NAU Analysis Finds Climate TRACE Underestimates US Urban Vehicle CO2 Emissions by 70% on Average Versus Vulcan Benchmark

Kevin Gurney's team compared Climate TRACE AI-derived city-scale vehicle CO2 to the Vulcan database, which integrates traffic counts and fuel sales with 14% uncertainty. The 70% average shortfall, plus prior power-plant undercounts, implies Climate TRACE misses over half of US urban fossil emissions. This compounds when cities rely on such data for net-zero targets under frameworks like the Global Covenant of Mayors. The pattern echoes documented AI training shortfalls in remote-sensing emissions models that lack ground-truth calibration at sub-national scales. Independent audits using atmospheric inversions from towers or satellites would expose whether the bias is US-specific or global, directly affecting Nationally Determined Contribution verification. Policymakers using Climate TRACE for 2025-2030 progress tracking risk over-crediting reductions unless corrected inventories are adopted within two budget cycles.