Cities are responsible for more than 70% of the global CO2 emissions and thus play an important role in mitigating climate change. Mayors and local governments have been taking measures to reduce urban CO2 emissions and to reach carbon neutrality. In order to evaluate their efforts, a series of high-resolution city-scale emission inventories were established. Top-down inversion modeling is a widely-used complementary solution to reduce the uncertainties in traditional bottom-up emission inventories. It combines atmospheric modeling and measurements to optimize the greenhouse gas estimates using Bayesian inference methods.The Mexico City Metropolitan Area (MCMA) is one of the largest megacities in the world. Its annual CO2 emissions have grown from 42.1 Mt to 66.0 Mt from 2012 to 2018. The Mexico government has planned to reduce 65.2 Mt CO2 emission during the period 2021-2030. To assess local CO2 emission reduction strategies, a French-Mexican project Mexico City regional Carbon impacts (MERCI-CO2) deployed a network of in-situ and column CO2 observation instruments in MCMA. The CO2 concentration gradients are assimilated in our inversion system based on the WRF-Chem model to improve the inventory estimates of CO2 emissions in and outside MCMA.Various options on meteorological drivers, domain sizes, physics, dynamics schemes and spectral nudging of the WRF modeling system over MCMA were quantitatively evaluated for model performance. A series of meteorological parameters were taken into account for the comparison between simulations and in-situ observations, LiDar analysis as well as WMO radiosonde observations. For the purpose of CO2 simulation, the most studied variables are those related to the dispersion of the ambient air, including air temperatures, wind speeds, wind directions and mixing heights. These sensitivity tests helped to define the optimal model configuration.The CO2 concentration maps over MCMA during 3 typical months (January, May and July) were reconstructed by the double-nesting 5-km resolution WRF-Chem model, coupled with the local emission inventories from UNAM and the global emission inventories ODIAC scaled by temporal scaling factors. The evaluation of CO2 simulations were based on CO2 in-situ measurements by PICARRO and column measurements (XCO2) by FTIR at an urban site UNA and at the background station ALZ. Along with the evaluation, we also analyzed the temporal and spatial distribution of CO2 signals, as well as the area impacted by anthropogenic fluxes and by biogenic fluxes. Based on our analysis, we assessed the potential of our network to constrain the urban emissions, defined the potential locations for future stations, and defined a “background index” to represent the suitability to build a background station.After the ground validation of the modeling system, we performed a 1-year inversion over the MCMA from 30 March 2018 to 30 March 2019. According to the assimilation of concentration gradients between the urban station and the rural station, the inversion adjusted the prior anthropogenic emission from UNAM and ODIAC estimates, in parallel with prior biogenic fluxes from the CASA model and background concentrations by CarbonTracker 2019B global inversion system. An ensemble of inversion configurations was constructed. The reference configuration optimizes three components: fossil fuel sources, biogenic fluxes and background concentrations to generate separate scaling factors for each block of 5 days. The sensitivity tests include several temporal error correlation length scales between continuous days, varying time windows over each day, a separation of the activity sectors (traffic), a filter over the MCMA, varying data screening and block sizes, to evaluate the performances of the inversion, and to specify the impact of our various configurations. The same system was also used to assimilate carbon monoxide concentrations, collected at the two stations since December 2018.