THE IMPACT OF OCEAN INITIAL CONDITIONS ON ENSO FORECASTING WITH A COUPLED MODEL

A coupled atmosphere-ocean GCM (general circulation model) has been de veloped for climate predictions on seasonal to interannual timescales. The atmosphere model is a global spectral GCM T30L18 and the ocean mo del is global on a 1 degrees grid. Initial conditions for the atmosphe re were obtained from National Meteorological Center (now known as the National Centers for Environmental Prediction) analyses, while those for the ocean came from three ocean data assimilation (DA) systems. On e systems a four-dimensional DA scheme that uses conventional SST obse rvations and vertical temperature profiles inserted into the ocean mod el and is forced from winds from an operational analysis. The other tw o initialization schemes are based on tile coupled model, both nudging the surface temperature toward observed SSTs and one nudging surface winds from an operational analysis. All three systems were run from 19 79 to 1988, saving the state of the ocean every month, thus initial co nditions may be obtained for any month during this period. The ocean h eat content from the three systems was examined, and it was found that a strong lag correlation between Nino-3 SST anomalies and equatorial thermocline displacements crisis. This suggests that, based on subsurf ace temperature field only, eastern tropical Pacific SST changes are p ossibly predictable at lead times of a year or more. It is this ''memo ry'' that is the physical basis for ENSO predictions. Using the couple d GCM, 13-month forecasts were made for seven January and seven July e ases, focusing on ENSO (El Nino-Southern Oscillation) prediction. The forecasts, whose ocean initial conditions contained subsurface thermal data, were successful in predicting the two EI Nino and two La Nina e vents during the decade, whereas the forecasts that utilized ocean ini tial conditions from the coupled model that were nudged toward surface wind fields and SST only, failed to predict the events. Despite the c oupled model's poor simulation of the annual cycle in the tropical Pac ific, the ENSO forecasts from the full DA were remarkably good.