![Hier klicken, um den Treffer aus der Auswahl zu entfernen](images/unchecked.gif) |
Titel |
ENSO forecasting based on noise sampling and low-frequency variability |
VerfasserIn |
Dmitri Kondrashov, Mickael Chekroun, Michael Ghil |
Konferenz |
EGU General Assembly 2010
|
Medientyp |
Artikel
|
Sprache |
Englisch
|
Digitales Dokument |
PDF |
Erschienen |
In: GRA - Volume 12 (2010) |
Datensatznummer |
250034850
|
|
|
|
Zusammenfassung |
The El-Niño/Southern-Oscillation (ENSO) phenomenon dominates interannual climate
signals in and around the Tropical Pacific Ocean and affects the atmospheric circulation and
air–sea interaction over many parts of the globe. In particular, these effects are significant
during ENSO’s extreme phases, El Niño and La Niña, and include large anomalies in rainfall
and temperatures. In practice, accurate long-term forecasting of ENSO beyond 6
months remains a challenge for current state-of-the art dynamical and statistical
models.
Kondrashov et al. (2005) developed an Empirical Mode Reduction (EMR) model of
ENSO based on monthly time series of sea surface temperature (SST) anomalies in a tropical
belt spanning the three major ocean basins. EMR is a methodology for constructing
stochastic models based on the observed evolution of selected climate fields; these models
represent unresolved processes as multivariate, spatially correlated stochastic forcing. In
EMR, multiple polynomial regression is used to estimate the nonlinear, deterministic
propagator of the dynamics, as well as multi-level additive stochastic forcing, directly from
the data set. The EMR-ENSO model has quite competitive forecast capabilities, which
are due to its nonlinear dynamical operator’s ability to capture ENSO’s leading
quasi-quadrennial and quasi-biennial oscillatory modes of low-frequency variability
(LFV).
In this talk we develop a new procedure for forecasting based on the ENSO-EMR model.
This procedure is based on two main ideas. The first idea is based on theoretical arguments
that show — subject to suitable technical assumptions on the stochastic process that generates
the noise — that, given a noise realization Ï (t) of length L, each "continuous snippet" of
length l of this realization, with l |
|
|
|
|
|