We perform a simultaneous data assimilation of surface CO2 fluxes and atmospheric CO2 concentrations along with meteorological variables using the Local Ensemble Transform Kalman Filter (LETKF) within an Observing System Simulation Experiments (OSSEs) framework. Surface CO2 fluxes are not observed, but estimated as “parameters”, augmenting the state vector of atmospheric CO2 concentrations with the surface fluxes, and using the ensemble Kalman filter to estimate the multivariate error covariance. The “localization of variables” method has already been shown to reduce sampling errors in the CO2 LETKF multivariate analysis system. In this paper, we focus on the impact of advanced inflation methods and vertical localization of column CO2 data on the analysis of CO2 variables. With both additive inflation and adaptive multiplicative inflation, we are able to obtain encouraging multiseasonal analyses of surface CO2 fluxes in addition to atmospheric CO2 and meteorological analyses. By contrast, the analysis performed with a standard fixed multiplicative inflation results in a poor estimation of surface CO2 fluxes which become “stuck” in time. In addition, we examine strategies for vertical localization in the assimilation of simulated CO2 from GOSAT (or OCO-2) that have nearly uniform sensitivity from the surface to the upper troposphere. Since atmospheric CO2 is forced by surface fluxes, its short-term variability should be largest near the surface layer. We take advantage of this by updating only the lower tropospheric CO2, rather than the full column. This results in a more accurate analysis of CO2 in terms of both RMS error and spatial patterns. Assimilating simulated CO2 ground-based observations and CO2 retrievals from GOSAT and AIRS with the enhanced LETKF, we obtain a rather accurate estimation of the evolving surface fluxes even in the absence of any a priori information.