Improved and Expanded Global Ocean Carbon Product Using Machine Learning

  • 15 July 2022
Improved and Expanded Global Ocean Carbon Product Using Machine Learning

Ocean processes remove CO2 from the atmosphere and reduce human impact on climate. For example, since the industrial revolution, oceans have removed approximately 37% of anthropogenic carbon emissions. The air-sea CO2 exchange can be estimated using analysis of observations and global ocean biogeochemical models (GOBMs), but taken separately, these both have significant regional and seasonal bias. The LDEO-Hybrid Physics Data product uses a combined approach of modeling, observations, and machine learning,  resulting in an improved long-term record of estimates of air-sea fluxes. This product’s incorporation of machine learning in modeling to advance our understanding of anthropogenic change aligns with the primary goals of the NOAA Artificial Intelligence Strategy.

The LDEO-Hybrid Physics Data product was previously published for 1982-2018, but a new study, partially funded by the Climate Program Office’s Climate Observations and Monitoring (COM) program, extends the estimate to 1959 - 2020. This research is part of a COM initiative to develop improved surface-atmosphere interaction representation in models. This expanded data product, published by researchers Val Bennington, Lucas Gloege, and Galen McKinley, indicates that the ocean carbon sink increased over the last 60 years, consistent with growth of atmospheric CO2. These results can be used to understand annual changes in the ocean carbon sink and impacts to global climate.

Read the article »




1315 East-West Highway Suite 100
Silver Spring, MD 20910


Americans’ health, security and economic wellbeing are tied to climate and weather. Every day, we see communities grappling with environmental challenges due to unusual or extreme events related to climate and weather.