Regional climate

Rangeland productivity is governed largely by the effects of the spatial and temporal distribution of precipitation on soil water availability and the climatic effects on evaporative demand (Campbell et al. Reference Campbell, Stafford Smith and Mckeon1997; Morgan Reference Morgan, Reynolds and Frame2005; Fay et al. Reference Fay, Kaufman, Nippert, Carlisle and Harper2008; Heisler-White et al. Reference Heisler-White, Blair, Kelly, Harmoney and Knapp2009). Most rangelands, but especially those on which productivity is strongly water-limited, should be sensitive to the positive effects of CO₂ enrichment on soil water content and plant water relations. Unless stomatal closure is compensated by atmospheric or other feedbacks, such as an increase in leaf area, CO₂ enrichment should slow canopy-level water loss (Polley et al. Reference Polley, Johnson, Fay and Sanabria2008) and the rate or extent of soil water depletion (Owensby et al. Reference Owensby, Ham, Knapp, Bremer and Auen1997; Polley et al. Reference Polley, Johnson and Derner2002a; Nelson et al. Reference Nelson, Morgan, LeCain, Mosier, Milchunas and Parton2004) and increase leaf and plant water use efficiency (Polley et al. Reference Polley, Johnson and Derner2002a; Fay et al. Reference Fay, Kelley, Procter, Hui, Jin, Jackson, Johnson and Polley2009). Indeed, variability in water availability may be the single most important factor mediating the biomass response of rangeland vegetation to CO₂ (Morgan et al. Reference Morgan, Pataki, Körner, Clark, Del Grosso, Grünzweig, Knapp, Mosier, Newton, Niklaus, Nippert, Nowak, Parton, Polley and Shaw2004b).

Positive effects of CO₂ enrichment on productivity generally increase as precipitation and water availability decline, especially when water is saved at elevated CO₂ while plants are active. In general, then, the CO₂ stimulation of biomass should be proportionally greater or more consistent on semi-arid than mesic ecosystems and, for a given ecosystem, during dry than wet years (Morgan et al. Reference Morgan, Pataki, Körner, Clark, Del Grosso, Grünzweig, Knapp, Mosier, Newton, Niklaus, Nippert, Nowak, Parton, Polley and Shaw2004b). For example, CO₂ enrichment increased biomass production proportionally more on semi-arid shortgrass steppe (Morgan et al. Reference Morgan, Mosier, Milchunas, LeCain, Nelson and Parton2004a) than on mesic tallgrass prairie (Owensby et al. Reference Owensby, Ham, Knapp, Bremer and Auen1997) and calcareous grassland in Switzerland (Niklaus & Körner Reference Niklaus and Körner2004) and, for each ecosystem, during relatively dry than wetter years. Morgan et al. (Reference Morgan, Pataki, Körner, Clark, Del Grosso, Grünzweig, Knapp, Mosier, Newton, Niklaus, Nippert, Nowak, Parton, Polley and Shaw2004b) found that the ratio of aboveground biomass production at elevated CO₂ to production at ambient CO₂ (biomass enhancement ratio) increased across these three grasslands as growing season precipitation declined from c. 750 mm to 250 mm (Fig. 2).

Fig. 2.

The ratio of aboveground plant biomass at elevated compared to ambient CO₂ (biomass enhancement ratio) in Kansas tallgrass prairie (closed square), Colorado shortgrass steppe (closed circle) and Swiss calcareous grasslands (open diamond) as a function of yearly precipitation. The fitted line has an R ² of 0·48, P=0·005. The figure is adapted from Morgan et al. (Reference Morgan, Pataki, Körner, Clark, Del Grosso, Grünzweig, Knapp, Mosier, Newton, Niklaus, Nippert, Nowak, Parton, Polley and Shaw2004b).

Positive effects of CO₂ enrichment on productivity do not consistently extend to arid ecosystems, however. CO₂ enrichment increased plant biomass in the Mojave Desert in the western USA only during a wet year, when the total precipitation was 2·4 times the long-term average for the site (Smith et al. Reference Smith, Huxman, Zitzer, Charlet, Housman, Coleman, Fenstermaker, Seemann and Nowak2000). These findings imply that a minimum level of precipitation or soil water availability exists below which CO₂ has little influence on productivity (Morgan et al. Reference Morgan, Pataki, Körner, Clark, Del Grosso, Grünzweig, Knapp, Mosier, Newton, Niklaus, Nippert, Nowak, Parton, Polley and Shaw2004b; Nowak et al. Reference Nowak, Ellsworth and Smith2004). CO₂ effects are negligible during most years in arid environments partly because a significant fraction of water loss is not under plant control. Much of precipitation probably is lost to evaporation rather than transpiration in arid and semi-arid ecosystems because evaporative demand of the atmosphere is high, plant cover is sparse, vegetation often must recover from extended periods of drought in order to appropriate water for growth, and most precipitation events are small (e.g. Sala & Lauenroth Reference Sala and Lauenroth1982; Wythers et al. Reference Wythers, Lauenroth and Paruelo1999).

Soil resources

Low N availability frequently limits plant productivity on rangeland ecosystems (e.g. Seastedt et al. Reference Seastedt, Briggs and Gibson1991) and may reduce or even eliminate any benefit of CO₂ enrichment for plant growth (Owensby et al. Reference Owensby, Auen and Coyne1994; Reich et al. Reference Reich, Hobbie, Lee, Ellsworth, West, Tilman, Knops, Naeem and Trost2006). Productivity is probably limited more often by N than by insufficiencies in other elements, but limits on the availability of any essential element of plants could constrain CO₂ effects on productivity.

Soil water content is a critical determinant of plant productivity, both because of its direct effect on plant water status and activity and its indirect effect in stimulating decomposition, mineralization, nutrient cycling and mass flow of elements to roots (Burke et al. Reference Burke, Lauenroth and Parton1997; Dijkstra et al. Reference Dijkstra, Pendall, Mosier, King, Milchunas and Morgan2008). The water-relations benefit of CO₂ to plant growth may result in part from an increase in N mineralization rates in wetter soils (Morgan et al. Reference Morgan, Pataki, Körner, Clark, Del Grosso, Grünzweig, Knapp, Mosier, Newton, Niklaus, Nippert, Nowak, Parton, Polley and Shaw2004b; Dijkstra et al. Reference Dijkstra, Pendall, Mosier, King, Milchunas and Morgan2008).

Disturbance regime

Many rangeland ecosystems are sustained in their current states by disturbances, the two most influential of which are grazing and fire. Disturbances, like other interactive controls, regulate CO₂ responses both directly and indirectly, the latter by affecting other controlling factors. Grazing and fire directly modify plant growth and physiology by removing plant tissues. Grazers can also serve as a vector for seed dispersal in faecal deposits (Kramp et al. Reference Kramp, Ansley and Tunnell1998; Brown & Archer Reference Brown and Archer1999), with the potential for profound changes in plant community structure. Many of the other effects of these disturbances on ecosystem processes are indirect, mediated through changes in soil resource levels or the biota.

Grazers strongly influence plant productivity on rangelands, but our understanding of how defoliation regulates productivity in response to changes in CO₂ levels is limited. It might be expected that defoliation would enhance the effects of changes in CO₂ levels on plant growth and speed plant recovery from defoliation by stimulating photosynthesis, at least in C₃ species. Recent results confirm this expectation (D. J. Augustine, personal communication). Defoliated plants of Pascopyrum smithii recovered more rapidly when grown at elevated levels of CO₂ than at ambient levels. Photosynthesis rates were doubled in defoliated compared to non-defoliated plants grown at 780 ppmv CO₂, but not in the plants grown at 400 ppmv CO₂. Similarly, increasing CO₂ to 235 ppmv above ambient levels enhanced the production of vegetation growing in monoliths extracted from semi-natural grassland in Wales when plants were defoliated six times per year (Harmens et al. Reference Harmens, Williams, Peters, Bambrick, Hopkins and Ashenden2004). CO₂ treatment had no effect on production of plants defoliated just twice per year. In contrast, results from other studies do not support the view that defoliation magnifies CO₂ effects on plant growth. For example, re-growth following defoliation was no greater at elevated than at ambient levels of CO₂ for a C₃ forb (Fajer et al. Reference Fajer, Bowers and Bazzaz1991) or a C₄ grass (Wilsey et al. Reference Wilsey, McNaughton and Coleman1994). Plants of three grass species from each of three grassland ecosystems showed similar responses to CO₂ when clipped and not clipped (Wilsey et al. Reference Wilsey, Coleman and McNaughton1997), as did vegetation on shortgrass steppe (Milchunas et al. Reference Milchunas, Mosier, Morgan, LeCain, King and Nelson2005) and a New Zealand pasture subjected to sheep grazing (Newton et al. Reference Newton, Allard, Carran, Lieffering, Nösberger, Long, Norby, Stitt, Hendrey and Blum2006). Herbivory may even reduce plant sensitivity to CO₂. Clipping reduced aboveground biomass of Arabidopsis thaliana more at elevated than at ambient CO₂, for instance (Lau & Tiffin Reference Lau and Tiffin2009).

Grazing may indirectly influence CO₂ effects on productivity by contributing to seed dispersal, seedling recruitment and, ultimately, vegetation change. Elevated CO₂ has been found to enhance seedling establishment on rangeland (Morgan et al. Reference Morgan, Mosier, Milchunas, LeCain, Nelson and Parton2004a) and mesic pasture (Edwards et al. Reference Edwards, Clark and Newton2001), but evidence for an effect of grazers is limited because of the difficulty of incorporating large animals into manipulative CO₂ experiments. In a rare CO₂ experiment with animals, Newton et al. (Reference Newton, Allard, Carran, Lieffering, Nösberger, Long, Norby, Stitt, Hendrey and Blum2006) found that grazing by sheep enhanced aboveground production of two important Trifolium species in CO₂-enriched New Zealand pasture. Enhanced recruitment under grazing appears to have been an important factor in the CO₂ response of one of the species, T. subterraneum.

Fire indirectly affects CO₂ response via changes in soil resources and vegetation. Fires volatilize substantial quantities of N (Seastedt et al. Reference Seastedt, Briggs and Gibson1991), hence frequent burning may constrain ecosystem responses to CO₂ by reinforcing N limitations on plant growth. Fire also can affect ecosystem processes by changing plant abundances or composition. Henry et al. (Reference Henry, Chiariello, Vitousek, Mooney and Field2006) found that summer wildfire removed the suppressive effect of elevated CO₂ on primary productivity of annual grassland by increasing production of forb species. Conversely, spring burning favours C₄ grasses at the expense of C₃ forbs on tallgrass prairie (Howe Reference Howe1994), potentially diminishing biomass response to CO₂. By increasing plant growth, CO₂ enrichment may also lessen the time period during which woody seedlings or saplings are most susceptible to fire and enhance woody incursion into grasslands (Bond Reference Bond2008), with potential consequences for future CO₂ response.

Functional groups

Results from several studies have shown that the response of productivity to CO₂ is regulated by the number or identities of species or species groups present in plant assemblages (Leadley et al. Reference Leadley, Niklaus, Stocker and Körner1999; Niklaus et al. Reference Niklaus, Leadley, Schmid and Körner2001; Reich et al. Reference Reich, Tilman, Naeem, Ellsworth, Knops, Craine, Wedin and Trost2004). Increasing the number of species or functional groups of species enhanced positive effects of CO₂ enrichment on aboveground and total biomass of herbaceous communities (Reich et al. Reference Reich, Tilman, Naeem, Ellsworth, Knops, Craine, Wedin and Trost2004). Similarly, Niklaus et al. (Reference Niklaus, Leadley, Schmid and Körner2001) reported a positive correlation between the diversity of grassland communities and biomass enhancement at high CO₂. Biomass may respond more to CO₂ in diverse than in depauperate communities simply because diverse communities are more likely to contain responsive species (Niklaus et al. Reference Niklaus, Leadley, Schmid and Körner2001). Alternatively, increasing diversity may increase CO₂ effects on biomass by maximizing plant capture of N, light and other resources that are required for greater production (Reich et al. Reference Reich, Tilman, Naeem, Ellsworth, Knops, Craine, Wedin and Trost2004).

Summary – regulation of CO₂ response

Results from manipulative experiments demonstrate that three interactive controls, regional climate, soil resources and functional groups of plants, strongly regulate CO₂ effects on rangeland productivity. The effects of the disturbance regime on CO₂ responses are variable in direction and are often inconsequential. The benefits of CO₂ enrichment for productivity generally increase as precipitation (regional climate) declines and species diversity (functional groups) increases, and decline as N availability decreases (soil resources). Consequently, differences in regional climate, functional groups and soil resources among rangeland ecosystems and among years for a given ecosystem contribute to variability in CO₂ effects on productivity. However, we must ask whether these controls on the response of a given ecosystem operate indefinitely without variation, or whether changes in controls caused by alterations in levels of CO₂ might feed back to amplify or dampen ecosystem sensitivity to CO₂.

Regional climate

Elevated CO₂ can lead to changes in canopy-level and local climatic factors that reduce water savings from a decrease in stomata conductance and, hence, the benefits of CO₂ enrichment for water-limited ecosystems. One of these negative feedbacks involves stomatal effects on leaf temperature. Partial stomatal closure reduces the transpiration rate and latent heat flux, leading to a rise in leaf temperature (Idso et al. Reference Idso, Kimball, Akin and Kridler1993; Kimball et al. Reference Kimball, Pinter, Garcia, LaMorte, Wall, Hunsaker, Wechsung, Wechsung and Kartschall1995). The saturation vapour pressure of air increases as temperature rises, causing an increase in the leaf-to-air vapour pressure gradient. This increase in the driving gradient for water loss tends to offset effects of reduced conductance on transpiration. Higher canopy temperatures and reduced transpiration contribute to a second negative feedback on plant regulation of transpiration. The vapour pressure deficit of air within and immediately above vegetation depends partly on transpiration. Slower transpiration tends to dry air in the canopy boundary layer, further increasing the vapour pressure gradient for transpiration. Bunce et al. (Reference Bunce, Wilson and Carlson1997) simulated these and other processes that regulate transpiration with a model calibrated using crop species. Simulations indicated that elevated CO₂ increased leaf temperature and the vapour pressure difference between leaves and air. This feedback, together with other feedbacks operative at regional scales, almost completely negated effects of a 20–60% decrease in canopy conductance on water loss. Negative feedbacks also have been shown to attenuate stomatal effects on rangeland transpiration. Elevated CO₂ often reduces leaf conductance to 20–50% of that measured at ambient CO₂, but rarely reduces evapotranspiration on grasslands by greater than 20% (Morgan et al. Reference Morgan, Pataki, Körner, Clark, Del Grosso, Grünzweig, Knapp, Mosier, Newton, Niklaus, Nippert, Nowak, Parton, Polley and Shaw2004b). Elevated CO₂ also may reduce the tolerance of photosynthesis to acute heat stress, especially among C₄ species and plants grown at supra-optimal temperatures (Hamilton et al. Reference Hamilton, Heckathorn, Joshi, Wang and Barua2008).

Any process that reduces the influence of plants on the amount of soil water that is returned to the atmosphere or the rate at which water is depleted from soil will limit the response of rangeland productivity to CO₂. These processes include those regulated by other interactive controls, such as soils and disturbance regimes. Drainage of water below the rooting zone of plants represents a negative feedback on biomass response to CO₂ that depends, in turn, on soil texture. Deep drainage almost doubled at elevated compared to pre-industrial CO₂ in model grassland assemblages of Israel during the wet part of the growing season (Grünzweig & Körner Reference Grünzweig and Körner2001). Similarly, CO₂ enrichment increased water storage at 0·75–1·05 m depth in semi-arid shortgrass steppe (Nelson et al. Reference Nelson, Morgan, LeCain, Mosier, Milchunas and Parton2004). Disturbances also can reduce plant regulation of water loss by reducing the cover of plants relative to bare soil.

Soil resources

Most terrestrial N occurs in organic forms that are not readily available to plants, hence rangeland responses to CO₂ depend partly on how quickly N cycles between organic and inorganic N compounds and whether changes associated with CO₂ enrichment accelerate or slow N cycling. The Progressive N Limitation (PNL) hypothesis holds that changes associated with CO₂ enrichment will create or reinforce N limitations on production by reducing N recycling to inorganic forms (Luo et al. Reference Luo, Su, Currie, Dukes, Finzi, Hartwig, Hungate, McMurtrie, Oren, Parton, Pataki, Shaw, Zak and Field2004). According to the PNL hypothesis, CO₂ enrichment may lead to N limitation of productivity even in ecosystems in which N availability does not currently prevent a growth response to CO₂. The PNL hypothesis arises from the observation that organic matter, whether in plants or soil, contains N and other elements in rather fixed ratios to C. To the extent that CO₂ enrichment increases plant production and C accumulation in ecosystems, it also increases the sequestration of N and other elements in organic matter (Fig. 3). As more N is sequestered over time, the availability of mineral N to plants could decline and limit CO₂ effects on productivity. It is the progressive decrease in N availability to plants, rather than the initial amount of available N, that distinguishes the PNL hypothesis. Consistent with the PNL hypothesis, two differently configured ecosystem models, G'DAY and DAYCENT, predict that CO₂-stimulation of net ecosystem production on tallgrass prairie and shortgrass steppe will decline over a 100-year period because of a decline in soil N availability (Pepper et al. Reference Pepper, Del Grosso, McMurtire and Parton2005).

Fig. 3.

Possible feedback effects of an increase in plant growth at elevated CO₂ on ecosystem N pools and N availability to plants. By increasing plant production, higher CO₂ increases N sequestration in plant biomass, litter pools and eventually SOM. CO₂ enrichment may increase or decrease the net rate at which N in SOM is mineralized to plant-available forms depending on whether microbial activity is ‘primed’ or inhibited by the extra C input to soil at elevated CO₂. Increasing CO₂ also may increase the pool of available and total N by stimulating N₂ fixation, reducing gaseous or liquid losses of N, or increasing retention of atmospheric N deposition.

Elevated CO₂ has been found to reduce soil N availability (Reich et al. Reference Reich, Hobbie, Lee, Ellsworth, West, Tilman, Knops, Naeem and Trost2006), but results from most experiments indicate that various processes operate at elevated CO₂ to reduce the strength of this negative feedback by either delaying the onset of N limitation or enhancing N accumulation in soil/plant systems. Nitrogen limitation may be delayed by maximizing the amount of C sequestered per unit of N, that is by maximizing the C:N ratio of organic matter. Organic matter accumulation is increased, for example, by transferring N from soil organic pools with relatively low C:N ratios to plants with higher C:N ratios, as demonstrated recently on mesic grassland (Gill et al. Reference Gill, Polley, Johnson, Anderson, Maherali and Jackson2002, Reference Gill, Anderson, Polley, Johnson and Jackson2006) and semi-arid shortgrass steppe (Pendall et al. Reference Pendall, Del Grosso, King, LeCain, Milchunas, Morgan, Moiser, Ojima, Parton, Tans and White2003). On mesic grassland, microbes mineralized resident soil organic matter (SOM) to meet their nutritional needs at high CO₂. The N released by decomposition of SOM contributed to an increase in total N content of plant tissues (Gill et al. Reference Gill, Anderson, Polley, Johnson and Jackson2006) and sustained an increase in plant productivity at high CO₂ (Polley et al. Reference Polley, Johnson and Derner2003). In the absence of additional N required to stabilize organic C in soil pools, however, much of the extra organic C that was added to soil at elevated CO₂ was dissipated through rapidly cycling soil pools and did not accumulate (Gill et al. Reference Gill, Polley, Johnson, Anderson, Maherali and Jackson2002). Decomposition rates of resident SOM more than doubled at elevated CO₂ on shortgrass steppe in Colorado (Pendall et al. Reference Pendall, Del Grosso, King, LeCain, Milchunas, Morgan, Moiser, Ojima, Parton, Tans and White2003). The N released by more rapid decomposition of SOM reduced N limitations on growth and contributed to an increase in productivity at elevated CO₂ (Dijkstra et al. Reference Dijkstra, Pendall, Mosier, King, Milchunas and Morgan2008).

Various other processes may increase N accumulation in soil/plant systems and contribute to a positive feedback on CO₂ responses. These processes include increased biological fixation of N, greater retention of atmospheric N deposition, reduced losses of N in gaseous or liquid forms, and more complete exploration of soil by expanded root systems (Luo et al. Reference Luo, Su, Currie, Dukes, Finzi, Hartwig, Hungate, McMurtrie, Oren, Parton, Pataki, Shaw, Zak and Field2004; Fig. 3). Perhaps the most rapid and direct manner by which CO₂ enrichment could increase N accumulation on rangelands is by stimulating fixation of atmospheric N₂ by the legume-Rhizobium symbiosis. The absolute amount of plant-available N that can be derived from additional N₂ fixation at elevated CO₂ is large. For example, enrichment to 600 ppmv CO₂ was found to increase the annual N yield of mixed grass/legume swards by 9–13% (Zanetti et al. Reference Zanetti, Hartwig, van Kessel, Lüscher, Hebeisen, Frehner, Fischer, Hendrey, Blum and Nösberger1997). All additional N at elevated CO₂ was derived from enhanced N₂ fixation. Similarly, increasing CO₂ by 80–150% above the ambient concentration increased N₂ fixation by the woody legume Acacia smallii Isely (huisache) by a factor of 3–4 (Polley et al. Reference Polley, Johnson and Mayeux1997). Elevated CO₂ stimulated N₂ fixation by increasing both the mass of root modules per plant and fixation per unit of nodule mass. In the absence of large N input from legumes, the N concentration of leaves or aboveground tissues declined on shortgrass steppe, tallgrass prairie, and mesic grassland at elevated CO₂ and on tallgrass prairie with warming, but total N content of aboveground tissues increased as plant biomass increased on these ecosystems and on annual grasslands (Owensby et al. Reference Owensby, Coyne and Auen1993; Hungate et al. Reference Hungate, Chapin, Zhong, Holland and Field1997; King et al. Reference King, Mosier, Morgan, LeCain, Milchunas and Parton2004; Wan et al. Reference Wan, Hui, Wallace and Luo2005; Gill et al. Reference Gill, Anderson, Polley, Johnson and Jackson2006; Dijkstra et al. Reference Dijkstra, Pendall, Mosier, King, Milchunas and Morgan2008). The implication is that plants accessed more, rather than less, N at elevated CO₂, thereby delaying the onset of N limitation to productivity. Pepper et al. (Reference Pepper, Del Grosso, McMurtire and Parton2005) point out that a modest and gradual increase in soil N may be sufficient to overcome PNL in the long term.

Disturbance regime

CO₂ enrichment may change the behaviour of grazers and possibly the intensity of grazing by altering forage quality (Polley et al. Reference Polley, Morgan, Campbell, Stafford Smith, Reddy and Hodges2000), with potential feedback effects on plant responses to CO₂. Plant N concentration usually declines at elevated CO₂ (Owensby et al. Reference Owensby, Coyne and Auen1993; Cotrufo et al. Reference Cotrufo, Ineson and Scott1998; Morgan et al. Reference Morgan, Mosier, Milchunas, LeCain, Nelson and Parton2004a), implying that CO₂ enrichment will reduce the crude protein content of forage. Conversely, elevated CO₂ may improve forage quality by increasing tissue concentrations of non-structural carbohydrates (Read et al. Reference Read, Morgan, Chatterton and Harrison1997). On shortgrass steppe, CO₂ enrichment reduced the crude protein concentration of autumn forage below critical maintenance levels for livestock in 3 out of 4 years and reduced the digestibility of forage by 14% in mid-season (Milchunas et al. Reference Milchunas, Mosier, Morgan, LeCain, King and Nelson2005). Significantly, the grass most favoured by CO₂ enrichment also had the lowest crude protein concentration. In addition, the unpalatable shrub Artemisia frigida exhibited the largest biomass increase of all plant species, a 40-fold increase during 5 years at elevated CO₂ (Morgan et al. Reference Morgan, Milchunas, LeCain, West and Mosier2007). The magnitude and even sign of the feedback of changes in forage quality on CO₂ responses will probably depend on whether changes in forage quality result in an increase or decrease in grazing intensity.

Increased CO₂ could increase fire frequency or intensity (Sage Reference Sage, Körner and Bazzaz1996) and thereby reinforce the effects of fire on ecosystem processes by increasing plant production, fuel load, or fuel flammability. The flammability of plants should increase if, as often observed, CO₂ enrichment increases the accumulation of non-structural carbohydrates in plant tissues (Read et al. Reference Read, Morgan, Chatterton and Harrison1997). Conversely, rising CO₂ could reduce the intensity or frequency of fires by delaying soil water depletion during dry periods (Morgan et al. Reference Morgan, Pataki, Körner, Clark, Del Grosso, Grünzweig, Knapp, Mosier, Newton, Niklaus, Nippert, Nowak, Parton, Polley and Shaw2004b) and increasing the water content of fuel or the ratio of live to dead tissues in plant stands. This suppressive effect of CO₂ on fire will probably be most important for rangelands on which dry intervals are relatively brief (Sage Reference Sage, Körner and Bazzaz1996). A delay of days to weeks in canopy senescence would probably have little effect on fire in rangelands which regularly experience extended periods of drought. Plant growth appears to be particularly responsive to CO₂ among several fire-adapted annual grasses, including Bromus tectorum and Avena barbata (Smith et al. Reference Smith, Strain and Sharkey1987; Jackson et al. Reference Jackson, Sala, Field and Mooney1994; Ziska et al. Reference Ziska, Reeves and Blank2005), evidence that CO₂ enrichment can promote a positive feedback on productivity by enhancing flammability or fuel load. On the other hand, more frequent fires may dampen CO₂ effects on productivity by reducing N availability to plants (Seastedt et al. Reference Seastedt, Briggs and Gibson1991) or increasing the fraction of water that is lost to evaporation.

Functional groups

CO₂ enrichment has been shown to alter plant abundances, and thus the potential response of productivity to CO₂, in most experiments with multi-species communities (Körner et al. Reference Körner, Diemer, Schäppi, Niklaus and Arnone1997; Niklaus et al. Reference Niklaus, Leadley, Schmid and Körner2001; Polley et al. Reference Polley, Johnson and Derner2003; Morgan et al. Reference Morgan, Milchunas, LeCain, West and Mosier2007). It might be predicted that CO₂ enrichment would increase the size or abundance of legumes and favour potentially more responsive C₃ over C₄ species, but CO₂ effects vary among ecosystems. Legumes were more abundant at elevated than ambient CO₂ in New Zealand pasture and Mediterranean grasslands (Edwards et al. Reference Edwards, Clark and Newton2001; Joel et al. Reference Joel, Chapin, Chiariello, Thayer and Field2001), but as a group responded little to CO₂ in calcareous grassland (Leadley et al. Reference Leadley, Niklaus, Stocker and Körner1999). Similarly, as predicted by differences in photosynthetic pathway (Polley Reference Polley1997), elevated CO₂ enhanced productivity of C₃ over C₄ grasses in shortgrass steppe (Morgan et al. Reference Morgan, Milchunas, LeCain, West and Mosier2007), but did not benefit C₃ grasses on C₄-dominated tallgrass prairie (Owensby et al. Reference Owensby, Ham, Knapp and Auen1999). Several factors likely contribute to inconsistencies in CO₂ effects on vegetation. First, the life stage of plants most affected by CO₂ differs among experiments. In relatively closed canopies of perennial species, species change probably results mostly from differences in the growth responses of established plants to CO₂. Over longer time periods or in other ecosystems, vegetation change results partly from differences in the responses of seed production (Edwards et al. Reference Edwards, Clark and Newton2001) or seedling recruitment and survivorship among species to CO₂ (Smith et al. Reference Smith, Huxman, Zitzer, Charlet, Housman, Coleman, Fenstermaker, Seemann and Nowak2000; Morgan et al. Reference Morgan, Mosier, Milchunas, LeCain, Nelson and Parton2004a). Secondly, it is likely that CO₂ effects on vegetation differ among ecosystems because factors that limit seedling establishment and plant growth differ among ecosystems or among years for a given ecosystem (Belote et al. Reference Belote, Weltzin and Norby2003). Species change in water-limited systems is often linked to CO₂ effects on soil water content (Morgan et al. Reference Morgan, Pataki, Körner, Clark, Del Grosso, Grünzweig, Knapp, Mosier, Newton, Niklaus, Nippert, Nowak, Parton, Polley and Shaw2004b). CO₂ enrichment changed species abundances in ecosystems as diverse as Swiss grassland and semi-arid shortgrass steppe by slowing soil water depletion and preferentially increasing seedling recruitment of certain species (Niklaus et al. Reference Niklaus, Leadley, Schmid and Körner2001; Morgan et al. Reference Morgan, Mosier, Milchunas, LeCain, Nelson and Parton2004a). In contrast, CO₂ had little effect on species abundances in C₄-dominated tallgrass prairie in Kansas, USA (Owensby et al. Reference Owensby, Ham, Knapp and Auen1999), presumably because the growth of the shorter C₃ species was limited by low light or N availability, or C₃ plants were incapable of exploiting the mid- to late-season improvement in soil water that occurred at elevated CO₂.

If the plants that benefit most from CO₂ enrichment are dominant species or become important contributors to communities, species change should increase the positive effect of higher CO₂ on productivity (e.g. Stöcklin et al. Reference Stöcklin, Schweizer and Körner1998). The link between the change in relative abundance of a species and its growth response to CO₂ should be strongest in communities that establish following disturbances and when plants of all species are of similar size. On the other hand, species change that is caused by factors other than CO₂ may reduce CO₂ effects on productivity. Elevating CO₂ increased biomass of newly planted grassland communities by increasing the growth of ruderal species (Niklaus et al. Reference Niklaus, Leadley, Schmid and Körner2001). The response of community biomass to CO₂ decreased over time, apparently because N availability declined following planting and favoured species that were more stress-tolerant but less CO₂-responsive than ruderals.

Unfortunately, there is little evidence from which to infer the impact of species feedbacks on productivity and other ecosystem processes. Biomass of C₃ species increased at the expense of C₄ grass biomass in mesic grassland, irrespective of CO₂ treatment (Mielnick et al. Reference Mielnick, Dugas, Johnson, Polley and Sanabria2001; Polley et al. Reference Polley, Johnson and Derner2003). This shift towards C₃ plants, physiologically more responsive than C₄ plants (Anderson et al. Reference Anderson, Maherali, Johnson, Polley and Jackson2001; Maherali et al. Reference Maherali, Reid, Polley, Johnson and Jackson2002), would be expected to increase plant growth response to CO₂ enrichment. Indeed, a temporary increase in CO₂ of 150–200 ppmv increased net ecosystem uptake of CO₂ more in C₃- than C₄-dominated vegetation (Fig. 4). The proportional increase in CO₂ uptake increased from 0 to 26% as the contribution of C₃ cover to total cover increased from 0·26 to 0·96 (Polley et al. Reference Polley, Dugas, Mielnick and Johnson2007). Species interactions also proved an especially strong mediator of the response of annual grassland to changes in the seasonality of precipitation (Suttle et al. Reference Suttle, Thomsen and Power2007). Feedbacks resulting from species shifts reversed the initial responses of productivity and plant and food web diversity to precipitation change within a few years.

Fig. 4.

The response of net ecosystem CO₂ exchange (NEE) of grassland to a 1–2-day change in CO₂ concentration of 175 ppmv. Mesic grassland was exposed for 3 years to a continuous gradient in CO₂ spanning sub-ambient to elevated concentrations in elongated chambers. Daytime NEE was measured continuously in each of 18 plots, 5 m long and 1 m wide, along the gradient. The amount by which NEE of each plot was greater at elevated than sub-ambient CO₂ following brief change in CO₂ (E-S) increased linearly as the ratio of C₃ plant cover to total cover increased. The graph is reproduced from Polley et al. (Reference Polley, Dugas, Mielnick and Johnson2007).

To the extent that species change reinforces CO₂ effects on productivity, the current composition or relative abundances of species may be considered to constrain CO₂ responses. There are at least two consequences of this constraint. First, the full response of ecosystem processes to CO₂ may not be evident in the short-term. Secondly, factors that reduce the species pool, limit the dispersal of propagules, or reduce plant establishment could limit CO₂ effects on productivity.