Why can't climate models capture the observed connection between seasonal snow cover and the Northern Annular Mode?

The suite of general circulation models (GCMs) in the Coupled Model Intercomparison Project (CMIP3) have been found not to reproduce the observed relationship between October Eurasian snow cover anomalies and the wintertime Northern Annular Mode (NAM). This apparent deficiency is reexamined here based on analysis of observational data and GCM simulations with prescribed snow forcing. Previous work has shown that in a comprehensive GCM in which an autumnal Siberian snow forcing is prescribed, a vertically propagating Rossby wave train is generated that propagates into the stratosphere, drives dynamical stratospheric warming and induces a negative NAM response that couples to the troposphere. The wave response and background climatological stationary wave must interfere constructively to achieve wave activity amplification into the stratosphere and the zonal mean stratosphere-troposphere NAM response. Using observational data, it is shown that constructive interference also occurs in the observed October Eurasian snow cover-NAM connection. This constructive interference peaks in December, corresponding to strong wave activity flux into the stratosphere two months after the snow cover anomalies in October. By contrast, the CMIP3 GCMs typically show a negative correlation between October Eurasian snow cover and December wave activity flux, which is related to destructive interference between the wave train associated with the snow and the background stationary wave. This work suggests that differences in the phasing of regionally forced waves can have a significant effect on stratospheric variability.