An open dataset reconstructs daily carbon dioxide emissions back to 1970

What happened

Researchers combined more than two million recent activity observations with machine-learning and degree-day methods to build daily fossil carbon-dioxide estimates for 1970–2024. The public product includes global, country or regional, and sectoral series for power, industry, transport, and buildings.

Why it matters

Daily timing can help researchers study how heat waves, cold snaps, outages, and mobility shocks change fossil-fuel use. The 1970–2018 record is reconstructed using relationships learned from 2019–2024, so it is not a direct daily measurement or a replacement for official inventories.

What to watch

• Independent validation by region and sector, transparent uncertainty, and sensitivity to long-term structural change.
• Post-2024 updates, a planned gridded product, and comparisons with official annual inventories.

Sources & evidence

• Global daily fossil CO2 emissions from 1970 to 2024Peer-reviewed Data Descriptor, Scientific Data (2026), published February 24, 2026. Recent activity inputs are combined with reconstructed historical daily series; the latter are estimates rather than direct observations.
• Global daily CO2 emissions datasetPublic Figshare dataset associated with the paper.
• Global daily CO2 emissions code repositoryPublic research code repository cited by the authors.

Topics

• Carbon Dioxide
• Open Data
• Emissions Tracking
• Climate Data
• Machine Learning