Quantifying the change in extreme seasonal precipitation events under global warming using a grand ensemble experiment

Estimates of future precipitation extremes are subject to much uncertainty. Uncertainties in emission rates, climate model structures, parameterisations and initial conditions add to the uncertainties in the prediction of extremes simply due to rarity of such events. Here, we produce probabilistic predictions of seasonal changes in extreme precipitation for regions across the globe using results from the climateprediction.net experiment. In this experiment, a coupled atmosphere-ocean global climate model was run in 'grand ensemble' mode for a transient integration from 1920 to 2080, varying parameter values, initial conditions and forcing scenarios. Here, we examine changes separately for each forcing scenario but examine the uncertainties introduced by different parameterisations and initial conditions. We use extreme value analysis to define extremes of precipitation. We fit the Generalized Extreme Value distribution to annual maxima using L-moments, to estimate extremes with return values of between 5 and 50 years for moving 30-yr time slices within the transient integration. We then apply the principle of equal weighting of the results from different models in the production of probability distributions of change for different regions of the globe. This allows us to establish which regions are most sensitive to the impacts of global warming on precipitation extremes, and the likely rates of change.