How Much Predictive Skill Is Contained in the Thermal Structure of an Oceanic GCM?

Abstract

The time history of upper-ocean temperatures in the tropical Pacific has been used as a predictor in a statistical prediction scheme to forecast SST anomalies in this region. The temperature variations were taken from the output of an oceanic general circulation model that was forced by observed winds for the period 1961 to 1985. Since such model data are presently used as initial conditions in prediction experiments with coupled ocean?atmosphere models, it is of particular interest to investigate up to what lead time tropical Pacific SST is predictable without the coupling of an atmosphere model to the ocean model. We compared our results with those obtained by the persistence forecast and with those obtained by using the wind stresses themselves as predictors in a statistical forecast model. It is shown that using the upper ocean temperatures from the ocean model forced by observed winds gives significantly better skills at lead times of 6 to 12 months compared to persistence and to the pure wind-stress model. Off-equatorial heat content anomalies at 5°N are shown to contribute significantly to the predictability at these lead times, while those at 12°N do not.