Assessing improvements in global ocean <i>p</i>CO₂ machine learning reconstructions with Southern Ocean autonomous sampling

Abstract

The Southern Ocean plays an important role in the exchange of carbon between the atmosphere and oceans and is a critical region for the ocean uptake of anthropogenic CO2. However, estimates of the Southern Ocean air–sea CO2 flux are highly uncertain due to limited data coverage. Increased sampling in winter and across meridional gradients in the Southern Ocean may improve machine learning (ML) reconstructions of global surface ocean pCO2. Here, we use a large ensemble test bed (LET) of Earth system models and the “pCO2-Residual” reconstruction method to assess improvements in pCO2 reconstruction fidelity that could be achieved with additional autonomous sampling in the Southern Ocean added to existing Surface Ocean CO2 Atlas (SOCAT) observations. The LET allows for a robust evaluation of the skill of pCO2 reconstructions in space and time through comparison to “model truth”. With only SOCAT sampling, Southern Ocean and global pCO2 are overestimated, and thus the ocean carbon sink is underestimated. Incorporating uncrewed surface vehicle (USV) sampling increases the spatial and seasonal coverage of observations within the Southern Ocean, leading to a decrease in the overestimation of pCO2. A modest number of additional observations in Southern Hemisphere winter and across meridional gradients in the Southern Ocean leads to an improvement in reconstruction bias and root-mean-squared error (RMSE) of as much as 86 % and 16 %, respectively, as compared to SOCAT sampling alone. Lastly, the large decadal variability of air–sea CO2 fluxes shown by SOCAT-only sampling may be partially attributable to undersampling of the Southern Ocean.