Measurements of the surface ocean fugacity of carbon dioxide (fCO2) provide an important constraint on the global ocean carbon sink, yet the gap-filling products developed so far to cope with the sparse observations are relatively coarse (1° × 1° by 1 month). Here, we overcome this limitation by using a novel combination of machine ...
Measurements of the surface ocean fugacity of carbon dioxide (fCO2) provide an important constraint on the global ocean carbon sink, yet the gap-filling products developed so far to cope with the sparse observations are relatively coarse (1° × 1° by 1 month). Here, we overcome this limitation by using a novel combination of machine learning-based methods and target transformations to estimate surface ocean fCO2 and the associated sea-air CO2 fluxes (FCO2) globally at a resolution of 8-day by 0.25° × 0.25° (8D) over the period 1982 through 2022. Globally, the method reconstructs fCO2 with accuracy similar to that of low-resolution methods (∼19 μatm), but improves it in the coastal ocean. Although global ocean CO2 uptake differs little, the 8D product captures 15% more variance in FCO2. Most of this increase comes from the better-represented subseasonal scale variability, which is largely driven by the better-resolved variability of the winds, but also contributed to by the better-resolved fCO2. The high-resolution fCO2 is also capable of capturing the signal of short-lived regional events such as hurricanes. For example, the 8D product reveals that fCO2 was at least 25 μatm lower in the wake of Hurricane Maria (2017), the result of a complex interplay between the decrease in temperature, the entrainment of carbon-rich waters, and an increase in primary production. By providing new insights into the role of higher frequency variations of the ocean carbon sink and the underlying processes, the 8D product fills an important gap.
