Data supporting The eMLR(C*) method to determine decadal changes in the global ocean storage of anthropogenic CO2

Abstract

The determination of the decadal change in anthropogenic CO2 in the global ocean from repeat hydrographic surveys represents a formidable challenge, which we address here by introducing a seamless new method. This method builds on the extended multiple linear regression (eMLR) approach [Friis et al., 2005] to identify the anthropogenic CO2 signal, but in order to improve the robustness of this method, we fit C* [Gruber and Sarmiento, 2002] rather than dissolved inorganic carbon and use a probabilistic method for the selection of the predictors. In order to account for the multi-year nature of the surveys, we adjust all C* observations of a particular observing period to a common reference year by assuming a transient steady-state. We finally use the eMLR models together with global gridded climatological distributions of the predictors to map the estimated change in anthropogenic CO2 to the global ocean. Testing this method with synthetic data generated from a hindcast simulation with an ocean model reveals that the method is able to reconstruct the change in anthropogenic CO2 with only a small global bias (<5%). Within ocean basins, the errors can be larger, mostly driven by changes in ocean circulation. Overall, we conclude from the model that the method has an accuracy of retrieving the column integrated change in anthropogenic CO2 of about 10% at the scale of whole ocean basins. We expect that this uncertainty needs to be doubled to about 20% when the change in anthropogenic CO2 is reconstructed from observations.