Local weather and body condition influence habitat use and movements on land of molting female southern elephant seals (Mirounga leonina)

Abstract Southern elephant seals (Mirounga leonina) are known to move and aggregate while molting, but little is known about their behavior on land during this time. In this study, 60 adult females were monitored (23 with GPS tags) during four molting seasons, between 2012 and 2016 at Kerguelen Archipelago, Indian Ocean. Population surveys were recorded each year (N = 230 daily counts), and habitat use was analyzed in relation to the stage of the molt and local weather. Based on stage of molt, habitat use, and movements on land, we classified the molt of elephant seals into three phases: (1) a “search phase” at the initial stage of molt when grass and wallow habitats were used and characterized by greater mean distances travelled on land per day compared with the two other phases; (2) a “resident phase”: during initial and mid‐stage of molt when animals were found in grass and wallow habitats but with less distance moved on land; and (3) a “termination phase” at the final stage of molt where grass and beach habitats were occupied with no change in distances. Windchill and solar radiation influenced individual distances moved per day (mean 590 ± 237.0 m) at the mid‐ and final stage of molt such that animals travelled greater distances on days of low windchill or high solar radiation. Individual variation in distance moved and relative habitat use were also linked to body mass index (BMI) at arrival on the colony, as females with higher BMI moved less and preferred beach habitat. Moreover, the individual rate of molt increased with the use of wallows. Aggregation rate tended to be negatively correlated with distances moved. We therefore suggest that individuals face an energetic trade‐off while molting, balancing energy expenditure between movement and thermoregulation.


| INTRODUC TI ON
Pinnipeds are highly adapted for an aquatic existence, but most species spend several weeks ashore each year for reproduction and molting (Le Boeuf & Laws, 1994;Williams & Worhty, 2002). Indeed, seasonal molting is characterized by hair renewal and results in an increase in haul-out frequency and duration (Ling, 1970). Molt duration varies among phocid seals as some species (e.g., elephant seals and monk seals) experience a "catastrophic" molt, replacing not only their hair but also the top layer of their epidermis (Parsons, Bauer, McCafferty, Simmonds, & Wright, 2013;Yochem & Stewart, 2009). Furthermore, these species are known to fast during the molt when ashore for several weeks (Kenyon & Rice, 1959;Ling, 1968).
Molting is a highly energy demanding phase as adult female elephant seals experience on average a mass loss of 5 kg/day (Boyd, Arnbom, & Fedak, 1993;Carlini, Marquez, Daneri, & Poljak, 1999;Hindell, Slip, & Burton, 1994). As they are fasting on land, elephant seals rely on their body reserves in the form of blubber (Boyd et al., 1993;Liwanag, Berta, Costa, Budge, & Williams, 2012;Worthy, Morris, Costa, & Le Boeuf, 1992). Moreover, blubber also has an important role as insulation against cold (White & Odell, 1971;Whittow, 1987). Molting elephant seals are exposed to different environmental constraints in the colony, resulting in heat exchange with their surroundings (Hind & Gurney, 1998;Schmidt-Nielsen, 1997). In particular, when seals haul-out, they use thermal windows on their trunk and flippers to dissipate heat (i.e., heat dissipation; Mauck, Bilgmann, Jones, Eysel, & Dehnhardt, 2003). During the molt, increased perfusion of peripheral tissues promotes hair growth and renewal of the epidermis (Ashwell-Erickson, Fay, Elsner, & Wartzok, 1986). This vasodilation bypasses the insulating blubber layer, increasing heat loss. Paterson et al. (2012) showed that heat losses of harbor seals vary when hauled out and reach a maximum corresponding to the peak stage of the molt. In order to facilitate rapid hair growth, elephant seals must maintain a relatively high peripheral temperature through perfusion of the skin surface and avoid vasoconstriction in response to environmental conditions. However, physiological mechanisms (such as vasomotor control of heat flow across the skin) can be supported by behavioral strategies to maintain thermal equilibrium (Heath, McGinnis, & Alcorn, 1976;White & Odell, 1971). Any energy-saving strategy used to minimize heat loss during molting would reduce the depletion of energy reserves in blubber, with possible fitness consequences. Indeed, previous studies on pinnipeds have shown that reproductive success is linked with higher body mass or better body condition in females (South African fur seals Arctocephalus pusillus: Guinet, Roux, Bonnet, & Mison, 1998; northern elephant seal Mirounga angustirostris : Crocker, Williams, Costa, & Le Boeuf, 2001 : Redman, Pomeroy, & Twiss, 2001). In phocids, some studies found a correlation between local weather conditions and haul-out behavior in polar or temperate environments (harbor seal Phoca vitulina: Pauli & Terhune, 1987;Watts, 1992;Grellier, Thompson, & Corpe, 1996). Mainly solar radiation, air temperature, and wind speed influenced haul-out numbers, while in other studies, there was no consistent effect of meteorological parameters. In elephant seals, behavior on land and physical activity was linked to weather conditions (northern elephant seal: Norris, Houser, & Crocker, 2010; southern elephant seal Mirounga leonina: Cruwys & Davis, 1994, 1995. In general, during the molt, elephant seals often aggregate in wallows (mud pools), when the nature of the colony's substrate allows their formation (Boyd et al., 1993;Carlini et al., 1999). Boyd et al. (1993) described the molt in two phases, with a "wallow phase" while elephant seals lose their old fur and skin, and a "shore phase" closer to the sea, when individuals have lost their old skin and are renewing their hair. However, movement patterns, distances travelled, and habitat use on land during the molt have not yet been extensively studied in this species.
The aim of this study was to investigate how the behavior of southern elephant seals on land may be driven by molt stage, environmental conditions, and body reserves at their arrival on land.
Specifically, we hypothesize that movements and habitat selection are influenced by stage of molt and local weather at a group and individual level, and by body condition (i.e., BMI: body mass index) at an individual level. We propose that different types of habitat (based on substrate type: grass, wallows, beach) offer different thermal or physical properties and therefore influence habitat use. In particular, we expect that for elephant seals, the peak of the molt would be characterized by the use of wallows and limited movement. In comparison, we hypothesize greater movements when weather is poor or for individuals with lower BMI.

| Study site
Data were collected during the molting season of adult females ( Figure 1) from end of December to early March (Table 1)  The site is composed of a coast of stones (basalt; Nicolaysen, Frey, Hodges, Weis, & Giret, 2000), sometimes covered with seaweeds (Durvillaea antarctica; Lawrence, 1986), overhung by hills of grass (mainly endemic Azorella selago and Acaena magellanica, with invasive Taraxacum officinale; Chapuis, Frenot, & Lebouvier, 2004) where wallows (mud pools without vegetation) are formed from seal aggregations. We divided the study site into three habitats based on substrate type: "beach," "grass," and "wallows." Based on the hypothesis that wallows are a specific habitat for molting, habitats "grass" and "beach" were grouped together to distinguish between "wallow" and "nonwallow" habitats for further analysis of habitat use.

| Population surveys
To study habitat use of females during the molt, two line transects on "nonwallow" habitats (136 × 10 m; T1: on habitat "grass" and T2: on habitat "beach") and one quadrat Q on "wallows" (200 × 200 m, Figure 2b) were selected and scanned daily (Table 1). During each scan, the total number of adult female elephant seals present was recorded, along with their molt stage assessed by the percentage (± 10%) of old hair surface shed (0%: no old hair shed to 100%: all old hair shed). We defined three stages of molt: "initial stage" (0-40% of old hair shed; still largely covered with old hair), "mid-stage" (50-80%; mainly bare skin exposed to air with old hair shed but new hair still ungrown), and "final stage" (90-100%; new hair growing).
From 2014, a second quadrat Q′ was added (500 × 500 m) which included all three habitats ( Figure 2b) to count the number of elephant seals present daily on each habitat type (Table 1)

| Individual data collection
Females were captured (Table 1) and anaesthetized (Chaise et al., 2017;McMahon, Burton, Slip, McLean, & Bester, 2000) on the colony at initial stage of molt. They were recaptured at the final stage of molt, when all old hair was shed.
An automatic weather station (MiniMet, Skye Instruments Ltd,

| GPS locations
CatTracks were set up to record latitude and longitude every 30 min.
Distance travelled (m) and speed (m/h) were calculated from GPS coordinates and time (GMT) of recorded positions. Mean distances per day were calculated excluding data from the day of capture and the day of recapture, to remove bias from anesthesia and capture.  to be due to position errors (i.e., 1.1% of data). Moreover, for movements with speed > 1000 m/h, GPS coordinates of departure and arrival were mapped (ArcGIS ® 10, ESRI) to check whether the track was associated with travel at sea. Hence, 0.7% of tracks were considered to be at sea movements and 98.2% of data were considered to be land movements.

| Weather data analysis
Given that weather data comprised multiple potentially colinear variables, we used a centered-scale principal component analysis (PCA), to account for variation in meteorological parameters (air temperature, relative humidity, solar radiation, and wind speed). To examine population differences in habitat use, we used meteorological data from Q′ counts (N = 36, Q′ counts with complete meteorological records). The first component (PC1) accounted for 45% of the variation and the second component (PC2) accounted for 31% of the variation. We retained "day" as a coordinate on the first two axes of the PCA. PC1 received a major positive loading from solar radiation and a major negative loading from relative humidity. PC2 received a major positive loading from air temperature and a major negative loading from wind speed (Supporting Information). The components were then transformed to binary factors, corresponding to a weather index, based on their respective median values. We distinguished days of "bad weather" (for days with components value < components median value; high relative humidity, low solar radiation, low air temperature, and high wind speed) from days of "good weather" (for days with components value > components median value; low relative humidity, high solar radiation, high air temperature, and low wind speed).
To examine individual differences in distance moved during the molt, we selected meteorological parameters from the automatic weather station (air temperature, relative humidity, wind speed, solar radiation, and precipitation    . Therefore, based on our observations, we defined early molting season from end of December to mid-January (25 December to 11 January), the peak of the molting season between mid-January and early February (23 January to 1 February) and late molting season from mid to end of February (11 February to 28 February).

| Density and habitat selection of elephant seals
Counts of elephant seals in the quadrat Q′ allowed us to compare density between different habitat types. In quadrat Q′, use of wallow and nonwallow habitats varied depending on the season.
In wallows, there were more seals during the early season (estimate ± SD = 2.34 ± 0.57, z = 4.133, p < .0001) and during the peak of the season (estimate ± SD = 2.53 ± 0.34, z = 7.516, p < .0001; Variation in habitat use (density per habitat in Q′) was influenced by weather and the molt season (early, peak, and late). There was no relationship between wallow selection and weather index (estimate ± SD = 0.03 ± 0.58, z = −0.05, p = .96). In contrast, nonwallow habitat selection was related to weather index ( Figure 5). The model indicated that elephant seals were relatively more abundant on grass during "good weather" days and more abundant on beach habitat during "bad weather" days (estimate ± SD = 2.00 ± 0.94, z = 2.124, p = .03). The model that best explained nonwallow habitat selection retained weather index and the molt season as explanatory variables. This indicated that elephant seals tended to select grass over beach during "good weather" days in the early season (estimate ± SD = 2.73 ± 1.52, z = 1.79, p = .07). The model estimated that grass was selected in early season (estimate ± SD = 3.74 ± 1.85, z = 2.03, p = .04) and during the peak of the season (estimate ± SD = 3.93 ± 1.19, z = 3.30, p = .001) and beach was selected during the late season with no influence of weather ( Figure 5). The model also showed a significant but weak effect of solar radiation on distances depending on molt stage: distances per day at the final stage increased when solar radiation was high (estimate ± SD = 0.0007 ± 0.0003, z = 2.55, p = .01).

| Movements during the molt
We found no correlation between distance moved by females per day with body mass loss (kg/day) during the molt (nonparametric Spearman's rank correlation: N = 19, r = −.04, p = .88).
Mean distances moved tended to be negatively correlated with initial mass at capture but this was nonsignificant (nonparametric

| Habitat selection and movements during the molt
The selection of wallows by females during initial and mid-stage of molt, independent of local weather conditions, suggests that wallow use is strongly linked to the molting process, as previously noted (Boyd et al., 1993;Laws, 1956). During initial and mid-stages of the molt, elephant seals are most prone to heat loss due to blood perfused bare skin exposed to ambient air (Paterson et al., 2012). We found a positive correlation between the rate of molt and the use of wallows. Individuals seen more often in wallows shed skin/fur more rapidly each day. We suggest two hypotheses to explain the apparent advantage of molting in wallows. Firstly, individuals would have an accelerated loss of old skin and hair due to the mechanical action of body scratching, as elephant seals move against each other in the middle of aggregations (Cruwys & Davis, 1995;Laws, 1956). Secondly, we suggest that wallows where seals closely aggregate would offer a favorable microclimate when exposed to solar radiation. Indeed, temperature measurements indicate that wallows are 4.5°C warmer than air (Chaise et al., in preparation) . This microclimate in a cold sub-Antarctic climate is favorable to molting: Firstly, because it minimizes heat loss (i.e., energy savings by local microclimate warming; Gilbert et al., 2010) and secondly it may allow the skin to reach an optimal temperature for epidermal cell growth (Feltz & Fay, 1966). Further analyses may confirm these hypotheses (Chaise et al., in preparation). We showed that wallow use was not influenced by weather conditions. This could be due to the fact that wallows represent an optimum but limited resource (Setsaas, Bester, Van Niekerk, Roux, & Hofmeyr, 2008), perhaps due to the topography of the colony site. Hence, skin removal would be the primary driver for wallow use and the thermal advantage of aggregation would be a secondary consequence of this habitat selection.
The two approaches used in our study, animal density counts per habitat and individual molt stage observations along transects, show that the visible molt process may be described in three phases, linked to different habitats. At the initial stage of molt, there appears to be a "search phase" for appropriate wallows to molt, which results in an increase in movements, as found in our study. Hence, at initial and mid-stages of molt (i.e., early season and peak season), wallows and grass are preferentially used. At the final molting stage (i.e., late were found mainly on the beach and on grass. We suggest that grass is a transition habitat, while seals migrate from beach sites to wallows after their arrival on land and migrate from wallows to beach at the final stage of their molt. We could also suppose that tidal cycles may affect the availability of beach habitat. Further study on movements of elephant seals linked to the variations of the foreshore area during the molt could explore this relationship. Movements were reduced during the mid and final stages of the molt. A decrease in physical activity (i.e., movements) during the main part of the molt would result in energy savings.
Our study is the first to investigate fine-scale movements of female elephant seals while molting. We showed that distance covered is around 362-1406 m/day. Therefore, elephant seals move more while molting on land than previously thought (Boyd et al., 1993;Laws, 1956). Moreover, we recorded individuals going to sea in 2012 and 2014. These movements occurred at the initial and final stages of their molt and more than 24 h after the day of capture. Our results suggest a more variable pattern of behavior of elephant seals while molting than previously described (Boyd et al., 1993;Setsaas et al., 2008).

| Influence of weather conditions
Weather conditions influenced habitat selection for nonwallow habitats (grass and beach). Several studies have described seals entering water during hot days in order to cool while on land, mainly during the breeding season (Codde et al., 2016;White & Odell, 1971;Whittow, 1987). In contrast, our results suggest that, during the molt, seals select grass over beach during "good weather" days. This difference could be due to the fact that the humid surface of the grass represented a cooler environment during warm and sunny days compared to the rocky shore (Chaise et al., in preparation). Moreover, habitat selection of elephant seals may be linked to haul-out topography.
Indeed, Cruwys and Davis (1994) described that only wallows which offered wind protection were selected by elephant seals. At the Pointe Suzanne study site, grass habitat is situated on hills, separated from beach by a few metres of steep gradient (cliff). Therefore, beach selection during "bad weather" days may depend on beach orientation and wind direction (Laws, 1956): Beach could offer a better shelter than the exposed grass surface. Further studies could confirm these findings. The selection of grass during the peak of the season, at mid-stage of molt, could be related to access to wallows.
The selection of beach in late season at the final stage of molt is likely due to their imminent departure to sea (Boyd et al., 1993) or a need to access sea water to drink (Redman et al., 2001).
Our results may suggest that molting elephant seals increased their movements in response to temperature sensation (i.e., windchill), in order to change habitat, as different habitats (topography, substrate) offer different thermal environments (Cruwys & Davis, 1995;Twiss et al., 2002), or to join an aggregation if they were not aggregated before (Cruwys & Davis, 1994). Indeed, we also found a tendency for elephant seals to decrease their distances moved per day when they were aggregated in groups. Our model also estimated an increase in individual distances moved each day when solar radiation was high at the final stage of molt. Despite effects of windchill, high solar radiation may therefore have influenced thermal comfort of elephant seals, leading animals to seek water for thermoregulation (Codde et al., 2016;Cruwys & Davis, 1995;White & Odell, 1971).
The fact that distances moved on land were influenced by weather conditions only for elephant seals at mid-to final stages of their molt may be due to the fact that location of suitable habitat for molting may have dominated movement patterns, rather than weather constraints. Alternatively, an overall decrease in insulation from the loss of old fur and decrease in blubber thickness may have made animals more sensitive to weather variations (White & Odell, 1971;Worthy et al., 1992).

| Body condition and individual strategies
Individual movements on land during the molt approximated 600 m/ day, a greater level of activity than previously noted for molting females (Boyd et al., 1993;Laws, 1956). Movement on land is energetically costly (Twiss, Caudron, Pomeroy, Thomas, & Mills, 2000), but we did not find evidence of relatively greater mass loss associated with increased movements.
Elephant seals may adapt their behavior to balance their energetic expenditure (skin and fur renewal, movements on land, thermoregulatory costs), resulting in individual strategies. For example, larger females have relatively less thermoregulatory costs than smaller females (with a higher surface-area-to-volume ratio; Schmidt-Nielsen, 1997), but energy expenditure may be greater when moving (Schmidt-Nielsen, 1997). During the molt, elephant seals may maximize energy savings either by decreasing physical activity but with less chance of changing habitat to reduce heat loss, or decreasing heat loss by moving habitat depending on weather conditions. Indeed, females with higher BMI were less active (negative correlation between initial body mass and initial BMI, and mean distances moved per day). Based on these results, we suggest that during the molt females with a higher BMI at arrival (i.e., greater thermal inertia and lower surface area:volume ratio) would be less sensitive to environmental conditions, while a female with a lower BMI at arrival may move from site to site when environmental conditions deteriorate in order to minimize cost of thermoregulation. Initial BMI was also inversely correlated with beach and grass habitat selection. Indeed, larger females or females in better condition tended to arrive later in the molting season (early and peak season: initial body mass = 305 ± 42.7 kg and BMIi = 57.6 ± 4.5 kg/m²; late season: initial body mass = 357 ± 73.9 kg and BMIi = 63.1 ± 6.2 kg/m²; Chaise et al. unpublished). In late season, wallows are emptied as most elephant seals at the final stage of molt selected beach sites. Laws (1956) suggested that molt duration in elephant seals appeared to increase with body size. This could be linked to the negative correlation between molt rate and beach use: Individual females spending more time on beach had slower rate of molt. Females that arrive late in the season, when wallows are emptied, would molt mostly on beach sites and benefit less from optimal molting habitat in wallows. These females would hence show an increase in molting duration (showing a slower molting rate), with concomitant higher fasting duration.
In conclusion, movements and habitat use of female elephant seals during their molt appear to show greater variability than previously documented. Aggregative behavior and habitat selection may be linked to energetic constraints faced while molting on land. Further studies that investigate energy expenditure and behavior during the molt may reveal greater detail on individual strategies of female elephant seals during this relatively understudied stage of their lifecycle.

ACK N OWLED G EM ENTS
The present research project was supported by the Institut Polaire

CO N FLI C T O F I NTE R E S T
The authors declare that they have no conflict of interest.

AUTH O R CO NTR I B UTI O N S
CG, AA, DM, SG, and LC conceived the ideas and designed methodology; CT, SG, LC, and WP collected the data; CT, LC, and IP analyzed the data; LC, CG, DM, AA, and MT led the writing of the manuscript.

DATA ACCE SS I B I LIT Y
No database from the manuscript has been made publicly available yet. Data are available upon request.

DATA A RCH I V I N G
We intend to archive data from the manuscript in Dryad Digital Repository: https://doi.org/10.5061/dryad.5ds04gg