Ocean carbon data products, code, papers
In the past several years, the McKinley group has developed two machine learning methods to interpolate sparse surface ocean fCO2 data from SOCAT to global coverage.
For fCO2-Residual (monthly 1982-2022), we set aside the temperature component of pCO2 prior to machine learning in order to focus the statistics on the more poorly constrained biological / physical component (Bennington et al. 2022, JAMES).
In LDEO-HPD (monthly 1959-2022), we use hindcast ocean models as a prior, and use machine learning to correct them toward the data (Gloege et al. 2022 JAMES, and Bennington et al. 2022, GRL). We contribute this product annually to the Global Carbon Budget.
In collaboration with colleagues at NOAA, we have also developed a final version of the LDEO/Takahashi climatology of fCO2 and net air-sea CO2 flux Fay et al. 2024.
To make these results easily available, we have developed a new website that describes the data products in detail and that provides links to the data products and their source code. This site will be updated annually.
Please check it out this new site and bookmark it for future reference!