habitats. Males are shallow-diving benthic predators in continental shelf habitats. Males gain six times more mass and acquire energy 4.1 times faster than females. High foraging success comes with a high mortality rate. Males are six times more likely to die than females. These foraging strategies and trade-offs are related to different energy demands and life-history strategies. Males use a foraging strategy with a high mortality risk to attain large body sizes necessary to compete for females, as only a fraction of the largest males ever mate. Females use a foraging strategy with a lower mortality risk, maximizing reproductive success by pupping annually over a long lifespan. Our results highlight how sex-specific traits can drive disparity in mortality rates and expand species' niche space. Further, trade-offs between foraging rewards and mortality risk can differentially affect each sex's ability to maximize fitness.


Foraging strategies
Northern elephant seals used sex-specific foraging strategies, differing in their movement patterns, foraging locations and dive behaviour (figure 1). Hierarchical clustering analysis (HCA) of 130 seals (14 males, 116 females) and 31 foraging variables resulted in three foraging groups-one male strategy and two female strategies ( post-breeding and post-moult; electronic supplementary material, tables S1-S3). Analyses of variance (ANOVAs) identified significant differences in the movement patterns and dive behaviour associated with each foraging strategy (table 1).

Overlap between strategies
The sexes showed little to no overlap in horizontal and vertical space use (table 2). When comparing twodimensional (2D) satellite tracks, females only showed a 4% overlap with male foraging ranges (95% utilization distribution overlap index (UDOI) = 0.002). Females had no overlap with male core foraging areas (50% UDOI = 0.00). When comparing three-dimensional (3D) satellite tracks and vertical dives, Table 2. Comparison of 2D and 3D foraging ranges (95% utilization distribution, UDs) and core foraging areas (50% UDs) and percentage of overlap of the foraging ranges and core foraging areas between male and female northern elephant seals. females had a 5% overlap with male foraging ranges (95% UDOI = 0.04). Females had a 6.3% overlap with male core foraging areas (50% UDOI = 0.05).

Discussion
Northern elephant seals have a broad ecological niche. Males and females, however, occupy separate subsets of the species' overall niche space. Northern elephant seals use sex-specific foraging strategies Table 3. Northern elephant seal foraging success variables (mean ± s.d.) for the three foraging strategies. Letters indicate significant differences between strategies based on post hoc pairwise contrasts ( p ≤ 0.05). that are characterized by specific movement patterns and dive behaviour. Males travel to coastal habitats along the eastern North Pacific coastline and feed continuously in small, localized foraging areas over the continental shelf. Males also take relatively shallow benthic foraging dives (approx. 230 m) to catch prey on or near the seafloor (e.g. sharks, cephalopods, fish; [38,39]). Conversely, females travel throughout vast swathes of the North Pacific Ocean and feed in open ocean habitats. Females have large foraging areas and often travel between multiple prey patches. Females use deep (approx. 500-600 m) pelagic foraging dives to consume mesopelagic prey (e.g. fishes, squids; [40]). Females' dives also have a strong diurnal pattern to track vertically migrating prey. Within the general female strategy, females show seasonal foraging patterns. Females on the post-breeding trip have smaller foraging areas, shorter foraging dives with fewer prey capture attempts, and lower dive efficiency than females on the post-moult trip. The two female strategies are probably related to the different trip lengths, seasonal changes in prey and the physiological demands of gestation experienced by post-moult females [40,41]. These foraging behaviours are largely consistent with previous work [27][28][29]31,40]. Southern elephant seals (Mirounga leonina), the sister taxa to northern elephant seals, also show similar sexual segregation in foraging behaviours [42][43][44][45].
Habitat-specifically, coastal or open ocean-leads to strong convergence in marine megafauna movement patterns [46][47][48]. The open ocean is a highly dynamic environment that contains patchily royalsocietypublishing.org/journal/rsos R. Soc. Open Sci. 9: 210522 distributed prey [46][47][48]. Consequently, pelagic predators often travel large daily distances that are interspersed with short periods of concentrated movements when animals feed in a prey patch [46,47]. Coastal habitats, by comparison, are stable with high prey concentrations [46][47][48]. Coastal predators often travel short daily distances and show high variability in vertical movements when feeding [29,[46][47][48][49]. Previous work suggests that both northern and southern elephant seals exhibit convergence in movement patterns and dive behaviour based on physical habitat [29][30][31][32][46][47][48][49]. Here, 6% of females travel to coastal habitats, like males. Unlike males, however, these females rarely, if ever, forage on the continental shelf. Instead, these coastal females feed in waters immediately adjacent to the shelf edge and use deep pelagic foraging dives to consume prey. Similarly, some males travel through pelagic habitats on their way to remote continental shelf habitat. However, they rarely, if ever, slow down until they begin benthic foraging over the continental shelf. Males and females, therefore, exhibit sexual segregation in their space use, which results in intraspecific niche divergence. Males, as the larger sex, need absolutely more energy than females to support and sustain their body masses. Males also expend 5-6 times more energy during the breeding season compared with females, with the largest and most dominant males using the most energy [50]. The males' huge energy deficit must be-and is-recovered by feeding in benthic continental shelf ecosystems. Males have higher foraging success than females, gaining up to six times as much mass and acquiring energy 4.1 times faster. Males' high foraging success, though, appears to come at the cost of an increased mortality rate.
Males are six times more likely to die on the at-sea foraging trips than females. It is difficult to determine the cause of mortality for wide-ranging animals that mostly die at sea [34]. As a result, the frequency of all potential causes of northern elephant seal at-sea mortality (e.g. disease, injury, marine debris, predation, ship strikes, starvation) is unknown [23,25,34,51,52]. Predation and starvation are thought to be the leading causes of mortality in adult northern elephant seals [23,25], as opposed to injuries sustained during intense breeding competitions [51]. Here, we hypothesize that males experience increased predation pressure because their foraging areas overlap closely with those of known predators that results in males' high mortality rate.
White sharks (Carcharodon carcharias) and killer whales (Orcinus orca)-the two primary predators of northern elephant seals-hunt in coastal habitats along the eastern North Pacific coast [53]. Two-thirds of male mortality events occur in coastal habitats compared with only 9% of female mortality events. Males are found in killer whale stomachs more frequently [54][55][56] and are documented with more white shark bites than females [57,58]. Other North Pacific marine mammals also experience high levels of white shark and killer whale predation in coastal habitats [59]. Conversely, we find no significant differences in body condition or foraging behaviour between seals that lived and died that would support the starvation or injury hypotheses. Therefore, predation may be the primary cause of sexspecific mortality in northern elephant seals, but this requires further investigation. Specifically, future research is needed to: (i) define and quantify different sources of at-sea mortality in northern elephant seals, and (ii) determine satellite transmitted space use variables associated with different sources of atsea mortality.
Our findings suggest that foraging strategy, potentially through sexual segregation of foraging habitat, drives sex-specific differences in northern elephant seal foraging success and mortality rates. Open ocean habitats provide sufficient prey resources for females to meet their physiological requirements while also appearing to minimize the risk of mortality. In contrast, males must feed in benthic continental ecosystems with abundant prey resources to meet their energetic requirements; however, these areas also appear to be associated with a higher risk of mortality.
Intraspecific niche divergence in northern elephant seals supports their different life-history strategies for maximizing fitness. Males reach physical and sexual maturity later in life and have lower annual survivorship than females [20,24,60]. Males must invest in growth to quickly obtain and then maintain the large body sizes required to be competitive in physical combats for access to mating opportunities and for protracted time spent onshore fasting [50,61]. Only a fraction of males that survive to adulthood-the largest and most dominant-successfully reproduce [20,60,61]. The extreme disparity in male reproductive success counterbalances the high mortality rate. A large proportion of males can die without affecting the population dynamics [60]. By contrast, females have higher annual survivorship and reach sexual maturity quicker than males [24,60]. The most successful females have long lifespans in which they reproduce annually until their death [24,25]. Females can meet their energetic demands in the pelagic environment and avoid the increased mortality risk, which prolongs their reproductive years.
Many sexually dimorphic taxa exhibit niche divergence, but it is often restricted to specific temporal periods around breeding and rearing offspring [18,19,62]. Male and female northern elephant seals, royalsocietypublishing.org/journal/rsos R. Soc. Open Sci. 9: 210522 however, operate in different ecological niches most of the year, except during the synchronous annual breeding season. Southern elephant seals also exhibit sexual dimorphism and sex-specific foraging strategies [42][43][44][45], with potential trade-offs between foraging habitat use and mortality rate [34]. Niche divergence may therefore have evolved early in elephant seals. A few closely related seals (e.g. grey seals, leopard seals) feature some form of sexual dimorphism with accompanying sex-specific differences in diet and/or foraging behaviour [63,64]. Most seals, however, do not [65,66]. Among pinnipeds (seals, sea lions and walruses) more broadly, some sexually dimorphic taxa show intrasexual differences in foraging patterns [63,67], while others do not [68,69]. It is therefore unlikely that sexual dimorphism and accompanying niche divergence represent the ancestral condition in pinnipeds. Instead, elephant seals provide a compelling example of how evolutionary and ecological processes can operate differently between the sexes.
Northern elephant seals exemplify many of the predictions associated with the evolution of sexual dimorphism. The consequences of niche divergence on fitness have remained elusive for many species [2,[6][7][8]. Here, we show that northern elephant seal foraging strategies represent a trade-off between mortality risk and foraging reward. These risk-reward foraging patterns of northern elephant seals match predictions associated with the predation-risk hypotheses, which have mostly been examined in ungulates with mixed results [70,71]. The male northern elephant seal strategy requires high foraging success to sustain a large body size required for reproductive success; however, this comes at the cost of low survival. Alternatively, the female strategy promotes fitness by maximizing lifespan to increase reproductive success. Our findings highlight the interplay between sexual dimorphism and sexspecific behavioural and life-history strategies. Further, we demonstrate that trade-offs between mortality risk and foraging reward probably affect fitness by driving differences in the sexes' mortality rates. Finally, these results emphasize the importance of studying the biology of both sexes, especially for sexually dimorphic species where the sexes occupy completely different subsets of the overall species' niche space. August-January). Seals were instrumented with 0.5 W ARGOS satellite transmitters (SPOT or SPLASH tags, Wildlife Computers or Conductivity-Temperature-Depth tags, Sea Mammal Research Unit), timedepth recorders (TDRs; MK9 or MK10-AF, Wildlife Computers) and VHF radio transmitters (Advanced Telemetry Systems) using quick-set epoxy, high-tension mesh netting and cable ties. We chemically immobilized seals using an established protocol (detailed description in [31]) to attach the instruments and collect morphometric data and tissue samples [28,31,72,73]. All female and some male seals were previously flipper tagged as part of ongoing demographic studies of northern elephant seals at Año Nuevo. For flipper-tagged seals, we had long-term individual life-history and resighting data [25]. For untagged seals, a flipper tag was inserted into the webbing of each hind flipper during sedation. The flipper tags aided in individual identification during the biannual haulouts. All instrumented seals also had their flipper tag number written on their sides in hair dye to facilitate visual identification upon their return to the colony. Upon the seal's return to the colony at the end of the foraging trip, seals were chemically immobilized to recover the instruments and to collect morphometric data and tissue samples.

Body composition
Each seal's body composition was measured on instrument deployment and recovery [28,30,31,73]. Length and girth measurements were taken at eight locations along the seal's body. Blubber thickness was measured with an ultrasound or a backfat meter at 12-18 locations along the body. We measured females' mass with a Dyna-Link digital scale attached to a tripod. Males' mass was estimated from the combination of lengths, girths and ultrasound measurements [50]. Mass was corrected for the time spent on shore. For females, mass change on shore was estimated using an equation derived from serial mass measurements of fasting seals from previous studies: mass change (kg d −1 ) 0.51 + royalsocietypublishing.org/journal/rsos R. Soc. Open Sci. 9: 210522 0.0076 × mass, n = 27, r 2 = 0.79, p < 0.01 [73]. When females arrived after the post-moult trip, the recovery procedure occurred after parturition, and her pup's mass was added to the female's mass. For males, mass change on shore was estimated using a metabolic rate of two times the predicted metabolic rate for a mammal of equal size during the moult and 3.1 times during the breeding season [50,74,75] and fat and protein contributions to metabolism from Crocker et al. [50]. Energy gain was estimated assuming the adipose tissue was 90% lipid and lean tissue was 27% protein, with a gross energy content when mobilized of 37.3 kJ g −1 for lipids and 23.5 kJ g −1 for protein [76,77]. These estimates of body condition have been validated with those from the dilution of isotopically labelled water [77].

Data analysis
Statistical analyses and data analyses were conducted in R v. 3.3.3 [78]. Tracking and dive data were processed using standard filtering techniques and protocols for processing the satellite transmitter and TDR data as described in detail in [30,31]. Specifically, we truncated raw ARGOS tracks to the departure and arrival times from the breeding colony as identified from the diving record or sightings database. A speed, distance and angle filter removed unlikely position estimates (thresholds: 12 km h −1 and 160°; argosfilter; [79]. The filter also examined secondary position calculations reported by ARGOS and replaced erroneous prior positions if speed/angle criteria were met [30,31]. Tracks were smoothed using a state-space model that yielded hourly estimates of position and incorporates estimates of at-sea ARGOS error (crawl; [80,81]). For females tracked in multiple years, we randomly removed repeat tracks so each seal was included in the analysis only once. All dives were georeferenced using the satellite track. Dive data were collected at sampling intervals between 1 and 8 s but were subsampled to 8 s for comparison. Dive behaviour was only analysed from recovered TDRs that recorded a full time series of data (Wildlife Computers MK9 or MK10-AF). The raw time series of depth measurements were analysed using a custom-written dive type script in Matlab (IKNOS Toolbox, Y Tremblay 2008, unpublished data; [30,31]). Identified dives were only retained if they exceeded 32 s in duration and 15 m in depth. Dives were classified into one of four types (pelagic foraging dives, benthic foraging dives, drift dives or transit dives) using a custom hierarchical classification program; this program was designed to detect the unique characteristics of each dive type (e.g. depth, shape, duration) as recorded by the TDR [30,31,36,72]. Females exhibit diel diving patterns, so we assigned each dive to day or night based on the solar zenith angle associated with each dive.

Foraging variables
We calculated the distance to the continental shelf edge (km) for each foraging location, which was defined by the 200 m depth isobath. We determined the proportion of time each seal spent feeding. We also calculated each seal's foraging region (km 2 ), defined as 95% contour area determined from the utilization distributions (UDs). The UDs were generated from kernel density analyses of 2D foraging locations (latitude + longitude) for each track using a 2 km cell size and default bandwidth in ArcGIS 10.3.1. Each track was assigned to a mesopelagic ecoregion ('ecoregion'; [35]) based on where the majority (greater than or equal to 50%) of foraging locations occurred. Each track was also assigned a habitat type [29,82]; tracks were categorized as coastal, coastal/open ocean or open ocean habitat. Lastly, each track was classified as 'focused' or 'throughout' based on the location of foraging points in relation to the furthest point from the breeding colony. 'Focused' trips occurred where feeding occurred at the most distant part of the track from the colony and less than five foraging locations were identified in other portions of the track. Trips were classified as 'throughout' when more than five foraging locations occurred outside the farthest point from the colony [83,84].
For each TDR record, we determined the proportion of the dive record in which seals used transit, pelagic foraging, drift and benthic foraging dives. For pelagic and benthic foraging dives, we calculated several dive metrics. We determined the mean maximum depth (m), bottom time (min) and the postdive surface interval (min). We also calculated the mean number of vertical excursions at the bottom of each dive, which represent prey capture attempts [37] and the mean dive efficiency (bottom time/dive duration). Values closer to 0 indicate lower dive efficiency and values closer to 1 indicate higher dive efficiency [23,72]. We examined data for deviations from normality using density and Q-Q plots and a Shapiro-Wilks test. We then assessed variance with F-tests. Welch two-sample t-tests were used to compare diurnal patterns in seal dive behaviour. When data were not normally distributed, we ran non-royalsocietypublishing.org/journal/rsos R. Soc. Open Sci. 9: 210522 parametric Mann-Whitney-Wilcoxon tests. We applied a Bonferroni correction to account for the multiple comparisons, and adjusted p-values were used to assess significance.

Foraging success
We measured absolute and relative metrics of foraging success in northern elephant seals. We measured the seal's body mass at departure (kg) and determined the seal's total mass gain over the foraging trip (kg). We also determined the seal's rate of mass gain over the entire trip (kg d −1 ) and the rate of mass gain relative to feeding time (kg d −1 ). Feeding events were identified from satellite tracking and dive behaviour data. First, we calculated transit rate from the interpolated satellite tracks and then filtered the data for transit speeds less than 2 km h −1 , which are associated with foraging dive behaviour [36]. We further filtered the data to only include foraging locations that were associated with a foraging dive type (either pelagic and/or benthic foraging dives; [30,36]). Finally, we then calculated the proportion of time each seal spent feeding relative to their total time at sea. We calculated the seal's total energy gain on a trip (MJ), their total rate of energy gain (MJ d −1 ), and the rate of energy gain relative to feeding time (MJ d −1 ) based on the seal's rates of mass gain and body composition. We also calculated the proportion of a seal's mass gain on the foraging trip in relation to its departure mass to account for the sex-specific differences in body size.

Mortality rate
For each seal with an instrument that stopped transmitting (71 out of 217 seals), we did the following: (i) determined whether the seal was seen alive in subsequent years using historical records from 2006 to present from Año Nuevo State Park; and (ii) examined satellite position quality across the duration of the trip to ascertain whether the tag malfunctioned (e.g. degraded location quality signal over time) or functioned properly (e.g. random distribution of location qualities throughout the trip). Seals were resighted daily at Año Nuevo during the breeding and moult haul-outs. Seals were resighted weekly the rest of each year. We collected data on age, behaviour and haul-out history using archived resight records from the 1980s to present for each seal. Ground surveys and/or aerial photographic surveys (plane, drone) are conducted at least once each year at each northern elephant seal breeding colony in the USA, providing annual census data [33]. Double flipper tagging and retagging individuals helped account for biases that could result from flipper tag loss and/or inability to read tags [25]. To reduce bias due to instrumented seals returning to other colonies, we receive regular reports from stranding networks and colleagues observing seals at other colonies whenever seals flipper tagged and/or instrumented at Año Nuevo emigrate to other colonies [25,60]. Previous studies of female seals also show that emigration/immigration most frequently occurs in juveniles, with 90-95% of adult northern elephant seals showing high site fidelity to their natal colony [60]. If the tag was working correctly at the last transmission time and that seal was never seen again, the seal was then presumed dead. For each seal that died, we calculated the following metrics for the last known location: latitude and longitude, transit rate (m s −1 ), distance to the continental shelf edge (km), ecoregion and portion of the trip (outward, farthest point or return). We used non-parametric Mann-Whitney-Wilcoxon tests to examine differences in the movement patterns between males and females that died. Adjusted p-values from a Bonferroni corrected were used to assess significance.

Statistical analyses
We conducted principal components analysis (PCA) on 31 foraging variables, encompassing seal movement patterns and dive behaviour (base-R, FactoMineR, missMDA, RColorBrewer, scales, tidyverse, gridExtra, here; [78,[85][86][87][88][89][90][91][92][93]). Seals missing the majority of data were excluded from the analysis, resulting in a dataset of 14 males and 116 females. Variables were centred and scaled prior to PCA. A scree plot was used to examine natural breaking points in the variance. Principal components (PCs) with eigenvalues greater than or equal to 1.0 and that explained greater than or equal to 10.0% of the variation were retained. A coefficient correlation analysis was used to assess the contribution of each variable to each PC axis. Three PCs explained 56.2% of the total variation, and all variables were significantly correlated with one or more PC axes (electronic supplementary material, tables S1 and S2). Principal components 1-3 were analysed using HCA to examine naturally occurring clusters of individuals in the dataset. We created a dissimilarity matrix based on Euclidean distances and performed an agglomerative HCA using 'hclust' and Ward's linking method on the PC scores royalsocietypublishing.org/journal/rsos R. Soc. Open Sci. 9: 210522 (tidyverse, cluster, factoextra, dendextend; [93][94][95][96]). Elbow and average silhouette methods were used to determine the optimal number of clusters and assign each seal to a cluster (electronic supplementary material, table S3).
We set up generalized linear models for each foraging variable to determine foraging variables that best discriminated among clusters. Generalized linear models had a Gaussian distribution and an identity-link function. Cluster was the predictor variable. Each candidate model was compared with a null model (intercept only) using likelihood ratio tests of the null and residual deviances. Models were ranked using the Akaike information criterion corrected for small sample sizes (AICc; AICcmodavg; [97]). We evaluated the fit of the best model using an ANOVA (Type II, Wald's test, car; [98]). We then used estimated marginal means to perform post hoc pairwise contrasts between each cluster and used Tukey's method for adjusting the p-value for multiple comparisons (emmeans; [99]).
We examined the overlap between male and female 2D and 3D foraging ranges and core foraging areas by comparing UDs; this same approach was recently used in a study of nine individual southern elephant seals [100]. We used kernel density estimation to determine the 95% and 50% UDs for each sex, which represented foraging range (km 2 ) and core foraging areas (km 2 ), respectively [100]. The maximum dive depth was determined for each 2D foraging location (latitude + longitude) to create the 3D dataset. We only included locations and dives associated with foraging to examine overlap in the foraging space. A data-based 'plug-in' bandwidth selector (Hpi) was calculated for each dataset, and 2D and 3D kernel density UDs were calculated for males and females (ks; KernSmooth; MASS; [101][102][103][104][105]). We also calculated the proportion of overlap in the area (km 2 , 2D-UD) and volume (km 3 , 3D-UD) between the sexes and calculated the UDOI, which provided a measure of space-sharing use where values close to 0 represented no overlap and 1 indicated complete overlap (misc3d; [106,107]).
We tested for differences in mortality rate between the sexes using logistic regression models with a binomial distribution and logit-link function; these models tested the distribution of the data against a logistic curve to account for the categorial nature of the response variables (i.e. 'survived' = 0, 'died' = 1; [108,109]). Variables in the full model included sex, individual, tagging year and foraging trip (breeding or moult). We ran all possible combinations of the model that included sex as an explanatory variable using glm in R (electronic supplementary material, table S4). We compared the null and residual deviances of the models using likelihood ratio tests (stats, [78]). We ranked models using the Akaike information criterion and considered models of biological importance when ΔAIC was less than or equal to 7 [110]. We then tested the difference between our best-fitting model and the observed data using a Hosmer-Lemeshow goodness-of-fit test (ResourceSelection; [111]). We used an ANOVA to evaluate the fit of the best model and identified the most important explanatory variables with a Wald's test (car; [98]). We also modelled the probability of dying by sex across the foraging trips for our best-fitting model (emmeans; [99]). Data accessibility. Life history, movement, dive behaviour, foraging success and mortality rate data for the northern elephant seals and the R scripts for data analysis and figure generation can be found on Dryad Digital Repository: https://doi.org/10.5061/dryad.c2fqz617f.