Nordic precipitation trends and North Atlantic circulation patterns in CMIP6 models

Recent observed trends indicate increasing future precipitation in Northern Europe, yet CMIP6 models show substantial spread in sign and magnitude of both historical and future precipitation trends in the region. This large model spread highlights the difficulty in accurately modelling regional precipitation in global climate models, and the low confidence in future precipitation trends. In this study, we evaluate recent historical winter precipitation trends in selected models of the CMIP6 ensemble by validating the models against reanalysis and observations, and examining how the trends are influenced by sea surface temperatures (SSTs) and the large-scale atmospheric circulation, particularly the North Atlantic storm track. We consider data from CMIP6 historical and AMIP experiments, the latter using prescribed observed SST and sea ice concentration. Additionally, we include a complimentary set of AMIP-style experiments with the Norwegian Earth System Model (NorESM2) where we explore the effect of different time-evolving SST fields. 

Results show that models yield consistent precipitation patterns, which enables an assessment of the role of North Atlantic SST in shaping Nordic precipitation trends, when averaging over multiple ensemble members. Differences in Nordic precipitation trends are tied to differences in the North Atlantic SST pattern evolution, with consistent changes in latent heat flux, atmospheric baroclinicity and lower tropospheric zonal wind patterns. Particularly, differences in SST in the North Atlantic warming hole and along the eastern US coast and close to Svalbard, the latter related to rapid warming from Arctic sea ice loss, can influence the precipitation trends. Employing the relationship between North Atlantic SST and precipitation trends in the Nordic region, we can better understand the large precipitation spread among climate models, and thus increase confidence in both historical and future precipitation in the models. Furthermore, while the majority of CMIP6 models provide just a single ensemble member for the AMIP experiment, our results demonstrate that using multiple ensemble members for the AMIP experiments is essential to account for internal variability and to achieve robust results.