Quantifying biases in a long-term field study: finding Wytham’s missing tits

Inferences drawn from long-term field studies of animals are vulnerable to biases in observability of different classes of individuals, which may lead to biases affecting ecological and evolutionary conclusions. Advancing statistical techniques are providing increasing evidence that individuals missing from population samples show consistent individual differences, although any resulting biases remain poorly understood. In this thesis, I investigate how biases can occur in estimating individual variation in reproductive success and residence behaviour within a natural great tit (Parus major) population. The primary objective of this thesis is to investigate whether individuals that are detected using standard sampling techniques are representative of the entire population. To do so, I use RFID technology to sample more completely across seasons, investigating individual detectability throughout the year. I show that females that fail early in their breeding attempt, before they are identified under standard sampling protocols, are more likely to be young and immigrants from outside the study population and breeding in poorer quality habitats. As a consequence, estimates of the fitness of immigrants versus residents, age effects on fitness, and the age and immigration specific effects of habitat quality derived from standard survey methods are likely to be underestimates. I show that the high failure rate amongst immigrant females may be a result of late arrival time to the breeding grounds. These findings imply that costs of dispersal might impact on reproductive success. Moreover, I show that the exclusion of early nest failures from analyses can lead to underestimates in the effects of dispersal on reproductive success; as a consequence, current work concluding that immigrants have lower fitness than residents is likely to underestimate the size of this effect. This thesis also provides some of the first baseline data tracking the presence or absence of passerines within a population across seasons. Our findings suggest that certain individuals are less likely to be detected as a function of their phenotype. We identify sex as a predictor of failure to be detected breeding, with males less likely to be detected. We also find that the likelihood of detection during successive breeding and winter surveys is repeatable across years. This adds to a gathering body of knowledge suggesting that sampling at certain times of the year selects for specific phenotypes. This thesis represents the first large-scale investigation of individuals that remain undetected in a long-term great tit population study, demonstrating that more attention should be given to the question of whether methods for surveying natural populations introduce systematic biases that influence conclusions about ecological and evolutionary processes.