There have been many experimental studies to evaluate the response of vegetation to the effect of increases in the partial pressure of carbon dioxide in the atmosphere (pCO(2)) that are expected to occur during the next century. This knowledge is important for the future protection of food supplies, for understanding changes in natural ecosystems and for quantifying the role of terrestrial plants in regulating the rate of change of pCO(2) and resulting changes in the global climate. Most of our knowledge about these effects has derived from experimental studies that have used open-top or closed chambers. These methods are subject to "chamber effects" caused by differences in energy balance and water relations that may significantly modify the response of vegetation to elevated pCO(2). The small plot sizes imposed by these techniques add other limitations both to interpretation of results and scope of investigations. Free-air carbon dioxide enrichment (FACE) provides an experimental technique for studying the effects of elevated PCO2 on vegetation and other ecosystem components in large unenclosed plots (> 20 m diameter). FACE avoids many modifications to the microclimate imposed by chamber methods and therefore provides some of the most reliable estimates of plant response to elevated pCO(2). Control of pCO(2) in large-scale FACE experiments has now been developed to an extent where performance is similar to that achieved with sophisticated closed-chamber facilities. Experience has shown that, when FACE facilities are fully utilised, the cost per unit of usable ground area enriched with CO2 is significantly lower than alternative methods. The large scale of FACE plots can support a range of integrated studies on the same material, thereby achieving a more complete analysis than has been possible with other methods of elevating pCO(2). This review considers the technical aspects of FACE methodology, outlines the major FACE experiments and summarises the advances in understanding of pCO(2) effects on ecosystems that it has allowed. Published data on large-scale FACE experiments with adequate plot replication are limited to experiments on four crop/vegetation types at three locations. FACE has been used for experiments on two crops, cotton and wheat, at Maricopa, Arizona, and on grassland species, principally ryegrass and clover, at Eschikon, Switzerland. The method has also been adapted for the first study of mature forest trees, loblolly pine at Duke Forest, North Carolina. A number of other large-scale FACE experiments are in progress and the method has been adapted for use in much smaller experimental plots. The results of the major FACE experiments represent important advances over understanding obtained from previous pCO(2) treatment methods. Most significant in terms of the global climate and atmosphere system are the clear observations with cotton and wheat crops that elevated pCO(2) increases the ratio of sensible : latent heat transfer and causes daytime warming of the surface vegetation. This results from decreased water use and loss, and has been evident at a range of scales. The scale of FACE plots has allowed quantitative and detailed studies of the dynamics of below-ground production and C accumulation in a range of systems, and all have shown surprisingly large increases. Of particular note are the increases observed in grassland grown with low N, where there was no response of the above-ground biomass, but an increased rate of turnover of leaves and input of surface litter. FACE has allowed cultivation of crops at a scale appropriate to agronomic trials and shown statistically significant increases in the yields of wheat, cotton and pasture crops, although some of these increases are less than suggested by chamber experiments. Against expectations, the FACE experiments at Maricopa have shown a greater relative increase in yield in crops grown under water shortage than in water-sufficient crops. Acclimatory loss of photosynthetic capacity has been widely anticipated to offset the increase in photosynthesis that follows initial transfer of vegetation to elevated pCO(2). None of the FACE experiments provides any evidence of such a loss; however, changes which will allow a re-optimisation of N distribution within plants have been reported. FACE methods have now been demonstrated to be feasible and effective within a range of crops and vegetation types. The information from past experiments has greatly improved our understanding of the impacts of global atmospheric change on terrestrial ecosystems.
