The University of Arizona

College of Agriculture and Life Sciences

Forecasting Ecology in a Changing World

Wednesday, February 21, 2018

Speaker:  Michael Dietze, Boston University
Date: Wednesday, February 21st, 2018
Time: 3:00-4:00 pm
Location: ENR2, S107
ABSTRACT: Ecosystems are responding to a multitude of changes that challenge our ability to understand and manage natural systems. As such, the practice of forecasting the state of ecosystems and their services is undergoing a shift away from long-term, deterministic projection, to include both a richer accounting and partitioning of uncertainties (better to be honestly uncertain than over-confidently wrong) and an increased emphasis on near-term forecasts that are updated as new data become available. Near-term forecasts provide the opportunity to iteratively cycle between performing analyses and updating predictions in light of new evidence. This iterative process of gaining feedback, building experience, and correcting models and methods is critical for improving forecasts. Iterative, near-term forecasting will accelerate ecological research, make it more relevant to society, and inform sustainable decision-making under high uncertainty and adaptive management. Herein I present a general framework for understanding ecological predictability. The implications of this framework are assessed conceptually and linked to classic questions in ecology, such as the relative importance of endogenous (density dependent) versus exogenous factors, stability versus drift, and the spatial scaling of processes. The framework is used to make a number of novel predictions and reframe approaches to experimental design, model selection, and hypothesis testing. Next, the quantitative application of the framework to partitioning uncertainties is illustrated using a short-term forecast of net ecosystem exchange. Finally, I advocate for a new comparative approach to studying predictability across different ecological systems and processes and lay out a number of hypotheses about what limits predictability and how these limits should scale in space and time.