CORAL REEFS AS CHAOS: AN ASSUMPTION-FREE, SYSTEM-STATE APPROACH TO CAUSALITY, DYNAMICS AND PREDICTIONS Stopnitzky, SK., Munch, SB., Potts, DC. January, 2016.
Although there is a high risk of continued reef loss on a global scale, responses to widespread stressors at local and regional scales indicate that resilience to chronic stress is possible but also variable. Coral reefs are complex systems that exhibit nonlinear behavior, including chaos, feedbacks, multistabilities, cascading effects, adaptation and emergent phenomena. Using traditional models to resolve these dynamical processes that control resilience is problematic due to error from excluded variables, incorrectly identifying mirage correlations as system drivers, and untestable assumptions about relationships between variables. Alternatively, a changing coral reef can be considered a trajectory through different states, whose change over time depends on previous states and is determined by a set of rules. We present a promising new technique for understanding and forecasting ecosystems that is adapted from single-species Empirical Dynamic Modeling, using time series data to reconstruct nonlinear state-space. This reconstruction preserves the topology of its chaotic attractor manifold, which represents a trajectory of linked variables through state-space, allowing us to correctly discern shared causal drivers from interactions. Without the need for error-inducing model assumptions, this approach also outperforms all other tested models for forecasting system dynamics. We show ecosystem-scale predictions from simulated and real data that demonstrate the tremendous value of this tool for improving our understanding and management of coral reefs in a changing world.