A prominent warm bias in surface temperature over the mid- and high-latitudes of the Southern hemisphere (SH) has been identified in many climate models. Similar biases are also pronounced in the DOE’s Energy Exascale Earth System Model (E3SM) versions 1 and 2. In this study, we evaluated an ensemble of atmosphere-only (AMIP) and coupled historical simulations during the period of 1979-2014 conducted with E3SM version 2. The long-term mean surface air temperature in the coupled E3SM simulations reveals large warm biases over the entire Southern Ocean and Antarctic region (locally exceeding 5°C), especially in the boreal winter season (December-January-February). Most of the biases arise from the ocean component and coupled atmosphere-ocean processes in E3SM as the biases in the same region are reduced by more than a factor of two in the AMIP simulations with prescribed observational SST. To understand the climatological patterns of surface temperature biases in the coupled simulations, atmospheric circulation anomalies associated with climate variability over SH in the boreal winter season including the El Ninõ-Southern Oscillation (ENSO), the low-level westerly jet, and Southern Annular Mode (SAM) are analyzed. The coupled E3SM simulations generally capture the dominant feature of atmospheric anomaly patterns reasonably well. However, the coupled E3SM produces an overly persistent SAM, an erroneous equatorward shift of the westerly jet, and erroneously strong covariations between ENSO, SAM, and global mean surface temperature. The associated deficiencies in the atmospheric circulation anomalies affect the simulated cloud processes and cloud-radiative effect, leading to biases in the surface temperature. Detailed analyses of the links between surface temperature biases and biases in atmospheric circulation anomalies will be presented at the conference.