From pattern to process : extracting ecology from time series

Abstract

Abstract Patterns in time series are observational records of past ecological processes, and identifying these processes remains a major challenge to population ecologists. Until recently, analysis of time series has concentrated on statistical description of density dependent structure of patterns, while the ecological factors that underlie the structure remain to a large part unknown. In this thesis, mechanistic properties of a model building on existing statistical framework are explored and discussed. By incorporating a two species trophic interaction and seasonality, the model may allow for direct ecological interpretation of density dependence. General model dynamics are explored and the biologically relevant parameter space identified. It is further shown that the relative length of summer/winter can have profound consequences for model dynamics, and very possibly force populations between regions of stable, cyclic and unstable dynamics. Supposing that the model can adequately describe the outcome of biological interactions, it is fitted to a large set of time series of voles from Hokkaido, Japan. From qualitative properties of six independent estimates of model coefficients, inferences are done about the parameter aggregates that constitute these coefficients. Inferences from two different ecological scenarios; predator-prey and grazer-vegetation, are then compared with the parameter space found appropriate from when studying the general model dynamics. Under specified assumptions, support is found for that signals in the time series are a result of an interaction between vole and its food resource.