Satellite observations of CO2 abundance in the upper troposphere can provide a major constraint for deriving the net carbon fluxes from tropical landmasses that is unavailable from current surface observations. Such global CO2 profiling with an uncertainty of about 1% (3 ppm) contains key longitudinal information needed to derive surface fluxes in a standard Bayesian inversion. Upper-tropospheric data available from flight-proven FTIR solar occultation measurements could provide comparable information to that from yet-to-be-demonstrated column CO2 observations, which have heretofore been the focus of carbon cycle studies. A strategy for improving CO2 source inversions with either type of satellite data should focus on tropical observations and on careful evaluation of possible sampling biases affecting the observational uncertainties. ¿ 2001 American Geophysical Union