Common Eider and large gull and nesting associations in coastal Labrador

Apparent nesting associations between avian egg predators and their prey have received much interest, with gulls and waterfowl receiving considerable attention. We examined the co-occurrence of breeding large gulls (Herring Gull (Larus argentatus Pontoppidan, 1763) and Great Black-backed Gull (Larus marinus L., 1758)) and Common Eiders (Somateria mollissima L., 1758) along the coast of Labrador from 1998 to 2003. Nest counts for large gulls and eiders were undertaken by ground crews on 45–109 islands each year, counting 79–283 and 721–3424 nests annually, respectively. Gulls were more likely to nest on an island with nesting eiders (69.4%) than without nesting eiders (38.4%), and the probability and numbers of gulls nesting on an island increased as eider colony size increased. Large gulls were 1.76 times more likely to occupy islands that had nesting eiders in the previous year, while eiders were equally likely to colonize islands that did or did not have nesting gulls in the previous year. Eiders were n...


Introduction
The relationship between breeding large gulls (Larus spp.) and Common Eiders (Somateria mollissima L., 1758) has received much consideration.Although it is recognized that large gulls are predators of eider eggs and ducklings, studies vary in their conclusions on how important gull predation is to breeding eiders.Some conclude that predation rates are high enough to lead to population declines (Mendehhall and Milne 1985;van Dijk 1986;Mawhinney et al. 1999) while others suggest that the impact of gull nest predation on eider populations is minimal, and that much of the predation seen is caused by disturbance at the colony or is focused on weakened ducklings (Götmark and Ahlund 1984;Swennen 1989;Bolduc and Guillemette 2003).Reduced predation rates on eider clutches have been seen within gull colonies (Gerell 1985), suggesting that eiders may nest preferentially with gulls to protect them from other predators, such as corvids or non-nesting gulls (Götmark and Åhlund 1988).
The importance of nesting eiders to breeding gulls, however, has received comparatively little attention.In New Brunswick, Common Eider eggs and ducklings made up a relatively small proportion of the diet of Great Black-backed Gulls (Larus marinus L., 1758) and Herring Gulls (Larus argentatus Pontoppidan, 1763) nesting close to eider colonies (Gilliland et al. 2004;Steenweg et al. 2011).Yet gulls and eiders are found nesting in association across much of their extensive ranges along the coastlines of boreal and Arctic waters of the northern hemisphere.
Compared to the large colonies in temperate nesting areas, in northern regions both eiders and large gulls generally nest in relatively small and widespread colonies (Gilchrist and Robertson 1999;Chaulk 2012;Robertson and Chaulk 2016).Additionally, nesting colonies are much more dynamic, with numerous islands being abandoned and new islands colonized each year by both gulls and eiders (Chaulk et al. 2006;Robertson and Chaulk 2016), possibly due the dynamics of snow melt and sea ice (Parker and Mehlum 1991;Chaulk et al. 2007).In this study D r a f t 4 we examine the relationships between colony occupancy and abundance of eiders and large gulls (Great Black-backed and Herring Gull) nesting along the central coast of Labrador, Canada.
Because of the dynamic nature of the colonies in Labrador, we also examine relationships across years to see how these species respond after nesting in association with one another or not.We sought to determine if eiders actively avoid or seek out gull colonies from year to year, and whether gulls preferentially move to islands that have had nesting Common Eiders.

Study area
Archipelagos near the communities of Nain and Hopedale were surveyed once each year from 1998-2003 and the one near Rigolet was surveyed once each year from 2000-2003.The extent of the study areas were as follows: Nain was approximately 3383 km 2 , containing 1000 islands ranging in size from 0.01 -44800 ha; Hopedale was approximately 566 km 2 , containing 650 islands ranging in size from 0.01 -3875 ha and Rigolet was approximately 2834 km 2 , containing 335 islands ranging in size from 0.02 to 5204 ha.Surveys were generally conducted in late June in Rigolet, and continued through to mid-July in Hopedale north to Nain, to coincide with the incubation period in each region (Chaulk et al. 2005a).Islands in the archipelagos are typically barren with sparse vegetation and very limited nesting cover.

Census methods
Archipelagos were surveyed for evidence of breeding eiders and gulls, and geodetic information (including island size) from these surveys was plotted on 1:50,000 digital base maps and linked to tables containing information on species occurrence and nest abundance.In all areas, islands were selected for study based on random sampling (Chaulk et al. 2004(Chaulk et al. , 2005a(Chaulk et al. , 2005b) ) and D r a f t 5 45-109 islands were surveyed annually (Chaulk et al. 2005a).We limited our sampling to islands that were estimated to be smaller than 30 ha; this removed 120 islands leaving 1855 islands available to be sampled (about 6%; Chaulk et al. 2005a); islands larger than 30 ha are likely to function as, and be perceived as, mainland habitat (Chaulk et al. 2006;Robertson and Chaulk 2016).In all, 125 islands were suitable for analysis, in that they included counts of gulls and eiders in 2 or more years.
Ground censuses were conducted using standard search method employed by the Canadian Wildlife Service (Nettleship 1976) and other researchers (Falardeau et al. 2003;Merkel 2004;Chaulk et al. 2005): two to four people systematically walked over the islands searching for signs of eider and gull nesting.Colony sizes were based on all detected nests (active, recently depredated, etc.); older remnants of nests from previous years were rare and easily recognized for both eider and gulls, and not counted.Depredated eider nests were defined as those nests that were clearly established in the current year (i.e. containing fresh down), and/or contained depredated eggs (eiders do not line their nests with down until eggs are laid, with most down being laid during incubation).This is a relatively rough index of nest depredation, and is certainly an underestimate, as some depredated nests would not be detected or even identifiable as such, but it serves as an index to compare among islands.Islands in the four archipelagos had limited cover and were for the most part barren, so that both nesting birds and unattended nests were easily detected (Robertson and Chaulk 2016).In all cases an island colony was considered occupied if it contained at least one nest.The primary gulls nesting with eiders in the region were Great Black-Backed and Herring Gull, and of these the Great Black-backed Gulls were more widespread.These two gull species were grouped to improve sample size for analysis and because they were not reliably identified to species in the field (Robertson and Chaulk 2016).
These two species share a number of ecological similarities, in that they both likely perceive D r a f t 6 eiders (eggs and ducklings) as a food source, and both species are likely to be perceived by eiders as clutch and duckling predators (Bourget 1973;Götmark and Åhlund 1988).Although this survey was timed to coincide with the mid to late incubation period of eiders, we felt this survey overlapped the timing of the two gulls species reasonably well to make our counts representative; most nests still contained unhatched gull eggs during our surveys (Robertson and Chaulk 2016).

Gull-eider associations
Nest counts were used to investigate breeding associations between eiders and large gulls.
The correlation between island size and breeding numbers of eiders and gulls was examined but was found to be negligible and so was not considered in further analysis (eider: r 98 = 0.026, P = 0. 80, gulls: r 109 = 0.031, P = 0.75).For analyses involving incidence (presence or absence), each trial refers to one island in one survey year; a successful trial indicates the presence of the species in question.We used generalized linear models (GLMs) with a binomial error structure to test associations between eider and gull incidence, with year as factor and abundance of the other species nested within year as a covariate to address the possible dependency in nest counts on the same island across years.We present P values for Hosmer-Lemeshow goodness of fit, which assesses the fit of the logistic model against actual outcomes and is a form of Pearson chi-square statistic.P > 0.05 suggests that the model fits the data well (Hosmer and Lemeshow 2000;Peng et al. 2002).Linear models with a similar structure were used for analyses considering abundance as the response.We also examined the relationship between the numbers of depredated eider nests found and gull nest abundance using linear models.As another approach to address the potential dependency in eider and gull numbers on the same islands across years we also examined similar models treating island as random effect.These models led to the same general conclusions as the models with abundance nested within year.

D r a f t 7
We used Program MARK (White and Burnham 1999) to calculate colonization and extinction rates of eider and gull colonies across years (see Barbraud et al. 2003 as an example).
Our model considered that in each survey year, each island was in one of four states: unoccupied, occupied by eiders, occupied by gulls or occupied by both species.Survival rates were forced to 1.0, as islands cannot be lost to the study.In this analysis, our goal was not to search for the bestfitting models among a candidate model set, but rather to estimate specific transition rates (i.e.colonization or extinction rates) and to test whether they were different depending on whether the other species was present in the previous year.Specifically, we assessed whether gulls or eiders became more likely to colonize an island if the other species had nested there in the previous year, and whether eiders were more likely to abandon a colony if gulls had been nesting on the island in the previous year.All probabilities are presented as ± 1 SE, and simple χ 2 statistics are used to examine whether rates were different.All statistical tests were two-tailed and the critical alpha was set at 0.05.

Results
Of the 125 islands included in the analysis, across all years the numbers of gull nests on occupied islands ranged from 1 to 95, with a mean of 3.2 and a median of 1 (see Robertson and Chaulk 2016).For eiders, numbers of nests on occupied islands ranged from annual means of 3.3 to 145 nests (see Chaulk et al. 2005a).
Gulls were present on islands 69.4% (261 of 376 island-year combinations) of the time when nesting eiders were present, but were only present 38.4% (56 of 146 island-year combinations) of the time on islands when eiders were not nesting (χ 2 1 = 42.5, P < 0.001).Binary logistic regression modelling revealed that eider abundance nested within year was a significant D r a f t 8 positive predictor of gull incidence and the positive relationship was seen in every year (χ 2 6 = 39.7,P < 0.01, Hosmer-Lemeshow P = 0.95, Fig. 1), indicating that as eider numbers rose, we were more likely to find sympatric nesting by gulls.The global general linear model (F 6,413 = 17.7,P < 0.001, R 2 = 24.0%)revealed that eider abundance nested within year was a significant positive predictor of gull abundance, indicating that as eider colony size increased, so too did the numbers of gulls; this positive relationship was seen in every year (Fig. 2).Gull abundance nested within year was a significant negative predictor of eider nest depredation (χ 2 6 = 121.3,P < 0.01, Hosmer-Lemeshow P < 0.01, Fig. 3), indicating that as the number of gulls increased the proportion of eider nests that were depredated declined.However, as indicated by the significant Hosmer-Lemeshow statistic, this model did not fit the data very well, and the relationship was not consistent across years (Fig. 3).
We found that gulls preferentially colonized islands that had been occupied by eiders the previous year (probability of colonization if eiders were present in previous year: 0.302 ± 0.032, without eiders in previous year, 0.166 ± 0.041, χ 2 1 = 6.07,P = 0.01), whereas the converse trend was not significant (probability of eider colonization if gulls present in previous year: 0.141 ± 0.039, without gulls in previous year: 0.151 ± 0.036; χ 2 1 = 0.03, P = 0.85).Eiders also did not show evidence of preferentially abandoning colonies when they had been nesting sympatrically with gulls in the previous year (probability of colony abandonment if gulls present in the previous year: 0.045 ± 0.011, without gulls in previous year: 0.059 ± 0.020, χ 2 1 = 0.37, P = 0.54).

Discussion
Our results indicate that eiders and gulls in Labrador do nest in association with one another.A number of factors could explain these associations including; eiders and gulls prefer to D r a f t 9 nest on similar islands, that eiders are gaining protection for their clutches by nesting in association with gulls, and that gulls are preferentially nesting in association with eiders, presumably as potential prey for the gulls.
Previous studies have shown positive relationships with eider nest numbers and gull nest numbers (Gerell 1985;Götmark and Åhlund 1988).It is likely that eiders and gulls are drawn to similar nesting islands with certain environmental characteristics such as being offshore to avoid mammalian predators (Robertson 1995;Wyman et al. 2014).In Labrador, eiders show a strong preference for islands that do not trap ice in spring, and so would not be connected to mainland areas and accessible to mammalian predators (Chaulk et al. 2007;Chaulk and Mahoney 2012).
Presumably gulls would also be attracted to nest on these islands for the same protection from mammals.A relationship between gull and eider abundance could simply be a function of island size (e.g.Robertson 1995), where a restricted number of nesting territories are possible on smaller islands.However, we found no relationship between island size and eider or gull abundance.Most islands in our study area are in the 2-3 ha range (Chaulk et al. 2005b) islands which can easily support a few dozen nesting pairs of both species.
Nesting sequence has a role to play in the colony selection processes, but it is not clear what that role might be.Great Black-backed Gulls initiate nesting before eiders (Bourget 1973), but eiders are known to visit breeding islands before nesting is initiated (Goudie et al. 2000), so gulls may use the presence of prospecting eiders to identify potential colonies within a given year.Interestingly, because eiders usually initiate nesting later than gulls, it should be possible for them to avoid nesting within gull colonies, but they do not appear to do so.
Although based on a relatively poor fitting model and a coarse index of predation rates, we demonstrated the proportion of eider nests that showed evidence of egg predation was higher in smaller gull colonies, suggesting there may be disadvantages for eiders to nest alongside one  (Gerell 1985;van Dijk 1986;Götmark and Åhlund 1988), but may suffer higher egg loss during egg-laying (Götmark 1989), leading to only weak or no benefits to eiders.After hatching, it is not known whether this potential benefit is reduced or negated, as gulls may also take large numbers of recently hatched ducklings (Campbell 1975;Mendenhall and Milne 1985), and presumably, this predation level would be higher with increasing numbers of gulls.However, if there was extensive post-hatch depredation, it could be expected that eiders would try to avoid nesting in association with large gulls.In Labrador, eiders preferentially nest on islands in areas that do not trap spring sea ice (Chaulk et al. 2007), and are less accessible to terrestrial predators, such as Arctic Fox.Avoiding mammalian predators is probably a stronger driver for eiders (and gulls) in selecting nesting islands.
Our results regarding interannual associations suggest a previously unconsidered mechanism which may explain why eiders and gulls are often found nesting in association.Based on our results, eiders neither avoid nor select islands with nesting gulls, but rather, it is the gulls that preferentially nest with the eiders, by moving to islands that had nesting eiders in the previous year.Because gulls (especially Great Black-backed Gulls) initiate nesting before eiders (Bourget 1973), they may have to rely on evidence of nesting from the previous year to indicate whether eiders are likely to be present on the island in an upcoming season.This knowledge could be gained during foraging trips during incubation and the hatching periods of the gulls.
Large gulls have foraging ranges of 10s of kms (Thaxter et al. 2012), which would give them D r a f t 11 opportunity to examine nearby islands for the presence of nesting eiders; either as potential prey and/or indicators of suitable nesting islands.
Eiders, on the other hand, apparently do not preferentially abandon or colonize sites with nesting gulls in the previous year, and appear not to use the presence or absence of gulls as a determinant of whether to select a particular island for nesting.Götmark (1989) found weak to no relationship between eiders and gulls.One reason that eiders may not preferentially abandon islands which they shared with gulls the previous years is the relatively low number of gull nests on islands in Labrador.An island with a one or a few pairs of nesting gulls may not factor in the decision of eiders to use that island again.Large gull populations are declining in Labrador (Robertson and Chaulk 2016), while eider number are increasing (Chaulk et al. 2005a), further reducing the potential effect of gull predation on nesting eiders.
Compared to southern areas where fish and other marine resources may be locally abundant for gulls (Gilliland et al. 2004;Steenweg et al. 2011; but not always, see Bond 2016) nesting in association with eiders may be more important for gulls nesting in dispersed northern areas where prey resources may be limited in number and diversity.Except for highly localized sites around dispersed human settlements, anthropogenic food sources are absent in many northern areas.Although fish and scavenging opportunities probably remain important in the diet of northern nesting gulls, local bird colonies may represent a more reliable food source available throughout the nesting period (Schmutz and Hobson 1998;Veitch et al. 2016).
Due to time restrictions on surveys and the similarity of their nests, one constraint of our analysis was the necessity of pooling Great Black-Backed and Herring Gulls and considering them equivalently in the analysis.Both species are predators of eider eggs and ducklings and in many respects both gull species share similar nesting habitats and diets.However, Great Blackbacked Gulls are more predatory on birds than Herring Gulls (Good 1998;Veitch et al. 2016), so D r a f t 12 eiders may perceive these two species differently.Based on surveys of the Labrador coast in the late 1970s, most nesting gulls were in fact Great Black-backed Gulls (87.3%) in our study area (Nettleship and Glenn 1992).The results from our study are probably representing the interaction of Great Black-backed Gulls and eiders, but we recognize that there may be intraspecific differences in the population processes of these two larids, and therefore we suggest that future studies attempt to resolve these potential differences.
Our results, coupled with previous research, indicate the relationship among eiders and gulls is dynamic, complex and likely varies across the ranges of these species.As the human population in Arctic communities grows, more pressure on coastal resources can be expected, and the management of gull populations may become an issue.Careful study of the relationship between gulls and eiders will be needed to inform management options as it is currently not clear whether results from other regions are readily transferable to the situation in the Arctic.
Figure Captions

Figure 2 .
Figure 2. The relationship between nesting gull abundance (log 10) and nesting Common

FigureFigure 1 .Figure 2 .Figure 3 .
Figure 3. Relationship between gull nest abundance (log 10) and depredation rate of gulls in Labrador.Predation rates on eider nests may increase in smaller gull colonies because there are not enough gulls to deter predators like Arctic Foxes (Vulpes lagopus L., 1758) or Ravens (Corvus corax L., 1758).Eiders nesting with large gull colonies may have reduced egg predation by excluding other potential avian predators