Matthew P. Hammond
Department of Biology, McMaster University, Hamilton, ON L8S 4K1, CANADA.
Ecosystems and their components (e.g., organisms, physicochemical variables) are dynamic in space and time. This dynamism makes ecological change notoriously difficult to study and manage. This thesis therefore aims to develop new ways of using spatiotemporal information for inference and prediction. Applying theoretical and statistical concepts to patterns of spatiotemporal variation in aquatic ecosystems led to three discoveries that show promise as ecological applications. First, I show that temporal variability of an ecosystem process can be inferred from its spatial variability. This application may be the first quantitative form of the widely-used method, space- for-time substitution. Its use is supported by an analytical framework giving the conditions under which space is a good surrogate for time. Second, I demonstrate the use of spatiotemporal patterns to predict responses of variables when ecosystem fragments are connected. Connection leads to large shifts in spatiotemporal pattern and other response metrics (e.g., temporal variability) for variables showing asynchrony and concentration gradients among sites (e.g., populations). Meanwhile, these changes are minimal if variables exhibit synchrony and homogeneity across space (e.g., energetic variables). A final discovery is that temporal attributes like stability are strong predictors of persistent spatial variation – a pattern that reflects how reliably resource concentrations occur in the same locations. This finding suggests the potential of time- for-space substitution, where one or few well-resolved time series could be used to infer landscape patterns. All but one of the tested approaches were data efficient and broadly-applicable across ecosystems and ecological processes. They thus contribute new possibilities for prediction when data are scarce, as well as new perspectives on dynamics in multi-variable landscapes. Research here shows that work at the intersection of spatial and temporal pattern can strengthen the interpretation of ecosystem dynamics and, more generally, foster synthesis from populations to landscapes.