We constructed a model to analyze the interactions between land-use change and atmospheric CO2 during the recent past and for the future. The primary impact of the conversion of forested lands to cultivated lands is to increase atmospheric CO2, via losses of biomass and soil carbon to the atmosphere. This increase is likely to continue in the next decades, but its magnitude can vary according to each land-use scenario. We show that this first-order effect is further amplified by the correlated diminution of terrestrial sinks, because when croplands replace forests, the turnover time of excess carbon in the biosphere decreases, and hence the sink capacity of terrestrial ecosystems decreases. This effect acts to further increase by up to 100 ppm the CO2 level reached by 2100, and it is ofthe same order of magnitude, although smaller, than climate-carbon feedbacks. Uncertainties on the magnitude of this land-use induced effect are large, because of uncertainties in the sink role of terrestrial ecosystems in the future and because of uncertainties inherent to the modeling of land-use induced carbon emissions. Such an extra rise in atmospheric CO2 is however partially offset by the ocean reservoir and by sinks operating over undisturbed, pristine ecosystems, suggesting that conserving pristine forests with long turnover times might be efficient in mitigating the greenhouse effect