Seasonal predictability of African summer rainfall

Canonical Correlation Analysis (CCA) has been used to predict African rainfall. The model uses global SSTs as a predictor field. The results show that CCA predicts African rainfall better during strong El Nino events than during la Nina events. Part of this is the relatively weak SST anomalies over the tropical Pacific during cold episodes as opposed to the stronger SST anomalies during warm events. CCA being a linear statistical model with equal weight at all grid points, the higher the SST anomalies over an ocean basin, the more likely this ocean basin will influence the forecasts. Hence when the magnitude of the SST anomalies are comparable across the global ocean, conflicting signals in the SSTs can weaken the predictions. In this paper, results from CCA experiments are presented and compared with hindcasts from AGCMs during ENSO.