The hierarchical modelling framework represents a powerful and flexible framework for modelling and inference about ecological processes. It admits an explicit and formal representation of the data model into constituent components for observations and ecological process.
The model for the ecological process of interest (the 'process model'), describes variation (spatial, temporal, etc.) in the ecological process that is the object of inference. This process is manifest in some (typically unobservable, or only partially so) state variable, say z(i,t), e.g., abundance or occurrence at some point in space (i) and time (t). Whereas the model for the observations conditional on the ecological process (the 'observation model'), describes the probabilistic mechanisms by which the data are obtained. Whereas almost all classical methods focus exclusively on models that describe the sampling process, through the closely related probability distribution [data parameters], the incorporation of these two component models into a single unified model (referred to as a hierarchical or state-space model) results in a generic and flexible strategy for conducting inference about population and community structure from biological sampling data. In particular, while the [data, process, parameters] model may be very complex, the two component sub-models are typically very simple, even for some very complex data structures. This yields surprisingly simple solutions to some very complex problems.
Examples include: hierarchical models of simple counts; modelling individual heterogeneity in capture-recapture models; estimating community structure by modelling occurrence of species; wide variety of examples involving many taxa (birds, amphibians, mammals, insects, plants); and detailed explanations describing the implementation of hierarchical models using freely available software such as R and WinBUGS. Computing support is provided as technical appendices in an online companion site.