Understanding seasonal climate predictability in the Atlantic sector
This dissertation aims at understanding ocean-atmosphere interactions in the Atlantic basin, and how this coupling may lead to increased climate predictability on seasonal-to-interannual time scales. Two regions are studied: the South Atlantic convergence zone (SACZ), and the tropical Atlantic. We studied the SACZ during austral summer and separated its variability into forced and internal components. This was done by applying a signal-to-noise optimization procedure to an ensemble of integrations of the NCAR Community Climate Model (CCM3)forced with observed Sea Surface Temperature (SST). The analysis yielded two dominant responses: (1) a response to local Atlantic SST consisting of a dipole-like structure in precipitation close to the coast of South America; (2) a response to Pacific SST which manifests mainly in the upper-level circulation consisting of a northeastward shift of the SACZ during El Ni?o events. The land portion of the SACZ was found to be primarily dominated by internal variability, thereby having limited potential predictability at seasonal time scales. We studied two aspects of tropical Atlantic Variability (TAV). First, we investigated the effect of extratropical variability on the gradient mode. We found that the intensive Southern Hemisphere (SH) winter variability can play a pre-conditioning role in the onset of the interhemispheric anomalies in the deep tropics during boreal spring. This SH influence on TAV is contrasted with its northern counterpart that primarily comes from the North Atlantic Oscillation during boreal winter. Second, we explored the importance of ocean dynamics in the predictability of TAV. We used the CCM3 coupled to a slab ocean as a tier-one prediction system. The ocean processes are included as a statistical correction that parameterizes the heat transport due to anomalous linear ocean dynamics. The role of ocean dynamics was studied by comparing prediction runs with and without the correction. We showed that in the corrected region the corrected model outperforms the non-corrected one particularly at long lead times. Furthermore, when the model was initialized with global initial conditions, tropical Atlantic SST anomalies are skillfully predicted for lead times of up to six months. As result, the corrected model showed high skill in predicting rainfall in the ITCZ during boreal spring.