The possibility of trading greenhouse gas emission permits as a result of the Kyoto Protocol has spurred interest in developing land-based sinks for greenhouse gases. Extensive grazing lands that have the potential to develop substantial woody biomass are one obvious candidate for such activities. However, such activities need to consider the possible impacts on existing grazing and the possible impacts of continuing CO2 buildup in the atmosphere and resultant climate change. We used simulation models to investigate these issues in the mulga (Acacia aneura) woodlands of southwest Queensland. The simulation results suggest that this system can be managed to act as either a net source or a net sink of greenhouse gases under current climate and CO2 and under a range of global change scenarios. The key component in determining source or sink status is the management of the woody mulga. The most effective means of permanently increasing carbon stores and hence reducing net emissions is to exclude both burning and grazing. There are combinations of management regimes, such as excluding fire with light grazing, which, on average, allows productive grazing but transient carbon storage. The effects of increased CO2 on ecosystem carbon stores were unexpected. Carbon stores increased (7-17%) with doubling of CO2 only in those simulations where burning did not occur, but decreased when burnt. This occurred because the substantial increases in grass growth with doubling of CO2 (34-56%) enabled more fires, killing off the establishing cohorts needed to ensure continued carbon accumulation. On average, the doubling of atmospheric CO2 concentration increased grass growth by 44%, which is identical with mean literature values, suggesting that this result may be applicable in other ecosystems where fire has a similar function. A sensitivity analysis of the CO2 response of mulga showed only minor impacts. We discuss additional uncertainties and shortcomings.