Modelling the dominant climate signals around southern Africa

Abstract

In this study, statistical techniques are employed to decompose climate signals around southern Africa into the dominant temporal frequencies, with the aim of modelling and predicting area-averaged rainfall. In the rainfall time series over the period 1900–1999, the annual cycle accounts for 83% of variance. Residual spectral energy cascades from biennial (42%) to interannual (20%) to decadal bands (3%). Regional climate signals are revealed through a multi-taper singular value decomposition analysis of sea surface temperature and sea level pressure fields over the Atlantic and Indian Oceans, in conjunction with southern Africa rainfall. Rossby wave action in the South Indian Ocean dominates the biennial scale variability. El Nino-Southern Oscillation (ENSO) and related Indian Ocean dipole patterns are important for interannual variability. Significant sea temperature and pressure fluctuations occurring 6–12 months prior to rainfall contribute biennial and interannual indices to a multi-variate model that demonstrates useful predictive skill.