The contemporary global carbon cycle is dominated by perturbations from anthropogenic CO2 emissions. One approach to identify, quantify, and monitor anthropogenic emissions is to focus on intensely emitting urban areas. In this study, we compare the ability of different CO2 observing systems to constrain anthropogenic flux estimates in the Los Angeles megacity. We consider different observing system configurations based on existing observations and realistic near-term extensions of the current ad hoc network. We use a high-resolution regional model (Stochastic Time-Inverted Lagrangian Transport-Weather Research and Forecasting) to simulate different observations and observational network designs within and downwind of the Los Angeles (LA) basin. A Bayesian inverse method is employed to quantify the relative ability of each network to improve constraints on flux estimates. Ground-based column CO2 observations provide useful complementary information to surface observations due to lower sensitivity to localized dynamics, but column CO2 observations from a single site do not appear to provide sensitivity to emissions from the entire LA megacity. Surface observations from remote, downwind sites contain weak, sporadic urban signals and are complicated by other source/sink impacts, limiting their usefulness for quantifying urban fluxes in LA. We find a network of eight optimally located in-city surface observation sites provides the minimum sampling required for accurate monitoring of CO2 emissions in LA, and present a recommended baseline network design. We estimate that this network can distinguish fluxes on 8 week time scales and 10 km spatial scales to within ~12 g C m–2 d–1 (~10% of average peak fossil CO2 flux in the LA domain).