Making sense of uncertainty: insights into temporal variability and drivers of change in southern African rainfall

As with many regions in the world, southern Africa strongly relies on seasonal rains for livelihoods, ecosystems, food and water security, therefore understanding rainfall changes is vital for adaptation planning and climate service practitioners. Future changes in rainfall over southern Africa are highly uncertain, with different global climate models projecting wetter, drier and shifted rainfall seasons. Taking an ensemble average suggests poor model agreement on the direction of change and therefore a low mean change that is not statistically robust. Ensemble averages also blur boundaries between different sources of uncertainty and obscure significant aspects of rainfall over this region that we have a better (or at least reasonable) understanding and confidence about. Analysis focussing on large ensemble averages often does not account for the significant internal variability in the climate, which we find to be very large for this region.

We will show results from a recently published study (Kennedy-Asser et al., Climatic Change, 2026), using over 200 different global climate model simulations from CMIP5, CMIP6 and UKCP18 (global runs used for UK Climate Projections), highlighting how important internal variability has been in the past and continues to be in the future for southern Africa. All models are imperfect, each have strengths, weaknesses and significant biases that prevent us from fully constraining potential futures. By reframing analysis around temporal variability, we show there is better model agreement on scenarios of low change, where the future remains within the range of historic variability, than there is on significant change towards wetting or drying in future.

In addition, by analysing models individually, it is possible to construct Climate Process Chains and Climate Storylines that explain the mechanisms behind simulated responses and plausibly justify the divergent model signals. We will present results from an in-depth analysis of outputs from multiple ensemble members across four CMIP6 models that show contrasting futures for this region (CanESM5, CNRM-ESM2-1, HadGEM3-GC31-MM, IPSL-CM6A-LR). We explore linkages between regional rainfall and Indian Ocean sea surface temperatures, pressure systems, ENSO teleconnections and changes in the Angola Low, and demonstrate how these changes could result in wetting, drying or delayed rainfall seasons. Framing the analysis in this way highlights some important climate states and drivers that may be indicative of future change in seasonal rainfall in one direction or another.

This research is part of an interdisciplinary project, SALIENT (https://www.climatebristol.org/projects/salient/), combining climate science and modelling findings presented here with insights from risk communication research and structured expert judgement. Novel insights on the interdisciplinary research process and of policy relevance will also be presented.