Abstract

Evolutionary potential is dependent on the additive genetic variance displayed by important adaptive behavioral and life-history traits, such as phenology. However, the genetic variance of such traits may vary over the life span. Using a long-term study (1994–2008) of arrival date from spring migration in the long-lived common tern (Sterna hirundo), we examined changes in variances of this key phenological trait across reproductive stages (before, during, and after the first breeding event). Based on 5315 records from 1232 individuals, arrival date at the breeding grounds exhibited significant declines in phenotypic and additive genetic variances across reproductive stages. Canalization of this phenological trait and selective disappearances across reproductive stages are hypotheses to explain these declines. Canalization was revealed by significant reductions in phenotypic and additive genetic variations at progressive reproductive stages. Viability analyses rejected the selective disappearance hypothesis. Heritability of arrival date also declined with reproductive stage (from 0.23 to 0.11). Finally, arrival date was under fecundity selection for experienced breeders, suggesting a current influence on canalization via fecundity selection. These results highlight how fine-tuned quantitative genetic investigations can reveal canalization in behavioral traits, reflecting past and present selective forces, and refining predictive evolutionary potential.

Introduction

Because phenological traits often have high impacts on fitness (e.g., reviewed by Pulido 2007a; Lane et al. 2012), it is particularly important to explore processes that constrain their evolution or promote responses to environmental changes. Timing of migration and arrival dates at the breeding grounds represent a determinant of breeding success and a key trait for migrating birds, as they need to match reproduction and favorable conditions in seasonal breeding environments (e.g., Bety et al. 2004; Both et al. 2005). However, evolutionary potential of such behaviorally and ecologically important traits depends on the amount of additive genetic variance. In addition, additive genetic variance of quantitative traits can vary across the life span. For instance, additive genetic variance of timing of breeding in mute swans Cygnus olor changed with age (Charmantier et al. 2006) and heritability of size increased over the first 5 years of life in bighorn sheep Ovis canadensis (Wilson et al. 2005).

However, the large majority of studies on behavioral and life-history traits have neglected changes in additive genetic variance across life stages, although there is strong evidence from morphological traits of ontogenetic changes in both heritability and variance (Wilson and Reale 2006). In the case of spring migration timing, studies show that young inexperienced birds typically arrive from migration later than experienced breeders (e.g., Potti 1998; Stewart et al. 2002; Dittmann and Becker 2003; Becker et al. 2008a). Gains in experience with age produce changes in characteristics, such as duration of migration stopover (e.g., in the reed warbler Acrocephalus scirpaceus, Rguibi-Idrissi et al. 2003) and reproductive performance (reviewed by Forslund and Part 1995; Rebke et al. 2010). Thus, a process of canalization, that is an adaptive narrowing of variation with age (possibly due to stabilizing selection in the past, see below), might be likely for the timing of spring migration. Studies of avian phenology and migration have generally not focused on canalization (but see Pulido and Widmer 2005).

Canalization was historically used to explain adaptive morphogenesis during development, and was defined by Waddington (1942; definitions reviewed by Debat and David 2001) as “developmental reactions, adjusted so as to bring about one definite end-result regardless of minor variations in conditions during the course of reaction.” The existence of a canalized trait can be observed as a reduction in trait variance during ontogeny, “buffering the adult phenotype against any types of effects (genetic or environmental) experienced earlier in life” (Wilson and Reale 2006). For traits such as body size and weight, canalization during ontogeny is well known and called “compensatory” or “targeted” growth (Cheverud et al. 1983; Atchley 1984, 1987; Wilson and Reale 2006; Dmitriew et al. 2010). Compensatory growth is a convergence of variable growth trajectories into a reduced set of adult phenotypes, leading to a decrease of phenotypic and additive genetic variances over ontogeny (Wilson and Reale 2006). Processes that produce canalization, such as strong stabilizing selection (Wagner et al. 1997; Debat and David 2001; Meiklejohn and Hartl 2002), are likely to have occurred for traits that have a strong influence on fitness (Stearns and Kawecki 1994). Thus, the process of canalization might explain changes in variance of traits across life stages.

An alternative hypothesis to canalization for explaining reductions in phenotypic and genetic variances over different life-history stages is termed “selective disappearance” (van de Pol and Verhulst 2006). Selective disappearance or selective mortality occurs when stabilizing or directional viability selection acting on the focal trait depletes both phenotypic and genetic variances over life stages (Nussey et al. 2006, 2008). In the case of migration timing, a strong directional or stabilizing mortality on this phenological trait would produce lower survival of young birds arriving too early or too late at the colony, resulting in decreasing phenotypic and additive genetic variances of arrival date with older age. In order to clearly dissociate whether changes in variances are due to canalization (perhaps reflecting past stabilizing selection) or to selective disappearance (reflecting current viability selection), we need to test whether there is mortality selection acting on arrival date from migration, whether arrival date is heritable (viz., to see if a response to current selection is possible), and whether a pattern of canalization is revealed when the influence of current selective mortality is removed. In this way, influences of both current selection for reducing variation in migration arrival and historical canalization of the trait should be revealed because both processes might contribute to the narrowing of variation in traits with advancing life stages.

In this study, we used a long-term data set on common terns (Sterna hirundo) in Germany to address the question of evolutionary potential across life stages and plasticity of migration timing. The objectives of our study were 4-fold. First, we examined if the strong ontogenetic advance in timing of spring arrival with reproductive stages during the life of common terns (Ezard et al. 2007; Becker et al. 2008a) is linked to changes in phenotypic and additive genetic variations in arrival date. A reduction in phenotypic variation in arrival date with increasing reproductive stage is suggested by previous studies on common terns (Ezard et al. 2007), but changes in genetic variation have not been examined. Second, we tested the alternative hypotheses to explain a narrowing variance in arrival date over the life cycle, namely canalization and selective disappearance. Under the former, we predicted a persistent narrowing variation when controlling for selective disappearance; under the latter, we predicted stabilizing or directional viability selection (Nussey et al. 2006, 2008). The viability selection analysis was coupled with a fecundity selection analysis in order to assess the relationship between arrival date and fitness in this species. Third, when additive genetic and phenotypic variations in a trait change during the life cycle, it is unclear how estimates of heritability will change because they are a function of the 2 measures of variance. If additive genetic variance decreases more rapidly than phenotypic variance, heritability will decrease. Fourth, although canalization results in a decrease in phenotypic variation, phenotypic plasticity should increase phenotypic variation (Debat and David 2001). Thus, we also evaluated evidence of phenotypic plasticity in arrival date because adaptive phenotypic plasticity promotes responses to environmental changes.

Common terns are long-lived seabirds, with a maximum reported age of 33 years, and exhibit strong birth site, breeding site, and mate fidelity (Nisbet et al. 2002; Becker and Ludwigs 2004). The colony in Wilhelmshaven, Germany, has been monitored for more than 20 years, and passive transponders have provided high-quality longitudinal measures of the birds’ exact arrival dates from their spring migration, as well as pedigree and demographic information. Due to the existence of long-term and individual-based data with an extensive social pedigree, common terns provide an excellent biological model to assess changes across reproductive stages of phenotypic and additive genetic variances and evolutionary potential of arrival timing from spring migration.

Methods

Study site, data, and pedigree

From 1984 to 2010, a colony of about 90–530 pairs of common terns was monitored at Wilhelmshaven, on the German North Sea coast. Common terns migrate south in September to winter along West African coasts, and return to their breeding grounds in April (Becker and Ludwigs 2004). In Wilhelmshaven, 6 rectangular artificial islands of equal size (about 10.7 × 4.6 m), in a line and separated by 0.9 m, are used by the colony for breeding. These islands are protected against flooding and predators (mainly rats) by low concrete walls. Since 1992, every fledgling has been marked with a passive subcutaneous transponder (Becker and Wendeln 1997). Additionally, from 1992 to 1995, 101 adults were captured and marked. Resting platforms on the walls of the islands are fitted with antennae that record the presence of marked subadults, nonbreeding adults, and breeders. This system allows an exact record of the first day of presence in the colony of most individuals during each breeding season (termed “arrival date”). Arrival date is recorded as a Julian date, that is, the number of days since the first of January of the focal year.

Each year, monitoring started when the first terns were sighted (on about 10 April). For the reliability of the automatic detection system, see Becker et al. (2008b); the probability of resighting is close to 1 (Szostek and Becker 2012). The first egg is laid early in May. We included data collected between 1994 and 2008. Before 1994, the number of transponder antennae was too low to provide precise estimates of arrival date.

As reproductive stage was previously reported to strongly influence arrival date (Ludwigs and Becker 2002; Becker et al. 2008a), we defined 3 developmental stages with respect to reproduction: prospectors, first-time breeders, and experienced breeders. Prospectors are young individuals (at least 2 years old) that have never reproduced and visit the colony without breeding. We examined only 2-year-old prospectors (about 55% of prospecting birds) because this is the age at which the majority of the prospectors (93%) arrive for the first time at the natal colony and older prospectors mainly included individuals that failed to reproduce. The average age of first-time breeders was 3.5 years (±0.8 standard error [SE]). All birds with at least 1 clutch in a previous year were considered experienced breeders. The maximum age in our data set was 22 years and mean age was 7.3 years (±2.9 SE).

Our pedigree was based on surveys of the total colony every 2–3 days, when newly initiated clutches and newly laid eggs were located and marked. To identify parents of each nest, a portable antenna was placed around the nest for 1–2 days during the incubation period. Chicks were ringed on the day of hatching and checked every 2–3 days until death or fledging at about 26 days old (Becker and Wink 2003). At about 14 days of age, all chicks were marked with transponders (see above). The data set included 5315 records of arrival dates for 1232 individuals. The social pedigree consisted of 4023 individuals, of which 1336 form the original previously unmarked population, with 825 fathers and 800 mothers. Common terns exhibit a very low level of extra-pair copulations and fertilizations (respectively, 1.3% and 2.9%, see González-Solís et al. 2001), hence the social pedigree was a good approximation of the genetic pedigree. The maximum depth of the pedigrees was 4 generations. Numbers of links of paternities and maternities were both equal to 2377 because both parents of birds born in the colony were systematically identified.

Estimating repeatability of arrival date

As advised in a recent review on repeatability estimation (Nakagawa and Schielzeth 2010), we calculated repeatability estimates (i.e., the intraindividual correlation coefficient) for all data combined and for repeated observations on experienced breeders, using linear mixed models. Compared with a classic one-way analysis of variance (Lessells and Boag 1987; Falconer and Mackay 1996), a linear mixed model allows the calculation of an “adjusted repeatability” that controls for confounding fixed effects (Nakagawa and Schielzeth 2010). For the overall data set, adjusted repeatability was calculated by including reproductive stage as a fixed effect and individual identity as a random effect. For the estimation across experienced breeders only, sex, age, age² (Ludwigs and Becker 2002; Becker et al. 2008a), and breeding success in the previous year (a binary factor, “success” if the individual had at least 1 fledgling the previous breeding season, otherwise “failure”) were included as fixed effects and year of breeding, year of birth, and individual identity as random effects.

Selection analyses

We used 3 fitness components for our selection analyses: annual sum of fledglings SFl, survival of the focal individual to the next year Sv, and annual fitness Ft. The annual sum of fledglings corresponds to the sum of fledglings from the first clutch and any second clutch and/or replacement clutch during the focal year. Annual sum of fledglings was augmented by 1 to avoid 0 values. Following Qvarnström et al. (2006), annual fitness was calculated as:

 
formula

We used survival of the focal individual to the next year as an individual fitness component for the viability selection analyses. For each reproductive stage and sex, we obtained relative annual sum of fledglings, relative survival, and relative annual fitness, respectively, by dividing each individual’s measure of annual sum of fledgling, survival and annual fitness by the mean sum of fledglings per breeding attempt, the proportion of surviving individuals, and the mean annual fitness in the given year (see, e.g., McAdam and Boutin 2003; Garant et al. 2007). Arrival date was standardized (zero mean, unit variance) within each year. We estimated directional selection differentials using linear models for each sex and for each reproductive stage separately, and estimated quadratic selection differentials from models that included both a linear and a quadratic term to explore stabilizing/disruptive selection (Lande and Arnold 1983).

Statistical significance of the selection differentials was estimated using the raw data with generalized linear models 1) with Poisson error structure for fecundity and annual fitness and 2) with binomial error structure for viability selection for 2-year-old prospectors and first-time breeders. For experienced breeders, we used generalized linear mixed models with the identity of the individual as random effect. In order to test for differences between fecundity selection differentials in first-time breeders and experienced breeders for each sex, we tested the interaction between arrival date and reproductive stage, using data annually standardized with the 2 stages and generalized linear mixed models with Poisson error structure and the identity of the individual as random effect.

Quantitative genetic analyses

In order to partition the phenotypic variance in arrival date graphic and estimate variance components and heritability, we ran restricted maximum likelihood mixed models (known as “animal models,” Kruuk 2004) on the phenotypic and pedigree data. Using nonstandardized arrival date, we ran an animal model based on the complete data set with sex, age, age², and the reproductive stage as fixed effects. Then, we performed a separate animal model for each reproductive stage. Fixed effects in these single-stage models included sex in all models, age and age² for first-time breeders and experienced breeders (prospectors were all 2 years old), and breeding success in the previous year (SUCCESS) for experienced breeders.

In these animal models, the total phenotypic variance (graphic) of arrival date was partitioned into variances attributed to differences in years of breeding yi (graphic), years of birth (graphic), individual permanent environment pei (graphic), early development environment (brood of birth; graphic), maternal and paternal effects (graphic and graphic), additive genetic effects ai (graphic), and residual variance (graphic). The permanent environment variance is based on repeated measures of experienced breeders across multiple years and includes common effects of the individual on multiple arrival dates beyond the genetic effects. An individual is perfectly related to itself, but can also occupy a similar environment from 1 year to another, or the trait studied can be influenced repeatedly by early life conditions. Shared genetic and environment components can bias the estimation of heritability if the permanent environment variance is not considered in the analysis (Kruuk and Hadfield 2007). To test if these variances were significantly different from zero, we used a reduced model where the focal random effect was dropped. We compared the deviance from the reduced and complete models via a likelihood ratio test (LRT) with a single degree of freedom (Wilson et al. 2010). The variances attributed to differences between years of birth (graphic), early development environment (graphic), and maternal and paternal effects (graphic and graphic) were significant for none of the models and were therefore dropped from the results presented. Thus, the arrival date (AD) of any experienced breeder i was given as:

 
(1)
formula

where ei is a residual error term (having mean zero and variance graphic). In the mixed models using repeated data and accounting for individual differences, a significant year effect can be interpreted as indicative of phenotypic plasticity (Nakagawa and Schielzeth 2010).

After these univariate analyses, we performed a multivariate quantitative genetic analysis in which the arrival dates at different reproductive stages were treated as 3 different traits. Sex was included as a fixed effect and year as a random effect. A permanent environment variance was fitted only for the arrival date of experienced breeders, for which we had repeated individual records. This analysis provided additive genetic variances for the trait in each reproductive stage as in the univariate analyses but also phenotypic and additive genetic correlations for pairwise comparisons of arrival date for each combination of reproductive stages. In order to test whether additive genetic variances differed between reproductive stages, we compared this multivariate model with additive genetic variances unconstrained to models where graphic was constrained to be equal between 2 life stages using LRTs. In addition, Fligner–Killeen tests were applied to test whether phenotypic variances differed between reproductive stages. Finally, we performed a similar multivariate animal model but based on data containing only individuals that were in all of the 3 reproductive stages at least once. Comparison of models with genetic variance constrained versus unconstrained using LRT and Fligner–Killeen tests as above allowed tests of whether changes in variances persisted within the group of longest lived individuals that were observed throughout their lifetime.

All animal models were implemented in the program ASReml 3.0 (Gilmour et al. 2009). This software also provides SEs associated with variance components or functions of the variances such as heritability and genetic correlations.

Results

On average, experienced breeders arrived from spring migration on 28 April, first-time breeders 17 days later, and 2-year-old prospectors more than a month after that and about 2 months after experienced breeders (Table 1).

Table 1

Variance components and heritability estimates from a multivariate analysis of arrival date (number of days since the first of January) across reproductive stages: 2-year-old prospectors (2yPr), first-time breeders (FtBr), and experienced breeders (ExBr)

Data set  N  n  Mean  graphic  graphic  graphic  graphic  graphic  h² 
2yPr  918  918  178.5  284.00 (18.87)  188.25 (21.89)  30.38 (14.68)  65.39 (22.24)  0.23 (0.08) 
FtBr  604  604  135.2  158.20 (16.94)  96.63 (12.85)  34.04 (15.26)  27.51 (12.62)  0.17 (0.08) 
ExBr  2767  648  117.9  102.50 (4.30)  57.66 (1.77)  6.04 (2.52)  28.99 (5.38)  9.83 (5.09)  0.10 (0.05) 
Data set  N  n  Mean  graphic  graphic  graphic  graphic  graphic  h² 
2yPr  918  918  178.5  284.00 (18.87)  188.25 (21.89)  30.38 (14.68)  65.39 (22.24)  0.23 (0.08) 
FtBr  604  604  135.2  158.20 (16.94)  96.63 (12.85)  34.04 (15.26)  27.51 (12.62)  0.17 (0.08) 
ExBr  2767  648  117.9  102.50 (4.30)  57.66 (1.77)  6.04 (2.52)  28.99 (5.38)  9.83 (5.09)  0.10 (0.05) 

Sex was included as a fixed effect. Parentheses give the SEs of variances and heritabilities. N, total number of records; n, number of individuals; graphic, phenotypic variance; graphic, residual variance; graphic, annual variance (explained by differences between years); graphic, permanent environment variance; graphic, additive genetic variance; h², narrow-sense heritability. X, could not be estimated.

Arrival date on breeding grounds for common terns was repeatable, both using the overall repeatability adjusted for reproductive stage (among all individuals; 0.20±0.02 SE; one-tailed t-test; P < 0.0001; n = 1232; mean number of observations per individual = 4.31) or estimated among experienced breeders (adjusted for age and sex; 0.35±0.02 SE; one-tailed t-test; P < 0.0001; n = 648; mean number of observations per individual = 4.35).

When combining all phenotypic and pedigree data in an animal model, additive genetic variance was significant, yet the heritability estimate remained low (0.06±0.03) because of large residual and environmental variance components (Appendix S1). When splitting the data into 3 reproductive stages, the phenotypic variance was greatest for 2-year-old prospectors and least for experienced breeders (all comparisons with P ≤ 0.001). Variances in arrival date due to annual fluctuations and permanent environment effects were significant, respectively, for all reproductive stages and for experienced breeders. Additive genetic variance was significant for 2-year-old prospectors graphic = 11.22; P < 0.001), first-time breeders graphic = 6.66; P = 0.01), and experienced breeders graphic = 4.02; P = 0.045). Variance components estimated by the multivariate and the univariate analyses were similar (Table 1; Appendix S1). The univariate models, multivariate model, and multivariate model restricted to individuals that did not disappear before the experienced breeder stage confirmed the decreases of phenotypic and additive genetic variances across reproductive stages (Appendix S2). Indeed when comparing multivariate models where additive genetic variance was constrained to be equal between life stages, or unconstrained, we found significant differences in graphic between prospectors and experienced breeders graphic = 7.84; P < 0.01), but not between prospectors and first-time breeders graphic = 2.70; P = 0.10) nor between first-time breeders and experienced breeders graphic = 2.12; P = 0.15). In the multivariate model, including all the individuals, Fligner–Killeen tests confirmed a significant decrease in phenotypic variance with reproductive stage (all comparisons, P ≤ 0.0001). Heritability consistently decreased over reproductive stages in the univariate models and the multivariate model including all the individuals (Table 1; Appendix S1). In the restricted multivariate model, Fligner–Killeen tests showed significant differences in phenotypic variance between reproductive stages (all comparisons, P ≤ 0.0001). LRTs showed significant differences in graphic between first-time breeders and experienced breeders graphic = 6.64; P < 0.01) but not between prospectors and first-time breeders graphic = 0.3; P = 0.58) and differences that only approached significance between prospectors and experienced breeders graphic = 3.24; P < 0.07).

A multivariate animal model using arrival date of all 3 reproductive stages as dependent variables showed strongly significant phenotypic correlations in arrival date between all 3 stages, ranging from 0.14 to 0.81, and a significant additive genetic correlation in arrival date of 0.59 between the 2-year-old prospectors and the first-time breeders (Table 2). In the same model, variance in arrival date due to annual fluctuations accounted for 5.9–21.5% of the phenotypic variances, and permanent environment accounted for 28.2% of the phenotypic variance in arrival date of experienced breeders (Figure 1).

Table 2

Phenotypic and additive genetic covariance/variance/correlation matrices from the multivariate analysis for the 3 reproductive stages: 2-year-old prospectors (2yPr; sample size = 918), first-time breeders (FtBr; sample size = 604), and experienced breeders (ExBr; sample size = 2767 records on 648 individuals)

 2yPr FtBr ExBr 
Phenotypic covariances/variances/correlations 
2yPr 284.00±18.87 0.32±0.06*** 0.14±0.04** 
FtBr 66.97±13.90 158.20±16.94 0.81±0.13*** 
ExBr 24.35±7.02 42.98±6.37 102.50±4.30 
Additive genetic covariances/variances/correlations 
2yPr 65.39±22.24 0.59±0.240.20±0.28 
FtBr 25.06±2.05 27.51±12.62 0.48±0.29 
ExBr 5.05±0.68 7.81±1.31 9.83±5.09 
 2yPr FtBr ExBr 
Phenotypic covariances/variances/correlations 
2yPr 284.00±18.87 0.32±0.06*** 0.14±0.04** 
FtBr 66.97±13.90 158.20±16.94 0.81±0.13*** 
ExBr 24.35±7.02 42.98±6.37 102.50±4.30 
Additive genetic covariances/variances/correlations 
2yPr 65.39±22.24 0.59±0.240.20±0.28 
FtBr 25.06±2.05 27.51±12.62 0.48±0.29 
ExBr 5.05±0.68 7.81±1.31 9.83±5.09 

Elements below the diagonal are pairwise covariances; on the diagonal are variances; above the diagonal are pairwise correlations (in italics).

Significance of the correlations: *P < 0.05; **P < 0.001; ***P < 0.0001.

Figure 1

Proportions of the phenotypic variance associated with the different variance components for arrival date in common terns. Variance components were estimated by a trivariate analysis on arrival date of 3 reproductive stages: 2-year-old prospectors (2yPr), first-time breeders (FtBr), and experienced breeders (ExBr).

Figure 1

Proportions of the phenotypic variance associated with the different variance components for arrival date in common terns. Variance components were estimated by a trivariate analysis on arrival date of 3 reproductive stages: 2-year-old prospectors (2yPr), first-time breeders (FtBr), and experienced breeders (ExBr).

A lack of directional or stabilizing selection on survival of any reproductive stage was revealed by a complete absence of viability selection (viz., selective disappearance) on the distribution of arrival dates (Table 3). For both male and female experienced breeders, however, timing of arrival was significantly related to individual fitness due to significant negative directional fecundity selection for experienced breeders. Selection differentials with annual fitness did not significantly differ between first-time breeders and experienced breeders for both sexes (males, z = −0.67, P = 0.50; females, z = −0.94, P = 0.35). Selection differentials with annual sum of fledglings did not significantly differ between first-time breeders and experienced breeders for males (z = −1.04, P = 0.30) but tended to differ for females (z = −1.81, P = 0.07).

Table 3

Directional selection differentials (s) and quadratic terms (c²) (±SEs) on arrival date in common terns

Stage  Sex    Sv  SFl  Ft 
2yPr  Male  n  497 
sAD  −0.01±0.01     
graphic  −0.01±0.01     
Female  n  422 
sAD  0.00±0.01     
graphic  −0.01±0.01     
FtBr  Male  n  331  354  331 
sAD  −0.01±0.01  −0.05±0.02  −0.04±0.02 
graphic  0.01±0.01  0.01±0.01  0.02±0.02 
Female  n  271  300  271 
sAD  0.00±0.01  0.01±0.02  −0.01±0.02 
graphic  0.00±0.00  0.00±0.01  −0.01±0.02 
ExBr  Male  n  1216  1453  1216 
sAD  0.00±0.01  −0.10±0.01***  −0.07±0.01* 
graphic  0.00±0.01  0.02±0.01  0.02±0.01 
Female  n  1309  1531  1309 
sAD  −0.01±0.01  −0.08±0.01***  −0.06±0.01** 
graphic  0.00±0.01  0.01±0.00  0.01±0.01 
Stage  Sex    Sv  SFl  Ft 
2yPr  Male  n  497 
sAD  −0.01±0.01     
graphic  −0.01±0.01     
Female  n  422 
sAD  0.00±0.01     
graphic  −0.01±0.01     
FtBr  Male  n  331  354  331 
sAD  −0.01±0.01  −0.05±0.02  −0.04±0.02 
graphic  0.01±0.01  0.01±0.01  0.02±0.02 
Female  n  271  300  271 
sAD  0.00±0.01  0.01±0.02  −0.01±0.02 
graphic  0.00±0.00  0.00±0.01  −0.01±0.02 
ExBr  Male  n  1216  1453  1216 
sAD  0.00±0.01  −0.10±0.01***  −0.07±0.01* 
graphic  0.00±0.01  0.02±0.01  0.02±0.01 
Female  n  1309  1531  1309 
sAD  −0.01±0.01  −0.08±0.01***  −0.06±0.01** 
graphic  0.00±0.01  0.01±0.00  0.01±0.01 

The results are presented for the 3 reproductive stages: 2-year-old prospectors (2yPr), first-time breeders (FtBr), and experienced breeders (ExBr); and for all fitness components, survival (Sv), annual sum of fledglings (SFl), and annual fitness (Ft). n is the sample size for each sex and reproductive stage.

Significance: *P < 0.05; **P < 0.001; ***P < 0.0001.

Discussion

The long-term monitoring of arrival dates in common terns allowed us to document a decrease in phenotypic and additive genetic variations during the birds’ life cycle (Figure 1). The narrowing of variation in arrival date appeared to be due to canalization instead of selective disappearance, as evidenced by the lack of current directional or quadratic natural selection via viability as individuals proceed through their breeding careers. Further analysis also supported canalization because decreases of variances were still observed in the analysis even when the influence of selective mortality was removed by considering only individuals that survived all 3 reproductive stages. These are indirect tests of canalization because it is difficult or impossible to directly test for the presence of selection and adaptation in the past (Dobson 1985), hence such tests are the best presently available.

Interestingly, significant fecundity selection was evident (favoring earlier arrival dates) that could contribute to ongoing narrowing of phenotypic and genetic variations over reproductive stages. In addition to an influence of viability selection, the influence of fecundity selection is also possible. The response to this selection, however, depends in part on heritability of arrival date, which was moderate to low (Table 1, Appendices S1 and S2). Tests for canalization in behavioral or life-history traits are needed. For example, a similar decrease in phenotypic variance was previously described for tern reproductive performance (Rebke et al. 2010), and also might be due primarily to ontogenetic canalization. If the narrowing of variation in arrival date during the life cycle is a currently evolving phenomenon, a response to selection is only possible if significant additive genetic variation occurs.

Migratory experience and learning during early reproductive stages could lead to narrowing of variation in spring arrival date in experienced breeders, and thus provide a behavioral mechanism for canalization. Prospectors and first-time breeders may be extremely plastic in arrival date (Becker et al. 2008a), as suggested by greater year effects on phenotypic variances compared with experienced breeders. After the initial periods of increasing experience, birds might use acquired knowledge about migration routes and resources, and thereafter mainly repeat themselves. This idea is supported by the highly significant permanent environment effect of experienced adults. Such a pattern would also facilitate the synchronization between partners and thus promote pair fidelity (see, coefficient of assortative mating, Appendix S3, and González-Solís et al. 1999; Ludwig and Becker 2008). Parents might promote the learning process during early age of the offspring, through the long period of parental care (Schauroth and Becker 2008; Watson et al. 2012).

Early arrival from migration is usually considered advantageous for breeders (e.g., Kokko 1999; Bety et al. 2004; Neto and Gosler 2005; Teplitsky et al. 2011). In our study population, although previous analyses showed weak fecundity selection overall on arrival date (Ezard et al. 2007), our analyses showed that early breeders have higher numbers of fledglings, but mainly if they are experienced birds (Table 3). Conversely, young and inexperienced birds, which produce few or no offspring, were not under significant fecundity selection (though experienced breeding males and females were), yet had greater genetic variance and heritability in their migration arrival dates. Overall, the display of additive genetic variance restricted to the life stages where selection forces are weak suggests limited evolutionary potential for this character and little possibility of rapid evolutionary change.

However, positive genetic correlations between traits that are not subject to opposing selection pressures can promote and accelerate evolutionary change (Blows and Hoffmann 2005). There was a highly significant phenotypic correlation (0.81±0.13) between first-time breeders and experienced breeders but the additive genetic correlation was not significant (0.48±0.29). On the other hand, the phenotypic correlation between 2-year-old prospectors and first-time breeders was quite low (0.32±0.06) and the genetic correlation was significant and fairly strong (0.59±0.24) (Table 2). The moderate genetic correlations among arrival dates for different life cycle stages, far from unity, might indicate that arrival date is a somewhat diverse trait at different stages in the life cycle. This was also suggested by the overall heritability being less than separate stage-specific heritabilities (Roff and Fairbairn 2011). Thus, even though genetic correlations were positive, there is a limit to how rapid evolutionary changes of spring migratory timing may be in a context of environmental changes. Ultimately, an investigation of correlated characters such as body condition is probably necessary to complement our understanding of evolutionary potential in timing of migration (e.g., Blows 2007).

We estimated the heritability of spring migration arrival dates using an “animal model” approach, which efficiently uses the full power of a large pedigree (Kruuk 2004; Wilson et al. 2010). As expected, the heritability estimates from animal models (0.06–0.23, see Appendix S1) were lower than repeatability estimates (Dohm 2002), yet heritability estimates from parent–offspring regressions (0.27–0.40, see Appendix S3) were higher. Because these later estimates were in the same range as the mean estimate in previously published studies in long-distance migrants (0.39±0.27) (reviewed by Pulido 2007b), our results support the suspicion that previous heritability measures of timing of arrival date may have been overestimated. Teplitsky et al. (2011) used the animal model to estimate heritability of arrival date for migratory barn swallows Hirundo rustica at 0.11 (Spain) and 0.32 (Denmark). Although our animal model estimates of heritabilites were generally significant (Table 1, Appendices S1 and S2), they also declined over the life cycle. This occurred because as the phenotypic variance decreased, additive genetic variance also declined and a bit more rapidly.

Individuals also displayed moderate but significant variation in their migration timing between reproductive events, suggesting moderate plasticity. Permanent environment effects on variation in arrival dates were substantial and significant in the global and experienced breeders’ models. Such “permanent” effects are independent of additive genetic variance, and may reflect factors that produce consistency of individuals in repeated measurements. In our case, such effects might reflect the need for establishing the territory in due time for breeding and for the sexes of faithful pairs to arrive in synchrony (González-Solís et al. 1999), as well as previous experience with resources along the migration route. Arrival date from spring migration had a low but significant repeatability (0.20–0.35), which confirms previous results in common terns (Ezard et al. 2007; Becker et al. 2008a) and other species (e.g., Møller 2001; Pulido et al. 2001; Bety et al. 2004). On the other hand, variation that was not explained by repeatability (viz., 1 – R, at 80% overall and 65% for experienced breeders) was substantial, giving a rough idea of the possible amount of plasticity and its individual variation (Nakagawa and Schielzeth 2010).

When decomposing the phenotypic variance using animal models, the annual variance was significant, yet it declined with reproductive stage, from about 22% of the phenotypic variation in first-time breeders to less than 6% in experienced breeders. Nonetheless, individual adjustments occurred in spite of a lack of directional changes in mean arrival date between 1994 and 2006 in this population (Ezard et al. 2007). In fact, phenotypic plasticity might be a stronger influence than heritable genetic change on the way that arrival date responds to environmental changes. An important follow-up to our study would be to identify the environmental cues with annual fluctuations on which birds base their timing decisions: photoperiod, temperature, rain, winds, and food availability are likely candidates (review, Gordo 2007; see, e.g., Arizaga et al. 2011; Conklin and Battley 2011; Robson and Barriocanal 2011); and whether these cues are perceived on wintering grounds, during migration, in stopovers or at the breeding grounds.

Arrival from spring migration was related to reproductive success in our study population and appeared to be a canalized trait. This is in line with theoretical expectations that fitness-related traits should show a high degree of canalization (Stearns and Kawecki 1994). Additive genetic variances and covariances between characters reflect past evolutionary events in a population (e.g., McGuigan 2006; Arnold et al. 2008; Badyaev 2010) and help predict future evolutionary change. Our fine-grained analyses of additive genetic variance displayed across life stages suggest that canalization evolved in the past, possibly because of strong stabilizing selection (Wagner et al. 1997; Meiklejohn and Hartl 2002) but is not promoted by current stabilizing selection over the reproductive stage groups (Table 3). The phenotypic and genetic canalization of arrival date, coupled with life-stage-specific selection forces, may represent a constraint for future evolutionary changes of the trait (Gibson and Wagner 2000).

Supplementary Material

Supplementary material can be found at http://www.beheco.oxfordjournals.org/

Funding

The data collection was supported by the Deutsche Forschungsgemeinschaft (BE 916). C.M.A. was supported by a PhD stipend from the AXA Research Fund (contract 043408), and A.C. and F.S.D. by the Centre National de la Recherche Scientifique and the Agence Nationale de la Recherche (grant ANR-08-JCJC-0041-01).

This study was made possible by the long-term monitoring of the common tern population in Wilhelmshaven (Germany), which relied on the work of many researchers, students, and field assistants.

References

Arizaga
J
Arroyo
JL
Rodriguez
R
Martinez
A
San-Martin
I
Sallent
A
.
2011
.
Do blackcaps Sylvia atricapilla stopping over at a locality from southern Iberia refuel for crossing the Sahara?
Ardea
 .
58
:
71
85
.
Arnold
SJ
Bürger
R
Hohenlohe
PA
Ajie
BC
Jones
AG
.
2008
.
Understanding the evolution and stability of the G-matrix
.
Evolution
 .
62
:
2451
2461
.
Atchley
WR
.
1984
.
Ontogeny, timing of development, and genetic variance-covariance structure
.
Am Nat
 .
123
:
519
540
.
Atchley
WR
.
1987
.
Developmental quantitative genetics and the evolution of ontogenies
.
Evolution
 .
41
:
316
330
.
Badyaev
AV
.
2010
.
The beak of the other finch: coevolution of genetic covariance structure and developmental modularity during adaptive evolution
.
Philos Trans R Soc Lond B Biol Sci
 .
365
:
1111
1126
.
Becker
PH
Dittmann
T
Ludwigs
JD
Limmer
B
Ludwig
SC
Bauch
C
Braasch
A
Wendeln
H
.
2008
.
Timing of initial arrival at the breeding site predicts age at first reproduction in a long-lived migratory bird
.
Proc Natl Acad Sci USA
 .
105
:
12349
12352
.
Becker
PH
Ezard
THG
Ludwigs
JD
Sauer-Gurth
H
Wink
M
.
2008
.
Population sex ratio shift from fledging to recruitment: consequences for demography in a philopatric seabird
.
Oikos
 .
117
:
60
68
.
Becker
PH
Ludwigs
JD
.
2004
.
Sterna hirundo common tern
.
BWP Update
 .
6
:
91
137
.
Becker
PH
Wendeln
H
.
1997
.
A new application for transponders in population ecology of the common tern
.
Condor
 .
99
:
534
538
.
Becker
PH
Wink
M
.
2003
.
Influences of sex, sex composition of brood and hatching order on mass growth in common terns Sterna hirundo
.
Behav Ecol Sociobiol
 .
54
:
136
146
.
Bety
J
Giroux
JF
Gauthier
G
.
2004
.
Individual variation in timing of migration: causes and reproductive consequences in greater snow geese (Anser caerulescens atlanticus)
.
Behav Ecol Sociobiol
 .
57
:
1
8
.
Blows
MW
Hoffmann
AA
.
2005
.
A reassessment of genetic limits to evolutionary change
.
Ecology
 .
86
:
1371
1384
.
Blows
MW
.
2007
.
A tale of two matrices: multivariate approaches in evolutionary biology
.
J Evol Biol
 .
20
:
1
8
.
Both
C
Bijlsma
RG
Visser
ME
.
2005
.
Climatic effects on timing of spring migration and breeding in a long-distance migrant, the pied flycatcher Ficedula hypoleuca
.
J Avian Biol
 .
36
:
368
373
.
Charmantier
A
Perrins
C
McCleery
RH
Sheldon
BC
.
2006
.
Age-dependent genetic variance in a life-history trait in the mute swan
.
Proc R Soc Lond B Biol Sci
 .
273
:
225
232
.
Cheverud
JM
Leamy
LJ
Atchley
WR
Rutledge
JJ
.
1983
.
Quantitative genetics and the evolution of ontogeny. 1. Ontogenetic changes in quantitative genetic variance-components in randombred mice
.
Genet Res
 .
42
:
65
75
.
Conklin
JR
Battley
PF
.
2011
.
Impacts of wind on individual migration schedules of New Zealand bar-tailed godwits
.
Behav Ecol
 .
22
:
854
861
.
Debat
V
David
P
.
2001
.
Mapping phenotypes: canalization, plasticity and developmental stability
.
Trends Ecol Evol
 .
16
:
555
561
.
Dittmann
T
Becker
PH
.
2003
.
Sex, age, experience and condition as factors affecting arrival date in prospecting common terns, Sterna hirundo
.
Anim Behav
 .
65
:
981
Dmitriew
C
Blows
MW
Rowe
L
.
2010
.
Ontogenetic change in genetic variance in size depends on growth environment
.
Am Nat
 .
175
:
640
649
.
Dobson
FS
1985
.
The use of phylogeny in behavior and ecology
.
Evolution
 .
39
:
1384
1388
.
Dohm
MR
.
2002
.
Repeatability estimates do not always set an upper limit to heritability
.
Funct Ecol
 .
16
:
273
280
.
Ezard
TH
Becker
PH
Coulson
T
.
2007
.
Correlations between age, phenotype, and individual contribution to population growth in common terns
.
Ecology
 .
88
:
2496
2504
.
Falconer
DS
Mackay
TFC
.
1996
.
Introduction to quantitative genetics
.
4th edn
New York
:
Longman Inc
.
Forslund
P
Part
T
.
1995
.
Age and reproduction in birds - hypotheses and tests
.
Trends Ecol Evol
 .
10
:
374
378
.
Garant
D
Kruuk
LE
McCleery
RH
Sheldon
BC
.
2007
.
The effects of environmental heterogeneity on multivariate selection on reproductive traits in female great tits
.
Evolution
 .
61
:
1546
1559
.
Gibson
G
Wagner
G
.
2000
.
Canalization in evolutionary genetics: a stabilizing theory?
Bioessays
 .
22
:
372
380
.
Gilmour
AR
Gogel
BJ
Cullis
BR
Thompson
R
.
2009
.
ASReml user guide release 3.0 [Internet]
Hemel Hempstead (UK)
:
VSN International Ltd
Available from www.vsni.co.uk
GonzáLez-SolíS
J
Becker
PH
Wendeln
H
.
1999
.
Divorce and asynchronous arrival in common terns, Sterna hirundo
.
Anim Behav
 .
58
:
1123
1129
.
González-Solís
J
Sokolov
E
Becke
PH
.
2001
.
Courtship feedings, copulations and paternity in common terns, Sterna hirundo
.
Anim Behav
 .
61
:
1125
1132
.
Gordo
O
.
2007
.
Why are bird migration dates shifting? A review of weather and climate effects on avian migratory phenology
.
Clim Res
 .
35
:
37
58
.
Kokko
H
.
1999
.
Competition for early arrival in migratory birds
.
J Anim Ecol
 .
68
:
940
950
.
Kruuk
LEB
.
2004
.
Estimating genetic parameters in natural populations using the ‘animal model’
.
Philos Trans R Soc Lond B Biol Sci
 .
359
:
873
890
.
Kruuk
LEB
Hadfield
JD
.
2007
.
How to separate genetic and environmental causes of similarity between relatives
.
J Evol Biol
 .
20
:
1890
1903
.
Lande
R
Arnold
SJ
.
1983
.
The measurement of selection on correlated characters
.
Evolution
 .
37
:
1210
1226
.
Lane
JE
Kruuk
LEB
Charmantier
A
Murie
JO
Dobson
FS
.
2012
.
Delayed phenology and reduced fitness associated with climate change in a wild hibernator
.
Nature
 .
489
:
554
557
.
Lessells
CM
Boag
PT
.
1987
.
Unrepeatable repeatabilities - a common mistake
.
Auk
 .
104
:
116
121
.
Ludwig
SC
Becker
PH
.
2008
.
Supply and demand: causes and consequences of assortative mating in common terns Sterna hirundo
.
Behav Ecol Sociobiol
 .
62
:
1601
1611
.
Ludwigs
JD
Becker
PH
.
2002
.
The hurdle of recruitment: Influences of arrival date, colony experience and sex in the common tern Sterna hirundo
.
Ardea
 .
90
:
389
399
.
McAdam
AG
Boutin
S
.
2003
.
Variation in viability selection among cohorts of juvenile red squirrels (Tamiasciurus hudsonicus)
.
Evolution
 .
57
:
1689
1697
.
McGuigan
K
.
2006
.
Studying phenotypic evolution using multivariate quantitative genetics
.
Mol Ecol
 .
15
:
883
896
.
Meiklejohn
CD
Hartl
DL
.
2002
.
A single mode of canalization
.
Trends Ecol Evol
 .
17
:
468
473
.
Møller
AP
.
2001
.
Heritability of arrival date in a migratory bird
.
Proc R Soc Lond B Biol Sci
 .
268
:
203
206
.
Nakagawa
S
Schielzeth
H
.
2010
.
Repeatability for Gaussian and non-Gaussian data: a practical guide for biologists
.
Biol Rev
 .
85
:
935
956
.
Neto
JM
Gosler
AG
.
2005
.
Breeding biology of Savi’s warbler Locustella luscinioides in Protugal
.
Adea
 .
93
:
89
100
.
Nisbet
ICT
Apanis
V
Friar
MS
.
2002
.
Breeding performance of very old common terns
.
J Field Ornithol
 .
73
:
117
124
.
Nussey
DH
Coulson
T
Fsta-Bianchet
M
Gaillard
JM
.
2008
.
Measuring senescence in wild animal populations: towards a longitudinal approach
.
Funct Ecol
 .
22
:
393
406
.
Nussey
DH
Kruuk
LE
Donald
A
Fowlie
M
Clutton-Brock
TH
.
2006
.
The rate of senescence in maternal performance increases with early-life fecundity in red deer
.
Ecol Lett
 .
9
:
1342
1350
.
van de Pol
M
Verhulst
S
.
2006
.
Age-dependent traits: a new statistical model to separate within- and between- individual effects
.
Am Nat
 .
167
:
766
773
.
Potti
J
.
1998
.
Arrival time from spring migration in male pied flycatchers: individual consistency and familial resemblance
.
Condor
 .
100
:
702
708
.
Pulido
F
.
2007
.
Phenotypic changes in spring arrival: evolution, phenotypic plasticity, effects of weather and condition
.
Clim Res
 .
35
:
5
23
.
Pulido
F
.
2007
.
The genetics and evolution of avian migration
.
Bioscience
 .
57
:
165
174
.
Pulido
F
Berthold
P
Mohr
G
Querner
U
.
2001
.
Heritability of the timing of autumn migration in a natural bird population
.
Proc R Soc Lond B Biol Sci
 .
268
:
953
959
.
Pulido
F
Widmer
M
.
2005
.
Are long-distance migrants constrained in their evolutionary response to environmental change? Causes of variation in the timing of autumn migration in a blackcap (S-atricapilla) and two garden warbler (Sylvia borin) populations
. In:
Bauchinger
U
Goymann
W
JenniEiermann
S
, editors.
Bird hormones and bird migrations: analyzing hormones in droppings and egg yolks and assessing adaptations in long-distance migration
 .
New York
:
New York Academy of Sciences
p.
228
241
.
Qvarnström
A
Brommer
JE
Gustafsson
L
.
2006
.
Testing the genetics underlying the co-evolution of mate choice and ornament in the wild
.
Nature
 .
441
:
84
86
.
Rebke
M
Coulson
T
Becker
PH
Vaupel
JW
.
2010
.
Reproductive improvement and senescence in a long-lived bird
.
Proc Natl Acad Sci USA
 .
107
:
7841
7846
.
Rguibi-Idrissi
H
Julliard
R
Bairlein
F
.
2003
.
Variation in the stopover duration of Reed Warblers Acrocephalus scirpaceus in Morocco: effects of season, age and site
.
Ibis
 .
145
:
650
656
.
Robson
D
Barriocanal
C
.
2011
.
Ecological conditions in wintering and passage areas as determinants of timing of spring migration in trans-Saharan migratory birds
.
J Anim Ecol
 .
80
:
320
331
.
Roff
DA
Fairbairn
DJ
.
2011
.
Path analysis of the genetic integration of traits in the sand cricket: a novel use of BLUPs
.
J Evol Biol
 .
24
:
1857
1869
.
Schauroth
C
Becker
PH
.
2008
.
Post-fledging body mass increase in Common Terns Sterna hirundo: influence of age, sex and year
.
Ibis
 .
150
:
50
58
.
Stearns
SC
Kawecki
TJ
.
1994
.
Fitness sensitivity and the canalization of life-history traits
.
Evolution
 .
48
:
1438
1450
.
Stewart
RLM
Francis
CM
Massey
C
.
2002
.
Age-related differential timing of spring migration within sexes in passerines
.
Wilson Bull
 .
114
:
264
271
.
Szostek
L
Becker
PH
.
2012
.
Terns in trouble: demographic consequences of low breeding success and recruitment on a common tern population in the German Wadden Sea
.
J Ornithol
 .
153
:
313
326
.
Teplitsky
C
Mouawad
NG
Balbontin
J
de Lope
F
Møller
AP
.
2011
.
Quantitative genetics of mrigration syndromes: a study of two barn swallow populations
.
J Evol Biol
 .
24
:
2025
2039
.
Waddington
CH
.
1942
.
Canalization of development and the inheritance of acquired characters
.
Nature
 .
150
:
563
565
.
Wagner
GP
Booth
G
Bagheri
HC
.
1997
.
A population genetic theory of canalization
.
Evolution
 .
51
:
329
347
.
Watson
MJ
Spendelow
JA
Hatch
JJ
.
2012
.
Post-fledging brood and care division in the roseate tern (Sterna dougallii)
.
J Ethol
 .
30
:
29
34
.
Wilson
AJ
Kruuk
LE
Coltman
DW
.
2005
.
Ontogenetic patterns in heritable variation for body size: using random regression models in a wild ungulate population
.
Am Nat
 .
166
:
E177
E192
.
Wilson
AJ
Reale
D
.
2006
.
Ontogeny of additive and maternal genetic effects: lessons from domestic mammals
.
Am Nat
 .
167
:
E23
E38
.
Wilson
AJ
Réale
D
Clements
MN
Morrissey
MM
Postma
E
Walling
CA
Kruuk
LE
Nussey
DH
.
2010
.
An ecologist’s guide to the animal model
.
J Anim Ecol
 .
79
:
13
26
.

Author notes

Handling editor: Regina Macedo

Supplementary data