Model Dependence and Internal Variability in Projected Change of Extreme Precipitation over South America

This study investigates the response of climate extremes to 1.5 °C and 2.0 °C of warming using the HAPPI "Half a degree Additional warming, Projections, Prognosis and Impacts" ensemble, with a focus on precipitation. Climate model experiments with identically prescribed sea surface temperature and sea ice concentration used in HAPPI permits a large number of realizations enabling precise statistical description of climate extremes. The HAPPI experimental design also allows an assessment of uncertainty in the climate extremes' response due to model dependence and internal variability. We analyse and interpret the relative importance of uncertainty due to model differences and internal variability using a two-way analysis of variance ANOVA framework. Our results show that over South America, models disagree on the exact location of changes in precipitation owing to differences in their present-day climatology and internal variability, but when extreme indices are aggregated across regions, model agreement increases. We demonstrate that from a spatial probability perspective, the evidence for heavier precipitation is robust.