An assessment of CMIP6 climate signals and biases in temperature, precipitation and soil moisture over Europe

The CMIP6 projections constitute the basis of our latest understanding of the climate response to anthropogenic forcing. However, there is still considerable uncertainty in the projections, especially at the regional scale. One way to constrain the uncertainty is by comparing the models historical climate change signals against observations and investigate the physical reasons for divergences. Here, we assess the signal-to-noise ratio (S/N) of surface air temperature (SAT), precipitation (PREC) and soil moisture (SM) over Europe for a set of CMIP6 historical simulations and compare them against the E-OBS observational product and the ERA5 reanalysis. We found considerable divergences between the CMIP6 ensemble mean S/N and that of E-OBS and ERA5, as well as between ERA5 and E-OBS. The latter indicates that the S/N is affected by data coverage. We show that the differences among model signals are associated with different atmospheric circulation responses. We also investigate the potential relationships between the models' signals and climatological biases, and we found evidence that the models with a warm climatological bias in southern Europe tend to have smaller SAT signals (warm less). Finally, we found no apparent relationship between SM biases and the warming signal, suggesting that the mechanism by which SM-atmosphere interactions affect climate variability does not explain the mean changes. However, there is a tendency for models with higher SM to dry faster than models with lower SM.

Automated summary

The CMIP6 projections are the foundation of our latest understanding of the climate response to human-induced forcing, but there is still significant uncertainty, especially at the regional level. To constrain this uncertainty, we compared historical climate change signals from CMIP6 models with observations, and found divergences between the models and the observations, as well as among different observations. We also identified that these differences are associated with atmospheric circulation responses and revealed potential relationships between model signals and climatological biases.