Introduction

Some bird species are determinate egg-layers, where addition or removal of eggs during the laying sequence does not influence the final number of eggs that the female lays. Indeterminate layers on the other hand can alter their final clutch size if eggs are added or removed. Whether a female is a determinate or indeterminate layer can influence responses to egg predation, nest parasitism and other events with consequences for reproductive success. The occurrence, circumstances and adaptive significance of these reproductive patterns have been a focus of research for a long time (e.g. Cole 1917; Klomp 1970; Kennedy 1991; Haywood 1993). A group of particular interest is waterfowl (Anatidae), where the mechanisms of clutch control have been much debated and may vary among species (e.g. Barry 1962; Lack 1968; Heusmann et al. 1980; Andersson and Eriksson 1982; Rohwer 1984; Alisauskas and Ankney 1992; Arnold et al. 2002).

Accumulating most or all resources for egg production before starting the clutch, the common eider Somateria mollissima is perhaps the most extreme capital breeder among ducks, and among birds as a whole (e.g. Parker and Holm 1990; Meijer and Drent 1999). Experiments in this species have reached differing conclusions as regards clutch size flexibility. Adding or removing eggs in nine clutches of each treatment category, Swennen et al. (1993) found no significant effect on the number of eggs laid by the female and concluded that eiders are determinate layers, clutch size being fixed before the first egg is laid. In a larger experimental study, Erikstad and Bustnes (1994) likewise found no response to egg removal. In nests where eggs were added, however, females laid significantly fewer eggs: on average 0.5 and 0.9 eggs fewer in 2 years with mean clutch sizes of 4.4 and 4.3 eggs, respectively. There is also evidence of cluch reduction in response to egg addition by parasitic females in some birds (e.g. Lyon 1988), including eiders (Waldeck et al., in preparation).

Does the lack of response to egg removals mean that eider females cannot increase their egg production? Erikstad and Bustnes (1994) pointed out that their removal of eggs only after the clutch already contained three eggs was perhaps too late for the female to respond by laying more eggs. An eider egg takes about 1 week to produce from the start of rapid follicle growth until laying (Alisauskas and Ankney 1992), and the estimated average interval between eggs is 28 h (Watson et al. 1993). The female may therefore have more scope for laying extra eggs if the loss occurs early, especially at the one-egg stage. However, this might lead to nest desertion. Swennen et al. (1993) and Erikstad and Bustnes (1994) did not remove first eggs.

Here, we study arctic common eiders S. mollissima borealis breeding in Svalbard, with high natural predation risk for the first egg in the clutch, before the female sits permanently at the nest (Mehlum 1991b; Robertson and Cooke 1993; Swennen et al. 1993; Hanssen et al. 2002; Andersson and Waldeck 2006; Öst et al. 2008). After predation of their first eggs, some females did not desert the nest but continued laying in it. This, combined with protein fingerprinting of egg albumen verifying that the same female continued laying in the nest, allowed us to test female responses and clutch-size consequences of partial predation from the one-egg stage.

Materials and methods

We studied eiders breeding on Prins Heinrich Island (called PH Island below), a 400 × 150 m open, low moraine island 1 km SE of Ny-Ålesund, Kongsfjorden, West Spitsbergen (79°N, 12°E). Vegetation is sparse or nonexistent, and most nests are in open gravel or moss, or in a few cases under wooden boxes from earlier experiments, or on beach gravel. Old nest bowls are common, offering many suitable nest sites (Bjørn and Erikstad 1994). From the pre-laying period in late May until the end of the laying period in late June 2007 and 2008, we searched the island thoroughly for active nests twice a day, finding most new clutches at the one-egg stage. In 2007, there was one started cluch (3 eggs) at our first search, 24 May. In 2008, we began searching on 26 May, and also on that date there was one started nest (3 eggs, by the same marked female). We sampled new eggs for albumen, marked them individually with a non-toxic felt pen, and measured their length and width with calipers to the nearest 0.1 mm. Nest position was recorded with GPS (Garmin GPS Map 76).

Eiders usually lay 3–6 eggs and start incubation after the second or third egg (Swennen et al. 1993; Goudie et al. 2000; Hanssen et al. 2002). In the Kongsfjord area, most nests contain 2–4 eggs, but there is much egg predation and most females have laid more eggs, as reflected in their larger number of ruptured follicles (Ahlén and Andersson 1970). The clutch is incubated for 22–26 days (Goudie et al. 2000; Erikstad et al. 1993; Hanssen et al. 2002), during which females seldom and only briefly leave the nest (Korschgen 1977; Parker and Holm 1990). This is probably an adaptation that reduces nest predation (Andersson and Waldeck 2006). The egg predators during this study were glaucous gull (Larus hyperboreus) and arctic skua (Stercorarius parasiticus), as revealed by our observations of eider nests from hides and by digital video cameras in the colony.

We used protein fingerprinting (Andersson and Åhlund 2001) for analysing conspecific nest parasitism among the ~170 eider females nesting on Prins Heinrich Island (numbers vary considerably between years depending, e.g., on ice conditions; Mehlum 1991a). From each egg, we took a non-destructive sample of albumen for analysis by isoelectric focussing electrophoresis. This technique produces individual-specific band patterns that allows discrimination between females (see Andersson and Åhlund 2001 and, for eiders, Waldeck et al. 2004). Based on presence/absence of 75 different albumen bands in each egg, we could therefore determine if the same female continued laying in the nest after its first egg was depredated. Protein fingerprinting of egg albumen also showed that conspecific brood parasitism is frequent in this population, and in some cases another female took over the nest (Waldeck et al., in preparation).

Nest predation in eiders decreases with increasing vegetation cover (Öst et al. 2008), and to reduce nest predation we covered the eggs with moss before leaving. Females returned to the nest soon after we left. Criscuolo (2001) found that female eiders may abandon their nests if caught early in the incubation period. We therefore waited for at least 5 days after a female laid her last egg before catching her, using a fishing pole to sneak a nylon loop over her head. We measured both wings (from the carpal joint to the tip of the longest primary), head (back of head to bill tip) and both tarsi, weighed her and sampled 0.1 ml of blood from a leg.

Female condition can perhaps influence whether she deserts the nest after predation of its first egg or continues laying in it. We study this possibility by comparing the condition of females in several ways, based on egg volume, body mass and laying date. We estimate egg volume from measurements (in cm) as 0.51 × length × breadth2 (see Robertson and Cooke 1993). Comparison of egg size beween the females that stay with those that desert after predation of their first egg is based on the volume of that egg. It is usually similar in size to the next two eggs in the clutch, whereas later eggs are progressively smaller (e.g. Swennen and Van der Meer 1992; Waldeck and Andersson 2006).

Schamber et al. (2008) tested several predictors of body fat reserves in five species of waterfowl and found that body mass alone is a suitable simple condition index. Eider females lose much mass during the incubation period, 1/3 or even more (e.g. Korschgen 1977). To compare the mass of females weighed at different stages, we therefore standardised female mass to day 10 after her clutch completion. The mass m x of a female weighed x days after laying her last egg is translated to 10-day mass m 10  = m x  − b (x − 10), where b is the regression coefficient in the linear regression of body mass on day (x).

In 125 non-parasitised nests, we followed egg-laying from the first egg until the clutch was full (incubated, no new egg added in the last 3 days). In 106 of these nests, there was no predation during the laying sequence. In the other 19 nests, there was partial predation early in the laying sequence, the first (n = 11), second (4) or both eggs (4) being taken by predators before any later eggs were added, but the female continued laying in the nest. Because nest predation in some cases led to desertion after the clutch was full, we could only catch and weigh 83 of the 106 and 12 of the 19 females, respectively. In addition, we weighed 42 other females caught at the nest for another study, and included them in a regression analysis of body mass decline in relation to day after clutch completion (which is hence based on 83 + 12 + 42 = 137 females). The 42 females had almost identical mean body mass as the others (“Results”).

Predation and clutch size may depend on year and clutch start day (see also Erikstad and Bustnes 1994). Analysing the effect of partial predation on clutch size, we therefore included start day as covariate and year as random factor in a mixed ANCOVA model (type III sum of squares). We then removed non-significant parameters to increase power (Quinn and Keough 2002). Unless otherwise explained, statistical analyses are done in SPSS version 17.0.

Results

In the 2 years, we observed a total of 79 nests with early egg predation (and without conspecific nest parasitism) out of 286 started nests (not all sampled). Sixty of the 79 females deserted after predation of their first egg, but 19 females continued laying in the same nest, as verified by protein fingerprinting of the eggs. In both years, the staying females on average laid more eggs than females in non-predated nests (Table 1; Fig. 1). The mean (±SE) for the 2 years combined are 3.71 ± 0.07 eggs (range 2–5) in 106 non-predated nests, and 4.42 ± 0.16 (range 3–6) in 19 partially predated nests, i.e. a difference of about 0.7 eggs. There was no significant difference in clutch size variance between these 19 partially predated and the 106 non-predated nests (0.48 and 0.55, respectively; F = 1.15, two-tailed P = 0.77).

Table 1 Clutch size and laying date of common eiders on Prins Heinrich Island, Svalbard, 2007 and 2008
Full size table
Fig. 1
figure 1

Clutch size (including depredated eggs) of common eider females on Prins Heinrich Island 2007–2008, in nests without egg predation (diamonds), with one egg depredated (squares) and with two eggs depredated (triangles), in relation to clutch initiation date (0 = 31 May) (some data points represent several nests). Both predation and initiation date had significant effects on clutch size (see main text), which increased with the number of eggs depredated

Full size image

In the full ANCOVA model (“Materials and methods”), the random factor (year) was far from significant (P = 0.32), so we pooled the 2 years. In the reduced model, the interaction predation × start day (P = 0.19) was also removed. (The estimated regression line for nests with two eggs depredated is not parallell with the other two lines in Fig. 1. There is, however, no substantial evidence for interaction, as the deviating two-egg line is based on just four data points, and the 95% confidence interval of its slope (b = 0.06 ± 0.28) includes the slopes of the two other lines. See Engqvist 2005 for consideration of interactions in covariance analyses.) The final analysis shows that partial predation had a significant positive effect on clutch size (P < 0.001), whereas later initiation date had a negative effect (P < 0.009). For the 106 non-predated clutches, the linear regression of clutch size y on start date x (1 June is day 1) was y = 3.96 − 0.041 x (P = 0.006). Based on this approximate relationship, clutch size decreased from about 4.3 eggs on 25 May, when the first clutch had been completed, to 3.3 eggs on 20 June, when the last clutch was completed.

As the results from partially predated clutches came from the minority of females that did not desert the nest after predation at the one- or two-egg stage, one may ask if these females are atypical in other respects, for instance having higher condition (mass) than other females. The linear regression of female body mass (in kg) on day x after she laid her last egg was m = 1.87 − 0.022 x (P < 0.001, adjusted r 2 = 0.31, n = 137). Quadratic regression did not improve the fit (adjusted r 2 = 0.30). Mean (±SE) 10-day mass m 10 (see “Materials and methods”) of 83 females with non-predated clutches was 1.66 ± 0.012 kg, almost identical to that of the 12 staying females (with partial clutch predation) that were weighed: 1.66 ± 0.037 kg (P > 0.99, t test), and to that of the 42 additional females included in this analysis (see “Materials and methods”): 1.65 ± 0.025 kg (P = 0.80, t test) .

There was a weak positive relationship between m 10 and volume V of the female’s first egg, the linear regression (in cm3) being V = 79.8 + 11.5 m 10 (P = 0.012, r 2 = 0.06, n = 137). As egg volume was positively related to female mass, which reflects condition in waterfowl (“Materials and methods”), egg volume may also reflect female condition. The mean (±SE) volume of the first egg of the 19 staying females with partial clutch predation was 97.9 ± 1.2 cm3, that of 48 females with their first egg measured before they deserted after predation was 98.9 ± 0.97 cm3, and that of the first egg of the 106 females without nest predation was 99.5 ± 0.56 cm3. There was no significant difference in egg volume between these three categories of females (ANOVA, F = 0.5, P = 0.60).

Differences in condition or other aspects between staying and deserting females might also be reflected in their clutch initiation date. We tested this possibility in a mixed ANCOVA model. Staying and deserting females had similar initiation dates (2-year mean: day 5.6 and day 4.2, respectively, i.e. a difference of 1.4 days; see also Table 1). The interaction desertion × year was not significant (P = 0.35), nor was the random factor year itself (P = 0.49). With years pooled and interactions omitted, desertion still had no significant relation to clutch initiation date (P = 0.28), providing no evidence of other differences between staying and deserting and females.

Discussion

Clutch size determination in waterfowl has been much debated, one of the contentious issues being whether they are determinate or indeterminate layers (see “Introduction”). Our results suggest that many eider females lay an extra egg after predation of the first egg(s) in the clutch, and that clutch size is not fixed at the one-egg stage or earlier. Experimental egg removals in wild ducks have reported negative results (Rohwer 1984; Swennen et al. 1993; Erikstad and Bustnes 1994), but the removals may have been too late in the laying sequence for females to respond by laying extra eggs. We found increased egg production after early predation (usually at the one-egg stage) in the same species as studied by Swennen et al. (1993) and Erikstad and Bustnes (1994). These contrasting results suggest that the timing of egg loss in the laying sequence is crucial for the response of the female (see also Kennedy 1991).

It is noteworthy that eider females have often laid more eggs than is apparent from their clutch. On average, 19 examined females from Svalbard had more than twice as many ruptured follicles (6.2) as the number of eggs in their nest (2.6), probably because of early egg predation (Ahlén and Andersson 1970), which is high in arctic-breeding common eiders (e.g., Mehlum 1991a, b; Robertson and Cooke 1993; Andersson and Waldeck 2006).

There is also other evidence that clutch size in ducks is not fixed before the start of laying (e.g. by limited availability of nutrients for egg formation; Alisauskas and Ankney 1992). In mallards Anas platyrhynchos, many females that lost their nest during laying continued without interruption to produce eggs in a new nest, ending up by laying on average 12 eggs in total, compared to about 9 eggs in females without nest predation (Arnold et al. 2002). It therefore seems that egg-laying capacity and clutch size are not fixed at the start of laying. Duck females are often able to adjust their egg production in response to events that occur during the early part of the laying sequence. Remarkably, this applies also in the common eider, often regarded as a prime example of capital breeding, with nutrients for egg production accumulated before the start of laying (e.g. Parker and Holm 1990; Meijer and Drent 1999; Stephens et al. 2009).

Although ducks are indeterminate layers in the above sense, clutch size in geese, their close phylogenetic relatives (Livezey 1986), seems to be determined by the number of developing follicles (Barry 1962). An egg removal experiment in a semi-natural population of barnacle geese Branta leucopsis, where new eggs were removed from the first egg onwards, showed that females with all their eggs removed did not produce more eggs than control females with intact clutches (Williams et al. 1996). There is evidence that egg-laying in geese depends both on nutrients accumulated before the start of laying and on food intake during the laying period (Alisauskas and Ankney 1992; Meijer and Drent 1999; Gauthier et al. 2003).

Why did many females desert the nest after early egg predation while others continued laying? We do not know, but one possibility is that females that lost early eggs to predators but did not desert had higher initial condition (body mass) than others. Increased egg production of these non-deserting females may have reduced their body mass to similar levels as in other females (see “Results”). The similarity in egg sizes and clutch initiation dates between these two categories of females, however, provides no evidence of other differences between them. But staying females might be a biassed sample, for instance if females with few developing follicles are less likely to carry on nesting with few eggs and are more prone to desert after predation of the first egg than other females. Possible evidence against this alternative is that many females carry on incubating even as few as 1–2 eggs (Ahlén and Andersson 1970). Other differences in females with few developing follicles might also be expected, such as later laying start or smaller eggs, for which there was likewise no evidence. Still another possibility is that females which persisted laying after early egg predation were older and more experienced birds, with larger clutches (Öst and Stele 2010). However, experienced females also tend to lay earlier than others (Baillie and Milne 1982), but there was little difference in laying date between abandoned and other nests, casting doubt on this explanation. Yet, we cannot entirely rule out the possibility that females staying after predation of the first egg differed in some other way from those that deserted.

Another possible reason why some females deserted whereas others stayed is that the nature of predation differs between nests. Glaucous gulls search for eider nests and often find eggs by digging into the plant material with which the female covers the nest if temporarily leaving (own observations; see also Fast et al. 2010). She may then simply find the egg(s) gone when returning. Sometimes, however, one or more gulls approach the sitting female and try to chase or grab and pull her off the nest to get access to the eggs, as we documented with video camera. During a 48-h telescope watch of PH Island from the mainland, Campbell (1975) observed 34 attacks by glaucous gulls on eiders at 22 of 68 observed nests. Both nests from which eggs were taken were deserted 24 h later. Females repeatedly experiencing gull attacks also risk injury (see Waldeck and Andersson 2006), and may be more prone to desert after egg predation than females that simply miss an egg upon returning.

In summary, we present evidence that early egg loss through predation is followed by larger egg production in many eider females. Combined with the work of Erikstad and Bustnes (1994), our result indicates that eiders are indeterminate layers. The response to egg loss at the one-egg stage, which has not previously been analysed in ducks in the wild, suggests that indeterminate laying may be more general in this group than previously thought (see also Arnold et al. 2002), and that it occurs even in its perhaps most extreme capital breeder.