This study contributes to a broader effort of institutions toward improving seasonal forecasts over southern Africa. The primary objective is to understand where global models show shortcomings in their simulations, and how this impacts on their seasonal forecast skill. It is proposed that the skill of a model in simulating natural climate variability is an appropriate metric for a model's potential skill in seasonal forecasting. Thus the study investigates the performance of two global models in simulating the regional processes in relation to the processes variability, and how this is related to their forecast skill.