Robust spatially aggregated projections of climate extremes

Many climatic extremes are changing1–5, and decision-makers express a strong need for reliable information on further changes over the coming decades as a basis for adaptation strategies. Here, we demonstrate that for extremes stakeholders will have to deal with large irreducible uncertainties on local to regional scales as a result of internal variability, even if climate models improve rapidly. A multimember initial condition ensemble carried out with an Earth system model shows that trends towards more intense hot and less intense cold extremes may be masked or even reversed locally for the coming three to five decades even if greenhouse gas emissions rapidly increase. Likewise, despite a long-term trend towards more intense precipitation and longer dry spells, multidecadal trends of opposite sign cannot be excluded over many land points. However, extremes may dramatically change at a rate much larger than anticipated from the long-term signal. Despite these large irreducible uncertainties on the local scale, projections are remarkably consistent from an aggregated spatial probability perspective. Models agree that within only three decades about half of the land fraction will see significantly more intense hot extremes. We show that even in the short term the land fraction experiencing more intense precipitation events is larger than expected from internal variability. The proposed perspective yields valuable information for decision-makers and stakeholders at the international level. Significant changes to more hot and less cold extremes and record events have been observed over several regions1–3 and identified in globally aggregated approaches4,5. Attribution studies argue that anthropogenic influence has enhanced the probability of the occurrence of some types of temperature and precipitation extremes occurring6–11. Recent projections suggest that these trends continue along with rising anthropogenic greenhouse gas emissions12–16. Model simulations from the Coupled Model Intercomparison Project Phase 5 (CMIP5) project pronounced warming of the annual temperature maxima (TXx, hereafter referred to as hot extremes, see Methods) and minima (TNn, cold extremes) as well as widespread changes in maximum five-day accumulated precipitation (RX5day, heavy precipitation intensity) and annual maxima of consecutive number of dry days (CDD, dry spell length)17. The changes by mid-century (2041–2060) shown in Fig. 1 (left) based on an extended set of 25 CMIP5 models (see Supplementary Information) are largely consistent with those projected for the end of the twenty-first century in refs 12,13. Projections for all four extreme indices are associated with very large uncertainties12. At many grid points individual models simulate warming of hot and cold extremes that is almost twice the multimodel mean (Supplementary Fig. 1, right) and others show hardly any change or even a slight cooling at some grid points despite the strong global warming in