Comparison of EDGAR, ODIAC, and MEIC Grid CO2 Emission Inventories: Historical and Current Versions

High-resolution gridded CO2 emission inventories underpin carbon cycle research and facilitate bridging global and national climate mitigation targets with regional emission assessments and site-specific mitigation actions. Major datasets including ODIAC, EDGAR, MEIC-global and MEIC-China adopt inconsistent spatial proxies and accounting frameworks, causing notable discrepancies. This study compares historical and latest versions at the global, top five emitting national, and grid scales, analyzing cross-version/inter-inventory variations and grid scale effects. The results reveal global total emission deviations of only 7.97% and 2.03%, yet pronounced regional heterogeneity. Maximum discrepancies reach 17.58% in Russia and 44.68% in Japan; grid-scale deviations over 50% exceed 35%. Post-v2022 ODIAC and post-v8.0 EDGAR have reduced version uncertainty, whereas MEIC-China shows large inter-version differences of 14.52% and 17.88% from long update cycles. Grid scale exerts nonlinear impacts on discrepancies, with the most pronounced transitions observed around 2°: discrepancies tend to amplify at resolutions finer than 2°, whereas discrepancies gradually diminish and appear to stabilize at resolutions coarser than 2°. This work clarifies core inventory characteristics, informing data selection, fusion and precision carbon management.