Survival in king penguins Aptenodytes patagonicus : temporal and sex-specific effects of environmental variability

Share Embed


Descrição do Produto

Oecologia (2002) 132:509–516 DOI 10.1007/s00442-002-0985-6

P O P U L AT I O N E C O L O G Y

Olof Olsson · Henk P. van der Jeugd

Survival in king penguins Aptenodytes patagonicus: temporal and sex-specific effects of environmental variability

Received: 28 May 2001 / Accepted: 30 May 2002 / Published online: 12 July 2002 © Springer-Verlag 2002

Abstract We investigated annual adult survival rates of king penguins Aptenodytes patagonicus breeding at South Georgia during 6 years in relation to age/breeding experience, sex, and food availability. During the first 3 years of the study, when food availability was good, survival was 97.7% for experienced breeders, which confirmed the very high survival rates observed in penguins in general. In these years survival did not differ between the sexes, presumably because parental investment is shared equally between the sexes, and the sexual dimorphism is small in king penguins. Survival was lower for young, first-time breeders (83.0%). In experienced birds the annual survival rate decreased to 68–82% following a catastrophic year when food availability was extremely low. We address the question how parents balance their current investment in offspring against their chances to reproduce in the future. We argue that the high mortality rate among breeding individuals after the year of food stress provides support for previous suggestions that the response to increased costs in seabirds might be complex to predict and does not always follow intuitive expectations according to general life-history theory. We also found that females survived significantly less well than males following the bad year. We explain this result as follows: the male-biased sex ratio (56:44) that we observed in our study colony clearly does not reO. Olsson (✉) British Antarctic Survey, Natural Environment Research Council, Madingley Road, Cambridge, UK H.P. van der Jeugd Department of Zoology, Edward Grey Institute of Field Ornithology, University of Oxford, Oxford, OX1 3PS, UK H.P. van der Jeugd Evolutionary Biology Centre, Department of Animal Ecology, Norbyvägen 18D, 752 36 Uppsala, Sweden Present address: O. Olsson, Environmental Advisory Council, Ministry of the Environment, 103 33 Stockholm, Sweden, e-mail: [email protected], Tel.: +46-8-4052705, Fax: +46-8-204331

sult from lower female survival during normal conditions. An already existing skewed sex ratio forces males to delay the onset of breeding because of a lack of breeding partners. This in turn causes breeding females to be, on average, younger and less experienced than males and to have lower survival following a year of food shortage. In this study survival was linked with food availability and we suggest that this was connected to climatic/oceanographic features, such as the position of the Antarctic Polar Front Zone. We could, however, not verify this by anomalies in sea surface temperature data. Keywords Breeding experience · Cost of reproduction · Life history · Sex ratio · Survival

Introduction Seabirds are usually very long-lived and breed many times during their lifetime. However, they produce few offspring at each breeding occasion and, therefore, each breeding attempt constitutes a relatively small proportion of their lifetime reproductive effort (Clutton-Brock 1991; Stearns 1992). In general, life-history theory predicts that organisms with a high probability of adult survival should sacrifice their offspring (especially if they are few) rather than jeopardise their own survival (Armstrong and Robertsson 1988; Clark and Ydenberg 1990), e.g. if having to choose in a situation of food shortage. King penguins Aptenodytes patagonicus are longlived birds, with reported annual adult survival rates over 90% (Weimerskirch et al. 1992; Olsson 1996), laying at the most one clutch annually, which consists of only one egg. In a situation of severe food shortage, we would, therefore, expect them to cease rearing offspring rather than risk their own probability of survival. The uniquely high proportion of bi-parental care in birds makes the study of sex-specific investment in offspring particularly interesting among this group. Among birds, king penguins have an extremely long period of chick rearing; both parents spend about 1 year from start

510

of incubation to fledging of the chick (Stonehouse 1960; Barrat 1976). The female invests in the formation of the egg (which constitutes about 2–4% of her body mass) and the male takes a larger proportion of the incubation duties (two of the, normally, three spells) (Stonehouse 1960; Weimerskirch et al. 1992; Olsson 1995). The extensive rearing period of the chick seems, however, to be shared equally between the sexes (Olsson 1996). Normally, the sex investing most in the offspring is exposed to a higher risk of mortality (Lack 1968; Nur 1984). In addition, mortality may differ between the sexes as a result of sexual dimorphism (Promislow et al. 1992). In king penguins, we expect no sex-biased mortality since the parental investment appears to be about equal and the sexual dimorphism is small. In this paper, we investigate annual adult survival rates of king penguins breeding at South Georgia during a 6-year period when the availability of food was normal or good in all years except one, when it was very low. Hence, this variation in the environmental conditions offered a natural experiment where, based on general lifehistory theory, we had the opportunity to test two hypotheses on an extremely long-lived species with equal parental care: (1) cancelling of breeding efforts and maintenance of high adult survival as a response to adverse environmental conditions; and (2) no sex-biased mortality linked to increased costs of breeding for one sex in king penguins.

Materials and methods Marking and resighting This research was part of a general study of king penguins carried out in a colony of 150–200 individuals at Husvik, South Georgia (54°11′S, 36°40′W), from 1991 to 1996. Virtually all breeding birds in the colony were individually identifiable, using two marking systems: metal flipper rings (Lambournes, UK) on one flipper, which could be read using telescope or binoculars, and passive implant 32 mm transponder tags (TIRIS, Texas Instruments), inserted subcutaneously in the loose skin dorsally between the tail and a leg, for electronic identification. No loss of markings was detected (see results and Olsson 1997a). The small size of the colony allowed the bands of more than 90% of the birds to be read daily, including those in peripheral areas and beaches. All moulting birds were also checked daily but the probability of detecting the bands of these birds was generally lower. Each summer, except in 1991 (see below), bands were checked between October or November and March or April. Bands were also checked sporadically in early spring (September) and in some winters. In addition, from mid-December 1992 to the end of the summer of 1994, between 25% and 100% of the daily arrivals and departures were registered automatically by birds passing a gate on their way to and from the colony, where their identity (from the transponder), time and mass at passage were recorded. The effective resighting period was relatively short and varied depending on which phase in the breeding cycle each individual was in. Hence, all resighted birds were observed early each season during post-winter chick feeding, pre-breeding moult, or at the latest, early in the courtship period. All mortality in adult king penguins detected in this study occurred during winter. In the first summer (1991) the colony was visited daily for about 3 weeks in January/February, and approximately 70% of the breeding birds were banded and their breeding status was obtained. In the following summer the remaining birds belonging to

the colony were banded. Hence, compared to 1991, a relatively large proportion of those marked in 1992 were recruits of the year, and of those marked in subsequent summers virtually all were recruits. Only birds that had made a breeding attempt (egg laid) were included in the study. From 1992 onwards, birds were marked with both transponders and flipper bands. However, because no transponders were inserted in 1991, all birds already banded were caught again in 1992 to be equipped with a transponder as well. At the same time their bands were adjusted if showing any signs of deformation or opening, and thus with potential to injure the bird or to be lost. Such adjustments were also done subsequently when needed; and we made great efforts throughout the entire study to keep the bands of all birds in good condition. Contrary to experience elsewhere (Weimerskirch et al. 1992; van Heezik et al. 1994; Froget et al. 1998), no markings were lost in the present study (Olsson 1997a). To avoid disturbance, unmarked birds breeding in the colony were temporarily marked by applying permanent dye, using a long-reaching pump device, for subsequent recognition and marking outside the breeding area. To control for possible band loss, from 1992 virtually all birds were checked for a transponder by a mobile reading unit, before they were marked. Food availability Although we have no direct measurements of food availability, there is extensive evidence pointing to 1994 being a catastrophic year for the king penguins at South Georgia. This can be summarised as follows: 1. Duration of foraging trips was up to 100% longer compared to normal years (Olsson 1995), and the food carried back to the chicks was more digested (Olsson and North 1997), suggesting that it was caught at a greater distance from the colony. Our satellite tracking showed that king penguins fed up to about 500 km off the island in 1994 (Rodhouse et al. 1998; Kooyman et al. 1999). 2. In 1994 about 60% of the breeding attempts failed before hatching due to abandonment of the egg, compared to only about 10% in other years, and in 1994 none of the chicks that hatched were still alive by early April (Olsson 1995; Olsson 1997a). 3. The chicks from 1993, which fledged in 1994, had lower mass and fledged later than a previous cohort, and none of them was resighted in subsequent years, compared to ca. 50% of the previous cohort (Olsson 1997a). 4. The breeding cycle that included 1994 was significantly longer, and a lower proportion of the birds made late breeding attempts after their previous chick had fledged (Olsson and Brodin 1997). Moreover, nutritional conditions for higher predators around South Georgia in general were very poor in 1994. For many seabird species it was the worst year recorded in over 20 years of study (Croxall et al. 1999). All these observations strongly suggest that the availability of food was extremely poor through the summer of 1994 (and probably also the subsequent winter), compared to the other years of this study. Hence, if this poor year had any effects on the probability of birds’ survival, this should be evident in the subsequent year (1995). Sea surface temperature It is likely that the varying foraging conditions during this study are connected with varying oceanographic/climatologic conditions, such as the position and seasonal variation of the up-welling area in connection with the Antarctic Polar Front Zone (APFZ). Therefore we have investigated if there were obvious anomalies in sea surface temperatures (SST), obtained by satellite remote sensing, in the potential foraging areas of king penguins north of South Georgia in 1994. Based on satellite tracking of five birds from the study colony in 1994 (Rodhouse et al. 1998; Kooyman et al. 1999; Olsson et al.

511 Table 1 Capture-resighting matrix for king penguins Aptenodytes patagonicus at South Georgia Number released Year

Year when first seen again

Newly marked

Previously marked

Total

1992

1993

1994

1995

1996

Never

1991 1992 1993 1994 1995

50 23 9 11 10

– 49 66 73 68

50 72 75 84 78

49

0 66

0 0 73

0 0 0 68

0 0 0 0 70

1 6 2 16 8

Females 1991 1992 1993 1994 1995

40 23 12 18 6

– 39 56 65 59

40 62 68 83 65

39

0 56

0 0 65

0 0 0 59

0 0 0 0 58

1 6 5 24 7

Males

unpublished data) we defined the potential foraging area as covering 50–54°S by 33–40°W. In this area the monthly average SST was extracted each year for the entire study period, 1991–1996, using data from the US NOAA-satellite remote sensing programme (methods are described at http://www.cpc.noaa.gov/products/predictions/30 day/SSTs/ sst_clim. html). Survival analysis Although not necessarily breeding every year, no bird took a sabbatical year to the extent that it was not observed at least during moult or courtship (see Olsson 1997a for details). Therefore, the resighting rate in this study was always equal to one (Table 1). Consequently, there was no need to use software such as SURGE or MARK, which allow resighting rate to differ between years or groups of individuals, to model local annual survival rates. Instead, we used PROC CATMOD in SAS (SAS Institute 1990) to fit linear logistic models in which local annual survival was modelled as a function of age, sex and year. This procedure gives the same results as analyses in SURGE or MARK when the resighting rate is set to one, but offers the advantage that it is less tedious to fit certain complicated models. We used the program RELEASE to test whether the dataset fitted the Cormack-Jolly-Seber (CJS) model, in which survival varies with time only, and fates of individuals are independent (Burnham et al. 1987; Lebreton et al.1992). Because the resighting probability was one, off-diagonal elements in the capture-resighting matrix (Table 1) were absent. Therefore, only test 3Sr in RELEASE was relevant, all other tests being uninformative (and unnecessary). Because data were scarce in certain elements of test 3Sr, Fisher’s exact tests were applied instead of chi-square tests, and results were subsequently pooled by using the formula –2ΣlnPk, where Pk stands for the P value of a single Fischer’s exact test (Sokal and Rohlf 1995). This pooled value approximately follows a chi-square distribution with 2 k degrees of freedom. Since we wanted to investigate possible effects of sex on survival we fitted the CJS model for each sex separately, using a two-group model in RELEASE, and then pooled the results to obtain the overall fit. We used test 1 in RELEASE to test for a possible difference in survival between the sexes (Burnham et al. 1987). We started the analyses from a model considering effects of all variables of interest, plus interactions. We then modelled survival using the principle of parsimony, i.e. by gradually reducing the number of parameters in such a way that the model still provided a good representation of the data (Lebreton et al. 1992). Model selection was done on the basis of a modified Akaike’s Information Criterion (AICc), calculated as the deviance (–2 log likelihood) of the model plus twice the number of parameters, weighted for sample size (Akaike 1973; Anderson et al. 2000). We also performed Likelihood Ratio tests (LRT) to test for effects of different variables on

survival. We did this by comparing two models, one including and one omitting the variable of interest. We followed the model notation as used by Lebreton et al. (1992), where the symbol φ is used for survival, and subscripts a, s and t are used to denote effects of age, sex and time (in this case year), respectively. An asterisk between two subscripts indicates the presence of an interaction between the two effects, while a plus sign denotes that the effects are additive. To investigate the effect of the bad year 1994 on survival we replaced the variable time with the dummy variable b in one of the models we tested. This dummy variable allowed survival only to differ between this year and all other years. To investigate the effect of the bad year 1994 on the sex effect on survival we replaced the variable sex with the dummy variable sb that allowed a sex effect on survival only in this year, but not in all other years. Sex ratio To picture the availability of potential partners of the different sexes during courtship, we extracted data on sex ratio from the daily routine observations in 1991–1996. The data are based on all birds present in the colony that were in the courtship phase of their breeding cycle, including unmarked birds. Birds in other phases of the breeding cycle were not included. The sex ratio was examined on three days each season, two during the normal courtship period for early breeders and one at the median date of courtship for late breeders; except in 1993 when there were no late breeders (Olsson 1996). The time between the two dates for early breeders (1 and 18 December, respectively) was long enough to prevent individuals being included in the sample twice in one season.

Results Survival The aim of this study was to investigate survival rates between sexes and years in king penguins in relation to large-scale fluctuation in food availability. Figure 1 shows the annual survival rates for newly and previously marked males and females, respectively. All mortality occurred during winter when no field observations were made, i.e. between March or April and October or November. Test 3Sr in RELEASE indicated that the CJS model did not fit the data since there were differences in survival between birds captured for the first time and birds captured earlier (sexes pooled: χ2=25.66, df=13,

512

P=0.019). We therefore started our analyses using a model with two age classes, the first one spanning the first year after capture, the second one spanning all remaining years, model (φa2*t*s). The fit of this model was assessed by calculating the LRT between model (φa2*t*s) and model (φt*s), and then subtracting it from the chisquare value given in RELEASE for model (φt*s). Likewise, the appropriate number of degrees of freedom was found by subtracting the difference in the number of parameters between the two models from the number of degrees of freedom given by RELEASE. The approximate LRT thus obtained indicated that model (φa2*s*t) fit-

ted the data satisfactorily (χ2=0.165, df=5, P=0.999). Because in each year all unmarked new breeders were marked, these two age classes generally refer to firsttime breeders and experienced breeders, respectively, but they are no real age classes in the strict sense. Both test 1 in RELEASE (χ2=2.638, df=4, P=0.620) as well as the LRT between models (φa2*t+s) and (φa2*t) (Table 2, models 1, 2 and 3) indicated clearly that, considering the entire study period, there was no difference in survival between the sexes. Annual variation in survival, however, was highly significant, as well as the interaction between age and time (Table 2, models 4 and 5). We then proceeded with our modelling by testing specific hypotheses. First, we tested whether significant annual variation in survival was restricted to experienced breeders only. Probably due to low sample sizes for firsttime breeders, this simplification of the time effect was justified (Table 2, model 6). We then investigated the effect of the bad year 1994 on survival of experienced breeders by replacing the variable time with the dummy variable b (see Materials and methods). However, the LRT between this model and model (φa2*t) indicated that this simplification of the time effect was not supported by the data (Table 2, model 7). In other words, annual variation was not restricted to lower survival following the bad year. However, it was possible to eliminate the time variation during the first 3 years of the study, i.e. before the bad year (Table 2, model 8). Although earlier models indicated that survival did not differ between the sexes in general, we specifically tested whether survival following the bad year did differ between the sexes by replacing the sex effect with the dummy variable sb (see Materials and methods). The resulting model (Table 2,

Table 2 Survival (φ) of king penguins as a function of age (a), sex (s), and time (t) using capture-resighting data. Subscripts b and sb refer to dummy variables (see text). Upper section: general models. Lower section: models used to test specific hypotheses (see text). NP Number of identifiable parameters, DEV deviance, AICc modi-

fied Akaike’s Information Criterion. Bold type AIC values indicate the most parsimonious model for each section. Likelihood ratio tests are based on differences in deviance between two models. A non-significant likelihood ratio test indicates that the model with the lower number of identifiable parameters fits the data

Fig. 1 Annual survival of king penguins Aptenodytes patagonicus breeding on South Georgia, for previously marked birds, i.e. experienced breeders (left), newly marked birds, i.e. first-time breeders (right), and males (● ● ) and females (●). Only survival following the bad year (1994) is significantly lower for females than for males. Survival was analysed using a linear logistic model, and survival values were obtained from the initial model (φa2*s*t) (see text). Since resighting rate always equalled 1, no standard errors or confidence limits can be given for the annual survival estimates

Model

NP

DEV

AICc

Likelihood ratio test

General models: 1 (φa2*t*s)

18

395.060

432.100

2 (φa2*t+s)

11

405.361

427.758

3 (φa2*t)

9

407.097

425.367

4 (φa2+t)

6

419.976

432.101

5 (φa2)

2

467.128

471.146

Fits the data: χ2=0.17, df =5, P=0.999 Interaction between sex and time: 2 vs 1: χ2=10.30, df=7, P=0.172 Sex effect on survival: 3 vs 2: χ2=1.736, df=2, P=0.418 Interaction between age and time: 4 vs 3: χ2=12.88, df=3, P
Lihat lebih banyak...

Comentários

Copyright © 2017 DADOSPDF Inc.