Studies of plant–plant interactions at elevated atmospheric CO2 concentration, designed to link individual performance to the response of the entire community (Bazzaz & McConnaughay 1992), have used one of two experimental frameworks. In most studies, the response to CO2 of individual species has been evaluated for mixtures and for either monocultures or plants grown individually, then compared to the community response to CO2 (e.g. Bazzaz & Garbutt 1988; Williams, Garbutt & Bazzaz 1988). However, the response of multi-specific communities to CO2 cannot be scaled up from individual species because the effect of competition on species response to CO2 tends to be unpredictable (Körner 1995). Other studies have characterized the change in competitive relationships between species owing to elevated CO2 and its influence at the community level (e.g. Zangerl & Bazzaz 1984; Arnone & Körner 1995). Species biomass generally responded to CO2, with large interspecific differences in magnitude and direction of response, leading some authors to assert that community structure changes in elevated CO2 (e.g. Zangerl & Bazzaz 1984; Leadley & Stöcklin 1996). However, these studies did not test for the significance of interspecific differences or community structural responses to CO2 levels.
The aim of this paper is to clarify further the links between the performance of individual species and community response to elevated CO2. First, we reanalysed data from 20 experiments that compared the response to CO2 of species grown either as single plants or single species stands to that of the same species in mixed stands. Published data suggest that community response to CO2 is more dependent on those species with a low responsiveness to CO2 and a high initial biomass proportion in the mixture, than on highly responsive species with low initial biomass proportions (e.g. Schäppi & Körner 1996). Our hypothesis therefore was that the response to CO2 of a multi-species community should be predicted both from the responses of individual species to CO2 and from the species biomass in the initial mixture. Then, we proposed a method to test for significant community structural responses to elevated CO2, in relation to concomitant changes in species dominance and to significant interspecific differences in response to CO2.
Materials and methods
Twenty studies (Table 1) that document the response of biomass (19 studies) or Leaf Area Index (Teughels et al. 1995) of single plants, monocultures or species mixtures under ambient (300–390 p.p.m.) and elevated CO2 (600–700 p.p.m.) were selected. The species mixtures comprised from two to seven species. All the studies used to compare the structural responses of species mixtures included at least four species. Experimental growth conditions differed: plant densities varied according to the growth form of species and habitats were characterized by a range of resource availability, especially when studying the interaction between CO2 concentration and another resource. Species occurred at the same density in mixtures, except for undisturbed microcosms (Clark et al. 1997) or when mimicking densities recorded in natura (e.g. Leadley & Stöcklin 1996). Duration of experiments varied from 24 to more than 1000 days; this range was used to assess the influence of time on the results.
Table 1. . Summary of studies reviewed. Listed for each study are the number of species, atmospheric CO2 concentration, duration of experiment, plant density, level of resource in the habitat, types of communities (Com: I, isolated plants; M, monocultures; X, mixtures) and reference
PREDICTION OF MIXTURE RESPONSES TO CO2
Data from 13 experiments comparing mixtures and either monocultures or single plants (Table 1) were used to predict the responses of mixtures to CO2 from constituent species responses. For each experiment and for mixtures, monocultures or single plants, the response of each species to elevated CO2 was calculated as the ratio of its biomass recorded under elevated CO2 to that under ambient CO2. The species were grouped into classes according to this response ratio: species were classified as ‘stimulated’ if they had a response ratio higher than 1·10, ‘indifferent’ if it fell between 0·91 and 1·10 and ‘depressed’ if it was lower than 0·90. Species distributions were compared between mixtures and single species (either monocultures or single plants), using χ2-tests. Percentages of species changing classes between mixtures and single species were calculated.
The observed response to CO2 of the biomass of a mixture formed of j species, Sx, was calculated as:
where Bje and Bja were the biomass of a species j in mixture under elevated and ambient CO2, respectively. The estimated response of a mixture to CO2 was calculated, estimating what the individual biomass of constituent species should be in the mixture under elevated CO2 if species responses to CO2 were not changed by competition. The estimated biomass of a species j under elevated CO2 was the product of its biomass under ambient CO2 and its response to CO2, recorded from monoculture or isolated plants. Then, Sc was:
where Sj was the response to CO2 of a species j in monoculture or from isolated plants. Linear regressions were separately fitted to the relationships linking Sx and Sc, calculated from monocultures or isolated plants.
STRUCTURAL RESPONSES OF SPECIES MIXTURES TO CO2
Data from 11 experiments comparing mixtures grown under elevated and ambient CO2 (Table 1) were re-analysed to test for structural changes of these communities in response to elevated CO2. For each experiment and each CO2 level, species were ranked according to their biomass in mixture. The Spearman rank correlation coefficient was used to test for the significance of a change in species ranking between CO2 levels. The null hypothesis was that species ranking in the two CO2 levels were arrayed at random with respect to each other, i.e. that the Spearman coefficient would be zero. Rejection of the null hypothesis was interpreted as a lack of response of species hierarchy to CO2. When species hierarchy did not differ with CO2 levels, linear regressions were fitted to the relationships linking the species biomass at elevated and ambient CO2. The slope of a significant regression represented the average response of the mixture to elevated CO2. The value of the intercept varied with the range of interspecific differences in responses to CO2; its significance was tested using Student’s-t. An intercept not significantly different from zero meant that species did not significantly differ in their response to CO2. It was concluded that the structure of mixtures did not significantly differ with CO2 levels when species ranking did not respond to CO2 and linear regressions of species biomass had an intercept of zero and a non-zero slope.
PREDICTION OF MIXTURE RESPONSES TO CO2
Average species responses to elevated CO2 were similar whether grown individually (in monocultures or as single plants) or in species mixtures. This was the case (1Fig. 1a) for studies where individual species were grown in monocultures (responses to elevated CO2 of 1·22) and in mixtures (response = 1·29). It was also true (1Fig. 1b) for species grown as single plants (CO2 responses of 1·39) or as mixtures (response = 1·37). The difference between the two sets of data in average response of species grown in mixture is a result of the high stimulation by CO2 of Mediterranean legumes and tropical trees (1Fig. 1b: Reekie & Bazzaz 1989; M.-L. Navas & Y. Durand, unpublished data). The proportions of ‘stimulated’, ‘indifferent’ and ‘depressed’ species did not differ between monocultures or isolated plants and mixtures. However, 39% of species in monocultures and 60% of isolated plants changed CO2 response class in mixtures (Fig. 1).
Responses of species mixtures to CO2 could be predicted from constituent species responses recorded in either monocultures or isolated plants and from the biomass of constituent species recorded in mixture under ambient CO2 (comparison of monocultures and mixtures: r = 0·863, P < 0·01, 1Fig. 1c; isolated plants and mixtures: r = 0·978, P < 0·01, 1Fig. 1d). The slope of the regression calculated between the responses of mixtures observed and estimated from monocultures was lower than 1·0 because, in two studies, the dominant species in mixture under ambient CO2 were depressed under elevated CO2 (1Fig. 1c: Carter & Peterson 1983; Teughels et al. 1995). The regression was still significant excluding data of Wong & Osmond (1991) (1Fig. 1c) characterized by a higher stimulation by CO2 than other experiments. The lower values of Sx than Sc when calculated from responses of plants grown individually, suggested a weak depressing influence of interspecific competition on the response of communities to elevated CO2 (1Fig. 1d).
STRUCTURAL RESPONSES OF SPECIES MIXTURES TO CO2
Species were similarly ranked in both CO2 levels and showed a similar pattern of response to CO2 in 15 of the 21 mixtures reviewed (Table 2). This lack of significant structural response of mixtures to elevated CO2 was independent of the duration of studies or of the productivity of experimental habitats as four sets of mixtures had been submitted to elevated CO2 for more than 200 days and 10 had been cultivated under highly productive conditions (Tables 1 and 2). Significant structural responses were reported for six mixtures; they were also independent of the duration of studies and productivity of habitats.
Table 2. . Effect of elevated CO2 on mixture structure. Ranking of species forming mixtures is compared between CO2 levels (rS, the Spearman rank correlation coefficient is used for testing the null hypothesis that species ranking in the two samples are arrayed at random with respect to each other, i.e. that rS is zero). Equations of significant regressions linking Be and Ba, the biomass of species in mixtures measured under elevated and ambient CO2, respectively, and associated r2 and t (Student’s-t value for testing the null hypothesis that the regression line intercept equals zero) are given with level of significance. NS, non significant; (*) P < 0·10; *P < 0·05; **P < 0·01; ***P < 0·001. When communities have been grown under different resource levels, increasing resource availability is indicated with ascending numbers. References of studies are in ascending order of study duration: this was less than 100 days for studies 1–5 and more than 300 days for studies 9–11. 1, Bazzaz & Carlson (1984); 2, Zangerl & Bazzaz (1984); 3, Williams et al. (1988); 4, M.-L. Navas & Y. Durand, unpublished data; 5, Bazzaz & Garbutt (1988); 6, Reekie & Bazzaz (1989); 7, Leadley & Stöcklin (1996); 8, Lüscher et al. (1996); 9, Clark et al. (1997); 10, Arnone & Körner (1995); 11, Chiariello & Field (1996)
Competition largely altered the response of plant species to CO2, inducing a change in direction of response to CO2. This was so for 40% of species grown in monoculture then compared with mixtures and 60% of species grown as isolated plants then in mixtures (1Fig. 1a,b). These changes were unpredictable as they vary for a species between habitats of contrasted structure and resource availability (Bazzaz & McConnaughay 1992), and they have been related to an enhanced intensity of asymmetric competition when CO2 is elevated (Ackerly & Bazzaz 1995). Competition is asymmetric for mono-directional resources such as light when large plants suppress the growth of small plants more than they themselves are suppressed, with a greater effect than would be expected considering their relative sizes (Weiner & Solbrig 1984; Weiner & Thomas 1986). Asymmetric competition should be more intense under elevated CO2 than under ambient conditions because of enhanced competition for light. However, this hypothesis was not confirmed for mono-specific stands of birch (Wayne & Bazzaz 1997) and should be tested for multi-specific communities.
Such unpredictability of plant response to CO2 in competitive environments explains why the response of a community to CO2 cannot be scaled-up from species responses. However, the responses to CO2 of multi-specific communities can be successfully predicted from both component species responses to CO2 evaluated without interspecific competition and the component species biomass in initial mixtures (1Fig. 1c,d). This result suggests a similar influence of species responsiveness to CO2 and performance in competitive environment on the response of a community to elevated CO2.
Lower values of the observed responses of mixtures to CO2 than of their estimated responses calculated from plants grown individually, indicate a depressing effect of interspecific competition on the response of mixtures to CO2 (1Fig. 1d), as previously hypothesized by Bazzaz (1990). The influence of plant density on community response to elevated CO2 has been seldom studied (Du Cloux et al. 1987; Firbank et al. 1995; Schenk et al. 1995; Wayne & Bazzaz 1995; Retuerto, Rochefort & Woodward 1996) and never for multi-specific communities. Species are generally more stimulated by elevated CO2 when grown individually or at very low density than when grown in dense monocultures (but see Lolium perenne in Schenk et al. 1995), probably in relation to increasing light limitation by self-shading in dense stands (Schenk et al. 1995). This limitation of plant response to CO2 should be lower in multi-specific mixtures than in monocultures of same density because space is occupied more efficiently by leaves from several species than it is from one species (Teughels et al. 1995).
This study also shows that the structure of most mixtures does not respond significantly to elevated CO2, as species dominance does not change with CO2 levels in relation to lack of interspecific differences in response to CO2 (Table 2). This discrepancy with the previous assertion of a large effect of CO2 on the structure of mixtures (e.g. Williams et al. 1988; Drake 1992; Arnone & Körner 1995) can be explained in some cases. First of all, most of these studies focused on species biomass production without relating it to the biomass produced by the whole mixture. If the responses to CO2 of species comprising a mixture are similar, their proportions in the mixture will not be changed and the structure of the community will not differ between CO2 levels (e.g. data from Williams et al. 1988). If species do differ in response to CO2, their individual influence at the community level will depend also upon their biomass proportion, as already suggested. For example, undisturbed alpine grasslands did not change in structure after 3 years of growth under elevated CO2, because the only responsive species occurred at very low proportion (Schäppi & Körner 1996). This effect is emphasized in mixtures containing species occurring at varied proportions (e.g. Arnone & Körner 1995; Lüscher et al. 1996). However, the significance of the structural responses to CO2 of mixtures comprising less than four species cannot be tested using this framework; it can only be assumed when large interspecific differences in response to CO2 are recorded (e.g. data from Drake 1992; Owensby, Auen & Coyne 1994).
In some cases slight, even non-significant, changes in mixture structure under elevated CO2 can be of importance. Bazzaz et al. (1992) showed that small, non-significant changes in seed production of individuals under elevated CO2 could be significant at the population level when self-thinning differed between CO2 levels. As interspecific differences in response to CO2 of self-thinning and/or survivorship have been recorded (Woodward, Thompson & McKee 1991; Morse & Bazzaz 1994; Navas et al. 1997), slight interspecific differences in biomass response to CO2 would be magnified after some generations, i.e. in communities such as perennial grasslands or tree communities.
In conclusion, this review documents the major influence of the initial structure of multi-specific communities on their responses to elevated CO2, even in the long term and for experiments performed in high-productive habitats. This suggests that the structure of communities should not be changed dramatically by elevated CO2, except in the case of mixtures formed of a very low number of species occurring in similar proportions. However, more subtle changes of multi-specific communities, such as alterations in canopy structure (Reekie & Bazzaz 1989) or foliage composition (Navas et al. 1995), have been documented under elevated CO2. More precise analyses would be needed to evaluate their consequences on the functioning of communities under elevated CO2.
This work was initiated while I was on sabbatical leave at CSIRO, Division of Wildlife and Ecology, Canberra (funded by the ‘Ministère de l’Agriculture’ of France). The idea came from talks with M. Austin. Preliminary results were discussed at the Anglo-French 2nd Alliance Meeting held at Castleton, UK (funded by the French Foreign Office). I thank E. Garnier, M. Austin, C. Roumet, R. Hunt and P. Leadley for their comments and remarks, C. Smith and M. Lonsdale for revising a first version of the text very carefully. Unpublished data by M.-L. Navas and Y. Durand were obtained with the help of J. Peyre and J. Richarte (funded by the ‘Ministère de l’Agriculture’ of France, DGER no. 94/121).
Corresponding address: UFR de Biologie et Pathologie Vegetale ENSA-M, Place Viala, 34060 Montpellier Cedex 1, France.