In marine ecosystems, climatic fluctuations and other physical oceanographic variables affect the population dynamics of organisms by exerting an influence on their vital rates. This influence is mediated by drivers such as food availability and environmental conditions during the breeding period or migration (Duffy-Anderson et al., 2005; Lee, 2011). For example, large-scale climatic phenomena and oceanographic processes produce changes in water temperatures and currents that generate spatio-temporal variation in the production, distribution, and abundance of the prey consumed by marine predators such as seabirds (Durant et al., 2004). Consequently, these processes may affect the foraging ecology of seabirds (Navarro and Gonzalez-Solís, 2009; Weimerskirch et al., 2012) and influence their survival, reproduction and population dynamics (Jenouvrier et al., 2005; Genovart et al., 2013).

Like environmental variability, individual covariates such as age and sex also influence vital rates (Stearns, 1992). Age-specific differences in breeding performance have been observed in several seabirds (Nevoux et al., 2007; Vieyra et al., 2009; Pardo et al., 2013). In most cases young individuals perform less well than older individuals (Martin, 1995), thereby indicating improvement in performance with age leading to a peak and then decline (i.e., senescence) (Clutton-Brock, 1988). Several non-exclusive hypotheses have been proposed to explain these age-related variations in breeding performance: a progressive appearance or disappearance of phenotypes, age-related improvements in the skills needed for reproducing, and an optimization of reproductive effort (Forslund and Pärt, 1995).

Demographic parameters in wild populations are thus expected to be shaped by both individual covariates and environmental variability (e.g., Stearns, 2000; Nevoux et al., 2007; Oro et al., 2010). Despite this, few studies have ever examined at the same time how environmental conditions affect age patterns in breeding performance (Bunce et al., 2005; Pardo et al., 2013; Oro et al., 2014). Depending on environmental conditions, different age patterns can exist (see Oro et al., 2014): in most studies, either (a) differences among age classes in breeding performance were found to decrease under better environmental conditions (Sydeman et al., 1991; Laaksonen et al., 2002; Barbraud and Weimerskirch, 2005; Bunce et al., 2005) or (b) environmental conditions had no clear influence on breeding performance (Nevoux et al., 2007; Vieyra et al., 2009; Lee, 2011; Pardo et al., 2013); just one study found that differences between age-classes were greater under favorable environmental conditions (Oro et al., 2014).

Here we analyze 23 years of data from a population of Scopoli's Shearwater (Calonectris diomedea), a long-lived and highly monogamous seabird with extreme life-history traits that include high adult survival rates, low fecundity (females lay a single egg), intermittent breeding, and delayed sexual maturity (Thibault, 1994; Thibault et al., 1997). Shearwaters breed in the Mediterranean and, after reproducing, perform long trips to their Atlantic wintering grounds where there are important upwellings that provide abundant food resources (González-Solís et al., 2007). We tested the influence of environmental conditions at different spatial scales on two breeding parameters, laying date and egg volume, in individuals of different age classes. Our main aim was to test (a) whether breeding performance increased with age and (b) whether or not differences between age classes depend on environmental conditions at different spatial scales.

We found a positive effect of age (log-transformed) on the breeding parameters analyzed (Table 1). Older individuals tend to lay earlier and to produce larger eggs than young birds (Table 1, Figure 1, and Appendix B in Supplementary Material). We found a large temporal variability in both parameters. In both cases, the retained model, i.e., the one with the lowest AICc value, assumed an effect of age, year and their statistical interaction, indicating that the strength of the relationship between breeding performance and age was changing across the years (Models 1 and 1; Table 1). The coefficient of determination (R²) between the fitted and observed values of the selected models was 0.74 and 0.67 for laying date and egg volume analysis, respectively. Another important finding was that in both analyses the parameter accounting for across-individual variation was large. When we estimated the variance explained by Nakagawa and Schielzeth's (2013) method, we obtained R²_m = 0.39 and R²_c = 0.64 for laying date (Model 1) and R²_m = 0.28 and R²_c = 0.53 for egg volume analysis (Model 1). Thus, the variance explained by the random effect (i.e., the breeder) was 0.24 and 0.25 in laying date and egg volume, respectively. Environmental covariates were negatively associated with the parameters considered, but although this association was statistically significant, the covariates explained a relatively small part of the total temporal variability (NAO winter index explained about 24% of variance in laying date and SOI index 20% in egg volume; Table 2). For this reason none of the models that included climatic indexes were retained by the model selection procedure. Similarly, fishing discards taken as a proxy of food availability had a positive effect on both parameters, but the variance explained was low and this effect cannot be considered an important predictor of laying dates and egg volume. The year-by-age interaction term is indicating that differences between age classes were more marked under good environmental conditions (negative values of the climatic indexes; Figure 1).

As has been previously reported in other long-lived animals (Clutton-Brock, 1988; Sæther, 1990; Forslund and Pärt, 1995) breeding performance in Scopoli's Shearwater increases with age; younger breeders have later laying dates and smaller egg volumes than older breeders. Since egg size and laying dates are positively associated with breeding success in several bird species and also in Scopoli's Shearwaters (e.g., Ramos et al., 2003), it is expected that fertility also increases with age. This increase in reproductive performance could be mediated by a gain in breeding skills over the years, which would enhance reproduction, or alternatively by the existence of a selection filter (the so-called “selection hypothesis”), whereby older age classes include a higher percentage of good breeders than younger age classes (Curio, 1983). In Storm petrel Hydrobates pelagicus, for example, individuals that survive their first breeding attempt should have a greater likelihood of breeding successfully the following year, probably due to the progressive selection of higher-quality individuals (e.g., Sanz-Aguilar et al., 2008). Furthermore, a previous study of Scopoli's Shearwater in our study colony (Genovart et al., 2013) detected that a large number first-time breeders do not return to the colony, either because they die (i.e., indicating thus the inherent costs of first reproduction) or permanently disperse to other breeding sites. Aside from the increase in breeding performance with age, several studies of long-lived animals have reported a decline in performance as a result of senescence (Vieyra et al., 2009; Pardo et al., 2013; Oro et al., 2014). We did not find a similar pattern, however the lack of senescence in our study may either be due to the small sample size (i.e., a lack of statistical power owing that there were only 4 individuals older than 20 years of age), to an increase of skip breeding with age, or to within-cohort phenotypic selection hiding senescence patterns (Cam et al., 2002).

Beside the clear and positive effect of age, we found that breeding performance varied across individuals and over the study period, the last probably reflecting environmental variability. We investigated whether the temporal variability was mirroring the fluctuations of large-scale climatic indices (namely NAOw and SOI) or the amount of fishing discards (a proxy for food availability just before the breeding season). NAOw and SOI are expected to influence breeding success of Scopoli's shearwater indirectly through their effects on marine productivity and oceanic conditions at the wintering grounds. Their influence on breeding performances, however, can be complex as climatic conditions might affect birds' energetic reserves through for example adverse flying conditions (Knutson et al., 2008; Arizmendi-Mejía et al., 2013) or alternatively they might influence them indirectly by changing the availability of prey, i.e., water column mixing, increasing river run off, and primary productivity (Lloret et al., 2001; Ottersen et al., 2001). This complexity is reflected in the fact that environmental covariates alone, despite being significant, explained only up to 24% of the yearly variability in breeding performances. A stronger effect of environmental covariates has been found for other seabirds (e.g., Frederiksen et al., 2004; Tavecchia et al., 2007; Dunn and Winkler, 2010; Barbraud et al., 2011). Genovart et al. (2013) found that the NAO winter index explained up to 40% of the temporal variance in breeding success in Scopoli's shearwater, but they included birds of unknown age and also nests monitored in an additional colony of the western Mediterranean (Chafarinas Island). Here we have restricted the analysis to birds of known age because we were interested in investigating the age-dependent pattern of breeding performance and whether environmental variability might change that pattern each year. It is possible that the statistical signal is lost in a smaller data set.

Our results showed an important heterogeneity among individuals in both, egg volume and laying dates (approximately 25% of the total variance of the retained model). At present we do not know the reason for such heterogeneity. Human disturbance and terrestrial predators are known to trigger dispersal and determined breeding success in seabirds (Cam et al., 2004; Fernández-Chacón et al., 2013). The studied colony however is located in a protected site with no human disturbances or alien predators. We can thus exclude the influence of these factors as a potential cause for nest quality heterogeneity. Differences among individuals might be due to difference in body size, individual investment, or individual ability in exploiting anthropogenic food resources (i.e., fishing discards, Votier et al., 2010). This individual heterogeneity might explain part of the differences between age classes, a pattern that was more marked under good environmental conditions than during harsh years, as it has been recorded for Audouin's gull (Oro et al., 2014). Other process that might be involved in the differences between the two patterns (good vs. bad years) is intermittent breeding, which has been commonly found in the study species (Mougin et al., 1997; Genovart et al., 2013). Previous studies on other seabirds found that during harsh years, individuals in bad condition may skip breeding (Cam and Monnat, 2000; Pyle et al., 2001; Sanz-Aguilar et al., 2008) and therefore the differences between age classes in breeding performance might be more marked in good years. In Scopoli's shearwater, males do not feed females during courtship as other seabirds (Jouanin et al., 2001), therefore male quality does not affect females' body condition and can be ruled out as a possible explanation. However, female reproductive investment might depend on male quality (Cunningham and Russell, 2000), but we have not enough data to assess whether shearwaters exhibit assortative mating. Out of nine cases in which both members of the pair were of known age, only in one case, female and male had the same age. Nevertheless, when males were included in the data set, results did not change.

Our results show clearly that the interplay of environmental conditions and individual age influences the observed breeding performance of Scopoli's shearwaters. Neither a proxy of the local food availability nor the large-scale climatic indexes can explain the temporal variability in the age pattern of breeding performance. This complex pattern is further complicated by the potential effects of individual quality. Future research should focus on the role of local climatic indexes and direct measures of food availability as predictors of breeding performances in long-lived birds.

Funds were provided by the Balearic Government (FEDER program) and the Spanish Ministry of Economy and Competitiveness (ref. CGL2013-42203-R).