We investigate the potential of the Copernicus Anthropogenic Carbon Dioxide (CO2) Monitoring (CO2M) mission, a proposed constellation of CO2 imaging satellites, to estimate the CO2 emissions of a city on the example of Berlin, the capital of Germany. On average, Berlin emits about 20 MtCO(2) yr(-1) during satellite overpass (11:30 LT). The study uses synthetic satellite observations of a constellation of up to six satellites generated from 1 year of high-resolution atmospheric transport simulations. The emissions were estimated by (1) an analytical atmospheric inversion applied to the plume of Berlin simulated by the same model that was used to generate the synthetic observations and (2) a mass-balance approach that estimates the CO2 flux through multiple cross sections of the city plume detected by a plume detection algorithm. The plume was either detected from CO2 observations alone or from additional nitrogen dioxide (NO2) observations on the same platform. The two approaches were set up to span the range between (i) the optimistic assumption of a perfect transport model that provides an accurate prediction of plume location and CO2 background and (ii) the pessimistic assumption that plume location and background can only be determined reliably from the satellite observations. Often unfavorable meteorological conditions allowed us to successfully apply the analytical inversion to only 11 out of 61 overpasses per satellite per year on average. From a single overpass, the instantaneous emissions of Berlin could be estimated with an average precision of 3.0 to 4.2 Mtyr(-1) (15%-21% of emissions during overpass) depending on the assumed instrument noise ranging from 0.5 to 1.0 ppm. Applying the mass-balance approach required the detection of a sufficiently large plume, which on average was only possible on three overpasses per satellite per year when using CO2 observations for plume detection. This number doubled to six estimates when the plumes were detected from NO2 observations due to the better signal-to-noise ratio and lower sensitivity to clouds of the measurements. Compared to the analytical inversion, the mass-balance approach had a lower precision ranging from 8.1 to 10.7 Mtyr(-1) (40% to 53%), because it is affected by additional uncertainties introduced by the estimation of the location of the plume, the CO2 background field, and the wind speed within the plume. These uncertainties also resulted in systematic biases, especially without the NO2 observations. An additional source of bias was non-separable fluxes from outside of Berlin. Annual emissions were estimated by fitting a low-order periodic spline to the individual estimates to account for the seasonal variability of the emissions, but we did not account for the diurnal cycle of emissions, which is an additional source of uncertainty that is difficult to characterize. The analytical inversion was able to estimate annual emissions with an accuracy of < 1.1 Mtyr(-1) (< 6 %) even with only one satellite, but this assumes perfect knowledge of plume location and CO2 background. The accuracy was much smaller when applying the mass-balance approach, which determines plume location and background directly from the satellite observations. At least two satellites were necessary for the mass-balance approach to have a sufficiently large number of estimates distributed over the year to robustly fit a spline, but even then the accuracy was low (> 8 Mtyr(-1) (> 40 %)) when using the CO2 observa-tions alone. When using the NO2 observations to detect the plume, the accuracy could be greatly improved to 22% and 13% with two and three satellites, respectively. Using the complementary information provided by the CO2 and NO2 observations on the CO2M mission, it should be possible to quantify annual emissions of a city like Berlin with an accuracy of about 10% to 20%, even in the pessimistic case that plume location and CO2 background have to be determined from the observations alone. This requires, however, that the temporal coverage of the constellation is sufficiently high to resolve the temporal variability of emissions.
