Top-down and bottom-up influences on demographic rates of Antarctic fur seals Arctocephalus gazella

Authors


Correspondence author. E-mail: Schwarz@biology.ucsc.edu

Summary

  1. Two major drivers in population dynamics are bottom-up processes, such as environmental factors that affect foraging success, and the top-down impacts of predation.
  2. Many populations of marine mammal and seabird species appear to be declining in response to reductions in prey associated with the bottom-up effects of climate change. However, predation, which usually occurs at sea and is difficult to observe, may also play a key role.
  3. We analysed drivers of population dynamics of Antarctic fur seals Arctocephalus gazella at Cape Shirreff from 1997 to 2009, including a predator that targets pre-weaned pups and bottom-up environmental effects in an ecosystem particularly sensitive to small changes in temperature.
  4. We use Bayesian mark-recapture analysis to demonstrate that although large-scale environmental variability affects annual adult survival and reproduction, first year survival appears to be driving the current decline in this population (as defined by a decline in the annual number of pups born). Although the number of pups increased during the first third of the study, first year survival and recruitment of those pups in later years was very low. Such low survival may be driven by leopard seal Hydrurga leptonyx predation, particularly prior to weaning.
  5. Our results suggest that without leopard seal predation, this population would most likely increase in size, despite the observed bottom-up effects of climate changes on adult vital rates. More broadly, our results show how age-targeted predation could be a major factor in population decline of K-selected colonial breeders.

Introduction

Environmental change can have complex effects on populations, influencing population growth via interactions that impact individual fitness. At the same time, many species are subject to top-down predation. Few studies have attempted to address the two factors simultaneously in large mammals.

A changing environment can produce population-level increases or decreases through changes in available critical habitat (i.e. breeding habitat), alterations in the metabolic requirements of individuals (McDonald et al. 2012), and modifications in prey availability (Costa et al. 2010), with the latter defining bottom-up effects. In addition, there may be lags as changes in prey availability work their way through the food web, resulting in different temporal and spatial scales of prey availability across the different age classes (Le Bohec et al. 2008).

Although long-lived species, including mammals and seabirds, often have behavioural and physiological mechanisms that buffer them against variations in food resources occurring over large spatial and temporal scales (Costa 1993; Forcada, Trathan & Murphy 2008), some species are exhibiting dramatic declines in survival and reproductive rates attributed to bottom-up forcing over relatively short time periods. For example, vital rates of several marine bird species are correlated with environmental variability, including El Niño conditions, which may reduce prey availability (Barbraud & Weimerskirch 2001; Le Bohec et al. 2008; Rolland, Weimerskirch & Barbraud 2010; Baylis et al. 2012). However, none of these studies examined top-down effects, or the relative importance of top-down and bottom-up factors.

With respect to top-down effects, predation is often not random, with predators targeting the most easily caught groups when the energetic costs of prey acquisition are high (Carbone, Teacher & Rowcliffe 2007). If predators prefer younger animals (due to ease of capture, or increased availability) with higher reproductive potential, increases in predation will reduce the prey population growth rate unless the targeted species can compensate by increasing its reproductive rate. However, most long-lived species, and especially large mammals, have limited reproductive flexibility because of restrictions on the number of times a female can reproduce and low offspring-parent ratio per reproductive event. For example, predation has dramatic consequences on early survival of large terrestrial herbivores (Linnell, Aenes & Anderson 1995; Barber-Meyer, Mech & White 2008; Carstensen et al. 2009). Additional studies have shown variability in pre-reproductive female survival can be an important component of population growth, particularly when a species exhibits comparatively high and stable adult survival and limited reproductive flexibility (Gaillard, Festa-Bianchet & Yoccoz 1998; Gaillard et al. 2000). Indeed, both empirical data and modelling have shown that, in such cases, prey populations can shift from declining to increasing simply by removing predation during the first year (Banks, Newsome & Dickman 2000; Berger & Conner 2008; Yarkovich, Clark & Murrow 2011).

Antarctic fur seals Arctocephalus gazella are emblematic of this life-history strategy, with both bottom-up forcing and predation of young potentially limiting population growth. In South Georgia, where predation of young is low (Forcada et al. 2009), both pup production and population growth were associated with El Niño conditions, with variability in breeding female survival as the largest factor in growth rate variability (Forcada et al. 2005; Forcada, Trathan & Murphy 2008). In contrast, at a cove on Seal Island, Antarctica, leopard seals Hydrurga leptonyx may have taken between 32 and 69% of the annual pup cohort within the first 3 months of a 5 month pupping season (Boveng et al. 1998). In addition, by comparing trends in pup counts at sites with and without predation, Boveng et al. (1998) suggested leopard seal predation could contribute to declines of some colonies and result in lower population growth in the larger region of the South Shetland Islands.

This research seeks to quantify the relative effects of climate variability and predation on population growth of fur seals at Cape Shirreff, Livingston Island, South Shetland Islands (Fig. 1). Cape Shirreff is the southern-most breeding colony for the species, indicating environmental factors may limit population growth. At the same time, leopard seals frequently consume fur seal pups (Goebel et al. 2009). Preliminary analysis estimates roughly half of the 87 individual leopard seals identified since 1997 are resident or regular visitors during the breeding season, and leopard seal numbers have increased over the last decade, with a recent maximum count of 30 individuals hauled out at one time (U.S.–AMLR, unpublished data; Vera, Vargas & Torres 2004).

Figure 1.

Location of Cape Shirreff, Livingston Island, Antarctica with primary study area located in the square. General adult female fur seal winter range shaded in green (created by L. Hückstädt, unpublished data from M. L. Lea).

Materials and methods

We use Bayesian mark-recapture techniques to first quantify the effects of environmental variability on adult demographics. Next, on the basis of the findings and methods of Boveng et al. (1998), we measure the potential effects of leopard seal predation on early survival by assuming pup survival before entering the water (less than 20 days old) is a measure of non-predatory mortality. Lastly, we combine those results to determine the collective effects of both predation and the environment on Cape Shirreff fur seal population growth, while verifying our results against leopard seal energetic studies.

Data

We began a dedicated mark-resight programme for Antarctic fur seals at Cape Shirreff, Livingston Island, Antarctica (62°27′32″S, 60°47′19″W) in the austral summer of 1997/1998. Each summer, we individually mark roughly 500 pups and between 30 and 100 lactating females with flipper tags. Annual resight effort coincides with the start of pupping in mid-November and continues to early March, with daily resight effort in the main study area (see Supp. Info.- Tagging and resight methods). Beginning in 2000/2001, we extracted teeth from females tagged as adults to estimate age from cementum layers. A professional tooth ageing laboratory (Matson's Lab, Missoula, MT) counts dentition layers, and we consider age estimates from teeth to be precise and unbiased (see Supp. Info.- Teeth ageing technique and sampling bias).

We used mark-resight data of females from austral summer 1997/1998 through 2006/2007 to estimate annual demographic rates. Because of unintended sampling bias (see Supp. Info.- Teeth ageing technique and sampling bias), adult survival rates by age could only be calculated for years 2002 through 2006. The bias also meant age-structured reproductive rates (the fraction of females giving birth to live young) were calculated for all years, but data were restricted to only animals that survived to 2002/2003. We determined the potential effect of leopard seal predation on pup survival using daily resight data of female pup pairs with known pupping dates for austral summers 2004/2005 through 2007/2008.

Analysis: annual demographic rates in relation to environment

We developed a Bayesian multi-event, multistate method based on the standard Cormack-Jolly-Seber (CJS) mark-recapture model (Poole 2002) that also allowed for mark loss (Reisser et al. 2008) and temporary emigration (Dupuis 1995). In our case, we explicitly modelled the probability of an Antarctic fur seal losing one tag and the probability of moving in or out of the main study area (see Supp. Info.- Annual demographic rate analysis). Thus, we estimated five basic probabilities/parameters that define an animal's transition from one state to another state between sample periods: survival rate (s), reproductive rate (r), probability an animal moves in/out of the study area (m), probability of adult female detection (d) and probability of losing one tag (g).

In contrast to fur seals from South Georgia that stay within roughly 500 km of their breeding beaches year-round (Staniland et al. 2012), Cape Shirreff seals remain local during breeding season but travel thousands of kilometres offshore to forage in areas off the coast of Chile, north of the Falkland Islands, around South Georgia Island and in Pacific Ocean fronts between 45°S and 65°S during winter (Polito & Goebel 2010; Lea unpublished data) (Fig. 1). We examined the influence of environmental conditions by correlating adult survival and reproduction with an environmental index based on the temporal and spatial scale of the seals' foraging range as the mean of the monthly Antarctic Oscillation (AAO) between April and December of each year (http://www.cpc.ncep.noaa.gov/products/precip/CWlink/daily_ao_index/aao/aao.shtml). AAO values representing larger and smaller time periods within a year were also evaluated as well as annual lag periods between AAO values and demographic rates. However, the above AAO values with no lag showed the highest correlation and were the most biologically plausible given the seals' temporal and spatial foraging distribution. When estimating reproductive rates, the range and variability in average over-winter AAO during the studied time period were similar to all available years combined (1979–2009), while over-winter AAO values were slightly lower for years when survival rates were estimated (Supp. Info.- Fig. 1). Survival and reproductive rates (x) were estimated as a linear function of the AAO index with categorical age class (AC) as an independent covariate:

display math(eqn 1)

The interaction term between AC and AAO was analysed but removed because its posterior distribution centred on zero, and the Deviance Information Criteria (DIC) was higher for a model with this interaction term. Age classes were defined as 5–7, 8–16 and 17 + years old (see Supp. Info.- Annual demographic rate analysis). For comparison, we also estimated annual demographic rates independent of AAO values.

Low juvenile resight probabilities prevented a rigorous analysis of the influence of AAO on survival rates of 1–4 year olds. However, younger survival rates (age classes: 1 and 2–4) were estimated separately by year for years 1998–2005, maintaining between-year posterior correlations to project population counts while acknowledging high within-year uncertainty. In addition, small sample size limited the estimate of reproductive rate for ages three and four to one value across all years which we used in analyses of population growth.

Other transition probabilities were quantified to assess potential differences based on age or reproductive status (see Supp. Info. – Annual demographic rate analysis). However, changes in assumptions about movement, tag loss and tag detection by age, tag number, or reproductive status did not change the overall pattern in demographic rates with respect to AAO. Priors on estimated parameters β0, β1 and β2 were broad, uniform distributions wide enough to prevent truncation of posterior distributions. All other priors were uniform between 0 and 1.

Analysis: first year pup survival

Using a similar multi-event, multistate Bayesian approach, mark-resight analyses were used to estimate pup survival, pup sighting probability and female sighting probability. We used daily resights of adult females with and without pups and analysis of mark-recapture data of the pups. Transition probabilities between states were pup survival (s), pup detection (dp) and female detection (df). We made model assumptions based on potential temporal and individual differences in sighting probabilities derived from the animals' natural history (see Supp. Info.- First year pup survival). As the probability of detecting a pup given detection of its mother changes as pups become more independent, pup detection was defined as a cubic-logit function of pup age:

display math(eqn 2)

Based on DIC values, a cubic function fit better than models with fewer or more parameters. For the other two transition probabilities, we took a hierarchical approach, with pup survival probabilities on the daily scale (sa) and maternal detection probabilities on the individual scale (dfi). We also allowed for individual variability in the probability of sighting the female given the pup was alive or dead. (Clark et al. 2005).

Using daily counts of live pups, Boveng et al. (1998) showed that mortality in fur seal pups is constant when leopard seal predation is absent. They then used the non-predatory mortality rate as a control to estimate the proportion of pups taken by leopard seals at a nearby site. Instead of using a nearby site to estimate non-predatory mortality, we used early survival at the same site as a control. So, the analysis was also conducted grouping survival rate into age classes: 1–7, 8–20, and 21 + days old. Because pups do not enter the water until they are approximately 30 days old, we assumed survival from 8 to 20 days represented non-predation mortality, and any additional mortality measured at 96 days old was associated with predation. Ninety-six days was the maximum age at which researchers were present in the area, resighting pups and observing leopard seal predation.

To investigate the potential influence of higher first year survival on population growth (as defined by annual pup counts), the proportion of pups alive at 1 year old was also calculated by extending daily survival rate estimates out to 365 days using both the higher daily survival rate (8–20 days) and the lower daily survival rate measured on day 96. The lower rate corresponds to survival if predation were assumed to continue. To determine the feasibility of the assumption that additional mortality was due to predation, we estimated the number of leopard seals required to consume the estimated number of predation mortalities up to 96 days old if all leopard seals were either juvenile females with the lowest consumption rate (0·8 ± 0·15 stdev fur seal pups/day) or adult males with the highest consumption rate (1·3 ± 0·25 fur seals pups/day) (Forcada et al. 2009). We then calculated the number of additional days of predation required by the same number of leopard seals to reach the estimated lower first-year survival rate. To compare first year survival based on annual flipper tag resights with estimates from daily pup resights, results are reported as survival from 75 days old (mean age at which pups are generally flipper tagged) to 1 year old.

Analysis: population growth

To determine the effects of both potential top-down and bottom-up factors on this population, we estimated the population age distribution for 1997/1998 derived from pup counts from that year, knowledge of the system, and several assumptions. Using annual mark-recapture estimates of juvenile (age class 2–4) survival rates, reproductive rates ages 3 and 4 pooled for all years, and adult demographic rates as a function of AAO, the population was projected forward, and the number of estimated pups was recorded through 2009/2010. We projected the population using estimates of first year survival based on both higher and lower daily pup survival rates (see Supp. Info. – Age distribution and population projection). In both cases, inter-annual variability in juvenile survival rates is maintained. Model estimates of pup counts were then compared with field counts. In addition, posterior samples of demographic rates were used in a Leslie matrix to calculate growth rate sensitivity with high and low AAO values and high and low first year survival (see Supp. Info.- Sensitivity analysis). To verify our results, we also ran both sensitivity analysis and pup projections doubling the relationship between AAO and adult rates.

Posterior distributions and joint posterior samples of all variables were determined using the program MTG (Metropolis within Gibbs) developed by Daniel Goodman of Montana State University (Schwarz 2008). Standard practices ensured mixing, convergence and stationarity in posterior samples (see Supp. Info – MCMC sampling).

Results

Antarctic fur seals show age variation in demographic rates with prime-age (8–16 years old) females exhibiting the highest reproductive and survival estimates. Oldest females (17 + years old) exhibited the lowest rates, while young animals (5–7 years old) were intermediate (Fig. 2). Uncertainty was higher for the oldest age class due to small sample size (Fig. 2). Annual survival and reproductive rates of animals older than 4 years were slightly positively correlated with AAO (Fig. 2), but the relationship was stronger for survival rates [β2 = 0·37 (95% Cred. Int.: −0·09, 0·83)] than that for reproductive rates [β2 = 0·29 (−0·05, 0·63)]. Based on annual tag resight data, the mean probability a female survives from time of tagging (roughly 75 days old) to 1 year old is annually variable between 0·04 and 0·26 (Fig. 3). For results on tag loss, tag detection and movement probabilities, see Supp. Info. – Tables 4 and 5.

Figure 2.

Posterior distributions of survival and reproductive probabilities as a function of age class and Antarctic Oscillation (AAO). Solid lines are means and grey areas are 95% posterior intervals. Points with bars are posterior mean and 95% credible intervals with years assumed independent from AAO.

Figure 3.

Survival probabilities from 75 days to 1 year old by cohort. Results for cohorts 97/98–03/04 were based on annual resight data for all years and ages (filled circles). Results for cohorts 04/05–07/08 were based on first year survival assuming high daily survival rates (open triangles) and low daily survival rates (open circles). Points are means. Boxes are quartiles, and whiskers are 95% credible intervals.

When grouping pup survival by age class, rates were lower and annually variable for the oldest age class (21 + days old), whereas rates were higher and similar for the younger age classes (1–7 and 8–20 days old) (Table 1). As a result, both the mean proportion of pups surviving to 96 days old and the proportion of pups potentially lost to predation by 96 days old were annually variable (Table 2). When extending daily survival rates out to 365 days old, mean cohort survivorship from 75 days to 1 year old was between 0·18 and 0·28 (Fig. 3). For results on pup and female daily detection probability, see Supp. Info. – Fig. 2 and 3.

Table 1. Daily probability of pup survival grouped by age and year Marginal means with 95% credible intervals in parentheses
Breeding seasonAge class (days)
1–78–2021+
04/050·998 (0·994,1·000)0·998 (0·996,1·000)0·992 (0·990,0·994)
05/060·998 (0·995,1·000)0·999 (0·997,1·000)0·991 (0·989,0·993)
06/070·999 (0·996,1·000)0·999 (0·997,1·000)0·988 (0·985,0·991)
07/080·997 (0·993,1·000)0·997 (0·994,0·999)0·987 (0·984,0·990)
Table 2. Estimated and measured survival rates from birth to 96 days old based on daily survival rates and sample size (n). Also, the estimated proportion of fur seals lost to predation at 96 days old and the number of leopard seals required to produce that estimated loss. Lastly, the additional number of predation days needed for the same number of leopard seals to produce the difference in low and high annual survival rate estimates. Marginal means with 95% posterior intervals in parentheses
Breeding season n Survival rate (birth to 96 days)Potential proportion consumedNumber of leopard seals neededAdditional predation days
MeasuredAssumed highJuvenile femalesAdult males
04/051740.51 (0.43,0.59)0.80 (0.65,0.90)0.28 (0.12,0.41)11.0 (4.3,19.2)6.8 (2.7,11.9)27.0 (0.0,101.0)
05/061730.49 (0.40,0.57)0.87 (0.76,0.93)0.38 (0.25,0.49)13.9 (8.0,22.3)8.6 (4.9,13.9)39.7 (0.0,112.7)
06/071660.43 (0.33,0.52)0.85 (0.71,0.93)0.42 (0.27,0.54)15.2 (8.5,25.1)9.4 (5.2,15.5)26.4 (0.0,84.7)
07/081430.36 (0.27,0.45)0.71 (0.55,0.83)0.35 (0.18,0.50)10.6 (4.9,18.6)6.6 (3.0,11.2)3.0 (0.0,28.0)

Depending on energetic needs and year, the average number of leopard seals needed to consume the estimated number of pups lost to predation at 96 days old was between 6·6 and 15·2 leopard seals per day (Table 2). If the same number of leopard seals continues to consume fur seals at the same rate after 96 days old, they would need an average of 3·0 to 39·7 additional predation days to produce the annual survival rates estimated from daily rates before 96 days old (Table 2).

Pup count estimates using low first-year survival rates closely matched increasing and then declining annual female pup counts [corr = 0·75 (95% Cred. Int.: 0·52, 0·93)] (Fig. 4). Although variability in adult demographic rates did produce some inter-annual variability in pup counts, pup counts were more likely to increase with high first year survival, and such model estimates were negatively correlated with observed pup counts [corr = −0·63 (−0·76, −0·41)]. Correlations are even stronger when limited to years later than 2000 when field pup counts were generally in decline: 0·85 (0·61, 0·97) with low first year survival and −0·85 (−0·95, −0·60) for high first year survival (Fig. 4).

Figure 4.

Estimated annual female pup count based on model estimates with measured low first year survival (blue) and assumed high first year survival (orange) compared to field counts (red points). Lines are means and shaded areas are 95% posterior intervals.

The sensitivity of population growth rate varied with different levels of first year survival and AAO values. Population growth rate is most sensitive to first year survival when it is low, whereas sensitivity is more evenly spread across survival rates of age classes 1 through 16 when first year survival is high (Table 3). In addition, differing AAO values did not change sensitivity estimates by age class when first year survival was high (Table 3). Patterns in pup counts and sensitivity analysis did not change when doubling the relationship between AAO and adult rates.

Table 3. Sensitivity analysis with high and low AAO values defining survival and reproduction for ages five and older and high and low first year survival rates. See Supp. Info. – Sensitivity analysis for more details. Marginal means with 95% posterior intervals in parentheses
 Age classHigh 1st year survivalcLow 1st year survivald
High AAOaLow AAObHigh AAOLow AAO
  1. a

    0.614 in 2001.

  2. b

    −0.831 in 2002.

  3. c

    0.56 (0.15,0.90).

  4. d

    0.04 (0.01,0.08) in 2001; 0.10 (0.05,0.19) in 2002.

Survival10.20 (0.10,0.40)0.22 (0.12,0.45)0.83 (0.22,1.67)0.60 (0.29,0.96)
2–40.34 (0.20,0.42)0.35 (0.23,0.43)0.09 (0.04,0.16)0.20 (0.14,0.25)
5-70.23 (0.14,0.26)0.25 (0.19,0.27)0.09 (0.03,0.15)0.18 (0.13,0.22)
8–160.28 (0.21,0.33)0.26 (0.19,0.33)0.20 (0.08,0.31)0.32 (0.27,0.34)
17+0.09 (0.02,0.34)0.04 (0.01,0.17)0.56 (0.27,0.85)0.19 (0.07,0.40)
Reproduction3–40.08 (0.01,0.15)0.10 (0.02,0.18)0.00 (0.00,0.01)0.02 (0.01,0.03)
5–70.08 (0.01,0.14)0.10 (0.03,0.17)0.00 (0.00,0.01)0.02 (0.01,0.04)
8–160.10 (0.04,0.13)0.12 (0.08,0.14)0.02 (0.00,0.05)0.07 (0.03,0.11)
17+0.04 (0.02,0.07)0.03 (0.01,0.07)0.06 (0.02,0.08)0.07 (0.06,0.09)

Discussion

Our goal was to measure both the bottom-up effects of environment and the potential top-down effects of leopard seal predation on Cape Shirreff Antarctic fur seal population growth. Both sensitivity analysis on population growth and projection estimates of pup counts show that although adult demographic rates vary as a function of over-winter AAO, first year survival is currently the primary driver of growth for this population. Given the analyses presented here, coupled with preliminary data on leopard seal behaviour, counts and attendance, leopard seal predation prior to weaning could have caused the observed low recruitment of fur seals in this area.

Two studies have hypothesized that summer environmental conditions near the breeding colony influence adult vital rates. Food limitations during a relatively short lactation period may impose fitness costs that reduce maternal survival as well as prevent implantation, the latter resulting in lower pupping rates the following year (Forcada et al. 2005; Forcada, Trathan & Murphy 2008). Although that may be the case, over-winter foraging conditions will determine if a female can compensate for lost energy reserves due to rearing offspring as well as provide nutrients for a developing foetus (Hiruki-Raring et al. 2012). Thus, large-scale winter conditions far from the breeding colony may have consequences for adult reproductive and survival rates, similar to what is seen in some colonial seabirds (Le Bohec et al. 2008; Baylis et al. 2012).

Environmental conditions measured by mean winter AAO values were trending downwards during the period of this study, and adult survival and reproductive rates decreased with decreasing AAO. Many climate models predict the AAO index will decline (IPCC 2007), and our results suggest that this would lead to reduced adult fur seal survival and reproduction and further lower population growth. It follows that survival of younger animals (1–4 years old) would further drive population growth, and as the probability of a female surviving and returning to her birth site after age 4 years is low, it is unlikely this group would recover from consistently poor environmental conditions. However, our analysis is restricted to adult female survival rates as a function of AAO. As juvenile survival rates most likely also respond to environmental perturbation (Beauplet et al. 2005), a declining AAO may lead to even further reduction in Cape Shirreff Antarctic fur seal population growth. In addition, the small number of years studied here may not entirely represent the long-term relationship between AAO and fur seal demographic rates or completely capture the periodicity and autocorrelation of AAO. However, effectively doubling the strength of the AAO vs. adult vital rate relationship did not change the pattern seen in sensitivity analysis or pup projections: first year survival is currently driving growth of this population.

Six lines of evidence support the hypothesis that leopard seal predation is the cause of low first year survival for Cape Shirreff fur seals. First, in South Georgia where leopard seals are not present during the summer, the majority of pup mortality occurs prior to pups entering the water (Forcada, Trathan & Murphy 2008). Our results show pup mortality is higher after they enter the water in an area where leopard seals are present (Table 1), as previously observed at Elephant Island (Boveng et al. 1998). Second, although separated spatially and temporally, Cape Shirreff annual pup count declines from 2002 to 2009 (6·0 ± 7·7%year−1; mean and std) and model predictions (decline 6·0 ± 6·6%year−1) are similar to what was seen at Elephant Island when leopard seals were present (decline 4·2 ± 1·0%year−1) (Boveng et al. 1998). Third, the range of annual mean fraction of pups taken by leopard seals at around 75 days old measured at Elephant Island is similar, but slightly higher compared to this study (0·32–0·69 and 0·22–0·35 respectively) (Boveng et al. 1998). Fourth, mean cohort survivorship from 75 days to 1 year old extrapolated from high daily pup survival rates (pooled mean: 0·56, mean of annual means: 0·51, range: 0·45–0·82) is comparable to mean survivorship from South Georgia where leopard seal predation is negligible (0·52) (Forcada, Trathan & Murphy 2008). Fifth, although uncertainty was higher when estimating survival from 75 days to 1 year old using daily survival rates, values were comparable to estimates based on annual mark-recapture analysis (Fig. 3).

Lastly, leopard seal estimates fall within a reasonable range of what we know from preliminary leopard seal data. Namely, roughly half of all tagged leopard seals appear to remain in the area for the majority of the breeding season, and the maximum number of hauled out leopard seals has been counted at 30. Regardless of the age structure of the leopard seal population, 15 leopard seals could consume the number of additional pup mortalities estimated at 96 days old (Table 2). Unweaned fur seal pups will remain vulnerable to leopard seals an additional 24 days before they can potentially move out of leopard seal habitat. Depending on the year, the probability the additional mortalities occurred prior to weaning are between 0·34 and 0·97. However, the degree of winter habitat overlap for these two groups is not known, and predation events may occur throughout the year. Leopard seal scat and stable isotope analysis, tracking data, mark-recapture analysis and detailed behavioural studies will allow us to determine temporal and spatial aspects of leopard seal prey type (including prey switching and prey age classes), attendance patterns and consumption rates.

Only a few marine mammal studies have begun to empirically explore population-level effects when predation is targeted on pre-breeders. Horning & Mellish (2012) showed killer whale Orcinus orca predation accounted for 10 of 11 documented deaths in juvenile Steller sea lions Eumetopias jubatus, and low recruitment prevents some populations from recovering. Based on energetics models, killer whales in the Subantarctic Prince Edward Islands could theoretically consume the entire Antarctic fur seal pup population within a matter of days or potentially reduce the pup population by 10–40% within a year, depending on different prey selection strategies (Reisinger, de Bruyn & Bester 2011). In general, our research, along with the above-mentioned studies, suggest predation on pre-breeders has the potential to create rapid and dramatic changes in marine mammal populations.

Colony breeders may be particularly vulnerable to predation of newborns as the naïve group provides a large and reliable food source for predators. While parental protection and predator saturation can reduce such risk (Ainley et al. 2005), some species do not guard their young and require time to physiologically develop effective foraging and fleeing capabilities (flying in birds and diving in marine air breathers) (Noren et al. 2001; Fowler et al. 2007; Tome & Denac 2012). In addition, physiological and environmental limitations may keep reproductive rates low that could have otherwise compensated for young lost to predation (Bonin et al. 2012). Therefore, future studies investigating the impacts of predation on K-selected species, particularly colonial breeders, need to consider age or size-specific selection by the predator.

Although AAO has an effect, we found that fur seal population growth at Cape Shirreff is more sensitive to top-down processes in the form of leopard seal predation. Most importantly, we have shown that both top-down and bottom-up forces can be important population drivers for Antarctic fur seals, and the way in which the two forces affect population growth can vary from one colony to another. Bottom-up processes appear to be more important at Bird Island, South Georgia (Forcada, Trathan & Murphy 2008; Forcada et al. 2009). South Georgia is at the northern extreme of the leopard seals' range, so other conditions may influence the leopard seals’ impact on South Georgia fur seals. Or, as seen in Boveng et al. (1998), South Georgia leopard seals may consume fur seal pups at nearby, unstudied colonies where hunting conditions are more favourable. In addition, complex interactions may have created a trophic cascade where leopard seals are now more abundant and targeting fur seals at Cape Shirreff, as has been demonstrated in some non-pinniped systems (Anthony et al. 2008; Harley 2011). In fact, this research, along with additional leopard seal research, could show environmental changes are potentially driving top-down predation on fur seals. If so, fur seal populations could decline more rapidly than anticipated as a result of environmental change. In general, future studies need to consider entire ecosystem shifts when investigating effects of environmental change, which could have both bottom-up and top-down effects on a species.

Acknowledgements

Data were collected by researchers and technicians at NOAA, Southwest Fisheries Science Center, Antarctic Ecosystem Research Division. Research was conducted under MMPA permit 774-1847 to the U.S. AMLR Program. Data collection was also supported by the NSF Office of Polar Programs grant ANT-0838937. Analysis and manuscript preparation was supported by the E & P Sound and Marine Life Joint Industry Project of the International Association of Oil and Gas Producers.

Ancillary

Advertisement