This study assesses the performance of an atmospheric GCM forced with persisted SSTs in simulating austral summer precipitation at smaller spatial (regional) scales. Two statistical recalibration techniques of differing technical complexity are then presented and compared to get an idea as to which method among them is best suitable for southern Africa. The two regression-based methods applied in recalibrating the ECHAM4.5 GCM output during austral summer in southern Africa are based on model output statistics (MOS) using principal components regression (PCR) and canonical correlation analysis (CCA) to statistically link archived records of the GCM to regional rainfall over much of Africa south of the equator. A linear statistical model linking near-global sea-surface temperatures (SSTs) to regional rainfall is also developed. Southern Africa is divided into 18 homogeneous regions using cluster analysis. The potential predictive skill of summer precipitation over each region from raw-GCM ensembles, the linear statistical and MOS models is evaluated using the relative operating characteristics (ROC) score and the ranked probability skill score computed over a 12-year retroactive period 1989/90–2000/01. The MOS technique outperforms the raw GCM ensembles and the linear statistical model in certain cases. On many occasions, the PCR-MOS performs better than CCA-MOS but the former does not show clear superiority over the latter method because the two methods are in a broad sense performing the same task. The need to recalibrate GCM predictions at regional scales to improve their skill at smaller spatial scales is demonstrated in this study.
Dissertation (MSc (Meteorology))--University of Pretoria, 2007.