Estimating change in extreme European precipitation using a multimodel ensemble

[1] Using the results from multimodel ensembles enables the assessment of model uncertainty in present and future estimates of extremes and the production of probabilities for regional or local-scale change. Six regional climate model (RCM) integrations from the PRUDENCE ensemble are used together with extreme value analysis to assess changes to precipitation extremes over Europe by 2070-2100 under the SRES A2 emissions scenario, investigating the contribution of the formulations of global (GCM) and regional climate models to scenario uncertainty. RCM ability to simulate precipitation extremes is evaluated for a UK case study. RCMs are shown to underestimate 1 day return values but reasonably simulate longer-duration (5 or 10 day) extremes. A multimodel approach by which probabilities can be produced for regional or local-scale change in extremes is then developed. A key result is that all RCMs project increases in the magnitude of short- and long-duration extreme precipitation for most of Europe. Individual model projections vary considerably but are independent of changes in mean precipitation. The magnitude of change is strongly influenced by the driving GCM but moderated by the RCM, which also influences spatial pattern. Therefore, when designing future ensemble experiments (1) the number of GCMs should at least equal the number of RCMs and (2) if spatial pattern is important then integrations from different RCMs should be incorporated. For impact studies, both the resolution and number of models in the ensemble will influence projections of change. The use of a multimodel approach therefore provides more robust estimates.