Under-reporting of greenhouse gas emissions in U.S. cities

Fossil fuel CO2 emissions, the principal anthropogenic greenhouse gas, predominantly originate from cities, which are growing rapidly and risk locking in high-emissions pathways. Cities worldwide have adopted greenhouse gas mitigation goals and rely on self-reported inventories (SRIs), typically built from standard protocols, to set baselines and track progress. However, no systematic, peer-reviewed assessment of the accuracy of SRIs existed. This study addresses that gap by comparing SRIs from 48 U.S. cities with independent, research-grade estimates from the Vulcan v3.0 FFCO2 data product to evaluate the accuracy of city-reported emissions and to understand sources of discrepancy, with implications for urban mitigation planning and verification.

The paper situates city SRIs within established accounting protocols (e.g., ICLEI U.S. Community Protocol, Global Protocol for Community-Scale GHG Emissions, and Global Covenant of Mayors frameworks) and notes that despite their widespread use, SRIs have not been rigorously validated. Prior related work includes the Vulcan project’s high-resolution U.S. FFCO2 estimates and their consistency with atmospheric 14CO2-based assessments and the INFLUX experiment in Indianapolis, which demonstrated close agreement between bottom-up estimates and atmospheric inversions. The study contrasts its approach with Nangini et al. (2019), which archived city-reported data (largely unstandardized and insufficiently documented) and provided limited, low-resolution comparisons to an older Vulcan version, underscoring the need for a careful, scope-aligned, sectorally detailed, and temporally matched evaluation.

• Sample selection: Compiled 48 U.S. cities based on prior literature (e.g., CDP-listed SRIs), highest emitting cities from Vulcan 2011, and availability of adequately documented SRIs. Cities lacking documentation or outside the Vulcan v3.0 time window (2010–2015) were excluded, with four exceptions near the window (Columbus 2017; New Orleans 2007; Longmont 2016; Salt Lake City 2009) mapped to the closest Vulcan year (2010 or 2014); sensitivity tests showed negligible effect on mean statistics.
• Independent emissions dataset: Used Vulcan v3.0, which quantifies complete U.S. FFCO2 emissions at point (e.g., power plants, factories), line (roadways), and polygon (census block-group) scales for 2010–2015, derived from local fuel, activity, and flux monitoring data. Vulcan v3.0 agrees within 1.5% with atmospheric 14CO2-based national estimates and within 3% (3-year mean) to 6% (annual) with INFLUX Indianapolis atmospheric assessments.
• Geographic and sectoral alignment: Extracted Vulcan emissions using unambiguous administrative city boundaries. Ensured alignment between SRIs and Vulcan in terms of gases, scopes, and sectors. When SRIs omitted entire sectors (e.g., airport, railroad, nonroad), those sectors were removed from the Vulcan comparison. Sub-sector omissions within SRIs (e.g., certain fuels, specific facilities) were not adjusted and considered part of the observed differences.
• Scope alignment and gas consistency: Focused on scope 1 emissions (within-boundary sources), consistent with atmospheric inversion applications. SRIs were examined to isolate scope 1 sector emissions. Scope 2/3 elements were removed where possible. Waste-related emissions (often non-CO2) were removed from SRIs to match Vulcan’s FFCO2-only coverage. After alignment, retained emissions represented on average 86.4% of original SRI totals.
• Source category treatment: Vulcan includes all emissions within the city boundary, including onroad emissions from all road segments within the boundary, marine emissions in city waters, and aviation emissions from LTO (up to 3000 ft AGL). Many SRIs differ by allocating only resident-activity portions of marine/air travel or using fuel sales or travel demand models for onroad.
• Metrics and statistics: Computed relative difference (RD) as (Vulcan − SRI)/average(Vulcan, SRI) and absolute (unsigned) relative difference (MAD). Also computed absolute difference (AD = positive RD) and the mean across-sector absolute difference (MASAD) for each city. Evaluated regional patterns (East vs West), city-size dependence (Pearson correlation of AD vs city total), and sector-specific discrepancies (onroad; stationary: residential + commercial + industrial; other transportation: aviation + marine + nonroad + rail). Considered Vulcan’s 95% confidence interval when interpreting total RD per city.

• Average discrepancy: Mean RD across 48 cities is +18.3% (Vulcan > SRI), with MAD of 29.1%; Vulcan urban emissions 95% CI for mean RD: +5.2% to +31.7%.
• Under-reporting prevalence: 37 of 48 cities report fewer emissions than Vulcan; among these, mean RD is +30.7%.
• Aggregate impact: Summed difference across all 48 cities is 19,076,760 tC/year (≈ 2015 Massachusetts emissions). Extrapolated nationally, the difference is 129,219,255 tC/year, 23.5% larger than California’s 2015 emissions.
• Regional patterns: Mean RD is smaller in the West (+11.0%) than the East (+25.0%), while MAD is similar (30.3% vs 28.0%) due to more negative differences in Western cities.
• Scale independence: Little correlation between AD and total city emissions (Pearson’s R = 0.01).
• Sectoral inconsistencies: Despite close total agreement in some cases, sectors diverge strongly. Mean MASAD across cities: 50.3%. Example: Detroit total RD +1.3% but MASAD 44.2%, with onroad RD −28.5% and stationary RD +38.7%.
• Sector-specific AD means (all cities): onroad 28.1% (N=46), stationary 37.9% (N=43), other transportation 82.6% (N=27). For cities under-reporting relative to Vulcan: onroad 25.7% (N=28), stationary 44.9% (N=23), other transportation 102.2% (N=11).
• Likely causes: SRI omission of petroleum fuel use and industrial/commercial point sources; differing treatment of marine and aviation emissions; varying onroad estimation methods (fuel sales vs travel models) versus Vulcan’s within-boundary, comprehensive accounting.

The comparison directly addresses the question of SRI accuracy by showing systematic under-reporting relative to an independently validated, high-resolution FFCO2 dataset. The magnitude and sectoral spread of discrepancies imply that many cities’ baselines are biased, complicating target-setting and progress assessment. For example, Indianapolis’ building sector reduction target could be misinterpreted given a 26.9% underestimate found here. The findings suggest SRIs’ methodological choices (scope mixing, sector omissions, differing transportation accounting) drive much of the divergence. The authors advocate a comprehensive urban GHG information system that integrates detailed bottom-up inventories (e.g., Vulcan) with tiered atmospheric observations (ground, airborne, satellite) and modeling, developed collaboratively with local governments. Such a system can provide consistent, spatially and functionally resolved emissions, enable tracking via atmospheric verification, and ensure mass conservation across scales, thereby improving policy design and evaluation.

This study provides a systematic, peer-reviewed assessment showing that U.S. city self-reported inventories tend to under-report FFCO2 emissions, on average by 18.3%, with substantial sector-level discrepancies. The results call into question the reliability of SRIs for baseline setting and mitigation tracking in their current form and highlight the need for standardized, scope-aligned, and atmospherically calibrated urban emissions systems. Future work should enhance SRI documentation and boundary clarity, expand integrated bottom-up/top-down systems across more cities and internationally, refine sectoral attribution (especially transportation and industrial point sources), and continue independent atmospheric validation to support transparent, effective urban mitigation.

• Documentation gaps in SRIs limited precise alignment of system boundaries and sector definitions; better documentation could alter difference statistics.
• Scope differences: SRIs often mix scopes 1, 2, and 3 and include non-CO2 (e.g., waste), whereas Vulcan quantifies scope 1 FFCO2 only; although adjustments were made (e.g., removing waste), residual inconsistencies may remain.
• Sub-sector omissions in SRIs (e.g., certain fuels, facilities) were not adjusted and contribute to observed differences.
• Temporal mismatches for four cities (mapped to nearest Vulcan year) introduce minor uncertainty, though sensitivity analysis suggested negligible impact on means.
• The sample (48 cities) covers a substantial but incomplete fraction of U.S. urban emissions; extrapolation to all cities adds uncertainty.
• Attribution of city/sector differences was often not fully resolvable due to limited SRI detail.