Air-Sea CO2 Fluxes in the Southern Ocean: Lessons Learned From the Comparison Between Models and SOCCOM Data

Although the Southern Ocean is recognized as a major contributor to the global ocean carbon sink, its intensity is still very much debated. So far, studies have shown that observational estimates and models generally agree on annually-averaged air-sea fluxes of carbon dioxide (CO₂), but strongly disagree on the seasonal phasing and amplitude of the fluxes. However, the poor observational sampling of the Southern Ocean and associated bias towards summer and spring lower the confidence in some of these comparisons. Here we revisit the performance of models in light of the new observational estimates of air-sea CO₂ fluxes made as part of the Southern Ocean Carbon and Climate Observations and Modeling (SOCCOM) project. SOCCOM estimates are year-round and provide a large spatial coverage of the Southern Ocean. These estimates were compared to air-sea CO₂ fluxes from eight Climate Model Intercomparison Project 5 (CMIP5) models, and from the Southern Ocean State Estimate (SOSE) and the high-resolution climate model CM2.6. Fluxes were integrated into six distinct provinces using a method that offers a simple procedure for detecting fronts from both modelled and observational fields. Of all the provinces, the Polar Antarctic Zone (PAZ) is where models and observations are found to disagree most on both the intensity and sign of the fluxes. In the PAZ, models could be sorted in three broad categories according to their representation of the seasonal cycle of the fluxes. None of these categories though accurately represents the strong observed outgassing of CO₂ in winter. In models whose seasonal cycle of fluxes is out-of-phase to that from observations, temperature is the main driver of fluxes instead of dissolved inorganic carbon. Causes for these discrepancies are related to the representation of the westerly winds and of the mixed layer seasonal cycle. Neither the high-resolution CM2.6 model nor the data-assimilating SOSE model show significant improvement in the representation of CO₂ fluxes compared to the CMIP5 models, suggesting some fundamental level of disagreement between models and observations.