Constraining urban fossil fuel CO2 emissions in Seoul using combined ground and satellite observations with Bayesian inverse modelling

Abstract

Accurate carbon emission estimates are essential for achieving net zero targets by 2050. The Bayesian inverse method, combined with atmospheric carbon dioxide (CO2) measurements and a transport model, can serve as an independent verification approach to improve accuracy. We developed a Bayesian inverse modelling framework using ground- and space-based measurements and applied it to Seoul to test the framework and constrain its fossil fuel CO2 emissions. By leveraging the high temporal resolution of ground-based in situ observations and the broad spatial coverage of satellite data, we improved the accuracy of emission estimates. Our results indicate the spatiotemporal variability of posterior emissions increased significantly, enabling us to track CO2 fluctuations and assess the impact of carbon reduction policies over time and space. The mean absolute error between simulated and observed CO2 enhancements decreased by 55 %, indicating improved agreement. We investigated the performance of the inverse model through a sensitivity analysis that considered different observational network configurations. Uncertainty reductions varied with the type of observations used: 19.2 % when all observations were included, 18.7 % using only ground-based sites, and 6 %-8.4 % when using only OCO-2 or OCO-3 satellite data, highlighting the complementary contributions of ground and space-based measurements. The analysis also showed that assumptions about background concentrations, biogenic fluxes, and prior emission uncertainties can alter posterior results, demonstrating the importance of model configuration. The framework shows strong potential for application in other cities and can support the development of effective climate mitigation policies.