Tianyi Guan

Abstract

Historical greenhouse gas databases are revised after publication, but database release is rarely tested as a source of analytical sensitivity. We compared country or territory–sector–year methane and nitrous oxide estimates across three releases of the Emissions Database for Global Atmospheric Research: v8.0, 2024 GHG and 2025 GHG. Primary comparisons used positive–positive records that were structurally comparable across all three releases for 1990–2020. We quantified same-year absolute symmetric revisions and tested whether release choice changed predefined trend categories based on normalized Theil–Sen slopes or tie-aware country rankings. Across 80,837 methane and 86,978 nitrous oxide matched records in each sequential comparison, median revisions were zero or near zero, but 90th-percentile revisions ranged from 5.5% to 40.7% and emission-weighted absolute-revision footprints from 5.1% to 13.3%. At the prespecified 0.5% yr⁻¹ classification threshold, 2.72–6.52% of eligible trend series changed category, accounting for 0.97–4.97% of pair-symmetric mean annual emissions within the eligible trend support. Median Kendall rank agreement was high (0.986–0.997), yet 29–36 of 84–92 panels met a predefined ranking-sensitivity criterion and 12 of 352 eligible panels had disjoint tie-aware sets of top-ranked countries. Thus, successive releases preserved most central summaries while changing some conclusion-bearing country–sector trends and rankings. Because historical trend labels and leading-emitter rankings summarize comparative emissions evidence, exact release identity should be reported, archived and tested when they carry a conclusion. Inter-release differences alone do not identify error, improvement or policy effect.

Keywords: greenhouse gas inventories; historical revisions; release sensitivity; methane; nitrous oxide; environmental informatics; reproducibility