The Southern Ocean forms a key component of the global carbon budget, taking up about 1.0 Pg C yr−1 of anthropogenic CO2 emitted annually (∼10.7 ± 0.5 Pg C yr−1 for 2012). However, despite its importance, it still remains undersampled with respect to surface ocean carbon flux variability, resulting in weak con- straints for ocean carbon and carbon – climate models. As a result, atmospheric inversion and coupled physics-biogeochemical ocean models still play a central role in constraining the air-sea CO2 fluxes in the Southern Ocean. A recent synthesis study (Lenton et al., 2013a), however, showed that although ocean biogeochemical models (OBGMs) agree on the mean annual flux of CO2 in the Southern Ocean, they dis- agree on both amplitude and phasing of the seasonal cycle and compare poorly to observations. In this study, we develop and present a methodological framework to diagnose the controls on the seasonal variability of sea-air CO2 fluxes in model outputs relative to observations. We test this framework by comparing the NEMO-PISCES ocean model ORCA2-LIM2-PISCES to the Takahashi 2009 (T09) CO2 dataset. Here we demonstrate that the seasonal cycle anomaly for CO2 fluxes in ORCA2LP is linked to an underes- timation of winter convective CO2 entrainment as well as the impact of biological CO2 uptake during the spring-summer season, relative to T09 observations. This resulted in sea surface temperature (SST) be- coming the dominant driver of seasonal scale of the partial pressure of CO2 (pCO2) variability and hence of the differences in the seasonality of CO2 sea-air flux between the model and observations.