Differences in anthropogenic greenhouse gas emissions estimates explained

We examine differences in global and national greenhouse gas (GHG) emissions estimates, focusing on the role of varying system boundaries and conceptual approaches in driving these variations. Despite consensus among assessments and datasets that GHG emissions continue to increase and that trends are far from aligned with the Paris Agreement goals, estimates can differ significantly. Our review finds three main reasons for these differences. First, datasets vary in their coverage of gases, sectors and countries; second, there are different approaches to defining “anthropogenic” emissions and removals in the land use, land-use change and forestry (LULUCF) sector; and third, the Paris Agreement doesn’t cover all relevant sources of emissions, including the cement carbonation sink and ozone depleting substances. As different assessments have different objectives, they may deal with these issues differently. We highlight three assessment conventions that report or use emissions data: those focused on interpreting national progress, policies and pledges under the Paris Agreement; those consistent with integrated assessment modelling (IAM) benchmarks of emissions under different warming scenarios; and those consistent with climate forcing assessments. Considering annual average emissions over the period 2014 to 2023, we show global totals of 44.4 GtCO ₂e yr ⁻¹ [90 % CI ± 4.9], 54.5 GtCO ₂e yr ⁻¹ [90 % CI ± 5.6], and 56.4 GtCO ₂e yr ⁻¹ [90 % CI ± 5.7] for these three conventions, respectively. We suggest that users of GHG emissions data increase transparency in their decision criteria for choosing datasets and setting the scope of an assessment. The data used in this study to make Figs. 9–14 is available at: https://doi.org/10.5281/zenodo.15126539 (Lamb, 2026).