On the inaccuracy of CMIP6 models in capturing the observed long-term variability of the NAO

This study assesses how the CMIP6 simulations capture the non-stationarity of the main source of winter climate variability in the Euro-Atlantic region, the North Atlantic Oscillation (NAO), observed in the recent past.

For that purpose, we characterise the NAO long-term variability in climate reanalysis, analysing their features in several 30-year periods since 1851; and we evaluate whether CMIP6 historical simulations capture all the observed NAO “types”. Although the literature sometimes assumes that the NAO pattern is stationary, three groups of NAO pattern have been proved in the reanalyses depend on the location of their Action Centres (ACs): 1) the north AC locates over Iceland and the south AC in Azores, 2) the north AC locates over Southern Greenland and the south AC in the Western Mediterranean, and 3) the north AC locates over Northern Scandinavia and the south AC in the Azores.

Our main finding is that the NAO long-term variability is not accurately captured by all CMIP6 models. In particular, the overestimation of the NAO group 3 is remarkable in most simulations. This NAO group mainly represents the last decades, which the literature has addressed with much interest for its exceptional features (e.g. NAO+ strengthening and northeastward shift of its north AC), and which has been generally associated with the anthropogenic warmer climate. We also found underestimation of NAO group 2.

We have also found that each NAO group could be associated with precipitation anomalies in Europe. For example, the NAO group 3 implies drier(wet) conditions in the south(north). While group 2 implies the opposite pattern of anomalies. Therefore, we have reason to suggest that the lack of accuracy of models reproducing the non-stationarity of NAO may explain some of the bias in the expected changes of winter precipitation in Europe for future scenarios.