A1044
Title: The predictability of sea surface temperatures in El Nino regions
Authors: Tommaso Proietti - University of Roma Tor Vergata (Italy) [presenting]
Alessandro Giovannelli - University of Tor Vergata - School of Economics (Italy)
Abstract: The El Nino Southern Oscillation induces the alternation of persistent warming and cooling of sea surface temperatures along the equator in the east-central Pacific. We investigate the predictability and the persistence of the El Nino cycle by considering sea surface temperature anomalies averaged over four rectangular areas that subdivide the equatorial Pacific (Nino 1, 2, 3-4, and 4 regions). Predictability is defined both in terms of the prediction error variance and by the mutual information between past and future. By means of the decomposition of the mutual information, we can assess the information gains arising from modelling the time series jointly. Persistence measures the strength of the auto- and cross-covariances by comparing the long-run variance to the unconditional variance of the series. Our methodology is based on regularized univariate and multivariate Levinson-type algorithms for estimating the autocovariance and the cross-covariance matrix of the series.
