Reconstruction and prediction of the ocean carbon sink variations

The oceans take up about 1/4 of the anthropogenic carbon emissions, therefore, they largely mitigate the atmospheric CO₂ growth and the corresponding global warming. In the presence of an increasing trend following the emissions, the ocean carbon sink also shows prominent variations on interannual to decadal time-scales. It challenges the future prediction/projection of changes in the ocean carbon sink and the global carbon cycle. Previous studies on decadal predictions underline the advantages in predicting evolution of the climate such as the earth’s temperature and ocean meridional circulation by initializing the climate models with observations. We extend the climate model predictions to the Earth System by including the ocean biogeochemistry and carbon cycle processes.

We assimilate atmosphere and ocean reanalysis of physical data from the ECMWF into the decadal prediction system based on the Max Planck Institute Earth System Model (MPI-ESM). The reconstructed ocean carbon sink variations in the assimilation are consistent with the data-based estimates. Furthermore, the retrospective predictions starting from the assimilation states, which is close to the observations, show a predictive skill of the global ocean carbon sink up to 2 years. The regional predictability is higher up to 5 years. As the ocean carbon sink is mainly modulated by the variable ocean surface pCO₂, we further separate the ocean surface pCO₂into thermal and non-thermal drivers to investigate the mechanisms in maintaining the predictability of the ocean carbon sink. The thermal driver affected directly by the temperature maintains the shorter-term (of <3 years) predictability of the ocean carbon sink. The non-thermal driver, which is not directly affected by the temperature but by the ocean circulation and biology, maintains the longer-term (of >3 years) predictability of the ocean carbon sink.

Li, H., et al. (2019). Predicting the variable ocean carbon sink. Science Advances, 5: eaav6471.