Controls on present-day and future rainfall over southern Africa in coupled climate models

Global climate change will likely alter rainfall at a regional scale. This is a particular concern in southern Africa, where rainfall is central to socioeconomic wellbeing and where we do not understand the full complexity of the climate system. Climate models are the only tools available to estimate future rainfall change, but in the southern African summer rainfall season the latest generation of general circulation models disagree on the sign and magnitude of this change. Some models project small increases in rainfall (~20 mm.season-1) while other simulate large decreases (~100 mm.season^-1), and this diversity occurs in the context of large positive rainfall biases (up to 300%) in the present day. Regional policymakers are thus left with the task of designing strategies for climate change adaptation across a wide range of possible futures. With a view to assessing model reliability, this thesis addresses this problem by investigating the physical processes which underlie present day rainfall biases over southern Africa, and which explain the range in future rainfall projections. In tropical and eastern regions of southern Africa, present day intermodel variability in the strength of an important regional circulation feature, the Angola Low, accounts for 40 to 60% of intermodel rainfall variability. Meanwhile, in the subtropical region, intermodel variability in the strength of anomalous northeasterlies across the high topography of Tanzania and Malawi explains 72% of the intermodel spread in rainfall. These findings suggest that improving the representation of the Angola Low, topography and the northeasterly flow should be targets for model development. In the future projections, the model diversity in rainfall response is associated with differences in the large scale adjustment of the tropical atmosphere to warming. In models projecting the highest magnitude drying, the northern tropical oceans warm at a faster rate than the tropical mean (+0.5 K), and this is associated with enhanced subsidence and greater atmospheric stability over southern Africa. However, the most severe projections of drying (>60 mm.season^-1) only occur in models with large biases in the present day circulation, tropical sea-surface temperatures, topography and rainfall. The thesis therefore suggests that extreme drying is an unlikely scenario.