Predicting changes in community composition and ecosystem functioning from plant traits: revisiting the Holy Grail

Authors

  • S. Lavorel,

    Corresponding author
    1. Centre d’ Ecologie Fonctionnelle et Evolutive, CNRS UPR 9056, 1919 route de Mende, 34293 Montpellier Cedex 5, France
      *Author to whom correspondence should be addressed. E-mail: lavorel@cefe.cnrs-mop.fr
    Search for more papers by this author
  • E. Garnier

    1. Centre d’ Ecologie Fonctionnelle et Evolutive, CNRS UPR 9056, 1919 route de Mende, 34293 Montpellier Cedex 5, France
    Search for more papers by this author

*Author to whom correspondence should be addressed. E-mail: lavorel@cefe.cnrs-mop.fr

Summary

  • 1The concept of plant functional type proposes that species can be grouped according to common responses to the environment and/or common effects on ecosystem processes. However, the knowledge of relationships between traits associated with the response of plants to environmental factors such as resources and disturbances (response traits), and traits that determine effects of plants on ecosystem functions (effect traits), such as biogeochemical cycling or propensity to disturbance, remains rudimentary.
  • 2We present a framework using concepts and results from community ecology, ecosystem ecology and evolutionary biology to provide this linkage. Ecosystem functioning is the end result of the operation of multiple environmental filters in a hierarchy of scales which, by selecting individuals with appropriate responses, result in assemblages with varying trait composition. Functional linkages and trade-offs among traits, each of which relates to one or several processes, determine whether or not filtering by different factors gives a match, and whether ecosystem effects can be easily deduced from the knowledge of the filters.
  • 3To illustrate this framework we analyse a set of key environmental factors and ecosystem processes. While traits associated with response to nutrient gradients strongly overlapped with those determining net primary production, little direct overlap was found between response to fire and flammability.
  • 4We hypothesize that these patterns reflect general trends. Responses to resource availability would be determined by traits that are also involved in biogeochemical cycling, because both these responses and effects are driven by the trade-off between acquisition and conservation. On the other hand, regeneration and demographic traits associated with response to disturbance, which are known to have little connection with adult traits involved in plant ecophysiology, would be of little relevance to ecosystem processes.
  • 5This framework is likely to be broadly applicable, although caution must be exercised to use trait linkages and trade-offs appropriate to the scale, environmental conditions and evolutionary context. It may direct the selection of plant functional types for vegetation models at a range of scales, and help with the design of experimental studies of relationships between plant diversity and ecosystem properties.

Introduction

The quest for general rules associating species and environmental conditions and, in particular, the search for associations between abiotic and biotic factors and species characteristics (recurrent patterns of species specialization, Grime 1979; life-history strategies, Southwood 1988) has concerned community ecologists for decades. Recently, the desire to simultaneously predict vegetation responses to global change factors and changes in important terrestrial ecosystem functions (such as biogeochemical cycles, invasion resistance, stability in the face of disturbance) has revived this concept of plant functional classification. A conceptual framework and methods have been developed to predict changes in ecosystem processes such as biogeochemical cycling by considering the role of plant traits in ecosystem structure and processes.

Initial conceptual and large-scale vegetation models (Smith et al. 1997; Woodward & Cramer 1996) assumed that grouping plants a priori, based on knowledge of their function or on observed correlations among traits, would make it possible to directly predict changes in ecosystem processes from projected changes in plant composition in response to global change. These assumptions, which reflected the belief that functional effect groups (species with a similar effect on one or several ecosystem functions, Gitay & Noble 1997; Walker et al. 1999) and functional response groups (groups of species with a similar response to a particular environmental factor such as resource availability, disturbance or CO2, Gitay & Noble 1997; Lavorel et al. 1997) should coincide, have remained problematic despite sustained efforts on concepts and terminology (Gitay & Noble 1997; Lavorel & Garnier 2001; Lavorel et al. 1997).

A priori functional effect groups based on taxonomy (grasses, legumes, non-legume forbs) and/or coarse descriptions of function (N-fixing, phenology, life cycle, photosynthetic pathway) have also been used in experiments documenting the effects of functional diversity on ecosystem functioning (Díaz & Cabido 2001). In contrast, recent tests of functional redundancy and ecosystem resilience (Walker et al. 1999) have clearly distinguished between effect and response. These experimental studies examined the hypothesis that response to environmental change should cause species composition turnover but leave biogeochemical cycling unchanged, especially when effect groups are species-rich. Finally, Chapin et al. (2000) proposed a conceptual framework where modifications of species composition resulting from environmental change translate into modifications of ecosystem functioning via changes in the representation of species traits (Fig. 1b).

Figure 1.

Conceptual framework for effects of environmental changes on plant community structure (or biodiversity) and ecosystem functioning. (a) Simplification of the filter theory of Keddy (1992) and Woodward & Diament (1991), where response of community structure to environmental conditions is the result of species response traits. (b) Summary of the framework of Chapin et al. (2000) predicting the ecosystem consequences of environmental changes via species effect traits. (c) Representation of the proposed conceptual framework that articulates environmental response and ecosystem effects through varying degrees of overlap between relevant traits.

The objectives of this paper are to:

  • 1briefly summarize the rationale, approaches and traits for classifications of plant responses and functional effects;
  • 2examine commonalities and differences between response traits and effect traits underlying these groupings for an example set of key environmental factors and ecosystem processes;
  • 3propose a conceptual framework that links traits associated with responses to those that determine effects on ecosystems – the aim of this framework is to integrate analyses of response traits in relation to environmental and/or biotic factors with analyses of functional effects of species, and hence trait composition, in order to analyse the effects of environmental changes on ecosystem processes.

Functional Response and Functional Effect Groups

Plant Responses to Environmental Changes

We first focus on organism-centred issues, and aim to understand the adaptive significance of traits or combinations of traits in order to predict the responses of organisms to environmental factors. Plant communities can be seen as the result of a hierarchy of abiotic (climatic, resource availability, disturbance) and biotic (competition, predation, mutualisms) filters that successively constrain which species and traits, from a regionally available pool, can persist at a site (Fig. 1a; Keddy 1992; see review by Díaz et al. 1999). Following this model, we should be able to predict the trait pool of plant communities by combining knowledge of the nature and strength of different filters with that of response traits for each of these filters. For each filter, the corresponding principal environmental factor defines response groups and hence species that are retained. The effects of changes in abiotic factors, such as climate, atmospheric CO2 concentration and disturbance, could then be modelled as changes in the strength of these different filters (Díaz et al. 1999; Woodward & Diament 1991).

Recent analyses of the significance of selected traits for plant responses to environmental factors have considered ‘soft’ traits, which are easy to measure for a large number of species and sites, but are not necessarily explicitly related to a specific functional mechanism; and ‘hard’ traits, usually less accessible but with a direct functional role (Hodgson et al. 1999; Table 1). These commonly include:

Table 1.  Traits relevant to ecosystem response and effects
Soft traitMechanism and corresponding hard traits
Environmental responseEcosystem effect
NutrientsFirePrimary productivityFlammability
  1. We selected two environmental factors, nutrient gradient (resource) and fire (disturbance), and two ecosystem functions, net primary production (NPP) and flammability. The left-hand column lists soft traits, or substitute measurements when these are not available, that are relevant to these responses or effects. The table presents the specific functions and, when available, the hard trait corresponding to a given soft trait – response or effect association. Our analyses focused on identifying traits that are common to both response and effect, which can be found occupying the same sections of the table. Lists of traits and corresponding processes are compiled from the literature (see Lavorel 2002 for general review).

  2. Response to nutrient gradient: based on Chapin (1980); Grime (1979); Poorter & Garnier (1999); Reich, Walters & Ellsworth (1992).

  3. Response to fire: based on Bond & Midgley (2001); Noble & Slatyer (1980).

  4. Primary productivity: based on Chapin (1993); Monteith (1977); Saugier, Roy & Mooney (2001).

  5. Flammability: based on Bond & Midgley (1995); Schwilk & Ackerly (2001); Whelan (1995).

Whole plant
Growth formAbsorption: rooting depthAvoidance: plant/bud height (physical escape), phenology (temporal escape)C stock: standing biomassWater status: access to deep water
 Decomposition: lignin contentRegeneration: strategy (resprout vs. seed)Resource capture: standing biomass 
   Growth: growth rate 
Life span Tolerance: longevity  
Maturation age Regeneration  
Period of photosynthetic activity Avoidance: phenologyLight capture 
Mass of underground reserves Tolerance: resprouting abilityC stock: carbon sink 
Whole shoot
Shoot height Avoidance: physical escapeC stock: standing biomassFire spread
   Light captureCompetition: stand structure
Canopy architecture Avoidance: physical escapeLight capture: leaf area index (LAI)Water stress: LAI
   C fixation: whole-plant instantaneous photosynthetic rateFire spread
   Growth: growth rateMicroclimate in understorey: LAI
    Allocation: fuel size distribution
Bark thickness Tolerance: tissue protection  
Leaf
Specific leaf area (SLA)Conservation: residence timeTolerance: resprout RGRC fixation: leaf instantaneous photosynthetic rate 
 Decomposition: lignin contentRegeneration: seedling RGR  
 Growth: growth rate   
Dry matter contentConservation: residence time Growth: growth rateTissue composition: water content
 Decomposition: lignin content?  Water status: osmotic potential (drought tolerance)
N concentrationConservation: residence timeTolerance: resprout RGRC fixation: leaf instantaneous photosynthetic rateAllocation: fuel size distribution?
 Decomposition: C : N ratioRegeneration: seedling RGRGrowth: growth rate 
   Allocation: C : N ratio? 
Leaf life spanConservation: residence timeTolerance: resprout RGRC fixation: cumulated photosynthesis 
  Regeneration: seedling RGRGrowth: growth rate 
Leaf phenology Avoidance: timing of leaf sheddingC fixation: cumulated photosynthesis 
Photosynthetic pathway  C fixation: leaf instantaneous photosynthetic rate 
Natural 15N abundanceAbsorption: root distribution  Water status: access to deep water?
Root
Rooting depthAbsorption: root distribution, root lengthTolerance: resprouting abilityC stock: root massWater status: access to deep water?
Specific root lengthAbsorption: instantaneous absorption rate C and nutrient uptake: specific absorption rate 
   Growth: growth rate 
Process of N captureN capture: N2 fixation/N absorption C and nutrient uptake: specific absorption rate 
   Growth: growth rate 
Seed
Seed mass Regeneration: fecundity, seedling establishment, dispersal  
Seed shape Regeneration: seed bank persistence  
Germination phenology Regeneration: phenology with respect to timing of fire and growing season  
Germination tests Regeneration: germination cues  

Natural gradients combine variations in climate, resource and disturbance. These underlying simple gradients may be explicit, as for analyses of response to altitude (Pavón et al. 2000) or agricultural disturbance (Kleyer 1999). On the other hand, little may be quantified about the nature and amount of environmental variation along complex natural gradients, as in the case of succession (Bazzaz 1996; Prach, Pyšek & Šmilauer 1997). In all cases, the combination of multiple factors and a lack of knowledge about key factors impede the interpretation and prediction of plant distributions, because traits associated with different single gradients can be independent – filters do not match well, as for water vs. nutrient stress (Cunningham, Summerhayes & Westoby 1999); or one filter involves adult and the other regeneration traits (Leishman & Westoby 1992; Shipley et al. 1989).

Triangular models, such as Grime's (1979) plant strategy (C-S-R) scheme and Westoby's (1998) leaf-height-seed model, have helped elucidate these issues by explicitly using soil resource availability and disturbance as two orthogonal dimensions for plant classification. These models for responses to combined factors are underpinned by the hypothesis of trade-offs and correlations among plant traits. More generally, the analysis of plant functional types has been guided by the recognition that plants are constrained in their performance for alternative functions, such as resource capture and conservation (Chapin, Autumn & Pugnaire 1993; Grime 1979; Poorter & Garnier 1999); acquisition of different resources such as light and water (Smith & Huston 1989) or light and nutrients (Tilman 1988); or growth and reproduction (Silvertown et al. 1993; Solbrig 1993). Growth forms are the ultimate expression of these trade-offs, and of the links between key plant traits and plant response and function (Chapin 1993).

Groups, Traits and Biogeochemical Cycles

We now turn to the issue of how organisms affect the functioning of ecosystems. We restrict our analysis to a snapshot view of ecosystems, where species composition is assumed to be stable and the primary functions considered are fluxes of energy and matter. We concentrate on how species affect components of biogeochemical cycles, taking net primary productivity (NPP) as an example. From an ecosystem perspective, all green plants convert inorganic resources to organic matter and belong to a single functional group: primary producers. The relevant question is therefore how biogeochemical cycles, and NPP in particular, are regulated by particular species groups and/or traits.

To address this issue, many studies have examined differences in NPP among communities of varying composition, without necessarily basing their a priori groupings on plant traits. These include the following, in order of increasing refinement of species grouping.

The impact of species traits on NPP has been studied empirically in a number of cases (Eckardt et al. 1977; Herbert et al. 1999; see below). Complex, mechanistic models have also been developed which involve detailed calculations of carbon balance components, and explicitly involve traits (mostly hard) (Ågren et al. 1991). The general formulation of NPP proposed by Monteith (1977) could theoretically be used to assess the impact of some relevant traits on NPP. However, it is more suited to a whole-stand analysis where the properties of individual species are not taken into account, and it has been widely used for crop production and monospecific stands (reviewed by Sinclair & Muchow 2000; but see Saugier, Roy & Mooney 2001 for applications to natural vegetation). Recently, Chapin (1993); Chapin et al. (1996) proposed a different approach to NPP, more suitable for analysis of the effect of functional diversity on biogeochemical cycling in natural, species-rich ecosystems. This formulation explicitly incorporates species traits central to plant functioning in a continuous fashion. Based on these ideas, we develop below an expression of NPP, whose exact form differs from that proposed by Chapin et al. (1996). NPP is the sum of the productivity of individual species in the community, and can be written as:

image(eqn 1)

where Ni is the number of individuals of species i per unit ground area; Mf and Mo are the final and initial average biomasses of individuals of species i; and ΔT is the period over which NPP is assessed (1 year in many cases). Now Mfi can be written as:

image(eqn 2)

where RGRi and (tf – to)i are the average relative growth rate and period of active growth of species i, respectively. Combining equations 1 and 2:

image(eqn 3)

According to equation 3, NPP is controlled by the relative initial biomass of each species in the community (Ni × Moi; its carbon stock), the integrated functioning of each species (the outcome of carbon assimilation, nutrient uptake, allocation, etc.), and its phenology (duration of active growth period).

On a broad scale, when different biomes or vegetation types are compared, differences in NPP are strongly correlated to the first factor in equation 3, standing biomass or height of the dominant species (Chapin et al. 1996; Saugier et al. 2001) – two closely related variables at this scale of comparison (Niklas & Enquist 2001). The similarity of broad relationships involving organisms as different as unicellular algae and terrestrial macrophytes have led Niklas & Enquist (2001) to argue that NPP depended mostly on community standing biomass, not on species composition. However, when biomes of similar physiognomy found under different climates are compared, substantial differences in NPP can still be observed (as also seen in the data of Niklas & Enquist 2001), and are a likely consequence of differences in metabolic activity and length of growing season (second and third factors in equation 3). At a more local scale, the influence of biomass weakens and the physiology of the species predominates (Table 2). The role of phenology in NPP has not been thoroughly assessed, but some examples show that increasing the growing season either of individual plants (Jackson et al. 2001) or at the community level (for example in mixtures of C3 and C4 species, Epstein et al. 1999) could significantly increase NPP.

Table 2.  Nitrogen mineralization rate, above-ground net primary productivity (ANPP) and leaf characteristics of dominant species taken from various vegetation types in Central Europe
Vegetation type/dominant speciesNet N mineralization (kg N ha−1 year−1)Minimum above-ground biomass (g m−2)ANPP (g m−2 year−1)Mean SLA (m2 kg−1)Mean leaf N concentration (mg g−1)Estimated Amax (nmol g−1 s−1)
Ellenberg/Poorter
Sand dunes 12–19  ≈0  90 9·913·5 67·2
Heath 11–30  ≈700* 21018·716·7124
Chalk grasslands 20–30  ≈0 33021·315·7130
Fertilized meadows130–160  ≈0108031·836·1328
Aerts and co-workers
Wet heathland:
Erica tetralix  4·4 600376 8·012·6 54·8
Molinia caerulea  7·8 11786721·319·3152
Dry heathland:
Calluna vulgaris  6·2 710540 8·0nana
Molinia caerulea 10·9  5661422·714·0125

This approach could probably be extended to other processes, such as the rate of litter decomposition, provided that a model linking the decomposition constant of litter and traits of either the litter itself (its chemical composition, Heal et al. 1997) or living leaves (Cornelissen et al. 1999) can be developed.

From Community Response to Ecosystem Functioning: a Framework and Some Case Studies

A Framework to Link Response and Effect Groups

As outlined in the two previous sections, the two types of functional classifications – response groups and effect groups – correspond to different approaches and, to some extent, traits measured. Physiological, harder traits at the individual level are more commonly used for effect groups (Chapin 1993; Herbert et al. 1999), whereas response groups are identified through community-level studies of changes in soft, morphological or behavioural traits in response to abiotic or biotic factors. The frequent use of lists of traits that do not overlap makes it difficult to reconcile the two types of classifications (Weiher et al. 1999; but see Hodgson et al. 1999), which is needed to build a more comprehensive framework of response–effect linkages onto the ideas of Chapin et al. (2000).

To achieve this, we need to match attribute lists for responses with the known effects of some of these attributes (or their correlates) on ecosystem processes. From a global perspective, the primary environmental factors determining plant community structure are resources and disturbances. Although a wider range of axes should be considered for a comprehensive study, for illustration we restricted our comparative analysis of response and effect traits to one type of resource (soil nutrients) and one important disturbance (fire). Similarly, we chose two main ecosystem effects (primary productivity and flammability). Traits were selected based on the literature and on expert opinion synthesized during a workshop (Building a Global Key of Plant Functional Types; http://gcte.org/focus2/). We favoured traits that are continuous (SLA, shoot height) rather than categorical, although some traits, such as life form or photosynthetic pathway, are by definition categorical. Where possible we preferred soft traits, but in some instances where this knowledge is not yet available, harder traits or phenomenological surrogates had to be chosen. In an attempt to examine the mechanisms responsible for differing degrees of overlap between responses and effects, we also listed specific functions associated with each trait. For instance, disturbance response is classically split into three processes: avoidance, tolerance and regeneration, and the specific relevance of these for different fire-response traits was listed. Similarly, for traits associated with primary productivity we noted which of the three components of equation 3 they were relevant to. Having done this, we were able to construct a table listing relevant functions (in terms of responses and effects) for each trait (Table 1). Our comparison then examined which traits were associated with at least one response process and at least one effect process.

A first examination of Table 1 reveals that the resource axis shows maximum overlap between response and effect traits, whereas overlaps for the disturbance axis are more limited. To address the causes of differing degrees of trait overlap between response and effects, we must analyse the specific functions for the traits involved. Analyses are presented in the following section, first for the relationship between resource gradients and productivity, then for the apparently tenuous relationship between fire tolerance and flammability. We then expand our investigation to consider general causes of independence between disturbance response and ecosystem effects, and finally discuss the extent to which overlaps between response and function can be inferred from trait linkages.

How Do Species Traits and Ecosystem Primary Productivity Covary Along Gradients of Soil Nutrient Availability?

Early syntheses on changes in species traits along nutrient gradients (Chapin 1980; Grime 1979) recognized that species from nutrient-rich habitats tend to be inherently fast-growing, with rapid resource capture and fast turnover of organs leading to poor internal conservation of resources, while the reverse is true for species from nutrient-poor habitats (Table 1). More recently, a series of quantitative traits has been associated with this fundamental trade-off in plant function (Grime et al. 1997; Poorter & Garnier 1999; Reich et al. 1992). Fast-growing species from nutrient-rich habitats usually have a combination of high SLA; high tissue nutrient concentration (in particular, N); low tissue density and cell wall content; high rates of carbon and nutrient uptake; and short-lived leaves. Opposite traits characterize species from nutrient-poor habitats in which the mean residence time of nutrients tends to be maximized through longer organ longevity (in particular, leaf) and/or higher resorption of nutrients from senescing organs (Aerts & Chapin 2000; Garnier & Aronson 1998).

At the ecosystem level, limitation of primary production by nutrient availability – particularly N, on which we concentrate here – is widespread (Aerts & Chapin 2000; Vitousek & Howarth 1991). Net mineralization rates of N vary tremendously among ecosystems, from 0 to 0·5 kg N m−2 year−1 in arctic tundra or bogs, to 300 kg N m−2 year−1 in ruderal vegetation (Ellenberg 1977; Larcher 1995). These rates correlate positively with above-ground net primary productivity (ANPP) in several ecosystems (Hunt et al. 1988; Reich et al. 1997). Table 2 shows two examples for different types of vegetation in northern Central Europe: in the first, a tenfold variation in mineralization rates from sand dunes to fertilized meadows (Ellenberg 1977) was associated with a 12-fold increase in ANPP (Poorter & de Jong 1999); in the second, increases in mineralization rates associated with a shift in dominants from Erica tetralix and Calluna vulgaris to the perennial grass Molinia caerulea (van Vuuren et al. 1992) paralleled the increase in ANPP (Aerts 1989; Aerts & Berendse 1989).

Is there a relationship between these changes in NPP along gradients of N availability and the shifts in plant traits highlighted above? We are not aware of any study in natural vegetation in which the variation in NPP along gradients of N availability was analysed according to equation 3. Indirect assessments of the different factors can be deduced for the two sets of experiments presented in Table 2. The impact of length of growing season on NPP is likely to be low in these cases, as all data were collected from communities with a similar length of growing season (Aerts 1989; Aerts & Berendse 1989). Contrary to the prediction of Niklas & Enquist (2001), standing biomass had no consistent effect on ANPP in either case. In the study by Aerts and co-workers, the vegetation with the highest initial standing biomass (dominated by Ericaceae) showed the lowest ANPP. By contrast, consistent differences in leaf structure and function were found among species. In nutrient-rich habitats, species tend to have leaves with high SLA, N concentration and photosynthetic rates per unit dry mass (Table 2). These traits and process are positively related to whole-plant RGR (Poorter & Garnier 1999; Reich et al. 1992), the physiological factor in equation 3. We therefore conclude that, in this case, differences in ANPP would mainly result from differences in physiology of the dominant species at the different sites.

We cannot rule out the possibility that length of growing season and standing biomass may play a role in other instances. However, the evidence presented here suggests that the species traits involved in changes in primary productivity of ecosystems along a nutrient gradient (effect traits) strongly overlap with those involved in the response of species to the same gradient (response traits). These traits include primary traits such as SLA, as well as some of their correlates (leaf dry matter content, leaf longevity), or other (physiological) traits which represent a scaling mechanism from environmental factor to function (photosynthetic rate, C : N ratio).

Are Fire-Tolerant Ecosystems More Flammable?

In contrast to the case of nutrients and productivity, traits determining ecosystem flammability showed little direct overlap with traits associated with response to fire (Table 1).

Fire tolerance is related to alternative sets of traits that allow plants to avoid fire entirely (height taller than flames, annuals with dormant seeds during the fire season); or to tolerate fire by surviving and regrowing vigorously (thick bark, resprouting ability resulting from investment in underground reserves) or regenerating from seeds (canopy or soil seed banks, fast growth rate, fast maturation). The relative importance of the two tolerance mechanisms, seeding or sprouting, has been related to fire frequency and intensity (Bellingham & Sparrow 2000; but see Pausas 2001), and to resource availability (Bond & Midgley 2001).

Ecosystem flammability is a complex emergent property which involves two successive processes and traits linked with each (Whelan 1995). First, fire must ignite plant material, then it must propagate through the canopy and/or understorey. Traits relating to the ignition phase (flammability) are those determining tissue moisture, such as water content and traits conferring drought resistance; and, to a debated extent, chemical composition – volatiles, waxes and resins increase flammability while high lignin and mineral content decrease it. Fire spread relates to two characteristics: the energy produced by the initial burning, which depends largely on the same traits as flammability; and the spatial distribution of fuel, which is a consequence of total biomass accumulation and its spatial arrangement. Traits promoting biomass accumulation through rapid growth and slow decomposition therefore increase flammability. Architecture and structural traits then determine the spatial distribution of this biomass. A high surface area-to-volume ratio of organs (fine foliage, thin branches), low stature, wide lateral spread, low canopy density and retention of dead branches are attributes that increase fire propagation. Two additional effects of individual plant traits on communities must finally be considered for a complete picture of flammability: the effects of species on soil moisture and temperature in the understorey; and their effects on stand vertical structure and density.

This list of traits relevant to ecosystem flammability shows little of the overlap with traits relating to fire response that would be expected from evolutionary arguments about the co-occurrence of flammability and fire tolerance (Bond & Midgley 1995). Phylogenetically independent contrasts among pine species have shown associations between flammability-enhancing traits and fire tolerance (Schwilk & Ackerly 2001). Models have demonstrated that these can be sustainable if flammability-promoting traits warrant additional fitness benefits such as secondary compounds with antiherbivore benefits, increased growth rate associated with early and prolific branching, or enhanced carbon gains for resprouters (Bond & Midgley 1995). Alternatively, the occurrence of flammability-enhancing traits in a species could merely reflect response to factors other than fire, such as herbivory and drought.

Consequently, if functional linkages exist between traits promoting fire tolerance and those involved in ecosystem flammability, they would have to be mostly indirect, through character associations or trade-offs. For instance, a high growth rate is required to increase the success of seed regeneration after fire, and is also associated with canopy architectures with many thin stems and high surface : volume ratios. The association between sprouting and drought tolerance (which allows low water potential and hence increases flammability), both of which result from investment in large underground structures, is another example. However, inferring such response–effects associations from a series of correlations of varying strengths needs to be done with caution. Closer investigations using phylogenetically independent analyses across floras evolved in high vs. low fire regimes, or sites with high vs. low resources, are needed to explore this issue further.

From Community Response to Ecosystem Functioning: Some Generalizations

Independence Between Disturbance Response and Ecosystem Processes: A General Rule?

The limited degree of convergence between disturbance response traits and ecosystem effect traits – seen in Table 1 for response to fire and flammability or primary productivity – is not surprising, for several reasons. As hypothesized within plant strategy schemes (Grime 1979; Westoby 1998), the disturbance and resource axes are mostly independent and therefore relate to distinct trait sets. Indeed, disturbance response relates to a great extent to regeneration traits whose lack of correlation with adult traits, more relevant to response to environmental resource factors, is notorious (Leishman & Westoby 1992; Shipley et al. 1989). Seed mass has been used as the single soft trait that simultaneously captures aspects of regeneration (dispersal, seed persistence, recruitment success, Thompson, Band & Hodgson 1993; Westoby, Jurado & Leishman 1992) and seedling response to environmental stress (drought, Jurado & Westoby 1992; shading, Walters & Reich 2000). Disturbance response also involves demographic rather than physiological traits. Although Silvertown et al. (1993) proposed to match Grime's plant strategies with demographic strategies described by allocation to survival (stress-tolerant), fecundity (ruderal) and growth (competitor), their scheme was only a first step from which more general relationships between demographic and physiological traits or their soft proxies remain to be established.

The relative independence of traits relevant to disturbance response and those involved in ecosystem effects supports the redundancy and insurance hypotheses (Lawton & Brown 1993; Walker 1992). If disturbance selects species according to traits that are unconnected with functional effects, then biogeochemical cycles will be maintained. In Australian semiarid rangelands, the distribution of effects traits (plant height, SLA, longevity, total biomass, leaf litter quality) was unchanged in heavily compared to lightly grazed communities (Walker et al. 1999). Larger numbers of quantitative comparisons of effect traits before and after disturbance are needed to establish the range of applicability of this hypothesis in terms of disturbance characteristics, initial diversity and abiotic conditions.

Trait Correlations and Trade-Offs: Usefulness and Limitations

In the previous discussion we pointed out potential linkages between response and effect traits, either directly through shared traits or through trait correlation. Plant strategy schemes and other proposed functional trait lists have emphasized the usefulness of correlation among traits, either to infer process from easily measured structure (Hodgson et al. 1999), or to capture several response or effect processes with few traits (Weiher et al. 1999; Westoby 1998). However, many correlations that hold true over entire floras or across steep environmental gradients may no longer be relevant to less contrasted ranges of conditions, or within species assemblages that present little variation in those key traits (such as assemblages with a single life form, McIntyre et al. 1999). In addition to the weakening of correlations when trait ranges are narrow, the use of surrogate correlated traits can lead to a loss of information unique to particular traits. This is particularly true for loose correlations, such as those involving seed mass to infer persistence in the seed bank or dispersal ability (Weiher et al. 1999).

Growth forms provide the most interesting and extreme example of the usefulness and limitations of trait correlation patterns (Chapin 1993). Many analyses of trait correlations within floras have highlighted that growth form captures patterns of variation in several important functional traits, and is one of the best correlates of plant regional distributions (Chapin et al. 1996a; Díaz & Cabido 1997; Leishman & Westoby 1992; Raunkiaer 1934). Growth form can then be a surrogate for other traits in vegetation containing sufficiently wide variations in growth form. This observation has been applied for models of vegetation response to global climate that use classifications based on subdivisions of growth forms, assuming that key traits for biogeochemical cycling are constant within them (Foley et al. 1996; Steffen et al. 1996). However, we know little about the mechanisms by which response and effect traits are represented by growth form, other than the resource capture–conservation trade-off. Raunkiaer"s (1934) life-form classification was initially designed with a well defined function in mind – response to cold winters – related to a well identified trait – the position of dormant meristems over this unfavourable period. Nevertheless, it is also relevant to other environmental responses (Table 1), but relationships between dormant meristem position and traits relevant to these responses need to be clarified. Life form has some relationships with stem height, leaf characteristics (Garnier et al. 2001) and, to some extent, phenology, and hence with primary productivity. It is also a robust predictor of responses to disturbance (McIntyre et al. 1999) as it reflects strategies and associated traits for avoidance (phenology and position of vulnerable meristems), tolerance (resprouting capacity), and regeneration through seeds or underground organs.

Conclusion

We present a conceptual framework aiming to unify a series of ideas drawn from community (Keddy 1992); ecosystem (Chapin, Autumn & Pugnaire 1993; Eckardt et al. 1977); and evolutionary (Solbrig 1993) ecology. Although precursors of this framework have been presented previously (Chapin et al. 2000; Díaz et al. 1999; Grime 2001), we have now linked disconnected conceptual elements and a suite of empirical data. Our framework is built on three key tenets (Fig. 1), as follows.

  • 1The keystone hypothesis is that traits can simultaneously explain individual plant responses to biotic and abiotic factors, and ecosystem effects such as biogeochemical cycling and propensity to disturbance.
  • 2Ecosystem functioning is the end result of the operation of multiple filters at a hierarchy of scales which, by assembling individuals with appropriate responses, result in communities with varying trait composition (Woodward & Diament 1991; Keddy 1992). Ecosystem functioning is predictable from composition if those traits involved in the response to environmental filters can be used to estimate ecosystem processes. Rather than discontinuous classifications into functional types, the use of continuous traits representing changes in the intensity of processes is likely to make this linking more operational.
  • 3Functional linkages and trade-offs among traits that each relate to one or several processes determine whether filtering by different factors matches or not, and whether ecosystem effects can be deduced easily from knowledge of the filters. However, linkages and trade-offs must be used with caution, depending on the scale, environmental conditions and evolutionary context.

Although the detailed examples presented here focus on two well documented linkages between response and effects, similar analyses should be developed for other environmental factors and ecosystem processes, including nutrient or water gradients and cycling, and grazing and palatability. We predict that overlap between traits determining response and those determining effects will be most common for biogeochemistry, where ecosystem fluxes may be calculated by scaling-up from individual physiological traits (Schimel et al. 1995) because the traits concerned relate to fundamental resource capture–conservation trade-offs. On the other hand, we expect little convergence between traits and processes associated with disturbance response and those relating to ecosystem functioning. Cases of overlap (leaf area and leaf toughness) correspond with traits that are involved in the acquisition–conservation trade-off, the exact primary function of which has been debated.

Our framework offers potential for developing a better understanding both of the role of biodiversity in ecosystem functioning, and of their coupled vulnerabilities to local and global environmental changes. It should assist researchers in selecting traits covering a broad range of processes which could be used to design observations, experiments and models. Finally, although the framework is presented here at ecosystem level, its applicability to the landscape and larger scales needs to be considered.

Acknowledgements

This work resulted from the GCTE workshops ‘Plant Functional Types: Concepts and Applications’ and ‘Building a Global Key of Plant Functional Types’, sponsored by the French Ecophysiological Society and the Dutch Global Change Committee, respectively. Sandra Díaz and Karl Grigulis helped substantially with successive drafts of the manuscript.

Ancillary

Advertisement