CarboScope pCO2 Interpolation: Method Summary

The sea-air CO2 flux estimates are primarily based on spatio-temporal interpolation of the pCO2 data, but also compatible with the dynamics of mixed-layer carbon content (including the buffer effect). Firstly, the sea-air CO2 fluxes and the pCO2 field are numerically linked to each other and to the spatio-temporal field of ocean-internal carbon sources/sinks through parametrizations of sea-air gas exchange, solubility, and carbonate chemistry, as well as a budget equation for the mixed-layer dissolved inorganic carbon (DIC) content. Then, the ocean-internal carbon sources/sinks are adjusted to optimally fit the pCO2 field to the pCO2 observations. Spatio-temporal interpolation is achieved by enforcing the source/sink adjustments to be smooth (through a Bayesian a-priori autocorrelation with prescribed spatial and temporal de-correlation scales). The temporal interpolation is overlaid by the inherent relaxation timescales of the mixed-layer carbon budget. Though the process parametrizations are driven by SST, wind speed, mixed-layer depth (MLD) climatology, alkalinity climatology, and some auxiliary variables, this external variability only determines features not constrained by the pCO2 observations (e.g. day-to-day variations, or variability in data-void areas/periods), while the estimated pCO2 field in well-constrained areas/periods is determined by the observed signals (i.e., there is no regression against the driver fields).