LDEO pCO2-Residual Method

The ocean reduces human impacts on global climate by absorbing and sequestering CO₂ from the atmosphere. To quantify global, time-resolved air-sea CO₂ fluxes, surface ocean pCO₂ is needed. A common approach for estimating full-coverage pCO₂ is to train a machine learning algorithm on sparse in situ pCO₂ data and associated physical and biogeochemical observations. Though these associated variables have understood relationships to pCO₂, it is often unclear how they drive pCO₂ outputs. Here, we make two advances that enhance connections between physical understanding and reconstructed pCO₂. First, we apply pre-processing to the pCO₂ data to remove the direct effect of temperature. This enhances the biogeochemical/physical component of pCO₂ in the target variable and reduces the complexity that the machine learning must disentangle. Second, we demonstrate that the resulting algorithm has physically understandable connections between input data and the output biogeochemical/physical component of pCO₂. The final pCO₂ reconstruction agrees modestly better with independent data than most other approaches. Uncertainties in the reconstructed pCO₂ and impacts on the estimated CO₂ fluxes are quantified. Uncertainty in piston velocity drives substantial flux uncertainties in some regions, but does not increase globally integrated estimates of uncertainty in CO₂ fluxes from observation-based products. Our reconstructed CO₂ fluxes show larger interannual variability than smoother neural network approaches, but a lesser trend since 2005.