Evidence for a ‘plant community economics spectrum’ driven by nutrient and water limitations in a Mediterranean rangeland of southern France

Authors


Correspondence author. CSIC Instituto de Recursos Naturales y Agrobiología de Sevilla, Av. Reina Mercedes 10, 41012 Seville, Spain. E-mail: imperez@irnase.csic.es

Summary

  1. Plant species composition and community functional structure (i.e. trait composition at the community level) result from a hierarchy of environmental filters that constrain which species and traits tend to be dominant in a given habitat.
  2. We quantified variation in community functional structure along natural gradients of soil resources using several above- and below-ground parameters and explored links among these attributes to determine whether plant resource economics can be applied at the community level in a Mediterranean rangeland of southern France.
  3. Limitation by nitrogen, soil water and soil depth were the main ecological factors driving the functional response at the community level. Most of the community functional parameters considered in this study were more dependent on nitrogen limitation than on the other two factors, mostly related with the acquisition–conservation trade-off at both the leaf and the root level.
  4. We found a strong coordination between above-ground and below-ground components, with a high level of concordance along the resource gradients explored. As an example, tissue dry matter content – both in leaves and roots – was positively related to nitrogen limitation. These findings indicate that the leaf economic spectrum paradigm (resource conservation in resource-poor habitats versus resource acquisition in resource-rich habitats) can be extrapolated to the below-ground component and extends to a plant community spectrum.
  5. Changes in the functional structure of communities were promoted by two complementary components of variation: (i) the replacement of species with highly contrasting resource-use strategies and, to a lesser extent, (ii) the intraspecific variation in several above-ground traits.
  6. Synthesis. This study showed that soil water and nutrient limitations are the main drivers controlling functional community structure in the Mediterranean rangelands studied and that shifts in this structure were mainly due to species turnover. In addition, we provided evidence for a plant community economics spectrum, based on a strong coordination between above- and below-ground components in these resource-limited communities.

Introduction

Plant species composition and community functional structure (i.e. trait composition at the community level) result from a hierarchy of abiotic (climate, resource availability and disturbances) and biotic (competition, predation) filters that constrain which species and traits tend to be dominant in a given habitat (Woodward & Diament 1991; Keddy 1992; Díaz, Cabido & Casanoves 1999; Laliberté et al. 2012). The ability of a plant species to grow and reproduce under certain abiotic and biotic conditions usually relates to specific morphological and physiological traits (concept of plant strategy; Grime 2001), which can co-vary or vary independently in response to different factors (Chapin, Autumn & Pugnaire 1993; Grime 2001; Westoby et al. 2002). Resource availability (particularly soil water and nutrient availability) has been recognized as one of the main ecological factors driving plant strategy selection (e.g. Aerts & Chapin 2000; Sandel et al. 2010). A common way of analysing the relationships between resources and plant strategies is to compare resource-rich versus resource-limited environments (e.g. Fonseca et al. 2000; Wright & Westoby 2001; Quétier, Thébault & Lavorel 2007; Laliberté et al. 2012) for a limited (one or two) number of resources available at highly contrasting levels. In natural conditions, however, plants are exposed to a variety of constraints, which constitute a multidimensional space where many abiotic and biotic factors act simultaneously and interactively (Ibañez & Schupp 2001; Gómez 2004). Detailed studies quantifying combined and interactive effects of soil resources under field conditions are scant, and most of them are focused on single species (but see Gross et al. 2008; Ordoñez et al. 2009; Sonnier, Shipley & Navas 2010). Field studies conducted at the community level explicitly measuring continuous gradients of abiotic factors are therefore essential to determine how species and community functional structure segregate along natural environmental gradients.

To better understand plant adaptations to soil resource scarcity, most studies on trait-environment linkages have focused on leaf traits. A good example of functional strategy selection in response to resource availability is the trade-off between acquisition and conservation of resources, which can be captured from the range of trait variation that defines the leaf economics spectrum (Grime et al. 1997; Wright et al. 2004). Species with high specific leaf area (SLA), low-density tissues and high leaf N concentration tend to have rapid resource capture and high relative growth rates (e.g. Wright & Westoby 2001; Ruíz-Robleto & Villar 2005; Poorter & Garnier 2007) that allow plants to be more dominant in moist and fertile areas (Reich et al. 1999; Grime et al. 1997). Opposite attributes characterize species with an efficient resource conservation (Chapin, Autumn & Pugnaire 1993; Wright et al. 2004), allowing the plant to minimize nutrient loss and increase competitive abilities in dry and nutrient-poor environments (Hobbie 1992; Aerts 1995). However, this trade-off has seldom been applied to other organs of the plant, such as the below-ground component, and thereby the extent to which it can be generalized to whole-plant strategies remains largely untested (Craine et al. 2005; Tjoelker et al. 2005). There is therefore a need to explore how below-ground traits vary along environmental gradients in conjunction with those pertaining to the above-ground component (Holdaway et al. 2011).

Recent studies (Freschet et al. 2010; Holdaway et al. 2011; Birouste et al. 2012) have empirically demonstrated that the trade-off pertaining to acquisition and conservation of resources can be extrapolated to other plant organs such as stems or roots, although other studies have revealed partially inconsistent cross-species correspondences between leaf and root traits (e.g. Tjoelker et al. 2005; Withington et al. 2006; Liu et al. 2010). Here, we test whether the coordination between leaf and root traits found by Freschet et al. (2010) could be also applied in Mediterranean rangelands at the scale of the whole community, using above- and below-ground community functional parameters (sensu Violle et al. 2007). Above-ground community functional parameters were calculated by combining average trait values of individual species with the relative abundance of these species, whereas those pertaining to the below-ground component were measured directly by collecting deep soil cores containing multi-species root samples of the whole community.

Changes in the functional structure of communities may be due to the replacement of species with different trait values (interspecific variation), to within-species changes in trait values (intraspecific variation) or to a combination of these two sources of variability (Garnier et al. 2004). There is growing evidence that intraspecific variability in certain functional traits might be relatively large (Albert et al. 2010a; Leps et al. 2011), being sometimes as important as the interspecific variation (Messier, McGill & Lechowicz 2010). This within-species variability, which can play a crucial role on community dynamics and ecosystem functioning (e.g. Booth & Grime 2003; Crutsinger et al. 2006), has been rarely considered in studies exploring community functional traits along environmental gradients. Here, we quantified variation in multiple community functional parameters along field resource gradients and examined the relative importance of species turnover versus intraspecific variation on the changes of these community functional parameters.

To summarize, the main objectives of this study were (i) to examine the relative importance of environmental factors (especially soil water and nutrient limitation, but also soil depth, texture and chemistry) on the functional matrix of multiple leaf and root traits; (ii) to explore whether the leaf economics spectrum can be expanded to the below-ground component and applied at the community level; and (iii) to determine the relative importance of species turnover versus intraspecific variation on community functional parameters along the explored natural gradient of the main environmental factors. The information provided by this study will enable us to better understand the drivers of community assembly and plant adaptations to soil resource scarcity in Mediterranean rangelands. In addition, our findings could be used as a valuable tool to infer the impact of changing environmental conditions due to global change on species composition and community functional structure in these resource-limited communities.

Materials and methods

Study area

The study was conducted at the INRA La Fage experimental station (43°55′ N, 3°05′ E, 790 m a.s.l.), located on a limestone plateau (Larzac Causse), 100 km north-west of Montpellier (France). Climate is sub-humid Mediterranean type, with cool, wet winters alternating with warm, dry summers. Mean annual precipitation ranges from 680 to 1790 mm (mean of 1070 mm over the previous 36 years); rainfall occurs mainly during spring and autumn. Mean monthly temperature varies from 1 to 19 °C (mean annual of 9.5 °C over the same period). Bedrock is mostly formed of Jurassic limestones, which compose a mosaic of calcareous and dolomitic rangelands dominated by herbaceous species with scattered shrubs such as Buxus sempervirens or Juniperus communis. For the last 35 years, the vegetation has been grazed by a sheep herd (Romane breed), raised outdoors all year long for meat production (see Molenat et al. 2005 for details).

Experimental design and plant sampling

In February 2008, 12 plots (5 × 5 m) located on dolomitic rendzines were selected to span the widest possible range of soil resources and conditions, from shallow and dry plots to deep and moist plots (Appendix S1). The distance among plots ranged from 100 to 1500 m. The selected plots were fenced for the duration of the experiment to avoid sheep grazing.

In May 2008, during the peak of vegetative growth, 8 0.25 m2 quadrats were randomly placed in each plot to determine species composition and frequency. In each quadrat, we identified and ranked the most dominant species in terms of biomass using the visual, non-destructive ‘Botanal’ method (Tothill et al. 1992). Botanal is widely used to estimate pasture yield and species composition in grassland communities (Tothill et al. 1992; see Lavorel et al. 2008 for comparison with other methods).

Characterization of the soil environment

A large number of soil characteristics were measured in each of the 12 experimental plots (Appendix S1). Soil volumetric water content was measured using a capacitance soil moisture probe (DIVINER 2000, Sentek Pty Ltd, Stepney, SA, Australia), which provides a complete profile of soil humidity every 10 cm in depth. In each plot, three permanent tubes (diameter 6 cm) were vertically inserted to maximum soil depth; measurements of soil moisture were taken weekly from May to July 2008. To calibrate the Diviner probe, three soil cores (diameter 5 cm) by plot were collected in the middle of May, in the vicinity of each tube. Each core was divided into several 10-cm sections, which were used to determine soil humidity using the gravimetric method. Soil gravimetric and volumetric humidity were related, and two calibration equations were developed according to the percentage of clay in soil (r = 0.98 and < 0.001 when clays > 25%; r = 0.80 and < 0.001 for the remaining soils). From these measurements, we derived an integrative variable of the whole soil profile for each of the 12 experimental plots, namely soil water storage (SWS, expressed in mm of water). SWS represents the amount of water available for plants and was calculated as the sum of all the water content values (measured for each 10-cm-depth section) from the soil surface to a soil depth corresponding to the 95% of rooting depth (see below). Due to the high correlation between mean SWS values in spring and summer (r = 0.99 and < 0.001), we used the average SWS for the whole sampling period (May–July) in statistical analyses.

Soil physico-chemical properties were assessed from three soil cores collected in each plot using a 5-cm-diameter auger, to the maximum reachable soil depth. Soil cores were taken each 10 cm depth and then bulked by pedological horizons. In the laboratory, samples were air-dried, crushed and sieved; the < 2-mm fraction was analysed by the ‘Laboratoire d′Analyses des Sols’ of the National Institute for Agronomic Research (INRA, Arras, France) using standard procedures (Afnor 1994). Nine soil properties were determined: soil texture, pHwater, cation exchange capacity (CEC), available phosphorus (using the Olsen method), total organic matter, total nitrogen and C:N ratio. Percentage of clay was used as a representative measurement of textural characteristics in statistical analyses. In addition, the amount of nitrogen and phosphorus that a given soil can release and make available for plants was assessed using nitrogen (NNI) and phosphorus (PNI) nutrition indices, respectively, as recommended by Garnier et al. (2007). NNI was calculated as the ratio between the actual nitrogen concentration of above-ground biomass and the critical nitrogen concentration (i.e. the N content allowing potential growth), as proposed by Lemaire & Gastal (1997). This critical N concentration has been previously calculated and validated in multi-species grasslands from N dilution curves (Duru, Lemaire & Cruz 1997), which represent decreasing relationships between N concentration and above-ground biomass. PNI, which depends on nitrogen concentration of above-ground biomass, was calculated following Duru & Ducrocq (1997). Nitrogen concentration was determined on additional plant biomass harvested at the end of the growing season (4 July 2008) in four 0.5 × 0.5 m quadrats placed in each of the 12 plots (see details below). These indices reflect thus the growth limitation by nutrient supply and allow quantifying changes in soil nutrient availability for plants (cf. Garnier et al. 2007).

Soil depth was estimated from the soil cores (6–9 replicates by plot) collected for measuring soil water content (gravimetric method), soil physico-chemical characteristics and below-ground traits (see details below).

Plant measurements

Above-ground functional parameters

Species selected for trait measurements were the most abundant ones, which collectively made up more than 80% of the maximum standing biomass of each community. Measurements taken on species present in different plots allowed us to quantify intraspecific variability along the explored environmental gradients; traits were thus measured on a total of 45 populations belonging to 14 different species (Table 1). All measurements were conducted in May 2008 during the peak of vegetative growth, on healthy plants, located in well-lit microsites. Six quantitative above-ground traits were selected for their known or hypothesized responses to the factors studied. For plant sampling, we followed the criteria defined in Cornelissen et al. (2003).

Table 1. Mean frequency and standard deviation of the most abundant species in the 12 experimental plots. Plots have been separated as a function of their soil depth: shallow (20–30 cm), intermediate (30–40 cm) and deep plots (> 50 cm). Species and plots selected for trait measurements (i.e. those that collectively made up more than 80% of the maximum standing biomass of the community) have been highlighted with bold letters
SpeciesDeep plotsIntermediate plotsShallow plots
Achillea millefolium 0.9 ± 1.90.00.0
Anagallis arvensis 0.00.00.7 ± 1.3
Brachypodium pinnatum 2.1 ± 3.0 0.00.0
Bromus erectus 51.2 ± 12.5 43.5 ± 16.6 4.2 ± 8.5
Carex flacca 9.3 ± 18.6 0.00.0
Carex humilis 10.7 ± 14.3 10.3 ± 9.1 3.8 ± 5.4
Carex halleriana 2.9 ± 4.1 2.2 ± 4.4 0.0
Cirsium acaule 0.30.00.0
Festuca christiani-bernardii 0.0 19.0 ± 12.8 41.2 ± 24.3
Genista hispanica 0.9 ± 1.90.00.0
Helianthemum apenninum 0.00.0 6.1 ± 4.9
Helianthemum canum 0.01.3 ± 2.6 10.2 ± 12.7
Helianthemum nummularium 0.3 ± 0.51.1 ± 1.30.3 ± 0.5
Hieracium pilosella 0.9 ± 1.20.7 ± 1.30.0
Inula montana 0.3 ± 0.50.00.0
Lotus corniculatus 0.7 ± 0.50.00.0
Myosotis ramosissima 0.3 ± 0.50.00.0
Onobrychis viciifolia 2.9 ± 4.1 0.00.0
Ononis striata 0.00.9 ± 1.22.5 ± 4.2
Plantago lanceolata 4.3 ± 8.6 0.00.0
Poa bulbosa 0.00.1 ± 0.30.9 ± 1.5
Potentilla neumanniana 4.3 ± 3.3 8.0 ± 8.0 15.0 ± 12.0
Ranunculus gramineus 0.00.00.3 ± 0.5
Ranunculus sp.0.3 ± 0.50.00.0
Sanguisorba minor 0.00.3 ± 0.5 11.1 ± 22.3
Sedum sp.0.00.00.5 ± 0.6
Seseli montanum 0.00.00.3 ± 0.5
Sherardia arvensis 0.00.01.4 ± 2.9
Stipa pennata 0.7 ± 1.3 8.4 ± 12.9 0.0
Thymus dolomiticus 0.00.01.2 ± 1.8
Trifolium sp.0.4 ± 0.80.00.0
Trifolium sp. II0.1 ± 0.30.00.0
Vulpia myuros 6.3 ± 6.7 1.6 ± 3.20.1 ± 0.3

A total of 30 different individuals per species and plot were randomly selected to measure maximum reproductive height.

Water-saturated specific leaf area (SLA; leaf area per unit of dry leaf mass; m2 kg−1), leaf dry matter content (LDMC; dry mass per unit of water-saturated fresh mass; mg g−1) and leaf thickness (LT; μm) were determined on blades of 10 individuals per species among the 30 selected for maximum reproductive height measurements, following the protocol described by Garnier et al. (2001). Leaf projected area was determined with an area meter (Delta-T Devices, model MK2, Cambridge, UK). Leaf thickness was measured with a linear variable displacement transducer, taking from 5 to 10 measurements per blade depending on the species-specific leaf size. All the leaf samples were weighed and oven-dried at 60 °C for 48 h and then re-weighed.

Leaf nitrogen concentration (LNC; mg g−1) and leaf carbon isotope ratio (δ13C; ‰, values relative to PDB standard, which provides a time-integrated measure of intrinsic water use efficiency; Farquhar, O'Leary & Berry 1982) were measured on four samples obtained by pooling the 10 original leaves used for SLA and LDMC. Bulked samples were ground, and LNC was determined with an elemental analyser (Carlo Erba Instruments, model EA 1108, Milan, Italy). The δ13C was determined using a CHN elemental analyser coupled to an isotope mass spectrometer.

To scale up from species to community level, all these above-ground traits were weighted by the relative abundance of the species at each experimental plot to calculate community weighted means (CWMa sensu Garnier et al. (2004) – the subscript a refers to above-ground parameters). The CWMa were calculated with or without considering the intraspecific variability to infer the relative magnitude of this source of variability (see details below), although we recognize that this is sometimes underestimated when only the most dominant species are considered into analyses.

Below-ground functional parameters

Due to the difficulty of separating root systems of different species, below-ground functional parameters (BFP) were measured directly at the community level. Three randomly distributed soil cores (5 cm diameter and 0.2–1 m length according to soil depth) per plot were collected at the end of the growing season (July 2008). Cores were divided into sections of 10 cm depth from the surface to the maximum soil depth. In the laboratory, roots belonging to the different species present in the plant community were carefully washed free of soil in water and a representative sub-sample of fresh roots from each soil depth interval was scanned at 400 dpi [see Hummel et al. (2007) for methodological details]. The digital images were used to determine length, area, volume (as the sum of volumes for the different diameter classes) and mean diameter of roots using image analysis software (Winrhizo ver. 2003b, Regent Instruments Inc., Quebec, Canada). The root material harvested was immediately weighed, oven-dried at 60 °C for 48 h and then re-weighed. Three below-ground functional parameters (BFP) were calculated from these measurements: specific root area (SRA; root area per unit of dry root mass; m2 g−1), tissue mass density (TMDr; the ratio of root dry mass to fresh volume; g cm−3) and root dry matter content (RDMC; root dry mass per unit of root fresh mass; mg g−1). In this study, SRA was preferred to SRL since it is the actual analogue of SLA. In addition, it was highly correlated with SRL (r = 0.91; < 0.001). All these below-ground parameters were weighted by the relative biomass of their 10-cm sections to calculate community parameter means for the whole soil profile (WSP).

display math

Where i is the soil depth interval divided into 10-cm sections, ranging from 0–10 cm (1) to the maximum reachable soil depth (n).

Root biomass distribution along the whole soil profile was used to calculate the rooting depth (m), that is, the soil depth that contains the 95% of the total dry root biomass. Finally, we calculated the root mass fraction (RMF, root dry mass per unit of total plant dry mass; g g−1), which indicates the proportional biomass investment in the below-ground compartment. Above-ground biomass was determined by harvesting all the plants present in four 0.5 × 0.5 m quadrats placed in each of the 12 plots, close to the location where the soil cores were taken. The above-ground fraction was oven-dried (at 60 °C for 48 h) and then weighed with a precision of 0.0001 g. For the calculation of RMF, above-ground and below-ground fractions were previously standardized per cm2 of area sampled.

A list of all the environmental factors and community functional parameters measured in this study is given in the Table 2.

Table 2. Environmental factors and community functional parameters measured in this study, with their abbreviations and units when relevant
VariableAbbreviationUnit
Environmental factors
Soil depthcm
Proportion of claySoil clay%
pHwaterpH
Cation exchange capacityCEC
Soil water storage available for plantsSWSmm
Plant-available phosphorusSoil Pg kg−1
Soil total nitrogenSoil Ng kg−1
Soil organic matterg kg−1
C:N ratio
Nitrogen nutrition indexNNI%
Phosphorus nutrition indexPNI%
Functional community parameters
Above-ground parameters
Plant reproductive height (community weighted mean)Plant rep. heightcwmcm
Specific leaf area (community weighted mean)SLAcwmm2 kg−1
Leaf dry matter content (community weighted mean)LDMCcwmmg g−1
Leaf thickness (community weighted mean)Leaf thicknesscwmμm
Leaf nitrogen concentration (community weighted mean)LNCcwmmg g−1
Carbon isotope ratio (community weighted mean)δ13Ccwm
Below-ground parameters
Root mass fractionRMF
95% rooting depthRooting depthm
Specific root area (whole soil profile)SRAwspm2 kg−1
Root dry matter content (whole soil profile)RDMCwspmg g−1
Tissue mass density of roots (whole soil profile)TMDrwspg cm−3

Data analyses

Analyses of environmental factors

To explore the level of dependence among environmental variables, Pearson′s correlation analyses were conducted among the 11 soil factors considered in the study. The criterion false discovery rate (FDR) was set at the 5% level to control the inflation of type I error derived from repeated testing (Garcia 2004). When necessary, some variables were log transformed prior to analyses to fulfil assumptions of normality and homoscedasticity. Normality was tested using the Shapiro–Wilk test. These analyses were carried out using Statistica v. 6 (StatSoft Inc. 2001).

Models of community functional parameters

The 11 functional parameters assessed at the community level were modelled independently as a function of the 11 environmental factors considered in the study, using maximum likelihood techniques. We tested three alternative functions that encapsulated different responses of communities to abiotic factors, covering a wide range of possible forms: a linear response (linear model), an exponential response (exponential model) and a saturating response (Michaelis–Menten model; see equations in Table 3). These functions are commonly used to model parameters related with plant growth and morphology at the individual level (e.g. Pacala et al. 1994; Gómez-Aparicio et al. 2008; Pérez-Ramos et al. 2010). We first tested models for each environmental factor and function independently, and the best of the three models was compared with the null model, which assumes no effect of any factor. Second, to test for among-factors interactions, we fitted two-factor models using those environmental factors that had an effect on the different functional parameters when evaluated independently. We tried alternative models in which the second environmental factor was added either additively or multiplicatively. Models including more than two environmental factors are not presented here due to their lower empirical support.

Table 3. Summary of the best-fitted models analysing community functional parameters (‘specific’ averages) in response to environmental factors. Only the models with better empirical support than null are shown, ranked from best to poorest fits. Additive interactions between two factors are noted as (+). The best-supported model and their equivalents (ΔAIC ≤ 2) have been highlighted with bold letters for each community parameter. The signs of the relationships (positive or negative) between each dependent variable and the selected predictors are also indicated, separated by commas. Model Forms: LIN, linear model; EXP, exponential model; MM, Michaelis–Menten model; null, null model
Dependent variablePredictorsModelRelation R 2 AICΔAIC
  1. The equations of the different functions fitted in the models calibrated for this study are:

  2. (i) Linear additive:

    display math
  3. (ii) Linear multiplicative:

    display math
  4. (iii) Exponential additive:

    display math
  5. (iii) Exponential multiplicative:

    display math
  6. (iii) Michaelis–Menten multiplicative:

    display math
  7. where a, b, c and d are parameter estimates that maximized the likelihood function, and Factors Ai, Bi and Ci are the selected predictor variables for each experimental plot ‘i’.

Above-ground parameters
Plant reproductive heightcwm (cm) Soil depth MM + 0.63 95.62 0.00
SWS MM + 0.61 96.38 0.76
Null 103.978.35
SLAcwm (m2 kg−1) NNI LIN + 0.69 120.10 0.00
SWS + NNI LIN +,+ 0.74 122.03 1.93
SWSLIN+0.55124.534.44
Soil depthLIN+0.46126.606.51
Null 130.3910.29
LDMCcwm (mg g−1) NNI EXP 0.55 121.26 0.00
SWSLIN0.37125.364.10
Null 127.396.13
Leaf thicknesscwm (µm) NNI EXP + 0.67 137.02 0.00
C:N ratioEXP0.39144.337.31
Soil depthEXP0.37144.677.65
Null 146.609.58
LNCcwm (mg g−1) NNI MM + 0.53 71.06 0.00
Null 76.505.45
δ13Ccwm SWS LIN 0.65 118.72 0.00
Soil depthLIN0.52122.593.86
CECLIN0.50122.914.19
ClaysLIN0.48123.444.72
NNILIN0.33126.537.80
C:N ratioLIN0.31126.868.14
null 198.8780.14
Below-ground parameters
RMF NNI EXP 0.69 84.71 0.00
Soil depthLIN0.4192.477.76
null 95.0210.32
95% Rooting depth (m) SWS LIN + 0.86 89.02 0.00
Soil depthLIN+0.61101.5612.55
NNILIN+0.56103.1814.16
null 109.2320.21
SRAwsp (m2 kg−1) C:N ratio LIN + 0.48 47.98 0.00
null 52.074.09
RDMCwsp (mg g−1) NNI EXP 0.68 127.74 0.00
C:N ratioEXP0.58130.893.15
null 137.679.93
TMDrwsp (g cm−3) C:N ratio EXP 0.51 103.50 0.00
null 108.494.98

Models were parameterized with maximum likelihood (Edwards 1992), using a simulating annealing algorithm. Competing models to predict functional parameters were selected with the Akaike Information Criterion corrected for small sample sizes (AICc) (Burnham & Anderson 2002) as a measure of goodness of fit: the lower the AIC value, the better the model. The absolute magnitude of the differences in AICc (ΔAIC) between alternative models provided an objective measure of the strength of empirical support for each one of them. Models with ΔAIC between 0 and 2 were considered to have equivalent and substantial empirical support. The R2 of the regression of observed versus predicted was used as a quantitative measure of goodness of fit of each alternative model. All models were implemented using the likelihood package version 1.1 for R and software written specifically for this study in R v 2.5.0 (R Development Core Team 2006, Vienna, Austria).

Links among community functional parameters

The same above-described modelling approach was used to examine the relationships among the 11 community functional parameters considered in this study, both within and between the above-ground and the below-ground parameters.

Intraspecific variability versus species turnover

To infer the relative magnitude of intraspecific variability versus species turnover on changes in aboveground community functional parameters, we used the method recently proposed by Leps et al. (2011). For this purpose, we calculated two types of CWMa: (i) ‘specific’ average traits, using trait values of each species within each plot (i.e. taking into account intraspecific – interplot – variability); and (ii) ‘fixed’ trait values, using mean trait values of each species along the whole resource gradient (i.e. site-independent trait values). The variation of ‘specific’ trait values across environments can be caused by both species turnover and intraspecific trait variability, whereas the response of community trait averages using only ‘fixed’ trait values is solely affected by changes in species turnover. To estimate the pure effects of the intraspecific variability, we computed a new community parameter based on the differences between ‘specific’ and ‘fixed’ average traits, according to Leps et al. (2011). We explored ‘CWMa – environment’ linkages for the three types of community parameters (‘specific’, ‘fixed’ and ‘intraspecific’ variability) using the same statistical model approach presented above. To quantify how much variability is accounted for by each individual component (species turnover or intraspecific variability), we used the method based on the Sum of Squares (SS) decomposition (see details in Leps et al. 2011). Since the fixed and intraspecific effects do not always vary independently, we also considered the effect of covariation by calculating SS of covariation as followed: covSS = SSspecific−SSfixed−SSintraspecific variability.

Results

Variation of environmental factors

Environmental factors were highly variable among the 12 selected plots (Appendix S1). For example, soil depth ranged from 0.19 to 0.95 m, soil water storage (SWS) varied between 6 and 147 mm, and nitrogen nutrition index (NNI) from 20 to 49%. Soil water availability varied strongly with depth. Thus, the deeper layers (60─70 cm) were moister throughout the sampled period; intermediate (30–40 cm) and superficial (0–10 cm) layers showed similar values during the rainy season (spring), but the decrease in water content was more pronounced in the upper 10 cm as water input by rainfall was diminishing (Appendix S2).

We found significant correlations among several environmental factors (Appendix S3). On the one hand, deeper soils showed higher water content (SWS) and were less limited in nitrogen available for plants (higher NNI). However, after conducting the false discovery rate test, SWS and NNI were not significantly correlated (Appendix S3). On the other hand, a higher content in clays led to lower values of pH and consequently to a higher cation exchange capacity (Appendix S3).

Functional community structure along resource gradients

Both above- and below-ground parameters were affected by three main environmental factors: soil depth, soil water availability and nitrogen supply (see Table 3).

Soil depth was positively related to plant reproductive height, appearing as the best explicative predictor of this whole-plant attribute (Fig. 1a and Table 3). In addition, soil depth played a secondary role on some leaf and root parameters, being positively related with SLAcwm and rooting depth and negatively with leaf thicknesscwm and δ13Ccwm (Table 3).

Figure 1.

Relationships between the main analogous above- and below-ground community parameters and the best-supported environmental predictors. Dashed lines represent the best-fitted models (see Table 3).

Soil water storage was positively correlated with plant rep. heightcwm and SLAcwm (Fig. 1c and Table 3) and negatively with δ13Ccwm, reflecting higher investment in photosynthetic leaf surface per unit of leaf mass and lower water use efficiency under improved soil water conditions. Communities on moister soils developed a deeper root system (i.e. higher values of 95% rooting depth) than communities growing on drier soils (Fig. 1b and Table 3).

Nitrogen limitation was the best predictor of many community functional parameters involved in the acquisition–conservation trade-off (Table 3). Communities that were less N limited showed higher values of SLAcwm and LNCcwm (Fig. 1c and Table 3), lower values of leaf thicknesscwm, leaf and root dry matter content (LDMCcwm and RDMCwsp, respectively; Fig. 1e,f), and a lower proportion of biomass allocated to roots (RMF). In addition, some below-ground parameters – commonly related to root foraging ability – were influenced by the C/N ratio. This soil factor, which can be considered as an indicator of quality of the organic matter (Heal, Anderson & Swift 1997), was positively correlated with SRAwsp and negatively with TMDrwsp (Fig. 1d and Table 3).

Links among community functional parameters

Regarding the above-ground parameters, SLAcwm and LNCcwm were positively correlated among them, and both of them decreased linearly with increasing LDMCcwm, leaf thicknesscwm and δ13Ccwm (Table 4). Similarly, for the below-ground parameters, communities with lower values of SRAwsp showed denser tissues in roots (i.e. higher values of RDMCwsp and TMDrwsp; Table 4).

Table 4. Summary of the models analysing the relationships among all the community functional parameters (‘specific’ averages) considered in this study. The R2 values, the signs of the relations (positive or negative) and the best models (LIN, linear; EXP, exponential; MM, Michaelis–Menten) are showed for all the possible combinations. The strength of empirical support (based on the absolute magnitude of the differences in AICc with the null model) is indicated as follows: *2 ≤ ΔAIC ≤ 5; **5 ≤ ΔAIC ≤ 10; ***10 ≤ ΔAIC
 SLAcwmLDMCcwmLeaf thicknesscwmLNCcwmδ13CcwmRMF95% Rooting depthSRAwspRDMCwspTMDrwsp
Plant rep. heightcwm0.230.020.270.250.130.39* [−,LIN]0.260.160.150.24
SLAcwm0.76*** [−,LIN]0.63** [−,LIN]0.72*** [+,MM]0.64** [−,LIN]0.72** [−,LIN]0.75*** [+,LIN]0.59** [+,LIN]0.76*** [−,LIN]0.41* [−,LIN]
LDMCcwm0.44* [+,LIN]0.40* [−,LIN]0.60*** [+,LIN]0.58** [+,LIN]0.55** [−,LIN]0.330.59** [+,LIN]0.15
Leaf thicknesscwm0.62** [−,LIN]0.42*** [+,LIN]0.59*** [+,LIN]0.51* [−,EXP]0.310.76*** [+,EXP]0.17
LNCcwm0.230.67** [−,LIN]0.270.40* [+,LIN]0.62** [−,LIN]0.26
δ13Ccwm0.47* [+,LIN]0.70*** [−,LIN]0.46*** [−,LIN]0.38*** [+,LIN]0.21
RMF0.45* [−,LIN]041* [−,LIN]0.75*** [+,LIN]0.30
95% Rooting depth0.260.320.16
SRAwsp0.44* [−,LIN]0.86*** [−,EXP]
RDMCwsp 0.35

We also found multiple relationships with better empirical support than the null model when we compared leaf and root parameters (Table 4). First, SLAcwm was correlated with all the below-ground parameters considered in this study. As an example, SLAcwm was positively and linearly related with its equivalent below-ground parameter, the specific root area (Fig. 2a). Second, LDMCcwm and leaf thicknesscwm, two leaf parameters commonly associated to a conservative resource-use strategy, were positively correlated with RDMCwsp (Fig. 2b) and biomass allocation to roots (Table 4). Finally, models including δ13Ccwm and most of the below-ground parameters had strong empirical support; thus, communities with high water use efficiency (high δ13Ccwm) allocated more biomass to shallow roots, which showed low SRAwsp and high RDMCwsp (Table 4).

Figure 2.

Links between analogous leaf and root parameters at the community level: specific leaf area (SLAcwm) versus specific root area (SRAwsp; panel A), and leaf dry matter content (LDMCcwm) versus root dry matter content (RDMCwsp; panel B).

Intraspecific variability versus species turnover

Our results indicate that changes in community above-ground parameters along the explored resource gradients were mainly promoted by changes in species composition but also by the intraspecific variability of trait values.

Decomposition of total variability in individual components demonstrates that the among-plot trait variation caused by species turnover was much higher than that due to intraspecific variability (Fig. 3). The relative importance of each source of variability differed slightly among the six CWMa considered in this study. Thus, intraspecific variability comprised a relatively important part of variation in δ13Ccwm and LDMCcwm (particularly along the NNI gradient), while it was practically negligible for the other four traits (especially for leaf thicknesscwm; Fig. 3).

Figure 3.

Decomposition of the total variability (i.e. ‘specific’ averages) in the six above-ground traits studied along the main environmental gradients (soil depth, SWS and NNI). Black part of the columns corresponds to species turnover (‘fixed averages’); open part to intraspecific variability effects and grey part denotes the effect of covariation between both sources of variability.

The total variation was strongly increased by a positive covariation between turnover and intraspecific variability effects in most of the CWMa, except for LNCcwm and leaf thicknesscwm for which covariation was sometimes negative (Fig. 3).

The larger relative importance of species turnover on CWMa variation was also supported by the fact that species composition substantially varied throughout the 12 experimental plots. As an example, the frequency of Bromus erectus increased with soil depth and resource availability, whereas Festuca christiani-bernardii was more abundant in shallow and resource-poor environments (Table 1).

Discussion

Community functional structure along resource gradients

The studied rangeland communities showed large differences in trait composition in response to field resources over very short distances. Most above- and below-ground community parameters considered in this study were driven by three main factors: nutrient limitation, soil water availability and, to a lesser extent, soil depth (which was strongly correlated with the other two factors).

Nutrient limitation

In spite of the widely recognized importance of water as one of the main limiting resources for plants in Mediterranean ecosystems, most of the community functional parameters considered in this study were more dependent on nitrogen limitation than on soil water availability. We recognize that, since both soil resources were positively correlated, our approach does not allow us to isolate completely pure resource effects on community functional structure. In fact, covariation of resources is a known difficulty in explaining the role played by individual resources under natural conditions (García et al. 2006). However, the fact that models including NNI had in general much stronger empirical support than models including soil water suggests that nutrient limitation was the major driver of the functional structure in the studied rangelands communities. Similarly, a recent study at a global scale demonstrated that several community leaf parameters, such as SLA or LNC, were more dependent on soil nutrient status than on other climatic variables responsible for soil moisture (Ordoñez et al. 2009).

Our data confirmed that those functional parameters determining the ability of the community to respond along natural gradients of soil fertility were mostly related to the acquisition–conservation trade-off (Chapin, Autumn & Pugnaire 1993; Grime 2001; Wright et al. 2004; Laliberté et al. 2012) and demonstrated that this trade-off governing plant resource economy can be applied at both the leaf and the root level. Communities growing in less nitrogen-limited soils had higher values of specific leaf area (SLAcwm) and greater amounts of leaf nitrogen per unit of mass (LNCcwm), two leaf traits commonly associated with rapid resource capture and high plant relative growth rate at the species level (e.g. Wright & Westoby 2001; Poorter & Garnier 2007). Opposite attributes characterized N-limited communities, which produced high-density tissues, at both the leaf (with higher values of leaf thickness and dry mass content) and the root level (with greater values of TMDrwsp and RDMCwsp). This suite of correlated traits, which can be grouped within a sclerophylly syndrome, has been interpreted as an adaptation for efficient nutrient conservation at the species level (e.g. Fonseca et al. 2000; Lavorel & Garnier 2002; Garnier et al. 2007). Our study shows that this syndrome can be scaled up and interpreted in the same way at the community level, with potential importance for inferring ecosystem functioning (e.g. Díaz & Cabido 2001; Lavorel & Garnier 2002).

Remarkably, communities were also influenced by soil C/N ratio, a factor generally used as an indicator of quality of organic matter. Thus, communities growing in soils with more recalcitrant organic matter (with higher values of C/N ratio) maximized below-ground resource uptake by increasing the volume and length of soil explored per unit of root mass (with higher values of SRAwsp and SRLwsp) (Comas & Eissenstat 2004). Other previous studies along fertility gradients suggested that N limitation favoured plants with higher values of length/mass ratio and root production (Roy & Singh 1994; Partel & Wilson 2001).

Interestingly, the above-discussed functional parameters were strongly affected by the nitrogen nutrition index (NNI) and/or the C/N ratio, whereas no models better than null were found when soil total N concentration or available phosphorus (both P Olsen and PNI) was tested. Three important conclusions can be derived from these results: (i) the stronger relative importance of soil N in comparison with soil P (Vitousek & Howarth 1991); (ii) the inadequacy of using total concentrations of soil nutrients, which do not give a satisfactory estimation of the nutrient pool actually available for plants (Vitousek & Howarth 1991; Ordoñez et al. 2009); and (iii) the importance of exploring NNI and C/N separately due to their different functional associations to community parameters.

Soil water availability

Communities were adapted to water shortage through a conservative resource-use strategy. As expected, drier conditions promoted the predominance of plant communities with low SLAcwm and higher values of δ13Ccwm, indicating a more efficient use of water (Farquhar, O'Leary & Berry 1982), consistent with other previous studies (e.g. Cunningham, Summerhayes & Westoby 1999; De Bello et al. 2009).

Regarding the below-ground component, communities developing in moister sites were deeply rooted. Similarly, Schenk & Jackson (2002) reported that maximum rooting depths tend to increase with soil water availability. In our system, this could be partially explained by the scarcity of deep soils in drier microsites, as indicated by the strong positive correlation between both environmental factors (soil depth and SWS). Rooting depth was better related to soil water availability than to the nutrition index. This may be because nutrient uptake usually takes place in upper soil layers (Jobbágy & Jackson 2001), whereas a relatively small proportion of roots in deeper soil layers can contribute substantially to water uptake (Jackson et al. 1996).

Based on the current changes observed in these functional parameters along the explored soil moisture gradient, we could suggest potential shifts in the functional structure of the community under future environmental scenarios (Suding et al. 2008). In Mediterranean rangeland communities, where soil water availability seems to be one of the main ecological factors that drive functional trait composition, it could be expected that the drier soil conditions predicted by global change models promote the predominance of plant communities with a conservative resource-use strategy. Thus, plants with reduced and sclerophyllous leaves, shallow root systems and a more efficient use of water will likely be favoured against fast-growing species characterized by a combination of high SLA, low-density tissues and deep root systems.

Links among community functional parameters: evidence for A plant community economics spectrum

The results from this study revealed parallel variations in leaf and root functional parameters at the community level. Thus, communities that produced thinner leaves with high surface area/mass ratios (SLAcwm) and large tissue N concentration also produced low-density roots with high area/mass ratios (SRAwsp), usually linked with efficient nutrient uptake and high relative growth rate (Reich et al. 1998a; Eissenstat et al. 2000; Wahl & Ryser 2000). Consequently, these plant communities characterized by a fast resource acquisition presented a more wasteful use of water, as indicated by their lower values of δ13Ccwm.

Interestingly, we also found a strong coordination at the community level between above-ground and below-ground components. In particular, SLAcwm was positively correlated with certain below-ground parameters involved in nutrient acquisition (e.g. SRAwsp, rooting depth) and negatively correlated with others commonly associated to resource conservation such as RMF and RDMCwsp. The positive correlation between SLAcwm and SRAwsp was consistent with recent studies (Reich et al. 1998b; Wright & Westoby 1999; Withington et al. 2006; Holdaway et al. 2011; Birouste et al. 2012), but contrast with others where both types of traits, measured at the species level, were unrelated (Tjoelker et al. 2005; Laughlin et al. 2010) or even negatively correlated (Lambers & Poorter 1992). These results would imply promising perspectives for functional ecology as they enable us to predict several community root parameters, which are difficult to measure in natural conditions, from easily measurable community leaf parameters. As an example, our findings showed that tissue dry matter content, both in leaves (LDMCcwm) and roots (RDMCwsp), could be used as a consistent leaf and root response trait to nitrogen limitation. Thus, nutrient-limited soils promoted the predominance of communities with functional parameters commonly associated to a conservative resource-use strategy, having high-density tissues at both the leaf and the root level. Similarly, Craine et al. (2001) demonstrated that herbaceous species occupying frequently low-N environments had greater tissue density in leaves and roots, which might be associated to higher nutrient use efficiency.

The high coordination between above- and below-ground components supports our initial hypothesis that leaf and root parameters involved in the resource acquisition–conservation trade-off tend to co-vary. These findings indicate that the paradigm on leaf economic spectrum can be extrapolated to the root system and provide evidence for a plant community economics spectrum, thereby generalizing the plant economics spectrum (sensu Freschet et al. 2010) to the community level.

Intraspecific variability versus species replacement

Another significant contribution of this study is the exploration of the relative importance of the two main sources of variability in community functional parameters along field resource gradients. Soil depth and resource limitation acted as environmental filters controlling the functional structure of communities by means of two main underlying mechanisms: (i) promoting the turnover of species with highly contrasting trait values (interspecific variation) and, to a lesser extent, (ii) forcing a strong intraspecific variability (phenotypic and/or genotypic plasticity) in several above-ground traits. Decomposition of total variability in individual components (species turnover and intraspecific variability) indicated that changes in most of the community functional parameters considered in this study were mainly due to species turnover. Thus, dry and nutrient-limited soils promoted the predominance of species with a conservative resource-use strategy such as Festuca christiani-bernardii, whereas moist and fertile areas increased the frequency of fast-growing species characterized by a rapid resource capture (e.g. Bromus erectus).

Interestingly, our results indicate that intraspecific variability also played a significant role on the community functional structure in response to soil depth and resource availability. The relative importance of this source of variability was more pronounced in several above-ground traits, such as LDMCcwm or δ13Ccwm, indicating that they are subjected to strong dependence on soil resources. This high intraspecific variability likely enabled certain species to survive, grow and reproduce under higher variety of environmental conditions (Joshi et al. 2001; Byars, Papst & Hoffmann 2007). In spite of the higher relative importance of species turnover compared with the intraspecific variability, our findings suggest the importance of considering this source of variability in several above-ground traits for a better understanding of plant strategies and ecological patterns along environmental gradients, as proposed by recent studies (Albert et al. 2010a,b; Messier, McGill & Lechowicz 2010).

The novel information provided by this study contributes to our understanding on the drivers of community functional structure and species assembly in Mediterranean rangelands, soil water and nutrient limitation promoting strong changes in functional community structure along the explored field resource gradients, mainly due to species turnover. In addition, we provide evidence for a plant community economics spectrum driven by nutrient and water availability, based on a strong coordination between above- and below-ground components in these resource-limited communities. Our community-level approach constitutes a useful tool for identifying and categorizing plant communities based on their functional attributes, predicting their responses to changes in environmental conditions as well as inferring ecosystem properties of ecological and agricultural importance (e.g. Díaz & Cabido 2001; Lavorel & Garnier 2002; Ansquer et al. 2009).

Acknowledgements

We thank the technical staff of the experimental station ‘INRA La Fage’ for the facilities and support provided during the fieldwork. We are tremendously grateful to Guillaume Coulouma for the pedological characterization of the selected plots and to Virginie Pons for her invaluable laboratory and field assistance. We also thank Ezequiel Zamora, Alison Munson, Elena Kazakou, Olivier Flores and Baptiste Testi for their help in field sampling and plant measurements. This study was supported by a postdoctoral MEC contract to I.M.P.R. and by the DIVHERBE project (French National Programme ECOGER). This is a paper from GDR 2574 (‘TRAITS’) of CNRS.

Ancillary