Equation-free Mechanistic Ecosystem Forecasting

Abstract

Mechanistic understanding is important for effective policy and management recommendations on climate, epidemiology, financial regulation, and much else. Two classic approaches have been commonly used for this purpose: correlation analysis and parametric models using a set of assumed equations. For approach 1, we face the long-lasting problem that correlation does not imply causation. For approach 2, we encounter the difficulty that we do not know the exact set of equations and ecosystems are complex.? Here, we show that the objective is better addressed using an alternative equation-free approach based on nonlinear state space reconstruction. This approach can distinguish causality from correlation and provide better forecasting skills for ecosystem dynamics, thus leading to mechanistic understanding.