1. Habitat use can influence individual performance in a wide range of animals, either immediately or through carry-over effects in subsequent seasons. Given that many animal species also show consistent individual differences in reproductive success, it seems plausible that individuals may have consistent patterns of habitat use representing individual specializations, with concomitant fitness consequences.
2. Stable-carbon isotope ratios from a range of tissues were used to discern individual consistency in habitat use along a terrestrial–aquatic gradient in a long-distance migrant, the Bewick’s swan (Cygnus columbianus bewickii). These individual specialisations represented <15% of the isotopic breadth of the population for the majority of individuals and were seen to persist throughout autumn migration and overwintering until aquatic habitats were no longer available.
3. Individual foraging specialisations were then used to demonstrate two consecutive carry-over effects associated with macroscale habitat segregation: consequences of breeding season processes for autumn habitat use; and consequences of autumn habitat use for future reproductive success. Adults that were successful breeders in the year of capture used terrestrial habitats significantly more than adults that were not successful, revealing a substantial cost of reproduction and extended parental care. Use of aquatic habitats during autumn was, however, associated with increased body condition prior to spring migration; and increased subsequent breeding success in adults that had been unsuccessful the year before. Yet adults that were successful breeders in the year of capture remained the most likely to be successful the following year, despite their use of terrestrial habitats.
4. Our results uniquely demonstrate not only individual foraging specializations throughout the migration period, but also that processes during breeding and autumn migration, mediated by individual consistency, may play a fundamental role in the population dynamics of long-distance migrants. These findings, therefore, highlight the importance of long-term consistency to our understanding of habitat function, interindividual differences in fitness, population dynamics and the evolution of migratory strategies.
Empirical evidence from a wide range of taxa suggests that an animal’s performance is profoundly influenced by its habitat use and resulting diet choice (e.g. McLoughlin et al. 2007; Smith, Reitsma & Marra 2010). These impacts may be borne within a given season; for example, breeding habitat may affect offspring fitness (e.g. Brown & Shine 2004). However, it is increasingly recognized that consequences of ‘decisions’ made at one stage in an animal’s life may not always manifest immediately. Processes also operate among seasons, such that the conditions experienced at one stage in the annual cycle may have significant ‘carry-over effects’ on performance in subsequent seasons (Harrison et al. 2011). Thus, current patterns of habitat occupancy are affected by previous life-history events and may go on to affect future performance (Norris & Marra 2007).
At the individual level, carry-over effects that influence processes during the breeding season have dominated research to date (Harrison et al. 2011). In particular, the quality of habitats used throughout winter has been shown to entail significant ramifications for individual condition (e.g. Bearhop et al. 2004b; Inger et al. 2008), migratory timing (Marra, Hobson & Holmes 1998), reproductive success (Norris et al. 2004) and ultimately the dynamics of migratory populations (Norris & Marra 2007). However, migratory animals need to acquire appropriate resources at multiple locations throughout their annual cycle, including wintering, breeding and interstitial staging sites. Habitat use at each of these locations, therefore, has the potential to affect the performance of individuals in subsequent seasons (Harrison et al. 2011). In particular, conditions experienced during migration are likely to play a critical role in population dynamics. Indeed, a limited number of studies from single time points or locations during spring migration have demonstrated a link between habitat use and reproductive success (Ebbinge & Spaans 1995; Prop & Black 1998; Baker et al. 2004; Inger et al. 2008). Thus, habitat use in both winter and spring has been found to induce carry-over effects that drive interindividual differences in performance.
Consistent differences in individual breeding success are common among animal species, and consequently, the majority of a population’s recruits are often produced by a minority of adults (Newton 1989). Such a skewed distribution of lifetime reproductive success is particularly apparent in long-lived, iteroparous species (Clutton-Brock & Sheldon 2010). Together with the demonstration of carry-over effects from habitats used in winter and habitats used in spring, the notion of heterogeneity in lifetime reproductive success raises the possibility that individuals may have consistent patterns of habitat use, with concomitant fitness consequences, that represent individual specializations (e.g. Bolnick et al. 2003; Vander Zanden et al. 2010). If this is the case, understanding individual consistency of habitat use throughout the annual cycle will be fundamental to gaining insight into the importance of different foraging habitats, carry-over effects on individual fitness and population dynamics and the evolution of migratory strategies. Yet, individual patterns of habitat use from breeding through migration to overwintering and the impact of processes within the breeding season on these habitat decisions remain relatively unknown (Inger et al. 2010; Bogdanova et al. 2011).
Accurate identification of individual habitat specialization requires longitudinal records of habitat use (Newsome et al. 2009). Such records have, until recently, been virtually impossible to define, particularly in migratory species that travel thousands of kilometres over the course of the annual cycle. However, because the isotopic composition of a consumer’s tissues ultimately reflect the composition of its diet, and hence the habitat in which it has been foraging, analysis of the stable isotope composition of an animal’s tissues enables the quantification of assimilated diet at the individual level (Peterson & Fry 1987). Furthermore, information spanning a range of temporal scales can be obtained by carefully selecting particular tissues for analysis. For metabolically inert tissues, such as feathers, the isotopic composition of the diet during the (brief) period of tissue synthesis will be permanently archived (Hobson & Norris 2008). Consequently, if the moult chronology of a species is known, samples from several different feather tracts can provide sequential information on the diet and habitat selection of individuals. Conversely, the isotopic composition of metabolically active tissues, where synthesis is ongoing, represents a moving window of information on individual diet, the breadth of which depends on the turnover rate of the tissue in question (Hobson & Norris 2008).
Bewick’s swans (Cygnus columbianus bewickii, Yarrell) are known to forage in both terrestrial and aquatic habitats, with a reported preference for the below-ground parts of aquatic macrophytes during their c. 3500 km autumn migration and arrival on their wintering grounds (van Eerden et al. 1997; Klaassen et al. 2010a). Terrestrial and aquatic plants differ considerably in their stable-carbon isotope composition (Finlay & Kendall 2007; Milligan, Pretzlaw & Humphries 2010), and hence, the isotopic composition of swans’ tissues can be used to infer their habitat use across the aquatic-terrestrial continuum (Nolet et al. 2000; Milligan, Pretzlaw & Humphries 2010). Using the Bewick’s swan as a model long-distance migrant, we retrospectively test the hypothesis that individuals show a consistent pattern of habitat usage from their summer breeding grounds, throughout autumn migration and on their wintering grounds. Given that Bewick’s swans also exhibit extended parental care, such that juveniles remain with their parents throughout autumn migration and overwintering, we go on to assess the consequences of breeding status for individual patterns of habitat use following breeding, as well as the potential for successful breeding in the following summer.
Materials and methods
Bewick’s swans are long-distance migrants, breeding in the Russian Arctic and wintering in north-west Europe. In autumn, the majority of the population rapidly migrates from Pechora Delta (NW Russia) via a stopover in the Baltic states (particularly in Estonia), to the Netherlands and the British Isles (Rees 2006). The population returns northwards in early spring, via stopover sites in Estonia and Dvina Bay area of the White Sea (Nolet et al. 2001). In captive Bewick’s swans, moult has been recorded to extend from late August through until early December, with primaries moulted first (August), followed by dorsal body feathers (October), ventral body feathers (September–November) and rectrices (November–December) (B.A. Nolet, unpublished data). Given that aquatic plants are generally composed of a relatively higher ratio of 13C to 12C than those in terrestrial ecosystems (Finlay & Kendall 2007; Milligan, Pretzlaw & Humphries 2010), information on stable-carbon isotope composition in tissues can be used to infer habitat use across the aquatic-terrestrial continuum (Nolet et al. 2000; Milligan, Pretzlaw & Humphries 2010). By combining these isotopic measurements with knowledge of moult chronology and turnover rates for avian blood plasma (half-life time of c. 4·3 days in herbivorous waterfowl) and red blood cells (half-life time of c. 30 days) (Klaassen et al. 2010b; Hahn et al. in press), we examined patterns of individual specialization by integrating information on stable-carbon isotope composition of sequentially-synthesized feathers and blood components.
Plants representative of the Bewick’s swan diet were sampled along their migratory flyway at times of the year when the birds were present. On their Arctic breeding grounds, the above-ground parts of plants eaten by Bewick’s swans foraging in the aquatic habitats of the Pechora Bay (Potamogeton perfoliatus and Potamogeton pectinatus) and the terrestrial habitat of the adjacent tundra (Carex aquatilis, C. lachanelii and C. rariflora) were collected. On the Dutch wintering grounds, the below-ground parts of their primary aquatic food source (P. pectinatus) and their main terrestrial food sources, including above-ground parts of Lolium perenne, Phragmites australis and Secale cereale and below-ground parts of P. australis and Beta vulgaris were collected. Salix viminalis was also sampled in the Netherlands in August, as a surrogate for the willow eaten on the tundra at this time. In addition, below-ground parts of aquatic plants (P. pectinatus) and above-ground parts of terrestrial plants (P. australis and Scirpus lacustris) were sampled in the Mud’yug area of the Dvina Bay (White Sea) during the spring migration period. All plant samples were dried at 60 °C until they reached constant weight before analysis.
To confirm that differences in isotopic composition between tissues in wild birds reflect dietary changes at the time of tissue synthesis, rather than any temporal variation in isotopic discrimination between feather tracts, samples from a captive population of Bewick’s swans were examined. Carcasses of six Bewick’s swans that died in captivity between 1999 and 2005 were collected within 12 h of death and stored at −20 °C in a sealed plastic bag until dissection. These birds had been held in outdoor aviaries for >2 years prior to their death, where they were fed a diet containing 50% grain mixture and 50% chicken mash (HAVENS Voeders, Maashees, the Netherlands) ad libitum. Carcasses were dissected, with swan biometrics, age and sex recorded. Feather samples from inner primary (p1), central rectrix (r5), ventral and dorsal body feathers were also collected; each stored in individual bags. Blood was collected from the heart cavity and stored at −20 °C until analysis.
To examine the isotopic composition of tissues synthesized throughout the autumn migration and winter period in the wild, the carcases of 16 free-living Bewick’s swans found across the Netherlands in either January or February of 1995–2004 were collected. Many appeared to have died following collision with power lines or motor vehicles; the most common cause of death for birds found along the flyway (Rees 2006). All carcasses were placed at −20 °C in a sealed plastic bag until dissection. Carcasses were dissected and sampled as above, with the additional sampling of an outer primary (p9). Sex was subsequently verified using the 2550F and 2718R primers (Fridolfsson & Ellegren 1999) on DNA isolated from this blood (Gentra Systems DNA isolation kit, Minneapolis, MN, USA).
Live bird tissues
In addition to carcasses collected on the wintering grounds, a reduced number of tissues were also collected from live birds on the breeding and wintering grounds. A total of 48 live birds were sampled on the Pechora delta in August 1996 (n = 33) and 1998 (n = 15), prior to departure on autumn (southward) migration to their nonbreeding grounds. In both years, birds were captured during the flightless phase of moult using a hook from a small boat, either on the aquatic foraging grounds of the Pechora Bay (n = 13) or after herding the birds on small tundra lakes (n = 35). Birds were aged as juveniles, yearlings or adults on the basis of plumage; sexed on the basis of cloacal examination; and had a dorsal body feather collected.
Approximately 6–8 weeks after arrival on their Dutch wintering grounds, a total of 182 live birds were sampled over five successive winters (2005–2009 inclusive). These birds were captured by means of cannon netting on sugar beet fields used for foraging. All swans were weighed (to the nearest 50 g) and their skull and wing length measured. Birds were aged (as described earlier) and marked with a yellow plastic neckband bearing an individual four-digit code. A dorsal body feather was collected from each individual, as well as c. 1 mL of whole blood from the tarsal vein. Blood samples were immediately placed in a vial containing a commercially prepared clot activator (Greiner Bio-one, Kremsmünster, Austria) and, c. 6 h later, centrifuged. Red blood cells were stored in 70% ethanol, and together with serum samples, maintained at −20 °C until analysis. Blood samples were used to determine the sex of individuals as described previously.
Live bird observations
The breeding status of adults (‘with offspring’ or ‘without offspring’) and a score of their body condition, referred to as an abdominal profile index (API), were assessed during repeated focal scans throughout the winter of capture (van Gils et al. 2007). These focal scans involved careful observation of a focal collared swan for a half-hour period through a 20–60× spotting telescope. Breeding status in the year of capture was observed for 119 of the 133 adults captured (89%), of which there were 50 birds (42%) for whom breeding status was also recorded in the following year. API scores, a measure frequently used in waterfowl ecology to estimate a bird’s abdominal fat storage (Bowler 1994; Madsen & Klaassen 2006), were observed prior to departure on spring migration for 57 of the 133 adult swans (43%).
Stable isotope analysis
All dried plant samples were ground to fine powder using an analytical mill (mesh size <1 mm). The feather samples were cleaned with hexane to remove any contamination and air-dried under a fume hood. Whole blood (from carcases) and red blood cells (from live birds) were freeze-dried for 2 days, as were 5 μL aliquots of each serum sample in preweighed tin cups. For each tissue and food sample, subsamples of 200–500 μg were analysed in a Euro EA 3000 elemental analyzer (Eurovector, Milan, Italy) coupled through a Finnigan con-flo III interface to a Finnigan Delta V Advantage isotope ratio mass spectrometer (Thermo Scientific, Bremen, Germany). Stable isotope ratios are reported using the typical delta notation, in parts per mile (‰) such that δ13C= [(Rsample/Rstandard)−1] × 1000, where R is the ratio of 13C/12C, and Rstandard is the international reference Vienna PeeDee limestone (PDB). Reproducibility based on replicate measurements of a casein standard (n = 144) during the period of measurements was 0·13‰ (SD). To make the δ13C values of tissues comparable with the diet samples, the following discrimination factors that have been experimentally obtained from our captive population of Bewick’s swans were subtracted: −0·69 for red blood cells and whole blood; +1·52 for dorsal and ventral body feathers and rectricies; +0·64 for primary feathers; and −0·09 for serum (Hoye 2011; Hahn et al. unpublished data).
All δ13C values from swan tissues were log10 transformed (after adding 35) prior to analysis in statistical models. Generalized linear mixed models were used to test for the effect of habitat on δ13C composition of diet, with location along the flyway as a covariate. Differences in δ13C between tissues were tested using repeated measures anova, and differences between breeding status groups were tested using repeated measures GLMs, with sex as a covariate. We also included all biologically relevant two-way interactions. We used a backward selection procedure in which nonsignificant interactions and factors (P > 0·10) were successively removed from each statistical model to yield the most simplified model. The fit of each simplified model was acceptable for all GLMs (lack of fit test: P > 0·05). GLMs and anovas were conducted in spss version 17.0.
To assess the proportion of diet coming from aquatic sources between different breeding status groups, we used the Bayesian mixing model SIAR (Stable Isotope Analysis in R, Parnell et al. 2010) with 200 000 iterations (R 2.13.1, http://www.r-project.org) for each breeding status group for each tissue. Values were calculated on the basis of δ13C values of tissues and dietary items from aquatic and terrestrial habitats. Nitrogen stable isotope composition did not reveal any discrimination between the two habitats (Fig. S1, Supporting information) and was, therefore, excluded from the SIAR analyses. A scaled mass index of body condition was used as it has been shown to accurately predict variations in fat and protein reserves as well as other body components across a range of vertebrate taxa (Peig & Green 2009). This technique standardizes body mass to a fixed value of body size based on the scaling relationship between mass and length. The first principle component from an analysis of maximum wing chord and skull length explained 71·6% of the variation, and the arithmetic mean of this PC1 score was used as the fixed length value. Values represent mean ± SE throughout the text and figures unless otherwise stated.
The δ13C composition of food plants used by Bewick’s swans in aquatic habitats (−17·2 ‰ ± 0·64) was significantly higher than that of food plants from terrestrial habitats (−27·5‰ ± 0·42; GLM r2 = 0·85, F1,49 = 152·86, P < 0·001). There was no effect of location along the flyway (breeding, wintering or spring staging grounds; F2,49 = 1·61, P = 0·21), however, there was an interaction between habitat and location (F2,49 = 10·72, P < 0·001), with P. pectinatus on the wintering grounds having a higher δ13C (−15·0 ± 0·85) than the aquatic vegetation from the breeding (−19·3 ± 1·47) or spring staging (−18·9 ± 0·93) locations.
Birds held in captivity showed consistent δ13C composition across tissues (Repeated measures anova: F4,1 = 17·47, P = 0·18; Fig. 1). Carcasses of free-living birds, however, showed a bimodal distribution of δ13C in all tissues (Fig. 1). While there was no significant difference between tissues (Repeated measures anova: F5,11 = 2·90, P = 0·07), feathers spanning the migratory period (August–November) maintained consistent δ13C values at the individual level, whereas blood, reflecting diet and habitat choice in January and February, showed considerably lower δ13C values across all individuals regardless of the habitat used during autumn migration (Fig. 1). Indeed, individual carcasses showed a relatively narrow isotopic range across all feathers synthesized during migration (median within-individual variation = 2·84‰; Table 1), accounting for just 19% of the between-individual variation.
Table 1. Difference in δ13C (‰) between tissues from individual Bewick’s swans found dead (‘Carcasses’), or captured and released (‘Live birds’), during overwintering in the Netherlands. Tissues from carcasses include primaries 1 and 9, dorsal body feather, ventral body feather and rectrix feather. Tissues from live birds include dorsal body feathers and red blood cells. Population = all individuals from each sampling method
Minimum range (‰)
Maximum range (‰)
Maximum range (‰)
Live bird tissues
On the breeding grounds dorsal body feathers of birds caught on the tundra had lower δ13C values (−24·83 ± 0·39) than feathers of birds caught in the aquatic habitat (−21·0 ± 0·64; t46 = 5·17, P < 0·001). The δ13C composition of these feathers did not differ between birds caught on the breeding grounds and those caught on the wintering grounds (t103 = −1·40, P = 0·16; Fig. 2), with the majority of the population primarily displaying signatures between the terrestrial and aquatic extremes of dietary δ13C. Repeated measures anova revealed that δ13C composition differed between tissues for birds captured on the wintering grounds (F2,125 = 89·4, P < 0·001), with blood cells showing higher δ13C (−23·0 ± 0·24) than either dorsal body feathers (−24·2 ± 0·24) or serum (−25·8 ± 0·24; Fig. 2). At the individual level, although 66% of the birds’ red blood cells showed higher δ13C than their dorsal body feathers, there was a positive relationship between the δ13C of an individual’s tissues from one time point to the next (feathers & red blood cells: r2 = 0·11, F1,145 = 18·04, P < 0·001; red blood cells & serum: r2 = 0·33, F1,144 = 69·31, P < 0·001). Furthermore, isotopic signatures of tissues reflecting the migratory period indicate that the majority of the population shifted <2·8‰ between synthesizing their back feathers (October) and red blood cells (November–December; Table 1), which represents only a fraction (15%) of the total isotopic range recorded for the population. There was, however, a minor portion of the population (c. 20%) that exhibited a 5‰ or greater intertissue difference in δ13C; indicative of within-season shifts in habitat. Only two individuals (1%) showed an intertissue difference of 10‰ or greater.
Habitat use and fitness proxies
There were consistent differences in the composition of δ13C of dorsal body feathers, red blood cells and serum of birds caught on the wintering grounds on the basis of breeding status (repeated measures GLM with a significant effect of tissue: F2,126 = 90·10, P < 0·001 and breeding status: F2,126 = 4·17, P = 0·018). Analyses using SIAR mixed models revealed that adults without offspring (singles or pairs) foraged in aquatic habitats 35·4% (±0·03%) of the time, compared with aquatic foraging of 23·4% (±0·03%) and 22·7% (±0·03%) for adults with offspring and the offspring themselves, respectively (Fig. 3).
Body condition of adults at the time of capture showed no relationship with δ13C of feathers or blood components (e.g. red blood cells: GLM r2 = 0·13, F1,100 = 1·04, P = 0·31), but did differ both between sexes (F1,100 = 10·60, P = 0·002, males higher) and on the basis of breeding status (F1,100 = 3·97, P = 0·05, adults with offspring higher than adults without offspring). There were no differences in the body condition of juveniles between the sexes. API scores of adults seen in the 2–3 week prior to departure on spring migration did not show any relationship with δ13C of feathers (October), or δ13C of red blood cells (November–December), but did relate to the habitat used in the days prior to capture (δ13C of serum: F1,56 = 5·16, P = 0·03, GLM r2 = 0·20), with individuals making greater use of aquatic habitats (i.e. higher δ13C of serum) seen to depart in slightly better condition regardless of their breeding status (F1,56 = 2·48, P = 0·12) when the effect of year (F1,56 = 4·68, P = 0·04) and the Julian date of the observation (F1,56 = 3·89, P = 0·05) were taken into account.
The breeding success of the Bewick’s swan population is generally very low (Rees 2006). Indeed, only 26% adults had offspring in the year of capture. Despite this, adults that had offspring in the year of capture were more likely to be successful breeders (i.e. return with offspring) the following winter (46%; n = 6) than were adults without offspring in the year of capture (11%; n = 4; Fisher’s exact test P = 0·012; Fig. 4a). Habitat use during migration (δ13C of red blood cells) did not have a significant effect on breeding success the following summer (logistic regression r2 = 0·43; χ2 = 1·58, P = 0·21) when the effect of breeding status (χ2 = 7·14, P = 0·01) was taken into account. However, there was a significant interaction between migratory habitat and breeding status, such that greater use of aquatic habitats by adults without offspring in the year of capture was associated with increased breeding success the following summer (χ2 = 4·08, P = 0·04; Fig. 4b).
Each autumn billions of birds migrate between their breeding and wintering grounds and yet little is known about the conditions experienced throughout this migration, or how events on the breeding grounds affect both migration and subsequent performance (Bogdanova et al. 2011). To our knowledge, our study presents a unique insight into individual habitat use throughout autumn migration, and the relationship between breeding success and subsequent patterns of habitat use, condition and future breeding potential.
Bewick’s swans were previously thought to rely almost entirely on submerged macrophytes as a food resource at autumn staging sites before moving onto agricultural fields in mid-winter (Beekman, van Eerden & Dirksen 1991). However, δ13C composition of tissues spanning the period of autumn migration reveal that, at the population level, Bewick’s swans showed substantial variation in foraging habitat, occupying nearly all of the ‘isotopic space’ created by the potential dietary items. Indeed, δ13C composition of newly synthesized feathers from swans caught on the breeding grounds demonstrates that while predominant use of the habitat of capture is distinguishable, individual Bewick’s swans utilize a combination of these habitats along a continuum from entirely aquatic to entirely terrestrial. Towards the end of autumn migration, there was a slight increase in the use of aquatic habitats, as indicated by red blood cell δ13C of swans caught on the wintering grounds; however, this was within the context of a predominantly terrestrial foraging habitat. Moreover, the δ13C range of the population was many fold broader than the δ13C range for the vast majority of individuals, demonstrating that the ‘generalist’ Bewick’s swan population is composed of individual specialists (Type I or Type A generalists, sensu Van Valen 1965; Bearhop et al. 2004a; Vander Zanden et al. 2010), with the majority of the population maintaining habitat specializations that span thousands of kilometres along the flyway. These specializations are only seen to breakdown on the wintering grounds, with blood from free-living carcasses and serum from live birds revealing almost exclusive use of terrestrial habitats during winter. At this time, aquatic habitats may have frozen over or have been depleted to the extent that they no longer provide a viable alternative (Nolet et al. 2002). Similar individual specializations, albeit from samples that aggregate multiple seasons rather than assessing multiple tissues from within a season, have been seen in marine mammals (Newsome et al. 2009) and marine reptiles (Vander Zanden et al. 2010), reinforcing the suggestion that individual specialization may be a common phenomenon in so-called generalist populations (Bolnick et al. 2003).
In addition to the maintenance of individual specializations in habitat use during autumn migration, our data demonstrate that the composition of these specializations was related to individual breeding status in the preceding summer. Strikingly, families (adults and their offspring) showed a disproportionate reliance on terrestrial habitats for the duration of migration. These isotopic findings are in line with the direct observation of families exhibiting lower overall attendance in aquatic compared with terrestrial (mainly agricultural) habitats (van Eerden et al. 1997) and suggest that processes during the breeding season influence foraging along the entire flyway.
Social status has been shown to be a major driver of intrapopulation polymorphism in habitat use across a wide range of animal species, with dominant individuals having greater access to high-quality habitats than subordinate conspecifics (Marra, Hobson & Holmes 1998; Harrison et al. 2011). On the contrary, we find that adult Bewick’s swans with juveniles (which are dominant; Badzinski 2003; Klaassen, Nolet & Bankert 2006; Rees 2006) show considerably reduced use of aquatic habitats throughout autumn migration. Depending on the level of food depletion, an aquatic diet or a mixed diet (terrestrial during the day, aquatic during the night) has been shown to be the most profitable in terms of net energy gain, even when flights to and from a roost are considered (Nolet et al. 2002). Furthermore, aquatic vegetation, particularly below-ground tubers of P. pectinatus, contain more protein than the crop remains available in terrestrial habitats (14·8% vs. 6·6% of organic matter; van Eerden et al. 1997), collectively suggesting that the predominantly terrestrial diet of families may be suboptimal. Corroborating this suggestion, we found that individuals that used aquatic habitats to a greater extent, or later into the winter, (lower serum δ13C) were in better condition in February just prior to departure on spring migration. Moreover, while adults that had offspring in the year of capture were more likely to return with offspring the following winter, a small number of adults that were without offspring in the year of capture returned with offspring the following winter. These ‘upgraded’ adults had made use of aquatic habitats during the autumn migration and overwintering period of the year of capture, whereas adults that were without offspring in both years had not. Finally, aquatic habitats have been suggested to entail a reduced risk of predation for Bewick’s swans, as their main predators are terrestrial (e.g. foxes; Nolet et al. 2002; Rees 2006). Together these findings reinforce the notion that greater use of aquatic habitats may be beneficial, but that current reproduction may limit parents’ use of these habitats.
For adults, a successful breeding season entails providing parental care throughout the ensuing autumn migration, which may constrain patterns of habitat use. Three potential mechanisms could underlie such constraints. Firstly, parents may be constrained by the foraging inefficiencies of their juveniles (Inger et al. 2010). In fact, as a result of morphological differences, rather than experience, juvenile Bewick’s swans have lower instantaneous intake rates than adults when foraging in aquatic environments, such that terrestrial habitats may be more profitable (Gyimesi, van Lith & Nolet 2010). Secondly, fidelity may play a major role in habitat choice. Fidelity to particular sites or habitats has been shown to increase individual fitness across a range of vertebrate taxa, even when energy gain is not maximized, with familiarity cited as the major mechanism underlying such benefits (Hoover 2003; Bradshaw et al. 2004). Finally, the resource polymorphism may be driven by decreased intraspecific competition (Bolnick et al. 2003). Although families are thought to be dominant in direct interactions, timing may be against them when it comes to accessing depletable aquatic food resources during autumn migration. Similar to many migratory species, parents and their offspring generally depart the breeding grounds later, take longer to complete autumn migration and hence arrive 2–4 weeks later than adults without offspring (Beekman, van Eerden & Dirksen 1991; Rees 2006). An earlier departure may enable adults without offspring to exploit alternative staging areas (e.g. Bogdanova et al. 2011), potentially by depleting the availability of aquatic macrophytes at each site before the families arrive. Such indirect competition would also exacerbate the foraging inefficiencies of juveniles and the benefits of fidelity.
Each of the mechanisms presented earlier suggest a carry-over effect, where adults with offspring are, as a result of their breeding success, either opting for or forced to use a habitat, that is, to our current knowledge, suboptimal. Such macroscale segregation during migration provides a feasible means for the observed differences in arrival and departure condition and winter habitat use seen in Brent geese (Branta bernicla horta; Inger et al. 2010) and potentially other migrants settling in winter habitats of disparate quality (e.g. Marra 2000). Yet, as with many iteroparous species, recruitment to the Bewick’s swan population is heavily skewed, with a relatively small number of birds breeding successfully over a number of years (Rees 2006). The consistent success of adults with offspring, despite their use of terrestrial habitats, is in agreement with the notion that adverse environments often have a lesser impact on successful individuals (Clutton-Brock & Sheldon 2010). It may also be that there are as yet undocumented benefits to avoiding aquatic habitats. For instance, hosts often encounter their pathogens while feeding (Hall et al. 2007), and aquatic habitats have been suggested to sustain higher pathogen density and diversity (Piersma 1997), although there has been little empirical work in this area.
To our knowledge, this is the first demonstration of both individual specialization throughout the migration period, and links between migratory habitat use, preceding breeding performance and future breeding potential. These results suggest that processes during breeding and autumn migration, although mediated by individual consistency, may play a fundamental role in the population dynamics of long-distance migrants, especially those with extended parental care. Given the potential for indirect competition between families and adults without offspring, we also highlight the importance of understanding individual strategies of habitat use across the entire population. Finally, the consistent patterns of habitat use and skewed reproductive success of Bewick’s swans underscore the importance of conserving habitats on the basis of function rather than occupancy rates.
We are grateful for the co-operation with Andrei Taskayev and Vladimir Volodin and field assistance of Yuri Mineyev, Sergey Petrusjenko, Andrei Glotov, Sasha Kuznetsov, Helen Hangelbroek, Oscar Langevoord and Thijs de Boer for making the work in the Pechora delta possible. We would also like to thank Kees Oosterbeek, Wim Tijsen, Symen Deuzeman, Naomi Huig, Jan van Gils, Thijs de Boer and Peter de Vries for valuable assistance with swan capture and behavioural observations on the wintering grounds. We thank the large network of volunteer ring-readers who located the neck-collared swans on their wintering grounds, Harry Korthals for stable isotope analysis, Christa Mateman for molecular sexing and Bart van Lith for assistance with the captive birds. We also thank Chris Tonra and an anonymous reviewer for valuable comments on an earlier version of this manuscript. Swans were handled under ringing permits of the Bird Ringing Centre of Russia, and approvals CL04.02, CL06.06 and CL08.05 from the Animal Experimentation Committee (DEC) of the Royal Netherlands Academy of Arts and Sciences (KNAW). This study was supported by the Research Council for Earth and Life Sciences (ALW) with financial aid from the Netherlands Organization for Scientific Research (NWO; grant 851.40.073 and 047.002.008) and the German Research Council (DFG grant HA4437/1-1). This is publication 5154 of the Netherlands Institute of Ecology (NIOO-KNAW).