The threat posed by the increasing concentration of carbon dioxide (CO 2 ) in the atmosphere motivates a detailed and precise estimation of CO 2 emissions and removals over the globe. This study refines the spatial resolution of the CAMS/LSCE inversion system, achieving a global resolution of 0.7°latitude and 1.4°l ongitude, or three times as many grid boxes as the current operational setup. In a 2-year inversion assimilating the midday clear-sky retrievals of the column-averaged dry air mole fraction of carbon dioxide (XCO 2 ) from NASA's second Orbiting Carbon Observatory (OCO-2), the elevated resolution demonstrates an improvement in the representation of atmospheric CO 2 , particularly at the synoptic timescale, as validated against independent surface measurements. Vertical profiles of the CO 2 concentration differ slightly above 22 km between resolutions compared to AirCore profiles, and highlight differences in the vertical distribution of CO 2 between resolutions. However, this disparity is not evident for XCO 2 , as evaluated against independent reference ground-based observations. Global and regional estimates of natural fluxes for 2015-2016 are similar between the two resolutions, but with North America exhibiting a higher natural sink at high resolution for 2016. Overall, both inversions seem to yield reasonable estimates of global and regional natural carbon fluxes. The increase in calculation time is less than the increase in the number of operations and in the volume of input data, revealing greater efficiency of the code executed on a graphics processing unit. This allows us to make this higher resolution the new standard for the CAMS/LSCE system.

Plain Language Summary Human activities have significantly increased the amount of carbon dioxide (CO 2 ) in the atmosphere, a major driver of climate change. Accurately quantifying CO 2 emissions and removals, known as fluxes, is crucial for implementing effective mitigation strategies. Inverse models are computer programs that analyze large amounts of CO 2 observations to estimate surface fluxes that best match these observations in space and time. While satellites provide extremely precise CO 2 observations all around the Earth, most inverse models lack the resolution to fully utilize these data at a large scale. Our study doubled the horizontal resolution of our inverse model, enhancing its performance and spatial precision when using data from the OCO-2 satellite. Thanks to graphics processing unit acceleration, the computational cost remained manageable. This improved resolution is now being implemented in the European Copernicus Atmosphere Monitoring Service, with ongoing efforts to further improve the resolution. This advancement promises a more detailed understanding of global CO 2 dynamics, supporting climate change mitigation efforts.