I. Nonlinear dynamics

Being able to forecast into future is one of the ultimate goals of science. If so, conversely, forecast capability may be used effectively to examine system dynamics. 

Based on out-of-sample forecast skill, we employ nonlinear time series techniques to investigate system dynamics. These methods are rooted in the theory of state-space reconstruction of the system attractor. These techniques have ability to distinguish low-dimensional nonlinear dynamics from high-dimensional linear noise in natural time series. 

Using these nonlinear methods, we investigate the relative contributions of intrinsic and extrinsic processes to the regulation of biological populations, detect potential abrupt ecosystem changes in response to environmental changes, characterize neural feedback control of flying behavior of drosophila, and are developing more applications. 

 Moreover, we extend these methods to detect causal relationships among ecosystem components. These methods are applied to improve ecosystem management, for example, forecasting fisheries stock sizes in response to climate changes and fishing. These methods were originally developed by Prof. George Sugihara (Scripps Institution of Oceanography, USA).