Use of stochastic models for the prognosis of qualitative transitions in ENSO dynamics
Abstract
We consider the problem of qualitative transitions prognosis in ENSO system behavior from observed time series. For this purpose the approach based on construction of parameterized stochastic models of discrete evolution operator is proposed. These models have a form of superposition of deterministic function and stochastic component in a form of state-dependent random function. Artificial neural networks with certain priors are used for parameterization of the models. The ability of the approach to provide prognosis for times greater than observation time interval is demonstrated on time series taken from intermediate complexity ENSO models: Galanti-Tziperman (JAS, 2000) DDE model and Jin and Neelin (JAS, 1993) PDE model; both models are derived as simplification of Cane-Zebiak coupled ocean-atmospheric model of ENSO (MWR, 1987). Slow drifts of control parameters were introduced to these ENSO models to simulate slowly changing external conditions of the system. Prognosis of qualitative behavior from scalar time series is shown, including prognosis of PDFs evolution, spectral density evolution and prognosis of main qualitative transitions.
- Publication:
-
AGU Fall Meeting Abstracts
- Pub Date:
- December 2011
- Bibcode:
- 2011AGUFMNG52A..08M
- Keywords:
-
- 4410 NONLINEAR GEOPHYSICS / Bifurcations and attractors