Integrating Ocean Subsurface Temperatures in Statistical ENSO Forecasts

Abstract

Subsurface characteristics of oceans have recently become of interest to climate modelers. Here subsurface information has been linked to the evolution of the El Niño?Southern Oscillation (ENSO) in a simple statistical formulation. The hypothesis proposed is that the inclusion of subsurface ocean heat content in a persistence-based representation of ENSO results in an increase in prediction skill. The subsurface temperature field is represented by anomalies in the 20°C isotherm (Z20) in the Indian and Pacific Oceans. Using a cross-validation approach, the first two empirical orthogonal functions (EOFs) of the Z20 anomalies are derived, but only the second EOF is used as a predictor. The first EOF is found to be representative of the mature ENSO signal while the second EOF shows characteristics that are precursory to an ENSO event. When included in a persistence-based prediction scheme, the second EOF enhances the skill of ENSO hindcasts up to a lead time of 15 months. Results are compared with another model that uses the second EOF of the SST anomalies in the tropical Pacific Ocean and persistence as predictors. Cross-validated hindcasts from the isotherm-based scheme are generally more skillful than those obtained from the persistence and SST-based prediction schemes. Hindcasts of cold events are particularly close to the observed values even at long lags. Major improvements occur for predictions made during boreal winter and spring months when the addition of subsurface information resulted in predictions that are not greatly affected by the damping effect of the ?spring barrier.?