No evidence for fitness signatures consistent with increasing trophic mismatch over 30 years in a population of European shag Phalacrocorax aristotelis

Abstract

As temperatures rise, timing of reproduction is changing at different rates across trophic levels, potentially resulting in asynchrony between consumers and their resources. The match-mismatch hypothesis (MMH) suggests that trophic asynchrony will have negative impacts on average productivity of consumers. It is also thought to lead to selection on timing of breeding, as the most asynchronous individuals will show the greatest reductions in fitness. Using a 30-year individual-level dataset of breeding phenology and success from a population of European shags on the Isle of May, Scotland, we tested a series of predictions consistent with the hypothesis that fitness impacts of trophic asynchrony are increasing. These predictions quantified changes in average annual breeding success and strength of selection on timing of breeding, over time and in relation to rising sea surface temperature (SST) and diet composition. Annual average (population) breeding success was negatively correlated with average lay date yet showed no trend over time, or in relation to increasing SST or the proportion of principal prey in the diet, as would be expected if trophic mismatch was increasing. At the individual level, we found evidence for stabilising selection and directional selection for earlier breeding, although the earliest birds were not the most productive. However, selection for earlier laying did not strengthen over time, or in relation to SST or slope of the seasonal shift in diet from principal to secondary prey. We found that the optimum lay date advanced by almost 4 weeks during the study, and that the population mean lay date tracked this shift. Our results indicate that average performance correlates with absolute timing of breeding of the population, and there is selection for earlier laying at the individual level. However, we found no fitness signatures of a change in the impact of climate-induced trophic mismatch, and evidence that shags are tracking long-term shifts in optimum timing. This suggests that if asynchrony is present in this system, breeding success is not impacted. Our approach highlights the advantages of examining variation at both population and individual levels when assessing evidence for fitness impacts of trophic asynchrony.

Keywords: Ammodytes marinus; breeding phenology; environmental change; lesser sandeel; long-term study; match-mismatch hypothesis; stabilising selection; trophic asynchrony.

FIGURE 1

Schematic of requirements of the match–mismatch hypothesis (a, b, vertical arrows in [a] represent asynchrony) and consequences for breeding success (c, d). The impacts of asynchrony on breeding success may increase in relation to year or temperature (e, f). Asynchrony may have consequences on diet (g, h) with impacts on breeding success (i, j). Left‐hand plots (a, c, e, g and i) show expected outcomes at the population (between year) level (hereafter BSp) and right‐hand plots (b, d, f, h and j) show expected outcomes at the individual (within year) level (hereafter BSi). All asynchrony predictions are generated under the assumption that historically timings were synchronous in the average year. Red lines are representative of resource, and black lines are representative of consumer. In (a), consumer phenology is described under two scenarios, where there is no trend but substantial inter‐annual variation in timing (Scenario A), and where timing is responding linearly but more slowly than the resource (Scenario B). In (b), solid lines indicate synchronous years, and dashed lines indicate asynchronous years. In (c) negative slope corresponds to the dotted consumer line in (Scenario A in a), positive slope corresponds to dashed consumer line in (Scenario B in a) and constant slope (solid line) corresponds to synchronous consumer and resource. In (f and j), a more negative y‐value represents stronger directional selection on laying date

FIGURE 2

(a) sea surface temperature (SST) as a function of year, (b) lay date as a function of year and c) variation in lay date as a function of SST. Red dots depict annual mean lay dates. Black lines indicate average trends in SST (a) and lay date (b) over time, and lay date over SST (c). Solid slope estimates represent significant trends and dashed slope estimates represent insignificant trends. Ordinal day refers to number of days after January 1st, allowing for leap years. Dashed lines represent 95% confidence intervals around the slope estimate

FIGURE 3

The relationship between lay date and breeding success on the data scale (a) at the between‐year level (BSp) and (b) at the within year level (BSi). Points in (a) are mean values from the data, red line corresponds to the slope across annual means estimated from the core model and estimates the change in mean fitness, and grey area corresponds to 95% credible intervals. Ordinal day refers to number of days after January 1st, allowing for leap years. Black lines in (b) correspond to best linear unbiased predictors of the within‐year slopes estimated in different years and the red line is the average within‐year slope (estimated from the fixed effects), with all coefficients taken from the core model (intercept based on the average annual lay date). See Figure S1 for a projection of these slopes on the logit scale. Solid red lines indicate significant slopes

FIGURE 4

Predicted effects of year (a), sea surface temperature (SST) in the current year (c) and proportion of principal prey in diet (e) on mean population‐level breeding success (back transformed from model output). Predictions for the effects of year (b), SST in the current year (d) and the within‐year change in the proportion of principal diet (f) on the strength of directional selection on relative lay date (i.e. the slope of individual breeding success regressed on relative lay date). Black points in (a, c and e) represent annual mean estimates. Black points in (b, d and f) represent annual predictions. Red lines indicate posterior mean response and grey areas represent 95% credible intervals from the year model (a, b), the SST model (c, d), the sandeel model (e) and the bivariate model (f). For (a) and (c), model predictions are made correcting for the mean annual lay date. Dashed red lines indicate non‐significant slopes

FIGURE 5

Between‐year (a) and within‐year (b) proportions of sandeels in the diet of shags during the chick‐rearing period. In (a), each point represents a yearly mean of the proportion of 1+ group sandeels in the total sandeel biomass and mean date of sample collection in that year (Ordinal day). The red line is the estimate from the core model of the change in diet proportion with mean lay date and back‐transformed from the logit scale, the grey area corresponds to 95% credible intervals. In (b), within‐year changes are represented with grey lines and the average within‐year slope across all years with a red line; all from the core model. Dashed red line indicates non‐significant slope estimate, and solid red line indicates significant slope estimate