Clearly we need more data to constrain global circulation models if we want to reduce the uncertainty in OAE predictions. We hope that this work will inspire, direct and inform future data gathering and experiments.

Finally we tried to simply turning off all the sophisticated biology modeling in ECCO-Darwin (i.e. the Darwin part). It turned out that this made virtually no difference to the OAE uptake trajectory - OAE-driven CO2 uptake appears to be primarily a function of bulk transport and gas exchange.

We also looked at the difference attributable to the divergent prediction of the horizontal plume trajectory (even if the gas exchange parametrization had been equal) and find that plume trajectory is significant. This is especially true of plumes end up going under ice or towards a high-wind zone.

We tried to figure out what other aspects of each model are responsible for the observed differences (other than subduction). The first suspect is the parameterization of wind speed and carbonate chemistry. The difference in wind params (and therefore gas exchange velocity) was the major influence.

It's interesting too that the models disagreed more for coastal injection sites but agreed better for sites further off-shore. This might have to do with the complexity of the near-shore water movements.

The inter-model variation is greater than the inter-annual variation within a model.

Unsurprisingly there is a strong correlation with the rate of subduction experienced by the alkalinity plume. We find this is the primary driver of the differences. If the model predicts faster subduction, the OAE efficiency suffers more.