Empirical dynamical system modeling of ENSO using nonlinear inverse techniques

A statistical technique is presented that allows for the empirical derivati on of dynamical system equations from data. It is based on multiple nonpara metric regression analysis and is applicable to a broad class of physical s ystems. It is applied to differential delay equations as well as to ordinar y differential equations. The aim of this paper is to illustrate this techn ique in the context of the El Nino-Southern Oscillation (ENSO) phenomenon. A set of reduced models is derived from an intermediate coupled atmosphere- ocean model of the tropical Pacific and from a state-of-the-art coupled gen eral circulation model simulation. The analysis in this paper focuses on th e dimensionality issue as well as on the role of nonlinearities. The empiri cal technique presented in this study helps to identify key ENSO processes and to explain physical peculiarities of ENSO simulations.