Abstract
This study evaluates simulated Antarctic sea ice edge (SIE) variability and its teleconnections in three global coupled climate models (GISS, NCAR and GFDL) against the observations. All models do a reasonable job in simulating the seasonal advance and retreat of the Antarctic sea ice fields. The simulated GISS and NCAR SIE distributions are in agreement with observations in summer and autumn, whereas the GFDL model does best in spring and winter. A common problem is the poor simulation of the observed SIE in the Weddell Sea. All models are not particularly good at simulating the observed regionally varying SIE trends. A comparison of dominant empirical orthogonal function modes of surface air temperature (SAT) variability in each model associated with observed modes show that the models generally capture features of the more prominent covarying spatial patterns such as an El Niño–southern oscillation (ENSO)‐like pattern in the tropical Pacific. The simulated teleconnection patterns between detrended Antarctic SIE anomalies and detrended global SAT anomalies in each model are evaluated for comparison with observed teleconnection patterns. All models capture the ENSO‐like phenomenon to some degree. Also, the GISS and NCAR models capture the Antarctic dipole pattern and meridional banding structure through the Pacific. The Antarctic SIE regions showing the strongest extrapolar teleconnections differ among the models and between the models and observations. Almost all models miss the observed polar–extrapolar teleconnections in the central Indian, western extreme of the tropical and southern Pacific, and over the tropical continents. Copyright © 2002 Royal Meteorological Society.