3rd EURO-CORDEX workshop on GCM evaluation and selection, June 28, 2021

At this workshop the efforts of the Task Team on CMIP6 GCM selection and ensemble design were presented. Based on existing literature and some in-house analysis GCMs are evaluated for their availability, plausibility, future change and independence. The outcomes are then displayed in tabular format so that modelling teams may make a well informed choice with respect to which GCMs to downscale. A number of community level decisions were taken and the Task Team will take these into account as they work towards establishing a list of recommended GCMs for EURO-CORDEX.

The community took the following decisions:

  • Optimum number of models: 8-10. N.B., this is a total number of GCMs, not the number of GCMs each group should downscale
  • Prioritization of SSPs: First priority are SSP1-2.6 and SSP3-7.0; Second priority are SSP2-4.5 and SSP5-8.5, following recommendations of the global CORDEX community
  • Do we include a criteria on observational constraints: Yes(overwhelmingly so)
  • Do we include one or two models that fall outside the observationally constrained range but are otherwise well performing (e.g. for worst case risk assessment): Yes (overwhelmingly so)
  • Do we include models with multiple realizations to explore internal variability: Mixed. The feeling is that this is important but maybe not a first priority due to the resources required. These are also opportunities to employ hybrid dynamical-statistical downscaling approaches (e.g. emulators) for this issue. The community settled on a decision to make sure we keep a few models with multiple realizations (EC-Earth, IPSL, MPI)