Skip to main content
Advertisement
Browse Subject Areas
?

Click through the PLOS taxonomy to find articles in your field.

For more information about PLOS Subject Areas, click here.

  • Loading metrics

Relationships between Long-Term Demography and Weather in a Sub-Arctic Population of Common Eider

Abstract

Effects of local weather on individuals and populations are key drivers of wildlife responses to climatic changes. However, studies often do not last long enough to identify weather conditions that influence demographic processes, or to capture rare but extreme weather events at appropriate scales. In Iceland, farmers collect nest down of wild common eider Somateria mollissima and many farmers count nests within colonies annually, which reflects annual variation in the number of breeding females. We collated these data for 17 colonies. Synchrony in breeding numbers was generally low between colonies. We evaluated 1) demographic relationships with weather in nesting colonies of common eider across Iceland during 1900–2007; and 2) impacts of episodic weather events (aberrantly cold seasons or years) on subsequent breeding numbers. Except for episodic events, breeding numbers within a colony generally had no relationship to local weather conditions in the preceding year. However, common eider are sexually mature at 2–3 years of age and we found a 3-year time lag between summer weather and breeding numbers for three colonies, indicating a positive effect of higher pressure, drier summers for one colony, and a negative effect of warmer, calmer summers for two colonies. These findings may represent weather effects on duckling production and subsequent recruitment. Weather effects were mostly limited to a few aberrant years causing reductions in breeding numbers, i.e. declines in several colonies followed severe winters (1918) and some years with high NAO (1992, 1995). In terms of life history, adult survival generally is high and stable and probably only markedly affected by inclement weather or aberrantly bad years. Conversely, breeding propensity of adults and duckling production probably do respond more to annual weather variations; i.e. unfavorable winter conditions for adults increase probability of death or skipped breeding, whereas favorable summers can promote boom years for recruitment.

Introduction

Global climate change is predicted to increase the Earth’s mean annual temperature by 2–4°C in the next decades [1], [2] and influence the frequency of weather extremes, such as severe storms and droughts. The resulting impacts on animal populations are likely to be significant at both regional and global scales [3], [4], [5], [6], [7], [8], [9].

In several migratory bird species, the timing of spring migration and subsequent initiation of breeding have been shown to occur progressively earlier in response to weather changes [6], [7], [10], [11]. Among long-lived seabirds, changes in local weather conditions have also been shown to influence nest initiation dates [12] and the subsequent timing of prey availability for nestlings [13], chick growth and fledging rates [14], [15] and adult survival rates [16]. Extreme weather events can also strongly influence populations [17], [18], [19], although their relative rarity can make their effects difficult to quantify. Among bird populations, prolonged periods of severe weather, successive harsh winters, prolonged droughts or episodic weather events such as storms or hurricanes can directly influence breeding success or mortality rates [20], [21]. Altered frequencies of episodic weather could thus be a key mechanism affecting population-level responses to climatic changes.

Arctic and sub-arctic nesting species are particularly vulnerable to global climate change because climate will change disproportionately more at the highest latitudes [21], [22]. The short summer at northern latitudes predisposes species to complete breeding in a particularly short time and renesting potential is more limited for Arctic-nesting species than it is for species nesting at temperate latitudes [23]. For long-lived species with high adult survival and low reproductive output, the long-term data series which are necessary to reveal the effects of weather conditions on fitness are rarely available.

Common eider Somateria mollissima in Iceland nest under sub-arctic weather conditions: and thus, can experience relatively pronounced weather fluctuations [24], [25]. Previous studies have indicated that weather conditions in the months prior to nesting can influence breeding numbers, first nest initiation dates, nest site selection, clutch sizes and nest success [26], [27], [28], [29]. Weather conditions can also influence eider body condition and phenology because: 1) this species uses endogenous reserves for incubation [30], [31]; 2) is a short-distance migrant, both breeding and wintering at high latitudes [26], [27], [32] and so is exposed to high latitude weather conditions year round; and 3) environmental variation (such as weather) can induce physiological changes which can manifest as differing breeding trade-offs, such as nest site selection, nest desertion, yolk hormone levels and immune function [31], [32], [33]. Previous studies have shown that the North-Atlantic Oscillation index [34] (hereafter NAO) is positively correlated with body condition of breeding female common eider and the proportion of juveniles in hunting bags [31], [35]. These correlations with regional-scale NAO indices are likely to result from local-scale effects of weather conditions on duckling productivity and/or recruitment but the details of these relationships are unknown. Lastly, effects of weather may vary within breeding seasons; unfavorable (cold/wet or warm/dry) conditions early in spring may negatively affect nest success whereas favorable conditions in late spring (warm/wet) can positively affect nest success [29].

We collated information on nest numbers for 17 different eider colonies spanning 30–100 years. Our objective was to determine: 1) the spatial scale at which weather conditions may be associated with nest numbers of eiders within and between colonies, i.e. determine if relationships with weather occur consistently in all colonies; 2) whether local weather conditions influence eider nest numbers directly (effects within same year, i.e. lag = 0) or via recruitment (lagged effects 2–3 years later); and 3) the effects of rare episodic weather events on nest numbers in subsequent years. Improved understanding of the potential significance of episodic or extreme weather events to bird numbers is particularly important in light of present climate change predictions. These long-term data span generations of the long-lived common eider and encompass a wide range of weather conditions. Thus, these data provide an unusual opportunity to explore the short and long-term effects of both rare extreme weather events and more subtle weather variation on population processes at different spatial scales.

Climatic models for the Arctic and sub-Arctic regions generally not only predict increased mean temperatures and precipitation, but also more variation in these parameters [36]. Regarding predictions for effects of weather on eiders, we paid special attention to two possible scenarios, both of which are predicted for Iceland in future climate predictions for the next 50–100 years [37]. Firstly, we considered increased winter mildness, which may offer a mixed blessing for common eiders [38]: while common eider females arrive in better body condition at breeding grounds following warmer winters [35], their favorite winter food, blue mussels (Mytilus edulis) provide a better food supply during colder winters [39]. Secondly, we considered the effect of summer wetness on recruitment, because wet summers are negative for duckling production via increased predation [40]. Therefore we predicted 1) that nest numbers increased following milder winters because of the generally better body condition of females; or 2) that nest numbers decreased following milder winters because of reduced food supply; 3) that a negative effect on recruitment was observed in the abundance of nesting females 2–3 years after wet summers; 4) that aberrantly cold years or harsh winters had a greater effect on common eider nesting abundance than more normal variation in weather conditions.

Methods

Ethics Statement

We used existing data, which were sent to us from eider farmers at 17 locations in southwest, west, northwest and north Iceland. Common eider are a wild and free-ranging species in Iceland, the down is collected from the nests during late incubation or after the birds have left the colonies [41]. Down collection during late incubation does not affect incubation temperatures [42]. Eider farming is regulated by Icelandic law no. 84/1989, for the purpose of which common eider is protected by law (currently no. 64/1994) from hunting and egg collection. The common eider has been completely protected from hunting in Iceland since 1849 and from egg collection since 1787 because of the economic importance of down collection [41]. However, the effects of the protection act on population dynamics, or any other population changes during the 18th or 19th century have not been quantified for this population.

Study Area

Climate.

Iceland has a relatively mild climate relative to its latitude because of the tempering effect of Atlantic Ocean currents, which flow along the southern and the western coast [25]. However, the climate is sensitive to changes in storm occurrence with mild Atlantic air coming into contact with cooler Arctic air. Rainfall generally is higher in the southern and western parts than in the northern part of Iceland. A positive North Atlantic Oscillation Index (NAO), which indicates a prevailing western wind in the Northern Atlantic, often means increased precipitation whereas NAO has no relationship with temperature in Iceland [43].

The average temperature in Iceland increased by 0.7°C per century since the early 1800s [37]. The warming during the 20th Century was comparable to the global climate change, although since 1975, annual mean temperature has increased by 0.35°C, which is somewhat greater than the global average increase. It is predicted that the annual mean temperature will increase by 1°C (0 to 2°C) by 2050. Precipitation in Iceland likely will increase during the 21st century by 0.4–0.8% per decade or 2–3% per every 1°C of increased temperatures [37]. Predicted future warming will be most pronounced during the winter but least during the summer.

Study species.

The common eider is Iceland’s largest waterfowl population and not listed on the Red List (bird species listed as LR by IUCN standards, or of greater concern) issued by the Icelandic Institute of Natural History [44]. The estimated breeding population size is 300.000 pairs, with a wintering population of 847,000 (566,000–1,127,000, 95% confidence limits) individuals [44], [45]. Nest initiation begins 1 May at the earliest but the last nests are initiated in mid-June. Clutch size generally is 3–5 eggs which is similar to that in other populations [26], [41], [42], [46], [47], [48]. Earlier nesting is related to increased clutch sizes [26], [35], [49]. In Iceland, male eiders remain with females by their nests into mid-incubation, unlike that reported for other populations [50]. Many common eider colonies have been husbanded for eider down collection since before the 19th century, often maintained by the same families who have kept records of numbers of nests throughout the 20th century.

Data Collection

We collated data on the number of nests per year (hereafter breeding numbers) from 17 common eider colonies from southwest, northwest and north Iceland (Fig. S1). The length of the time series differed between colonies; the longest spans 101 years (1906–2007) and the shortest spans 26 years (1981–2007) (Table 1). Nests were counted in a standardized way by experienced observers (eider farmers) who survey in great detail in order to ensure that every nest is located, as the eider down that they gather from the nests is a source of important revenue.

thumbnail
Table 1. Common eider colonies, location, series length (years), relevant weather stations and covariance structures used to estimate correlations between breeding numbers and local weather conditions, for common eider Somateria mollissima breeding in Iceland.

https://doi.org/10.1371/journal.pone.0067093.t001

All data come from colonies which are used for eider down collection. The high revenues available from eider down mean that farmers typically undertake management aimed at encouraging high nesting densities, such as controlling numbers of avian and mammalian predators, and providing nest shelters. The extent of management activities differs between colonies, regions, owners or care-takers, and generations of owners, and depends on local conditions. However, there is no indication that variation in colony management varies systematically with weather conditions.

Statistical Procedures

Our analyses involved five types of statistical procedures. We primarily used time-series analyses, including 1) autoregressive integrated moving averages analysis; 2) pair-wise cross-correlation analysis; and 3) impact analysis with ARIMA noise models and deterministic intervention functions. The first two were used to determine if there were correlations between time-series of weather and nest counts. Impact analysis was used to determine if individual years caused marked changes in the nest counts, such as turns, interventions or new trends. Prior to running the time-series analysis, we used 4) principle components analysis (PCA) to reduce the number of weather variables and create linear indices (PC scores) of types of seasons (for example, “rainy summers” or “cold, windy winters”). Lastly, we used 5) false discovery rates significance thresholds (which lowers P-value thresholds below α = 0.05, scaled with number of comparisons made) to consider all correlations in unison to avoid reporting spurious results.

Our analysis of population synchrony and correlations between weather and breeding numbers followed a time-series approach [51], using autoregressive integrated moving average analysis developed by Box and Jenkins (PROC ARIMA) [51], [52]. This method requires that series are de-trended or differenced so that each series has a constant dispersion around a single mean (hereafter stationarity). We first tested series for non-stationarity using the Dickey-Fuller test and found that all series only became stationary following differentiation [51]. In most cases, the models fit a given series without a covariance structure. Where needed, as indicated by examination of autocorrelation functions, the covariance structures used were either, first- and second order autoregressive (AR1,1 & AR1,2 ) or first-order moving averages (MA1,1) (Table 1).

Population synchrony.

We assessed the possibility of population synchrony in our data since we used 17 colonies between 4 and 330 km apart, and a high level of synchrony could indicate the presence of a population driver operating on a larger scale, such as weather conditions. Following Ranta et al. [53] and Koenig [54] we used a comparison of the temporal dynamics of common eider nests in the 16 colonies spanning 1977–2007, i.e. the 31 years in which all colonies had available data. We then made all pair- wise comparisons, using cross-correlation analysis with time lag = 0 among the 16 colonies, yielding 120 cross-correlation coefficients. We used PROC ARIMA in SAS to obtain cross-correlation coefficients between each pair of colonies. To estimate the degree of synchrony, we: 1) compared our distribution of correlation coefficients (R) to those of Ranta et al [53]; and 2) plotted a Mantel graph [54], where distance between all pairs of colonies are plotted against the respective correlation coefficients. We interpreted results as follows: correlation coefficients 0.5 or higher were interpreted as indicating a high degree of synchrony between the respective colonies, whereas values less than 0.3 (P = 0.05 at approximately R = 0.35) were interpreted as indicating relatively little synchrony between colonies. Note that a significant relationship between distance (km) and correlation coefficients also would suggest synchrony in the data, but that a lack of such relationship does not necessarily indicate an absence of synchrony.

Relating weather to breeding numbers.

We examined the effects of variation in local weather conditions on breeding numbers by analyzing cross-correlations with indices of local weather [26], [28], and the regional-scale mid-winter NAO-index, in PROC ARIMA. In order to explore the effects of weather conditions during each of the four seasons, monthly averages of local weather data (temperature, precipitation, atmospheric pressure and wind speed) were averaged for each season (winter: December-March; spring: April-May; summer: June-September; autumn: October-November) and then indexed using a principal components analysis (PCA). The PCA not only reduced number of variables within each season but, more importantly, accounted for combined effects and interplay of weather variables within each season [26], [28], [55]. PC scores that explained ≥25% of the variation in the four weather variables were included in subsequent correlations with eider demography.

We used the weather station in Stykkishólmur, West Iceland, which is representative for all of Iceland [24], [56] and most of our data (11 of 17 colonies) come from nearby colonies in the bay of Breiðafjörður, West Iceland (Fig. S1). However, we also examined relationships using weather data in the vicinity of six colonies furthest from Stykkishólmur (see Table 1). Weather recording in Stykkishólmur began in 1846 [25] but data on wind speed were not available until spring of 1949. Thus, our correlation analysis of local weather parameters included only temperature, precipitation and atmospheric pressure for these periods (Table 1). For the 101 year long Brokey Islands series, we compared analyses of: 1) the period 1906–2007 with three weather variables; and 2) two analyses: 1901–1950 with three weather variables and 1951–2007 with four weather variables (Table 1). These approaches yielded very similar results and thus the first results from 1906–2007 are presented for Brokey Islands.

Time series starting at similar times were analyzed from a common starting date (1961; 5 series, 1977; 10 series, Table 1) for consistency. Upon detecting a significant cross-correlation at the 5% level for at least one of 17 colonies, we used the corresponding P-values from all the colonies to calculate the False Discovery Rate significance thresholds (FDR) [57] against which P-values of tests were evaluated. Here, we present findings from the classical one-stage method for FDR, following [57]. Finally, we used P-values 1) from both available colonies when calculating FDRs for 1900–1930, 2) from all six available colonies when calculating FDRs for 1961–2007 and 3) from all 16 available time-series when calculating FDR for the period 1977–2007.

Time-lagged relationships between weather and breeding numbers.

Common eiders start breeding at 2–3 years of age [58] and female common eiders are highly philopatric to the maternal breeding area [59], [60]. Consequently, changes in the size of breeding colonies, as a consequence of weather effects on duckling production, may not be apparent for up to three years. Based on the recruitment age of the common eiders, we considered 0 (no lag), 1, 2, and 3 year lagged effects of local weather on breeding numbers, to explore effects of weather on recruitment [27]. We saw no biological reason for weather to be correlated with eider numbers at lag 4–5 years or later.

Impact analysis of severe weather events.

Episodic weather events (i.e. severe or unusual weather that persists for a limited length of time) can potentially affect breeding numbers for years after a weather event. Prolonged periods of severe weather (i.e. harsh winters or droughts) or episodic weather events may cause widespread breeding failure or mortality and can result in population declines [20]. However, such single, possibly rare, events may not be indicated by a correlation analysis on a long time-series. Therefore, we used impact analysis on breeding numbers, following [51] who considered effects of single events (interventions, or “shocks to the system” [61]). For example, the winter of 1918 was the most catastrophic weather event during the 20th Century in Iceland. Impact analysis assesses the response in a time-series to discrete events or interventions that occur perhaps once during the span of a time-series. Impact response models are formulated as a regression function, with a response series as a dependent variable, and the regression model independent variables comprised by an ARIMA noise model and a deterministic (dummy variable for the impact event) intervention function [51].

Here, years of potential impact are coded as 1, whereas all other years are coded as zeros. We selected years for impact analysis based on NAO and historical years. For the NAO, we considered using the standard deviation (SD), multiplied by 2, which is a common practice to select “extreme” values. The 2×SD was 3.87 for 1990–2007 (the combined study period). However, when we compared NAO years with other historical weather events, we noted that NAO years of ≈3.3 to ≈3.7 coincided with these other weather events and thus, generally were remarkable enough to be considered. Thus, we settled on NAO values of 3.3 or greater to be inserted as impact years (1903, 1983, 1989–1990, and 1992) as well as negative values of −3.3 or lower (1917, 1936, 1963, 1969, and 1996). We then added remaining historical weather events [25]; the very cold year of 1918, the years 1965–1971 & 1979 during which sea-ice was unusually close to Iceland, and the warmer years of 1928–1946 [56]. Successive impact years (for example 1917–1918 and 1994–1995) were treated as a single impact event.

Effects of density-dependence.

Lastly, we tested for possible density-dependence in breeding numbers by examining correlations between breeding numbers in year y with those in years y+1 to y+4. According to this approach, a significant cross-correlation would indicate that density-dependence was operating within each of the colonies. As before, we used PROC ARIMA to test for these relationships.

Results

Inter-annual synchrony between colonies was low. Of 120 pair-wise correlations, 88 (73%) had a cross-correlation coefficients <0.3 (Fig. 1a), indicating low synchrony between colonies. Of the remaining 32 cross-correlation coefficients, 26 were significant at the α = 0.05 level (cross-correlation coefficients ≥0.35), indicating high synchrony between colonies. The mean correlation coefficient was 0.213; with the highest correlation coefficient = 0.666. The Mantel graph (Fig. 1b) showed a highly variable distribution of correlations with respect to distance. In itself, the lack of a correlation did not mean that synchrony is low. Due to the overall low synchrony between Icelandic eider colonies, we did not pool any of the 17 time-series.

thumbnail
Figure 1. Frequency distribution of common eider population synchrony.

Synchrony was measured by pair-wise cross-correlation coefficients of breeding numbers with lag = 0, among 16 colonies (120 correlations) in Iceland 1977–2007. Low synchrony was inferred because of 120 pair-wise correlations, 88 (73%) had a cross-correlation coefficient <0.3, which were not significant at the α = 0.05 level. The mean correlation coefficient was 0.213; with the highest correlation coefficient = 0.666.

https://doi.org/10.1371/journal.pone.0067093.g001

The trends in numbers of breeding common eiders across the 17 colonies were variable, but several colonies showed a trend for increasing numbers after the 1960s or 1970s and also experienced declines in numbers from the 1990s onwards (Fig. 2). The two oldest time series, Rifgirðingar and Brokey Islands began in 1901 and 1906 respectively, and both remained stable at around 500–1000 pairs until 1918 when both populations crashed to ∼60% of their previous peak number before steadily increasing again until 1930 (Fig. 2).

thumbnail
Figure 2. Variation in the numbers of breeding common eider Somateria mollissima in 17 colonies in Iceland.

Breeding numbers are shown during (A) 1901–2007, (B) 1963–2007, (C, D, E) 1978–2007. Y-axis shows breeding numbers which were standardized by setting the first year of each series at 100%. Note the differing scales on y- and x axis.

https://doi.org/10.1371/journal.pone.0067093.g002

Cross-correlations of Breeding Numbers with Local Weather

We found only two cross-correlations between local weather and breeding numbers within 1 year (lag = 0). Firstly, at Rifgirðingar, there was a significant correlation between summer weather conditions and breeding numbers (Summer-PC2, t = 2.30, P = 0.02, FDR-derived significance threshold = 0.025), with warm previous summers (positive PC scores, Table S1) being associated with greater breeding numbers (Fig. 3A) and cooler previous summers associated with decreased breeding numbers. Between 1919 and 1930, the population at Rifgirðingar had not recovered from the sharp decline that occurred in 1918 (see also impact analysis below). Secondly, in Svefneyjar (but not in any of the other colonies, including neighbouring colonies) final models for breeding numbers included a positive correlation with autumn weather (Autumn-PC1, t = 2.78, P = 0.0054, FDR-derived significance threshold = 0.0071). Autumn-PC1 indicates that a relatively warm and calm autumn prior to breeding (loadings 0.74 on temperature and −0.61 on wind speed, Appendix 2) was positively correlated with breeding numbers (Fig. 3B), whereas lowered breeding numbers followed cooler, windier autumns.

thumbnail
Figure 3. Relationship between annual changes in breeding numbers and weather.

In A) summer weather (see PC score loadings in Table S1) and breeding numbers (no. of nests) of common eider at Rifgirðingar, West Iceland 1900–1930; B) autumn weather and breeding numbers of common eider at Svefneyjar, West Iceland 1961–2007. Note the differing scales on y- and x axis.

https://doi.org/10.1371/journal.pone.0067093.g003

In Brokey Islands, the longest time-series, there were no relationships between variation in local weather conditions and breeding numbers. We included this series within all time periods considered in this study because of its length. Correlations with weather for Brokey Islands produced the same non-significant results, regardless of the periods considered or whether the PC scores were based on three weather variables (1906–2007 full series, 1906–1930 for comparison with Rifgirðingar, 1906–1949) or four weather variables (1950–2007).

We found few lagged (breeding numbers 1–3 years after weather) correlations between local weather (indexed by the PCA) and breeding numbers within colony. However, in three colonies, there were significant correlations between weather in a given summer and breeding numbers recorded 3 years following that summer (Fig. 4). There were two kinds of lagged correlations with summer weather: 1) a positive effect of Summer-PC2 (where positive PC scores indicated higher pressure, drier summers 3 years previously and negative scores indicated lower pressure, wetter summers; see Table S1) on breeding numbers three years later at Flatey 1977–2007 in Breiðafjörður (Fig. 4B t = 3.84, P = 0.0001, FDR-derived significance threshold = 0.0031) and 2) a negative effect of Summer-PC1 (where positive PC scores indicated warmer, calmer summers 3 years previously and negative scores indicated cooler, windier summers; see Table S1) on breeding numbers three years later at Þyrill 1977–2007 (t = −3.87, P = 0.0001, FDR-derived significance threshold = 0.0031) in southwest Iceland (Fig. 4A) and at Hrísey 1961–2007 in North Iceland (Fig. 4C); Hrísey (t = −2.79, P = 0.0053, FDR-derived significance threshold = 0.0063).

thumbnail
Figure 4. Relationship between summer weather and breeding numbers (no.of nests) 3 years later.

Shown are three common eider colonies in Iceland 1977–2007 (A, B) and 1961–2007 (C). Graphs on the left-hand side have Summer-PC2 on the X-axis whereas graphs on the right-hand side have Summer PC1 on the X-axis (see explanations on top of each column of graphs). Note the differing scales on y- and x axis.

https://doi.org/10.1371/journal.pone.0067093.g004

We found no correlations between midwinter NAO and breeding numbers for 1901–1930 (Rifgirðingar and Brokey Islands), 1906–2007 (Brokey Islands) or for series ranging 1961–2007 at lags 0–3 years, except for Fuglavík-Norðurkot where a 3-year positive lagged effect was found (t = 2.72, P = 0.0066, FDR-derived significance threshold = 0.0071) (Fig. 5). For the set of series ranging 1977–2007, we found no correlations at any time lags between NAO and breeding numbers for seven colonies.

thumbnail
Figure 5. Only one colony showed a relationship with NAO.

Relationship between mid-winter (DJFM) North Atlantic Oscillation index and breeding numbers (no. of nests) of common eider at Fuglavík-Norðurkot in southwest Iceland 1961–2007.

https://doi.org/10.1371/journal.pone.0067093.g005

Impact of Episodic Weather Events

Years considered as impact years were all numerical NAO values of 3.3 or greater (positive 1903, 1983, 1990, 1992, and 1995; negative 1917, 1936, 1963, 1969, and 1996) and remaining historical weather events in Iceland, including the very coldest year 1918 (see Fig. 3A), sea-ice years of 1965–1971 & 1979 and the warm years 1928–1946.

The impact analysis indicated a series of influential years with aberrant weather affecting breeding numbers in subsequent years (summary in Table 2). The very cold year of 1918 had a strong negative impact in both series that included the year (Rifgirðingar 1900–1930: t = −7.83, P = 0.0001, FDR-derived significance threshold = 0.025; and Brokey Islands 1906–2007: t = −4.48, P = 0.0001, FDR-derived significance threshold = 0.05).

thumbnail
Table 2. Summary of impact analyses for 17 common eider colonies in Iceland 1900–2007, which evaluated impact of certain years on breeding numbers of common eider within each colony.

https://doi.org/10.1371/journal.pone.0067093.t002

In series ranging 1977–2007, two years with extreme NAO values (1992, and 1996; NAO >3.3 or NAO<−3.3) were each indicated at least once but never in any of the longer series (see summary of impact year findings in Table 2). The year 1992 (Yearly NAO higher than 3.3) impacted 1) Flatey 1977–2007: t = 3.85, P = 0.0001 FDR-derived significance threshold = 0.0031; 2) Inneyjar 1977–2007: t = −3.54, P = 0.0004, FDR-derived significance threshold = 0.0063. The year 1996 (NAO lower than −3.3) impacted Þyrill 1977–2007: t = 3.82, P = 0.0001, FDR-derived significance threshold = 0.0031.

Test for Effects of Density-dependence

There were no relationships found between breeding numbers for year y and y+1, to y+4 within any of the colonies (FDR-derived significance thresholds P = 0.0063 for series 1958–2007 and longer, and P = 0.0031 for 1977–2007). Thus, given the method and data, we report that density-dependence generally was not present in breeding numbers in this study.

Discussion

Of our four predictions, one was met with a twist, i.e. in that there was a negative effect on nest numbers was observed 2–3 years following warmer, calmer (but not wet, as predicted) summers, albeit only for 2 of 17 colonies. Furthermore, one colony showed a positive relationship with drier summers 3 years earlier. Negative effects of impact years, aberrantly cold years or harsher winters were observed for the cold year 1918 and NAO years in the 1990s. In addition to these relationships, weather effects on common eider in Iceland are important with respect to arrival dates to the breeding colony and the subsequent clutch size [26]. The 3-year time lag suggests that duckling production may affect some local populations via recruitment, although the evidence for such a relationship was limited to 3 of 17 colonies. Weather conditions have been implicated as a synchronizing agent in spatially separated populations [62], but we found limited effects of weather on local population size among relatively unsynchronized sub-populations that breed within 330 km of each other. In this study, the lack of a spatial relationship between distance and correlation coefficients confirms the relatively minor effects of weather on eider nest numbers. We found no relationships between NAO and common eider in Iceland for 16 of 17 colonies, the lone exception being Fuglavík-Norðurkot 1977–2006, see also [27]. According to the impact analysis, the cold winter of 1918 had a significant, catastrophic impact on the eider population whereas the population seemed robust to most other episodic years considered.

Our findings indicate a low level inter-annual population synchrony for common eider in Iceland (lag = 0) during 1977–2007. In comparison, see: 1) Ranta et al. 1997 [53], where 50 of 55 (90.9%) of cross-correlation coefficients were 0.5 or higher and the mean cross-correlation coefficient was 0.716: these findings were interpreted as a high overall population synchrony; 2) Koenig 1999 [54] where most coefficients at shorter distances (left-hand side of figure) ranged 0.50 to 0.70, whereas our ranged 0.0 to 0.50. Separate treatment of colonies in our analysis also is likely to be appropriate because common eiders are highly philopatric with up to 98% of females returning to their natal colony every year [59], [60], [63], [64].

We found no relationships between breeding numbers in year t and later years. Given the method and data, this suggests that density-dependence is generally not evident in nest numbers of the Icelandic common eider population, although further studies, using other variables than nest counts are needed to confirm such a statement. Density dependence is highly relevant to interpretation of recruitment indices, particularly when a time-lag is used as a proxy for recruitment, such as in [27] and this study. If a population is under high density-dependence (for example, there are relatively high nest densities at the colony), low recruitment years matter little because recruits are competing for few open nest sites anyway. Conversely, if density matters little, local breeding numbers can be highly influenced by number of recruits.

A Locally Important, Lagged Effect of Weather on Recruitment

Effects on bird populations can be classified as 1) tub effects, which influence birds in the non-breeding season; and 2) tap effects, which influence breeding output [65]. To date, reported tub effects on common eider in Iceland operated mostly on arrival date and clutch size via body condition [26]. Here, we report possible, localized tap effects: a 3-year lagged correlation between weather and breeding numbers. We attribute these correlations to variation in duckling production or subsequent recruitment of 3 year old females into the respective breeding populations. Such a relationship should not be observed for any colony unless the number of breeding females is somewhat dependent on recruitment.

Duckling mortality is probably relatively more important than hatching failure in limiting breeding success prior to fledging in common eider [66]. Furthermore, females remain faithful to their natal colony and males pair with females from their natal colony more often than would be expected at random [59], [60]. Thus, colony-specific production will partly determine recruitment to a given colony because both sexes display philopatry, although males do so to a lesser extent or not at all [60], [66].

In Flatey, drier summers were positively related to breeding numbers three years later. At Þyrill and Hrísey, warmer, calmer summers were negatively related to breeding numbers three years later. In Scotland, predation on common eider ducklings was more probable in rainy, windy weather than in fine, calm weather and such weather accounted for over 90% of all variation in duckling survival during eight of nine years [40]. In the Netherlands, duckling production was positively related to high temperatures and “calm” weather at the end of May/beginning of June, although there was considerable scatter in the data [66]. Although weather may affect duckling production, food supply is probably the most important limiting factor [67], [68] and conditions during the breeding season can also influence subsequent return rates of adult breeders [68], [69].

Life History of Common Eider

Like other sea-ducks, the life-history strategy of common eider is essentially that of a long-lived species, with stable, high adult survival but low annual breeding output, although with a potential for high production during favorable years [29], [46], [70], [71,]. Adult survival generally is 0.87 or higher and constant over time for many populations [28], [72]) although not all of them [73]. Adult survival may have less impact on actual population dynamics than recruitment or duckling survival, which may vary considerably over time [50].

Costs of reproduction can become higher in unfavorable years and individual quality probably varies with environmental conditions [74]. Arctic environments provide common eider with a variable environment in most parts of the species range. However, such environmental stochasticity probably is relatively lower for Iceland, which is relatively sub-Arctic compared to Arctic Canada, Svalbard or Greenland. For example, initiation of nesting in Iceland is independent of sea ice, unlike that in Canada or Svalbard [49] and occurs in early or mid-May, rather than late May or June. Furthermore, duckling production generally seems high in Iceland; in Breiðafjörður 2007–2012 it was 0.4 ducklings/females or higher for all years except 2011 and was 0.6 or higher for three of six years (JEJ unpublished data).

During heightened exposures to avian cholera, large clutches were associated with lower survival of breeding females in East Bay, Nunavut [71]. While avian cholera is thus far unknown to Iceland, a similar but more general pattern of consequences of unfavorable conditions on breeding output may apply to this system. Common eiders are well known to skip breeding in some years, which can be seen as a strategy to preserve residual reproductive value, or to employ a trade-off of costs of current reproduction relative to reduced future opportunities or even reduced survival [46], [74]. In our data, non-breeding years are detectable within certain colonies, and many of those occur simultaneously among colonies. However, long-term implications of such years for population dynamics probably have been limited in Iceland, because during skip years, breeding numbers will decline sharply for one year, and generally return to the previous year’s number one year later, with the time-series usually continuing their previous trend for years to come. The source of environmental stochasticity that prompts such years remains unknown, although reduced food availability, unfavorable climate conditions or disease have been implicated as potential drivers [46] along with weather conditions.

Why is Weather Important in some Colonies – and not Others?

We observed low levels of inter-annual synchrony in the Icelandic common eider population, i.e. the likely absence of a Moran effect and other synchronizing agents, such as dispersal or a synchronized food supply [62]. Such low levels of synchrony indicate that local effects, perhaps via interactions of weather and local topography or resources, are more dominant in this population. The long life expectancy of this species could however mask changes because of high philopatry and the reproductive tactics among breeding females [46]. The lack of relationships between breeding numbers and weather variables indicates that eider colonies are robust in terms of inter-annual changes, perhaps because females can skip breeding in unfavorable years to return in later years.

Despite having limited effects on breeding numbers, weather can affect the arrival, behavior or physiology of common eider throughout the annual cycle [26], [27], [28], [35], [75]. Variation in individual decisions regarding movements, nest site selection, skipped breeding seasons and presence or absence of a floater population [38], [46], [76] may further contribute to inter-colony variation in demography. For common eider, weather probably affects: 1) adult survival but only during severe, impact years, most probably during harsher winters although it is doubtful that milder winters are entirely beneficial; 2) adult breeding propensity, winter via winter body condition or last year’s breeding conditions [31], [32], [33]. Furthermore, negative effects of rain or drought on duckling production would represent direct effects of weather on recruitment, although such effects are probably much localized [40].

Effects of weather may be dependent on adjacent topography. For instance, a dry summer may have quite different effects on the intertidal feeding habitat at low lying offshore islands, such as Flatey, in comparison with colonies close to mountain ranges, such as Þyrill or Hrísey, where there is abundant freshwater runoff from the mountains. Adjacent mountain ranges also may provide shelter from wind or influence amount of precipitation. During the duckling stage, shelter from wind is locally important in areas which face the open ocean (such as Fuglavík-Norðurkot, Flatey), whereas sheltered colonies inside fjords or inlets are often little affected. Lastly, breeding numbers are influenced by factors other than weather, including food availability, some of which may be weather-related. Furthermore, anthropogenic factors, such as protective efforts by the eider farmers, human depopulation events, changed farming practices, disturbance or farm abandonment can affect eider nest numbers.

Supporting Information

Figure S1.

(map) Numbers indicate locations of 17 colonies in Iceland which provided breeding numbers of common eider (Somateria mollissima) used in this study; numbers refer to numbers of colonies in Table 1. Names are the weather stations considered in this study.

https://doi.org/10.1371/journal.pone.0067093.s001

(TIF)

Table S1.

Principal components analysis of weather data from Stykkishólmur, West Iceland, for different periods where eider nests were counted in Iceland. The first two principal scores (PC) were used as indices of local weather. The highest loading for each variable in each analysis is indicated in bold.

https://doi.org/10.1371/journal.pone.0067093.s002

(DOC)

Acknowledgments

We thank the eider farmers and their families, (farms are listed in Table 1) who gave access to their nest count data, without their cooperation this study would not have been possible: Atli Vigfússon, Ásgeir Gunnar Jónsson, Ásdís Ásmundsdóttir, Bergur Hjaltalín, Páll Hjaltalín, Jón Jakobsson, Úlla Knudsen, Erla Friðriksdóttir, Friðrik Jónsson, Gissur Tryggvason, Hafsteinn Guðmundsson, Svanhildur Jónsdóttir, Daði Sigurþórsson, Sigurþór Guðmundsson, Ágúst Bjartmars, Zófónías Þorvaldsson, Smári Lúðvíksson, Ásrún Jóhannesdóttir, Ólafur K. Nielsen, Bjarni Hákonarson, Jóhannes G. Gíslason, Þorvaldur Björnsson, Þórður Örn Kristjánsson, and Sigurður K. Eiríksson. Michael D. Kaller advised us on statistics. We thank Sébastien Descamps, Jonas Knape and two anonymous reviewers for their helpful comments that greatly improved this manuscript.

Author Contributions

Conceived and designed the experiments: JEJ TGG JAG AG AP. Analyzed the data: JEJ. Wrote the paper: JEJ TGG JAG AG AP. Data collation: JEJ UKP.

References

  1. 1. ACIA (2005) Impacts of a warming climate: Arctic climate impact assessment. Cambridge: Cambridge University Press. 1042 p.
  2. 2. Hansen J, Sato M, Ruedy R, Lo K, Lea DW, et al. (2006) Global temperature change. Proceedings of the National Academy of Sciences 103: 14288–14293.
  3. 3. Parmesan C, Yohe G (2003) A globally coherent fingerprint of climate change impacts across natural systems. Nature 421: 37–42.
  4. 4. Anders A, Post E (2006) Distribution-wide effects of climate on population densities of a declining migratory landbird. Journal of Animal Ecology 75: 221–227.
  5. 5. Both C, Bouwhuis S, Lessels CM, Visser ME (2006) Climate change and population declines in a long-distance migratory bird. Nature 44: 81–83.
  6. 6. Jonzen N, Linden A, Ergon T, Knudsen E, Vik JO, et al. (2006) Rapid advance of spring arrival dates in long-distance migratory birds. Science 312: 1959–1961.
  7. 7. Gordo O (2007) Why are bird migration dates shifting? A review of weather and climate effects on avian migratory phenology. Climate Research 35: 37–58.
  8. 8. Both C, Van Turnhout CAM, Biljsma RG, Siepel H, Van Strien AJ, et al. (2010) Avian population consequences of climate change are most severe for long-distance migrants in seasonal habitats. Proceeding of the Royal Society London B 277: 1259–1266.
  9. 9. Jiguet F, Devictor V, Ottvall R, Van Turnhout C, Van der Jeugd H, et al. (2010) Bird population trends are linearly affected by climatic change along species thermal ranges. Proceeding of the Royal Society London B 277: 3601–3608.
  10. 10. Marra PP, Francis CM, Mulvihill RS, Moore FR (2005) The influence of climate on the timing and rate of spring bird migration. Oecologia 142: 307–315.
  11. 11. Gunnarsson TG, Gill JA, Atkinson PW, Gélinaud G, Potts PM, et al. (2006) Population-scale drivers of individual arrival times in migratory birds. Journal of Animal Ecology 75: 1119–1127.
  12. 12. Moe B, Stempniewicz L, Jakubas D, Angelier F, Chastel O, et al. (2009) Climate change and phenological responses of two seabird species breeding in the high-Arctic. Marine Ecology Progress Series 393: 235–246.
  13. 13. Watanuki Y, Ito M, Deguchi T, Minobe S (2009) Climate-forced seasonal mismatch between the hatching of rhinoceros auklets and the availability of anchovy. Marine Ecology Progress Series 393: 259–271.
  14. 14. Thompson PM, Ollason JC (2001) Lagged effects of ocean climate change on fulmar population dynamics. Nature 413: 417–420.
  15. 15. Gjerdrum C, Vallee AMJ, St Clair CC, Bertram DF, Ryder JL, et al. (2003) Tufted puffin reproduction reveals ocean climate variability. Proceedings of the National Academy of Sciences 100: 9377–9382.
  16. 16. Grosbois V, Thompson PM (2005) North Atlantic climate variation influences survival in adult fulmars. Oikos 109: 273–290.
  17. 17. Parmesan C, Root TL, Willig MR (2000) Impacts of extreme weather and climate on terrestrial biota. Bulletin of the American Meteorological Society 81: 443–450.
  18. 18. Frederiksen M, Daunt F, Harris MP, Wanless S (2008) The demographic impact of extreme events: stochastic weather drives survival and population dynamics in a long-lived seabird. Journal of Animal Ecology 77: 1020–1029.
  19. 19. van de Pol M, Ens BJ, Heg D, Brouwer L, Krol J, et al. (2010) Do changes in the frequency, magnitude and timing of extreme climatic events threaten the population viability of coastal birds? Journal of Applied Ecology 47: 720–730.
  20. 20. Newton I (1998) Population limitation in birds. San Diego: Academic Press. 597 p.
  21. 21. McFarlane Tranquilla L, Hedd A, Burke C, Montevecchi WA, Regular PM, et al. (2010) High Arctic sea ice conditions influence marine birds wintering in Low Arctic regions. Estuarine, Coastal and Shelf Science 89: 97–106.
  22. 22. Robinson SA (2009) Introduction: Climate change biology at the ends of the Earth-International Polar year special issue. Global Change Biology 15: 1615–1617.
  23. 23. Lepage DG, Gauthier G, Menu S (2000) Reproductive consequences of egg-laying decisions in snow geese. Journal of Animal Ecology 69: 414–427.
  24. 24. Einarsson MA (1984) Climate of Iceland. In: van Loon H, editor. World Survey of Climatology, 15, Climates of the Oceans. Amsterdam: Elsevier. 673–697.
  25. 25. Hanna E, Jónsson T, Box JE (2004) An analysis of Icelandic climate since the nineteenth century. International Journal of Climatology 24: 1193–1210.
  26. 26. Jónsson JE, Gardarsson A, Gill JA, Petersen A, Gunnarsson TG (2009) Seasonal weather effects on a subarctic capital breeder: common eiders in Iceland over 55 years. Climate Research 38: 237–248.
  27. 27. D’Alba L, Monaghan P, Nager R (2010) Advances in laying date and increasing population size in relation to climate change in an Icelandic eider colony. Ibis 152: 19–28.
  28. 28. Jónsson JE, Lúðvíksson S (2012) A choice of two different islands: is switching nest sites related to weather or increasing nesting density in Common Eider? Ornis Fennica 89: in press. Ornis Fennica Online early website.Available: http://www.ornisfennica.org/pdf/early/Jonsson.pdf Accessed 2013 May 21.
  29. 29. Iles DT, Rockwell RF, Matulonis P, Robertson GJ, Abraham KF, et al. (2013) Predators, alternative prey, and climate influence annual breeding 1 success of a long-lived sea duck. Journal of Animal Ecology 82: 683–693.
  30. 30. Klaassen M, Abraham KF, Jefferies RL, Vrtiska M (2006) Factors affecting the site of investment, and the reliance on savings for arctic breeders: the capital-income dichotomy revisited. Ardea 94: 371–384.
  31. 31. Descamps S, Yoccoz NG, Gaillard J-M, Gilchrist HG, Erikstad KE, et al. (2010) Detecting population heterogeneity in effects of North Atlantic oscillations on seabird body condition: get into the rhythm. Oikos 119: 1526–1536.
  32. 32. Love OP, Gilchrist HG, Descamps S, Semeniuk CAD, Bety J (2010) Pre-laying climatic cues can time reproduction to optimally match offspring hatching and ice conditions in an Arctic marine bird. Oecologia 164: 277–286.
  33. 33. Sénéchal É, Bety J, Gilchrist HG (2010) Interactions between lay date, clutch size, and postlaying energetic needs in a capital breeder. Behavioral Ecology 22: 162–168.
  34. 34. Hurrell JW, Kushnir Y, Visbeck M (2001) The North Atlantic Oscillation. Science 291: 603–605.
  35. 35. Lehikoinen A, Kilpi M, Öst M (2006) Winter climate affects subsequent breeding success of common eiders. Global Change Biology 12: 1355–1365.
  36. 36. Post E., Forchhammer M.C., Bret-Harte M.S., Callaghan T.V., Christensen T.R., et al. (2009) Ecological dynamics across the arctic associated with recent climate change. Science, 325, 1355–1358.
  37. 37. Björnsson H, Sveinbjörnsdóttir AE, Daníelsdóttir AK et al. (2008) Hnattrænar loftslagsbreytingar og áhrif þeirra á Íslandi – Skýrsla vísindanefndar um loftslagsbreytingar. Umhverfisráðuneytið. [Global climate change and its effects in Iceland. Report published by the Ministry of Environment, Iceland] (In Icelandic). Website of the Icelandic Ministry of the Environment.Available: http://www.umhverfisraduneyti.is/media/PDF_skrar/visindanefndloftslagsbreytingar.pdf Accessed 2013 May 21.
  38. 38. Ekroos JFox AD, Christensen TK, Petersen IK, Kilpi M, et al. (2011) Declines amongst breeding eider numbers in the Baltic/Wadden Sea flyway. Ornis Fennica 89: 81–90.
  39. 39. Beukema JJ (1993) Increased mortality in alternative bivalve prey during a period when the tidal flats of the Dutch Wadden sea were devoid of mussels. Netherlands Journal of Sea Research 31: 395–406.
  40. 40. Mendenhall VM, Milne H (1985) Factors affecting duckling survival of eiders Somateria mollissima in northeast Scotland. Ibis 127: 148–158.
  41. 41. Jónsson J (editor) (2001) Æðarfugl og æðarrækt á Íslandi. [Common eider and common eider husbandry in Iceland] Reykjavík: Skrudda. (In Icelandic). 528 p.
  42. 42. Kristjánsson TO, Jónsson JE (2011) Effects of down collection on incubation temperature, nesting behaviour and hatching success of common eiders (Somateria mollissima) in west Iceland. Polar Biology 34: 985–994.
  43. 43. Jónsson T (2010) Ísland og nao-fyrirbrigðið. Hver eru áhrif þess? [Effects of NAO in Iceland] (In Icelandic) Website of the Icelandic Meteorological Office. Available: http://www.vedur.is/vedur/frodleikur/greinar/nr/1803. Accessed 2013 May 21.
  44. 44. Institute of Natural History (2000) Válisti 2: Birds [Red List II: Birds] 103 p. (In Icelandic with an English Summary)
  45. 45. Gardarsson A (2009) Fjöldi æðarfugls, hávellu, toppandar og stokkandar á grunnsævi að vetri. [Numbers of common eider, long-tailed duck, red-breasted merganser and mallard, wintering on the coast of Iceland] Bliki 30: 49–54 (In Icelandic with an English Summary).
  46. 46. 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.
  47. 47. Robertson GJ (1995) Factors affecting nest site selection and nesting success in the common eider Somateria mollissima. Ibis 137: 109–115.
  48. 48. Erikstad KE, Tveraa T (1995) Does the cost of incubation set limits to clutch size in common eiders Somateria mollissima? Oecologia 103: 270–274.
  49. 49. Mehlum F (2012) Effects of sea ice on breeding numbers and clutch size of a high arctic population of the common eider Somateria mollissima. Polar Science 6: 143–153.
  50. 50. Iles D.T. (2012) Drivers of nest success in common eiders and impacts on population dynamics. M.S. thesis, Utah State University, Logan, Utah.
  51. 51. Yaffee R (2000) Introduction to time-series analysis and forecasting. Amsterdam: Academic Press. 528 p.
  52. 52. SAS Institute (2001) SAS/SYSTAT useŕs Guide. Version 8. SAS Institute Inc., Cary.
  53. 53. Ranta E, Kaitala V, Lindström J, Helle E (1997) The Moran effect and synchrony in population dynamics. Oikos 78: 136–142.
  54. 54. Koenig WD (1999) Spatial autocorrelation of ecological phenomena. Trends in Ecology and Evolution 14: 22–26.
  55. 55. Traylor JJ, Alisauskas RT (2006) Effects of intrinsic and extrinsic factors on survival of white-winged scoter (Melanitta fusca deglandi) ducklings. Auk 123: 67–81.
  56. 56. Jónsson T (2006) Hitafar á Íslandi eftir 1800. [Temperatures in Iceland from 1800]. (In Icelandic) Icelandic climate. Website of the Icelandic Meteorological Office. Available: http://www.vedur.is/loftslag/loftslag/fra1800/hitafar/. Accessed 2013 May 21.
  57. 57. Pike N (2011) Using false discovery rates for multiple comparisons in ecology and evolution. Methods Ecology Evolution 2: 229–332.
  58. 58. Hario M, Rintala J (2009) Age of first breeding in the common eider Somateria m. mollissima population in the northern Baltic Sea. Ornis Fennica 86: 81–88.
  59. 59. Swennen C (1990) Dispersal and migratory movements of eiders Somateria mollissima breeding in the Netherlands. Ornis Scandinavica 21: 17–27.
  60. 60. Tiedemann R, von Kistowski KG, Noer H (1999) On sex-specific dispersal and mating tactics in the common eider Somateria mollissima as inferred from the genetic structure of breeding colonies. Behaviour 136: 1145–1155.
  61. 61. Box GEP, Tiao GC (1975) Intervention analysis with applications to economic and environmental problems. Journal of the American Statistical Association 70: 70–79.
  62. 62. Liebhold A, Koenig WD, Bjørnstad ON (2004) Spatial synchrony in population dynamics. Annual Review of Ecology, Evolution, and Systematics 35: 467–490.
  63. 63. Wakeley JS, Mendall HL (1976) Migrational homing and survival of adult female eiders in Maine. Journal of Wildlife Management 40: 15–21.
  64. 64. Baillie SR, Milne H (1989) Movements of eiders Somateria mollissima on the east coast of Britain. Ibis 131: 321–335.
  65. 65. Sæther BE, Sutherland WJ, Engen S (2004) Climate influences on avian population dynamics. In: Møller AP, Berthold P, Fiedler W, editors. Advances in ecological research: birds and climate change, 35. Amsterdam: Elsevier. 185–209.
  66. 66. Swennen C (1991) Fledgling production of eiders Somateria mollissima in the Netherlands. Journal für Ornithologie 132: 427–437.
  67. 67. Gardarsson A, Einarsson A (1994) Responses of breeding duck populations to changes in food-supply. Hydrobiologia 280: 15–27.
  68. 68. Gardarsson A, Einarsson A (1997) Numbers and production of Eurasian wigeon in relation to conditions in a breeding area, Lake Myvatn, Iceland. Journal of Animal Ecology 66: 439–451.
  69. 69. Bustnes JO, Eriksstad KE (1993) Site fidelity in breeding common eider Somateria mollissima females. Ornis Fennica 70: 11–16.
  70. 70. Wilson H.M., Flint P.L., Moran, C.L. & Powell, A.N. (2007) Survival of breeding pacific common eiders on the Yukon-Kuskokwim Delta, Alaska. The Journal of Wildlife Management, 71, 403–410.
  71. 71. Descamps S, Gilchrist HG, Bety J, Buttler EI, Forbes MR (2009) Costs of reproduction in a long-lived bird: large clutch size is associated with low survival in the presence of a highly virulent disease. Biological Letters 5: 278–281.
  72. 72. Hario M, Mazerolle MJ, Saurola P (2009) Survival of female common eiders Somateria m. mollissima in a declining population of the northern Baltic Sea. Oecologia 159: 747–756.
  73. 73. Coulson JC (2010) A long-term study of the population dynamics of common eiders Somateria mollissima: why do several parameters fluctuate markedly? Bird Study 57: 1–18.
  74. 74. Robert A, Paiva VH, Bolton M, Jiguet F, Bried J (2012) The interaction between reproductive cost and individual quality is mediated by oceanic conditions in a long-lived bird. Ecology 93: 1944–1952.
  75. 75. Jónsson JE, Gardarsson A (2001) The pair formation in relation to climate: mallard, Eurasian wigeon and Eurasian teal wintering in Iceland. Wildfowl 52: 55–68.
  76. 76. Bolduc F, Guillemette M, Titman RD (2005) Nesting success of common eiders Somateria mollissima as influenced by nest-site and female characteristics in the Gulf of the St. Lawrence. Wildlife Biology 11: 273–279.