Statistical prediction of ENSO (Nino 3) using sub-surface temperature data

A number of statistical schemes for predicting the evolution of the El Nino - Southern Oscillation (ENSO) have been developed in recent years. These tend to show some skill out to 9 to 12 months from late in the southern autumn, but only limited skill for a few months from late summer through the so-called predictability barrier''. More recently statistical models utilizing sub - surface temperature data have shown improvement of skill over persistence through this autumn period. Empirical Orthogonal Function analysis is used to extract the dominant signals in the sub-surface variability. The resulting statistical model shows similar skill to that obtained using other simple indices of sub-surface temperature, such as the warm water volume or average depth of the 20 degrees C isotherm, or from coupled ocean atmosphere models. These statistical models can therefore be used as a more stringent benchmark against which the complex coupled dynamical models are assessed.