An assessment of the skill of downscaled GCM outputs in simulating historical patterns of rainfall variability in South Africa

Abstract

Uncertainties associated with General Circulation Models (GCMs) and the downscaling methods used for regional or local scale hydrological modelling can result in substantial differences in estimates of future water resources availability. This paper assesses the skill of nine statistically downscaled GCMs in reproducing historical climate for 15 catchments in five regions of South Africa. The identification of skilled GCMs may reduce the uncertainty in future predictions and the focus is on rainfall skill as the GCMs show very similar patterns of change in temperature. The skill tests were designed to assess whether the GCMs are able to realistically reproduce precipitation distribution statistics and patterns of seasonality, persistence and extremes. Some models are consistently less skilful for the regions assessed, while some are generally more skilful with some regionally specific exceptions. There are differences in the GCMs skill across the different regions and in the skill ranking between coastal areas and inland regions. However, only a limited reduction in uncertainty is achieved when using only the downscaled GCM outputs identified as being skilled in a hydrological model for one of the regions. Further modelling studies are required to determine the general applicability of this observation.