From Global to Regional Climate Models: Consistency assessment of the hydrological cycle

The German research projects NUKLEUS (Actionable local climate information for Germany)  and UDAG (Updating the data basis for adaptation to climate change in Germany) in collaboration with the CLM Community recently generated new RCM ensembles from downscaling of CMIP6 GCMs for EURO-CORDEX (at 12km) and additionally on the convection-permitting scale (3km) for Central Europe. The NUKLEUS ensemble consists of three GCMs from the EURO-CORDEX Balanced Ensemble Matrix (BEM) initiative, each dynamically downscaled by three RCM (ICON-CLM, COSMO CLM and REMO) for the ssp370 scenario. The UDAG ensemble encompasses a downscaling with just ICON-CLM, but for all six BEM GCMs and multiple ssp-scenarios. 

In this contribution, we evaluate the NUKLEUS ensemble to assess the propagation of biases and the consistency of climate change signals along different GCM-RCM models chain. Biases are examined through analyses of individual components of the hydrological cycle, which plays a central role in land–atmosphere flux interactions and whose realistic representation is crucial for reliable climate projections, alongside near-surface air temperature and sensible heat fluxes. ERA5 is used as a reference to explore model behavior over both land and ocean. The analysis is conducted on the EUR-12 grid over Europe in winter and summer, with detailed focus on Central Europe and the Mediterranean.

Regarding biases, results show partial consistency between GCMs and RCMs. RCMs generally provide a more consistent representation of the hydrological cycle than GCMs. However, given large deviations of GCM forcing data from observations, e.g., sea surface temperature, can cause RCMs to adjust surface fluxes, either reducing inherited biases or generating new ones. These adaptations occur regionally, especially in MIROC6-driven simulations over the Mediterranean.

Regarding the climate change signal, the biases affect the representation of the hydrological cycle, with the strongest impact in the Mediterranean. Here, MIROC6-driven simulations show an intensification of the hydrological cycle particularly pronounced in future projections. Despite such model-specific issues, RCMs generally produce narrower climate change signals than their driving GCMs. 

These results highlight that the optimal GCM selection for downscaling is region-specific and that GCM–RCM model chains should be analysed with caution in ocean-influenced areas.