Input uncertainty overshadows cropland greenhouse gas accounting and mitigation tracking

Abstract Tracking agricultural greenhouse gas (GHG) mitigation to achieve the Paris Agreement goal requires robust emission baselines. Here we test the impact of choice of input datasets on emission baselines. We tested 1,284 source-specific input-dataset combinations across11,100 gridded emission maps through an IPCC Tier 1/2 framework at 5 arcmin resolution, covering cropland GHG from nitrogen inputs (synthetic fertilizer, manure and crop residues), rice paddies, and drained peatland, to characterize the dataset-driven uncertainty. The range of total emissions across input datasets was 1.9 Gt CO2eq yr-1, comparable to the central estimate of total cropland GHG emissions circa 2020 (2.4 Gt CO2eq yr-1) and to the assessed annual technical mitigation potential of these sources by 2050 (1.8 Gt CO2eq yr-1). For countries responsible for 99% of rice CH₄, 72% of nutrient N₂O and 50% of peatland emissions among those with an assessed mitigation potential, the variation of national emission estimates across input datasets exceeded the country's technical mitigation potential for the same source. This makes it extremely challenging to confidently monitor reductions in emissions from these sources. Activity dataset choice was the leading contributor to variance for every source (78-99% of ANOVA sum-of-squares), substantially exceeding climate, hydrology, or emission-factor zonation. The uncertainty carried a distinct spatial fingerprint, concentrated in identifiable sub-national hotspots: the Indo-Gangetic Plain for nutrient N₂O, monsoon-Asia rice basins for CH₄, and tropical peat-forest margins for peatland CO₂. These findings show an urgent need to harmonize and standardize agricultural activity datasets, because consistent input data are an essential prerequisite to reliably track national cropland mitigation.