Complementarity as a mechanism of coexistence between functional groups of grasses


  • N GROSS,

    Corresponding author
    1. Laboratoire d’Ecologie Alpine (LECA), UMR 5553 CNRS – Université Joseph Fourier, BP 53, F-38041 Grenoble, France,
    2. Station Alpine Joseph Fourier (SAJF), UMS 2579 CNRS – Université Joseph Fourier, BP 53, F-38041 Grenoble, France,
      *Author to whom correspondence should be addressed: Nicolas Gross. Tel.: +33 4 76 63 54 38. Fax: +33 4 76 51 46 73. E-mail:
    Search for more papers by this author
  • K. N SUDING,

    1. Department of Ecology and Evolutionary Biology – University of California Irvine, Irvine, CA, 92697–2525, USA, and
    Search for more papers by this author

    1. Laboratoire d’Ecologie Alpine (LECA), UMR 5553 CNRS – Université Joseph Fourier, BP 53, F-38041 Grenoble, France,
    2. Station Alpine Joseph Fourier (SAJF), UMS 2579 CNRS – Université Joseph Fourier, BP 53, F-38041 Grenoble, France,
    Search for more papers by this author

    1. Centre d’Ecologie Fonctionnelle et Evolutive (CEFE), UMR 5175, CNRS – 1919, Route de Mende, 34293 Montpellier, Cedex 5, France
    Search for more papers by this author

*Author to whom correspondence should be addressed: Nicolas Gross. Tel.: +33 4 76 63 54 38. Fax: +33 4 76 51 46 73. E-mail:


  • 1Increasing functional diversity often leads to an increase in ecosystem productivity in the form of overyielding. While the mechanisms (i.e. complementarity or facilitation) that underlie overyielding provide strong insights into species coexistence and community assembly, they are rarely tested. In subalpine grasslands, traditional management through manuring and hay-making results in intermediate productivity that is associated with high functional diversity. This functional diversity results from the coexistence between conservative plant species (with slow growth rates, low specific leaf area) and exploitative species (with fast growth rates, high specific leaf area).
  • 2We hypothesized that overyielding occurs among these two functional groups and tested whether complementarity or facilitation can explain overyielding. Using three perennial grass species per functional group, we compared single and mixed functional group mesocosms at low and intermediate levels of fertilization to test the occurrence of overyielding. Additionally, we measured the outcomes of biotic interactions among these two functional groups by manipulating plant density.
  • 3After two growing seasons, we found evidence of overyielding under intermediate levels of fertility. Overyielding was associated with a reduction of competition intensity when both functional groups were grown together. These results suggest that complementarity, as evidenced by a decrease in competition intensity, rather than facilitation, explains the observed overyielding. Indeed, we found evidence for complementarity for light and modification of nutrient use as possible mechanisms for the overyielding.
  • 4Synthesis. Complementarity between functional groups might be an important mechanism enhancing functional diversity, particularly in harsh environments at intermediate rather than low fertility.


In the past decade, many experiments have shown that increasing functional diversity can lead to an increase in ecosystem productivity, usually termed overyielding (Tilman et al. 1997; Hector et al. 1999; see Hooper et al. 2005 for review). In addition to sampling effects (Huston 1997; Loreau 1998), overyielding can be caused by increasing functional complementarity and/or facilitation (Hooper et al. 2005). If species are able to use different resources, or if they can use the same resource but at different times or in different locations, then complementarity can increase overall resource utilization (Berendse 1982; Sala et al. 1989; Naeem et al. 1994). Similarly, if some species ameliorate harsh conditions and increase resource availability for other groups of species, then facilitation can enhance ecosystem productivity (Mulder et al. 2001; Hooper & Dukes 2004).

While biodiversity experiments have allowed rapid progress in our understanding of the role of functional diversity in community structure (Fargione et al. 2003) and ecosystem functioning (Tilman et al. 1997), the results of these experiments have been strongly debated (Huston 1997; Loreau et al. 2001; Hooper et al. 2005). Most cases of overyielding have been related to effects of one particular plant functional group, nitrogen-fixers (e.g. Tilman et al. 1997; Hooper 1998; Hector et al. 1999). It is unclear whether overyielding may also occur between other functional groups (but see van Ruijven & Berendse 2003, 2005). Furthermore, although facilitation and complementarity are the most likely mechanisms contributing to overyielding, the design of most biodiversity experiments is not suited to properly test which of these mechanisms is the primary driver of overyielding (Huston 1997; Hooper et al. 2005). Finally, biodiversity experiments are not designed to test how biodiversity effects change along fertility gradients. For instance, it is unclear how nutrient availability affects mechanisms of overyielding (Fridley 2002, 2003).

Plant diversity often shows a hump-shaped response to productivity (Mittelbach et al. 2001) with decreased diversity associated with increased productivity in benign environments (Rajaniemi 2003) and an opposite relation in harsh environments (Gross et al. 2000; Suding et al. 2005). This response is typical in moderately productive European subalpine grasslands where traditional management combining fertilization and mowing has increased productivity as well as species or functional diversity (Tasser & Tappeiner 2002; Quétier et al. 2007). This high functional diversity results from the coexistence between conservative species (characterized by a slow growth rate and low specific leaf area; Diaz et al. 2004; Wright et al. 2004) and exploitative species (characterized by a fast growth rate and high specific leaf area) (Quétier et al. 2007).

Although conceptual competition models predict that conservative species are excluded from more fertile sites by competition with exploitative species (Grime 1977; Wedin & Tilman 1993), fertilization may promote coexistence of these two functional groups by two distinct mechanisms. Fertilization may increase the facilitative effect of vegetation by increasing the size of plants (Mulder et al. 2001; Callaway et al. 2002). Additionally, fertilization may limit competition for soil resources (Wedin & Tilman 1993) and promote complementarity for light (Fridley 2002, 2003; Kahmen et al. 2006).

In this study, we used a pot experiment to test whether overyielding occurred between conservative and exploitive grasses that coexist at subalpine hay meadows of intermediate productivity. We hypothesize that increasing fertility promotes overyielding in this system by (i) increasing facilitation through increased biomass or (ii) promoting complementarity in resource use.


study site

The experiment was located at the experimental garden of the Station Alpine Joseph Fourier, Lautaret Pass, central French Alps (Villar d’Arêne, 45.04° N, 6.34° E, elevation 2100 m). The climate is subalpine with a pronounced continental influence. Mean annual precipitation is 956 mm and mean monthly temperatures range between –7.4 °C in February and 19.5 °C in July The growing season starts after snowmelt, between mid-April and early May, and finishes at the end of September.

species and functional group definition

We studied two functional groups, in which we selected grass species that represented conservative and exploitative strategies. We selected three conservative and three exploitative perennial grass species, on the basis of their specific leaf area (SLA) and relative growth rate (RGR) measured under optimal (no resource limitation) conditions (see Gross et al. 2007; Appendix S2 in Supplementary Material). Exploitative species, Dactylis glomerata L., Agrostis capillaris (L.) P. De Beauvois and Poa alpina L. are characterized by a high relative growth rate and SLA. Conservative species, Festuca paniculata (L.) Schinz et Thellung, Sesleria caerulea (L.) Arduino and Bromus erectus (L.) are characterized by a low growth rate and SLA. In fertilized hay meadows at the study site, conservative species represent 45 ± 12% and exploitative species represent 55 ± 17% of total biomass. In unfertilized grasslands conservative species dominate, with 80 ± 8% total biomass (data from Quétier et al. 2007). In this study we focused on grass species, which make up from 50% to 80% of total cover of subalpine grasslands at our study site (Gross et al. 2007). Subalpine grasslands are exclusively dominated by perennial vegetation where recruitment events are rare (Zeiter et al. 2006). For this reason, we chose to focus on the adult stage and used tillers collected from the field for the experiment.

experimental design

We conducted a 2-year pot experiment where neighbour interactions and fertilization were manipulated in a factorial design (Fig. 1). The overall design consisted of nine planting schemes crossed with two levels of fertility replicated eight times in a randomized block design, for a total of 144 pots.

Figure 1.

Experimental design where density was varied to assess the importance of positive and negative interactions. In low-density treatments, one plant per pot was grown, and in high-density treatments six individuals were grown per pot. At high density, there were three types of mixtures: ‘intra’ indicates intrafunctional mixture, where each of the three species of a single functional group were grown with two individuals; ‘inter’ indicates interfunctional mixture, where functional groups were grown together with one individual per species. Symbols show individual species and their position in mixtures. Dark symbols are species from the conservative functional group and clear symbols indicate species from the exploitative functional group. Arrows highlight the statistical comparisons conducted in this study: comparison among high-density treatments tested for overyielding; comparison between low- and high-density treatments estimated the intensity and direction of plant–plant interactions (LNRR, see methods for details). All treatments were repeated under fertilized and unfertilized conditions.

Species were grown in pots either at low or high density. In the low-density treatment, the six species were grown individually. In the high-density treatment, six individuals were planted in a circle with 3-cm space between each individual. Two types of mixtures in which functional groups were grown either alone (intrafunctional mixture) or together (interfunctional mixture) are realized. Intra-functional mixtures were composed of a single functional group (either exploitative or conservative species); two individuals per species were used (Fig. 1), totalling six individuals per pot. Inter-functional mixtures included species from the two functional groups. The six species were planted in a random pattern with the constraint that species groups alternated with one another to promote interfunctional interactions. Thus, we planted species at low-density (six treatments) and high-density (two intrafunctional and one interfunctional), for a total of nine planting schemes.

The experiment was established between 28 June and 5 July 2003 by planting field-collected tillers of each species. Before planting, each tiller and roots were cut (3 cm and 5 cm length, respectively). Homogeneous tillers were planted following the design described in Fig. 1, in 15-L pots (33 cm diameter, 26.3 cm deep), filled with a soil composed of two-thirds sand, one-quarter calcinated clay and one-twelfth commercial potting compost (Fertiligène®). The pH of the soil used in the experiment was similar to soil pH measured in adjacent fields (pot pH, 7.2 ± 0.5; field pH = 7.0 ± 0.3; P > 0.05). Pots were placed outside in the experimental garden; they were moved regularly within and between blocks throughout the course of the experiment, thereby making the spatial design fully random. Although water limitation may occur in the field during the summer (N. Gross et al., unpublished data), we chose to water pots daily with an automatic system to avoid any confounding effect and tested only the effect of fertilization in our experiment. We assumed that pots were not water-limited. During winter, pots were buried to protect roots from frost.

Half of the pots were fertilized by adding 15 g year−1 of a commercial slow release fertilizer per pot (12% N, 12% P, 17% K, 2% Mg), which mimicked intermediate levels of fertility in fertilized grasslands. Previous studies have shown that in fertilized hay meadows, P does not limit plant growth and nitrogen availability is on average equal to 78 mg of mineral N per kg of soil (Tosca & Labroue 1986; Quétier et al. 2007; Robson et al. 2007). The nitrogen availability in the unfertilized pots (6.1 mg of mineral N per kg of soil) corresponded to the lowest level of nutrients measured in unfertilized grasslands at the study site (Quétier et al. 2007; Robson et al. 2007).

harvest and measurements

At the peak of biomass production during the second growing season (30 July 2004), plant height and light interception were measured for each species in each pot. Light interception was quantified at 2 cm above ground with a LI-190 Quantum Sensor (LI-COR®) under full sun between 11.00 and 14.00 for 20 random points per pot. We found no statistical differences in light interception between the fertilized mixture (71 ± 3%) and in the fertilized meadows at peak biomass (71 ± 1%; P > 0.05), measured in a previous study (Quétier et al. 2007). Five leaves per individual were randomly selected from the top of the plant canopy for subsequent chemical analysis. Leaves were dried at 60 °C for 72 h, ground and analysed with a CHN microanalyser (Carlo Erba 1500) to determine leaf nitrogen concentration for each species in each treatment.

In August 2004, at the end of the second growing season, all pots were harvested. Shoots and roots were washed carefully under water. Shoots were collected for each species whereas below-ground biomass was taken without separating roots by species. Shoots and roots were dried for 48 h at 60 °C and weighed. Root density in fertilized mixture pots (851 ± 70 g m−2) was comparable to that measured in the field (913 ± 137 g/m2; P > 0.05) (T. M. Robson et al., unpublished data). In interfunctional mixtures, root biomass of the exploitative functional group was determined using the near infra-red spectrometry (NIRS) method following Roumet et al. (2006) (see Appendix S3 for protocol and results of NIRS analysis). Root biomass of the conservative functional group was determined from the difference between total root biomass and predicted NIRS below-ground biomass of the exploitative functional group.

The total amount of nitrogen in leaf biomass for each species was determined by multiplying leaf nitrogen concentration (LNC) by total leaf biomass (Van Ruijven & Berendse 2005). We then determined the amount of leaf biomass produced per unit of nitrogen as proposed by van Ruijven & Berendse (2005) as an estimation of leaf nutrient use efficiency (LNUE).

overyielding calculations

Overyielding was assessed by comparing biomass of interfunctional mixtures with biomass of intrafunctional mixtures (Fig. 1). Two indices were used to address different aspects of overyielding. The first index, relative yield total (RYT), was calculated as follows:


where F is the total number of functional groups, and

RYi = Oi/Mi

where Oi is the biomass of functional group i in interfunctional mixture (three plants) and Mi is the intrafunctional biomass of i (six plants). RYT > 1 indicates overyielding. It is one of the most common metrics for assessing overyielding (Hooper 1998; Hooper & Dukes 2004). While this index characterizes overyielding at the mixture level, it does not address the specific response of functional groups and does not allow the rejection of the sampling effect (Hector 2006).

The second index Di estimated the proportional deviation of the observed biomass in interfunctional mixtures from its expected value in intrafunctional mixtures for each functional group (Loreau 1998):

Di = (Oi – piMi)/piMi

where piMi is the expected biomass of one functional group in intrafunctional mixture, where pi is the proportion of functional group i in interfunctional mixture (pi = 0.5 in this study).

Because different functional groups can be affected differently by changes in functional composition, Di quantifies the response of each functional group. When Di > 0 for all functional groups, there is transgressive overyielding (Loreau 1998), i.e. each functional group produces more biomass when grown with the other functional group than when grown alone. This provides a sufficient condition to unambiguously reject sampling effects (Loreau 1998). We calculated these indices with above-ground, below-ground and total biomass for each functional group.

outcomes of biotic interactions

Biotic interactions were quantified by comparing species of the different functional groups grown individually with species grown at high density in intra- or interfunctional mixtures (Fig. 1). We used a common competition index, the natural log response ratio for functional groups (LNRR). Because facilitation and competition operate simultaneously (Oksanen et al. 2006), this index measures the net outcome of biotic interactions (Suding et al. 2003):


where BM(i) with competition is the biomass of the functional group i (i.e. exploitative or conservative) and λ is the mixture type (intra- and interfunctional). In intrafunctional mixture BM(i) with competition is piMi. In interfunctional mixture BM(i) with competition is Oi. BM(i) estimated without competition is the sum of species biomass from functional group i produced in low density pots (Fig. 1). When LNRRinteraction < 0, net effects of neighbours are negative (interactions are dominated by competition) and when LNRRinteraction > 0, net effects of neighbours are positive (interactions are dominated by facilitation).

We also calculated indices at the species level to test whether species responses within a particular functional group were consistent with the group's aggregate response. This comparison was based only on responses of above-ground biomass because below-ground biomass could not be estimated for individual species.

statistical analysis

Statistical analyses were conducted using the software JMP 5.0.1 (SAS Institute Inc., Cary, NC, USA). First, analyses were performed at the functional group level. To test experimental treatment effects on biomass production per pot, light incidence, leaf nitrogen biomass and nutrient use efficiency, we conducted a set of full factorial anova type III testing for combined Functional Mixture (Mixt.), Fertilization (Nut.) and Functional Groups (FG) effects. We did not include in this analysis the low-density treatment. ‘Mixture’ compared biomass production of different functional groups in intra- vs. interfunctional mixture. The assessment of overyielding was made by testing if the RYT value differed from 1 and/or DT and Di differed from 0 using a Student's t-test. We conducted one-way anovas on LNRR interaction, LNUE and leaf nitrogen biomass to test the effect of the different functional mixtures at each nutrient level.

In a second set of analyses at the species level, we used the same set of anova type III as for functional groups to test whether species within functional groups (FG) had similar responses to experimental treatments (Mixt. and Nut.) in terms of above-ground biomass production, total leaves nitrogen biomass and nutrient use efficiency.


biomass production and overyielding

The interfunctional mixture was dominated by the exploitative functional group, comprising over 60% of the above-ground biomass in unfertilized conditions (low fertility treatment) (significant difference across FG, P < 0.05) and over 80% with added nutrients (intermediate fertility treatment) (significant difference across FG, P < 0.0001; Fig. 2b). With fertilization, exploitative species comprised 65% of total root biomass per pot (FG significantly different P < 0.05), whereas without fertilization it made up only 52% of the total root biomass (FG was not significantly different).

Figure 2.

(a, b) Root and shoot biomass, (c, d) above-ground nitrogen mass, and (e, f) leaf nitrogen use efficiency (LNUE) for each conservative and exploitative functional group grown in INTER and INTRA functional mixtures, and in unfertilized (a, c, e) or fertilized (b, d, f) conditions. INTRA is a species mixture composed of three conservative species (Cons.) or three exploitative species (Expl.); INTER is a species mixture composed of six both conservative and exploitative species. One-way anovaposthoc test was used to compare mixture effects on total biomass for each group in each nutrient treatment. Abbreviations: ns, not significant, *, P < 0.05, **, P < 0.001, ***, P < 0.0001.

The effect of functional mixture on biomass production was highly dependent on fertilization (Table 1, Fig. 2a, b). In the low fertility treatment (no fertilization), functional group biomass was not affected by the type of functional mixture (i.e. whether there were one or two functional groups; Fig. 2a), even though the biomass of intrafunctional mixtures was always greater than that of interfunctional mixtures. At intermediate levels of fertility (fertilization), the type of functional mixture affected biomass production (Table 1, Fig. 2b). Fertilization increased the biomass of both exploitative and conservative functional groups but this increase was greater in interfunctional mixture. Fertilization affected allocation patterns of the two functional groups differentially (Table 1). The exploitative functional group increased above-ground biomass (P < 0.01) in the interfunctional mixture, whereas below-ground biomass was unaffected. In contrast, the conservative functional group increased below-ground biomass (P < 0.01) while above-ground biomass was not affected in the interfunctional mixture.

Table 1.  Effect of experimental treatments analysed at the functional group level for biomass production, light incidence, amount of nitrogen in leaf biomass and nutrient use efficiency using full factorial anova
Effectd.f.Total BMShoot BMRoot BMN leavesLNUE
F ratioPF ratioPF ratioPF ratioPF ratioP
  1. FG, functional group; Mixt., functional mixture; Nut., fertilization. We indicated degrees of freedom (d.f.) and Fisher ratio (F ratio); NS, non-significant effect; *P < 0.05; **P < 0.01; ***P < 0.0001.

FG 1100.96***120.97*** 15.59** 23.08*** 0.00NS
Mixt. 1  6.07*  5.49*  2.26NS305.40*** 2.61NS
FG × Mixt. 1  1.10NS  9.02**  5.70*102.50*** 0.00NS
Nut. 1240.60***174.84***138.13***638.21***19.73***
FG × Nut. 1 72.73*** 93.94***  7.92**248.36*** 2.52NS
Mixt. × Nut. 1  9.99**  9.10**  3.64NS 45.73*** 0.61NS
FG × Mixt. × Nut. 1  0.90NS 10.12**  8.14** 93.69*** 7.12*

The RYT values of unfertilized plants were below 1, suggesting no overyielding (Table 2). Consistent with this result, Di values for the exploitative functional group were not different from zero. The Di value for the conservative functional group was slightly negative. With fertilization, total biomass was significantly lower in intra- than in interspecific mixtures (Table 1, Fig. 2b). We observed a high positive value of RYT (P < 0.0001) (Table 2). The two functional groups showed positive Di values with fertilization, indicating the occurrence of overyielding for these two groups. These results were often, but not always, consistent with the response of below-ground or above-ground biomass examined individually (Table 2). Positive Di for the conservative functional group was mainly driven by an increase in root biomass in interfunctional mixture. In contrast, the positive Di value for total biomass of the exploitative group was explained by an increase in shoot biomass. Significant positive Di values for both functional groups in the fertilized mixture indicate that the sampling effect can be rejected.

Table 2.  Indices for assessing the degree of overyielding calculated for above-ground, below-ground and total biomass with (1) and without (0) fertilization
 Total biomassAbove-ground biomassBelow-ground biomass
Fertilization Fertilization Fertilization 
  1. Overyielding occurs when RYT > 1 and Di > 0 for conservative (Cons.) and exploitative (Expl.) species. We conducted one-way anova to test for significant effects of fertilization on overyielding. Fertilization column: 0, no fertilization; 1, fertilization. An asterisk in the Fertilization column indicates whether the fertilization treatment significantly changed RYT and Di value. Additionally, we conducted a Student's t-test to compare RYT values to 1 and Di to 0: NS, P > 0.05; *P < 0.05; **P < 0.001; ***P < 0.0001.

RYT0**0.81 ± 0.12**0*0.78 ± 0.12NS0*0.87 ± 0.12NS
11.51 ± 0.07***11.25 ± 0.10**11.61 ± 0.14***
D cons.0***–0.24 ± 0.10*0NS–0.18 ± 0.15NS0***–0.29 ± 0.12*
10.55 ± 0.13*1–0.15 ± 0.20NS11.06 ± 0.31**
D expl.0*–0.13 ± 0.16NS0*–0.24 ± 0.16NS0 NS0.04 ± 0.18NS
10.34 ± 0.11*10.45 ± 0.17*1–0.03 ± 0.10NS

effects of functional mixture on biotic interactions

The net outcomes of biotic interactions were negative in this experiment (Fig. 3), indicating the prevalence of competition rather than facilitation. Indeed, individuals from the low-density treatment (individuals grown alone) always produced more biomass than plants in high-density treatments. Without fertilization, the type of mixture had no effect on the outcomes of biotic interactions (Fig. 3a). With fertilization, outcomes were less negative in the interfunctional mixture for both functional groups than in their respective intrafunctional mixtures (Fig. 3b).

Figure 3.

Competition indices for conservative (Cons.) and exploitative (Expl.) groups using natural log response ratio (LNRR) for intrafunctional (INTRA) and interfunctional competition (INTER) mixtures in (a) unfertilized and (b) fertilized conditions We used one-way anovapost hoc test to test the effect of mixtures on competition intensities for each group in each nutrient treatment. Abbreviations: NS, differences between INTRA and INTER are not significant, *P < 0.05.

effects of functional mixture on nutrient use efficiency

Whether functional groups grew alone or in the presence of the other functional group had a strong effect on the total amount of nitrogen in leaf biomass and on leaf nutrient use efficiency (LNUE) (Fig. 2). Without fertilization, both functional groups had lower leaf nitrogen biomass when grown together as compared with when grown alone, but this difference was stronger for the conservative functional group than for the exploitative functional group (Fig. 2c). Despite this difference, the LNUE was not significantly affected by the type of functional mixture (Fig. 2e). With fertilization, total leaf nitrogen biomass decreased for the conservative functional group, but increased for the exploitative functional group, when grown together as compared with when they were grown alone (Table 1, Fig. 2d). The type of functional mixture affected the nutrient use efficiency (LNUE) in opposite ways for the two functional groups (Table 1; Fig. 2F). The conservative functional group showed increased LNUE in interfunctional mixture, whereas the exploitative functional group had decreased LNUE in the presence of the conservative group as compared with when grown alone.

light interception and species height

Light availability was strongly modified by the type of functional mixture as well as fertilization (Mixture, F2,39 = 15.44, P < 0.0001; Nutrient, F1,39 = 122.52, P = 0.0001; Mixture × Nutrient, F2.39 = 7.51, P < 0.001). As expected, fertilization decreased light availability (P < 0.0001) (Appendix S1). Without fertilization the light availability was not strongly affected by the type of mixtures. In contrast, functional groups had contrasting effects on light availability when fertilized (Appendix S1). The exploitative functional group had a strong effect on light, intercepting more than 80% of PAR when grown alone. The conservative functional group had a weaker effect on light levels, with less than 40% of light interception when grown alone. When both groups were grown in mixture, the canopy intercepted over 80% of PAR.

Species grown in inter- or intrafunctional mixtures with fertilization strongly differed in plant height (Appendix S1), but these differences did not correspond to functional group designations. The exploitative D. glomerata, and the two conservative B. erectus and F. paniculata were significantly taller (c. 20 cm height) than P. alpina and A. capillaris (exploitative) and S. caerulea (conservative) (c. 7 cm) (P < 0.0001). There was no significant difference in plant height between intra- and interfunctional mixtures (Mixture, P-value not significant for any species), indicating no plant elongation in the interfunctional mixture.

analysis at species level

Similar patterns of responses were observed at the species and functional group levels (Appendix S2). Consistent with analyses at the functional group level using above-ground biomass data, species from the two functional groups responded differently to nutrient addition and to the type of functional mixture. We observed non-significant responses without fertilization when comparing competition intensity in intra- vs. interfunctional competition (Table S2). With fertilization, exploitative species experienced decreased competition intensity in inter- vs. intrafunctional mixture (P < 0.05 for D. glomerata and P. alpina, non-significant effect for A. capillaris, P = 0.15; Table S2). In contrast, the above-ground biomass of conservative species was not affected by interfunctional competition, with the exception of S. caerulea, for which competition intensity significantly increased in interfunctional mixture (P < 0.01, Table S2). Species responses were also strongly consistent with their functional groups response for leaf nitrogen biomass and leaf nutrient use efficiency (Appendix S2).


In this study, we showed that fertilization leading to intermediate levels of fertility promotes transgressive overyielding between conservative and exploitative grass functional groups, supporting the idea that overyielding can occur without the presence of legumes (van Ruijven & Berendse 2003, 2005). Although grass species are usually considered as a same functional group when classifications are based on life form (Hooper et al. 2005), they differ considerably in their traits and their responses to environmental factors (Diaz et al. 2004; Al Haj Khaled et al. 2005; Gross et al. 2007). Classification based on functional traits rather than simple growth form is critical when examining species coexistence or ecosystem processes (Diaz et al. 2004; McGill et al. 2006; Shipley et al. 2006; Wright et al. 2006).

complementarity as a mechanism of overyielding between grass species

Consistent with previous studies conducted at similar altitudes (e.g. Choler et al. 2001; Callaway et al. 2002), facilitation was not detected in this study and the outcomes of biotic interactions were primarily negative (Fig. 3). Functional composition affected biomass productions only with fertilization (Fig. 2b), with interactions becoming less negative in interfunctional mixtures for both exploitative and conservative functional groups (Fig. 3b). This result apparently contrasts with competition models (e.g. Grime 1977; Wedin & Tilman 1993) that predict exclusion of conservative species by exploitative species in high fertility conditions. However, fertilization led to an intermediate level of fertility in subalpine grasslands, allowing the coexistence between the two functional groups (Grime 1977; Quétier et al. 2007). This result is also supported by field observations in harsh environments where diversity does not decrease with fertilization (Gross et al. 2000; Suding et al. 2005).

The reduction of negative interactions in interfunctional mixture could be explained both by an increase in facilitation or a decrease in competition (Hooper et al. 2005). However, in our study, biomass production of plants growing alone (low-density treatments) was always greater than plants grown in mixtures (high-density treatments), indicating the overall importance of competitive interactions for the growth of established individuals. Additionally, two mechanisms of complementarity (for light and nitrogen) may act to promote a decrease in competition intensity between the two functional groups. For these reasons, overyielding in our experiment is most parsimoniously interpreted as a consequence of complementarity effects between functional groups, rather than facilitation.

Two mechanisms of complementarity are likely to have caused the overyielding between the two functional groups. First, differences in height among species in fertilized interfunctional mixtures might promote light partitioning (Fig. S1b) (Naeem et al. 1994; Fridley 2002, 2003). In a previous study (Gross et al. 2007), we found that grass species with different heights differed in their shade tolerance. Growth of short species like A. capillaris, P. alpina and S. caerulea was not affected by shade whereas tall species like D. glomerata and F. paniculata were shade intolerant (Gross et al. 2007). In our experiment, shade-intolerant species overtopped short shade-tolerant species. Thus, complementarity for light, promoted by above-ground space partitioning and differences in shade tolerance, may occur between functional groups within grasslands as it does within forests. Secondly, the decrease in leaf nitrogen biomass for conservative species was compensated for by an increase in leaf production per gram of nitrogen (LNUE) (Fig. 2f). This result confirms a previous study (van Ruijven & Berendse 2005) where an increase in LNUE was observed as functional diversity increased. The decrease in total nitrogen in leaf biomass and the increase of LNUE for the conservative functional group might be due to its larger allocation to root biomass in interfunctional as compared with intrafunctional mixture (Fig. 2b).

Without fertilization, we found no evidence that conservative species are better competitors than exploitative species (Ryser & Lambers 1995). Competition was likely due to below-ground interactions as no light depletion was detected. Additionally, nitrogen tissue content for the two functional groups decreased when grown together (Fig. 2d). Mechanisms that explain dominance patterns at unproductive sites may require longer periods than two growing seasons to be expressed. It is indeed not rare to find a shift of productivity and species abundance in long-term experiments (van Ruijven & Berendse 2005). Conservative species could ultimately dominate at low fertility sites because exploitative species are not nutrient stress-tolerant (Grime 1977). Alternatively, conservative species could build a high stature through time due to nutrient conservation (Aerts & Vanderpeijl 1993) and exclude exploitative species by competition for space (Elberse & Berendse 1993).

relevance of the functional group approach

In this study, responses at the functional group level were consistent with responses of species within their own group (Appendix S1), confirming the existence of two distinct functional strategies among the six grass species (Gross et al. 2007). Consistency between the species and functional group levels was even stronger for responses to interfunctional mixtures of total nitrogen in leaf biomass and LNUE. Responses of conservative and exploitative species tended to be opposing, highlighting the contrasting nutrient economies for conservative and exploitative species (Aerts & Vanderpeijl 1993). Our study supported the relevance of the functional groups approach to understanding species interactions and coexistence (Suding et al. 2003; McGill et al. 2006; Lavorel et al. 2007).

Idiosyncratic behaviour of a species within its functional group is in no way contradictory with the functional group approach, but rather provides additional insight into coexistence mechanisms. The behaviour of particular species within functional groups may inform us of the existence of other trade-offs linked to other sets of traits (Suding et al. 2003; Ackerly 2004; Grime 2006). For instance, the short-statured species S. caerulea showed an original response in interfunctional mixture with fertilization within the conservative group. Differences in plant height may reflect differences in competitive ability within this group (Gross et al. 2007). Within exploitative species, plant heights were linked with different shade tolerances (Gross et al. 2007) and may lead to light partitioning within the group.


This study showed overyielding between conservative and exploitative grasses from subalpine grasslands at intermediate rather than low level of fertility. Our results suggest that complementarity, resulting in a reduction in competitive intensity, is likely to explain this overyielding. Different mechanisms of complementarity may have occurred simultaneously in this study. Our results suggest that both light partitioning (Fridley 2002, 2003) and modification of leaf nutrient use efficiency (van Ruijven & Berendse 2005) may explain overyielding, species coexistence and resulting high functional diversity in fertilized subalpine grasslands. Although overyielding among grass species is likely to explain the high functional richness of subalpine grasslands, other mechanisms linked with water use strategy (N. Gross et al., unpublished data) or acting at the regeneration stage (Quétier et al. 2007) may also play important roles in subalpine grasslands. Future field studies are needed to quantify and understand the ecological role of complementarity, especially in harsh environments with intermediate fertility where diversity does not decrease with fertilization (Gross et al. 2000; Rajaniemi 2003; Suding et al. 2005; Quétier et al. 2007).


This study was supported by the GEOTRAITS project of the French ACI-ECCO programme and CNRS GDR 2574 Utiliterres. We thank M. Chausson, M. Enjalbal and C. Poillot for technical assistance during the experiment; G. Girard for chemical analysis; R. Hurstel, R. Douzet, S. Aubert and all the staff of the SAJF; F. Quétier and F. Grassein for light interception data, R. Joppre for NIRS analysis and T. M. Robson for roots data in the field; M. L. Navas and A. Bouasria for discussions; and I. Ashton, S. Harpole, P. Choler, P. Liancourt, H. Cornelissen and the two anonymous reviewers for their valuable comments during the preparation of the manuscript.