The dynamical processes that contribute to the seasonal prediction of the tropical Atlantic sea-surface temperature (SST) anomalies from boreal winter into spring are explored with an atmospheric general circulation model coupled to a slab ocean. Taking advantage of the reduced-physics model that effectively isolates thermodynamic feedbacks from dynamic feedbacks, we examine the joint effect of local thermodynamic feedback and the remote influence of El Nino-Southern Oscillation (ENSO) on the prediction of SST anomalies by conducting large ensembles of prediction runs. These prediction experiments yield the following findings: (1) in the northwestern part of the tropical Atlantic, the positive feedback between the surface heat flux and SST can play an important role in enhancing the predictability of the SST; (2) the remote influence from Pacific ENSO can enhance the SST predictability through a constructive interference with the local thermodynamic feedback, but can also make the SST prediction more difficult when the interference is destructive; (3) ocean dynamics plays a fundamental role for prediction of SST anomalies in the equatorial and south tropical Atlantic. To shed further light on the importance of the ocean dynamics, a statistical procedure of parameterizing the important ocean dynamics is developed within a linear dynamical framework. Prediction experiments with the parameterized ocean dynamics included in the simple coupled model result in an improved forecast skill in predicting the cross-equatorial SST gradient, which subsequently lead to a high skill of the model in predicting seasonal rainfall anomalies associated with variations in the Intertropical Convergence Zone during boreal spring. A diagnostic study suggests that the vertical advection of heat due to anomalous Ekman pumping/suction is a dominant contributing factor for causing equatorial SST anomalies, thereby, a major element of predictable dynamics in this region. (c) 2004 Elsevier B.V. All rights reserved.
930BWTimes Cited:3Cited References Count:33