SEARCH

SEARCH BY CITATION

Keywords:

  • Biodiversity;
  • Ecosystem function;
  • Functional effects traits;
  • Plant functional groups;
  • Temperate grasslands

Abstract

  1. Top of page
  2. Abstract
  3. Introduction
  4. Suggested methodology
  5. Example: The DIRECT experiment
  6. Discussion
  7. Acknowledgements
  8. Author contributions
  9. References
  10. Supporting Information

Aim

Biodiversity–ecosystem function (BDEF) experiments commonly group species into arbitrary a priori functional groups, e.g. the grass/forb/legume (GFL) classification. As a result, the causes of functional group diversity effects are often poorly understood. This paper presents a new process that uses functional trait data to create customized plant functional groups that can be tailored to address specific questions. This method is illustrated throughout with an example taken from a temperate mesotrophic grassland in southern England.

Location

Silwood Park, Berkshire, UK.

Methods

The method described applies divisive hierarchical cluster analysis to plant functional trait data (from either field or greenhouse conditions) in order to cluster species into a user-specified number of groups. In our example, this was done using unweighted traits with clear links to C and N cycling. To ensure between-group variance had been maximized, we used a linear discriminant analysis. ANOVA should also be used to compare the mean trait values of groups, in order to make specific hypotheses regarding the effect that each group has upon ecosystem functioning. We compared the resulting groups with the GFL classification to see which was more likely to deliver functionally distinct groups.

Results

The resulting groups had discrete functional characteristics, so simple hypotheses could be formulated. These groups also appeared to show stronger trait value differences than the GFL classification. Results from the experiment demonstrate that hypothesized removal effects on function were supported, thus validating our approach.

Conclusions

The method described is applicable to a wide range of communities and is able to recognize functionally distinct groups of species. General use of this approach could result in a more mechanistic understanding of biodiversity–ecosystem function relationships as it can establish experimentally validated links between functional effects traits and ecosystem functioning.


Nomenclature
USDA PLANTS Database

(2010)

Introduction

  1. Top of page
  2. Abstract
  3. Introduction
  4. Suggested methodology
  5. Example: The DIRECT experiment
  6. Discussion
  7. Acknowledgements
  8. Author contributions
  9. References
  10. Supporting Information

Concerns over the implications of global biodiversity decline for ecosystem functioning and associated services has led to a proliferation of biodiversity and ecosystem function (BDEF) experiments, particularly in grassland ecosystems (Hooper et al. 2005; Cardinale et al. 2006; Hillebrand & Matthiessen 2009). These have generally found that the number of species and functional groups influences function, but mechanistic understanding of this relationship is incomplete, with the species traits that influence function often remaining unidentified.

Another drawback of traditional BDEF experimental designs is that it is often impractical to establish large-scale, plot-based experiments containing all possible combinations of species from a given species pool. Therefore, there are both practical and scientific advantages to simplifying species richness by reducing it to functional groupings based upon morphological, physiological and/or phenological traits (Díaz & Cabido 1997; Lavorel & Garnier 2002; Petchey 2004). While there have been many calls for a universal protocol for both trait measurement and the classification of functional groups (Lavorel et al. 1997; Tilman 1999; Cornelissen et al. 2003; Naeem & Wright 2003; Harrington et al. 2010), species are distributed in a continuous and multivariate trait space in which there are relatively few discrete clusters (Craine et al. 2002; Wright et al. 2004). Defining this trait space is therefore a major challenge in ecology, and data are not yet available for a global functional classification (Kattge et al. 2011). Therefore, in the meantime, it is useful to develop our understanding of how traits and suites of traits affect functioning, and to advance this understanding the tailoring of methods to specific research questions is often required. For example, in grassland studies (the modal system for BDEF research) many field experiments have used the grass/forb/legume (GFL) classification to categorize species (Hooper 1998; Wright et al. 2006). This classification is convenient but lacks a sound mechanistic basis with respect to many functions, and makes no concession to the high likelihood of substantial within-group variation in functionally important traits, or the possibility of strong trait overlap between groups. We therefore feel that there is a need for groups based upon specific measured characteristics of the species present in the experimental system, with variation in between-group trait properties that is higher than within-group properties (Lavorel & Garnier 2002).

An alternative approach to understanding BDEF relationships has been to use a combination of trait and species abundance measures to characterize the functional trait distribution of a community in situ. This is then correlated with measures of ecosystem function taken from these communities (Díaz et al. 1999b, 2007; Craine et al. 2002; Garnier et al. 2004; Lavorel et al. 2008). There have been a number of recent studies that demonstrate good correlations between functional effects traits and ecosystem functioning; for example, in landscapes where a shift in management intensity occurs (Díaz et al.1999b; Kleyer 1999; Cornwell & Ackerly 2009). However, a major problem with this approach is that it is difficult to isolate the effects of plant traits from those of co-varying environmental drivers of both traits and function (de Vries et al. 2012). Thus there are questions of baseline null hypotheses with transect studies, as Yan et al. (2012) found when attempting to separate trait differences from environmental variation along a transect: using mean trait values as a null value leads to problems when trying to model intra-specific variation.

Here, we describe a new systematic process for the formulation of a priori functional effects groupings for use in BDEF experiments. This method combines the strengths of the two aforementioned approaches by allowing for the experimental manipulation of trait distributions in the field. The approach can therefore be used to investigate the mechanistic basis of relationships seen in traditional BDEF experiments, or to test whether trait–function relationships observed in correlative studies are genuinely causal. The approach presented advocates functional diversity manipulations that have a strong hypothetical link to changes in function. The method presented is therefore intended to provide mechanistic insight into how species possessing suites of trait values that commonly occur in natural systems affect a range of ecosystem processes. We feel that there is a place for bespoke functional groupings in large-scale, plot-based studies. This has worked well in studies such as the Jena experiment (Roscher et al. 2004), but so far the methods used for grouping species have not been standardized, and a there has been little formal discussion of possible methods.

Our protocol also builds upon work describing the calculation of functional diversity measures (e.g. Petchey & Gaston 2002) to offer a set of groupings where trait values of each species are known, and mean trait values for each group are statistically different. Therefore, precise hypotheses can be made regarding ecosystem response to the removal of certain groups. We can then make informed inferences regarding the effect of functional group identity, which could be equally or more important than functional group richness. This capability is currently lacking in many BDEF experiments (Balvanera et al. 2006). An additional benefit of this approach is that if species composition is known then the functional diversity differences created by the experimenter can be treated as a continuous gradient of functional traits or functional diversity in analysis, allowing for relationships between these properties and function to be established.

We offer this process as a successor to the studies of Ellenberg et al. (1991) and Grime (1988), as it aims to group species into the most functionally homogeneous clusters possible, disregarding taxonomic and morphological associations. In doing so, we acknowledge that this approach is more suitable for explaining biogeochemical processes than functions based upon co-evolved species interactions (e.g. pollination) or interactions with ecosystem physiognomy (e.g. those involving habitat selection). We should also stress that the method is suitable for trait- and trait syndrome-based hypothesis testing but does not allow for the exploration of ‘idiosyncratic’ individual species effects (Díaz et al. 2007).

We illustrate our approach with an example of a field experiment investigating the modification of plant functional diversity on the N and water cycles of a grassland ecosystem affected by climate change. Our example focuses on plant traits and terrestrial ecosystems, but the process should be applicable to a wide range of systems. Nevertheless, tailoring of the approach will be required, and the approach presented here should be viewed as a set of guidelines and a discussion of trait grouping methodologies in the light of recent research rather than a fully standardized method.

Suggested methodology

  1. Top of page
  2. Abstract
  3. Introduction
  4. Suggested methodology
  5. Example: The DIRECT experiment
  6. Discussion
  7. Acknowledgements
  8. Author contributions
  9. References
  10. Supporting Information

Choosing traits

The first step in generating tailored functional effects groups is to define the functions of interest and select a group of functional traits that are hypothesized to drive or determine them (Hillebrand & Matthiessen 2009). While there are multiple definitions of the word ‘trait’, we define functional traits using the response–effect framework (Lavorel & Garnier 2002), which defines response traits as those that determine response to environmental conditions, and effects traits as those that affect ecosystem processes. This varies from the definition of Violle et al. (2007), which focuses on morpho-physio–physiological traits that impact ‘fitness’ of the individual, with cascading effects upon community performance and ecosystem functioning. From this, hypotheses from Violle's definition would be concerned with finding the optimum trait values that determine ‘community performance’, rather than to directly link trait distributions to ecosystem function using an experimental approach such as we advocate here. This is an alternative objective for BDEF studies, and many aspects of previous classifications could be incorporated, such as Raunkiaer (1934), Ellenberg et al. (1991) or Grime's C-S-R (Grime 1988).

It is unlikely that a large experiment would be performed to test established trait–function relationships, but to clarify this point an example would be the selection of traits relating to leaf nutrient concentration or toughness if the researcher was interested in explaining decomposability (Garnier et al. 2004; McLaren & Turkington 2010). A less well established relationship would be the linking of root traits to below-ground properties such as soil stabilization and rhizo-deposition rates. Correlation between many traits may be unavoidable, but direct correlation should be avoided, for example specific leaf area (SLA) and leaf dry matter content (LDMC) describe very similar characteristics, so only one should be included. Other correlations are frequent, such as between leaf N content (LNC), SLA and photosynthetic rate, but between them they will describe a large range of different processes, and do not directly measure the same aspect of function.

If a wide range of functions is to be measured and explained, more traits are needed to form meaningful functional groups. There is likely to be some overlap where a single trait could be linked to very different functions, e.g. if herbivory and C cycling were to be measured, it would be wise to include traits based upon leaf C:N ratios and secondary chemistry, which could explain both palatability and C inputs to the system. If traits are given equal weighting in cluster analysis, this will generate bias towards more correlated traits, so we recommend testing for correlations before commencing the cluster analysis. As a rule of thumb, we recommend against including two traits with an r-value of >0.7. Trait choice will also depend upon community type and the functions in question, but they should be practical to measure, with strong theoretical support for their ability to describe the relevant functions (Westoby 1998; Wright et al. 2006). This may involve a mixture of physiological, morphological and life-history traits (Díaz et al. 1999a, 2004).

There are several published examples of systematic evaluations of trait and process linkages that provide a good overview of the trait selection process (Klumpp & Soussana 2009; de Bello et al. 2010). Failure to identify groups closely linked to function is likely to result in a weak capacity to explain ecosystem processes. The functional diversity (FD) work presented by Petchey & Gaston (2002) indicates that if processes are measured with no consideration of the trait variation in species present, they are often poorly explained, and these authors question whether a single functional classification scheme can describe a wide range of functions. Creating many different functional effects classifications, and accordingly functional group manipulations to describe different responses, is likely to be unfeasible in a field experiment (although not in model simulations), so using a set of the most comprehensive traits possible, given the constraint that all are linked to functions of interest, is likely to be the most practical alternative. Trait selection may also be limited by the availability of trait measures; certain traits, such as SLA and leaf N content, have been measured extensively and are available for a large number of species (Kattge et al. 2011), and where resources are limited these may need to be used instead of the ‘ideal’ trait to avoid the need for laborious trait measurement programmes. This will come at a cost to the experimenter's ability to formulate clear hypotheses about the functional consequences of removal.

Defining the species pool

The next step in the process is to delimit the species pool from which the groups are drawn. Deciding which species to include can be problematic, particularly if the study system is open to invasion. It is further complicated by the fact that the majority of field experiments using these methods will undergo successional change. Perhaps the simplest and most cost-effective method is to begin with all the species present in the field site at the beginning of the study and to classify these. Then, when new species invade, the experimenter should measure their traits and use this information to add them to the original groups on a similarity basis. Alternatively, a randomly selected pool of 18–32 potential species from the habitat could be included, which has been favoured by traditional BDEF practitioners (Tilman et al. 1996; Hector et al. 1999; Roscher et al. 2004; Fornara & Tilman 2008), including a certain number of representatives from grass and forb groups. However, the realism we advocate in this method is likely to be lost. In cases where communities are artificially assembled it may be sensible to use species that associate frequently and that are typical of the site environmental conditions, e.g. the plant species found in a class of the UK's National Vegetation Classification (NVC; Rodwell 1992).

Obtaining a traits database

Trait data can be obtained from database or literature sources (Fitter & Peat 1994; Kleyer et al. 2008; Royal Botanic Gardens, Kew 2008; USDA 2010; Kattge et al. 2011), or measured directly. Database trait values can be very useful as they are drawn from several studies, thus potentially increasing the reliability of the estimate, and can save a lot of time and expense (Lavorel et al. 2008). However, there are potential drawbacks of using such values, particularly the potential lack of standardization associated with multiple data contributors. While standardized protocols have been suggested (Grime et al. 1997; Cornelissen et al. 2003), standardization is still generally lacking in plant trait measurement studies, which differ in growth conditions and substrate, season of measurement, genotype and trait measurement protocols. In addition, not all species are represented in the databases. In many cases it may be advantageous for the experimenter to create their own traits database using local plants and soil, and keeping growth conditions as similar to the site as possible.

A second problem of using trait values is that a single trait value does not represent the full intra-specific trait variability (ITV) of a species, or its expression in field conditions (Albert et al. 2010a). This extremely complex problem has concerned ecologists for some time (Bolnick et al. 2011) and is discussed in detail elsewhere (Albert et al. 2010a,b, 2011). In communities of similar or highly plastic species, ITV could be a major barrier to the success of our approach, and other functional traits approaches, and there is no simple solution to dealing with ITV at present (e.g. a set of rules in which trait values are adjusted according to their context). However, in an analysis of 2.6 million trait data entries drawn from 69 000 species and using a range of different methodologies, it was found that 60–98% of variation between functional traits was inter-specific, meaning that species trait values should be reliable predictors of species effect, at least in communities of dissimilar species. The only traits where ITV exceeded BTV (between species trait variation) were leaf P content, leaf N per unit area and photosynthesis per unit leaf N content (Kattge et al. 2011). Accordingly, these traits should not be used as a basis of functional group classification unless their variation and plasticity is used as a trait in its own right (e.g. leaf P trait plasticity may explain temporal variability in soil phosphate availability).

Attempting to capture the full range of a species' trait distribution in the field could lead to ‘noisy’ data, so the worker would be better advised to measure several individuals of species in constant conditions (Díaz et al.1999b). A key tenet of trait measurement is that traits are likely to vary across abiotic gradients, and so conditions for growth should be kept as similar to the field site(s) as possible (Díaz et al. 1999b; Lavorel & Garnier 2002).

Using divisive hierarchical cluster analysis to create functional effects groups

Once a list of trait values for all species has been established, the next step is classification into functional groups. There are many possible ways of clustering species, e.g. using an ordination method such as principle components analysis (PCA) or multidimensional scaling (MDS) to assign axis values to the species, which are then used in cluster analysis. While the classification of functional groups may be best achieved via an unconstrained approach such as PCA, this may not be desirable in the case of experimental manipulations of functional diversity in which there are constraints of space, cost and time. To identify the effects of functional groups and their interactions, combinatorial designs featuring all possible combinations of functional groups are required (such as BIODEPTH, Hector et al. 1999). An efficient means of ensuring that the number of functional groups derived is that desired, is to use divisive hierarchical cluster analysis (Jongman et al. 1995; Shaw 2003). This can allow for a specified number of clusters of similar values to be identified within multivariate data sets. The alternative, agglomerative cluster analysis, is a ‘bottom-up’ method, where pairs of similar species are grouped, then the pairs are joined with other pairs, and so on. However, this will uniformly result in even-numbered groups (Roleček et al. 2009). Cluster analysis also creates a dendrogram (Fig. 1), allowing visualization of relationships between species. If assumptions about the relationship between traits and function are correct, and trait values are representative of those expressed by species at the study site, then the groups derived from this analysis should have more discrete and predictable effects on ecosystem function than random species assemblages.

image

Figure 1. Functional dendrogram of species found in Silwood Park, Berkshire, UK, created using functional effects traits in a divisive hierarchical cluster analysis. Species added later using dissimilarity indices are placed next to their group. The dashed lines divide the species into functional groups.

Download figure to PowerPoint

In order to create groups, trait means for each species should be calculated. Cluster analysis can be carried out using S-PLUS 6.0 (Insightful, Gothenburg, SE) or the freeware R2.13.1 (R Foundation for Statistical Computing, Vienna, AT; Appendix S2). TWINSPAN, one of the commonest programs used for clustering, is not recommended for this purpose as it only clusters into powers of two (2, 4, 8, 16, etc.), making it inappropriate if an uneven number of groups is desired for the experimental design. Ordinal variables should be treated as ranked factors. Weighting of traits is somewhat contentious, and there should be strong justification for using it. For example, if responses concerned with N cycling are the focus of the experiment, traits such as leaf N content or N fixation capacity could be double weighted (Petchey & Gaston 2002; Roscher et al. 2004), although this could lead to groups that exclusively describe that trait. This may be a problem in some communities, particularly since legumes do not always fix N (Serraj et al. 1999). Another weighting option is to weight down traits that show a stronger ITV, thus limiting ITVs capacity to interfere with the formation of functionally distinct groupings. Similarly, ITV could be dealt with by assigning species the same value if their values are found to be not significantly different in a preliminary analysis.

As clustering is based upon dissimilarity matrices, an appropriate distance measure must be chosen. The simplest (and default in most packages) is Euclidean distance, which calculates the straight-line distance between the trait combinations of every species in Cartesian space (n-species dimensions), and creates a matrix of these distances (MjM Software Design, OR, US). Euclidean distance emphasizes outliers, so the data should be standardized to have a mean of 0 and variance of 1. Most other distance measures have fairly rigid requirements and compute data in less intuitive ways. For example, Sørenson (Bray–Curtis index) dissimilarity gives less weight to outliers, but it is recommended for ecological community data as it gives proportions based on overlap of two communities, so is better for describing community turnover (MjM Software Design). Some other distance measures do not use straight-line methods, such as the Manhattan distance. This uses a ‘rectilinear’ pattern, moving in a grid-like fashion across Cartesian space. It is possible to use it for clustering, but less intuitive than Euclidean distance, which takes the shortest path. One other measure that is commonly used with clustering is Ward's distance. This is based upon finding small sums of squares of the distance between individuals in Cartesian space (an ANOVA approach); while it is not very sensitive to outliers, it is widely used and a valid alternative to Euclidean distance (Shaw 2003). Finally, divisive clustering analysis is commonly used with categorical data sets, although continuous data or a mix of the two can also be used. Plant trait values are likely to be mostly continuous, and must be standardized prior to the analysis.

When carrying out this analysis in S-PLUS, it requires the user to set the number of groups required before clustering begins. This is likely to be desired where the researcher seeks to limit the number of groups to a manageable number, e.g. in a combinatorial design. In R, the grouping occurs in a post-hoc fashion. Decision between these methods is important: the number of groups formed by the analysis will have a large impact on hypothesis formulation and the results obtained from the study.

Establishing the groups in the field

The optimal method for generating an experimental functional diversity gradient from the groups identified will depend upon the type of species in question and the functions to be studied, i.e. whether species are removed from an existing community or whether an artificially assembled community is established. Clearly, it is not feasible to rapidly establish late-successional ecosystems comprised of species with long generation times, so here species removals or removal simulations (e.g. Bunker et al. 2005) are the only options. When our method is employed, weeding to establish functional groups should cause a directional change in community-level trait values and their distribution, although shifts in trait distribution can only be expected for traits that show significant differences between groups. In other cases, niche space may be made available by the removal or exclusion of a functional group, and the remaining species that are most similar to those removed are likely to utilize this space and increase disproportionately. Where this occurs, the observed change in trait values following functional group exclusion may be less than expected. However, where very discrete functional groups are present, such a shift is unlikely and areas of niche space will remain unused and consequently affect function. When trait data are combined with a survey of species abundances in experimental units, a number of metrics, including functional diversity (Petchey & Gaston 2002), community-weighted means (Garnier et al.2004; Lavorel et al.2008) and dissimilarity measures (Hillebrand & Matthiessen 2009), can be calculated and used to estimate the impact of the functional group manipulation on community-level functional properties. These can also be used as explanatory variables in statistical models describing function. This will offer a more mechanistic understanding of which traits drive particular functions.

Assigning new species to existing functional groups

In open ecosystems, new species may colonize the experiment or emerge from stasis (e.g. seed bank), which were not considered when originally delimiting the species pool of the site. There are two options when a new species invades: the first is to remove the colonizing species; this is simple and may be preferable where the new species has functional properties that are distinct from the existing groups (e.g. the entry of an N fixer into a system that did not contain them), but risks reducing ecological realism (a key challenge in BDEF research; Naeem 2008), e.g. the species may be a potential new dominant. The other option is to obtain trait data for the new species and allocate data to the most appropriate of the existing functional groups. An invader that is functionally distinct would result in the convergence of group trait means (although in multidimensional trait space this is less likely). We offer a means of testing which group the invader would be best aligned with, but in some cases, e.g. if it is very different from all extant species, the experimenter should use their discretion as to whether the species should be excluded altogether.

To assign new species to functional groups, we suggest that a mean trait value is calculated for each functional group, to compare with trait values of the new species. This a posteriori integration into groups can be achieved by using dissimilarity indices to add the new species to the dendrogram. Dissimilarity values follow the same principles as a cluster analysis; again Euclidean distance measures are recommended to arrange the data in multidimensional space. Dissimilarity indices can be calculated using R2.13.1, choosing Euclidean distances and standardizing the data as before (Appendix S2). Ordinal variables should be converted to a continuous format by averaging values across species and labelling them as numeric, not factor, values. The new species is then assigned to the group with the lowest dissimilarity value, and the new trait means calculated. The mean trait value for the functional group must be adjusted with each new species.

Validation of the functional effects groups

An important criticism of hierarchical cluster analyses is that there is no measure of whether the groups identified are the most effective combination to explain function, or whether they are statistically different from one another. For this, we recommend using linear discriminant analysis (LDA) in the MASS library of R. LDA is an a posteriori method of verifying that each species has been allocated to the most appropriate group. LDA tests the within-group covariance matrix of standardized traits, and generates a probability of each species being in the most appropriate group, i.e. it generates a percentage similarity of the species mean trait values to each of the group mean trait values. The highest probability generated is taken to be the group the species belongs in. If the analysis finds that almost every species is appropriately categorized, this is strong justification for the groupings. High percentages of correctly allocated species in the LDA confirm that the functional groups are as discrete as possible. If there are some species that the LDA suggests are misclassified, close inspection is needed; it is possible that the species was classified on the basis of a single trait. It is the user's discretion whether this trait is particularly important to the functions of interest, if not, the analysis could be repeated with the trait removed and results re-evaluated (Appendix S2).

Before hypotheses about the effects of functional group removal can be formulated, it is important to check that there are quantifiable differences between the trait means of the groups. If there are no clear differences between the groups, this suggests that there is too much functional divergence or there are too many groups, and the outcome of manipulating these groups in a system could be confounding or inconclusive. In some other widely used classifications, such as GFL, functional groups may not exhibit clear trait differences, and so increased functional group richness may not always result in increased functional diversity; in some cases it may even decrease it (Naeem & Wright 2003). Simple analyses such as one-way ANOVA allow for evaluation of differences in trait means across functional groupings. Experiments evaluating the effect of traits upon ecosystem functions should be as realistic as possible in order to have real-world relevance (Hillebrand & Matthiessen 2009). However, if there are some species that are fully intermediate between groups or form their own distinct single species group, it is up to the experimenter to decide whether to forego realism and exclude these from the experiment.

Generating hypotheses about functional group exclusion

Identification of functional group differences, coupled with hypothesized trait–function relationships, enables predictions about the consequences of functional group removal. Despite these hypotheses, it is possible that the removal of a functional group will not change functional properties, e.g. if the group removed was rare, or if trait expression in the remaining species shifted to encompass the trait identity of the lost group (Walker et al. 1999).

To measure the realized trait value differences across treatments in the field, we suggest using community-weighted means (CWM; Garnier et al. 2004). A CWM weights species trait values by relative field abundances (e.g. taken from quadrat measures) to give a plot-level weighted mean trait value. For an integrated view of treatment effects on the distribution of all traits measured, there are various metrics including Rao's Q and FD (Rao 1982; Petchey & Gaston 2002; Botta-Dukát 2005). These metrics can then be related to ecosystem process measures (Garnier et al. 2004; Díaz et al. 2007).

Example: The DIRECT experiment

  1. Top of page
  2. Abstract
  3. Introduction
  4. Suggested methodology
  5. Example: The DIRECT experiment
  6. Discussion
  7. Acknowledgements
  8. Author contributions
  9. References
  10. Supporting Information

We now illustrate this approach with an example, the DIRECT experiment (DIversity, Rainfall and Elemental Cycling in a Terrestrial ecosystem). DIRECT was set up in 2008 on mesotrophic grassland in southeast England. We created tailored functional groupings to assess the role of plant functional diversity in ecosystem response to climate (precipitation) change, focussing on the cycling of C, N and water and their temporal dynamics as the ecosystem functions of interest (after Pacala & Kinzig 2001). We used an existing species list from Silwood Park, Berkshire, UK (Crawley 2005), where our field site is located (for details of growing conditions and trait protocols, see Appendix S1) to delimit the species pool. All species names conform to the USDA nomenclature (USDA, NRCS 2010). The experiment began in summer 2008 and consisted of a rainfall treatment and a functional group composition treatment (Fry et al. 2013). There were 56 plots in total (two climate treatments × seven combinations of plant functional groups × four blocks). Most results of these treatment effects on ecosystem function are presented elsewhere (Fry et al. 2013).

We selected eight traits with established links to C, N and water cycling and the temporal dynamics of these processes. Temperate grassland species have been extensively studied, particularly with regard to effects traits, with the result that there are comprehensive databases, standardized protocols and consistently observed (but not always experimentally evaluated) relationships between traits and function (Grime et al. 1997; Cornelissen et al. 2003; Kleyer et al. 2008; Kattge et al. 2011). We used specific plant area (plant area/mass; SPA – similar to SLA), leaf N content (LNC), N fixation ability (as a binary factor), leaf stomatal conductance rates (Gs), above- and below-ground biomass (AGB and BGB), leaf photosynthetic rate (A) and perennation (a ranked variable ranging from annual (1) to perennial (5) based on classifications of the USDA trait database (USDA, NRCS 2010). Our justification for trait choice was that in herbaceous and grass species with non-woody stems, SPA is a proxy for stem and leaf thickness, and high values could be associated with low decomposability, high relative growth rate and high leaf construction costs (Reich et al. 1998; Keddy et al. 2002), thus potentially explaining many other processes (Wilson et al. 1999). Relative tissue thickness compared with area also correlates with photosynthetic capacity, because thicker, smaller leaves need higher light levels to achieve the same amount of photosynthetic C fixation (Evans & Poorter 2001). LNC is closely linked to litter quality, and therefore positively correlated with decomposability and soil nutrient turnover (Cheng et al. 2010). Nitrogen fixation plays a key role in ecosystem N inputs and availability (Hartwig 1998). Gs is a potential proxy for ecosystem evapotranspiration rate and water-use efficiency (WUE), which may indicate the capacity to preserve water under soil moisture stress (Hsiao & Acevedo 1974). AGB and BGB describe potential C allocation and are associated with plant nutrient and water uptake (McConnaughay & Bazzaz 1991; Díaz et al. 2004; Fornara & Tilman 2008; de Bello et al. 2010). A can be considered an indirect measure of C fixation (Gilmanov et al. 2009). Perennation is likely to correspond to the temporal dynamics of the processes outlined above.

We tested for correlations between traits using Pearson's product–moment correlation coefficient (Table 1), then carried out a hierarchical cluster analysis as described above using S-PLUS 6.0, selecting three groups as this was the number considered to be manageable in a replicated combinatorial experiment (seven combinations), given the resources available. The groups returned consisted of perennial grasses, forbs and legumes (group 1), caespitose (bunch) grasses and tall forbs (group 2), and annual forbs, grasses and legumes (group 3). The species were closely clustered on LDA ordination axes (Fig. 2), indicating that there is strong similarity within groups, particularly groups 1 and 3. Only ~10% of species had a < 50% agreement with the LDA, and so appeared misclassified. The removal of each trait in turn did not create a change in accuracy of group allocation (Table 2), suggesting that no single trait had undue weight in the analysis.

Table 1. Correlations of traits using Pearson's product moment coefficient, for 57 grassland species. AGB, above-ground biomass; BGB, below-ground biomass; SPA, specific plant area; LNC, leaf N content; Gs, stomatal conductance; A, photosynthetic rate. Perennation and legume are not presented because of they are categorical/binary variables that are inappropriate for correlation analysis
 BGBAGBSPALNCGs
BGB     
AGB0.42    
SPA0.410.14   
LNC0.240.130.29  
Gs0.110.090.030.07 
A 0.030.40.050.120.56
Table 2. Functional group validation using linear discriminant analysis to test the robustness of the species allocation using divisive hierarchical cluster analysis. Percentages describe the proportion of species that have been allocated to the ‘correct’ group by the discriminant analysis. The removal of each trait individually determines whether any trait has a disproportionate influence on the cluster analysis and subsequent groupings. AGB, above-ground biomass; BGB, below-ground biomass; SPA, specific plant area; LNC, leaf N content; Gs, stomatal conductance
Trait group removedGroup 1Group 2Group 3
All Traits Present92%78%89%
AGB80%66%89%
BGB92%67%82%
SPA84%78%82%
LNC84%67%82%
Gs80%67%86%
Photosynthetic rate84%100%82%
Perennation72%67%93%
image

Figure 2. Ordination plot showing the percentage of species ‘correctly assigned’ to functional groups using linear discriminant analysis (LDA). Axis 1 explains 58.3% of the variation, axis 2 explains the remaining 41.7%.

Download figure to PowerPoint

One-way ANOVAs revealed significant differences between the mean trait values of all three groups (Fig. 3). For example, groups 1 and 3 had similar AGB and BGB, but differed in their values for SPA. The ANOVAs justified the group 2 classification, which appeared rather dispersed on the LDA ordination (Fig. 2), by illustrating that high AGB and BGB, and low LNC distinguish this group from the others (Fig. 3). Group 1 had the highest average SPA, so the loss of this group would be expected to result in lower community light capture and slower decomposition rates due to the loss of thin leaves with little structural tissue content. This is also the group with the highest proportion of perennial species, and so biomass stocks would be expected to be less variable across the year where it was present. Whilst most species in group 2 were caespitose (bunch) grasses, two herbs (Lapsana communis L. and Galium aparine L.) were also included, although no legumes were present. Species from this group had a high biomass both above- and below-ground and lower LNCs than the other two groups, so its elimination would be expected to increase the average quality of litter inputs and increase litter decomposition and N mineralization rates. Group 3 was mainly comprised of annual forbs, but also included numerous legumes and three annual grass species (Bromus sterilis L., Bromus hordeaceus L., Poa annua L.). This group had the fewest perennial species, and so communities containing them may see a high seasonal variability in standing biomass. Legumes were allocated to both FG1 (2/24 spp.) and FG3 (6/28 spp.), likely based upon their longevity.

image

Figure 3. Back-transformed trait differences between functional groups comparing the DIRECT with the grass–forb–legume (GFL) classification. (a) Above-ground Biomass, (b) Below-ground Biomass, (c) Specific Plant Area, (d) Leaf N Content, (e) Perenniality (1 = annual, 3 = biennial, 5 = perennial. 2 and 4 refer to either plastic life habits or an intermediate duration). Asterisks above bars indicate a significant difference (*< 0.05, **< 0.01, ***< 0.001), assessed using Tukey's HSD test. Error bars signify SE of the mean. Differences in Gs and A were not significant and are not presented here.

Download figure to PowerPoint

We analysed whether our classification created stronger functional trait differences between groups than the GFL classification. This consisted of allocating our species into grass, forb and legume groups and carrying out one-way ANOVAs to test for significant trait differences (Fig. 3). By categorizing species by traits, the differences in groups are stronger and could potentially describe more functions than GFL. In the GFL classification, the only discernible trait differences between groups were for biomass, which was higher in the grass group both above- and below-ground, and a there was also a weak difference in longevity. There were no significant differences between groups for SPA or LNC, which are both implicated in a large number of ecosystem processes. Additionally, at no point were the forbs and legumes different in terms of traits, so it is unlikely that their removal would cause measurable differences in nutrient and water cycling, in this system at least.

We established an experiment where every permutation of the three levels of the functional diversity treatments was represented (except the absence of all three) to test the relationship between functional groups and the ecosystem process(es) in question (Fry et al. 2013). We established functional group treatments by removing young plants from a recently ploughed field undergoing secondary succession; this kept community assemblage as natural as possible, while minimizing disturbance effects. After functional groups were created, in October 2008, we tested whether trait differences were sustained in the field. We also used abundance-weighted multivariate trait analyses to test if there were significant differences in trait dissimilarity between plots. If a plot contained species with similar trait values, it would have a lower value of the dissimilarity metric (e.g. Rao's Q; Rao 1982; Botta-Dukát 2005), while a community of functionally different species would have a higher value. A simple hypothesis would be that dissimilarity would increase with increasing functional group diversity. We used the program f-diversity (Casanoves et al. 2011) to calculate Rao's Q and FD (Petchey & Gaston 2002), with Gower dissimilarity. We found that there were clear differences in trait dissimilarity when different groups or combinations of groups were present (Fig. 4) but that our hypothesis was supported – the presence of functional group 1 reduced community dissimilarity, probably via the suppression of group 2 and 3 species, a finding that emphasizes the importance of interactions between functional groups in determining realized trait distributions and the need to use these realized trait distributions in analysis when explaining function. This result also indicates that increasing functional group richness and measuring ecosystem function on this basis alone could lead to the assumption that ‘functional diversity’ does not have a strong effect upon function. However, this is dangerous because, as we demonstrate, a closer inspection of the trait distributions of the groups is necessary for detailed mechanistic understanding.

image

Figure 4. Metrics of trait similarity in functional group treatments measured in October 2008, evaluated using one-way ANOVA. Fg corresponds to functional groups present in the plots. Functional group 1 refers to perennials, group 2 to caespitose grasses and tall forbs, and group 3 to annual plants. Letters refer to differences at the < 0.05 level evaluated using Tukey's HSD. (a) Weighted FD (Functional diversity; Petchey & Gaston 2002), (b) Rao's Q (Rao 1982; Botta-Dukát 2005). Error bars signify ±1 SEM.

Download figure to PowerPoint

In order to test the effect of the groups upon ecosystem function, we set up a litterbag study using 2 g dried Arrhenatherum elatius litter in 1-mm aperture mesh bags. These were fixed to the soil using staples in each plot in December 2009, and a bag was removed from each after 3 and then 6 mo. Living tissue was removed, then the content dried at 80 °C for 24 h, and weighed. The percentage mass loss over the period was calculated and an ANOVA using the seven levels of diversity as the factor term was carried out on the arcsine-transformed mass loss data. There was a very highly significant difference in mass loss at both of the time points measured, with higher decomposition rates overall where FG1 (perennials) were present (Fig. 5). This result is consistent with what was hypothesized and can be understood in terms of trait differences between the groups. FG1 species have a higher SPA and leaf N content (Fig. 3) than those of the other groups, meaning that these plots were used to receiving high-quality litter inputs that are likely to have stimulated microbial activity. A comparison of the pattern of the bars describing functional diversity (Fig. 4) and the bars describing decomposition rates (Fig. 5) appears to show reciprocal relationships between Rao's Q and weighted FD, and decomposition, indicating that there is a close relationship that is based on weighted traits and not functional richness.

image

Figure 5. Decomposition rates of Arrhenatherum elatius in a litterbag trial lasting 6 mo. Fg refers to functional groups present in the plots. The first litterbag harvest was after 3 mo. (a) December–March 2009, F6,37 = 7.87, < 0.001; the second after 6 mo (b) December–June 2009, F6,37 = 7.12, < 0.001. Error bars denote ±1 SEM. Letters refer to differences at the < 0.05 level evaluated using Tukey's HSD.

Download figure to PowerPoint

Discussion

  1. Top of page
  2. Abstract
  3. Introduction
  4. Suggested methodology
  5. Example: The DIRECT experiment
  6. Discussion
  7. Acknowledgements
  8. Author contributions
  9. References
  10. Supporting Information

The demarcation of groups of species with similar trait characteristics means that if the chosen traits are linked closely to function then the removal of groups will have different effects upon ecosystem function (Hillebrand & Matthiessen 2009). In presenting our new approach, we hope to show that field experiments based on rigorously tested trait groupings offer substantial benefits over a priori groupings with regard to their capacity to predict ecosystem function. We feel that hierarchical cluster analysis improves upon a priori groupings by allocating species to groups in a way that is unbiased towards any one trait and is applicable to many ecosystems. It also offers a method of grouping that results in clear, significant differences between traits, which can be mechanistically linked to function. This approach differs from the mainstream of past BDEF research, as most BDEF studies have focused on the functional implications of functional group richness rather than the role of specific traits and trait suites. Follow-up with post-hoc tests such as LDA provides confirmation of the appropriateness of allocations. One criticism of an approach in which groupings are tailor-made concerns the general applicability of results it delivers. However, in anything other than an analysis of the global flora, there are likely to be species missed from the classification. Nevertheless, there is the problem of how wide to make the classification; clearly there is a trade-off to be found in terms of an ‘all species, all functions’ approach in which the identification of underlying traits is difficult, and a ‘one-trait, one function’ approach. Neither is correct as both offer advantages in terms of generality and detailed mechanistic understanding, respectively.

In the DIRECT study the species groups each had a different set of trait means from the other two groups. While there was some overlap, e.g. AGB in FG1 and FG3 was not statistically significant, the selection of traits enabled differences to be noted in other areas, thus allowing for clear and testable predictions for their effect on ecosystem function and supporting the requirement for many traits to be used. A forerunner to our approach is that employed in the Jena experiment, a long-term field study that used customized functional groups to predict the role of functional diversity in a grassland with respect to a wide range of ecosystem functions, properties and services (Roscher et al. 2004). Their clustering method produced groupings that closely align with the GFL classification. Such approaches, also used for grassland ecosystems, have often used morphological traits, such as plant height and leaf size, that lack a clear theoretical link to most ecosystem functions.

A limitation of the trait grouping approach is that it does not account for the functional consequences of species interactions. It is very difficult to predict, particularly when species are free to join the community, what changes of form and function will occur due to competition and various stresses (Petchey 2004; Maestre et al. 2010). In our study, traits were chosen carefully, using a wealth of knowledge about each, but nevertheless there remain doubts of the efficacy of the trait-based approach to explain compositional controls of ecosystem function (Lavorel & Garnier 2002; Petchey 2004); the few large-scale BDEF studies that have used this approach indicate that traits are not silver bullets that fully explain how plants influence their environment (Wright et al. 2006; Díaz et al. 2007). The optimal way of defining compositional effects on function has yet to be established, but future approaches may incorporate careful screenings of potentially important species/trait interactions and responses to abiotic factors, which will then be pared back to the most effective set of descriptors. Careful consideration also needs to be afforded to the traits themselves. We chose to include root biomass as a trait because of the importance of below-ground processes in determining ecosystem function. However, root biomass correlates weakly with many below-ground functions, and above-ground traits often have better explanatory power over soil moisture and nutritional status, so there is a pressing need for higher quality below-ground trait measures (Wright et al. 2006; Kattge et al. 2011).

In the DIRECT experiment we aimed to compile the most comprehensive traits database with the resources and time available, using so-called ‘soft traits’, because often the most useful and accurate traits are too costly, time consuming or otherwise impractical to easily measure (Hodgson et al. 1998). A case in point would be key root traits (e.g. root exudation rates), many of which are difficult to quantify (Hishi 2007). Soft traits such as root length and root biomass are used as proxies for these traits (Poorter & de Jong 1999), and there is a large body of literature to support them.

Our system of deriving functional trait groupings is extremely flexible with regard to traits, ecosystems, functions and trophic levels, making it potentially applicable to a wide range of both terrestrial and aquatic systems. The method demonstrated few functional differences between the traditional GFL classification and showed that functional identity is a stronger predictor of ecosystem function than diversity or functional group richness. This offers a wide range of new possibilities for BDEF studies, and we hope that this will allow more mechanistic explorations of ecosystem functioning in the future.

Acknowledgements

  1. Top of page
  2. Abstract
  3. Introduction
  4. Suggested methodology
  5. Example: The DIRECT experiment
  6. Discussion
  7. Acknowledgements
  8. Author contributions
  9. References
  10. Supporting Information

The authors would like to thank A. Kendall, A. Pagan, T. Sloan and K. Prior for help with trait data collection. Thanks also to Dr D. Fontaneto for advice on LDA. ELF was funded by a PhD studentship from the Grantham Institute for Climate Change at Imperial College, London. Funding support from the Big Lottery Fund's Open Air Laboratories Project, NERC PopNET via the Centre for Population Biology and a British Ecological Society small project grant awarded to PM is also gratefully acknowledged. Thanks to C.E. Timothy Paine, A. Prinzing and anonymous reviewers for valuable comments on previous versions of this manuscript.

Author contributions

  1. Top of page
  2. Abstract
  3. Introduction
  4. Suggested methodology
  5. Example: The DIRECT experiment
  6. Discussion
  7. Acknowledgements
  8. Author contributions
  9. References
  10. Supporting Information

PM, SAP conceived and designed the study, ELF carried out practical work, PM, ELF carried out analyses, ELFwrote the manuscript with contributions from PM, SAP.

References

  1. Top of page
  2. Abstract
  3. Introduction
  4. Suggested methodology
  5. Example: The DIRECT experiment
  6. Discussion
  7. Acknowledgements
  8. Author contributions
  9. References
  10. Supporting Information
  • Albert, C.H., Thuiller, W., Yoccoz, N.G., Douzet, R., Aubert, S. & Lavorel, S. 2010a. A multi-trait approach reveals the structure and the relative importance of intra- vs. interspecific variability in plant traits. Functional Ecology 24: 11921201.
  • Albert, C.H., Thuiller, W., Yoccoz, N.G., Soudant, A., Boucher, F., Saccone, P. & Lavorel, S. 2010b. Intraspecific functional variability: extent, structure and sources of variation. Journal of Ecology 98: 604613.
  • Albert, C.H., Grassein, F., Schurr, F.M., Vieilledent, G. & Violle, C. 2011. When and how should intraspecific variability be considered in trait-based plant ecology? Perspectives in Plant Ecology, Evolution and Systematics 13: 217225.
  • Balvanera, P., Pfisterer, A.B., Buchmann, N., He, J.-S., Nakashizuka, T., Raffaelli, D. & Schmid, B. 2006. Quantifying the evidence for biodiversity effects on ecosystem functioning and services. Ecology Letters 9: 11461156.
  • Bolnick, D.I., Amarasekare, P., Araújo, M.S., Bürger, R., Levine, J., Novak, M., Rudolf, V.H.W., Schreiber, S.J., Urban, M.C. & Vasseur, D.A. 2011. Why intraspecific trait variation matters in community ecology. Trends in Ecology and Evolution 26: 183192.
  • Botta-Dukát, Z. 2005. Rao's quadratic entropy as a measure of functional diversity based on multiple traits. Journal of Vegetation Science 16: 533540.
  • Bunker, D.E., DeClerck, F., Bradford, J.C., Colwell, R.K., Perfecto, I., Phillips, O.L., Sankaran, M. & Naeem, S. 2005. Species loss and aboveground carbon storage in a tropical forest. Science 310: 1029103.
  • Cardinale, B.J., Srivastava, D.S., Duffy, J.E., Wright, J.P., Downing, A.L., Sankaran, M. & Jouseau, C. 2006. Effects of biodiversity on the functioning of trophic groups and ecosystems. Nature 7114: 989992.
  • Casanoves, F., Pla, L., Di Rienzo, J.A. & Díaz, S. 2011. FDiversity: a software package for the integrated analysis of functional diversity. Methods in Ecology and Evolution 2: 233237.
  • Cheng, X.L., Luo, Y.Q., Su, B., Zhou, X.H., Niu, S.L., Sherry, R., Weng, E.S. & Zhang, Q.F. 2010. Experimental warming and clipping altered litter carbon and nitrogen dynamics in a tallgrass prairie. Agriculture, Ecosystems and Environment 138: 206213.
  • Cornelissen, J.H.C., Lavorel, S., Garnier, E., Díaz, S., Buchmann, N., Gurvich, D.E., Reich, P.B., terSteege, H., Morgan, H.D., van der Heijden, M.G.A., Pausas, J.G. & Poorter, H. 2003. A handbook of protocols for standardised and easy measurement of plant functional traits worldwide. Australian Journal of Botany 51: 335380.
  • Cornwell, W.K. & Ackerly, D.D. 2009. Community assembly and shifts in plant trait distributions across an environmental gradient in coastal California. Ecological Monographs 79: 109126.
  • Craine, J.M., Tilman, D., Wedin, D., Reich, P., Tjoelker, M. & Knops, J. 2002. Functional traits, productivity and effects on nitrogen cycling of 33 grassland species. Functional Ecology 16: 563574.
  • Crawley, M.J. 2005. The flora of Berkshire: including those parts of modern Oxfordshire that lie to the South of the River Thames. Brambleby Books, Bedfordshire, UK.
  • de Bello, F., Lavorel, S., Dìaz, S., Harrington, R., Cornelissen, J.H.C., Bardgett, R.D., Berg, M.P., Cipriotti, P., Feld, C.K., Hering, D., da Silva, P.M., Potts, S.G., Sandin, L., Sousa, J.P., Storkey, J., Wardle, D.A. & Harrison, P.A. 2010. Towards an assessment of multiple ecosystem processes and services via functional traits. Biodiversity Conservation 19: 28732893.
  • Díaz, S. & Cabido, M. 1997. Plant functional types and ecosystem function in relation to global change. Journal of Vegetation Science 8: 463474.
  • Díaz, S., Cabido, M. & Casanoves, F. 1999a. Functional implications of trait–environment linkages in plant communities. In: Weiher, E. & Keddy, P. (eds.) Ecological assembly rules: perspectives, advances, retreats, pp. 338362. Cambridge University Press, Cambridge, UK.
  • Díaz, S., Cabido, M., Zak, M., Martinez Carretero, E. & Aranibar, J. 1999b. Plant functional traits, ecosystem structure and land-use history along a climatic gradient in central-western Argentina. Journal of Vegetation Science 10: 651660.
  • Díaz, S., Hodgson, J., Thompson, K., Cabido, M., Cornelissen, J., Jalili, A., Montserrat-Marti, G., Grime, J., Zarrinkamar, F., Asri, Y., Band, S., Basconcelo, S., Castro-Diez, P., Funes, G., Hamzehee, B., Khoshnevi, M., Perez-Harguindeguy, N., Perez-Rontom, M., Shirvany, F., Vendramini, F., Yazdani, S., Abbas-Azimi, R., Bogaard, A., Boustani, S., Charles, M., Dehghan, M., de Torres-Espuny, L., Falczuk, V., Guerrero-Campo, J., Hynd, A., Jones, G., Kowsary, E., Kazemi-Saeed, F., Maestro-Martinez, M., Romo-Diez, A., Shaw, S., Siavash, B., Villar-Salvador, P. & Zak, M. 2004. The plant traits that drive ecosystems: evidence from three continents. Journal of Vegetation Science 15: 295304.
  • Díaz, S., Lavorel, S., McIntyre, S., Falczuk, V., Casanoves, F., Milchunas, D.G., Skarpe, C., Rusch, G., Sternberg, M., Noy-Meir, I., Landsberg, J., Zhang, W., Clark, H. & Campbell, B.D. 2007. Plant trait responses to grazing: a global synthesis. Global Change Biology 13: 313341.
  • Ellenberg, H., Weber, H.E., Dull, R., Wirth, V., Werner, W. & Paulisen, D. 1991. Zeigerwerte von Pflanzen in Mitteleuropa [Indicator values of plants in Central Europe] Scripta Geobotanics Erich Goltze KG. Gottingen, DE.
  • Evans, J.R. & Poorter, H. 2001. Photosynthetic acclimation of plants to growth irradiance: the relative importance of specific leaf area and nitrogen partitioning in maximizing carbon gain. Plant, Cell and Environment 24: 755767.
  • Fitter, A.H. & Peat, H.J. 1994. The ecological flora database. Journal of Ecology 82: 415425.
  • Fornara, D.A. & Tilman, D. 2008. Plant functional composition influences rates of soil carbon and nitrogen accumulation. Journal of Ecology 96: 314322.
  • Fry, E.L., Manning, P., Allen, D.G.P., Hurst, A., Everwand, G., Rimmler, M. & Power, S.A. 2013. Plant functional group composition modifies the effects of precipitation change on grassland ecosystem function. PLoS ONE 8: article e57027.
  • Garnier, E., Cortez, J., Bill, G., Navas, M.L., Roumet, C., Debussche, M., Laurent, G., Blanchard, A., Aubry, D., Bellmann, A., Neill, C. & Toussaint, J.P. 2004. Plant functional markers capture ecosystem properties during secondary succession. Ecology 85: 26302637.
  • Gilmanov, T.G., Allard, V., Baldocchi, D., Beziat, P., Ceschia, E., Cellier, P. & Soussana, J.F. 2009. Productivity, respiration, and light-response parameters of world grassland and agro-ecosystems derived from flux-tower measurements. Rangeland Ecology and Management 63: 1639.
  • Grime, J.P. 1988. The C-S-R model of primary plant strategies – origins, implications and tests. In: Gottlieb, L.D. & Jain, S.K. (eds.) Plant evolutionary biology, pp. 371393. Chapman & Hall, London, UK.
  • Grime, J.P., Thompson, K., Hunt, R., Hodgson, J.G., Cornelissen, J.H.C., Rorison, I.H., Hendry, G.A.F., Ashenden, T.W., Askew, A.P., Band, S.R., Booth, R.E., Bossard, C.C., Campbell, B.D., Cooper, J.E.L., Davison, A.W., Gupta, P.L., Hall, W., Hand, D.W., Hannah, M.A., Hillier, S.H., Hodkinson, D.J., Jalili, A., Liu, Z., Mackey, J.M.L., Matthews, N., Mowforth, M.A., Neal, A.M., Reader, R.J., Reiling, K., Ross Fraser, W., Spencer, R.E., Sutton, F., Tasker, D.E., Thorpe, P.C. & Whitehouse, J. 1997. Integrated screening validates primary axes of specialisation in plants. Oikos 79: 259281.
  • Harrington, R., Anton, C., Dawson, T.P., de Bello, F., Feld, C.K., Haslett, J.R., Kluvankova-Oravska, T., Kontogianni, A., Lavorel, S., Luck, G.W., Rounsevell, M.D.A., Samways, M.J., Settele, J., Skourtos, M., Spangenberg, J.H., Vandewalle, M., Zobel, M. & Harrison, P.A. 2010. Ecosystem services and biodiversity conservation: concepts and a glossary. Biodiversity Conservation 19: 27732790.
  • Hartwig, U.A. 1998. The regulation of symbiotic N2 fixation: a conceptual model of N feedback from the ecosystem to the gene expression level. Perspectives in Plant Ecology, Evolution and Systematics 1: 92120.
  • Hector, A., Schmid, B., Beierkuhnlein, C., Caldeira, M.C., Diemer, M., Dimitrakopoulos, P.G., Finn, J., Freitas, H., Giller, P.S., Good, J., Harris, R., Högberg, P., Huss-Danell, K., Joshi, J., Jumpponen, A., Körner, C., Leadley, P.W., Loreau, M., Minns, A., Mulder, C.P.H., O'Donovan, G., Otway, S.J., Pereira, J.S., Prinz, A., Read, D.J., Scherer-Lorenzen, M., Schulze, E.-D., Siamantziouras, A.-S.D., Spehn, E.M., Terry, A.C., Troumbis, A.Y., Woodward, F.I., Yachi, S. & Lawton, J.H. 1999. Plant diversity and productivity experiments in European grasslands. Science 286: 11231127.
  • Hillebrand, H. & Matthiessen, B. 2009. Biodiversity in a complex world: consolidation and progress in functional biodiversity research. Ecology Letters 12: 14051419.
  • Hishi, T. 2007. Heterogeneity of individual roots within the fine root architecture: causal links between physiological and ecosystem functions. Journal of Forest Research 12: 126133.
  • Hodgson, J.G., Thompson, K., Wilson, P.J. & Bogaard, A. 1998. Does biodiversity determine ecosystem function? The Ecotron revisited. Functional Ecology 12: 843848.
  • Hooper, D.U. 1998. The role of complementarity and competition in ecosystem responses to variation in plant diversity. Ecology 79: 704719.
  • Hooper, D.U., Chapin, F.S., Ewel, J.J., Hector, A., Inchausti, P., Lavorel, S., Lawton, J.H., Lodge, D.M., Loreau, M., Naeem, S., Schmid, B., Setala, H., Symstad, A.J., Vandermeer, J. & Wardle, D.A. 2005. Effects of biodiversity on ecosystem functioning: A consensus of current knowledge. Ecological Monographs 75: 335.
  • Hsiao, T.C. & Acevedo, E. 1974. Plant responses to water deficits, water-use efficiency, and drought resistance. Agricultural Meteorology 14: 5984.
  • Jongman, R.H.G., Ter Braak, C.J.F. & Van Tongeren, O.F.R. 1995. Data analysis in community and landscape ecology. Cambridge University Press, Cambridge, UK.
  • Kattge, J., Díaz, S., Lavorel, S., Prentice, I.C., Leadley, P., Bönisch, G., Garnier, E., Westoby, M., Reich, P.B., Wright, I.J., Cornelissen, J.H.C., Violle, C., Harrison, S.P., Van Bodegom, P.M., Reichstein, M., Enquist, B.J., Soudzilovskaia, N.A., Ackerly, D.D., Anand, M., Atkin, O., Bahn, M., Baker, T.R., Baldocchi, D., Bekker, R., Blanco, C.C., Blonder, B., Bond, W.J., Bradstock, R., Bunker, D.E., Casanoves, F., Cavender-Bares, J., Chambers, J.Q., Chapin, F.S. III, Chave, J., Coomes, D., Cornwell, W.K., Craine, J.M., Dobrin, B.H., Duarte, L., Durka, W., Elser, J., Esser, G., Estiarte, M., Fagan, W.F., Fang, J., Fernández-Méndez, F., Fidelis, A., Finegan, B., Flores, O., Ford, H., Frank, D., Freschet, G.T., Fyllas, N.M., Gallagher, R.V., Green, W.A., Gutierrez, A.G., Hickler, T., Higgins, S.I., Hodgson, J.G., Jalili, A., Jansen, S., Joly, C.A., Kerkhoff, A.J., Kirkup, D., Kitajima, K., Kleyer, M., Klotz, S., Knops, J.M.H., Kramer, K., Kühn, I., Kurokawa, H., Laughlin, D., Lee, T.D., Leishman, M., Lens, F., Lenz, T., Lewis, S.L., Lloyd, J., Llusià, J., Louault, F., Ma, S., Mahecha, M.D., Manning, P., Massad, T., Medlyn, B.E., Messier, J., Moles, A.T., Müller, S.C., Nadrowski, K., Naeem, S., Niinemets, Ü., Nöllert, S., Nüske, A., Ogaya, R., Oleksyn, J., Onipchenko, V.G., Onoda, Y., Ordoñez, J., Overbeck, G., Ozinga, W.A., Patiño, S., Paula, S., Pausas, J.G., Peñuelas, J., Phillips, O.L., Pillar, V., Poorter, H., Poorter, L., Poschlod, P., Prinzing, A., Proulx, R., Rammig, A., Reinsch, S., Reu, B., Sack, L., Salgado-Negret, B., Sardans, J., Shiodera, S., Shipley, B., Siefert, A., Sosinski, E., Soussana, J.-F., Swaine, E., Swenson, N., Thompson, K., Thornton, P., Waldram, M., Weiher, E., White, M., White, S., Wright, S.J., Yguel, B., Zaehle, S., Zanne, A.E. & Wirth, C. 2011. TRY – a global database of plant traits. Global Change Biology 17: 29052935.
  • Keddy, P., Nielsen, K., Weiher, E. & Lawson, R. 2002. Relative competitive performance of 63 species of terrestrial herbaceous plants. Journal of Vegetation Science 13: 516.
  • Kjeldahl, J. 1883. Neue Methode zur Bestimmung des Stickstoffs in organischen Körpern. Medd Carlsberg Lab. Zeitschrift fur analytische Chemie 22: 366.
  • Kleyer, M. 1999. Distribution of plant functional types along gradients of disturbance intensity and resource supply in an agricultural landscape. Journal of Vegetation Science 10: 697708.
  • Kleyer, M., Bekker, R., Knevel, I., Bakker, J., Thompson, K., Sonnenschein, M., Poschlod, P., Van Groenendael, J., Klimei, L., Klimeiovi, J., Klotz, S., Rusch, G., Hermy, M., Adriaens, D., Boedeltje, G., Bossuyt, B., Dannemann, A., Endels, P., Gatzenberger, L., Hodgson, J., Jackel, A.K., Kuhn, I., Kunzmann, D., Ozinga, W., Ramermann, C., Stadler, M., Schlegelmilch, J., Steendam, H., Tackenberg, O., Wilmann, B., Cornelissen, J., Eriksson, O., Garnier, E. & Peco, B. 2008. The LEDA Traitbase: a database of life-history traits of the Northwest European flora. Journal of Ecology 96: 12661274.
  • Klumpp, K. & Soussana, J.-F. 2009. Using functional traits to predict grassland ecosystem change: a mathematical test of the response-and-effect trait approach. Global Change Biology 15: 29212934.
  • Lavorel, S. & Garnier, E. 2002. Predicting changes in community composition and ecosystem functioning from plant traits: revisiting the Holy Grail. Functional Ecology 16: 545556.
  • Lavorel, S., McIntyre, S., Landsberg, J. & Forbes, T.D.A. 1997. Plant functional classifications: from general groups to specific groups based on response to disturbance. Trends in Ecology and Evolution 12: 474478.
  • Lavorel, S., Grigulis, K., McIntyre, S., Williams, N.S.G., Garden, D., Dorrough, J., Berman, S., Quetier, F., Thebault, A. & Bonis, A. 2008. Assessing functional diversity in the field – methodology matters!. Functional Ecology 22: 134147.
  • Maestre, F.T., Bowker, M.A., Escolar, C., Puche, M.D., Soliveres, S., Maltez-Mouro, S., Garcia-Palacios, P., Castillo-Monroy, A.P., Martinez, I. & Escudero, A. 2010. Do biotic interactions modulate ecosystem functioning along stress gradients? Insights from semi-arid plant and biological soil crust communities. Philosophical Transactions of the Royal Society of London B, Biological Sciences 365: 20572070.
  • McConnaughay, K.D.M. & Bazzaz, F.A. 1991. Is physical space a soil resource? Ecology 72: 94103.
  • McLaren, J.R. & Turkington, R. 2010. Plant functional group identity differentially affects leaf and root decomposition. Global Change Biology 16: 30753084.
  • Naeem, S. 2008. Advancing realism in biodiversity research. Trends in Ecology and Evolution 23: 414416.
  • Naeem, S. & Wright, J.P. 2003. Disentangling biodiversity effects on ecosystem functioning: deriving solutions to a seemingly insurmountable problem. Ecology Letters 6: 567579.
  • Pacala, S.W. & Kinzig, A.P. 2001. Introduction to Theory and the Common Ecosystem Model. In: Kinzig, A.P., Pacala, S.W. & Tilman, D. (eds.) The functional consequences of biodiversity: empirical progress and theoretical extensions, pp. 169174. Princeton University Press, Princeton, NJ, US.
  • Petchey, O.L. 2004. On the statistical significance of functional diversity effects. Functional Ecology 18: 297303.
  • Petchey, O.L. & Gaston, K.J. 2002. Functional diversity (FD), species richness and community composition. Ecology Letters 5: 402411.
  • Poorter, H. & de Jong, R. 1999. A comparison of specific leaf area, chemical composition and leaf construction costs of field plants from 15 habitats differing in productivity. New Phytologist 143: 163176.
  • Rao, C.R. 1982. Diversity and dissimilarity coefficients: a unified approach. Theoretical Population Biology 21: 2443.
  • Raunkiaer, C. 1934. The life forms of plants and statistical plant geography. Clarendon Press, Oxford, UK.
  • Reich, P.B., Walters, M.B., Ellsworth, D.S., Vose, J.M., Volin, J.C., Gresham, C. & Bowman, W.D. 1998. Relationships of leaf dark respiration to leaf nitrogen, specific leaf area and leaf life-span: a test across biomes and functional groups. Oecologia 114: 471482.
  • Rodwell, J.S. 1992. British plant communities: grassland and Montane Communities. Cambridge University Press, Cambridge, UK.
  • Roleček, J., Tichý, L., Zelený, D. & Chytrý, M. 2009. Modified TWINSPAN classification in which the hierarchy respects cluster heterogeneity. Journal of Vegetation Science 20: 596602.
  • Roscher, C., Schumacher, J., Baade, J., Wilcke, W., Gleixner, G., Weisser, W.W., Schmid, B. & Schulze, E.D. 2004. The role of biodiversity for element cycling and trophic interactions: an experimental approach in a grassland community. Basic and Applied Ecology 5: 107121.
  • Royal Botanic Gardens, Kew 2008. Seed Information Database (SID). www.rbgkew.org.uk/sid (accessed 15th March 2008).
  • Serraj, R., Sinclair, T.R. & Purcell, L.C. 1999. Symbiotic N2 fixation response to drought. Journal of Experimental Botany 50: 143155.
  • Shaw, P.J.A. 2003. Multivariate statistics for the environmental science. Hodder-Arnold, London, UK.
  • Tilman, D. 1999. The ecological consequences of changes in biodiversity: a search for general principles. Ecology 80: 14551474.
  • Tilman, D., Wedin, D. & Knops, J. 1996. Productivity and sustainability influenced by biodiversity in grassland ecosystems. Nature 379: 718720.
  • Unit of Comparative Plant Ecology, Sheffield 2005. Leaf Area Program. The University of Sheffield, UK (accessed 29th April 2008).
  • USDA, NRCS 2010. The PLANTS Database. National Plant Data Center, Baton Rouge, LA, US (accessed 2nd June 2008).
  • Violle, C., Navas, M.-L., Vile, D., Kazakou, E., Fortunel, C., Hummel, I. & Garnier, E. 2007. Let the concept of trait be functional!. Oikos 116: 882892.
  • de Vries, F.T., Manning, P., Tallowin, J.R.B., Mortimer, S.R., Pilgrim, E.S., Harrison, K.A., Hobbs, P.J., Quirk, H., Shipley, B., Cornelissen, J.H.C., Kattge, J. & Bardgett, R.D. 2012. Abiotic drivers and plant traits explain landscape-scale patterns in soil microbial communities. Ecology Letters 15: 12301239.
  • Walker, B., Kinzig, A. & Langridge, J. 1999. Plant attribute diversity, resilience, and ecosystem function: the nature and significance of dominant and minor species. Ecosystems 2: 95113.
  • Westoby, M. 1998. A leaf–height–seed (LHS) plant ecology strategy scheme. Plant and Soil 199: 213227.
  • Wilson, P.J., Thompson, K. & Hodgson, J.G. 1999. Conventional functional classification schemes underestimate the relationship with ecosystem functioning. New Phytologist 143: 155162.
  • Wright, I.J., Reich, P.B., Westoby, M., Ackerly, D.D., Baruch, Z., Bongers, F., Cavender-Bares, J., Chapin, F.S., Cornelissen, J.H.C., Diemer, M., Flexas, J., Garnier, E., Groom, P.K., Gulias, J., Hikosaka, K., Lamont, B.B., Lee, T., Lee, W., Lusk, C., Midgley, J.J., Navas, M.-L., Niinemets, Ü., Oleksyn, J., Osada, N., Poorter, H., Poot, P., Prior, L., Pyankov, V.I., Roumet, C., Thomas, S.C., Tjoelker, M.G., Veneklaas, E.J. & Villar, R. 2004. The world-wide leaf economics spectrum. Nature 428: 821827.
  • Wright, J.P., Naeem, S., Hector, A., Lehman, C., Reich, P.B., Schmid, B. & Tilman, D. 2006. Conventional functional classification schemes underestimate the relationship with ecosystem functioning. Ecology Letters 9: 111120.
  • Yan, B., Zhang, J., Liu, Y., Li, Z., Huang, X., Yang, W. & Prinzing, A. 2012. Trait assembly of woody plants in communities across sub-alpine gradients: Identifying the role of limiting similarity. Journal of Vegetation Science 23: 698708.

Supporting Information

  1. Top of page
  2. Abstract
  3. Introduction
  4. Suggested methodology
  5. Example: The DIRECT experiment
  6. Discussion
  7. Acknowledgements
  8. Author contributions
  9. References
  10. Supporting Information
FilenameFormatSizeDescription
jvs12068-sup-0001-AppendixS1-S2.pdfapplication/PDF181KAppendix S1. Methods for obtaining trait data.
jvs12068-sup-0002-AppendixS2.txtplain text document3KAppendix S2. R code for methods.
jvs12068-sup-0003-AppendixS1-S2.docxWord document22K 

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.