Impact of model resolution on the Euro-Atlantic weather regimes' representation

A multi-model set of atmosphere-only ensemble simulations forced by observed Sea Surface Temperatures (SSTs) is studied to assess the following scientific questions: o Is the model able to represent the observed weather regimes over the Euro-Atlantic region? o Does the representation improve increasing resolution and adding stochastic physics? How and why? A model's resolution describes the smallest scale at which physical processes are resolved, with any smaller scale processes being approximated in some way. Weather regimes could be understood as envelopes of daily atmospheric variability with typical persistence, spatially well defined and limited in number. Weather regimes over the Euro-Atlantic region are computed via k-means clustering to DJF Z500 daily data for observations and simulations in the same way. The k-means algorithm is applied to the first 4 Principal Components (PCs), which explain more than 50% of the variance. The representation of weather regimes is assessed comparing it with the observed one, by means of the following metrics: significance of cluster partition (how well the clusters are defined and separated), frequency of weather regime occurrence and pattern correlation relative to observations. Uncertainty in the metrics refers to the difference in the metrics between ERA-Interim and NCEP/NCAR reanalysis. Clustering analysis on the model data identifies the standard 4 regimes for the Euro-Atlantic region: The positive phase of the North Atlantic Oscillation (NAO+), the Scandinavian Blocking (BLCK), the Atlantic Ridge (AR) and the negative phase of the North Atlantic Oscillation (NAO-). In general, models underestimate NAO+ and BLCK frequency and overestimates AR and NAO- frequency. However, there is a big variability among the ensemble members: certain single members can have frequency closer to the observations with respect to the unique partition case, where all the ensemble members are considered together. Overall the improvement due to increased resolution is evident in all the three climate models considered in this study.