v11i3 Kania

Accurate and transparent greenhouse gas (GHG) emissions data is crucial for effective climate mitigation, yet existing reporting systems remain inconsistent and difficult to verify. Carbon accounting has emerged to give structure and legitimacy to these measurement efforts by mandating rules for affected GHG emitters through programs such as the Greenhouse Gas Reporting Program (GHGRP) by the United States government. Historically, these programs have relied on self-reporting, significantly limiting the verifiability of corporate-reported data. In contrast, emerging non-profit organizations such as Climate TRACE (CT) estimate facility-level GHG emissions using satellite-based remote sensing. This study quantifies facility-level discrepancies between these datasets and identifies their key drivers. To do so, I systematically matched and compared facilities from both datasets based on reported emissions. I identified key predictors of these discrepancies using machine learning and Bayesian inference. My findings reveal that facility-specific effects drive most observed disparities, while parent-company and geographic influences play a secondary role. Industry-wide effects contribute minimally, with reporting year and total emissions volume having no significant impact. These results suggest that discrepancies stem from isolated inaccuracies rather than systemic errors, underscoring the need for hybrid verification frameworks that integrate self-reported (bottom-up) inventories with independent satellite-based monitoring (top-down) to enhance emissions transparency and accountability.