Abstract/Summary

Southern Africa and Southern South America have experienced recent extremes in dry and wet rainy seasons which have caused severe socio-economic damages. Selected past extreme events are here studied, to estimate how human activity has changed the risk of the occurrence of such events, by applying an event attribution approach (Stott et al., 2004)comprising global climate models of Coupled Model Intercomparison Project 5 (CMIP5). Our assessment shows that models' representation of mean precipitation variability over Southern South America is not adequate to make a robust attribution statement about seasonal rainfall extremes in this region. Over Southern Africa, we show that unusually dry austral summers as occurred during 2002/2003 have become more likely, whereas unusually wet austral summers like that of 1999/2000 have become less likely due to anthropogenic climate change. There is some tentative evidence that the risk of extreme high 5-day precipitation totals (as observed in 1999/2000) have increased in the region. These results are consistent with CMIP5 models projecting a general drying trend over SAF during December–January–February (DJF) but also an increase in atmospheric moisture availability to feed heavy rainfall events when they do occur. Bootstrapping the confidence intervals of the fraction of attributable risk has demonstrated estimates of attributable risk are very uncertain, if the events are very rare. The study highlights some of the challenges in making an event attribution study for precipitation using seasonal precipitation and extreme 5-day precipitation totals and considering natural drivers such as ENSO in coupled ocean–atmosphere models.