Explaining patterns of age-specific performance

Abstract

Individual life histories are frequently studied to gain insight into the mechanisms of ageing. However, various challenges complicate the accurate quantification of age-specific variation in fitness. In this thesis I develop and apply methods to accurately characterise patterns of ageing, and to explain why such patterns arise. All mammals and birds have an upper bound on litter size, and for many species this limit is quite low. In addition, in many species, not all individuals breed at every possible opportunity. Reproduction should consequently be considered as two processes: whether an individual breeds or not and the number of offspring produced. These processes mean that reproduction in many species does not follow a Poisson process as is often assumed in analyses of breeding performance. A more appropriate model for a repeated ordinal response like annual reproductive success is a proportional odds model with a random intercept for individuals. Such a model has not previously been used in ecology or evolutionary biology. I apply this model to analyse age and temporal variation in the number of fledglings produced annually by male and female common terns (Sterna hirundo). I use data collected from this intensively studied, long-lived species, repeatedly throughout the thesis. The proportional odds analysis reveal that reproductive performance in females initially increased with age, before declining as individuals began to senesce. But why does this pattern arise? Is it purely an effect of getting physiologically older or are other processes involved? I estimate the effect of the length of time spent with the current partner using the common tern data. Despite the quality of the data, it is not always obvious if unmarked partners are new or not. I use a hierarchical Bayesian model of the steps that lead to the number of fledglings. Modelling this complicated process requires a complex model, but results show that no substantial amount of observed age-related patterns in reproductive performance can be attributed to length of pair bond. While the proportional odds and Bayesian analyses account for repeated measures on individuals they do not account for compositional change. Such a change in the composition of the population caused by heterogeneity between individuals can mask true rates of individual change. I develop a novel retrospective decomposition method related to the Price equation to address this issue. The equation gives the exact contributions of selective disappearance and average change in individual performance among survivors to the aggregate change at the level of the population. This equation can be extended by including a term for the compositional change due to selective appearance of individuals in the study population. I apply this decomposition to the common tern dataset to disentangle whether apparent increases and decreases in reproductive performance with age reflect genuine changes within individuals or are an artefact of compositional change in a heterogeneous population. I show an improvement in average reproductive performance of individuals over most of adult life and give support for reproductive senescence at old ages. I show that the contribution of compositional change is of minor importance, suggesting that population-level averages accurately capture the individual-level ageing process well. Can the decomposition method I develop be applied to other systems? Does it lead to similar conclusions? I apply it to two different datasets dealing with functioning at old age in humans: the ability to live independently in the Danish 1905-cohort, and cognitive functioning for people aged 80 and older participating in the Chinese Longitudinal Health and Longevity Survey. In both studies I reveal that average individual functioning declines at old ages. Although the decline is also apparent at the population level it is less strong due to the tendency of individuals with lower functioning to drop out earlier. Finally, I illustrate the general use of the decomposition by applying it to epidemiological and economic studies in the appendix. Overall, I find that reproductive performance improves over many age classes before senescence begins. Numerous processes can influence rates of age-related change, with results apparently specific to the trait and population under study.