Survival of female common eiders Somateria m. mollissima in a declining population of the northern Baltic Sea

POPULATION ECOLOGY - ORIGINAL PAPER

Survival of female common eiders Somateria m. mollissimain a declining population of the northern Baltic Sea

Martti Hario Æ Marc J. Mazerolle Æ Pertti Saurola

Received: 20 February 2008 / Accepted: 8 December 2008 / Published online: 13 January 2009

� Springer-Verlag 2009

Abstract In long-lived species, adult survival is the

population parameter having the highest elasticity, and

therefore, it can be expected to be least affected by climatic

variations. We studied the dynamics and survival of

breeding female common eiders Somateria mollissima

mollissima in the Baltic Sea from 1960 to 2007. Using nest

censuses and capture–recapture methods, we investigated:

(1) the annual apparent survival (u) of breeding females,

(2) the survival-mediated population fluctuation, (3)

weather effects on survival, and (4) long-term population

trends. Based on capture histories of 6,393 females, aver-

age u was 0.882 (95% confidence interval 0.864, 0.899).

We found no relationship between population growth rate

and survival. Furthermore, the highest ranking models,

based on Akaike’s information criterion, indicated no

effect of the North Atlantic Oscillation index on the u of

eider females. Population size, assessed from nest counts,

has been steadily declining since 1985. Despite the long

time series (48 years), the overall variation in the u rates

remained comparatively narrow, at maximum ranging only

10% between 2 consecutive years. Results imply that

declining female survival is not the driving force behind

the population decline, and we hypothesize that the overall

poor fledging success and the consequent low recruitment

explain the decreasing trend of nest densities since 1985.

Keywords Adult female apparent survival �Mark-recapture � North Atlantic Oscillation �Common eider � Baltic Sea

Introduction

Seabirds are considered the paradigm of extreme K-selec-

ted species with low reproductive rates and long life

expectancies (Lack 1967; reviewed in Croxall and Rothery

1991). Their reproductive output is small, but is compen-

sated for by high adult survival, leading to comparatively

stable population development. The consistently high adult

survival leaves few vacancies, and K-selected populations

remain permanently close to the carrying capacity of their

habitat.

Like sea ducks in general, the common eider Somateria

mollissima is thought to exhibit life history patterns typical

of K-species (Goudie et al. 1994). Annual survival of adult

birds has been fairly high, the average ranging from 0.85 to

0.95 in most studies published so far (see below). The

reproductive rate, on the other hand, can be very low owing

to frequent breeding failures (brood losses rather than low

initial output in the form of small clutch size; Hario et al.

2005). Years with low reproduction may occur in succes-

sion and, as a whole, may outnumber those with high

reproduction. Thus, understanding population processes in

the common eider requires long-term data, especially for

Communicated by Peter Banks.

M. Hario (&)

Finnish Game and Fisheries Research Institute,

Soderskar Game Research Station,

P.O. Box 2, 00791 Helsinki, Finland

e-mail: [email protected]

M. J. Mazerolle

Centre d’Etude de la Foret,

Universite du Quebec en Abitibi-Temiscamingue,

445 boulevard de l’Universite, Rouyn-Noranda,

QC J9X 5E9, Canada

P. Saurola

Ringing Centre, Finnish Museum of Natural History,

University of Helsinki, P.O. Box 17, 00014 Helsinki, Finland

123

Oecologia (2009) 159:747–756

DOI 10.1007/s00442-008-1265-x

studies examining the role of survival in population regu-

lation and carrying capacity of the habitat. As rates of

reproduction vary around a low average, population growth

rates are believed to be most sensitive to changes in adult

survival (Goudie et al. 1994; Wilson et al. 2007). Yet,

despite its importance, the impact of adult survival on

population growth rate has seldom been reported in long-

term eider data sets (but see Coulson 1984; Kats et al.

2007).

In this paper, we explore population processes in a 48-

year data set of common eiders from a discrete population

in the Baltic Sea. We have previously shown that fledgling

production is the trait bearing the highest population

regulatory effect among the breeding parameters of this

population and found significant direct density dependence

(Hario and Rintala 2006). Here, we aim to investigate the

role of adult female survival in population dynamics, and

whether or not survival responds to large-scale climatic

variations. The climatic variable chosen here was the well-

studied North Atlantic Oscillation (NAO) index. This index

is based on the winter (December–March) difference

between the normalized sea level pressure between Lisbon,

Portugal, and Reykjavik, Iceland. Positive values of the

index reflect stronger than average west winds over the

middle latitudes resulting in milder winter temperatures

than normally encountered, whereas negative ones indicate

colder winter conditions. The NAO captures a large part of

the interannual variation in the climate of the northern

Atlantic and adjoining landmasses (Hurrell et al. 2003). We

expect that NAO measures weather conditions that possibly

affect seabirds’ survival rates, either directly (e.g., severe

weather causing seabird ‘‘wrecks’’) or indirectly (e.g.,

anomalies in food acquisition lead to lowering body con-

dition and lowering resistance against secondary

pathogens; Harvell et al. 1999; Friend et al. 2001).

Our study population is part of the large Baltic/Wadden

Sea flyway population, reportedly having declined from ca

1.2 million individuals in 1991 to ca. 760,000 in 2000, a

36% reduction in 10 years (Desholm et al. 2002). This is a

big change in an environment that is comparatively stable

with no catastrophically massive changes in food avail-

ability for seabirds, like those brought about by extreme El

Nino events in the Pacific (e.g., Oedekoven et al. 2001).

Yet, climatic factors in the Atlantic influence several

oceanographic and ecological parameters in the Baltic Sea

(Hanninen et al. 2000), although these changes progress

slowly and become visible only gradually (such as the

effects of the eutrophication; Ronka et al. 2005).

The reasons behind the population decline are so far

only speculative (Desholm et al. 2002). Pinpointing crucial

factors is hampered by the absence of information on the

mechanism of the eider decline, whether due to increasing

mortality or decreasing natality. Emigration can be ruled

out as a factor, because there has been no evidence of

large-scale movements of female eiders out from the

Baltic/Wadden Sea area (Desholm et al. 2002). The bulk of

this population breeds in Finland, Sweden, and Estonia,

and overwinters in the Danish Straits, 1,000 km southwest

from the core breeding areas (Noer 1991). No long-term

documentation of basic dynamic parameters has yet been

published for this population. Here, we make the first

attempt to document survival rates from extensive mark-

recapture data of Baltic common eiders and assess long-

term population trends.

Materials and methods

Study area

The fieldwork was conducted at the Soderskar Game

Research Station, central Gulf of Finland (60�070N,

25�250E), northern Baltic Sea. Soderskar archipelago is a

bird sanctuary consisting of 25 islands and islets. Most of

them are small rocky skerries with patchy grassy vegeta-

tion, averaging 1.14 ha in size. The larger ones, averaging

3.35 ha, have dense juniper Juniperus communis cover and

small stands of aspen Populus tremula and Norway spruce

Picea abies. The total land area of the study islands

amounts to approximately 0.5 km2, and water area of the

whole sanctuary amounts to 13.3 km2 (for detailed

description, see Hario and Selin 1988). The sanctuary is

closed to the general public from May to August, and, due

to active management, there are no mammalian predators

or raptors present during the eiders’ breeding period. The

only nest predators are hooded crows Corvus corone

cornix, but they prey upon single eggs during egg laying,

not entire clutches, and their numbers have always

remained limited (from three to five pairs in the entire

sanctuary) due to the territorial behavior of the species. In

the event of partial nest losses (e.g., single eggs lost to

crows), eider females select another nest cup where they

will continue laying; they do not leave the area or give up

laying (personal observations). Partial nest losses cannot

create confounded patterns to the encounter probabilities in

our recapture data.

Nest abundance

As a measure of population size, we used the annual

number of active nests (nest abundance). Being a capital

breeder (sensu Drent and Daan 1980), the common eider

produces only one clutch in a given year. Thus, the number

of active nests becomes a measure of density, as the size of

the study area was the same every year. Furthermore, all

the islets were searched thoroughly once or twice to locate

748 Oecologia (2009) 159:747–756

123

all the nests each year. As the islands are small and the total

land area is only 0.5 km2, it is relatively easy to find all the

nests, even in the juniper cover. The fieldwork has been

conducted by the same people for more than 30 years (and

by their predecessors for nearly 20 years). During surveys,

nests were always covered with down after the females had

been flushed. In this report, we use the nest abundance data

from 1960 to 2007.

Eider capture–recapture

Female eiders were captured and ringed, or recaptured on

their nests during a 2-week period in May and June each

year from 1960 to 2007 using a long-handled dipnet.

Capture was conducted during the nest count, starting not

earlier than at day 20 of population incubation (median

date). The median incubation date was known from direct

observations at nests of a plot studied intensively (for

additional details, see Hario and Selin 1988). We chose to

undertake our study at least 20 days into the incubation

period in order to minimize the possibility of nest desertion

(sensu Laurila and Hario 1988). Median date of the 20th

incubation day was 21 May for our study period, ranging

24 days (11 May–4 June). Timing of breeding varied

between years according to the ice disappearance (median

date of ice disappearance was 15 April from 1960 to 2007,

range 39 days). Female eiders on the northern breeding

grounds do not start laying until the ice connection to the

mainland has broken up (Hario and Selin 1988). During

this study, we only sampled breeding females (i.e., those at

nests) because there was no logistically feasible way to

catch the non-nesting portion of the population (i.e.,

prospecting immatures). As a result, our study focuses on

the population of genuine female breeders.

Once recruited, the females seem to be faithful to the

same breeding islet for years (only three out of 255 females

controlled at nests in adjacent archipelagos within

10–20 km had previously nested at Soderskar; M. Hario,

and K. Selin, unpublished data). The mean annual capture

success was 52.9% for incubating females [median 53.3%,

SE = 1.84; 95% confidence interval (CI): 56.6, 49.2)] The

range of variation is fairly small except for the 1960s (95%

CI: 69.2, 44.7, mean 57.0%), when the population was still

scarce. The retrieval of individuals might have been better

in those early days.

Modeling density and climatic effects on survival

We used mark-recapture models to investigate density

dependence and weather effects on the annual apparent

survival (u) of eiders. Adopting an information-theoretic

approach (Burnham and Anderson 2002), we formulated a

number of candidate models (see specific hypotheses

below). Some included the annual number of nests (nest

abundance, standardized by subtracting the mean and

dividing by SD) as a covariate influencing the u of eiders,

to evaluate density dependence. Because weather variables

can influence life history traits in animals (e.g., Sandvik

et al. 2005; Stenseth and Mysterud 2005), other models

also included the NAO index (Hurrell 2006). We consid-

ered the annual NAO index corresponding to survival of

the same year, as well as with the survival 1, 2, 3, and

4 years later (i.e., lags 1–4 year). We limited our investi-

gation to lags of 4 years because climate conditions beyond

this time frame were presumably less likely to influence

eider survival. Values of the covariates were entered

directly in the design matrix and all models were computed

using program MARK 5.1 (White and Burnham 1999).

We started by searching for a suitable general model

from which we would build our set of candidate models.

We fit the Cormack–Jolly–Seber (CJS) model to our data,

but goodness-of-fit tests suggested substantial lack-of-fit

(Release test 2 ? test 3, v2 = 1,941.83, df = 373,

P \ 0.0001; U-CARE tests 2.Cl ? 2.Ct ? 3.Sm ? 3.Sr,

v2 = 2,428.54, df = 490, P \ 0.0001). Test 2.Ct indicated

strong deviations from expected values (v2 = 696.36,

df = 45, P \ 0.0001). Though this test was originally

designed to detect immediate trap response, it can also

indicate Markovian or non-random temporary emigration

(Pradel 1993; Schaub et al. 2004; Choquet et al. 2005).

In our data set, female eiders were captured during a

single period each year and encounters of some individuals

were separated by 1–20 years. Because this pattern can be

due to temporary emigration or deferred breeding, we

attempted to alleviate this problem by removing outliers

(individuals that remained undetected for 2 years or more),

but this only marginally improved fit. Though Pollock’s

robust design (e.g., Pollock et al. 1990; Williams et al.

2002) allows one to directly estimate emigration, we could

not use this method because the data had not been collected

in this framework. As a result, we opted for the approach

suggested by Pradel (1993), which consists of creating, for

a given individual, as many encounter histories as there are

events of capture for the same individual. This is achieved

by deleting the individual from the sample following its

second capture, treating it as a removal, and then creating a

new encounter history starting with the occasion of this

capture. These steps are repeated for each capture until we

reach the end of the original encounter history. Each new

encounter history has, at the most, two encounters per

history. In essence, this conversion results in transforming

‘‘time elapsed since previous capture’’ in the original data

set into ‘‘time elapsed since marking’’ in the new data set.

To automate this process, one can either use a small

program (cnv2pradel.exe) developed by J. E. Hines

(USGS Patuxent Wildlife Research Center, Laurel, Md.) or

Oecologia (2009) 159:747–756 749

123

U-CARE 2.2 (Choquet et al. 2005). Once all original

encounter histories were converted as above, we fit a general

model with time-dependent u and different capture proba-

bilities (p and p0) for periods that did and did not follow a

capture, respectively. This model, which we denote as

utpm(t?t) (notation sensu Lebreton et al. 1992), had additive

time and trap effects on capture and corresponds to model

ITEC? of Pradel (1993). A likelihood-ratio test between

model ITEC? and CJS indicated the presence of trap

dependence (v2 = 693.288, df = 1, P \ 0.0001).

We could not implement typical goodness-of-fit

approaches such as the parametric bootstrap or median

variance inflection factor (c) to assess the fit of our

models because Program MARK excludes losses on

captures, whereas our data were coded with a large

number of removals (see Pradel 1993 conversion above).

As a surrogate, we computed a goodness-of-fit test based

on the Release v2 for the CJS model of the full data set

(original capture histories), and on the likelihood-ratio test

(v2) between models ITEC? and CJS for the coded

encounter histories. Here, the test suggested substantial

overdispersion (v2 = 1,394.1154, df = 372, P \ 0.001)

with a cð Þ of 3.75. Burnham and Anderson (2002, p. 68)

advise that we should expect c\4 from models that are

appropriate for the data. As a further check, we compared

the results with different increments of c to assess the

consistency of model ranking and its effect on our

conclusions.

From our general model utpm(t?t) (ITEC?), we

developed a set of candidate models to test the following

eight hypotheses: (1) constant survival, (2) time-depen-

dent survival, (3) linear relationship of survival on nest

abundance, (4) linear relationship of survival on the NAO

index, (5) linear relationship of survival on the NAO

index lagged by 1 year, (6) linear relationship of survival

on the NAO index lagged by 2 years, (7) linear rela-

tionship of survival on the NAO index lagged by 3 years,

and (8) linear relationship of survival on the NAO index

lagged by 4 years. For each hypothesis, we applied each

of the following constraints on p: (1) both capture prob-

abilities held constant (p for periods following a capture

and p0 for periods that did not); (2) both capture proba-

bilities time dependent; (3) constant capture probability of

eiders in periods following an encounter, time-dependent

capture of eiders in periods not following an encounter;

(4) time-dependent capture probability following an

encounter, constant probability of capture not following

an encounter. This yielded a total of 32 candidate models

(8 survival hypotheses 9 4 types of constraints on p). We

proceeded with the quasi Akaike’s information criterion

corrected for small sample sizes (AICc) to rank our can-

didate model set and adjusted the SEs of the estimates byffiffiffi

cp

(Burnham and Anderson 2002).

Population trends and climatic effects

on nest abundance

We analyzed the long-term nest data using regressions with

autoregressive errors from the gls function of program R

(Ihaka and Gentleman 1996; Pinheiro and Bates 2000). The

first analysis consisted of assessing the trends across years.

Nest abundance was natural log transformed prior to ana-

lyzes to homogenize variances. Inspection of the partial

autocorrelation plot suggested an autoregressive process of

order 1 (i.e., nest abundance at a given year depended on

the nest abundance the year before). Plots of residuals

(quantile–quantile plots, residuals vs. fitted values) indi-

cated good model fit.

In a second analysis, we investigated the effects of the

NAO index at lags of 0–4 years on the log number of

nests with regressions for correlated errors (as defined

above). Here again, we inspected the residual plots to

ensure that the assumptions were met and the fit was

adequate. Given that it is inappropriate to compare time

series models with different fixed effects (here, the NAO

index at different lags), we could not rank the autore-

gressive models among each other using the AICc

(Pinheiro and Bates 2000).

Finally, we used regressions with autoregressive errors

to assess the effect of female eider survival (estimated from

mark-recapture analyzes described above) on the popula-

tion growth rate of the same year. Here, we defined the

population growth rate as the logarithmic change in the

nest abundance from year t to t ? 1. It is worth noting that

growth rate is assigned to the year t, not t ? 1. Thus,

growth rate of a given year is plotted against the survival of

the same period. As our population shows long-term trends

(first increase, then decrease), we may expect a positive

relationship between density and survival. We explored to

what extent this holds, i.e., whether survival bears a

demonstrable effect on population trends.

Results

Annual apparent survival

A total of 6,393 females were captured and marked during

this study from 1960 to 2007. Given the data and the set of

candidate models, model utpm(t?t) (ITEC?) had virtually

all the support as being the most parsimonious, with an

Akaike weight of 0.92 after applying the variance inflation

factor of 3.75. Model ranking remained the same for values

of c from 1 to 3.75 with model utpm(t?t) consistently having

Akaike weights [0.92. The evidence ratio (ER) between

Akaike weights of model utpm(t?t) and the second-ranked

model [uNAO lag3pm(t?t)] was 17 for c ¼ 3:75; and this

750 Oecologia (2009) 159:747–756

123

value increased markedly with lower values of c (e.g., with

c ¼ 3:5; ER = 499; with c ¼ 3:25; ER = 24998). Thus,

we based our inference on model utpm(t?t) (sensu Burnham

and Anderson 2002). The u of female eiders ranged from

0.614 to 1, with an average of 0.882 (Fig. 1a), whereas the

probability of capture on a given year was higher for

individuals encountered the year before than those not seen

the year before (Fig. 1b). Given the very low weight of

models based on weather conditions or nest abundance,

there was no definite effect of these parameters on apparent

survival (Table 1).

Nest abundance and population growth rate

The index of population size (i.e., the number of eider

nests) in our study area generally increased from 1960 to

1985 (peak year) but has been decreasing steadily ever

since (Fig. 2). To quantify these relationships, we divided

the nest count data into two periods corresponding to

population rise and decline, namely increasing years

(1960–1985) and decreasing years (1986–2007), and we

analyzed each using generalized least-squares regression

with autoregressive errors (Table 2). The analysis for the

1960–1985 period indicated a strong positive trend of the

log number of nests (byear = 0.0519, 95% CI 0.0332,

0.0706), whereas that for the 1986–2007 period showed a

strong negative trend (byear = -0.0473, 95% CI -0.0630,

-0.0315).

We did not detect any relationship between eider nest

abundance and the NAO index at any lag (Table 3), indi-

cating no direct effects of winter meteorological conditions

on eider nest abundance. Finally, we did not detect any

relationship between population growth rate, expressed as

the change in nest abundance from year t to year t ? 1, and

eider apparent survival (beider survival = -0.146, SE =

0.294, 95% CI -0.738, 0.446), suggesting that survival has

only a limited effect on population fluctuations (Fig. 3).

Discussion

Our survival estimates are consistent with most previous

estimates on eiders (see Table 4). With a few exceptions,

1960 1970 1980 1990 2000 2010

0.5

0.6

0.7

0.8

0.9

1.0

Year

App

aren

t an

nual

sur

viva

l rat

e

1960 1970 1980 1990 2000 2010

0.1

0.3

0.5

0.7

Year

Pro

babi

lity

of c

aptu

re

a

b

Fig. 1a, b Probabilities of survival and capture of female common

eiders Somateria mollissima at the sampling site at Soderskar, Finland

between 1960 and 2007. Estimates are based on model utpm(t?t)

(ITEC?) consisting of time-dependent survival (ut), and additive

effects of trap response and time on the capture probability [pm(t?t)].

Note that the parameters for the last occasion for this model are not

separately identifiable and are not shown. a Annual apparent survival

of eiders of model ITEC?. Error bars denote 95% confidence

intervals around the estimates for each year. b Probability of capture

across years for individuals encountered the previous year (solid line)

or not encountered the previous year (dashed line) based on model

ITEC?

Table 1 Eight highest ranking capture-mark-recapture models of our

candidate set based on the quasi Akaike’s information criterion cor-

rected for small sample sizes (QAICc) [overdispersion parameter,

variance inflection factor (c) = 3.75)]

Modela Number of

parametersbDelta

QAICcc

QAICc

weight

utpm(t?t) 95 0 0.92

uNAO lag3pm(t?t) 51 5.68 0.05

uNAO lag4pm(t?t) 51 8.28 0.01

unest abundancepm(t?t) 51 10.07 0.01

uNAO lag2pm(t?t) 51 10.96 0.00

uconstantpm(t?t) 50 17.26 0.00

uNAO lag1pm(t?t) 51 17.55 0.00

uNAOpm(t?t) 51 19.14 0.00

a Model notation is based on Lebreton et al. (1992). For instance,

model unest abundancepm(t?t) indicates that apparent survival (u) of

female eiders is dependent on nest abundance, whereas the capture

probability is different for periods that do and do not follow captures

(additive time-dependence)b Includes cc QAICc of the highest-ranked model = 20,188.64

Oecologia (2009) 159:747–756 751

123

these estimates vary from 0.826 to 0.895 (range 7%) across

broad geographic areas and subspecies (see also Wilson

et al. 2007). The estimates are fairly high and invariant,

typical of long-lived seabirds. Indeed, the percentage

change in survival between years during the population

decline in our data averaged only 0.3% (range -8.2, 10.6).

In addition to those tabulated in Table 4, Baillie and

Milne (1982) mention a mean annual survival of 0.96 in

adult females in Scotland, but give no further details. Most

populations studied so far are migratory and subject to

hunting. Only populations on the British Isles and in the

Netherlands are sedentary and protected, which may be the

reason for their survival estimates ranking the highest

of all.

Effect of survival on population fluctuations

There was no significant relationship between the popula-

tion growth rate and adult survival. It is not an exceptional

result, but found earlier in several bird species (reviewed

e.g., by Paradis et al. 2002). Several possible scenarios

could explain this pattern. For instance, survival may be

expressing the vitality of the population, being the result

rather than the cause of a population change.

An alternative explanation is the fairly constant bagging

of eiders, the most important quarry species among sea

ducks in the Baltic Sea, on the Danish hunting grounds.

The bags in Denmark exceeded 130,000 individuals per

year in the 1980s, but have declined since then to 83,000

individuals due to fewer hunters (Christensen 2005). Noer

et al. (1995) have shown that, despite the increase in size of

the Baltic eider population in the 1970s and 1980s, the

Danish harvest rate remained at a constant level, enabling

the population to grow. Currently, while a population

decrease takes place, hunting pressure in Denmark is

1960 1970 1980 1990 2000 2010

050

010

0015

0020

0025

00

Year

Num

ber o

f nes

ts

Fig. 2 Yearly eider nest abundance at the Soderskar site in 1960–

2007

Table 2 Population trends based on the eider nest count data divided

into a period of increase (1960–1985) and of decrease (1986–2007)

assessed from regressions with autoregressive errors (AR1)

Period Estimate

of year

SE 95% CIa Pseudo

R2b

1960–1985 0.0519 0.0091 0.0332, 0.0706 0.93

1986–2007 -0.0473 0.0076 -0.0630, -0.0314 0.91

CI confidence intervala The 95% CIs indicate strong trends of the log number of nests

across year for the two periodsb Note that the pseudo R2 for each model was computed as the square

of the correlation between predicted and observed values

Table 3 Effect of the North Atlantic Oscillation (NAO) index at

different lags (0–4 years) on the log number of eider nests assessed

from regressions with AR1

Model Estimate

of NAO

SE 95% CIa Pseudo

R2b

NAO lag0 0.0066 0.0071 -0.0078, 0.0209 0.17

NAO lag1 year -0.0044 0.0073 -0.0192, 0.0104 0.17

NAO lag2 years 0.0029 0.0074 -0.0120, 0.0177 0.166

NAO lag3 years 0.0057 0.0073 -0.0089, 0.0203 0.14

NAO lag4 years -0.0035 0.0072 -0.0179, 0.0109 0.11

a The 95% CIs indicate that nest abundance is independent of the

NAO index at all lagsb Note that the pseudo R2 for each model was computed as the square

of the correlation between predicted and observed values

0.6 0.7 0.8 0.9 1.0

-0.3

-0.2

-0.1

0.0

0.1

0.2

Apparent annual survival

Pop

ulat

ion

grow

th

Fig. 3 Weak evidence for an effect of apparent annual survival on

population growth rate, expressed as the log of nest abundance from tto t ? 1 at the Soderskar site between 1960 and 2007

752 Oecologia (2009) 159:747–756

123

lessened by the fall in popularity of eider hunting

(Christensen 2005). Thus, harvesting has removed a

smaller rather than a larger share of the high densities and

is presently paralleling the population decrease. Such

mortality accentuates fluctuations instead of promoting

population regulation. However, this needs to be studied in

detail, because hunting mortality may have only been

compensated for by decreased density-dependent ‘‘natural’’

mortality factors, unknown so far (see Nichols 1991;

Kokko 2001).

Finally, whatever the causality behind the relatively

small range of variation in apparent annual female survival

rates, results suggest that female survival is not the main

driving force behind the population changes. In a previous

study, Hario and Rintala (2006) showed that annual

reproductive rate (fledglings/female) at Soderskar was

strongly related to population growth rate. The proportion

of explainable variance was 61%, whereas the corre-

sponding figure is only 2% in the current survival data

(Fig. 3). Based on the evidence, we may conclude that

changes in fecundity rather than those in survival bear the

strongest effect on changes in the Soderskar population

size.

The u (estimate ± SE) tended to remain higher

(0.896 ± 0.004) during the era of population increase

(1960–1985) and slightly lower (0.863 ± 0.011) during the

population decrease (1986–2007). Following recommen-

dations from a reviewer, we added an additional survival

model a posteriori to our initial model set to test for dif-

ferences in survival across the above-mentioned periods of

population fluctuations. This model, termed u(period)pm(t?t),

consisted of constant survival for the period of population

growth (1960–1985), and different but constant survival

for the period of population decline (1986–2007). Model

u(period)pm(t?t) ranked highest when compared to the other

models of our model set. Apparent survival estimates

(±SE) for the periods of population growth and decline

were 0.906 ± 0.004 and 0.870 ± 0.005, respectively.

However, model ITEC? regained support with decreasing

c; having virtually all support (Akaike weight [0.93) with

c\3:

Heterogeneity in recapture probabilities

We found that the probability of capture in a given year

was higher for females encountered the year before than

those not seen the year before. It is commonly appreciated

that this sort of heterogeneity in recapture probabilities,

when not accounted for, may bias survival estimates,

especially towards the end of a study period (Prevot-

Julliard et al. 1998; Kery et al. 2006). This clearly justifies

our method of dealing with capture heterogeneity in this

paper. The observed dichotomy in our capture probabilities

might simply stem from the differential catchability of

females, some being more difficult to trap at nests than others

and thus more liable to remain undetected for longer periods.

A similar individual heterogeneity in recapture rates has

been found in several other studies. It has been attributed to

spatial and temporal bias of observation (Kery et al. 2006),

to differences in quality and accessibility of nest sites

(Sandvik et al. 2005), and to strategic decisions to skip

breeding following a breeding (or pairing) failure (e.g.,

Grosbois and Thompson 2005). Whilst outside the scope of

this study, our results suggest that further investigations

into this issue could uncover the mechanisms creating the

observed heterogeneity in recapture probabilities. How-

ever, this would require a study design differing from ours.

Climate effects

Considering the close connection of winter NAO to the

freshwater runoff into the Baltic Sea and its subsequent

Table 4 Published survival estimates of adult female common eiders, with associated SEs, range, or, alternatively, 95% CIs

Country Years Annual

survival

SE Range (–) or

95% CI (;)

Subspecies Source

Canada 1964–1972 0.826 0.010 0.806; 0.846 dresseri Reed (1975)

US 1964–1973 0.886 0.076 0.641–0.971 dresseri Wakeley and Mendall (1976)a

England 1958–1980 0.895 0.015 0.756–1.00 mollissima Coulson (1984)

US 1976–1992 0.873 0.016 0.842; 0.904 dresseri Krementz et al. (1996)

The Netherlands 1969–1987 0.957 – 0.92–0.98 mollissima Swennen (2002)

Finland 1982–1989 0.864 0.012 0.796–0.911 mollissima Hario and Selin (2002)

Norway 1986–1999 0.85 – 0.77; 0.90 mollissima Yoccoz et al. (2002)

US 1994–2004 0.892 0.022 0.842; 0.928 v-nigrum Wilson et al. (2007)

The Netherlands 1974–2003 0.927 0.008 0.911; 0.943 mollissima Kats et al. (2007)

Finland 1960–2007 0.882 0.009 0.614–1.00 mollissima This study

a As reanalyzed and presented by Goudie et al. (2000)

Oecologia (2009) 159:747–756 753

123

effects on water salinity (Hanninen et al. 2000), we would

have expected a relationship between NAO and eider sur-

vival. The salinity of the water is the ultimate control of

faunal and floral composition in the Baltic Sea (Hanninen

et al. 2000). Being at an historical low during the current

era of positive NAOs (HELCOM 2007), the salinity is

thought to have a detrimental effect on many benthic

species in the Baltic biota including the blue mussel

Mytilus edulis, the staple food of eiders (Westerbom 2006).

Yet, we found no evidence of a climate effect on eider

survival when measured with winter NAO index. In other

studies showing effects of the NAO on seabird survival,

some have suggested a causal link with wind force

affecting directly the birds’ food availability or indirectly

the food quality, namely by increasing the predator-prey

interactions in pelagic fishes (Sandvik et al. 2005; reviewed

in Grosbois and Thompson 2005). Interestingly, several

studies demonstrate a negative and lagged effect, but the

actual mechanisms through which NAO influences survival

remain unknown.

Stressors of survival and reproduction

The Baltic–Wadden Sea populations of common eiders

have been in a marked decline over the past 10–20 years. At

present, it remains unclear whether the various subpopula-

tions in this area have declined due to common causes or as

a result of independent factors (Desholm et al. 2002).

Our results point to the central role of reproductive rate

as the population regulating factor. As reported by Hario

and Rintala (2006), the fledging rate at Soderskar tended

to decrease with time. This was principally an outcome

of the extremely low fledging success during the steeply

decreasing phase of the population size in 1986–1996,

when the per capita fledging rate was 0.052 ± 0.013 (SE).

These were years of mass mortalities of small ducklings.

During the period of population increase from 1967 to

1985, the fledging rate was 8 times larger than during the

population decrease, viz. 0.411 ± 0.063. As clutch size

and duckling rate (i.e., downy young entering the water)

did not change, the losses during brood rearing regulated

the population (Hario and Rintala 2006). In 1997 and 1998,

the fledging rate again improved (exceeding 0.411) tem-

porarily halting the decline.

It is commonly understood that even small changes in

adult survival may have large implications for population

development in long-lived animals. Our main results reveal

a fairly high and invariant survival in adult females up to

the end of the 1980s in the Gulf of Finland. However,

apparent survival decreased afterwards with increased

variability. Yet, the mean annual decrease of 0.3% in

survival during 1986–2007 cannot fully explain the mean

annual decline of 4% in population density of those years.

Here, we cannot exclude the possibility of deferred

breeding due to environmental stressors in certain years

(e.g., in 1991; see Fig. 2). Unfortunately, we have no

evidence to substantiate this assumption because we

worked on recapture data alone. Nonetheless, the consis-

tent and steep decline in population density from 1995 to

1999, even when survival was improving, makes extensive

deferred breeding improbable. Indeed, Kats et al. (2007)

reported an instant and total population recovery in only

2 years in the Netherlands as soon as non-breeders re-

entered the colony.

Potential stressors for survival and reproduction include

viruses, parasites, and lead poisoning, in addition to dete-

riorating nutritional condition (increasing metabolic stress

during fasting), all of which have been associated with poor

reproductive success and die-offs of adults in the Gulf of

Finland during the last 15–20 years (Hollmen 2002;

Hollmen et al. 2000, 2001; Franson et al. 2000, 2002).

Eider ducklings have especially suffered from increasingly

large viral infections (Hollmen et al. 2002). We suggest

that most microparasite infections hit the eider reproduc-

tion rates hard because of the undeveloped immuno-

competence of small ducklings. Adult survival is less

affected and remains more predictable than duckling sur-

vival, because adults are survivors from earlier infections

and have already developed an immune response. If this is

correct, it could partly explain the delayed and gradual

change in adult survival compared to the large and drastic

variations in fledging rate. This is also consistent with the

expected life history pattern of a K-selected species under

selection pressures that minimize the variation in the traits

having the greatest repercussions on fitness (such as adult

survival; Stearns 1992).

Acknowledgements Our sincere thanks go out to all those assisting

in the field work at Soderskar, and especially to K. Selin and

P. Muuronen who summarized the yearly statistics. Software and

thoughtful advice from J. E. Hines, as well as comments from

J. D. Nichols, J. Rintala, and M. Ronka enhanced this manuscript.

The long-term field work was made possible by the funding of a

governmental research institute, the Finnish Game and Fisheries

Research Institute.

References

Baillie SR, Milne H (1982) The influence of female age on breeding

in the eider Somateria mollissima. Bird Study 29:55–66

Burnham KP, Anderson DR (2002) Model selection and multimodel

inference: a practical information-theoretic approach, 2nd edn.

Springer, New York

Choquet R, Reboulet A-M, Lebreton J-D, Gimenez O, Pradel R

(2005) U-CARE 2.2 User’s manual (Utilities CApture-REcap-

ture). CEFE, Montpellier

Christensen TK (2005) Factors affecting the bag size of the common

eider Somateria mollissima in Denmark, 1980–2000. Wildl Biol

11:89–99

754 Oecologia (2009) 159:747–756

123

Coulson JC (1984) The population dynamics of the eider duck

Somateria mollissima and evidence of extensive non-breeding by

adult ducks. Ibis 126:525–543

Croxall JP, Rothery P (1991) Population regulation of seabirds:

implications of their demography for conservation. In: Perrins

CM, Lebreton J-D, Hirons GJM (eds) Bird population studies.

Oxford University Press, Oxford, pp 272–296

Desholm M, Christensen TK, Scheiffarth G, Hario M, Andersson A,

Ens B, Camphuisen CJ, Nilsson L, Waltho CM, Lorentsen S-H,

Kuresoo A, Kats RKH, Fleet DM, Fox AD (2002) Status of the

Baltic/Wadden Sea population of the common eider Somateriam. mollissima. Wildfowl 53:167–203

Drent RH, Daan S (1980) The prudent parent: energetic adjustments

in avian breeding. Ardea 68:225–252

Franson JC, Hollmen T, Poppenga RH, Hario M, Kilpi M (2000)

Metals and trace elements in tissues of common eiders (Soma-teria mollissima) from the Finnish archipelago. Ornis Fenn

77:57–63

Franson JC, Hollmen T, Hario M, Kilpi M, Finley DL (2002) Lead

and delta-aminolevulinic acid dehydratase in blood of common

eiders (Somateria mollissima) from the Finnish archipelago.

Ornis Fenn 79:87–91

Friend M, McLean RG, Dein FJ (2001) Disease emergence in birds:

challenges for the twenty-first century. Auk 118:290–303

Goudie RI, Brault S, Conant B, Kondratyev AV, Petersen, MR,

Vermeer K (1994) The status of seaducks in the North Pacific

Rim: toward their conservation and management. Trans 59th N

Am Wildl Nat Res Conf, pp 27–49

Goudie RI, Robertson GJ, Reed A (2000) Common eider (Somateriamollissima). In: Poole A, Gill F (eds) The birds of North

America, no. 546. The Birds of North America, Philadelphia,

pp 1–32

Grosbois V, Thompson PM (2005) North Atlantic climate variation

influences survival in adult fulmars. Oikos 109:273–290

Hanninen J, Vuorinen I, Hjelt P (2000) Climatic factors in the

Atlantic control the oceanographic and ecological changes in the

Baltic Sea. Limnol Oceanogr 45:703–710

Hario M, Rintala J (2006) Fledgling production and population trends

in Finnish common eiders (Somateria m. mollissima)—evidence

for density dependence. Can J Zool 84:1038–1046

Hario M, Selin K (1988) Thirty-year trends in an eider population:

timing of breeding, clutch size, and nest site preferences. Finn

Game Res 45:3–10

Hario M, Selin K (2002) Cohort-specific differences in reproductive

output in a declining common eider Somateria mollissimapopulation. Danish Rev Game Biol 16:33–45

Hario M, Rintala J, Selin K (2005) Fledgling production as the factor

regulating the population of the common eider (in Finnish with

English summary). Suomen Riista 51:6–26

Harvell CD, Kim K, Burkholder JM, Colwell RR, Epstein PR, Grimes

DJ, Hofmann EE, Lipp EK, Osterhaus ADME, Overstreet RM,

Porter JW, Smith GW, Vasta GR (1999) Emerging marine

diseases—climate links and anthropogenic factors. Science

285:1505–1510

HELCOM (2007) Climate change in the Baltic Sea area. Thematic

Assessment 2007. http://www.havet.nu/?d=186&id=54640279

Hollmen T (2002) Biomarkers of health and disease in common

eiders (Somateria mollissima) in Finland. Doctoral thesis,

Department of Basic Veterinary Sciences, University of Helsinki

Hollmen T, Franson JC, Docherty DE, Kilpi M, Hario M, Creekmore

LH, Petersen MR (2000) Infectious bursal disease virus

antibodies in eider ducks and herring gulls. Condor 102:688–691

Hollmen T, Franson JC, Hario M, Sankari S, Kilpi M, Lindstrom K

(2001) Use of serum biochemistry to evaluate nutritional status

and health of incubating common eiders (Somateria mollissima)

in Finland. Physiol Biochem Zool 74:333–342

Hollmen T, Franson JC, Kilpi M, Docherty DE, Hansen WR, Hario M

(2002) Isolation and characterization of a reovirus from common

eiders (Somateria mollissima) from Finland. Avian Dis 46:478–

484

Hurrell JW (2006) North Atlantic Oscillation (NAO) indices infor-

mation. Available from http://www.cgd.ucar.edu/cas/jhurrell/

indices.html. Accessed 11 Jan 2006

Hurrell JW, Kushnir Y, Ottersen G, Visbeck M (2003) An overview

of the North Atlantic Oscillation. In: Hurrell JW, Kushnir Y,

Ottersen G, Visbeck M (eds) The North Atlantic Oscillation:

climate significance and environmental impact. Geophys

Monogr Ser 134:1–35

Ihaka R, Gentleman R (1996) R: a language for data analysis and

graphics. J Comput Graph Stat 5:299–314

Kats RKH, Drent RH, Ens BJ, Duiven P, Swennen C, Van der Meer J

(2007) Rise and fall of the nesting population of the Common

Eider Somateria mollissima in the Netherlands since 1906: a

demographic reconstruction distinguishing between catastrophic

mortality events and non-breeding. Biol Conserv (to be

submitted)

Kery M, Madsen J, Lebreton JD (2006) Survival of Svalbard pink-

footed geese Anser brachyrhynchus in relation to winter climate,

density and land-use. J Anim Ecol 75:1172–1181

Kokko H (2001) Optimal and suboptimal use of compensatory

responses to harvesting: timing of hunting as an example. Wildl

Biol 7:141–150

Krementz DG, Hines JE, Caithamer DF (1996) Survival and recovery

rates of American eiders in eastern North America. J Wildl

Manage 60:855–862

Lack D (1967) The natural regulation of animal numbers. Clarendon

Press, Oxford

Laurila T, Hario M (1988) Environmental and genetic factors

influencing clutch size, egg volume, date of laying and female

weight in the common eider Somateria mollissima. Finn Game

Res 45:19–30

Lebreton JD, Burnham KP, Clobert J, Anderson DR (1992) Modeling

survival and testing biological hypotheses using marked animals:

a unified approach with case-studies. Ecol Monogr 62:67–118

Nichols JD (1991) Responses of North American duck populations to

exploitation. In: Perrins CM, Lebreton JD, Hirons GJM (eds)

Bird population studies. Oxford University Press, Oxford,

pp 498–525

Noer H (1991) Distributions and movements of eider Somateriamollissima populations wintering in Danish waters, analysed

from ringing recoveries. Danish Rev Game Biol 14(3):1–32

Noer H, Clausager I, Asferg T (1995) The bag of eider Somateriamollissima in Denmark 1958–1990. Danish Rev Game Biol

14(5):1–24

Oedekoven CS, Ainley DG, Spear LB (2001) Variable responses of

seabirds to change in marine climate: California current, 1985–

1994. Mar Ecol Progr Ser 212:265–281

Paradis E, Baillie SR, Sutherland WJ, Gregory RD (2002) Exploring

density dependent relationships in demographic parameters in

populations of birds at a large spatial scale. Oikos 97:293–307

Pinheiro JC, Bates DM (2000) Mixed-effects models in S and

S-PLUS. Springer, New York

Pollock KH, Nichols JD, Brownie C, Hines JE (1990) Statistical

inference for capture-recapture experiments. Wildl Monogr

107:1–98

Pradel R (1993) Flexibility in survival analysis from recapture data:

handling trap-dependence. In: Lebreton J-D, North PM (eds)

Marked individuals in the study of bird population. Birkhauser,

Berlin, pp 29–37

Prevot-Julliard A-C, Lebreton J-D, Pradel R (1998) Re-evaluation of

adult survival of black-headed gulls (Larus ridibundus) in

presence of capture heterogeneity. Auk 115:85–95

Oecologia (2009) 159:747–756 755

123

Reed A (1975) Migration, homing, and mortality of breeding female

eiders Somateria mollissima dresseri of the St Lawrence estuary,

Quebec. Ornis Scand 6:41–47

Ronka MTH, Saari CLV, Lehikoinen EA, Suomela J, Hakkila K

(2005) Environmental changes and population trends of breeding

waterfowl in northern Baltic Sea. Ann Zool Fenn 42:587–602

Sandvik H, Erikstad KE, Barrett RT, Yoccoz NG (2005) The effect of

climate on adult survival in five species of North Atlantic

seabirds. J Anim Ecol 74:817–831

Schaub M, Gimenez O, Schmidt BR, Pradel R (2004) Estimating

survival and temporary emigration in the multistate capture-

recapture framework. Ecology 85:2107–2113

Stearns S (1992) The evolution of life histories. Oxford University

Press, Oxford

Stenseth NC, Mysterud A (2005) Weather packages: finding the right

scale and composition of climate in ecology. J Anim Ecol

74:1195–1198

Swennen C (2002) Development and population dynamics of

common eider Somateria mollissima colonies in the Netherlands.

Danish Rev Game Biol 16:63–74

Wakeley JS, Mendall HL (1976) Migrational homing and survival of

adult female eiders in Maine. J Wildl Manage 40:15–21

Westerbom M (2006) Population dynamics of blue mussels in a

variable environment at the edge of their range. Doctoral thesis,

Faculty of Biosciences, Department of Biological and Environ-

mental Sciences, University of Helsinki

White GC, Burnham KP (1999) Program MARK: survival estimation

from populations of marked animals. Bird Study 46[Suppl]:

120–138

Williams BK, Nichols JD, Conroy MJ (2002) Analysis and manage-

ment of animal populations. Academic Press, New York

Wilson HM, Flint PL, Moran CL, Powell AN (2007) Survival of

breeding pacific common eiders on the Yukon–Kuskokwim

Delta, Alaska. J Wildl Manage 71:403–410

Yoccoz NG, Erikstad KE, Bustnes JO, Hanssen SA, Tveraa T (2002)

Costs of reproduction in common eiders (Somateria mollissima):

an assessment of relationships between reproductive effort and

future survival and reproduction based on observational and

experimental studies. J Appl Stat 29:57–64

756 Oecologia (2009) 159:747–756

123