Selective Impacts of Prior, Background-Concentration, and Observation-Error Assumptions on Urban CO₂ Inversion Estimates

Abstract. Urban-scale CO₂ inversions are increasingly used to constrain fossil fuel CO₂ emissions; however, posterior estimates can depend on the methodological assumptions used to define key inversion parameters. How these assumptions selectively affect different aspects of posterior information remains unclear. Here we quantify the impacts of prior emissions, background concentrations, and observation-error covariance assumptions on urban CO₂ inversion results. Using wintertime Seoul, South Korea, as a case study to minimize biogenic flux uncertainties, we conducted 18 Bayesian inversions by combining two prior emission inventories, three background concentration definitions, and three observation-error estimation methods. We found that these assumptions affect distinct components of posterior emissions: the spatial distribution of posterior estimates was primarily controlled by the choice of prior, whereas total emissions and their temporal variability were more strongly influenced by observation-error and background-concentration assumptions. The spread in posterior estimates arising from these methodological assumptions ranged from 25.6 % to 51.5 % of total emissions, exceeding the posterior uncertainty estimated within individual inversion configurations. This result indicates that uncertainty estimates from a single inversion setup may not fully capture the methodological uncertainty associated with input assumptions. Using parameter sensitivity and an uncertainty reduction metric, we further identified a relatively robust configuration for the Seoul winter daytime case and evaluated whether the inversion reduced the spread among the estimates relative to the prior spread. Our results demonstrate that methodological assumptions can selectively shape urban CO₂ inversion estimates and highlight the need to assess their impacts when interpreting posterior emissions.