Are Simulated Ocean Deoxygenation Rates Consistent with the Observational Reconstructions?

It is currently debated whether Earth system models (ESMs) can reproduce observation-based long-term changes in global and regional deoxygenation rates. Both models and observations include uncertainties, which must be considered when evaluating their consistency. Based on 14 ESMs and 6 observational datasets, the models’ climatological annual mean oxygen matches observations well near the surface. However, significant biases remain in the tropics and in the thermocline. Based on the same set of models and three time-varying observation-based datasets, the models tend to underestimate deoxygenation trends from 1965 to 2014, except for the North Atlantic basin. However, the small number of observational datasets limits this conclusion. One dataset appears to significantly underestimate deoxygenation due to sparse data coverage. This review highlights the need for improvements in model process representations and the development of more observation-based, quality-controlled datasets to better constrain and interpret oxygen changes in the ocean. ▪ Uncertainties of dissolved oxygen field in CMIP6 ESMs and observational reconstructions are quantified. ▪ The ESMs can skillfully reproduce long-term average oxygen near the surface, but challenges remain in the thermocline and tropics. ▪ The ESMs underestimate the deoxygenation trends except for the North Atlantic basin.