Investigating Labrador Sea's persistent surface O2 anomaly using observations and biogeochemical model results

Deviations of surface ocean dissolved oxygen (O₂) from equilibrium with the atmosphere should be rectified about twenty times more quickly than deviations of dissolved carbon dioxide (CO₂). Therefore, persistent O₂ disequilibria in the Labrador Sea, while CO₂ is close to equilibrium, has been a matter of interest to many previous works. Here we investigate this phenomenon by using a novel analytical technique, the ‘CORS (Carbon Dioxide and Oxygen Relative to Saturation) method’, and also by using more data than was available previously. We compare observations to results from a model we developed for the Labrador Sea which combines plankton ecology with biogeochemical cycling of oxygen, carbon and nitrogen. In contrast to earlier works which mostly considered individual factors in isolation, here we used the model, together with data, to distinguish between the varying influences of several processes potentially contributing to the long-lasting O₂ undersaturation: mixed layer depth, duration of mixed layer deepening, convection, entrainment and bottom water O₂ content. Our model experiments confirm that, for the same gas exchange rate, the effects on surface O₂ concentration differ significantly among the identified drivers. Our results suggest that prolonged surface O₂ undersaturation is not always dependent on the extreme winter mixed layer depths, but rather that even moderately deep mixed layers (e.g. 300 m), when prolonged and in conjunction with continuous entrainment of oxygen-depleted deep water, can also drive persistent surface O₂ anomalies. An implication of our results is that regions in the North Atlantic with maximum winter mixed layer depths of only a few hundred metres should also show persistent surface O₂ undersaturation. We further reveal that convection in deep water formation regions produces trendlines that do not pass through the origin of a plot of CO₂ vs. O₂ deviations which have previously been thought to indicate erroneous data.