High mortality rates in a juvenile free‐ranging marine predator and links to dive and forage ability

Abstract High juvenile mortality rates are typical of many long‐lived marine vertebrate predators. Insufficient development in dive and forage ability is considered a key driver of this. However, direct links to survival outcome are sparse, particularly in free‐ranging marine animals that may not return to land. In this study, we conduct exploratory investigations toward early mortality in juvenile southern elephant seals Mirounga leonina. Twenty postweaning pups were equipped with (a) a new‐generation satellite relay data tag, capable of remotely transmitting fine‐scale behavioral movements from accelerometers, and (b) a location transmitting only tag (so that mortality events could be distinguished from device failures). Individuals were followed during their first trip at sea (until mortality or return to land). Two analyses were conducted. First, the behavioral movements and encountered environmental conditions of nonsurviving pups were individually compared to temporally concurrent observations from grouped survivors. Second, common causes of mortality were investigated using Cox's proportional hazard regression and penalized shrinkage techniques. Nine individuals died (two females and seven males) and 11 survived (eight females and three males). All but one individual died before the return phase of their first trip at sea, and all but one were negatively buoyant. Causes of death were variable, although common factors included increased horizontal travel speeds and distances, decreased development in dive and forage ability, and habitat type visited (lower sea surface temperatures and decreased total [eddy] kinetic energy). For long‐lived marine vertebrate predators, such as the southern elephant seal, the first few months of life following independence represent a critical period, when small deviations in behavior from the norm appear sufficient to increase mortality risk. Survival rates may subsequently be particularly vulnerable to changes in climate and environment, which will have concomitant consequences on the demography and dynamics of populations.


| INTRODUC TI ON
For long-lived marine vertebrate predators, insufficient dive and forage ability are considered key drivers of mortality during juvenile and immature stages (Daunt, Afanasyev, Adam, Croxall, & Wanless, 2007;Orgeret, Weimerskirch, & Bost, 2016). In many species, these need to develop rapidly, as individuals quickly transition from full parental care and provisioning to complete independence (Breed, Bowen, & Leonard, 2013;Costa, 1991;Reiter, Stinson, & Boeuf, 1978). The first trip made at sea alone may thus prove a critical time for survival, as individuals must either learn, learned, or have inherited advanced dive and forage abilities so as to be able to exploit the large heterogeneous environments typically inhabited (Carter et al., 2017;de Grissac, Bartumeus, Cox, & Weimerskirch, 2017;Orgeret et al., 2016;Riotte-Lambert & Weimerskirch, 2013). This may include physiological capability and foraging efficiency, knowledge of habitats and prey distribution, alongside predator avoidance strategies. Moreover, during this time, perturbations in environmental conditions can have disproportionate impacts on fitness, as individuals are already pushing their capabilities as far as they will go, leaving little room to adapt (Burns, 1999;Lea et al., 2009). Indeed, in the early years of life following transition to independence, survival rates often increase with age as dive and forage abilities advance (McMahon, New, Fairley, Hindell, & Burton, 2015;Oro, Torres, Rodriguez, & Drummond, 2010;Pistorius & Bester, 2002).

Much of what is known about the factors impacting first-year
survival stems from studies linking resightings data to an individual's condition at departure from its natal site. For example, in pinnipeds, increased girth, length, weight, and level/quality of parental provision are known to impact first-year survival (Baker, 2008;Hall, McConnell, & Barker, 2001;McMahon et al., 2015), which likely relates to initial dive capabilities, development, and subsequent restrictions on foraging ability (Hindell et al., 1999;Irvine, Hindell, Hoff, & Burton, 2000). Juvenile survival rates have also been linked to climatic indices (e.g., El Nino Southern Oscillation), possibly due to changes in the abundance and distribution of resources, which can impact parental provisioning during lactation (and thus initial condition), alongside prey availability post-weaning (Beauplet, Barbraud, Chambellant, & Guinet, 2005;. However, beyond such broadscale correlations there is a sparsity of studies describing and quantifying the fine-scale dive and forage capabilities of juvenile marine predators in relation to survival (Hazen et al., 2012). This reflects difficulties in obtaining detailed behavioral measurements of far-ranging individuals that may not return to land.
Recent advances in bio-logging technologies have substantially increased our ability to observe the fine-scale behaviors, movements, and physiology of marine species (Hart & Hyrenbach, 2009;Hazen et al., 2012;Volpov et al., 2015). Moreover, new methods in data abstraction and transmission mean device retrieval is no longer obligatory to obtain the information these instruments record (Cox et al., 2018). For pinniped species, it is now possible to remotely track a suite of dive and foraging metrics such as prey catch attempts (PrCA), swimming effort, dive shape, and surface interval (Cox et al., 2018;Heerah, Cox, Blevin, Guinet, & Charrassin, 2019;Photopoulou, Lovell, Fedak, Thomas, & Matthiopoulos, 2015). As such, the early foraging strategies of free-ranging juveniles that may not return to land can now be observed and, when coupled with double tagging techniques (Drymon & Wells, 2017;Hays, Bradshaw, James, Lovell, & Sims, 2007), examined in relation to survival outcome to gain a more mechanistic understanding of how the two may be related.
Here, we present results from a unique dataset obtained via a new generation of satellite relay data tags, capable of recording and remotely transmitting fine-scale behavioral movements of 20 juvenile southern elephant seals Mirounga leonina during their first trip at sea following weaning. We use these data to conduct exploratory investigations of the potential factors influencing early mortality events. Specifically, we (a) individually compare time-series data of behavioral movements and encountered environmental conditions of non-surviving individuals to those of grouped survivors (which successfully completed their first trip at sea following weaning) and (b) statistically assess potential common causes of mortality across the sample population.

| Tag specifications
The DSA tag measured 86 × 85 × 29 mm and weighed 192 g, which is <1% of the mean mass of the juvenile southern elephant seals in our study. The design of the device was similar to that from investigative studies demonstrating that these tags do not adversely impact mass gain or survival of equipped individuals (McMahon et al., 2008). It comprised an Argos transmitter, pressure sensor (1 Hz sampling rate, resolution 0.5 m ± 1 m +1% of a reading), tri-axial accelerometer (16 Hz sampling rate), thermistor (1 Hz sampling rate, resolution 0.1°C ± 0.1°C), and wet-dry sensor. These functioned intermittently such that one complete dive (max depth >15 m and duration >60 s) was sampled every ~2.25 hr. Measurements were processed onboard the tag to provide a "per-dive" summary, which was later transmitted (alongside an Argos-derived location) via the Argos satellite system. This included the depths and durations of five broken-stick segments of a dive, alongside the total swimming effort (summed filtered accelerations associated with flipper movements across the lateral axis of the accelerometer; Richard et al., 2014) and time spent in PrCA behaviors (total time during which accelerations reflected "jerk"-like movements; Viviant, Trites, Rosen, Monestiez, & Guinet, 2010) across each segment. Surface intervals between each sampled dive and the next were also transmitted, alongside sea surface temperature (SST; at ~2-3 m). An in-depth description of onboard processing algorithms and DSA tag functionality can be found in Cox et al. (2018). The Argos only transmitting SPOT tag measured 72 × 54 × 24 mm and weighed 119 g. From this, an Argos-derived location was obtained several times per day, such that at least one location was obtained each day.

| Survival estimates
The survival outcome of each individual was estimated using the absolute difference in time between the dates of the last transmissions of the DSA and SPOT tags. Two discrete groups were identified. The first represents individuals whose tags ceased to transmit within a short time of one another (<30 hr), likely reflecting a mortality event. The second represents individuals with a large gap between the times of last tag transmissions (>528 hr/22 days; see Table S1 in Supporting Information S.1), reflecting either retrieval of the DSA tag on return to Kerguelen Islands or its possible failure (SPOT tags always out-transmitted DSA tags, and were left attached to animals after return to Kerguelen Islands). While it is noted that there is a small chance both tags could have failed simultaneously (thus presenting a false mortality event), we consider the likelihood of this negligible, since within the surviving group, for all but one individual, tags continued to function until return to Kerguelen Islands (reflecting the completion of the first trip at sea following weaning and a successful survival outcome). For the surviving individual that did not return to Kerguelen (ID 140066; see Table S1 in Supporting Information S.1), the first of the two tags failed after 176 days at sea and the second 22 days later after a total of 198 days at sea.

| Pre-analytical data preparation
To investigate causes of mortality, for all individuals, we generated indices representative of (a) an individual's morphology and departure information, (b) horizontal movements, (c) fine-scale accelerationand dive-based foraging behaviors, (d) changes in drift rate/body condition, and (e) encountered environmental conditions (Table 1).

| Morphology and departure information
Differences in survival rates between females and males were investigated alongside the influence of departure date, weight, and condition. Departure dates were the day an individual commenced a continual trajectory away from Kerguelen Islands (~0-25 days after tag deployment; Supporting Information S.1). Departure weights were estimated from an individual's weight at tag deployment, following Guinet (1994) as weight (t) = weight (t−1) − (0.0048(weight (t−1) ) + 0.3031), where t is time in days from tag deployment to pup departure. Individuals that left Kerguelen on the same day as tag deployment had the same deployment and departure weights. Departure conditions were taken as extrapolated drift rates (departure drift rate; see details below under "Drift rates/body condition").

| Horizontal movement metrics
Argos location data were not equally positioned in time and varied in quality (specified location class errors; www.argos -system. com). To obtain filtered location estimates at a temporal resolution of 1 hr, all data were processed using a hierarchical first difference correlated random walk state-space model (SSM; Jonsen, Flemming, & Myers, 2005; Supporting Information S.2). At-sea horizontal swimming behaviors were then corrected to remove trajectory distortion caused by ocean currents (Gaspar et al., 2006;Supporting Information S.2). Following this, for each individual, the daily maximum corrected speeds observed across a single hourly period were extracted alongside the total corrected distances traveled in a day. The first and last days of each track were excluded to ensure all days encompassed a full 24 hr of tracking data and were thus comparable. To assess differences between how individuals orientate themselves to ocean currents (i.e., with or against), the daily circular mean deviation of a pup's real heading from the concurrent ocean current direction was calculated.
These values were rescaled so as bearings 180-360 ran from 180 (against ocean currents) to 0 (with ocean currents), which is consistent with those of original bearing 0-180. Departure weight (kg) estimated following Guinet (1994) as weight (t) = weight (t−1) − (0.0048(weight (t−1) ) + 0.3031), where t is time in days from tag deployment to pup departure Departure drift rate Departure drift rate (m/s) taken as an indicator of departure condition (b) Horizontal movement metrics (time series at day scale)

Max speed
The daily maximum current corrected speed (m/s) between two hourly filtered locations Distance swam The total current corrected distance swam (km) across a 24-hr period

Current deviation
The daily mean difference between corrected swimming direction and the direction of ocean currents ( o ). Current deviations were rescaled so as bearings 180-360 ran from 180 (against ocean currents) to 0 (with ocean currents), which is consistent with those of original bearing 0-180 (c) Acceleration and dive summaries (time series at dive scale)

Dive depth
The maximum depth attained during a dive (m)

Dive duration
The total duration of a dive (s)

Scaled surface interval
The surface interval divided by the total duration of a dive (s) PrCA rate Total time spent in prey catch attempt (PrCA) behaviors (total time during which accelerations reflected "jerk"-like movements) divided by total duration of a dive (s)

Scaled bottom duration
The total bottom duration divided by the total duration of a dive (s)

Swim effort
The total swimming effort (summed filtered accelerations associated with flipper movements across the lateral axis of the accelerometer) of the descent and ascent phases of a dive divided by the total corresponding durations of these phases (m/ s 3 )

| Transmitted acceleration and dive summaries
For each dive performed by each individual, the maximum dive depth was taken alongside the total dive duration. Surface intervals were scaled by dive duration, as were times spent in PrCA behaviors.
Descent, bottom, and ascent dive phases were identified following Cox et al. (2018). For each dive performed by each individual, the bottom duration scaled by the total dive duration was taken. In addition, the average swimming effort required to transit to and from the bottom, hunting phase, of a dive was estimated as the sum of the swimming efforts of descent and ascent phases divided by the total duration of these phases.

| Drift rates/body condition
Southern elephant seals regularly perform resting dives, where individuals cease active movement and "drift" in the water column (Biuw, McConnell, Bradshaw, & Fedak, 2003;Gordine, Fedak, & Boehme, 2015;Mitani et al., 2010). Vertical movement during these periods can be used to make inferences on an individual's buoyancy, and thus body composition and condition (i.e., ratio of high-density lean to low-density lipid tissue; Biuw et al., 2003). Animals that are positively buoyant should be in "better" condition than those that are negatively buoyant (due to increased fat reserves). During their first trip at sea, juvenile southern elephant seals are still developing and growing, and as such, changes in buoyancy could also reflect the acquisition of lean muscle in addition to changes in fat reserves (Biuw et al., 2003;Orgeret, Cox, Weimerskirch, & Guinet, 2019). To identify drift dives, we used a modified stepwise filtering process similar to that described in Biuw et al., (2003) and Gordine et al., (2015), but that also took into account information from accelerometer transmissions (Supporting Information S.3). Vertical movement rates in each drift segment were then calculated. Following this, to obtain daily drift rate estimates (as drift dives are not always performed daily), we used functional data analysis (Yang, Zhu, Choi, & Cox, 2016; Supporting Information S.3). Daily changes in drift rate were then estimated.

| Environmental conditions
To investigate the impact of storms on juvenile survival, wind and wave data were extracted from the ERA-Interim global atmospheric reanalysis (Dee et al., 2011)

| Data analyses
Due to a limited sample size and high level of noise/variability in the data (i.e., individuals may die for a range of reasons), investigations toward causes of early mortality were predominantly exploratory in nature. Prior to all analyses, we verified that sampling bias (i.e.,

| Temporal patterns in survival
Kaplan-Meier survival functions were generated using the survival package in R (Therneau & Lumley, 2017), and used to assess the

| Individual comparisons between nonsurviving pups and grouped survivors
For each day, for each predictor for which time-series data were available (Table 1), daily estimates for each non-surviving individual were plotted against the daily medians of the entire surviving group, alongside the 2.5%, 25%, 75%, and 97.5% quantiles. This allowed us to visually assess behavioral patterns and environmental conditions encountered in relation to time since departure (and ontogenetic processes), and the time of death (i.e., were behaviors consistently different between surviving and non-surviving individuals, or did a change occur in the days and/or weeks immediately prior to death).
For predictors not already generated at a day scale (i.e., that were generated at the dive scale; Table 1), daily summaries were calculated for use only in these visual assessments (i.e., maximum daily dive depth, maximum daily dive duration, daily mean scaled surface interval, daily mean PrCA rate, daily mean scaled bottom duration, and daily mean swim effort).

| Statistical assessment of potential common causes of mortality across the sampled population
To assess common causes of mortality across the entire sampled population, a two-step modeling approach was applied, similar to that of joint distribution modeling frameworks (e.g., Henderson, Diggle, & Dobson, 2000), albeit modified so a range of parameter estimation techniques could be applied. First, for each potential predictor for which time-series data were available (at either a daily or dive scale; proportional hazards (ph) regression models (Cox, 1972) were fitted via the survival package in R (Therneau & Lumley, 2017). Cox ph models are non-parametric and assess how candidate predictors influence the hazard rate of a particular event happening (here death). A positive estimate for a predictor reflects an increased hazard rate, and thus negative influence on survival. Models were fitted for each candidate predictor one at a time (Table 1). The response variable (survival) was specified as described in the generation of Kaplan-Meier survival functions above.
Significance against a null model was tested for using one-way analysis of variance at p < .05. The proportionality assumption was checked using the Schoenfeld residuals test (Grambsch & Therneau, 1994).
The second and third methods considered all candidate variables in tandem and selected key predictors using "shrinkage" techniques, which are both well suited to high-dimensional studies and can be tuned to deal with collinearity (Pavlou, Ambler, Seaman, Iorio, & Omar, 2016;Tibshirani, 1997;Zou & Hastie, 2005). Such methods work by shrinking all coefficients toward zero and adding a penalty on their size, such that only the stronger signals retain non-zero coefficients. The first of these was implemented using the glmnet package in R (Friedman, Hastie, Simon, Qian, & Tibshirani, 2017), which fits a Cox ph model regularized by an elastic net penalty (Friedman, Hastie, & Tibshirani, 2010;Simon, Friedman, Hastie, & Tibshirani, 2011). An optimal shrinkage penalty of 0.317 was selected by eight-folds cross-validation (Simon et al., 2011).
The second shrinkage-based analysis (and third applied method) used a regularized horseshoe prior (Piironen & Vehtari, 2017) and was implemented in Stan (Carpenter et al., 2017; Supporting Information S.6). The advantage of the regularized horseshoe is to have explicit control on the prior for the number of non-null predictor variables, which was set to 10 (based on the outputs of the two previous investigations). Survival time was modeled as a normal distribution (on a log scale), which corresponds to a non-monotonic hazard. Such a structure is appropriate since after leaving Kerguelen Islands, mortality hazard can be expected to increase after weaning (as energy capital from maternal provisioning depletes), and then decrease as pups learn to forage (Orgeret et al., 2019). To assess true and false predictor detection rates of the model of this third method (and determine its suitability to our dataset), a retrospective power analysis was also performed (Supporting Information S.7).

| RE SULTS
Of the 20 tracked pups, nine died (two females and seven males) and 11 survived (eight females and three males) until at least the end of their first trip at sea following weaning (Figure 1, Supporting Information S.1, Table S1). All individuals bar two (one survivor and one non-survivor) headed in a southeastward direction upon departing Kerguelen Islands (Figure 2). No broad-scale spatial segregation was apparent between surviving and non-surviving pups ( Figure 2 and Supporting Information S.5).

| Temporal patterns in survival
All mortality events occurred between 8.6 and 112 days after departure from Kerguelen Islands (mean = 54 ± 33.5 days, ±standard deviation), at a straight-line distance of between 521 and 2,468 km from Kerguelen Islands (mean = 1,534.3 ± 604.4 km; Figures 1 and   3). For all but two individuals, this happened before (i.e., at) the distal point of a trip (Figures 1 and 2). For the two individuals that com- Kerguelen Islands (mean = 2,036.9 ± 534.4 km). Overall, weanling survival rates during their first trip at sea were 55% (males 30% and females 70%).

| Individual comparisons between non-surviving pups and grouped survivors
Visual comparisons of behaviors and encountered environmental conditions of non-surviving pups to those of grouped survivors suggested both marked differences (at times) and high heterogeneity (Table 2; see Supporting Information S.8 for all plots used in these assessments).
This was particularly prominent for the lightest individual at departure (pup 140074; Table 2). Specifically, maximum daily dive depths and durations alongside mean bottom times and PrCA rates were low ( Figure 4). This was accompanied by increased daily mean surface intervals and swimming efforts, alongside total daily distances traveled Finally, six of the nine pups that died foraged predominantly in waters cooler than those generally visited by surviving individuals (Table 2, Figure 9). For three of these individuals (and an additional one that remained in waters of SSTs similar to those visited by survivors), maximum TKE values were also reduced (Supporting Information S.9).

| Statistical analyses of common causes of mortality across the sampled population
Eight predictors of survival outcome were selected by at least two of the three implemented methods (Table 3). In all instances, magnitude and sign agreement was good. Variable selection cutoffs for the horseshoe analysis were a parameter estimate of at least 0.05 and a shrinkage less than 0.855 (further details in Supporting Information S.10). Results from the retrospective power analysis performed on this method suggest the likelihood of a selected predictor reflecting a true signal to be 50% (Supporting Information S.7).
Male survival was lower than female survival. Survival decreased with increased daily maximum speeds and distances traveled. These results were consistent across all three methods (Table 3). A positive influence of an increase in dive duration through time on survival was identified by the Cox ph and elastic net analyses, but was one of the weaker predictors from the horseshoe analysis (relatively small parameter estimate and high shrinkage; Table 3). The four other predictors deemed to impact survival were identified by both the elastic net and horseshoe analyses, but yielded p-values from the Cox ph analysis that were above the .05 cutoff (but note that these were all between .06 and .15). There was an influence of a decrease through time in PrCA rates, such that a greater decrease increased survival probability. Individuals with swim efforts that increased through time were more likely to die than those whose swim efforts did not (or did so less). Finally, individuals that typically visited waters characterized by increased SSTs and TKE estimates were more likely to survive than those visiting waters with decreased estimates.

| D ISCUSS I ON
In this study, we capitalized on new techniques in the abstraction and transmission of accelerometer and dive data, and coupled these to double tagging methods to investigate and gain a more mechanistic understanding of how the fine-scale behaviors of juvenile southern elephant seals impact survival outcome during their first trip at sea following F I G U R E 1 Individual tracking ranges from departure from Kerguelen Islands to either time of mortality or return to Kerguelen Islands (gray bar components are outward trip phases and black return). Note only one non-survivor commenced a distinct return trip phase, which lasted 49 days (1,084.4 km) before the individual died weaning. Using a cutting-edge yet conservative analytical approach, we provide novel insight toward the causes of early mortality in this farranging marine predator, which we suggest results from a mix of reduced dive and forage ability alongside habitat quality and possible predation. This is one of few studies to use bio-logging technologies (and particularly accelerometers) to assess the factors impacting survival in a longlived marine vertebrate predator, and the methods we employ have the potential to be applied across a number of other taxa to gain novel insight of the drivers of individual mortality and population dynamics.

| Overall survival rates and sex dependency
Despite our small sample size, survival rates were similar to published estimates for 0-to 1st-year southern elephant seals (McMahon et al., , 2015Pistorius & Bester, 2002

| Initial condition at departure
In contrast to previous findings (e.g., Baker, 2008;Hall et al., 2001;McMahon et al., 2003;McMahon et al., 2015), we did not find departure condition (i.e., drift rate, weight, or date (better condition individuals may leave later); Arnbom et al., 1993, Hindell et al., 1999 to be indicative of survival outcome. This may be reflective of our small sample size and/or range of weights/initial departure conditions (Table 2 and Supporting Information S.11). For example, the maximum estimated departure weight in our sample was 96.1 kg, but studies elsewhere on juvenile elephant seals often include individuals exceeding 135 kg (McMahon, Burton, & Bester, 2000;McMahon et al., 2003).
Nonetheless, of all non-surviving individuals, the pup that exhibited the most prominent difference in dive and forage ability to that of the surviving group was also the lightest (individual 140074; Figure 4; weight at departure ~44.6 kg vs. mean of all pups ~79.9 ± 17.7 kg).

F I G U R E 4 Time series behaviors for individual 140074.
Daily values (red) for (a) maximum dive depths, (b) maximum dive durations, (c) mean scaled bottom times, (d) mean prey catch attempt (PrCA) rates, (e) mean log-scaled surface intervals, (f) mean swim efforts, (g) maximum speeds, and (h) total distances swam. Large light gray bands represent concurrent 2.5%-97.5% quantiles of survival datasets, nested dark gray bands the 25%-75% quantiles, and dark gray lines the median

| Changes in condition at sea, temporal patterns in mortality, and possible predation
Across most individuals (surviving and non-surviving), there was a marked and rapid decrease in drift rates following departure from Kerguelen Islands, which likely reflected the depletion of fat reserves obtained via maternal provisioning prior to weaning (Figure 8; Biuw et al., 2003;Orgeret et al., 2019) and was supportive of our choice of mortality hazard in the horseshoe analysis. Following this initial decrease, all pups experienced a prolonged period of negative buoyancy. While we did not detect direct links between drift rates (either initial or change through time) and survival outcome, all but two pups that died did so during this time. For the 11 that survived, all but three individuals returned to positive buoyancy before returning to Kerguelen Islands. As such, during early life, individuals appear more vulnerable to the adverse consequences of fatigue and starvation when negatively buoyant (and likely of poor body condition). For southern elephant seals, during the initial, outward phase of the first trip at sea, the rapid development of dive and forage skills is thus crucial as fat reserves gained from maternal provisioning diminish (McConnell, Fedak, Burton, Engelhard, & Reijnders, 2002), and a return to positive buoyancy following this appears an important determinant in the decision to end the first trip at sea and return to land (Orgeret et al., 2019). This may also reflect the sufficient development of muscular lean tissue, after which excess energy is stored as fat. The one pup (140071) that died while positively buoyant during return to Kerguelen Islands was performing well right up until the days immediately prior to death (Figures 7 and 8). Here, an abrupt behavioral change occurred, possibly supportive of a predation event (which juvenile pinnipeds may be particularly vulnerable to; Horning & Mellish, 2012

| Dive and forage ability
Individuals that experienced early mortality appeared to exhibit slower development in dive capability compared to surviving pups.
A slowing in the rate at which dive durations increased through time was linked to reduced survival, as was an increase in swimming effort through time (i.e., individuals are using more energy to do less). This pattern appeared most prominent for pups 140067, 140074, 140076, and 140078. Temporal improvements in dive capability (e.g., duration) are a key component of diving marine predator ontogeny, as juveniles begin life with abilities that are significantly reduced compared to adults (Burns, 1999;Carter et al., 2017;Orgeret et al., 2019Orgeret et al., , 2016. Our observations support this F I G U R E 8 Drift rates of non-survivors (colors correspond to individuals), with times of death (and corresponding drift rates) indicated by black stars. The large light gray band represents the 2.5%-97.5% quantiles of the survival dataset, the nested dark gray band the 25%-75% quantiles, and the dark gray line the daily median F I G U R E 9 Sea surface temperatures (SSTs) of survivors and non-survivors. Plot (a), time series for six individuals (140061, 140064, 140065, 140074, 140076, and 140078) that appear to visit waters of reduced SSTs. Here, the large light gray band represents the 2.5%-97.5% quantiles of the survival dataset, the nested dark gray band the 25%-75% quantiles, and the dark gray line the daily median.

TA B L E 3 (Continued)
and suggest that failure to develop in a timely manner increases mortality risk.
In addition, non-surviving individuals moved faster (horizontally) than survivors, which may betray reduced foraging ability.
Upon location of profitable foraging grounds, animals tend to reduce travel speeds and increase turning angles, so as to focus efforts in areas where prey are more likely to be encountered (Fauchald & Tveraa, 2003). PrCA rates increased when distances traveled decreased (Figure 10) Figure 10), which suggests that even if an individual manages to increase its PrCA rate after a period when it is low, it cannot "catch up." We also found individuals that reduced their average PrCA rates through time had a better survival outcome than those that did not, or did so to a lesser extent. This was not generally accompanied by further decreases in buoyancy/ drift rate, and in some cases coincided with a positive change in buoyancy/drift rate (Figure 8-from ~day 50 onward). Such a trend supports a change in diet composition and/or improvement of forage ability. For example, in contrast to older individuals that feed predominantly on fish and squid (Slip, 1995), juvenile southern elephant seals have a high proportion of crustaceans (e.g., krill) in their diet (Lubcker et al., 2017;Walters et al., 2014), while the size of other preys (e.g., squid) is generally smaller (Field, Bradshaw, Hoff, Burton, & Hindell, 2007;Slip, 1995). As foraging skills improve, diet composition may change as individuals are able to expand/select/switch to other, possibly more nutrient-rich prey sizes and/or types, thus requiring less prey items to be caught, resulting in a decrease in PrCA rate (Chaigne, Authier, Richard, Cherel, & Guinet, 2013;Field et al., 2007;Walters et al., 2014). Additionally/ alternatively, initial high PrCA rates may include a larger number of unsuccessful attempts, which would decrease in frequency as individuals hone their capture techniques.

| Encountered environmental conditions
Non-surviving juvenile southern elephant seals from Kerguelen Islands encountered waters of lower temperature and reduced TKE compared to the surviving group. Here, PrCA rates were typically reduced ( Figure 10 and Supporting Information S.12). While confounding processes (e.g., temporal changes in foraging success and/or prey quality) suggest increased PrCA rates are not necessarily directly linked to better foraging (Orgeret et al., 2019;Richard et al., 2016), combined with increased survival rates these patterns suggest foraging habitats associated with increased SSTs and TKE estimates are more suitable for southern elephant seals during this time in their development.
Across the spatial domain visited by our tracked southern elephant seal pups, areas of higher SST and increased TKE correspond roughly to the southern edge of the polar front, where dynamic habitats associated with frontal zones and eddy edges are more likely to be found (Bost et al., 2009;Supporting Information S.13). Both juvenile and adult southern elephant seals are known to forage within such regions (Cotté, d'Ovidio, Dragon, Guinet, & Lévy, 2015;Dragon, Monestiez, Bar-Hen, & Guinet, 2010;Field, Bradshaw, Burton, & Hindell, 2004;Tosh et al., 2012), where increased prey availability and abundance can be found (Abrahms et al., 2018;Bost et al., 2009;Cotté et al., 2015). These habitats may thus be particularly important during early life, when juveniles need to compensate reduced dive and forage ability (Burns, 1999). Our results support this and suggest that failure to locate such regions increases mortality risk.
All but two pups (one survivor and one non-survivor) departed Kerguelen in a southeastward direction, and broad-scale spatial distributions between the surviving and non-surviving groups were similar with no evidence of distinct segregation, further highlighting the importance of fine-scale habitat use within a larger common area. Whether the successful location of such features is down to chance or driven by individual innate knowledge and/or other cues is unknown.

| CON CLUS IONS
The first few months following independence represent a particularly critical time in the life cycle of long-lived marine predators such as the southern elephant seal. Results from this study both support previous assertions that juvenile mortality rates during this period are high and additionally provide direct evidence that this is driven, at least partly, by insufficient development of dive and forage ability. Furthermore, the habitat type encountered, as depicted by SST and TKE, appears to be important to survival outcome as does predation. In all instances, chance seems to play a role, and individuals need only deviate slightly from the norm to increase their mortality risk. As such, survival rates in juvenile southern elephant seals (and possibly other long-lived marine predators) may be particularly vulnerable to even small changes in climate and environment, which will have concomitant consequences on demography and population dynamics. and was supported by the IPEV program 109 (PI Dr Henri Weimerskirsch).

CO N FLI C T O F I NTE R E S T
None declared.