GOBAI-O2: Training Data for Global Ocean Dissolved Oxygen Models

Global ocean data from quality-controlled Biogeochemical Argo and GLODAP sources, used to train machine learning models. The dataset was compiled by Jonathan D. Sharp of the University of Washington to generate the GOBAI-O2 gridded product of ocean interior dissolved oxygen concentrations. It is derived from autonomous float and shipboard observation programs.

Use Cases

• Training ML models to predict dissolved oxygen concentrations based on temperature and salinity maps.
• Validating three-dimensional ocean biogeochemical models using quality-controlled in-situ data.
• Studying global ocean deoxygenation trends by leveraging data from the Argo and GLODAP programs.

Strengths

• Data is quality-controlled from authoritative sources like the Biogeochemical Argo database and GLODAP.
• Explicitly designed for training machine learning models applied to a specific, published product (GOBAI-O2).

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count and file size are unknown, which may limit suitability assessment.
• Last update date is unknown; freshness unverified.

Provenance

Source

Biogeochemical Argo database and GLODAPv2.2022 data product.

Collection Method

Quality-controlled data compilation from autonomous floats and shipboard observations.

Geography

Global ocean