Speaker
Description
There is a clear need for robust tools for prediction and inference of ecological dynamics that do not depend on precisely knowing how ecosystems work. Data-driven methods such as empirical dynamic modeling (EDM) allow us to learn dynamics with minimal assumptions. Here, I will introduce the basic ideas of EDM and then discuss two recently developed approaches to incorporating external driving variables that expand our capacity to predict responses to environmental change. In the first, we constrain EDM by modeling dynamics on a "metabolic time step" using the Metabolic Theory of Ecology. This approach improved forecast accuracy in 18 of 19 empirical ectotherm time series (by 19% on average), with the largest gains in more seasonal environments. In the second, we addressed the need to anticipate qualitative changes in dynamics or ‘regime shifts’ in response to a slowly changing driver. Specifically, by incorporating data on the putative driver, we used EDM to reconstruct bifurcation diagrams and infer tipping points in a suite of simulated systems. We applied this approach to empirical time series to predict transitions to, and away from, chaotic behavior as exogenous forcing is changed in an experimental community.
