Dynamical Evolution of the Polar Annular Mode in CMIP5 Models

Wintertime intraseasonal variability in the troposphere and stratospheric polar vortex are linked via annular modes. The Northern Annular Mode (NAM), representing variability in the strength of the polar vortex and accounting for ~70% of the observed variability, is the dominant mode and is well studied. The secondary mode, accounting for ~15 % of observed variability, is the Polar Annular Mode (PAM) and represents variability in the north-south position of the polar vortex. Because the PAM has only recently been identified, it remains less well studied, primarily in the observational record. The ability of state-of-the-art climate models to accurately represent the NAM has been thoroughly examined. However, the simulation of PAM in models has received little focus. In particular, studies of PAM in the models have focused only on the ability of models to accurately reproduce the gross structure and relative percentage of stratospheric variability explained. To fully assess the simulated polar vortex variability, both the gross features and the underlying dynamics of the simulations must be compared with the observations. To assess the veracity of PAM dynamics in current models, a suite of runs from different members of the CMIP5 experiment are analyzed. For each model, discrete large-amplitude positive and negative events occurring in the absence of NAM events are identified. Using composite analyses, composite time evolutions are constructed for positive and negative events. Using these composites, the zonal-mean dynamics, structural evolution, and tropospheric impact are assessed. Key differences in tropospheric wave activity both before and during the event for positive and negative events are highlighted. Models are also binned into 'high-top' and 'low-top' categories based on the vertical resolution of the model stratosphere. High-top and low-top averages are calculated to test for biases in the ability to accurately simulate observed stratospheric variability among models with different vertical resolutions.