Hindcasts of Tropical Atlantic SST Gradient and South American Precipitation: The Influences of the ENSO Forcing and the Atlantic Preconditioning

Abstract

Hindcast experiments for the tropical Atlantic sea surface temperature (SST) gradient G1, defined as tropical North Atlantic SST anomaly minus tropical South Atlantic SST anomaly, are performed using an atmospheric general circulation model coupled to a mixed layer ocean over the Atlantic to quantify the contributions of the El Niño?Southern Oscillation (ENSO) forcing and the preconditioning in the Atlantic to G1 in boreal spring. The results confirm previous observational analyses that, in the years with a persistent ENSO SST anomaly from boreal winter to spring, the ENSO forcing plays a primary role in determining the tendency of G1 from winter to spring and the sign of G1 in late spring. In the hindcasts, the initial perturbations in Atlantic SST in boreal winter are found to generally persist beyond a season, leaving a secondary but nonnegligible contribution to the predicted Atlantic SST gradient in spring. For 1993/94, a neutral year with a large preexisting G1 in winter, the hindcast using the information of Atlantic preconditioning alone is found to reproduce the observed G1 in spring. The seasonal predictability in precipitation over South America is examined in the hindcast experiments. For the recent events that can be validated with high-quality observations, the hindcasts produced dryness in boreal spring 1983, wetness in spring 1996, and wetness in spring 1994 over northern Brazil that are qualitatively consistent with observations. An inclusion of the Atlantic preconditioning is found to help the prediction of South American rainfall in boreal spring. For the ENSO years, discrepancies remain between the hindcast and observed precipitation anomalies over northern and equatorial South America, an error that is partially attributed to the biased atmospheric response to ENSO forcing in the model. The hindcast of the 1993/94 neutral year does not suffer this error. It constitutes an intriguing example of useful seasonal forecast of G1 and South American rainfall anomalies without ENSO.