ENSO variability and prediction in a coupled ocean-atmosphere model

Abstract

We describe a numerically efficient coupled ocean – atmosphere model consisting of two dynamical levels in both the atmosphere and ocean components with simplified physics and coupling. The atmospheric domain is global, but the ocean domain approximates the Pacific Ocean basin. The model has a realistic climatology in both components, and ENSO-like variability is displayed in multi-decadal integrations. The model has been evaluated in hindcasts over the period 1981 to 2000. Model forecasts were initialised with initial conditions derived from an analysis run using a simple nudging technique to assimilate observed anomalies in the wind surface stress and upper level ocean temperatures. The model displays good skill, out to six or seven months, in forecasting the observed tropical upper oceanic temperatures and zonal current anomalies during the major El Niño and La Niña events which have occurred. Forecasts initiated between January and May tend to have the largest root mean square (RMS) errors and the most skilful forecasts occur for those initialised between June and November. Predictions of the atmospheric fields tend to have skill mainly during the first month but extending out to two or three months during major ENSO events. There is evidence of an annual cycle of error growth in the atmospheric forecasts. Overall, the model performs reasonably well during 1981 to 2000.