Evaluation of the capability of global and regional climate models to reproduce the temporal clustering in heavy precipitation across Europe

Evaluating the capability of global climate models (GCMs) and regional climate models (RCMs) to reproduce historical records represents a way to build confidence in their applicability for future projections. Here we focus on their capability in reproducing the temporal clustering of heavy precipitation events, which manifest itself through the alternating of quiet and active periods (in terms of number of events). More specifically, we investigate the connection between the occurrence of heavy precipitation events across Europe and four climate modes [the Arctic Oscillation (AO), the North Atlantic Oscillation (NAO), the East Atlantic (EA) pattern, and the Scandinavia pattern (SCAND)]. We use outputs from GCMs part of the Coupled Model Intercomparison Project Phase 5 (CMIP5) and the large ensemble Community Earth System Model (CESM), and from RCMs part of the EURO-CORDEX with two resolutions (0.11° and 0.44°). We use a peak over threshold (POT) method to identify heavy precipitation events, and Cox regression to relate the occurrence of the events to the climate modes. We find that the GCMs and RCMs can capture the temporal clustering in heavy precipitation over Europe in terms of these four climate modes. GCMs can better reproduce the relationship between heavy precipitation and AO/SCAND than NAO/EA; moreover, intra-model uncertainties are smaller than the inter-model ones in most of the cases. RCMs well reproduce the observed relationship for the four climate modes; we also find that RCMs with higher spatial resolution do not necessarily have a better performance compared to the lower resolution ones.