Audit Scope and Deviation Magnitude: 70% Average Undercount Across 260 Cities, Over 90% in Individual Cities
The audit targeted the urban on-road vehicle (cars and trucks) CO2 sector of the global greenhouse gas database released by the Climate TRACE consortium, co-founded by former U.S. Vice President Al Gore. A team led by Professor Kevin Gurney of the School of Informatics, Computing, and Cyber Systems (SICCS) at Northern Arizona University compared it city-by-city against the Vulcan on-road emissions database developed in his laboratory, covering 260 U.S. cities. As disclosed in the source: Climate TRACE urban vehicle CO2 estimates averaged 70% lower than Vulcan, with individual cities such as Indianapolis and Nashville lower by more than 90%.
Benchmark System and Algorithm Architecture: Vulcan's Official-Traffic-Record Calibration Chain vs. Climate TRACE's AI Route
Vulcan, developed by Gurney's laboratory, is calibrated using official traffic records and energy consumption data (source disclosure), i.e., it reconstructs city-level road emissions bottom-up anchored to auditable administrative statistics such as traffic counts, vehicle fleet composition, and fuel sales. Bilal Aslam, SICCS postdoc and co-investigator, stated that while Vulcan on-road data is not perfect, with about 14% uncertainty, this is far lower than the differences found across the 260-city comparison (source disclosure).
Field Comparison Data and Extrapolation: From the 260-City Sample to Global Data Credibility Risk
The measured comparison chain is clear: the sample covers on-road vehicle CO2 in 260 U.S. cities, compared city-by-city across the two databases; results show a 70% average undercount, worst cases (Indianapolis, Nashville) above 90%, and a Vulcan benchmark uncertainty of about 14% (all source disclosures). The paper was published in Environmental Research Letters; authors include Gurney, SICCS postdoc Bilal Aslam, and research associate Pawlok Dass.
Remediation Recommendations for AI Emission Inventories and Domestic Implementation Points
The paper outlines several recommendations intended to strengthen and improve Climate TRACE's work (source disclosure), pointing to three essentials: transparency (publicly auditable methods and data chains), expert review (peer-review mechanisms covering the algorithm pipeline), and rigorous scientific method (closed-loop validation against independent benchmark databases). Gurney's warning targets governance directly: substandard emission data mislead decision makers and risk eroding public trust in climate action capability (source disclosure).