Hierarchy of root functional trait values and plasticity drive early-stage competition for water and phosphorus among grasses

Authors


Summary

  1. The link between species' functional traits and competitive abilities has been described as a major factor structuring plant communities. However, two diverging hypotheses have been proposed to explain this process: competition-trait similarity and competition-trait hierarchy.

  2. We performed a greenhouse experiment to determine whether grasses' root foraging strategies, from acquisitive or conservative functional groups, are linked to plant competitive ability and to test which hypothesis better explains interactions during the early stage of grass establishment under contrasting growth conditions.

  3. Two grass species of each functional group were grown with and without a neighbour under two levels of water and phosphorus supplies. Three functional traits related to plant competitive ability were measured on all plants grown without neighbours: specific root length (SRL), root phosphorus use efficiency and root length density. Above-ground biomass was measured on plants grown with and without neighbours to evaluate the intensity of plant interaction.

  4. We demonstrated that for the three traits the intensity of interaction is driven mainly by hierarchical trait distance, that is, trait distance between target and neighbour, and not by trait similarity. Growth conditions strongly affected the significance of the relation between hierarchical distances and competition intensity. For the SRL hierarchical distance, this effect may be due to the most competitive species (with high SRL) being strongly impacted by water shortage, which modified the competitive hierarchy. Trait plasticity in response to stresses also appeared an important factor influencing the competitive ability of species, that is, species with the most plastic SRL in response to P stress were also the most competitive under P stress.

  5. A strong hierarchy exists among grasses' competitive abilities in non-limiting growth conditions that is linked to their root functional traits and investment in the root system. Consequently, our results support the trait hierarchy hypothesis in its ability to describe competitive interaction among grasses during early stages of establishment.

  6. Our study provides evidence that root functional hierarchical trait distance and plasticity explain how grasses interact with their neighbours. This distance enables species to be ranked according to their competitive ability; however, this ranking may be influenced by the growth conditions and traits considered.

Introduction

Understanding the processes that structure natural communities is one of the oldest pursuits in ecological research (Diamond 1975). Among the processes proposed to explain the presence of species within a community are their abilities to migrate into the community (i.e. dispersal ability; Hubbell 2001) and to tolerate abiotic (Weiher & Keddy 1999) and biotic (Kraft & Ackerly 2010) environments. These processes are described as filters that select, from a regional species pool, species that are able to migrate into the community, then select those that can survive in the abiotic environment and finally select those most resistant to biotic interaction (Lortie, Brooker & Choler 2004). Competition for resources is currently recognized as a large part of the filtering effect in plant communities (Grace & Tilman 1990; Keddy et al. 2002; Kraft & Ackerly 2010; Wang et al. 2010). Most studies focusing on competition impacts on community structure have been based on a common assumption: species with similar ecological strategies, generally closely related phylogenetically, compete more intensely for resources than species with different strategies (Darwin 1859). Since Macarthur & Levins (1967), it has been assumed that species with similar strategies have similar ecological niches and thus are less likely to coexist (i.e. the ‘competition-relatedness’ hypothesis; Cahill et al. 2008).

Functional traits have been commonly used to describe plant strategies (Grime et al. 1997; Díaz et al. 2004) and to test or develop hypotheses about the impact of competition on community structure (Grime et al. 1997; Cahill et al. 2008; Mayfield & Levine 2010). As reported by Kunstler et al. (2012), combination of the competition-relatedness hypothesis with the functional trait approach leads to the ‘competition-trait similarity’ hypothesis, which predicts that competitive interaction will increase with increasing trait similarity in interacting plants (Cahill et al. 2008). Under this hypothesis, competitive interactions lead to an increase in trait divergence within a community by excluding the less competitive species having equivalent trait values. However, the presence of species with similar trait values in communities (Mayfield & Levine 2010) suggests that when these traits are related to competitive ability (i.e. ability to acquire and use common limiting resources; Westoby et al. 2002), and in the absence of niche segregation (Chesson 2000), competition is likely to be stronger between species differing in functional traits. In this case, the competitive interaction between species is positively related to the hierarchical difference in their trait attributes (i.e. the trait values they display). Kunstler et al. (2012) named this hypothesis the ‘competition-trait hierarchy’ hypothesis. In this case, less competitive species are strongly impacted by other species to the point that they are ultimately excluded from the environment, leading to trait convergence at the community level (Mayfield & Levine 2010). We hypothesize that better understanding of the link between plant functional traits and competitive ability can help distinguish between the ‘competition-trait similarity’ and ‘competition-trait hierarchy’ hypotheses, which lead to opposite patterns in plant community structuring.

To discuss the validity of these two hypotheses, we propose testing them using root functional traits. Traits commonly used in competition studies are above-ground functional traits (Gross et al. 2007; Keddy & Nielsen 2009; Kunstler et al. 2012), for example leaf functional traits, giving priority to the competition for light in well-established communities. Root functional traits have received less attention, even though roots are responsible for water and nutrient acquisition, which is supposed to be the main process driving early-stage competitive interactions (Belcher, Keddy & Twolan-Strutt 1995). In the few studies using root characteristics to explain plant interactions, the focus largely rested on selective root placement, that is, a species' ability to preferentially place roots in relatively nutrient-rich patches (Hodge et al. 1999; Mommer et al. 2011). However, recent work highlights that root functional traits tell us more about plant competitive ability than root placement does; specific root length (SRL), which is the root length produced per unit of dry root mass, appears to be the functional trait best related to plant resource-acquisition capacity (Mommer et al. 2011) and respiration rate (Tjoelker et al. 2005). Plants with high SRL also have high nutrient acquisition and growth rates (Comas & Eissenstat 2004). As a consequence, they may compete for soil resources better than plants with low SRL (Mommer et al. 2011). Root length density (RLD), which is the length of roots present per unit of soil volume, is a functional trait related to a plant's ability to explore the soil and locate and acquire resources. Modelling approaches have shown that plants with high RLD can acquire resources at a higher rate, are more competitive and exclude other species from the habitat (Raynaud & Leadley 2005). In well-established European grasslands, plants can be ranked along a gradient between two opposing functional groups: an acquisitive one with high SRL and low RLD and a conservative one with low SRL and high RLD (Craine et al. 2002; Fort, Jouany & Cruz 2012). Results from Mommer et al. (2011) do not support this opposition between RLD and SRL values, perhaps because plants in their study were grown for only 4 months. We hypothesize that during this short period slow-growing species (low SRL) could not produce higher RLD than fast-growing species (high SRL), which were not impacted by their higher root turnover (McCormack et al. 2012). On the basis of the literature, we hypothesized that fast-growing acquisitive species were competitive species well-adapted to non-stressful habitats, while conservative ones were stress-tolerant species adapted to face stressful habitats but were poor competitors.

Species identity is not the only source of functional trait variability; environment may also influence plant trait attributes and simultaneously influence plant competitive abilities. This is why Kunstler et al. (2012) recommend also considering intraspecific variability of traits when analysing the link between plant trait attributes and plant competitive ability, particularly if this variability is linked to growth conditions. Additionally, trait variability may cause plant-strategy differentiation within particular habitats and hence be responsible for niche differentiation, determining whether species coexist within a habitat or not (Chesson 2000). Root system plasticity also appeared an important factor determining plant competitive abilities. Superior competitors are supposed to have more plastic root systems, which allows rapid projection of roots into soil nutrient patches (Callaway, Pennings & Richards 2003). Moreover, Padilla et al. (2013) and Gastal, Dawson & Thornton (2010) showed that high root trait plasticity allowed species to withstand water and nitrogen stress better. However, little is known about the consequences of root trait plasticity on competitive abilities despite its importance in plant nutrition (Callaway, Pennings & Richards 2003).

As a result, we evaluated the extent to which environment changes species' functional trait attributes and the link between plant functional trait attributes, their variation in response to stress and the type and intensity of plant interactions. To do so, we grew four grass species in a greenhouse under a combination of two levels of P- and water availability (four growth conditions). These two factors were chosen because they are, in addition to N, the three main factors limiting plant growth (White & Brown 2010). Moreover, P availability highly depends on soil water content because P has low mobility in soil, and diffusion in soil solution is the major mechanism governing its transfer between soil and roots (Stroia, Morel & Jouany 2007).

To distinguish between the competition-trait similarity and competition-trait hierarchy hypotheses, we focused on root functional traits describing the soil colonization strategy (i.e. RLD) and investment by root length (i.e. SRL). To analyse plant P use in root growth, we propose a trait related to plants' abilities to optimize P use for resource acquisition : root P use efficiency (RPUE), expressed as the root length produced per unit mass of P. To determine which competition hypothesis better describes early-stage interactions between grasses, our objectives were (i) to test whether the type and intensity of interaction between grass species are related to their root functional trait distance (absolute and hierarchical), trait plasticity or phylogenetic distance and (ii) to evaluate whether growth conditions influence trait distances and the relation between root trait distance or phylogenetic distance and the intensity of interaction between a pair of grass species.

Materials and methods

Plant species

We chose species from the Poaceae family to avoid the effects of greatly differing root tissue structure (dicots vs. monocots) and/or functioning (N-fixing vs. non-fixing species) on functional trait values. Four grass species were selected (Bromus erectus, Dactylis glomerata, Holcus lanatus and Lolium arundinaceum) that belong to two functional groups. B. erectus and L. arundinaceum are considered conservative species, while D. glomerata and H. lanatus, with an opposite root trait syndrome, are considered acquisitive species (Fort, Jouany & Cruz 2012). Seeds of each species were collected from natural populations in Ercé, a French Pyrenean valley (42°50′N, 1°17′E; elevation 500–800 m). Plants were initially grown in a common garden in Auzeville, south-western France (43°31′N; 1°30′E; elevation 150 m), beginning in autumn 2004. To set up the experiment, tillers of the four species were collected from this common garden.

Experimental design

The experiment ran from 15th February to 6th June 2011. One individual of each species was grown: (i) individually in a monoculture (a single tiller was planted in the centre of each pot; four combinations); (ii) in an intragroup interaction (two species, both from the same functional group; four combinations); (iii) in an intergroup interaction (two species, one from each functional group; four combinations). For interaction treatments, the two tillers were positioned 2 cm from the centre of pots in opposite directions along the diameter. This design resulted in a set of ten different combinations, which was replicated four times within blocks (10 × 4 = 40 pots). Each set of replicates was subjected to one of four growth conditions resulting from the combination of two water and two P treatment levels: high water and high P availability, high water- and low P availability, low water- and high P availability and low water- and low P availability. To monitor functional traits and the change in plant interactions over time, we harvested five blocks three times at 1-month intervals. The first harvest occurred 4 weeks after differing water treatments began (50 days after planting), the second and third harvests occurred 86 and 113 days after planting, respectively. As a result, the experiment began with 960 plants grown in 600 pots (40 × 5 × 3).

Growth conditions

Plants were grown in 10-cm diameter × 1-m-deep pots, each containing 10 kg (dry weight basis) of soil. The soil used was a 1 : 1 mixture of sand and calcareous clay soil collected in a field near the common garden. This soil, which had not been fertilized during the previous 40 years, contained low available P and provided little P nutrition to the plants. The Olsen P value (Olsen et al. 1954) measured in the soil (7 mg P kg sol−1) lay far below the minimum threshold of non-limiting P availability (Morel et al. 2000).

All pots were maintained at 20 °C during the day and 17 °C during the night, and the day : night ratio was 16 h : 8 h. At the start of the experiment, each pot was fertilized with 5 g of N (NH4NO3). Later, N was applied monthly (1·24 g N pot−1) to prevent N shortage. Only pots of the high P-availability treatment were amended with commercial triple super phosphate (45% P2O5) to provide high P availability to plants (2 g P pot−1). To control the watering of all pots, soil moisture was monitored with 30 sensors (EC-5 DECAGON) randomly placed in 15 pots of each water-treatment level. To ensure adequate implantation of tillers, the water content of all pots was maintained close to field capacity by applying at least 15 mL of water twice a day during the first 3 weeks of the experiment. Afterwards, half of the pots were still watered twice a day until the soil moisture reached field capacity, and the amount was increased to compensate for water uptake by plants; it reached 110 mL per day by the end of the experiment. These pots constituted the high water-availability treatment. The water content of the other half of the pots was maintained close to the wilting point by watering when sensors indicated soil moisture below wilting point; a mean of 12 mL of water per day was applied during the entire experiment. These pots formed the low water-availability treatment.

Harvests and measurements

The duration of each harvest did not exceed 4 days to limit differences in growth time between harvested plants. At each harvest, shoots of all plants were clipped at the root base, dried at least 48 h at 60 °C and weighed to determine biomass. At the same time, monoculture pots were opened lengthwise and roots were carefully washed of soil particles under water and frozen (−18 °C) to conserve them. Before scanning (at 400 dpi with an Epson Perfection 4990 scanner), roots were defrosted in cold water, and a representative root subsample of each plant (more than 25% of the total root system) was stained with methylene blue (1%) for at least 12 h to improve the contrast. Scans were analysed using WinrhizoTM software (Regent Instrument Inc., Sainte-Foy, QC, Canada) to measure the total subsample root length. After scanning, the root subsample and the rest of the root system were dried at least 48 h at 60 °C and weighed. For the second and third harvests, non-stained roots were also dried, then ground to measure root P concentration and calculate root P content (Van Veldhoven & Mannaerts 1987). From these data, we calculated SRL (m g−1 = subsample root length/subsample root mass), RPUE (m mg−1 = subsample root length/subsample P content) and RLD (cm cm−3 = (SRL × 10 × total root mass)/total soil volume in the pot). We used total soil volume to calculate the RLD because roots of all plants reached the bottom of the pot before the first harvest.

Species pair distances

For each species pair, the hierarchical trait attribute distance was calculated as Tt−Tn, where Tt is the trait attribute of the target species in the competition and Tn the trait attribute of its neighbour. Since each plant in the pair could be considered as both target and neighbour, two hierarchical trait attribute distances were calculated per pair, using each plant as the target. This led to the calculation of twelve hierarchical trait attribute distances (two for each interaction treatment). These 12 distances were calculated for the four growth conditions for each of the three harvests to consider trait attribute change with environment and time. Absolute trait attribute distance was the absolute value of the hierarchical trait attribute distance of the two plants in interaction. Similarly, we also calculated for the three stressful growth conditions (low P- and high water availability, low water- and high P availability, low P- and low water availability) species trait plasticity as ((Ts−Tns)/Tns× 100, where Ts is the trait attribute of a species grown in a stressful condition and Tns is the trait attribute of the same species in the non-stressful growth condition (high P- and water availability). Afterwards, this allowed us to calculate trait plasticity distance, which is the absolute value of the trait plasticity of the target plant minus the trait plasticity of its neighbour. To calculate the phylogenetic distance, we developed a species-resolved phylogeny for the four species studied by combining published sequence data of the matK gene (Schaefer et al. 2011; Bruni et al. 2012). On the basis of this phylogenetic tree, we summed the branch length separating each species pair to calculate the phylogenetic distance and, as a result, estimate the evolutionary relatedness between species pairs.

Index calculation

A competition intensity index (CInt) was calculated to express the magnitude of the interactions between plants and their nature (i.e. competition, neutralism or facilitation). It is useful for comparing different pairs of plants experiencing the same growth condition or different growth conditions for the same pair of plants. We used the index formula proposed by Callaway et al. (2002):

display math

where, for a given species, growth condition and harvest, PT−N is the shoot biomass of the target plant without neighbours, PT+N is the mean shoot biomass of the target plant with neighbours and x is the maximum of PT+N or PT−N. A positive value indicates competition and a negative one facilitation. Moreover, the closer to 1 the absolute value of CInt, the stronger is the effect of the neighbour on the target plant. All plants with a dead neighbour (24 of the 740 planted in interactions) were excluded from analysis, since all plant death occurred during the implantation stage and was thus difficult to attribute to neighbour effects.

To assess the effect of abiotic stress among species, we calculated an abiotic stress intensity index on monoculture performance:

display math

where, for a given species and harvest, PT−S is mean shoot biomass of the plant growth without shortage of a given factor, PT+S is mean shoot biomass of the plant growth with shortage of a given factor and x is the maximum of PT−S or PT+S. The closer to 1 the value of abiotic stress intensity, the stronger is the effect of the shortage.

Data analysis

Relations between both hierarchical and absolute trait attribute distances and CInt values of species pairs were assessed by ancova. In these ancova, the CInt value of the target plant within each species pair were expressed as a function of the trait attribute distance between the two species in the pairs (hierarchical or absolute). Water treatment, P treatment and harvest were added as covariables (CInt and trait attribute distances were calculated for each growth condition and harvest). We used the same procedure for the phylogenetic distance, the only difference being that this distance did not change with growth conditions or harvest. We also used the same procedure for trait plasticity, expressing CInt as a function of the hierarchical trait plasticity distance of each species pair grown in stressful conditions. For each model, the assumptions of normality and homoscedasticity were tested. We used the post hoc Tukey-HSD test to identify significant differences between means. Variance analyses of species coordinates effect on the trait attributes of species were performed to determine the influence of phylogeny on species' trait attributes using eigenvectors associated with the phylogenetic tree of the four species (Diniz-Filho, de Sant'Ana & Bini 1998). All statistical analyses were performed with R 2.12.1 software (R Core Team 2012).

Results

Species attributes and abiotic factors

Analysis of root functional traits confirmed that the species differed in their below-ground resource-acquisition strategies (Table 1). Holcus lanatus and D. glomerata (acquisitive species) had the highest values of the three considered traits, while B. erectus and L. arundinaceum (conservative species) had the lowest (Table 1). These differences in root functional strategies were not linked to phylogenetic distance between species (variance analysis results SRL P = 0·82, RLD P = 0·61, RPUE P = 0·72).

Table 1. (a) Specific root length, (b) Root length density and (c) Root phosphorus use efficiency values of four grass species function of the growth conditions with all harvest combined (mean ± SE)
SpeciesHigh water and high PHigh water and low PLow water and high PLow water and Low P
  1. Different letters indicate significant differences within a line (anova and post hoc Tukey's HSD test, n = 228).

(a) Specific root length (m g−1)
Bromus erectus 164 ± 14 a199 ± 25 ab214 ± 19 b181 ± 17 a
Dactylis glomerata 226 ± 17 a238 ± 20 a212 ± 15 a210 ± 18 a
Holcus lanatus 254 ± 12 a346 ± 29 b253 ± 11 a252 ± 14 a
Lolium arundinaceum 150 ± 14 a164 ± 13 a144 ± 9 a164 ± 12 a
(b) Root length density (cm cm−3)
B. erectus 5·0 ± 1·4 b4·4 ± 1·1 b1·8 ± 0·4 a1·4 ± 0·3 a
D. glomerata 11·3 ± 2·0 b11 ± 2·3 b3·4 ± 0·2 a3·9 ± 0·3 a
H. lanatus 15·2 ± 2·2 b18·6 ± 3·5 b4·4 ± 0·1 a4·5 ± 0·3 a
L. arundinaceum 7·5 ± 1·5 b6·5 ± 1·2 b1·8 ± 0·2 a1·9 ± 0·2 a
(c) Root phosphorus use efficiency (m mg−1)
B. erectus 71 ± 6·6 a127 ± 10 b75 ± 10 a105 ± 17 b
D. glomerata 106 ± 8·6 a166 ± 13 b85 ± 10 a136 ± 10 ab
H. lanatus 100 ± 4 a263 ± 30 c82 ± 4·8 a150 ± 7·2 b
L. arundinaceum 70 ± 4·6 a101 ± 10 b47 ± 4·5 a99 ± 4·6 b

Differences in root functional strategies influenced species' above-ground biomass production and their ability to resist abiotic stress. The two species with the highest values of the three traits, that is, H. lanatus and D. glomerata, were also the most productive in high water- and P-availability growth conditions and the most impacted by water stress and the combined effect of water and P shortage (Table 2). The four species were equally impacted by P shortage despite their root trait differences (Table 2).

Table 2. Mean above-ground dry biomass at the third harvest of the four grass species studied, under high water and high P (mean ± SE), and mean index of stress intensity of the four grass species for all harvests combined (mean ± SE)
SpeciesAbove-ground dry biomass (g) in high water and high PIndex of stress intensity
Phosphorus stressWater stressWater and phosphorus stress
  1. Different letters indicate significant differences within a column (anova and post hoc Tukey's HSD test; n = 20 for above-ground biomass and n = 228 for the index of stress intensity).

Bromus erectus 17·5 ± 1·1 a0·46 ± 0·04 a0·36 ± 0·13 a0·47 ± 0·10 a
Dactylis glomerata 48·1 ± 2·5 c0·34 ± 0·03 a0·61 ± 0·06 b0·65 ± 0·07 bc
Holcus lanatus 48·8 ± 0·7 c0·43 ± 0·04 a0·66 ± 0·06 b0·72 ± 0·05 c
Lolium arundinaceum 38·7 ± 2·2 b0·35 ± 0·03 a0·44 ± 0·12 a0·58 ± 0·09 ab

Species trait attributes were impacted by growth conditions (Table 1). SRL appeared the less plastic trait; for example, H. lanatus and B. erectus were the only species that significantly increased their SRL in response to P and water shortage, respectively (Table 1a). P shortage had no effect on species' RLD values, while all species strongly decreased their RLD under water shortage (Table 1b). Finally, all species increased their RPUE under P shortage (with or without water shortage), with increases ranging from 29% (L. arundinaceum) to 164% (H. lanatus) (Table 1c).

Trait plasticity in response to growth conditions had significant effects on the trait distance of species pairs (Fig. 1). Trait plasticity in response to P shortage resulted in a higher mean absolute trait distance of the three traits considered. This was due mainly to a strong increase in values of the three traits of H. lanatus, while other species decreased their RLD and slightly increased their SRL and RPUE (Fig. 1). Under water shortage (with or without P shortage), the mean absolute trait distance of SRL and RPUE did not differ significantly from that for high water- and high P availability (Fig. 1a,c). Finally, under water shortage (with or without P shortage), trait plasticity induced a decrease in hierarchical trait distance of RLD (Table 1b).

Figure 1.

Effect of growth conditions on absolute distances of (a) specific root length (SRL), (b) root length density (RLD) and (c) root phosphorus use efficiency (RPUE) between pairs of species (mean ± SE). Growth conditions: NS = non-stressful, P = phosphorus stress, Water = water stress, P-Water = phosphorus and water stress. Different letters indicate significant differences between means (anova and post hocTukey's HSD test; n = 144 for RLD and SRL and n = 96 for RPUE).

Do hierarchical, absolute, phylogenetic or trait plasticity distances correlate with the intensity of interaction?

Hierarchical trait distance of the three root traits was highly correlated with the CInt value of interacting species pairs (Table 3a; Figs 2-4). Among the three ancovas, the best-fitting model included the RLD hierarchical distance (total R2 of the model = 0·65), while the other two models included the RPUE hierarchical distance or the SRL hierarchical distance and had an R2 of 0·49. Regression slopes were negative for all of these models (Figs 2-4), indicating that the larger and more negative the hierarchical trait distance between two interacting species, the more the neighbour negatively affected growth of the target species. Figs 2-4 also demonstrate that for the three traits, species with similar trait values (hierarchical distance close to zero) have intermediate CInt values, indicating no extreme competitive interactions for these species. At the same time, we did not observe significant relations between absolute distances of any traits or between phylogenetic distance and CInt values among species pairs (Table 3b,c).

Table 3. Results from four-way-multifactorial ancovas performed for the competition intensity index of species pairs as a function of their (a) hierarchical trait distance, (b) absolute trait distance and (c) phylogenetic distance. Traits considered are specific root length (SRL), root length density (RLD) and root phosphorus use efficiency (RPUE)
 SRLRLDRPUE
d.f. F P F P F P
(a) Hierarchical trait distance
Distance132·74 >0·001 92·46 >0·001 30·19 >0·001
Distance × water10·570·4529·10 >0·001 0·390·53
Distance × phosphorus12·520·118·64 >0·01 1·260·26
Distance × harvest2a0·060·810·010·910·370·54
Distance × water × phosphorus17·39 >0·01 3·97 0·04 18·58 >0·001
(b) Absolute trait distance
Distance11·160·281·710·190·150·71
Distance × water10·010·940·050·810·000·96
Distance × phosphorus10·110·740·020·860·030·85
Distance × harvest2a0·120·730·230·630·240·62
Distance × water × phosphorus10·160·690·000·940·050·82
 d.f. F P
  1. Bold values indicate statistical significance (P < 0·05); n = 144 for SRL and RLD and n = 98 for RPUE.

  2. a

    RPUE was measured only for the last two harvests, making the degrees of freedom for its harvest effect equal to one.

(c) Evolutionary distance
Distance11·380·24
Distance × water10·350·55
Distance × phosphorus10·000·97
Distance × harvest20·370·54
Distance × water × phosphorus10·250·62
Figure 2.

Index of competition intensity of target plants expressed as a function of the specific root length hierarchical distance under four growth conditions (a) high water and P availability, (b) high water and low P, (c) low water and high P, and (d) low water and P availability. Each point represents a species pair (target-neighbour) at one harvest. Significance of the regressions was assessed by Pearson's correlation test, n = 36 (12 species pairs × three harvests).

Figure 3.

Index of competition intensity of target plants expressed as a function of the root length density hierarchical distance under four growth conditions (a) high water and P availability, (b) high water and low P, (c) low water and high P, and (d) low water and P availability. Each point represents a species pair (target-neighbour) at one harvest. Significance of the regressions was assessed by Pearson's correlation test, n = 36 (12 species pairs × three harvests).

Figure 4.

Index of competition intensity of target plants expressed as a function of the root phosphorus use efficiency hierarchical distance under four growth conditions (a) high water and P availability, (b) high water and low P, (c) low water and high P, and (d) low water and P availability. Each point represents a species pair (target-neighbour) at one harvest. Significance of the regressions was assessed by Pearson's correlation test, n = 24 (12 species pairs × two harvests).

Root length density and RPUE plasticity distances were significantly but weakly correlated with the CInt value of interacting species pairs grown in stressful conditions, while the SRL plasticity distance was not (Table 4). As for trait hierarchical distance, these correlations were negative, indicating that for RLD and RPUE, the larger and more negative the trait plasticity distance between two interacting species the more the neighbour negatively affected growth of the target species.

Table 4. Correlation coefficients and their significance for linear regressions between trait plasticity distance of species pairs and their competition intensity in stressful growth conditions. Traits considered are specific root length (SRL), root length density (RLD) and root phosphorus use efficiency (RPUE)
Competition intensityTrait plasticity distance
SRLRLDRPUE
r P r P r P
  1. Bold values indicate statistical significance (P < 0·05); for all stressful growth conditions, n = 108 for SRL and RLD and n = 72 for RPUE.

All stressful growth conditions0·110·24−0·22 0·02 −0·25 0·03
High water low P−0·45 >0·01 −0·150·39−0·46 0·02
Low water high P0·45 >0·01 −0·59 >0·001 −0·45 0·03
Low water low P0·36 0·03 −0·230·180·250·22

Effects of growth conditions on relations between trait distances and competition intensity

For the three traits, growth conditions significantly influenced relations between hierarchical trait distances and CInt (Table 3a); the strongest relations (highest R2 and lowest P value) were found in high water- and P-availability treatment (Figs 2-4). For example, the interaction between water and P treatments significantly influenced the slope of regressions between SRL hierarchical distance and CInt values of species pairs (Table 3a, Fig. 2). In fact, the slope of the relations differed significantly from zero only for the high water-availability treatment (Fig. 2). Moreover, within the high water-availability treatment, slope and R2 were greater in the high P-availability treatment than the low P-availability treatment (Fig. 2a,b). As a result, the SRL hierarchical distance of species pairs was a good predictor of interaction outcomes only under conditions of high water- and P availability.

The relation between RLD and CInt was significantly influenced by growth conditions (Table 2) but remained negative and significant for all treatments (Fig. 3). Regression slopes differed significantly between high and low water-availability treatments (−0·18 and −0·68, respectively), indicating that even small RLD distances between interacting plants may have had a stronger effect on interaction outcomes in the low water-availability treatment than the high water-availability treatment. Conversely, within the high water-availability treatment, the P treatment had a significant effect on the regression slope (Table 3a). When P availability was low, the absolute value of the regression slope was lower (Fig. 3b,c), highlighting that in these conditions small RLD distances had little effect on the outcome of interaction. Interestingly, this effect of low P availability on the relation between RLD distance and competition intensity was not significant (P = 0·38) in the low water-availability treatment. Unlike the SRL hierarchical distance that for RLD appeared to predict well the competition intensity between plants, even under conditions of low water availability.

The relation between RPUE hierarchical distance and CInt was strongly influenced by the combined effect of water- and P-availability treatments (Table 2). Relations between these two parameters were negative and significant in all treatments except for the low water- and low P-availability treatments (Fig. 4). These results indicate that RPUE hierarchical distance may explain competitive intensity under conditions in which P availability can strongly influence plant growth. Regardless of the functional trait considered, the duration of growth did not change the relation between hierarchical trait distance and competition intensity of species pairs (Table 3).

As for hierarchical trait distance, growth conditions significantly influenced the relation between trait plasticity distance and CInt of species pairs (ancova: SRL: P < 0·001; RLD: P < 0·001; RPUE: P = 0·02). For SRL, the correlation was negative under P shortage and positive under water shortage. Thus, the more a species changed its SRL in response to P shortage the more it impacted its neighbour and the less it was impacted, while under water shortage, the more a species changed its SRL the less it impacted its neighbour and the more it was impacted (Table 4). Although relations between RLD and RPUE plasticity distances and CInt were also impacted by growth conditions, the latter did not change the sign of significant relations, which were negative (Table 4). Thus, under water shortage for RLD and P or water shortage for RPUE, species that changed their trait value more than their neighbour did were less impacted by the neighbour while more strongly impacting this neighbour.

Discussion

Hierarchical trait distances, trait plasticity distance and growth conditions drive neighbours' competitive effect

Results indicate that competition between pairs of grasses is linked to their hierarchical trait attribute distances but not to their functional similarity or phylogenetic relatedness. Since competition intensity coefficients are linked to the hierarchy and plasticity of SRL, RLD and RPUE, these three traits appear to be related to species' competitive abilities. Moreover, competition for soil resources appears to be a major driver of early-stage interactions between grasses. These results support the competition-trait hierarchy hypothesis for describing interactions between grasses, as recently suggested for trees (Kunstler et al. 2012). They also highlight that between-species differences in trait plasticity, in response to abiotic constraints, may induce differences in competitive abilities. Our results show that relations between hierarchical trait or trait plasticity distances and competition intensity were influenced by abiotic treatments, suggesting that growth conditions influence competitor hierarchies.

Species with high SRL are strong competitors under high water- and P-availability conditions, which is consistent with the idea that this trait attribute is displayed by competitive, fast-growing species able to acquire and efficiently use soil resources (Comas & Eissenstat 2004; Mommer et al. 2011). Similarly, within the Poaceae considered, the more a species increased its SRL in low P-availability conditions, the more it is a strong competitor, supporting the idea that trait plasticity is an important factor driving competitive ability (Callaway, Pennings & Richards 2003). High SRL allows species to be competitive in water-rich environments by acquiring the large amounts of resources necessary to sustain rapid development of above-ground organs (Mommer & Weemstra 2012). However, acquisitive fast-growing species appear to be more impacted by water shortage than conservative ones, perhaps explained by their high water demand to sustain high growth rates (Grime 1977). This strong impact of water shortage on the growth of acquisitive species may explain the lack of correlation between SRL hierarchical distance and competition intensity of species pairs in the low water-availability treatment. Under these conditions, acquisitive species were unable to tolerate stress and, as a consequence, their growth decreased more than that of conservative species. We hypothesize that this reduction in growth is responsible for decreasing competitive effects of acquisitive plants on their neighbours. This may explain why functionally contrasting species can coexist in Mediterranean grassland habitats with intermediate stress levels, under which acquisitive species can persist without competitively excluding conservative species (Bernard-Verdier et al. 2012). Moreover, species that change their SRL in response to water shortage are weak competitors in this condition; as a result, trait plasticity under water shortage does not denote high competitive ability.

Results show a strong relation between the RPUE and competition intensity of interacting species. Species with high RPUE are stronger competitors than species with lower RPUE, which is consistent with the idea that the more efficiently a species uses resources, the more competitive it becomes (Funk & Vitousek 2007). In this study, this relation depended greatly on growth conditions, specifically which factor(s) limited plant growth. As a result, one can argue that it is not plants' overall resource-use efficiency that drives their competitive ability, but their ability to efficiently use the most limiting resources in a given environment (Westoby et al. 2002). This is confirmed by the observation that species that increase their RPUE under P shortage are the most competitive in this condition, while those that decrease their RPUE are the most competitive under water shortage.

We demonstrated that regardless of the growth condition considered, species with high RLD were more competitive than those with low RLD. This finding supports model predictions of RLD as a main component of competitiveness in grass species (Craine 2006), which may explain some cases of competitive exclusion (Raynaud & Leadley 2005). Compared with the inclusion of SRL and RPUE, the advantage of including RLD in this type of approach is its integrative value. RLD considers not only the kind of root produced (SRL) but also a plant's production capacity (root biomass) in considered habitats, integrating the effects of abiotic conditions on plant growth. Regardless of the growth conditions, plants with higher RLD than their neighbours may be able to acquire more resources (Craine 2006; Mommer et al. 2011) and grow larger than their neighbours, thereby becoming more competitive. Moreover, the higher the plasticity of a species' RLD, the more strongly it competes, particularly under water shortage (Mommer et al. 2011). However, although acquisitive species had higher RLD in this short-term study, they did not when grown for more than one growing season (Craine et al. 2002; Fort, Jouany & Cruz 2012). It would be interesting to test in a long-term field study whether RLD is always correlated with plant competitive ability or whether this relation changes with time or with environmental factors. This is particularly pertinent when species are grown in field conditions, where the extent of root growth would be less limited in width but much more in depth than in pots. In this case, we can hypothesize that high RLD always allows high resource acquisition and higher competitive ability. However, in natural conditions, it also would be interesting to estimate whether the RLD of plants on shallow or deep soil layers is related to competitive ability that function to growth conditions, which is not relevant in pot experiments.

No evidence for the competition-trait similarity hypothesis

As discussed, when considering SRL, RLD and RPUE, the competition-trait hierarchy hypothesis describes interactions between grasses well under a variety of growth conditions. Moreover, even under growth conditions in which hierarchical trait distances failed to explain competition intensities between species pairs, we found no evidence that the competition-trait similarity or competition-relatedness hypotheses explained them better, since absolute trait distance or phylogenetic distance never correlated with the competition intensity index. In the case of phylogenetic distance, it is important to note that the strength of the analysis was limited because our experimental design (four species) provided only 12 phylogenetic distances, which were not modulated by growth conditions as functional traits were. This may explain the weakness of the relation between phylogenetic distance and competition intensity index. However, recent studies on trees (Uriarte et al. 2010; Kunstler et al. 2012), and more generally on vascular plants (Cahill et al. 2008), failed to support the competition-relatedness hypothesis. Moreover, competition has reduced trait variations and species richness in rich meadows (Grime 2006). These results show the influence of growth condition on the importance of niche similarity in driving plant competition at the neighbourhood scale (Kraft & Ackerly 2010).

In our study, differences in functional trait attributes did not appear to induce spatial niche segregation within pots (Chesson 2000). In optimal growth conditions, that is, high water- and P availability, the competitiveness of interacting species, summarized by trait attributes, was clearly the main factor driving interaction strength. However, it is important to note that, as previously mentioned, we worked with four grass species. Although they were chosen to maximize the range of trait values considered, they did not include all the trait variability possible in most natural grasslands. As a consequence, our results describe only the link between functional traits and plant interactions within the Poaceae family, which is the dominant family in grasslands (Gibson 2009). Further work is needed to test whether these relations between hierarchical trait and plasticity distances and competitive interaction between plants could be extended within and among other plant families.

Conclusion

Our results demonstrate that during early stages of implantation, the competition-trait hierarchy hypothesis better explains interaction strength between grasses than the competition-trait similarity hypothesis. We also showed that species may occupy different niches by having different trait values and tolerances for abiotic stress but interact strongly when put in competition within the same space. Our results show that investment in root production, the type of root produced and root trait plasticity are important factors defining plant competition ability. This work opens new perspectives in the field of plant-interaction studies by highlighting the importance of root resource-acquisition strategies on plant competition and coexistence.

Acknowledgements

This work was funded by the ANR project O2LA (ANR-09-STRA-09). F.F. was supported by a doctoral fellowship from the French Ministry of Education and Research. We are grateful to E. Lecloux for technical assistance and to P. Mouton, M. Rougier and E. Decorsière for their help measuring traits. We also thank C. Roumet and C. Picon-Cochard for advising and sharing their expertize on root functional trait studies. We also thank M. Corson for improving the English of the text.

Ancillary