• Steller sea lion;
  • Eumetopias jubatus;
  • neonatal survival


  1. Top of page
  2. Methods
  3. Results
  4. Discussion
  5. Acknowledgments
  6. Literature Cited

Neonatal survival of Steller sea lions (Eumetopias jubatus) are often considered inconsequential to their population dynamics. However, observations of dead animals on rookeries and in surrounding waters suggest that early mortality is not uncommon. This study used the natural markings of adult females in a mark and resighting framework to estimate the apparent survival (φ) of pups with the Cormack–Jolly–Seber model at two sites on Lowrie Island, Alaska from birth to 3 wk old. Estimates varied greatly by site and year; 2002 Area 5: inline image (95% CI: 0.199, 0.684; n= 21), 2002 Area 1: inline image (0.437, 0.916; n= 21), 2003 Area 5: inline image (0.414, 0.738; n= 56), and 2003 Area 1: inline image (0.695, 0.997; n= 32). The mean estimate across the four area × year combinations was inline image (0.569, 0.772). Survival was lowest on the first day of life and then leveled off at a higher rate. None of the four environmental covariates we considered (swell height, interaction of tide and swell heights, density, or birth date) were significantly related to neonatal survival. Our results suggest that estimates of first-year survival that do not account for mortality prior to dispersal from the natal rookery may significantly overestimate survival rate.

Steller sea lions (SSLs; Eumetopias jubatus) range from the coast of southern California, north through the Gulf of Alaska (GOA) and the Aleutian Islands to the Bering straight south around the Pacific Rim to northern Japan (King 1983, Loughlin et al. 1984). Kenyon and Rice (1961) estimated the total population to be between 240,000 and 300,000 in 1960. More recently, Trites and Larkin (1996) calculated from survey data that between the mid-1950s to the mid-1970s there were between 250,000 and 282,000 SSLs (including pups) in the GOA alone. By the early 1990s, however, the number of SSLs in the GOA had decreased by over 80% to <30,000 animals. Currently the GOA population is estimated around 38,000 animals (Angliss and Outlaw 2005).

Stocks of SSLs are divided based on phylogeographic information for management purposes into eastern and western populations at Cape Suckling (144 W) (Loughlin 1997). Telemetry and mark and recapture field studies also supported these stock designations based on high breeding-site fidelity and, except for males aged 1–4 yr, little movement between stocks (Calkins and Pitcher 1982, Raum-Suryan et al. 2002). However, there is recent evidence that the Forrester Island rookery complex, where this study occurred, has been at least in part established by females from the western stock as evidenced by mitochondrial DNA analysis (O'Corry-Crowe et al. 2005). The decline of the “endangered” (Angliss and Outlaw 2005) western population of SSLs over the last few decades has resulted in fishery restrictions that affected rural economies. The threatened eastern population has increased at an average annual rate of 3.1%/yr since the 1970s (Pitcher et al. 2007). Estimates of age-specific survival and reproductive probabilities from the 1970s (Calkins and Pitcher 1982) and 1980s (Calkins and Goodwin 1988) were based on necropsies of collected wild animals that are no longer practicable. Beginning in 1987, hot-branding of SSL neonates and resightings of branded animals by the National Marine Fisheries Service (NMFS) and the Alaska Department of Fish and Game (ADF&G) have provided estimates of age-specific vital rates through mark–recapture studies (Pendleton et al. 2006).

Several demographic analyses of SSL population dynamics have been conducted (York 1994, York et al. 1996, Gerber and VanBlaricom 2001, Holmes and York 2003, Gerber 2004, Eberhardt et al. 2005). All of these studies have used estimates of demographic rates from Calkins and Pitcher (1982). From these studies, it is apparent that the demographic factors affecting SSL population growth have changed over time. For example, declines in the 1980s were probably associated with severely low juvenile survivorship whereas declines in the 1990s appeared to be associated with disproportionately low fecundity (Holmes and York 2003). Even though survival during the first year of life is generally believed to have minimal effect on the population dynamics of long-lived marine mammals (Eberhardt 2002), a vital rate that is naturally variable and able to be affected by management may prove be important to a species' conservation (Noon and Biles 1990, Fefferman and Reed 2006). The survival component of SSLs representing the period from birth to dispersal to sea has not been rigorously studied.

From observations of dead pups on rookery study sites, and in the waters surrounding these areas, it is generally accepted that some level of neonatal mortality takes place before branding. Branding operations are timed for when the majority of females have given birth but before mother and pup pairs emigrate from their natal sites. By this time most pups are about 3 wk old. No quantitative studies have been conducted to estimate neonatal survival in this pre-branding period, nor have there been any attempts to quantify the causes of mortality.

The main objective of this research was to estimate neonatal survival rate from birth to 3 wk of age, the time at which branding of newborns takes place. Observations of mortalities and necropsies of dead pups found on rookeries have generally attributed the cause of death to starvation, drowning, or blunt trauma (Orr and Poulter 1967, Gentry 1970, Sandegren 1970, 1976). In addition to overall estimates of survival, we examined how these estimates are affected by several factors.


  1. Top of page
  2. Methods
  3. Results
  4. Discussion
  5. Acknowledgments
  6. Literature Cited

Our study took place from May to August 2002 and 2003, on Lowrie Island (54°53′N, 133°30′) a part of the Forrester Island Complex in southeast Alaska. Included in the Alaska Maritime National Wildlife Refuge, this island complex is the site of the largest SSL rookery in southeast Alaska. The island is heavily wooded with a rocky coastline. Intensive land-based research involving SSLs at Lowrie Island has been ongoing since 1992. ADF&G researchers sectioned the island and adjacent offshore sites into 14 different “areas” (also referred to here as “sites”) for conducting SSL age-composition counts in 1991.1

The two main study areas, named by ADF&G, “Area 1” and “Area 5,” were selected because good vantage points existed above the animals and because of their historically large concentrations of breeding animals. Covered blinds allowed observation of the animals without disturbance.

Marking Technique

Mark–resighting survival studies require that a portion of the observed population be marked. Direct capture and marking of newborn pups was infeasible because of the potential for disturbance to the rookery resulting in injury or separation of pups from their mothers. We instead used the unique scars, fungal patches, and other natural markings of individual focal females (Altmann 1974) as a proxy marker to follow the fate of their pups. The use of natural markings has been used successfully by other SSL researchers to follow the fate of adult animals (Gentry 1970, Sandegren 1970, Lewis 1987, Mamaev 1998, Millette and Trites 2003; see Fig. 1). Previously branded adult females were also used opportunistically.


Figure 1. An example of the natural markings used to follow Steller sea lion focal females and their associated pups on Lowrie Island in southeast Alaska in 2002 and 2003. A focal female with dry hair just seconds after giving birth in 2002 (top left), then 5 d later with wet hair (top right), and again in 2003 with dry hair (bottom left) and wet hair (bottom right).

Download figure to PowerPoint

Because litter size is almost always one and there is high spatial fidelity between mother and pup, we were able to use the individual marks of the female as a surrogate marker for its pup and use mark–resighting methods to estimate pup survival rates (see below). A similar approach has recently been used to estimate the proportion of females breeding in manatee populations using unique marks on females with young (Kendall et al. 2003). We felt confident in our use of a proxy marker because females have many unique body marks, fostering behavior is rare in SSLs, and adoptions are probably cases of mistaken identity (Reidman 1990). Females are generally aggressive toward any pup other than their own, occasionally biting or tossing pups that approach too close (Gentry 1970, personal observations, Sandegren 1970).

Using digital cameras with 70–200-mm zoom telephoto lenses, sometimes with 2× doublers or connected to a telescope, focal females were photographed from all available angles upon initial identification. In combination with photographic computer software and data storage, we used these photographs to identify and catalogue a large number of females. Field photos were enhanced by adjusting brightness and contrast to bring out markings, then printed (Hewlett Packard 930 c Deskjet, Palo Alto, CA, USA) onto heavy bond, bright white paper and cataloged along with sketches of each focal female to facilitate positive identification of each later resighting. For additional confirmation of sightings, we developed “best photos” folders for each female.

In 2002, rookery searches for focal females were conducted opportunistically in no systematic order. However, some individuals had natural markings that were more easily seen than others. Also, individuals could sometimes be found at different locations within the rookery on different days and sometimes changed locations within a day. Opportunistic searches may have introduced heterogeneity in resighting probabilities among individuals (Seber 1982) improving resighting probabilities for very well marked and more site-fidelic females over less well marked or site-fidelic females. To reduce this heterogeneity, in 2003, we equally distributed search effort within each area by dividing areas into numbered grid cells (Area 1 = 10 cells, Area 5 = 36 cells, and average cell size about 35 m2, based on natural rock features) and randomly ordered grid cells for searching on each occasion.

Each of the two sites was observed between 0800 and 1800, 6 d/wk by one to two observers, for approximately 5 h/d. One day each week we surveyed the entire perimeter of the island for females that may have emigrated out of the study area to determine the fates of their pups. We also searched beaches for dead pups to necropsy. During the all-island survey day, Areas 1 and 5 were searched opportunistically (both years) and only one sighting of each female and their status was recorded on that day.

Typically two 2-person teams worked at either Area 1 or 5 each day. Upon arriving at the study sites each morning, teams would note new females that had arrived overnight and determine if they were good candidates for cataloging. Females that were good candidates were those who possessed a minimum of three different natural marks (scars, wounds, fungal patches, etc.) that could serve as individual identifiers (see Fig. 1). Daily attempts were made to resight females that were “marked” and cataloged to see if they gave birth to a pup. All focal females were followed for a period of time (as noted below) upon each resighting. To estimate neonatal survival, females that gave birth were closely monitored to provide daily sighting information on the fate of their pup.

Resighting Methods

Upon resighting a focal female, a series of three verifications were recorded, one for each mark noted (see Fig. 1). Females were photographed daily and compared to reference photos for positive confirmation and archival purposes. Females that were not confirmed through observations of all three primary marks on a given day were closely examined on subsequent days and often identified by comparison with the daily archival photos. Some of these identifications were based on secondary identifying marks that could be retrieved from the photo archive.

After a focal female was located, we recorded the location of its pup and began a 30-min watch divided into three 10-min segments to confirm the status of the pup (Table 1). We watched each focal female for a few minutes then moved on to identify other focal females, checking back periodically and at the end of the 10-min segment to watch again to assess pup status. The purpose of these watches was to confirm the presence or absence of a female's pup, how closely the mother was associated with the pup, and whether the pup was still alive. A “with pup” status (Table 1) was assigned only if it could be determined that the pup in question was associated with its mother. Beginning at the time of birth, these observations formed the encounter history for each pup and generated the data needed to estimate daily survival rates.

Table 1.  Status of mother/pup pair for resighting observations.
NPFemale is sighted without a pup. No pup is nearby or suckling, touching, closely and consistently interacting with the female.
NPOFemale is sighted without a pup, but it may be nearby. The pup is not available to see due to the female's location or her pup is being hidden by a rock or other object.
WPFemale is with her pup. Female and pup form a distinct group. Pup is touching and consistently interacting with the female.
WP/ZSame as WP, but female and pup are sleeping during one or more observations.
SPup is suckling from female.
S?Female is definitely with her pup and it appears to be suckling, but due to an obstructed view the observer cannot be certain her pup is suckling.

In 2003, we changed our search methods somewhat and added 5-min scans of the rookery every 0.5 h. The purpose of these scans was to identify new females for marking, locate focal females that were overlooked during the grid search, and to note the status of females whose observation periods were finished, but which were not seen with a pup.

Age Determination of Neonates

Ages of newborn pups were determined by either directly observing the birth or noting “marked” females that previously had no pup and were then subsequently sighted with a pup at a later time (usually the next day). Upon resighting a focal female with a new pup every effort was made to confirm the age of the pup from other cues available at the scene, such as a fresh placenta or nearby blood on the rookery. Cues of young pup behavior such as uncoordinated movement, difficulty keeping the head up and a mother's urging of a pup to feed by dragging it toward her nipples were also taken into account (Sandegren 1970). Early in the pupping season this was much easier to observe since there were few mother–pup pairs on the rookery. Later in the season it was more difficult to identify individual pairs and greater vigilance was required. Only females that could be accurately assessed to have given birth within the last 48 h were included in the analysis. We did not believe that presence of an umbilical cord alone was sufficiently accurate to assign a birth date even though this method has been used by other researchers (e.g., Trites 1993, Merrick et al. 1995). We conducted a pilot study that revealed extensive variability in the number of days an umbilical cord stayed attached. Weather conditions, particularly rain or hot, dry weather, affected the duration of umbilical attachment by increasing or decreasing the attachment time, respectively.

Environmental Covariates

To investigate the possible relationship between environmental factors and the daily survival rates of the pups, we recorded information on several covariates (Table 2).

Table 2.  Covariates evaluated for their effects on neonatal survival until 21 d of age.
AreaDenotes groups of animals compared at the two sites on Lowry Island, Area 1 (north) and Area 5 (west)
Area 1Northern site on Lowry Island
Area 5Western site on Lowry Island
Day 1The first day of life
SwellHeight of waves above sea level
Swell2Swell squared
SurgeTide height multiplied by swell height
Surge2Tide*Swell interaction squared or (Tide*Swell)2
DensDensity of animals at each site
Den2Density squared
BdateDate of birth between 21 May and 15 June
Bdate2Date of birth squared
AgeAge of animal from birth to 21 d old. Since the encounter histories represent time as a factor of the daily aging of the pup, Age is congruous with time in typical Program MARK applications.
Age2Age squared

Swell—Storms are believed to be a leading cause of neonatal mortality (Gentry 1970, Sandegren 1970, Mate 1973, York 1994), though we are unaware of any quantitative measurements to establish causation. We measured swell height relative to the location of pups on the rookery using sea level as the baseline. These heights were recorded daily into the following five categories: 0 < 4 ft, 1 > 4 ft, 2 > 6 ft, 3 > 8 ft, and 4 > 10 ft.

Tide—High tides, in isolation, were not believed to have a negative effect on neonate survival. Only when high tides were associated with high swells were they considered a possible mortality factor. Tide measurements were taken directly from National Oceanic and Atmospheric Administration (NOAA), National Ocean Service height predictions for Forrester Island, Alaska.

Density—Total counts of adults, subadults, pups, and animals of unknown age were summed for each day. Figures for days when no counts were taken were averaged from the count data of the day before and day after.

Survival Estimation

We used the Cormack–Jolly–Seber (CJS) (Lebreton et al. 1992, Williams et al. 2002) open population model in Program MARK (White and Burnham 1999) to analyze the encounter history data and develop maximum likelihood, daily survival estimates for SSL neonates. We employed an information-theoretic approach (Burnham and Anderson 2002), where a set of competing models are constructed a priori and compared based on Akaike's information criterion (AIC; Akaike 1973) and the principle of parsimony (Box and Jenkins 1970). Similar to most open population studies, our methods estimate apparent survival (φa), and not true survival (φ). The quantity (1 −φa) equals the combined effects of emigration and mortality. Though we attempted to follow females presumed to have lost their pup throughout the study period, it is possible that a pup may have left the rookery and was never re-observed but yet survived until 21 d of age.

Twenty-two day encounter histories (providing 21 survival estimates) were constructed directly from the archived electronic daily data forms denoting the location and status of individual females. This form was transformed into an encounter history for each focal female's pup by replacing WP, WPZ, S?, and S with a “1” and each NP and NPO to a “0” (Table 1). If the female was not sighted on a given day, this event also became a “0.” Any focal female that could not unambiguously be recognized by her three identifying marks was either subsequently confirmed through photo analysis or the focal female was considered as not sighted for the day.

Encounter histories were censored at 22 d in order to obtain an unbiased estimate of φ rather than attempt to estimate survival over a longer interval that may be overly confounded by emigration from the rookery. Pups gather in groups at about 2 wk of age, sometimes playing in tide pools (Lewis 1987; personal observations). We were concerned that emigration would increase through time after pups begin to enter the sea to learn to swim with their mothers at around 24–32 d of age and becoming proficient swimmers between 36–41 d old (Sandegren 1970, Schusterman 1981).

All individual covariates, except swell height, were scaled in the input file to improve performance of the numerical algorithm to calculate the maximum likelihood parameter estimates (White and Burnham 1999). Each study female's pup was assigned a number based on its birth sequence. The remaining three individual covariates (Tide*Swell, Swell, and Density) were entered in a 21-d sequence coinciding with each encounter history day beginning with the pup's first day of life. Densities were entered as the total counts of all animals from each site for the day scaled by dividing by 100. Tide and swell interactions were entered as true tide heights for Forrester Island ( scaled by dividing by 10, and then multiplied by the categorical swell height. Areas 1 and 5 were delineated by a group dummy variable (0 or 1). Swell was entered categorically as noted in the Methods section as 0–4. All models incorporating a covariate were estimated using a logit link function.

Apparent survival estimates for the 3-wk interval were computed as the product of the 21 daily estimates of φ based on the top model from program MARK (Table 3). Given that the interval estimate is the product of 21 daily estimates, standard errors for the 3-wk estimate of φ were calculated via the delta method from the conditional variance–covariance matrix associated with each of the four sites/year resighting history combinations (Seber 2002). These estimates and associated standard errors were logit-transformed then back-transformed to produce unbiased, asymmetrical confidence intervals (Burnham et al. 1987).

Table 3.  Models of apparent daily survival probability (φ) until 21 d of age. Factor abbreviations are defined in Table 2.
ModelQAICcΔQAICcQAICc weightsModel likelihoodNumber of parametersQDeviance
{φ (Area*Year + Day 1) P (Area + Year + Age + Age2)}2,206.0600    0.250531     102,185.89
{φ (Area*Year + Day 1 + Surge + Surge2) P (Area + Year + Age + Age2)}2,207.3611.30050.130760.5219122,183.12
{φ (Area*Year + Day 1 + Swell + Swell2) P (Area + Year + Age + Age2)}2,207.5691.50890.117820.4703122,183.328
{φ (Area*Year + Day 1 + Dens) P (Area + Year + Age + Age2)}2,207.9681.90760.096520.3853112,185.764
{φ (Area*Year + Day 1 + Swell) P (Area + Year + Age + Age2)}2,207.9961.93580.095170.3799112,185.792
{φ (Area*Year + Day 1 + Bdate) P (Area + Year + Age + Age2)}2,208.0411.98050.093070.3715112,185.837
{φ (Area*Year + Day 1 + Surge) P (Area + Year + Age + Age2)}2,208.0581.99740.092280.3683112,185.854

Models were first estimated to evaluate differences in apparent survival based on sites and years allowing for heterogeneity in the resighting probabilities. Next, age effects on resighting probabilities were incorporated, then the inclusion of a time covariate that allowed for a unique survival estimate for the first day of life. Once a model that incorporated these effects was specified, the four covariates (birth date, density, surge, and swell) were incorporated both as linear and quadratic effects.

We used Program RELEASE (Burnham et al. 1987) to assess the potential for heterogeneity and over-dispersion in the data. Tests 2 and 3 of this program compute chi-square statistics and associated degrees of freedom that can be used to compute variance inflation factors useful in the analysis of data that may show a lack of independence. The variance inflation factor is used to adjust model selection with the quasi-likelihood AIC statistic (Burnham and Anderson 2002), and to adjust variances of parameter estimates. Justification of these procedures is provided by Burnham et al. (1987).

A Priori Models for Hypothesized Survival Effects

One of the primary precepts of the information-theoretic approach is that a set of candidate models be developed before any close examination of the data takes place to avoid using the data to propose likely models (Burnham and Anderson 2002). We based all candidate models on the accumulated published observations of mortality, the results of necropsies of dead pups (primarily starvation and sepsis) and our personal observations of mortality events at Lowrie Island from previous years.

Area 5 has a lower survival than area 1— This hypothesis was based on accumulated observations of presumed mortalities that occurred at Area 5, the fact that storms ravaged this area more frequently and the steep interface at the water's edge made it difficult for pups to make it ashore after being washed in.

Survival is lowest on the first day of life— We observed newborns swept away during storm events, tossed and bitten by aggressive females shortly after birth, and abandonment and separations of the mother pup pair before bonds could be established. These observations suggested that mortality may be significantly higher during the first day of life.

High swell events lower survival— Other researchers have proposed that storms are responsible for a portion of neonatal mortalities (Orr and Poulter 1967, Gentry 1970, Sandegren 1970, Mate 1973, King 1983). We measured swell height to account for storm activity.

High tide combined with high swell lowers survival— We noted that storm activity during higher tide cycles produced conditions where waves reached far up the rookery and washed pups some distance from the edge of the rookery into the ocean. High tides alone had no observed effect. We computed a variable called “surge,” estimated as the product of tide height in feet times a swell category based on the size of the swell at the base of the rookery (see below).

High rookery density lowers survival— Mortality rates are known to increase in other otariids with increasing density (e.g., Doidge et al. 1984, Harcourt 1992). We noted on Lowrie that as the rookery became more densely packed, near the mean pupping date of 4 June (Pitcher et al. 2001), pups often experienced aggression from neighboring females and mother pup separations were more common.

The time of birth affects the survival of neonates— Because of changing oceanic, weather, or social conditions, when birth occurs during the breeding season may affect neonatal survival.

Resighting probability (P) decreases over time— Immediately following birth and for the subsequent 3–13 d, females usually do not go to sea and can be found closely associated with their pups (Sandegren 1970, Higgins et al. 1988, Millette and Trites 2003). Because resighting is dependent on a clear association with their mother we predicted that resighting probability would drop off steadily as the pups became older.

Area 1 has a higher resighting probability than Area 5— The blind at Area 1 provided a clear view of the rookery and was located directly across from the animals. Most of the “marked” animals there stayed within our view throughout the study period. In contrast, at Area 5 there is a large cliff face to the northwest of the blind that partially blocked our view to the north where many females pupped. Some of the females that gave birth at Area 5 occasionally moved out of our view to this side and were not seen as regularly.


  1. Top of page
  2. Methods
  3. Results
  4. Discussion
  5. Acknowledgments
  6. Literature Cited

Parturition of focal females occurred from 20 May to 13 June (during a 25-d period) in 2002 and from 21 May to 15 June (during a 26-d period) in 2003. In 2002 we cataloged 146 females, 129 of which we resighted more than two times, and we were able to develop observation histories for 119 of these. In 2002, we had limited success in determining time of birth—only 42 pups with known birth dates were used in the analysis. In 2003 we cataloged 203 “marked” females, 179 of which we resighted more than two times and we were successful in developing observation histories for 143 of these. Many of these females left the site after they were marked. Of the 143 females, 45 were never sighted with a pup. Of the remaining 98, 10 were eventually removed from the analysis because: (1) after the female gave birth it moved to the periphery of the study site and could not be regularly resighted (n= 8), or (2) two females had such similar markings that we were unable to confidently discriminate among them. In summary, over two breeding seasons we were successful in developing encounter histories for 130 females with pups that had known birth dates, 42 in 2002 and 88 in 2003.

There were a total of four stillborn animals observed during the two seasons. In 2002, one stillborn was observed at Area 1. In 2003, two stillborns were observed at Area 5 and one at Area 1. None of the stillborn animals were born to focal females, so these mortalities were not included in the survival estimates.

Mark–Resight Model Results

Goodness of fit— Tests 2 and 3 from Program RELEASE (Burnham et al. 1987) produced a χ2 value of 164.9 with 120 df. The resulting variance inflation factor inline image of 1.3744 inline image was used to adjust the variance estimates.

Resighting probability (P)— Resighting probability was best modeled as a curvilinear, quadratic effect. As expected, resighting probability dropped off steadily over time at both sites, and Area 1 had a slightly higher probability of resighting than Area 5 (Fig. 2). All of the top models included the covariates Area + Year + Age + Age2 to explain heterogeneity in P (Table 3).


Figure 2. Comparison of daily resighting probabilities (P) at Areas 1 and 5 on Lowrie Island in southeast Alaska 2002 and 2003. Error bars depict 95% confidence intervals. Note that the estimates are offset slightly for clarity.

Download figure to PowerPoint

Apparent survival (φa) estimates. Several models were evaluated based on our a priori hypotheses. The best approximating φa models were ordered in terms of ΔQAICc (Table 3). Our top model was >1.3 ΔQAICc better than the second-best model and therefore we did not perform model averaging. All of the top models included Area and Year effects as their interaction term. In addition, the top models all contained a day 1 additive effect because survival on a pup's first day of life was significantly less than the remaining 20 d (Table 4). First-day survival (Table 4) ranged from a low of 0.900 (SE = 0.052, Area 5, 2002) to a high of 0.995 (±0.006, Area 1, 2003). Daily survival estimates after the first day ranged from a low of 0.963 (SE = 0.014) on Area 5 in 2002 to a high of 0.998 (±0.002) on Area 1 in 2003 (Table 4). Overall, 21-d survival estimates ranged from a low of 0.423 (±0.132) on Area 5 in 2002 to a high of 0.963 (±0.044) on Area 1 in 2003 (Table 5). Daily survival was consistently lower the first day following birth, and overall survival was lower in 2002 than 2003 and consistently lower at Area 5 (Table 4, 5).

Table 4.  Daily apparent survival estimates from the minimum QAICc model.
YearAreaDaysEstimateSE95% CI
Table 5.  Survival estimates for the 21-d neonatal interval including the first day of life.
YearAreaSample size (n)EstimateSE95% CIs
2002Mean 0.5840.09400.3960.749
2003Mean 0.7740.04830.6660.85 
MeanMean 0.6790.05240.5690.772

Inclusion of environmental covariates— The inclusion of covariates to explain additional heterogeneity in the daily survival rates was not strongly supported by our data (Table 6). This is evident in examination of the top seven models in Table 3, where the addition of linear and quadratic effects for Swell, Surge, Bday, or Dens ranked well below the top model (>1.3 ΔQAICc units lower). However, there was some indication that the Tide*Swell2 covariate may have some effect on lowering daily survival rates, particularly on Area 5. The inline image estimate for this covariate was negative and its 95% confidence interval narrowly overlapped 0 (Table 6).

Table 6.  Estimates and 95% confidence intervals (CI) of covariate coefficients inline image from the top two models (superscripted) presented in Table 3. Estimates are on the logit scale.
Variableinline imageSE(95% CI)
  1. 1Beta estimates from the top model (#1).

  2. 2Beta estimates from the second best model (#2).

1Phi Area−1.080.69(−2.44, 0.28)
1Phi Year2.071.35(−0.58, 4.72)
1Phi Area*Year−1.591.43(−4.38, 1.21)
1Day 1−1.070.54(−2.13, 0.00)
2Phi Surge2−0.260.16(−0.57, 0.05)


  1. Top of page
  2. Methods
  3. Results
  4. Discussion
  5. Acknowledgments
  6. Literature Cited

Possible Sources of Error in Field Methodology

Previous demographic analyses of SSLs have assumed a first-year survival estimate ranging from 0.776 to 0.782 (Calkins and Pitcher 1982, York 1994, Holmes and York 2003). Since our mean estimate for the first 21 d of life is less than this assumed annual estimate, it is important to discuss possible sources of error in our methods.

Detection and resighting biases— Due to darkness and poor visibility, we were unable to observe the rookery 24 h/d and it is therefore possible that we missed some females that gave birth and lost their pups shortly after birth. In our data set, these females would end up classified as “never sighted with a pup.” To the extent that this occurred our survival estimates would be biased high.

There is the possibility that some pups were washed from the main rookery sites and ended up on an offshore rock or a part of the island that was shielded from our view and survived to at least 3 wk of age despite our best efforts during all island searches. Permanent emigration during the survival interval cannot be discriminated from death in our estimates and can lead to a negative bias. We do not know how frequently this may have occurred. We often observed females apparently searching and calling for their pup after storm events and then resighted these females alone, repeatedly afterward, at their birthing sites. However, in one case, a focal female's pup was washed off the rookery during a storm at a time we believe she was at-sea foraging. We did not see her again with her pup until 5 d later when they both swam back to the birthing site. Emigration rates of mother–pup pairs are unknown for the 3-wk birthing interval, but we believe these events to be uncommon in this early stage since pups of this age usually do not enter the open sea unless washed in by a storm event or thrown in by an aggressive female (Gentry 1970, Sandegren 1970; personal observations). Though permanent emigration was confounded with survival, we believe any bias that was introduced to be small during this interval. During all island searches in both 2002 and 2003 we did not sight any focal-female with a pup outside of the study sites before her pup reached 3 wk of age. Within 3 wk of giving birth, only two females (both from Area 5) were sighted outside of the natal rookery and there were no pups with them at these sightings.

Not directly marking the newborns— Using the mother as the “marked” animal required meeting three assumptions: (1) females were not misidentified, (2) mother–pup pairs were properly categorized, and (3) pups were not adopted. As noted above, we believe our field protocol and careful elimination of unsubstantiated mother–pup liaisons helped to meet these assumptions.

If the identities of two or more females were confused, this would bias our estimates of the detection probabilities and the survival estimates. However, errors were minimized with the minimum “three distinct and unique marks” method described above, drawings of individual animals, and the use of a photo library to confirm identification. Any uncertain sightings were subsequently identified through the archived photos or eliminated from the analysis.

Pup adoption— Unlike elephant seals (Mirounga spp.) and some other pinnipeds, SSLs are not known to adopt or foster abandoned pups (as cited in Reidman and Le Boeuf 1982). Instead, they typically act aggressively toward any pup other than their own, especially in the several days before and following parturition. Female aggression toward pups is a leading cause of separations and injuries that lead to death (Gentry 1970, Sandegren 1970; personal observations). On several occasions we witnessed newborn pups being bitten and tossed by females, sometimes ending up in the ocean or with debilitating injuries.

Adoptions among SSLs are considered rare and accounts appear to be cases of misidentification rather than fostering behavior or altruistic acts (Sandegren 1970, Reidman 1990). We observed two instances in 2003 like the 17 June 2004 field journal entry below:

“Saw a female give birth to a pup that rolled down the rocks. She acted confused and carried the placenta like a pup. She did not recognize the pup even though it was only 20 feet away. Another female came down and claimed the pup and carried it up the rock. She allowed it to suckle. Eventually her pup came and suckled also.”

Based on our observations, we believe <1% of mother–pup pairs at the time of branding are cases of adoption. Since these events are rare and it is equally likely for a focal-female to adopt or loose a pup to another female, we believe they had negligible effects on our survival rate estimates.

Lack of fit of the data to the mark–resight model— CJS models assume that every “marked” pup had the same probability of resighting on each day and the same probability of surviving to the next time interval. However, our capture history data showed evidence of heterogeneity in both capture and resighting probabilities. Females that were on the periphery of a study area may have had a capture probability different from those in the center of the rookery, or other locations easier to survey. Resighting heterogeneity was difficult to assess a priori since some females were initially “marked” in easy to see locations but subsequently were displaced to the periphery by more dominant females. Further, a potential source for lack of independence in capture/resighting probabilities may have come from females that gave birth to pups at similar times. These study animals would have had similar attendance/resighting patterns and therefore parallel recapture probabilities.

The slight increase in resighting probability at the end of the 21-d sequence may have resulted from an increase in the nursing frequency of neonates. After the pups were ∼2 wk old, we would often see known females come ashore and call to their pup, followed by joining of the mother and pup and immediate nursing. This proved an effective way to recapture females and positively identify them as being with their pup.

Site selection by females— In both 2002 and 2003, Area 5 was colonized one to two wks earlier than Area 1 despite the fact that neonatal survival rates were lower in this Area (Table 5). Earlier occupancy of a rookery site with eventual lower neonatal survival is counterintuitive. We hypothesize that many of the early colonizers to Area 5 were young females that were less experienced at establishing and defending a prime birthing space within this site and in successfully rearing a pup. In contrast, Area 1 may have been occupied by older, more dominant and experienced females.

An alternative explanation for the lower survival rates from Area 5 is the possibility that this area experienced a higher emigration rate of mothers and pups. Emigration from the rookery prior to 21 d would be confounded with mortality and bias the survival rate estimates low. If emigration from Area 5 were higher, a greater proportion of pups should be sighted outside of Area 5 prior to completion of their 21-d survival interval. To test for a possible association between Area and emigration rate we compared the two areas in 2003 based on (1) all encounter histories and (2) encounter histories ending in zero (i.e., not sighted alive on the last day). We found no significant differences in the proportion of pups from the two areas sighted outside of their rookery based on all encounter histories (χ2= 0.3394; P= 0.5602) or with encounter histories ending in zero (χ2= 0.3395; P= 0.5274).

High mortality during the first day of life—De Villiers and Roux (1992) found that most pup mortality in South African fur seals (Arctocephalus pusillus pusillus) occurred shortly after birth and that survival rate increased over time. A similar pattern was noted in this study. First day mortalities to otherwise healthy SSL neonates may have occurred from several causes including injury from aggressive females, accidental trampling by fighting adult males, crushing during stampedes, being washed off the rookery and eventually drowning, or failure to develop the mother–pup bond leading to starvation. Failure of the mother–pup bond is known to be a cause of mortality in northern elephant seals (Mirounga angustirostris) (Reidman and Le Boeuf 1982). From our observations, the first few minutes after birth appear to be crucial to the development of the mother–pup bond. Once initial physical contact (suckling) and vocal and olfactory cues were exchanged, the bond seemed secure. If there was some disruption of this early contact, such as a fight that took the female's attention away, there was a greater chance the pair might not stay together. Other factors, such as the age and experience of the mother may play a role in the formation of these bonds. Smaller, and presumably younger, females were sometimes observed to quickly abandon their offspring after birth, appearing confused by the birthing event.

Demographic Implications

Estimates of age-specific survival and reproductive rates for SSLs are based on the ages of animals collected in the GOA from 1977 to 1978 (Calkins and Pitcher 1982). Because of inconsistency in the age structure used to estimate the vital rates, however, York (1994) adjusted the rate estimates based on a stable age distribution. These adjusted estimates have been used to evaluate SSL population dynamics and to explore the contribution of specific life history components to the observed population declines from the mid-1970s to the early 1990s (York 1994, Holmes and York 2003).

The contributions of changes in various demographic rates to the observed declines varied over time. The early declines were best explained by low juvenile (0–2 yr of age) survival, low adult (≥3 yr) survival during the late 1980s to early 1990s, and low fecundity throughout the 1990s (Holmes and York 2003). Seventy-two percent of the decline in population growth rate in the early 1980s was explained solely by the estimated declines in juvenile survival (York 1994). Over the entire period, however, population declines were most strongly related to declines in juvenile and adult survival. These results suggest that the factors affecting SSL population dynamics have changed over time.

Based on the data of Calkins and Pitcher (1982), first year survival was estimated to be 0.782 and equal to survival from ages 1–3 yr. Importantly, the first year survival estimate used in previously published demographic studies did not include any mortality that occurred during the breeding season (York 1994, Holmes and York 2003, Eberhardt et al. 2005). Based on the results of this study, previous estimates of first year survival appear to be biased high because they failed to account for mortality between birth and the time the pup leaves the birthing area. Combining our mean estimate (0.679) of apparent survival during the first 3 wk of life with the previous estimate of first year survival (0.782; York 1994) yields an adjusted estimate of 0.531.

The Leslie matrix reported by York (1994) was adjusted to yield a population growth rate = 1.0. Changing the first year survival rate estimate in York's Leslie matrix to 0.531 yields a= 0.961, suggesting a 4% annual rate of decline. However, the size of the eastern population of SSLs has apparently increased over the last several years (Angliss and Outlaw 2005, Pitcher et al., in press) suggesting that other survival rates may be greater than those previously published.

Management Insights

Most long-lived vertebrates, such as SSLs, show greatest sensitivity in population change (λ) to changes in adult survival rates (e.g., Eberhardt 2002, Gerber and Heppell 2004). However, population change can also be sensitive to life history attributes with smaller elasticities and sensitivity coefficients if those rates are naturally more variable. Our results suggest that neonatal survival varies among years and is variable among rookery sites, even among sites on the same island. Further studies of neonatal survival should gather data from several sites in order to better understand spatial variability and to identify those sites that may have consistently high neonatal survival rates. In addition, our results suggest that sustaining high levels of neonatal survival may be important to long-term population recovery.

At end of the season, with only a few animals remaining on the rookery at Area 1, only three dead pups were in sight from the blind. Without disturbing the few remaining animals, a survey of a portion of this same view-shed by foot resulted in a count of 26 dead pups. On Area 5, late also in the season, a foot-survey of a small portion of the rookery revealed nine dead pups in a single crack about 8 m long, unobservable from the blind. Since large storms commonly wash the rookery, many pups that die ashore are also swept away. At several times during the season we sighted many bloated pup carcasses floating in the water off of Area 5, and dead pups have been sighted by scuba divers on the ocean floor near Area 1 (personal observations). These observations suggest that studies using only direct observations of mortality over-estimate survival, and that mark and encounter studies of neonates will provide less biased estimates of rookery survival rates.

  • 1

    Personal communication from Ken Pitcher, Alaska Department of Fish and Game, 16141 Bridgewood Circle, Anchorage, AK 99516


  1. Top of page
  2. Methods
  3. Results
  4. Discussion
  5. Acknowledgments
  6. Literature Cited

This ADF&G research was conducted under MMP and ESA permit #358-1564 and ADFG ACUC 03-002; we thank the permitting agencies for their effort. We give sincere thanks to NMFS and ADF&G for providing funding and logistical support. Thanks to LGL Alaska Research for supporting draft preparations of this manuscript. Special thanks to: B. Dickson, J. Blakesley, R. Scherer, S. Converse, A. Yackel Adams, A. Franklin, B. Wunder, R. Robinette, D. Winkelman (all currently or formerly from CSU), K. Pitcher, L. Jemison, K. Hastings, L. Rea, K. Raum-Suryan, M. Rehburg, (all currently or formerly from ADF&G), T. Gelatt, T. Loughlin (both from NMFS), W. Kendall (USGS), H. Kaplan, A. Baylis, E. Schoen, G. Anderson, C. Mui, and B. Wilson (all ADF&G field crew members). Thanks to Loretta Curgus, John Woodbury, and Phil Smith for logistical and moral support. Thanks to Captain Wade Loofborough and the crew of ADF&G R/V Medeia for meals, transportation, and a welcome shower. Finally, we are grateful to Heather Kaplan for her edits and helpful suggestions that substantially strengthened this paper.

Literature Cited

  1. Top of page
  2. Methods
  3. Results
  4. Discussion
  5. Acknowledgments
  6. Literature Cited
  • Akaike, H. 1973. Information theory as an extension of the maximum likelihood principle. Pages 267281 in B. N.Petrov and F.Csaki, eds. Second International Symposium on Information Theory. Akademiai Kiado, Budapest , Hungary .
  • Altmann, J. 1974. Observation study of behaviour: Sampling methods. Behaviour 49:227265.
  • Angliss, R. P., and R. B. Outlaw. 2005. Alaska marine mammal stock assessments, 2005. U.S. Department of Commerce , NOAA Technical Memorandum NMFS-AFSC-161. 250 pp.
  • Box, G. E. P., and G. M. Jenkins. 1970. Time series analysis: Forecasting and control. Holden-Day, London , UK .
  • Burnham, K. P., and D. R. Anderson. 2002. Model selection and multimodel inference. A practical information-theoretic approach. 2nd edition. Springer-Verlag, New York , NY .
  • Burnham, K. P., D. R. Anderson, G. C. White, C. Brownie and K. H. Pollock. 1987. Design and analysis methods for fish survival experiments based on release-recapture. American Fisheries Society Monograph 5.
  • Calkins, D. G., and E. Goodwin. 1988. Investigations of the decline of Steller sea lions in the Gulf of Alaska. Alaska Department of Fish and Game. 76 pp. Available from Alaska Department of Fish and Game, 333 Raspberry Road, Anchorage, AK 99518.
  • Calkins, D. G., and K. W. Pitcher. 1982. Population assessment, ecology and trophic relationships of Steller sea lions in the Gulf of Alaska. Environmental assessment of the Alaskan Continental Shelf. Final Reports 19:455546.
  • De Villiers, D. J., and J. P. Roux. 1992. Mortality of newborn pups of the South African fur seal Arctocephalus pusillus pusillus in Namibia. South African Journal of Marine Science 12:881889.
  • Doidge, D. W., J. P. Croxall and J. R. Baker. 1984. Density-dependent pup mortality in the Antarctic fur seal Arctocephalus gazella at South Georgia. Journal of Zoology (London) 202:449460.
  • Eberhardt, L. L. 2002. A paradigm for population analysis of long-lived vertebrates. Ecology 83:28412854.
  • Eberhardt, L. L., J. L. Sease and D. P. Demaster. 2005. Projecting the trend of Steller sea lion populations in western Alaska. Marine Mammal Science 21:728738.
  • Fefferman, N. H, and J. M. Reed. 2006. A vital rate sensitivity analysis for nonstable age distributions and short-term planning. Journal of Wildlife Management 70:649656.
  • Gentry, R. L. 1970. Social behavior of the Steller sea lion. Ph.D. thesis, University of California Santa Cruz , Santa Cruz , CA . 113 pp.
  • Gerber, L. L. 2004. Steller sea lions (Eumetopias jubatus) in the Pacific Rim: Biological uncertainty and extinction risk. Pages 482490 in H. R.Akcakaya et al. , eds. Species conservation and management: Case studies. Oxford University Press, New York , NY .
  • Gerber, L. R., and S. S. Heppell. 2004. The use of demographic sensitivity analysis in marine species conservation planning. Biological Conservation 120:121128.
  • Gerber, L. R., and G. R. Vanblaricom. 2001. Implications of three viability models for the conservation status of the western population of Steller sea lions (Eumetopias jubatus). Biological Conservation 102:261269.
  • Harcourt, R. 1992. Factors affecting early mortality in the South American fur seal (Arctocephalus australis) in Peru: Density-related effects and predation. Journal of Zoology (London) 226:259270.
  • Higgins, L. V., D. P. Costa, A. C. Huntley and B. J. Le Boeuf. 1988. Behavior and physiological measurement of maternal investment in the Steller sea lion, Eumetopias jubatus. Marine Mammal Science 4:4458.
  • Holmes, E. E., and A. E. York. 2003. Using age structure to detect impacts on threatened populations: A case study with Steller sea lions. Conservation Biology 17:17941806.
  • King, J. E. 1983. Seals of the world. Cornell University Press, Ithaca , NY .
  • Kendall, W. L., J. E. Hines and J. D. Nichols. 2003. Adjusting multistate capture recapture models for misclassification bias: Manatee breeding proportions. Ecology 84:10581066.
  • Kenyon, K. W., and D. W. Rice. 1961. Abundance and distribution of the Steller sea lion. Journal of Mammalology 42:223234.
  • Lebreton, J. D., K. P. Burnham, J. Clobert and D. R. Anderson. 1992. Modeling survival and testing biological hypotheses using marked animals: A unified approach with case studies. Ecological Monographs 62:1118.
  • Lewis, J. P. 1987. An evaluation of a census-related disturbance of Steller sea lions. M.S. thesis, University of Alaska Fairbanks , Fairbanks , AK . 93 pp.
  • Loughlin, T. R. 1997. Using the phylogeographic method to identify Steller sea lion stocks. Pages 159171 in A.Dizon, S. J.Chivers and W. F.Perrin, eds. Molecular genetics of marine mammals. Special Publication No. 3, The Society for Marine Mammalogy, Lawrence , KS .
  • Loughlin, T. R., D. J. Rugh and C. H. Fiscus. 1984. Northern sea lion distribution and abundance: 1956–80. Journal of Wildlife Management 48:729740.
  • Mamaev, E. G. 1998. Non-invasive monitoring of a reproducing population of Steller's sea lion. 2nd International Symposium on Physiology and Ethology of Wild and Zoo Animals Berlin, 7–10 October. 92 pp.
  • Mate, B. R. 1973. Population kinetics and related ecology of the northern sea lion, Eumetopias jubatus, and the California sea lion, Zalophus californianus, along the Oregon coast. Ph.D. dissertation, Oregon State University , Corvallis , OR . 94 pp.
  • Merrick, R. L., R. Brown, D. G. Calkins and T. R. Loughlin. 1995. A comparison of Steller sea lion, Eumetopias jubatus, pup masses between rookeries with increasing and decreasing populations. Fishery Bulletin 94:753758.
  • Millette, L. M., and A. W. Trites. 2003. Maternal attendance patterns of Steller sea lions (Eumetopias jubatus) from stable and declining populations in Alaska. Canadian Journal of Zoology 81:340348.
  • Noon, B. R., and C. M. Biles. 1990. Mathematical demography of spotted owls in the Pacific Northwest. Journal of Wildlife Management 54:1827.
  • O'Corry-crowe G., T. Gelatt, K. W. Pitcher and B. Taylor. 2005. Crossing significant boundaries: Evidence of mixed-stock origins of new Steller sea lion, Eumetopias jubatus, rookeries in Southeast Alaska. Abstract, 16th Biennial Conference on the Biology of Marine Mammals, 12–16 December 2005 San Diego, CA.
  • Orr, R. T., and T. C. Poulter. 1967. Some observations on reproduction, growth, and social behavior in the Steller sea lion. Proceedings of the California Academy of Sciences 35:193226.
  • Pendleton, G. W., K. W. Pitcher, L. W. Fritz, A. E. York, K. L. Raum-Suryan, T. R. Loughlin, D. G. Calkins, K. K. Hastings and T. S. Gellatt. 2006. Survival of Steller sea lions in Alaska: A comparison of increasing and decreasing populations. Canadian Journal of Zoology 84:11631172.
  • Pitcher, K. W., V. N. Burkanov, D. G. Calkins, B. J. Le Boeuf, E. G. Mamaev, R. L. Merrick and G. W. Pendleton. 2001. Spatial and temporal variation in the timing of births of Steller sea lions in the Gulf of Alaska. Journal of Mammalogy 82:10471053.
  • Pitcher, K. W., P. F. Olesiuk, R. F. Brown, M. S. Lowry, S. J. Jeffries, J. L. Sease, W. L. Perryman, C. E. Stinchcomb and L. F. Lowry. 2007. Abundance and distribution of the eastern North Steller sea lion (Eumetopias jubatus) population. Fishery Bulletin 107:102115.
  • Raum-Suryan, K. L., K. W. Pitcher, D. G. Calkins, J. L. Sease and T. R. Loughlin. 2002. Dispersal, rookery fidelity and metapopulation structure of Steller sea lions (Eumetopias jubatus) in an increasing and a decreasing population in Alaska. Marine Mammal Science 18:746764.
  • Reidman, M. J., and B. J. Le Boeuf. 1982. Mother-pup separation and adoption in northern elephant seals. Behavioral Ecology and Sociobiology 11:203215.
  • Reidman, M. 1990. The pinnipeds: Seals, sea lions, and walruses. University of California Press, Berkeley , CA .
  • Sandegren, F. E. 1970. Breeding and maternal behavior of the Steller sea lion (Eumetopias jubatus) in Alaska. M.S. thesis, University of Alaska , Fairbanks , AK . 138 pp.
  • Sandegren, F. E. 1976. Courtship display, agonistic behavior and social dynamics in the Steller sea lion. Behaviour 57(1–2):136158.
  • Schusterman, R. J. 1981. Steller sea lion—Eumetopias jubatus (Schreber, 1776). Pages 119142 in S. H.Ridgway and R.J.Harrison, eds. Handbook of marine mammals. Volume 1. The walrus, sea lions, fur seal and sea otter. Academic Press, New York , NY .
  • Seber, G. A. F. 1982. The estimation of animal abundance and related parameters. Macmillan, New York , NY .
  • Seber, G. A. F. 2002. The estimation of animal abundance and related parameters. 2nd edition. Blackburn Press, Caldwell , NJ .
  • Trites, A. W. 1993. Biased estimates of fur seal pup mass: Origins and implications. Journal of Zoology (London) 229:515525.
  • Trites, A. W., and P. A. Larkin. 1996. Changes in the abundance of Steller sea lions (Eumetopias jubatus) in Alaska from 1956 to 1992: How many were there? Aquatic Mammals 22:153166.
  • White, G. C., and K. P. Burnham. 1999. Program MARK: Survival estimation from populations of marked animals. Bird Study 46 (Suppl.):120138.
  • Williams, B. K., J. D. Nichols and M. J. Conroy. 2002. Analysis and management of animal populations. Academic Press, San Diego .
  • York, A. E. 1994. The population dynamics of northern sea lions, 1975–1985. Marine Mammal Science 10:3851.
  • York, A. E., R. L. Merrick and T. R. Loughlin. 1996. An analysis of the Steller sea lion metapopulation in Alaska. Pages 259292 in D.R.McCullough, ed. Metapopulations and wildlife conservation. Island Press, Washington , DC .