The accurate quantification of anthropogenic carbon dioxide (CO₂) emissions in urban areas is hindered by high uncertainties in emission inventories. We assessed the spatial distributions of three anthropogenic CO₂ emission inventories in Shanghai, China—MEIC (0.25° × 0.25°), ODIAC (1 km × 1 km), and a local inventory (LOCAL) (4 km × 4 km)—and compared simulated CO₂ column concentrations (XCO₂) from WRF-CMAQ against OCO-3 satellite Snapshot Mode XCO₂ observations. Emissions differ by up to a factor of 2.6 among the inventories. ODIAC shows the highest emissions, particularly in densely populated areas, reaching 4.6 and 8.5 times for MEIC and LOCAL in the central area, respectively. Emission hotspots of ODIAC and MEIC are the city center, while those of LOCAL are point sources. Overall, by comparing the simulated XCO₂ values driven by three emission inventories and the WRF-CMAQ model with OCO-3 satellite XCO₂ observations, LOCAL demonstrates the highest accuracy with slight underestimation, whereas ODIAC overestimates the most. Regionally, ODIAC performs better in densely populated areas but overestimates by around 0.22 kt/d/km² in relatively sparsely populated districts. LOCAL underestimates by 0.39 kt/d/km² in the center area but is relatively accurate near point sources. Moreover, MEIC’s coarse resolution causes substantial regional errors. These findings provide critical insights into spatial variability and precision errors in emission inventories, which are essential for improving urban carbon inversion.

Satellite XCO₂ retrievals have been widely used in estimating fossil fuel carbon (FFC) emissions at point and urban scales. However, at the regional scale, it remains a significant challenge. Furthermore, current global and regional atmospheric inversions often overlook the uncertainties associated with FFC emissions. To meet the needs of the global carbon stocktake, we developed an inversion method based on Bayesian statistical theory and OCO-2 satellite XCO₂ observations to optimize FFC emissions alongside terrestrial ecosystem carbon fluxes (NEE). The methodology’s core is to distinguish the contributions of NEE and FFC to the observed concentrations using their different spatial distributions. We designed an observing system simulation experiment to invert the 2016 FFC emissions. The results showed that posterior FFC emissions were significantly optimized during the non-growing seasons in the regions with high emissions, with the optimization effect diminishing as emissions shrank. Average FFC emissions uncertainty reductions are in the range of 13–82% in the non-growing season for the eight largest emitting regions globally. By assuming the same uncertainty reduction for FFC emissions in both the growing and non-growing seasons, we can optimize annual emissions for high-emission areas. We believe this study provides a new idea for the inversion of FFC emissions at the regional scale, which is important for achieving the goal of carbon neutrality.

Monitoring and verifying anthropogenic greenhouse gases (GHG) Emissions at high spatiotemporal resolution with observation-based evidence is desirable for climate policymakers. A multiple-scale nested GHG assimilation system, named GCASv3, was developed for quantifying anthropogenic GHG emissions at high spatiotemporal resolution. GCASv3 uses a four level nested scheme and consists of one global module and one regional module. The global model is capable of assimilating XCO2 and XCH4 data to infer global CO2 flux and CH4 emissions at 10x10 resolution, while the regional module is able to assimilate ground and satellite GHG observations to quantify anthropogenic GHG emissions on national, regional and urban scales at 27km, 9km and 1km resolution respectively. This presentation describes briefly the framework and the major components of GCASv3. Anthropogenic CO2 emissions and CH4 emissions inferred by GCASv3 at different scales are presented and discussed.