How does large-scale nudging in a regional climate model contribute to improving the simulation of weather regimes and seasonal extremes over North America?

To determine the extent to which regional climate models (RCMs) preserve the large-scale atmospheric circulation of their driving fields, we investigate the ability of two RCM simulations to reproduce weather regimes over North America. Each RCM simulation is driven at its lateral boundaries by the ERA-Interim reanalysis, but one of them uses additional large-scale nudging (LSN) in the domain interior. Four weather regimes describing the variability of the large-scale atmospheric dynamics over North America are identified in winter and in summer. The analysis shows that for both seasons, the mean frequency of occurrence and persistence of the four weather regimes for the two RCM simulations are comparable to those of ERA-Interim. However, the frequency of false daily attributions of a specific regime on day-to-day and seasonal bases is significantly high, especially in summer, for the classic lateral-boundary driven simulation. Those false attributions are largely corrected with LSN. Using composite means for each weather regimes, substantial 2-m air temperature and precipitation anomalies associated to the large-scale atmospheric circulation are found. These anomalies are larger in winter than in summer. The validation of the simulations reveals that the 2-m air temperature bias is dependent on the weather regimes, especially in summer. Conversely, the precipitation bias varies significantly from one regime to another, especially in winter. Overall, the results suggest that a classic RCM simulates the mean statistics of the weather regimes well, but that LSN is necessary to reproduce daily weather regimes and seasonal anomalies that match with the driving field.