UC Berkeley

Life history variation is a general feature of natural populations. Most studies assume that local processes occur identically across individuals, ignoring any genetic or phenotypic variation in life history traits. In part, this is because a realistic treatment of individual heterogeneity results in very complex population models. Fitting models with individual heterogeneity to real data is further complicated by random effects in groups of the data, observations set at specific intervals, and the non-independence of data following a cohort of individuals through time. In this dissertation, I assume that individuals differ in the duration they spend in each developmental stage and also in the amount of time they live. Stage durations and survival times follow probability distributions with parameters specific to populations and stages. Parameters of these distributions may also include random effects when considering a subset of sampled populations and covariates such as temperature. In the first chapter I formulate a model and likelihood for variable development, using the time-to-event model framework. In the second chapter I use this model to ask whether field populations of herbivorous arthropods (Tetranychus pacificus) form host-associations on different cultivars of the same host species. In the third chapter I incorporate variable development with variable survival and ongoing reproduction in a stage-structured population model. I explore the ability of the approximate Bayesian computation framework to fit such a complex model to data, evaluating posterior distributions and model performance.