Southern Africa is characterized by a high degree of interannual climate variability, and human vulnerability. For mitigation and adaptation there is an urgent need for skillful seasonal forecasting of aspects of the climate relevant to the water sectors, industry, and subsistence and commercial farmers. The 2002/3 food crisis within southern Africa has highlighted this need.

An adaptation of the third Hadley Centre AGCM (HadAM3) has been implemented at the University of Cape Town in an experimental seasonal forecasting project. This paper presents the results from a 10 year (1990-1999) HadAM3 hindcast using the Reynolds sea surface temperature (SST) data set as a boundary forcing -- representing a seasonal forecast with perfect SST projections. The skill of HadAM3 in terms of simulating the summer season over the southern Africa region is evaluated, and aspects of the model pertinent to the problem of improving seasonal forecasts are discussed.

In an effort to reduce model biases in the region, HadAM3 has also been used with a mixed phase precipitation (MPP) scheme. The resulting changes in regional rainfall are assessed. Finally, examples of forecasts made using predicted SST anomalies with the MPP version of HadAM3 are assessed for the 2001/2 and 2002/3 summer seasons.