Assessing the physical processes controlling oxygen subduction in CMIP6 models

Approximately half of the total loss of upper ocean oxygen over the recent past has been driven by its reduced solubility in a warming ocean. The remainder can be explained by less well constrained changes to ocean circulation, mixing and biogeochemical processes. Model based studies have shown that the choice of mixing parameterisations as well as biogeochemical factors, can introduce substantial differences in the distribution of oxygen within an individual Earth System Model (ESM).  Model Intercomparison Projects such as the latest Coupled Model Intercomparison Project Phase 6 provide an opportunity to explore processes controlling oceanic oxygen in a multi-model framework. Here we apply an oxygen transport decomposition to seven CMIP6 ESMs, using a well-established framework for the transport of tracers from the surface mixed layer into the ocean interior. We show that despite a close agreement in the oxygen concentration at the mixed-layer base, the transports to the ocean interior vary greatly between models. ESMs with similar physical ocean model components are clearly identifiable based on the spatial distribution of oxygen transport, both in the globally integrated transport terms and their inter-annual variability. Applying this decomposition to CMIP6 pre-industrial control experiments, we find the total oxygen subduction ranges between +0.6 to +1.1PMol yr-1, in agreement with an observationally based estimate. Despite broad agreement in the total magnitude of oxygen subduction, the inter-model range for individual transport terms is often large (+0.69 PMol yr-1 to -0.23 PMol yr-1 for vertical advection), implying a high degree of model uncertainty as to the physical processes controlling interior oxygen. We also characterise variability in oxygen transport terms and find that interannual variability in advective transport depends on the term and the model family. Lateral advection displays the greatest model-model difference in interannual variability, by a factor of ~6 between the most and least variable model. Mixed-layer entrainment of oxygen shows closer agreement between models, with interannual variability in this term differing by a factor of ~1.4. We recommend that future model intercomparisons including ocean biogeochemistry archive the relevant transport, production and consumption terms for key biogeochemical variables such as oxygen.