The world-wide ‘fast–slow’ plant economics spectrum: a traits manifesto


  • Peter B. Reich

    Corresponding author
    1. Department of Forest Resources, University of Minnesota, St. Paul, MN , USA
    2. Hawkesbury Institute for the Environment, University of Western Sydney, Penrith, NSW , Australia
    Search for more papers by this author


  1. The leaf economics spectrum (LES) provides a useful framework for examining species strategies as shaped by their evolutionary history. However, that spectrum, as originally described, involved only two key resources (carbon and nutrients) and one of three economically important plant organs. Herein, I evaluate whether the economics spectrum idea can be broadly extended to water – the third key resource –stems, roots and entire plants and to individual, community and ecosystem scales. My overarching hypothesis is that strong selection along trait trade-off axes, in tandem with biophysical constraints, results in convergence for any taxon on a uniformly fast, medium or slow strategy (i.e. rates of resource acquisition and processing) for all organs and all resources.

  2. Evidence for economic trait spectra exists for stems and roots as well as leaves, and for traits related to water as well as carbon and nutrients. These apply generally within and across scales (within and across communities, climate zones, biomes and lineages).

  3. There are linkages across organs and coupling among resources, resulting in an integrated whole-plant economics spectrum. Species capable of moving water rapidly have low tissue density, short tissue life span and high rates of resource acquisition and flux at organ and individual scales. The reverse is true for species with the slow strategy. Different traits may be important in different conditions, but as being fast in one respect generally requires being fast in others, being fast or slow is a general feature of species.

  4. Economic traits influence performance and fitness consistent with trait-based theory about underlying adaptive mechanisms. Traits help explain differences in growth and survival across resource gradients and thus help explain the distribution of species and the assembly of communities across light, water and nutrient gradients. Traits scale up – fast traits are associated with faster rates of ecosystem processes such as decomposition or primary productivity, and slow traits with slow process rates.

  5. Synthesis. Traits matter. A single ‘fast–slow’ plant economics spectrum that integrates across leaves, stems and roots is a key feature of the plant universe and helps to explain individual ecological strategies, community assembly processes and the functioning of ecosystems.


Two central and intertwined goals in ecology and evolution are to understand trade-offs that underpin ecological strategies and to identify attributes of species that are responsible for those trade-offs (Raunkiaer 1934; Grime 1979; Chapin 1980; Noble & Slatyer 1980; Körner 2003; Westoby 1998; Craine 2009). ‘Strategy’ is defined here (sensu Westoby 1998) to mean ‘how a species sustains a population… (given that it is) operating in the presence of competing species, in varied landscapes and under regimes of disturbance’. The ‘gold standard’ in meeting these two goals is to develop a general theory that explains trade-offs in as many contexts as possible using the fewest but most critical attributes of species; this paper focuses on a subset of those attributes relevant to higher plants, the plant functional traits directly relevant to resource economics at organ, individual and ecosystem scales.

Plants possess characteristics, or trait values (traits hereafter), at tissue-to-organismal scales that reflect their evolutionary history and mould their performance (Grime 1977; Chapin 1980; Bond 1989; Lambers & Poorter 1992; Reich, Walters & Ellsworth 1992; Lavorel & Garnier 2002; Körner 2003; Reich et al. 2003; Westoby & Wright 2006; Cavender-Bares et al. 2009). Traits, including functional traits, therefore offer clues and insights regarding how and why a plant may behave as it does, where it grows and where it does not, how it interacts with other plants, and how it influences the abiotic and biotic environment around it (Fig. 1). An explicitly economic approach to plant functional trait ecology has deep roots, including the work of Bloom, Chapin and Mooney (1985), Givnish (1986) and others. Such an approach is attractive because it provides a conceptual framework that enables us to link the physiology and morphology of a taxon to its environmental and resource tolerance, as well as its contributions to ecosystem function.

Figure 1.

Illustration of traits, resources and linkages across scales. Relationships between organ traits shown in the expanded box; relationships across scales shown otherwise. Hypotheses (H1–H9) are shown; when in parenthesis, it suggests an indirect pathway. Note, although not shown, attributes at organism, community and ecosystem scales can be considered traits and represent the integration of organs at individual scale or the community-weighted mean and variance in the value of any given trait at the scale of community, ecosystem or landscape.

The terms ‘trait’ and ‘plant functional trait’ are widely and variably used, however, and encompass ecophysiological, life history, demographic, response and effect attributes, at organ, individual, population, community and ecosystem scales (Violle et al. 2007). Violle et al. (2007) define ‘functional traits’ as ‘morpho-physio-phenological traits which impact fitness indirectly via their effects on growth, reproduction and survival, the three components of individual performance’. Accordingly, functional traits are relevant to life-history theory, which addresses how selection acts to optimize fitness of organisms.

At the physiological level for individuals, trade-offs are caused by allocation of limited resources to one purpose vs. another; for example, individuals with lower reproductive effort may have a longer life span or vice versa. Herein, I focus on a subset of plant functional traits, those directly involved in the acquisition, processing and conservation of resources, and hence a subset that can be considered ‘economic’ from a resource analysis perspective. The resource economic traits that are the focus of this piece influence life history largely by influencing growth vs. survival trade-offs that impact performance (see Figures and associated references) across the continuum of low to high levels of resources (such as light, water or nutrients). As a consequence, this piece gives short shrift to plant traits specifically related to reproduction, such as seed size and dispersal, and related trade-offs, such as between colonization and competition.

It may be helpful to consider at what scales (spatial, taxonomic, biogeographical) ‘resource economics’ theory applies. There is substantial evidence that (i) predicted trait trade-offs occur and explain much about species performance and community assembly for different species within a given community (e.g. Reich, Walters & Ellsworth 1994; Poorter & Bongers 2006; Poorter et al. 2010), for different congeners across local landscapes (Cavender-Bares, Kitajima & Bazzaz 2004; Givnish, Montgomery & Goldstein 2004; Savage & Cavender-Bares 2012), and among populations within a species across landscapes (Oleksyn et al. 1998), (ii) that the predicted species trait trade-offs that occur in one site are predictably similar among other sites locally, regionally or globally (Reich, Walters & Ellsworth 1997; Wright et al. 2004), and (iii) that the aggregate traits of communities and ecosystems drive productivity and biogeochemical cycling (Garnier et al. 2004; Cornwell et al. 2008; Ollinger et al. 2008; Laughlin 2011; Reich 2012; Reich et al. 2012). Although I do not make a formal analysis of the scale dependency of economic spectrum theory, this review does address how it applies at a variety of hierarchical scales.

The focus on resource economics taken in this review sets it apart from prior work on plant traits and strategies. The approach taken herein is like the leaf-height-seed strategy scheme of Westoby (1998) in focusing on the traits themselves. It is narrower in the range of traits focused on (as it ignores height and seeds), but fuller in its exploration of traits directly involved in resource economics. Moreover, the functional trait approach adopted herein is radical in its simplicity because, as the leaf-height-seed strategy scheme of Westoby (1998), it considers traits themselves as the elements of plant strategy, rather than using traits to help assign species to a priori (and difficult to define) conceptual strategies, as the C-S-R scheme of Grime (1979). By using traits as the central elements, defined and quantifiable metrics can be related to other quantifiable metrics such as growth, abundance and distribution. Nonetheless, many of the mechanisms invoked along the C-S axis of the Grime triangle and the ‘leaf’ component of the Westoby LHS scheme are paralleled by findings from economics trait spectra studies (Grime 1979; Reich, Walters & Ellsworth 1997; Wright et al. 2004; Westoby 1998), despite different ways of describing and labelling those strategies.

Conceptual Premise

The premise of this piece is (i) that traits are central to coordinated trade-offs between resource acquisition and/or process rates on the one hand (i.e. productivity) and resource conservation (i.e. persistence) on the other that help determine where on a growth vs. survival trade-off, any taxon is located for a given set of conditions (e.g. Lambers & Poorter 1992; Wright et al. 2004, 2010; Kobe 1999), and (ii) that there is sufficient variation in time and space within (and across) communities, habitats and ecosystems (e.g. Wright et al. 2004; Liu et al. 2010) that every position along those trade-off surfaces represents potentially successful strategies. This premise is consistent with evidence and theory that species trade-off the ability to be effective exploitative resource competitors, that is, to grow quickly and thereby usurp relatively more resources when these are abundant, with the ability to avoid mortality under low-resource conditions (Grime 1979; Tilman 1982, 1987; Kobe et al. 1995; Pacala et al. 1996; Walters & Reich 1996; Kobe 1999; Aerts & Chapin 2000; Russo et al. 2005; Poorter & Bongers 2006; Dybzinski & Tilman 2007; Craine 2009; Kursar et al. 2009; Wright et al. 2010). Moreover, in many cases, species good at avoiding mortality at low resource supply also further reduce their supply under such conditions (Tilman & Wedin 1991; Reich et al. 2003). Herein, I build on prior plant trait economics research (e.g. Wright et al. 2004; Chave et al. 2009; Freschet et al. 2010, 2013; Mommer & Weemstra 2012) and ask whether trait relations and associated trade-offs are consonant among organ types and the ecologically most important plant resources, that is, whether a plant economics spectrum exists for all three major resources – carbon, water and nutrients – built on the foundations and linkages of leaf, stem and root economics spectra.

My overarching hypothesis is that the ubiquity of strong selection along trait and life-history trade-off axes, in tandem with biophysical constraints (Reich et al. 1999), results in convergence for any taxon on a uniformly fast, medium or slow strategy (i.e. having high, medium or low rates, respectively, of resource acquisition and processing) for all organs and all resources. I posit (i) that being fast at any (leaf, stem or root) organ level at acquiring or using C, nutrients or water requires being fast for the other resources at the same organ level, and (ii) being fast for all resources at any one organ level (e.g. the leaf level) requires being fast for all resources at the other organ levels (e.g. stem and root levels). Thus, despite different traits being of central importance in terms of selection under different conditions (e.g. among different resource and disturbance regimes) (Diaz, Cabido & Casanoves 1998), the coordination among traits, organs and resources results in fast or slow plants in different systems still converging on a fast or slow, respective, ecological strategy. For example, having low respiration may matter most in low light when C is most limiting, having low nutrient requirements may matter most when nutrients are limiting, and having drought tolerance may matter most in arid environments. Yet, in all cases, plants with the slow strategy will have low respiration, low nutrient concentrations, denser tissues and a lesser capacity to move and lose water. In other words, because of the small number of coupled resources of plant economics, a fast or slow strategy requires similar sets of leaf, root and stem traits regardless of whether the main limiting factor is light, N, P, water or temperature. In essence, the nature of plant integration sets the ‘fast–slow’ template that serves as the backbone for a more nuanced ‘low nutrient’ or ‘low light’ or ‘low water’ strategy. This over-arching hypothesis is built on five specific ideas, as follows.

First, selection is a key driver: being fast at acquiring and processing carbon, water or nutrients in leaves, stems or roots is advantageous only when acquiring and processing of all resources is fast for all organ systems, because otherwise plants will possess excess capacity which is costly and wasteful. Secondly, biophysics is a key constraint: being fast at processing of carbon, water and nutrients for leaves, stems or roots is possible only when processing of other resources is fast for all organ systems, because fast acquisition and processing of carbon requires fast acquisition and processing of water and nutrients, and fast acquisition and processing of water and nutrients requires fast acquisition and processing of carbon. As evolution and biophysics both set trade-off constraints (Reich et al. 1999; Körner 2013) and drive multiple resource acquisition to be coupled and linked among organ systems, these generally constrain plants to being generally fast or slow (or in-between) across all resources and organ systems.

Thirdly, having fast traits is advantageous in high-resource environments, but due to excess costs, is disadvantageous in low-resource and other low growth capacity (e.g. very cold) environments (Grime 1977, 1979). For light, nutrients and water, the ‘productive strategy’ implemented by rapid resource acquisition involves high use of resources (C, nutrients, and water) to rapidly acquire C (and as well nutrients and water). A plant must spend considerable resources (i.e. to build, deploy and use acquisitive canopy and root systems) to obtain more resources than its neighbours, and this strategy can be successful only when that investment is scaled to the available resources such that the return on investment is high enough to offset the costs of such resource investment. In other words, over-investing (relative to resource supply) in expensive resource acquisition machinery is not a viable strategy, and fast plants suffer from this problem when resources are scarce. Slow traits in contrast are advantageous in low-resource settings because resource conservation enhances survival, but being slow is itself a disadvantage at any point in space or time where resources are abundant, if others are faster. The slow return strategy involves considerable savings, such as reduced respiratory and turnover C loss in shade, reduced water loss in semi-arid conditions or reduced foraging costs and nutrient turnover costs in infertile conditions (Grime 1965; Reich, Walters & Ellsworth 1992; Craine 2009). Such savings allow plants to either tolerate low-resource conditions, eventually draw down resources to levels competitors cannot tolerate, or both (Tilman & Wedin 1991; Baltzer & Thomas 2007a; Dybzinski et al. 2011). These advantages and disadvantages result in a trade-off in performance (sensu ‘performance traits’, Violle et al. 2007) at high vs. low resource supply, as it is impossible to be equally successful at all resource supply levels.

Fourthly, spatial and temporal variation in resource supply and microenvironment are sufficient locally (within a stand) that taxa located all along the fast–slow trait axis are successful. Finally, the mean and variance of community-scale traits (‘effect’ traits; Lavorel & Garnier 2002) determine whether ecosystem-scale elemental fluxes and associated coupled plant-soil system processes are fast or slow. These effects might in some cases also exert selective pressure on individuals. For example, perhaps shade-tolerant species are selected to cast deep shade as canopy dominants, acidophilic species to acidify soils and low N-tolerant strong N competitors to drive down the available pool below levels other species can tolerate.

Central Questions

Given the above context, the goal of this paper is to explore trait variation between organs and across key resources, and the impact of this variation on individuals, community assembly processes and ecosystem-scale function. To do this, I focus on three central questions relevant to whether the LES idea (Reich, Walters & Ellsworth 1997; Wright et al. 2004) can be broadly extended to water – the third key resource –stems, roots and entire plants and to individual, community and ecosystem scales. The focus is largely on cross-species contrasts. Figure 1 diagrams the hypothesized relationships between these traits, properties and processes; throughout this paper, I will refer to these hypotheses using shorthand (e.g. H1 for Hypothesis 1, etc).

  1. Do economic trait spectra exist for stems and roots, as well as leaves (H1)? If so, do these apply to water as well as to carbon and nutrients? Do functional trait spectra for stems and roots mirror those for leaves (H1A to H1L)? Do trait spectra reflect evolutionary history, microenvironment and resource supply (H8–9)?
  2. Are economic traits correlated with performance measures and if so, consistent with hypothesized trait-based mechanisms (H2)? Does trait variation help explain differences between taxa in growth and survival, as well as the growth–survival trade-off, and thus the distribution of species and the assembly of communities across light, water, nutrient and thermal gradients (H3)?
  3. Does the aggregation of traits at the community and ecosystem scales help explain system-scale to biome-scale functioning, and feedbacks to biogeochemical processes (H4–7)? Would consideration of traits, and their incorporation into model logic, improve ecosystem- to global-scale models of vegetation change and C, water and nutrient cycling?

To assess the above questions, I review literature relevant to the hypotheses in Fig. 1. In doing so, I focus largely on traits with direct relevance to terrestrial resource economics (light, C, nutrients and water) and ignore other important traits. As a result, many traits relevant to reproduction (seed size, pollination, dispersal, germination, clonality, sprouting ability), disturbance (e.g. fire, wind, snowload, flooding), biotic interactions (e.g. herbivory, disease) and design (e.g. adult height, allometry, architecture) are given short shrift in this review, as are plasticity, plant size and colonization capacity. These omissions are in large part to keep this piece to a manageable length and focus. Moreover, despite their ecological importance, prior work suggests these other attributes largely add other additional layers, or axes, of complexity relative to individual performance, rather than undermining the central trends visited in this review.

This review covers territory addressed previously (e.g. Grime et al. 1997; Chapin 1980; Lavorel & Garnier 2002; Körner 2003; Diaz et al. 2004; Westoby 1998; Westoby & Wright 2006; Craine 2009), but differs from these in the integrated examination of leaf, stem and root traits relevant to water, carbon and nutrient economics, across hierarchical scales and processes, and in the degree to which it touches on plant hydraulics, phosphorus as a limiting element, and scaling from tissue to ecosystem and beyond. Thus, hopefully, this review can be a useful complement to the many earlier works it builds upon.

Trait–trait variation

The discovery of coordinated variation in the longevity, morphology [e.g. specific leaf area (SLA)], chemistry (e.g. [N], [P]) and metabolism (e.g. photosynthetic capacity) of leaves (e.g. Reich, Walters & Ellsworth 1992; Reich et al. 1999; Wright et al. 2004) helps explain species strategies, community assembly and ecosystem structure and function (e.g. Garnier et al. 2004; Wright et al. 2004; Poorter & Bongers 2006; Cornwell & Ackerly 2010; Kattge et al. 2011; Reich 2012; Körner 2013). Recently, trait ecology has embraced other traits and processes, such as leaf hydraulics, wood density, fine-root demography, senesced leaf decomposition and many others (see references below). In the following sections, I focus on what has been learned about co-variation in leaf economic traits, in other leaf traits and in stem and root traits, and about their variation across local- to macro-scale environmental gradients.

Original Leaf Economic Spectrum Traits

The LES (Table 1) quantified by Reich, Walters and Ellsworth (1992, 1997), Reich et al. (1999) and Wright et al. (2004) has as its core concept the productivity-persistence trade-off and contrasts inexpensive short-lived leaves with rapid return on C and nutrient (N, P) investment with costly long-lived leaves with slow returns on investment (H1). However, these examples represent the end-members of the spectrum, and LES studies have identified that a range of successful strategies exist in every community/ecosystem (H2–3) and that expected leaf trait differences between broad environmental gradients were more modest than originally hypothesized (H8), because of the diversity of successful strategies at local scales (Wright et al. 2004, 2005).

Table 1. List of abbreviations
AMArbuscular mycorrhizasNA
AareaLight-saturated photosynthetic capacity (at ambient CO2) per unit leaf areaμmol m−2 s−1
AmassLight-saturated photosynthetic capacity (at ambient CO2) per unit leaf massnmol g−1 s−1
AmaxLight-saturated photosynthetic capacityNA
GPPGross primary productivityg C m−2 year−1
KleafLeaf hydraulic conductancemmol m−2 s−1 MPa−1
kstemStem hydraulic conductivityvarious
LAILeaf area indexLeaf area per unit ground surface area
LESLeaf economics spectrumNA
LMALeaf dry mass per unit areag cm−2
MAPMean annual precipitationmm
MATMean annual temperature°C
RGRRelative growth rateg g−1 day−1
SLASpecific leaf aream−2 g−1
SRLSpecific root lengthm−1 g−1

A recent re-examination (Osnas et al. 2013) of the LES data in Wright et al. (2004) supported the fundamental role in leaf economics for the SLA vs. leaf life span trade-off and formalized the treatment of covariance of SLA with mass- and area-based expressions of leaf traits. Osnas et al. (2013) noted that the bi-variate relationship of mass-based photosynthesis (Amass) vs. mass-based N (Nmass = [N]) includes co-variation with SLA, because SLA varies positively with Nmass and thus the Amass vs. Nmass relation ‘includes’ effects on Amass due to SLA. Because Nmass and SLA are positively related and both independently influence Amass positively (Reich, Ellsworth & Walters 1998), the Amass vs. Nmass relationship is steeper than (and over-estimates) the part of that relation that is due strictly to N (Osnas et al. 2013). For similar reasons, the area-based Amax vs. N (Aarea vs. Narea) relationship underestimates the influence of N on Amax because SLA and Narea are negatively related (hence at rising Narea, the ‘low SLA’ effect drags down Aarea). A normalization-independent Amax vs. N relationship removes the SLA part of the effect (Osnas et al. 2013). This does not erase the fact that leaves with either higher Nmass or higher SLA do have higher Amass; nor that because of this, plus co-variance of Nmass and SLA, leaves with higher Nmass do have higher Amass. Instead, normalization-independent relations account for such covariances and highlight their hidden role when interpreting solely mass- or area-based bi-variate relations.

Additionally, when traits measured on a leaf area basis that are uncorrelated with SLA are converted to expression on a mass basis by multiplying by SLA (e.g. Amass = Aarea × SLA), strong positive correlations are generated with SLA and with other mass-normalized traits (Osnas et al. 2013). This shows that mass-based correlations could in theory arise from a random selection of area-based traits. However, there is little evidence to suggest that in nature, traits of co-occurring taxa are just random assemblages (and instead perhaps mass-based traits are selected to roughly equalize productivity per unit leaf area as well as return on unit mass investment, e.g. Falster et al. 2012). Additionally, Sack et al. (2013) suggest that while such linkages may be consistent with random mathematics, they still reflect physically based mechanistic processes relevant to trait integration and plant function. For example, all else being equal, thicker cell walls will decrease SLA and decrease Amass. The positive relationship of Amass with SLA is not trivial in meaning and will imply, for example, that low-SLA leaves will have lower return per mass investment per time (Westoby, Wright & Reich 2013).

Leaf diffusive conductance was considered a key trait in the leaf economic spectrum in some early LES papers (e.g. Reich, Walters & Ellsworth 1992; Reich et al. 1999; Wright, Reich & Westoby 2003), but received scant attention in others (Reich, Walters & Ellsworth 1997; Wright et al. 2004), perhaps to some extent because Amax and leaf diffusive conductance scale similarly among species. However, the relationship of photosynthesis to leaf diffusive conductance is of course by no means constant, and perhaps some economic trait research overemphasized the role of carboxylation capacity (associated with leaf [N]) and overlooked that of stomatal limitation (associated with stomatal conductance). Recent papers by Medlyn et al. (2011) and Prentice et al. (2011, 2014) are notable in showing how the close coupling of water and carbon flux and the importance of efficient (and perhaps optimal) use of both resources regulates water loss and carbon gain and allows accurate predictions of conductance as a function of environmental conditions (temperature, vapour pressure, aridity, [CO2]). These studies advance both our fundamental understanding of the links and trade-offs between water and carbon exchange, and the applicability of harnessing leaf physiology in a more sophisticated way in ecosystem- to global-scale models.

The LES incorporates physiology, ecology and evolution and can be viewed from each of these perspectives (e.g. Reich et al. 1999; Shipley, Vile & Garnier 2006; Donovan et al. 2011). Donovan et al. (2011) identified both abundant genetic variation for the LES traits and genetic correlations that are orthogonal to the main axes of LES trait co-variation. From this, they concluded that genetic constraints do not limit the LES trait combinations that can arise, and thus, natural selection is probably the most important factor influencing the evolution of the LES. Further exploration of these questions is presented elsewhere in this Special Feature, for instance, using Helianthus as a model system (Donovan et al. 2014) or evaluating the contributions of different evolutionary lineages to modern-day functional trait diversity (Cornwell et al. 2014).

Despite the primacy of SLA (along with leaf life span) in generating the leaf economic spectrum and its area- and mass-based expression, the focus on SLA has been questioned by a number of authors. For example, SLA has shortcomings as a measure of structure, given that it combines thickness and density, and may not effectively capture variation in important leaf mechanical or physical properties that are related to leaf life span, plant–herbivore interactions, litter decomposition and nutrient cycling (Garnier et al. 2004; Fortunel et al. 2009; Hodgson et al. 2011). However, based on analyses of almost 2000 species, Hodgson et al. (2011) concluded that SLA often discriminates between communities better than leaf dry matter content (dry leaf mass/water-saturated fresh leaf mass) because SLA is influenced by both shade and soil fertility, whereas leaf dry matter content largely reflects soil fertility and thus is a better predictor of soil fertility per se. Moreover, based on data for 2819 species from 90 sites world-wide, Onoda et al. (2011) noted that three components of mechanical resistance (work to shear, force to punch and force-to-tear) were all linearly correlated with 1/SLA, tissue density and lamina thickness, and surprisingly, much better correlated with the first than the latter two. This suggests that SLA better conveys meaningful information about leaf mechanical properties than either of its components. Although understanding the ecological roles of leaf density, thickness and dry matter content is important, neither Hodgson et al. (2011) nor Onoda et al. (2011) suggest that any of these is more informative than SLA as a general measure of leaf structure.

Leaf Traits Beyond the Original LES

A recent focus on leaf hydraulics bridges stem hydraulics and leaf economics. Leaf hydraulics should be important to both water and C economics, and their coupling, given the large fraction of total plant hydraulic resistance to water flow that occurs at the leaf level, and the fact that leaf gas exchange should be related to leaf xylem hydraulic traits due to the serial positioning of xylem and stomata in the flow path of water through the plant (H1) (Brodribb et al. 2005; Brodribb, Feild & Jordan 2007).

In studies of disparate species (widely varied phylogeny, leaf structure, leaf life span, phylogeny, geography), Brodribb et al. (2005) and Brodribb, Feild & Jordan (2007) found that leaf hydraulic conductance (Kleaf) was positively related to stomatal conductance and photosynthetic capacity (H1), indicating coordination of leaf-level liquid and vapour conductances and fluxes (Fig. 2). Mechanistically, higher Kleaf enables higher stomatal conductance, which in turn allows higher photosynthesis. The wider vessels of angiosperms enable higher Kleaf than do the narrower tracheids of conifers and ferns (Brodribb et al. 2005), although considerable heterogeneity occurs within each group.

Figure 2.

Relationships of leaf diffusive conductance (n = 58, r2 = 0.79) and photosynthetic capacity (n = 43, r2 = 0.93) to mean leaf hydraulic conductance (Kleaf), illustrating hypothesis H1A. Best fit relationship and 95% confidence intervals shown. Data for tropical and temperate angiosperms and gymnosperm trees, ferns and club mosses, redrawn from Brodribb et al. (2005) and Brodribb, Feild & Jordan (2007).

Brodribb, Feild and Jordan (2007) also advanced our understanding of the links between the structure of the leaf vein system, water transport and photosynthetic capacity. They suggested that the hydraulic and coupled photosynthetic performance of a leaf should be related to the length of mesophyll tissue to be traversed as water moves from a vein ending to the stomatal site of evaporation. This follows from the knowledge that hydraulic resistance to water flow is much higher passing through leaf mesophyll than through vein xylem; as a result, the distance water must flow through the mesophyll before evaporating (a function of the positioning of leaf minor veins) should regulate the leaf's hydraulic transport efficiency.

They produced evidence that strongly supports the hypothesis that the length of the hydraulic pathway through the mesophyll regulates Kleaf and thus indirectly Amax. Among a broad set of angiosperms, photosynthetic capacity is strongly correlated with vein density (also called vein length per area, Sack et al. 2013) and even more so with the proximity of veins to the evaporative surfaces of the leaf (as measured by the mean maximum mesophyll path length) (Brodribb, Feild & Jordan 2007). The influence of vein positioning over leaf water and carbon flux rates was similar across a range of species that differed dramatically in their evolutionary history and ecology. Leaf vein length per leaf area is analogous to SLA or specific root length (SRL) in the sense of being a measure of resource flux capacity (Sack et al. 2013) and thus also contributes to ‘fast–slow’ contrasts through its influence on leaf carbon and water fluxes. Although the total length of the transpiration pathway in trees can exceed 100 m, traits relevant to the last few tens of microns of that path play a key role in regulating the photosynthetic and hydraulic performance of the individual plant.

Thus, the links between leaf hydraulics and gas exchange appear to be strong, consistent with theoretical modelling linking the two (Katul, Leuning & Oren 2003). Moreover, the evolution of leaf vein systems that could support high water flux rates and other ‘fast return’ LES traits (Boyce et al. 2009; Brodribb & Feild 2010; Feild et al. 2011; Sack & Scoffoni 2013) may have been crucial to the evolution and successful spread of the angiosperms. The work of Brodribb and colleagues along with more recent work on this theme (Sack & Scoffoni 2013; Sack et al. 2013) provides strong support for the close coupling of leaf traits that regulate carbon and water flux (supporting H1A).

Stem Traits

The past decade has shown growth in understanding of stem traits and their importance. Chave et al. (2009) proposed a wood economics spectrum, based on the relationships of wood density (n = 8412 taxa) with mechanical and hydraulic properties. They posited that high wood density would be associated with a ‘slower’ potential to move water but with stronger and more flexible mechanical properties and greater protection from drought stress. The hypothesis was supported in several respects. Wood density is correlated with mechanical traits such as strength and bendability (Chave et al. 2009) and with performance (see below). Evidence demonstrates a considerable degree of coordination of stem hydraulic properties with wood density (Sobrado 1986; Meinzer et al. 2008a,b; Chave et al. 2009; Poorter et al. 2010; Russo et al. 2010; Zanne et al. 2010; Markesteijn et al. 2011). For example, in studies of tropical wet and dry forest trees, high wood density was associated with low stem hydraulic conductivity (kstem), whether expressed on a sapwood or leaf area basis (Figs 3 and 4).

Figure 3.

Relationships between leaf photosynthetic rate per unit area, maximum leaf-specific stem hydraulic conductivity (kstem), minimum daily leaf water potential (Ψmin) and wood density for 20 lowland tropical forest canopy tree species in Panama, illustrating H1A, H1C, H1G, H1L. Redrawn from data in Santiago et al. (2004). Values are species means. Best fit relationship and 95% confidence intervals shown.

Figure 4.

Relationships between vulnerability to cavitation (the water potential at 50% loss of conductivity, P50) and leaf-area-specific and sapwood-area-specific stem hydraulic conductivity (kstem) and wood density, for tropical dry forest trees in Bolivia. Redrawn from data in Markesteijn et al. (2011).

Additionally, a variety of stem hydraulic traits are correlated with each other and differ between species with the three radically different wood types: coniferous, diffuse-porous and ring-porous (Sperry, Meinzer & McCulloh 2008). The packing function describes a strong trade-off and upper limit to the relationship between conduit frequency (the number per xylem area) and conduit diameter, which holds within and across species with differing xylem anatomy (Sperry, Meinzer & McCulloh 2008; McCulloh et al. 2010; Zanne et al. 2010). Largely as a result of this trade-off, species with strikingly different xylem anatomy have similar scaling of leaf area and stem hydraulic conductivity with stem diameter of a branch or trunk segment (McCulloh et al. 2010), consistent with ideas embedded in metabolic scaling theory (e.g. Enquist et al. 2007; Savage et al. 2010; Sperry et al. 2012). Although they can be logically considered as plant traits, several other stem hydraulic measures (e.g. water potential at loss of 50% of conductivity) are discussed in the latter section on performance.

To date, differences between taxa in stem metabolism involving C fluxes have been addressed largely independently of studies of water flow. Stem dark respiration varies in relation to area : mass proportions and tissue [N] similarly as in leaves. For example (Fig. 5), stem dark respiration was correlated with stem N similarly in different plant groups (Reich et al. 2008) and both were, not surprisingly, lower in larger stems with a much lower surface area to mass ratio (and likely greater fraction of metabolically inactive wood).

Figure 5.

Mass-based dark respiration (nmol g−1 s−1) in relation to tissue nitrogen concentration (mmol g−1). Both are expressed on a logarithmic (base10) basis. Data are shown for different plant groups and for all plant groups pooled, with organ types labelled with different colours. All relationships, < 0.0001; R2 ranged from 0.53 to 0.80. Note to improve visibility of each panel, scales are not identical among panels. However, the axis ratios are the same; hence, slopes may be compared between panels. Illustrates Hypotheses H1B, H1D, H1F. From Reich et al. (2008).

The studies described above support the idea of an integrated stem economics spectrum for coupled water, C and N relations. There is close coupling of stem hydraulic, leaf hydraulic and leaf C flux dynamics (Brodribb et al. 2005; Brodribb, Feild & Jordan 2007; Meinzer et al. 2008a,b); leaf and canopy C and N dynamics (Wright et al. 2004; Ollinger et al. 2008; Reich 2012); stem hydraulic conductivity with leaf area (McCulloh et al. 2010); and leaf area with C fluxes (Reich 2012; Stark et al. 2012). Together these strongly support the idea that stem traits associated with C, N and water dynamics represent a unified stem economics spectrum (H1) that is also linked with broader C, N and water scaling processes at a range of hierarchical (organ, individual, stand) scales.

Fine-Root Traits

Several studies have assessed whether the LES is paralleled by a similar root economic spectrum (Eissenstat & Yanai 1997; Pregitzer et al. 1998, 2002; Reich et al. 1998a,b; Craine et al. 2005; Tjoelker et al. 2005; Withington et al. 2006; Freschet et al. 2010; McCormack et al. 2012). These studies have almost entirely focused on fine roots, which is also the case in this review. Variation in below-ground plant traits remains poorly quantified compared with leaf traits; hence, any conclusions are still rather more preliminary than final. Moreover, variation in root dimensionality and size (branching order, diameter, etc) complicates both the very definition of what a fine root is, as well as operational root censusing (Guo et al. 2008).

Candidate traits for a root economic spectrum include root [N], root respiration, root longevity, SRL (length of root per unit mass), root diameter and root architecture. SRL, sometimes considered analogous to SLA as an indicator of uptake potential per g investment, has been correlated with root diameter, dry matter content, high levels of branching and low tissue density (e.g. Craine et al. 2005; Comas & Eissenstat 2009; Freschet et al. 2010; Holdaway et al. 2011). This suite of morphological and structural traits has been correlated (i) with root [N] in some cases (Reich et al. 1998a,b; Craine et al. 2005; Tjoelker et al. 2005; McCormack et al. 2012), but not others (Pregitzer et al. 2002; Withington et al. 2006; Comas & Eissenstat 2009; McCormack et al. 2012), and (ii) often with root respiration (Reich et al. 1998b; Tjoelker et al. 2005; Makita et al. 2012). In turn, root [N] has been shown to correlate with root respiration (Reich et al. 1998b, 2003, 2008; Tjoelker et al. 2005; Chen et al. 2010) (Fig. 5); and both root [N] and root respiration have been shown to correlate with root lifespan (Tjoelker et al. 2005; Withington et al. 2006; McCormack et al. 2012). Moreover, data from temperate grasslands, woodlands and forests show a significant trade-off (Fig. 6) between root life span and root [N] (H1). Although not as uniformly or strongly coordinated as the LES, on balance the evidence indicates that a root economics spectrum exists and represents a ‘fast–slow’ trade-off and associated strategy axis.

Figure 6.

Relationship of tissue%nitrogen vs. life span (days) for fine roots (black circles) and leaves (open grey circles). Leaf data from Wright et al. (2004). Root data compiled from Reich et al. (2001); Tjoelker et al. (2005); Withington et al. (2006), McCormack et al. (2012). The relationships are significant for leaves (< 0.0001, r = −0.65, n = 706) and for roots (< 0.0001, r = −0.46, n = 68). The reduced major axis relations shown for both separately (roots, thicker black dashed line; leaves, thinner grey dashed line); the slopes are not significantly different, but at any given life span, roots on average have lower% nitrogen than leaves. Illustrates H1B, H1F.

There is rising interest in considering mycorrhizal status as an additional root trait (Fig. 1). Brundrett (2002) concluded that increasing mycorrhizal dependence should be considered a ‘slow’ strategy and be associated with lower SRL, less root branching and longer root life span. The occurrence, and often dominance, of ectomycorrhizal plants in infertile ecosystems has also led to suggestions that they comprise a low-nutrient (slow) trait syndrome (Read 1991; Cornelissen et al. 2001). To address these ideas, Koele et al. (2012) identified 19 evolutionary clades of ectomycorrhizal plants and used a data set comprising 11 466 samples across c. 3000 species to test whether there were consistent shifts in leaf nutrients with the evolution of ectomycorrhiza. There were significant differences in foliar [P] but not [N] between ectomycorrhizal and non-ectomycorrhizal species when not considering phylogeny. However, there was no evidence of consistent differences in [P] or [N] between ectomycorrhizal clades and their nearest non-ectomycorrhizal relatives. Thus, the hypothesis that ECM species are characterized by different leaf nutrient status was not supported. However, ectomycorrhizal species may be better able to acquire and use organic nutrients (Read, Leake & Perez-Moreno 2004; Wurzburger & Hendrick 2009; Orwin et al. 2011; Phillips, Midgley & Brozstek 2013), which could help explain their success in ecosystems with low nutrient availability. Implications for biogeochemical processes are addressed below.

Relationships among Leaf, Stem and Root Traits

Various lines of evidence (summarized above) support hypothesized axes (H1) of multiple trait variation for leaves, stems and roots, each viewed independently (e.g. Wright et al. 2004; Craine et al. 2005; Chave et al. 2009; Baraloto et al. 2010). This leads to the question of whether leaf, stem and root trait syndromes are coordinated – representing a single axis of variation – or whether these are largely independent. Strong integration of traits of all three tissue types would be predicted if parallel ‘productive vs. persistent’ tissue strategies are advantageous at the whole-plant scale. Results for a subarctic flora suggest that there is some degree of coordination of root, stem and leaf traits, supporting the idea of a whole-plant-based strategy (Freschet et al. 2010).

In a broad survey of leaf and wood tissues from 668 Neotropical tree species, Baraloto et al. (2010) confirmed the existence of leaf and wood trait spectra, but found trade-offs in leaf economics and stem economics spectra to be independent. The sets of traits examined in Baraloto et al. (2010) was quite limited (to wood density and water content, and bark thickness), so perhaps the results would differ in a more comprehensive contrast. For example, Brodribb & Feild (2000), Santiago et al. (2004), and Campanello, Gatti and Goldstein (2008) found strong coupling of kstem with leaf photosynthetic capacity across species and light environments, echoing the coupling of hydraulic and leaf economic traits at the leaf level (Brodribb et al. 2005; Brodribb, Feild & Jordan 2007) mentioned earlier. Additionally, across a set of Nothafagus species, leaf and stem hydraulic conductivity were coupled, and both were inversely correlated with wood density (Bucci et al. 2012) and in a subarctic flora stem and leaf, chemical and structural traits were strongly correlated (Freschet et al. 2010). Other evidence for coupling of leaf and stem traits comes from Choat, Sack & Holbrook (2007), Meinzer et al. (2008a,b), Blackman, Brodribb & Jordan (2010), and Markesteijn et al. (2011). They observed a coordination of wood density, stem hydraulic conductivity, leaf gas exchange rates and leaf water potential (e.g. Figs 3 and 4). Trees with denser wood had lower kstem, more negative leaf water potential at minimum daily (Ψmin), at the turgor loss point (ΨTLP) and at the point of vulnerability to cavitation, and lower rates of photosynthesis and transpiration. Savage and Cavender-Bares (2012) found similar trade-offs in a study of co-occurring willows and poplars, with species with denser wood having more negative ΨTLP. Not every detailed study found close association of stem and leaf traits, though; Ackerly (2004) found independent stem and leaf trait syndromes in a study of 20 co-occurring chaparral shrubs. However, the preponderance of studies above suggest that coupling of stem and leaf traits is likely more common than suggested by the Baraloto et al. (2010) analysis, and a meta-analysis of a comprehensively broad set of leaf and stem traits should be informative.

Similarly, root traits likely mirror stem and leaf traits to some degree. Positive correlation between species for pairs of traits (e.g. leaf vs. root [N]; SLA vs. SRL; leaf vs. root life span; leaf vs. root respiration) has been observed in some studies, but not all (e.g. Craine & Lee 2003; Craine et al. 2005; Tjoelker et al. 2005; Kerkhoff et al. 2006; Withington et al. 2006; Freschet et al. 2010; Liu et al. 2010). Additionally, in a survey of > 2500 measurements from 287 species, Reich et al. (2008) observed consistent mass-based dark respiration–nitrogen scaling for roots, stems and leaves examined separately. This was true for life-forms (woody, herbaceous plants) and phylogenetic groups (angiosperms, gymnosperms) examined separately or pooled (Fig. 5). No consistent differences in the slopes of the log–log scaling relations were observed between organs or between plant groups. Similarly, a new compilation that compares life span vs. [N] correlations for roots (n = 68) with those of leaves (n = 706) shows similar relations for leaves and roots (Fig. 6). The slope of tissue [N]–longevity did not differ significantly for leaves and roots, but at any common life span, leaves have higher [N] than roots. These results indicate that different organs have similar respiration [N] and longevity [N] relationships.

Other studies that measured leaf, stem or whole-plant traits as well as root traits also support the notion of trait coordination between organs. For example, McCormack et al. (2012) found shorter root life span and higher root [N] in faster-growing trees with lower wood density (all fast traits), and Comas and Eissenstat (2004) found faster-growing taxa to have smaller root diameter and high SRL than congeners. As higher root [N] is also associated with short root life span (Fig. 6) (Craine et al. 2002; Withington et al. 2006) and with high root respiration (Fig. 5) and high plant growth rates (Reich et al. 1998a,b, 2003; Tjoelker et al. 2005; Reich et al. 2008), it seems likely that fine-root chemistry, metabolism and duration are related to stem, leaf and plant traits as well. As with stems (Meinzer et al. 2010), the greater relative variation in size, order, age and architectural position of fine roots (Pregitzer et al. 2002; Guo et al. 2008) that is included in the measurements available in the literature (as compared to leaves, for which a narrow and standardized ontogenetic stage has been adopted) likely contributes to the greater uncertainty about the details of a root economics spectrum.

If the above findings are broadly applicable (and evidence suggests they are), they support the existence of a plant economics spectrum that applies to water as well as C and nutrients and that integrates across the leaf, stem and root systems. This spectrum suggests that plants at the ‘fast’ end of the productivity-persistence trade-off have high growth potential because they have high capacity to move water and to acquire and use nutrients and light to fix C, but build flimsy, disposable tissues (whether root, stem, or leaf) and are less tolerant of low resources (whether water, nutrients or light). In contrast, taxa with ‘slow’ traits are better protected from high C losses (low respiration, low leaf turnover rates) and drought stress (e.g. greater capacity to withstand low water potential without loss of turgor or hydraulic conductivity).

Traits, biogeography and individual performance

Traits and Climate Gradients

Given evidence of coupled economic traits for leaves, root, stems and whole plants, do these vary along large-scale environmental gradients (H8)? Interspecific relationships between plant C and nutrient economic traits on the one hand and temperature or precipitation gradients on the other have been identified, but explain surprisingly little of the total variances (Wright et al. 2004, 2005; ter Steege et al., 2006; Ordoñez et al. 2009).

For example, mean annual precipitation (MAP) and mean annual temperature (MAT) explained < 1% and 10% of global interspecific variation in SLA (Wright et al. 2004). The combination of four simple, widely available climate metrics (MAT, MAP, mean vapour pressure deficit and solar irradiance) explained only 5–20% of the overall interspecific variation in 5 LES traits (SLA, leaf life span, Amass, [N], [P]) at 175 sites (Reich, Wright & Lusk 2007). Similarly, only 1% to 13% of variance in three common measures of leaf mechanical strength could be explained by MAT or MAP (Onoda et al. 2011). Hydraulic margin of safety did not differ consistently across major global moisture gradients (Choat et al. 2012). Climate explains only a small fraction of trait variance in part because species with a variety of economic strategies are successful in communities and ecosystems all along these environmental gradients, reflecting high levels of local resource and microenvironmental niche diversity, as well as different alternative plant designs to make a living in similar microenvironments (Reich, Walters & Ellsworth 1997; Kobe 1999; Grime 2001; Wright et al. 2004; Ackerly & Cornwell 2007). Indeed, 38–67% of interspecific variation in dark respiration, leaf life span, Amass and [N] occurred between coexisting species within sites (Wright et al. 2004), 41–72% of variance in mechanical properties occurred within sites (Onoda et al. 2011), and 75–85% of SRL, root [N] and root N per length occurred within communities (Liu et al. 2010). Thus, the majority of the total variance of economic functional traits at organ scales is not explained by broad-scale climatic influences (see also Cornwell et al. 2008; Freschet et al. 2010). However, community-weighted mean traits may be better explained by climate. For example, climate explained from 38% to 55% of variance in a range of community-weighted mean traits including wood density, SRL, SLA and leaf [N] and P, across 19 sites spanning a 12 °C elevational range in MAT (Laughlin et al. 2011).

Moreover, although across all taxa, leaf life span is very poorly related to climate (see above), leaf life span of evergreen species decreases with MAT, whereas that of deciduous species increases (Wright et al. 2005; van Ommen Kloeke et al. 2012). The explanation for both is likely related to the length of the favourable portion of the season – for deciduous species that become leafless for a time each year, that period is as expected, closely related to the length of the unfavourable season (Kikuzawa et al. 2013). Why species that are evergreen should have longer leaf life span at lower MAT is not as obvious. One explanation, supported by optimization modelling, is that evergreen species need to increase the longevity of their foliage to optimize C gain (and offset the construction costs) as the favourable season becomes a shorter fraction of the year (Kikuzawa et al. 2013). The global pattern of increasing leaf life span in evergreen species with shorter, colder growing seasons is also observed within widely distributed boreal evergreen species (Reich et al. 2014) (Fig. 7). Intraspecific needle leaf life span increases by as much as 50–125% from the southern to northern regions of the boreal forest. The divergent impacts of growing season length on leaf life span of evergreen and deciduous species likely influence other LES traits, given their strong linkages with leaf life span.

Figure 7.

Intraspecific needle life span (mean among individuals at each site) in relation to mean annual temperature (MAT, °C) for five boreal conifers at between 19 and 78 sites (52 on average) across natural gradients in Eurasia (Pinus sylvestris) or North America (all others). Species include (from longest to shortest needle life span at low MAT) Picea mariana (open circles), Picea glauca (stars), Abies balsamea (closed circles), P. sylvestris (squares) and Pinus banksiana (triangles). Relations significant (< 0.01) for all species, mean R2 = 0.46. Illustrates H8. Data from Reich et al. (2014).

Global data illustrate correlations of leaf size and shape with climate that are considerably weaker for species means than for site means (Peppe et al. 2011; Royer et al. 2012), because species within a site vary considerably in leaf size and shape. Several assessments report that leaf size is not part of the LES (Ackerly & Reich 1999; Ackerly et al. 2002), and it is generally unclear what role leaf size and shape play in resource economics, despite decades of study (Nicotra et al. 2011).

In summary, economic traits are often weakly correlated (but see below for P) with climate at the species level (H8). Climate appears to exert some control on the average leaf (shape and economic) characteristics (hence stronger relations using site than species means), but many positions along leaf productivity-persistence trade-off axes are viable strategies in most communities and ecosystems.

The Unique Biogeography of Plant Phosphorus?

Perhaps the strongest biogeographical gradient for any economic plant trait exists for tissue [P] (and as a result, N : P ratio). From 19 to 58% of the global variation of N : P ratio in green and senesced foliage and in four fine-root size classes is related to MAT (H8) (McGroddy, Daufresne & Hedin 2004; Reich & Oleksyn 2004; Kerkhoff et al. 2005; Yuan, Chen & Reich 2011). This is likely due to a combination of biogeographical patterns of temperature and its seasonality, soil substrate type and age, erosion and occlusion (for P) and nutrient leaching (for N) (Reich & Oleksyn 2004; Lambers et al. 2008; Ordoñez et al. 2009; Vitousek et al. 2010; Turner & Condron 2013).

Soil substrate age gradients may be particularly notable at regional scales (Vitousek, Turner & Kitayama 1995; Lambers et al. 2008, 2013), especially along soil chronosequences which serve as powerful model systems (e.g. Vitousek, Turner & Kitayama 1995; Richardson et al. 2004; Laliberté et al. 2012). Along the Franz Josef chronosequence in New Zealand, community-level average traits (both abundance weighted and presence/absence) for root tissue density, branching architecture, [P], [N] and N : P were all correlated with soil age (Holdaway et al. 2011). The older soils, characterized by low P availability (Richardson et al. 2004), were inhabited by plant communities comprising species with higher root and leaf tissue density, low [N], low [P], high N : P and low Aarea (Turnbull et al. 2005; Whitehead et al. 2005; Holdaway et al. 2011) – all traits correlated with the ‘slow return’ economic strategy. Similar contrasts of coupled tissue chemistry and metabolism are seen in tropical forests varying in both N and P availability (Reich, Walters & Ellsworth 1994; Reich, Oleksyn & Wright 2009; Domingues et al. 2010). This view of the importance of soil P to plant economics is consistent with broad positive relations of leaf P to soil P (Ordoñez et al. 2009) and root P to soil P (Yuan, Chen & Reich 2011) as well as with studies showing low photosynthetic gain per unit leaf N in species inhabiting low-P sites (Reich, Walters & Ellsworth 1994; Reich, Oleksyn & Wright 2009; Domingues et al. 2010).

Clearly, whether the gradients are local, regional or continental, soil P gradients associated with soil substrate type or age are influential with respect to economic plant traits. In even more extremely P-impoverished (or strongly P-sorbed) soils, a broader syndrome of root traits exists (Lambers et al. 2008, 2013), with the increased abundance of species with carboxylate-releasing cluster roots or their equivalent that more efficiently scavenge P from soils than do mycorrhizal associates. Plants in such landscapes, especially in south-western Australia, are characterized by very low leaf phosphorus (P) concentrations, very high N : P ratios and very low SLA: fittingly, extremely ‘slow’ traits for an extremely low-resource environment (Lambers et al. 2013).

Traits and performance

The evidence above shows that coordinated multiple trait spectra exist for leaves, stems and roots considered separately (H1), that these organ-specific spectra are to some degree coordinated (H1), and that there is some relationship of these spectra to large gradients in climate and soils (H8). But do trait spectra reflect performance outcomes (Figs 8 and 9) in ecologically relevant settings (H2)? Moreover, what does trait coordination across organs and resources mean for a trait-based approach to performance, competition and community assembly/function? Because a single trait for a single organ likely represents a reasonable surrogate for a diversity of traits (of that organ, and of the other organs and the whole organism), a trait-based approach has a much greater chance of being of use. Clearly, as any single trait is only partially correlated with other traits, the greater the number of traits available and the greater relevance to the system and question at hand, the more powerful will be a trait-based analysis or model. Below I outline evidence about the role that traits play in individual performance and thus community assembly processes.

Figure 8.

Top panel (a) is the growth–survival trade-off for saplings expressed as the 95th percentile relative growth rate (RGR) (within species) vs. the survival rate over 5 years (% per 5 years) of the slowest growing 25% of individuals, for 103 tree species on Barro Colorado Island, Panama. Dotted line is regression line between the two; shaded area shows 95% confidence interval. Redrawn from Wright et al. (2010). Bottom panels show relationship between wood density and RGR (log-transformed, b), and mortality rate (log-transformed, c), for saplings in two tropical forest sites (Barro Colorado Island, Panama, white circles; and Pasoh, Malaysia, black circles). Correlations were significant (< 0.001), and the correlation coefficients ranged from between r2 = 0.13 and 0.19. From Chave et al. (2009).

Figure 9.

Relative abundance at high N availability in relation to performance at low N availability (left) and photosynthetic capacity (right). Left panel is the high N vs. low N availability performance trade-off for the most abundant species in temperate grasslands in a long-term experiment in Minnesota, USA (Tilman 1987). Data shown, averaged from 1991 to 2010, relative abundance per species (% of total above-ground biomass) for plants at high N (in the second highest N supply rate; 17 g N m−2 year−1) vs. an index of low N performance. The index is the shift in relative abundance measured comparing lowest N availability to the highest (relative abundance in ambient soil divided by relative abundance at highest N supply rate (27 g N m−2 year−1). Right panel shows relative abundance at second highest N supply in relation to the maximum light-saturated photosynthetic rate of the species growing in ambient soil. Data for relative abundances from Isbell et al. (2013) and for photosynthetic rates from Tjoelker et al. (2005).


The most abundant data linking traits to species performance and community assembly involve forests and resource variability, in particular light. There is evidence at the plant scale for a productivity-persistence trade-off; species with high maximum growth rate have higher mortality and are found in higher light (e.g. Kitajima 1994; Kobe et al. 1995; Walters & Reich 1996; Poorter & Bongers 2006; Wright et al. 2010) (Fig. 8). There is also considerable evidence that key traits of species in any given community vary significantly in relation to the typical light environment inhabited by each species and that such differences help explain performance differences that lead to those differing distributions (e.g. Grime 1965; Loach 1967; Kitajima 1994; Reich et al. 1995; Walters & Reich 1999; Lusk & Reich 2000; Poorter & Bongers 2006; Sterck, Poorter & Schieving 2006; Baltzer & Thomas 2007a,b; Poorter et al. 2008; Kitajima & Poorter 2010; Poorter et al. 2010; Wright et al. 2010; Lusk et al. 2011; Lusk & Jorgensen 2013) (Figs 10–12).

Figure 10.

Relationships between survival rate, height growth rate, leaf life span and an index of light habitats of juveniles (CEjuv) of 53 rain forest tree species in Boliva. Regression lines, coefficients of determination and significance levels are given. ***< 0.0001. Note the log–log scale. From Poorter & Bongers (2006).

Figure 11.

Relationships between 24-h period whole-plant light compensation point and sapling traits minimum instantaneous leaf-level light compensation point (LLCP), mass-based dark respiration rate, and leaf life span; as well as the relations between LLCP and dark respiration, and dark respiration and leaf life span. Each data point corresponds to the average values for a single species. Dotted line and associated solid lines represent best fits and 95% confidence intervals for the relationships. Redrawn from data in Baltzer and Thomas (2007a).

Figure 12.

Left panel; leaf dark respiration per unit leaf area for individuals growing in average light environments for 11 species of Hawaiian lobeliads (closed circles) and for 11 angiosperm and gymnosperm tree species in Minnesota (open circles), in relation to their average in situ daily photon flux density. Sampled leaves differ in light conditions experienced. Slopes and intercepts did not differ between the two groups so a single regression line (< 0.001, R2 = 0.83, with 95% confidence interval) is shown. Right panel, leaf dark respiration for the Minnesota species for plants grown at a standardized light (20% canopy openness), in relation to the minimum%canopy openness, as measured by the 95th percentile darkest individuals encountered. Respiration at 20% of canopy openness was determined from the parameters of log–log regressions of dark respiration rate vs. % canopy openness for each species individually. Thus, for this comparison, sampled leaves did not differ in light conditions experienced. A nonlinear regression fit (< 0.001, R2 = 0.76, with 95% confidence interval) is shown. Data for Hawaiian lobeliads from Givnish, Montgomery and Goldstein (2004) and for Minnesota tree species from Lusk and Reich (2000).

Species with ‘fast’ traits grow best and dominate in higher resource conditions, with ‘slow’ species surviving best (leading to eventual dominance) when resources are scarce and conservation of resources results in better growth and/or survival at low light (Figs 8–12). Across a gradient from higher to lower light requirements, species possess leaf and stem tissues that are longer-lived, tougher, denser and have lower [N], [P], Amax and dark respiration and lower hydraulic conductances (Reich et al. 1995, Walters & Reich 1999; Lusk & Reich 2000; Poorter & Bongers 2006; Meinzer et al. 2008a,b; Chave et al. 2009; Kitajima & Poorter 2010; Poorter et al. 2010) (Figs 10–12). Such species also have a lower whole-plant light compensation point (Fig. 11) (Baltzer & Thomas 2007a,b; Lusk & Jorgensen 2013). Such trait–environment relations also occur between closely related species. Givnish, Montgomery and Goldstein (2004) showed that the SLA, Amax, leaf respiration and light compensation point of 11 lobeliad species that occupy a wide range of light regimes in Hawaii all were positively coupled with the average photon flux density of their native habitats (Fig. 12), in accord with leaf economic theory.

Comparisons of traits and performance in natural field settings may be complicated though by differences in light microhabitats where species occur. Such differences may make it difficult to discern to what extent differences in traits and/or performance between species (e.g. Fig. 10) are due to species intrinsic differences or to influences of the differences in light microhabitat on traits and/or performance. Accounting for such light microhabitat differences can therefore be valuable in testing traits purported to be influential to performance. For example, by comparing 11 species across a wide range of light microhabitats in native forest, Lusk and Reich (2000) found that species differences in light habitat affinities were related to differences in respiration rate at a standardized light microhabitat. Species absent from more shaded microsites have higher respiration rates in a standardized light environment (Fig. 12). These results are consistent with many studies of plants in controlled experiments that found shade-tolerant species to have lower respiration rates and thus lower carbon losses than intolerants (H8, H2) (Loach 1967; Reich et al. 1998b).

Moreover, dark respiration is not only a component of shade tolerance, it may be among the most important traits conferring tolerance. A long-running debate asks whether shade tolerance is primarily a function of traits maximizing net C gain and growth in low light, or of traits minimizing C losses. Several recent papers suggest that conservation of energy at the leaf and plant scale (i.e. low respiratory losses, slow tissue turnover; slow traits par excellence) is more important to success in deep shade than maximizing efficiency of C gain at low light (Baltzer & Thomas 2007a,b; Lusk et al. 2011).


As expected, studies of strongly nutrient-limited systems provide the best examples of trait-based nutrient economic strategies, with slow traits associated with infertile conditions (Körner 2003; Craine 2009; Holdaway et al. 2011). Although data contrasting growth with mortality are scarce across nutrient gradients, species’ differential success across a long-term N availability gradient supports the notion of ‘fast vs. slow’ trait-based trade-offs that enable success at high resource supply or low, but not both (Fig. 9). Species with low tissue nutrient concentrations and high tissue longevity, which excel at tolerating low resources and/or suppressing resource supply to neighbours, are successful under low nutrient supply (Tilman 1987; Tilman & Wedin 1991; Aerts & Chapin 2000; Craine et al. 2002; Craine 2009; Holdaway et al. 2011; Mason et al. 2012).

The role of traits in relation to nutrient economy has been examined jointly with light in forested systems (e.g. Russo et al. 2005; Baltzer & Thomas 2007a,b; Holste, Kobe & Vriesendorp 2011). The pronounced growth-mortality trade-off (e.g. Fig. 8) that is a feature of forest taxa across light availability gradients (Kobe et al. 1995; Wright et al. 2010) is altered by soil nutrient supply, with mortality at a given growth rate being higher in less fertile soils, especially for the species with the ‘fastest’ resource strategies (Russo et al. 2005). The simplest explanation for this is that ‘fast’ traits are more costly in the face of any kind of resource shortfall. Challenges of measuring root traits and plant performance in relation to resource gradients in situ in the field, plus the potential multiple elemental limitations and complex role of mycorrhizal associations, continue to result in a still rather underdeveloped collective understanding of relations of below-ground traits to species strategies and nutrient economics, especially compared with what is known for leaf traits in relation to light and water.


There is ample evidence that the slow traits strategy is associated with drought tolerance (H1–2). At local scales, species rankings in daily minimum leaf water potential (Ψmin) mirror rankings in bulk leaf osmotic potential and the water potential at the turgor loss point (ΨTLP) (Meinzer et al. 2008a,b). Species differences in both Ψmin (Santiago et al. 2004; Meinzer et al. 2008a,b) and the water potential at loss of 50% of hydraulic conductance (Ψ50) (Maherali, Pockman & Jackson 2004; Blackman, Brodribb & Jordan 2010) also tend to track in parallel with each other and with water availability (H8). Ψmin and Ψ50 both vary inversely with leaf- and sapwood-specific kstem, vessel diameter and/or wood density (e.g. Santiago et al. 2004; Markesteijn et al. 2011) (Figs 3–5), but variably (e.g. Meinzer et al. 2010) and not always strongly (Blackman, Brodribb & Jordan 2010). Broad-scale species and community differences in ΨTLP and Ψ50 correlated with differences in broad-scale water availability (H8) (Figs 13 and 14) (Bartlett, Scoffoni & Sack 2012; Choat et al. 2012), although there is considerable variation between species within any given climate zone. Species with lower ΨTLP also have low Ψ50 (Cavender-Bares, Kitajima & Bazzaz 2004; Choat, Sack & Holbrook 2007; Meinzer et al. 2008a,b; Blackman, Brodribb & Jordan 2010). Campanello, Gatti and Goldstein (2008) also showed that species with greater leaf- and sapwood-specific kstem had higher photosynthesis and faster growth potential. The negative relationships seen often between wood density and both growth and mortality rate (Chave et al. 2009; Wright et al. 2010; Poorter et al. 2010) (but not always, Russo et al. 2010) are consistent with water economics as well, given the tendency of high wood density species to have low hydraulic conductance, low Ψ50 and low ΨTLP.

Figure 13.

Global data for pressure-volume parameters (osmotic potential at full turgor, left and at the turgor loss point, right), with mean ± standard error across biome categories, with inset plots of biome category means against the Priestly-Taylor coefficient of annual moisture availability (α). Biome categories: semi-desert, Mediterranean-type vegetation/dry temperate woodland, tropical dry and wet forest, temperate forest angiosperm and conifer, coastal vegetation, mangrove and crop herb. Data within biomes were separated into herb (H) vs. woody (W), or evergreen (E) vs. deciduous (D) when significantly different. π0 and πtlp showed separation of moist and dry biomes (dark and light bars respectively) and correlated with α across biomes (both r2 = 0.81, = 0.03–0.006). From Bartlett, Scoffoni and Sack (2012).

Figure 14.

Embolism resistance (ψ50) in relation to precipitation of the driest quarter for 384 angiosperm (open circles and dashed line) and 96 gymnosperm species (closed circle and solid line) from multiple sites across the globe. ψ50 is the water potential at loss of 50% of hydraulic conductance. The absolute value of the natural logarithm of ψ50 was significantly linearly related (< 0.0001) to precipitation of the driest quarter for both groups, with decreasing resistance to embolism corresponding to increasing rainfall (R2 = 0.15 in both groups). From data in Choat et al. (2012).

Collectively, these findings indicate that species that move and store water well have capacity to achieve higher C flux and growth rates, advantageous when conditions are good, but face greater mortality risk and are more vulnerable, in terms of their Ψ50 and ΨTLP. In contrast, species that cope well with drought tend to grow slowly, move and use less water and have dense tissues, all slow traits.

The difference between Ψmin and Ψ50 is a measure of the ‘safety margin’ for a plant in a given environment (Meinzer et al. 2009) and is thus another indicator of a plant's hydraulic strategy. A comprehensive synthesis (Choat et al. 2012) reported that, despite large differences in drought occurrence between forested biomes, angiosperms on average operate with narrow safety margins that, surprisingly, did not differ across enormous water availability gradients. In other words, the safety margin is ‘standardized’ to the water potentials typically experienced in any given region. Thus, angiosperm trees on average risk xylem failure during (locally) anomalously low rainfall in a manner that is largely independent of rainfall region and biome, suggesting a global convergence in the vulnerability of trees to drought (Choat et al. 2012). However, although species with the greatest embolism resistance were more common in drier climates, in every climate regime (and especially the dry ones), species exist with very wide ranges of both Ψ50 (Fig. 14) and safety margin (which I take as evidence that a range of strategies are successful). Moreover, although the Choat et al.'s (2012) synthesis found that the margin of safety varied little with precipitation gradients between angiosperms, within two well-studied systems (one in Tasmania, the other in Bolivia), species with the slow strategy did have greater margins of safety (Blackman, Brodribb & Jordan 2010; Markesteijn et al. 2011) as did gymnosperms in increasingly arid biomes (Choat et al. 2012). Whether safety margin is a trait that is part of a local or regional ‘fast–slow’ trade-off thus remains an open question.

Hydraulic traits also help explain species distributions (H2–3). Engelbrecht et al. (2007) showed that for 48 woody species at 122 sites spanning a rainfall gradient in Panama, niche differentiation with respect to soil water availability determined local- and regional-scale distributions of trees. Complementary studies suggest that traits explain these patterns. Kursar et al. (2009) reported that species differences in tolerance to low leaf water status were related to both their drought performance in the field and with their distribution across a gradient of water availability gradient (Fig. 15). Species that had lower stem hydraulic conductance wilted and died at lower (more negative) water potentials, had higher relative survival in droughted conditions and inhabited drier habitats. These results point to a causal link between hydraulic traits and performance in an ecological context. Studies elsewhere provide similar links of traits and performance. For example, for Tasmanian rain forest species, Ψ50 corresponded closely with an index of habitat distribution (the percentile of mean annual rainfall across each of the Tasmanian species’ geographical distribution) (Blackman, Brodribb & Jordan 2012) (Fig. 15).

Figure 15.

Left panel. The relationship between species tolerance of low leaf water status and their drought performance (Dp) in the forest understorey in Panama. Dp was defined as survival in non-irrigated conditions as a percentage of survival in irrigated conditions. Tolerance of low leaf water potential was assessed as the leaf water potential of severely wilted plants (SWψ). From Kursar et al. (2009). Right panel. The relationship between leaf vulnerability to cavitation (P50leaf) and the 5th percentile of mean annual rainfall across each of 18 Tasmanian species’ geographical distribution. Solid symbols represent montane rain forest species, while open symbols represent dry sclerophyll species. A hyperbolic curve was fitted through all the Tasmanian species data based on a theoretical intercept at −22 MPa. Closely related species pairs are denoted by enlarged symbols and connected by solid regression lines. A significant phylogenetically independent relationship was recorded between leaf vulnerability and climate (t-test; < 0.01). From Blackman, Brodribb and Jordan (2012).

The importance of hydraulic traits is not limited strictly to marked precipitation gradients. Savage and Cavender-Bares (2012) found that for a group of co-occurring willows, a set of traits that included ΨTLP and wood density varied in parallel with species abundances along a local (topographic) moisture gradient. Moreover, the weighted community means paralleled the species habitat affinities, suggesting that species with traits well matched to specific locations along the moisture gradient dominated those locations, supporting the idea that narrow niche breadth allowed coexistence and niche partitioning, consistent with a trait-based niche theory, but inconsistent with a neutral explanation.

There is also evidence of a trade-off between both growth rate and maximum leaf gas exchange with embolism risk associated with freezing tolerance for co-occurring Mediterranean oaks in France (Cavender-Bares et al. 2005), co-occurring evergreen angiosperms in Australia (Choat et al. 2011) and closely related North American willows and poplars (Savage & Cavender-Bares 2013). Cavender-Bares et al. 2005; Poorter et al. (2010), and Choat et al. (2011) found that shade tolerant, ‘dry habitat’ specialists and cold-tolerant species were all characterized by high wood and vessel density and small vessels, traits associated with slow growth and high water stress tolerance. This suggests that traits associated with high- vs. low-resource strategies may be similar for light, nutrients and water.

Traits and processes at community scales

Trait-based approaches offer compelling, if incomplete, frameworks to examine biotic interactions and community-scale traits (H3), and species-level differences in ecological traits play a key role in much of coexistence theory (Ackerly & Cornwell 2007; Ackerly & Cornwell 2009; Suding & Goldstein 2008). Trait-based approaches and models therefore offer promise regarding community assembly processes (e.g. Shipley, Vile & Garnier 2006; Suding et al. 2008; Dybzinski et al. 2011; Falster et al. 2011; Laughlin et al. 2012). For example, Adler et al. (2014) demonstrated that traits could explain variation in life history (including fitness) measures for more than 200 species from several terrestrial ecosystems.

A variety of models, experiments and observations have shown how trait differences between species indicative of capacity to pre-empt resources and reduce availability to competitors can explain the outcomes of competition between two or many species at a time (Tilman 1987; Tilman & Wedin 1991; Dybzinski & Tilman 2007; Dybzinski et al. 2011; Falster et al. 2011). For example, species with fast traits (e.g. Fig. 9, right panel) dominate at high N availability and usurp resources leading to reduced species diversity via competitive exclusion (Clark & Tilman 2008), but those with long root life span (Reich et al. 2001) and capacity to draw down soil N concentrations (Tilman & Wedin 1991) dominate at low N availability. Similar gradients in species abundances and traits (i.e. the ‘fast–slow’ contrast) in relation to soil P gradients support the links between traits, resources and biodiversity. For example, at low P supply in western Australia, plants with ‘slow’ traits dominate and fail to competitively exclude one another, leading to high diversity (Lambers et al. 2010, 2013), similar to low N situations in temperate grasslands (Clark & Tilman 2008). Several trait-based competition models have successfully modelled outcomes in multiple resource-limited conditions, a considerable achievement given our still nascent ability to understand multiple limitations and trait strategies (Sterck, Poorter & Schieving 2006; Farrior et al. 2013).

Trait-based biotic interactions can also be positive (facilitative); in fact the same traits can have positive and negative impacts on neighbours depending on conditions during a single season (Armas & Pugnaire 2005). Traits that lead to negative interactions through resource competition can also have positive (facilitative) impacts via amelioration of the microclimate under harsh abiotic conditions (Callaway & Walker 1997; Brooker et al. 2008; Valladares et al. 2008). Although likely more common in extreme environments, facilitation may be common, if overlooked, in more temperate and mesic environments (e.g. Montgomery, Palik & Reich 2010). Increased recognition that biotic interactions are a net effect of the positive and negative interactions has inspired the start of incorporating facilitation into trait-based community assembly theory (Schöb, Butterfield & Pugnaire 2012).

Phylogenetically oriented studies of co-occurring organisms have provided new tools and insight into the integrated exploration of community assembly, niche evolution and patterns of (taxonomic, trait and phylogenetic) diversity (Cavender-Bares, Ackerly & Kozak 2012). Despite abundant evidence of adaptive radiation of coordinated plant economic traits within lineages (e.g. Ackerly & Reich 1999; Cavender-Bares, Kitajima & Bazzaz 2004; Givnish, Montgomery & Goldstein 2004; Choat, Sack & Holbrook 2007; Savage & Cavender-Bares 2012), closely related taxa are often more alike in terms of their traits than would occur randomly, because of shared history (H9) (Cavender-Bares et al. 2009). Thus, convergent evolution of trait spectra across lineages (e.g. Ackerly & Reich 1999) and niche conservatism within lineages are not incompatible. This suggests that phylogenetic information cannot substitute for trait data, but that together both strands of data likely can explain more about ecological and evolutionary processes than either alone. For instance, both specialization within lineages and differences between lineages can contribute to extant trait patterns among taxa (Comas & Eissenstat 2009).

The tendency of related species to share form and function influences their broad habitat affinities as well as their local niche space occupancy and provides tools to help examine the filtering of functional traits into local-scale species assemblages and identify the ecological mechanisms governing community assembly processes (Cavender-Bares et al. 2009). Clade-based studies of the evolutionary and trait history of lineages can also help evaluate the role of traits in evolutionary diversification. For example, Savage and Cavender-Bares (2012) identified patterns of trait and phylogenetic structure of willow communities across a local hydrological gradient, providing evidence that the evolution of trait differences helped related species differentiate between niches even at a fine spatial scale and likely contributed to that diversification.

Kraft, Valencia and Ackerly (2008) and Kraft and Ackerly (2010) combined phylogenetic and trait data to examine niche-based vs. neutral theories of community assembly and coexistence in a species-rich tropical forest (H2–3, H9). They found that co-occurring trees are often less ecologically similar than niche-free (neutral) theory would predict, indicating that strategy differentiation between species contributes to the maintenance of high tropical forest diversity. They found evidence that some combination of trait-based strategy differentiation and/or enemy-mediated density dependence regulates species occurrence patterns at small scales (5–20 m), and species were not distributed randomly with respect to traits. However, the effect size of traits was modest; thus, the trait-based processes at play were weak, the statistical power of the tests was low, and/or the traits available were not the best proxies for detailed functional traits or trait combinations that influence fitness and thus population dynamics. The functional traits available to Kraft and Ackerly (2010) were wood density, SLA and leaf N; missing were any leaf metabolic (Kleaf, A, dark respiration) traits, or any root trait data at all. This is a common issue with tests of the importance of traits. If three or four traits explain a modest fraction of the variance in distribution (e.g. Kraft & Ackerly 2010) or in growth and survival (e.g. Wright et al. 2010), but these are not the traits one would choose if trait data was unconstrained, the power of the available data is not necessarily a good test of the power of a well-developed trait-based approach.

Evidence at community scale for simultaneous clustering and overdispersion in functional traits (Cavender-Bares, Kitajima & Bazzaz 2004; Swenson & Enquist 2009; Cavender-Bares & Reich 2012) indicates both the role of multiple processes and the signature of spatial scale on those processes. Swenson and colleagues have pioneered the consolidation of trait and inventory databases to examine how continental or global pools of functional trait diversity are filtered into regional-scale assemblages (e.g. Swenson & Enquist 2009; Swenson & Weiser 2010; Swenson et al. 2012a,b). Scaling traits up to the continental-scale enabled Swenson and Weiser (2010) to determine that the average trait values of temperate forest communities (in the eastern U.S.) were generally correlated with important climate metrics. In aggregate, these studies suggest that making full use of the promise of joint phylogeny–trait approaches to studies of community function and assembly is simultaneously a major challenge and opportunity.

Traits and processes at ecosystem scale and beyond

What are the consequences of leaf, stem and root spectra when aggregated to the system scale? I focus on two aspects here: first, the direct implications of plant traits for resource acquisition at the ecosystem scale (H4–5), and secondly, their indirect implications promulgated through impacts on below-ground community and biogeochemical processes (H6) (similar to ‘effects traits’, Lavorel & Garnier 2002).

Impact of Economic Trait Spectra on Ecosystem Processes and Cycles

The aggregated traits of co-occurring plants (e.g. canopies or root systems) regulate the uptake of all major resources (light, C, nutrients, water). Here, I consider the impact of both the average traits of a community [i.e. mass ratio hypothesis or community-weighted mean traits (Grime 1998)] and of trait heterogeneity. A straightforward hypothesis is that the size and chemistry of a canopy influence the ability to intercept light and the capacity to use that light to drive photosynthesis, and in parallel, the magnitude of water fluxes needed to support canopy photosynthesis. This hypothesis is supported by data from studies across multiple grassland (Garnier et al. 2004) and forest stands (Reich 2012) showing productivity correlates with community-weighted mean traits. For instance, ecosystem-scale maximum instantaneous canopy photosynthetic rate and annual net primary production are both joint functions of average leaf area index (LAI) and canopy [N] (H4) (Fig. 16, Reich 2012). This system-scale response mirrors the leaf-scale response (H1), whereby for a given SLA, leaves of higher [N] have greater instantaneous Amax, and for a given [N], those with higher SLA (which intercept more light per gram foliage) also have greater Amax. Thus, economic traits that lead to greater light harvesting and great photosynthetic potential lead to greater C uptake at leaf and stand scales, from second to year. Moreover, the average canopy [N] is influenced by trait-based competitive interactions that influence the abundance-weighted [N] (H2–4) (Dybzinski et al. 2013). Thus, traits of species and biotic interactions between species together regulate the scaling up of resource economics from leaf to ecosystem.

Figure 16.

Relationships, at different scales, of productivity to leaf area and nitrogen concentration in temperate forests. Except for panel ‘c’, the measures are equivalent to abundance-weighted, community mean traits. Relationships shown: (a) above-ground NPP per year in relation to leaf area index (LAI) and canopy% nitrogen (closed circles for 128 stands in Minnesota and Wisconsin, open circles for 18 stands in New Hampshire). (b) ANPP per day in relation to LAI and canopy [N] (data for 128 Minnesota and Wisconsin forests). (c) Instantaneous leaf-scale net photosynthetic capacity in relation to specific leaf area and leaf [N] for 296 tree species world-wide. (d) Relationship of maximum instantaneous ecosystem photosynthetic rate to LAI and canopy [N] for 33 forests. From Reich (2012).

The evolution in angiosperms of the fast traits that enable high leaf hydraulic and diffusive conductance, and thus high rates of transpiration and photosynthesis, also profoundly altered regional C, nutrient and water cycles (Boyce et al. 2009; Feild et al. 2011). The recycling of transpired water is an important source of rainfall, especially in the tropics, and climate modelling suggests that the tropics would be hotter, drier and more seasonal in the absence of the angiosperms (as explained, and the overall area of tropical rain forest would decline substantially (Boyce et al. 2009, 2010). Thus, the rise to ecological dominance of high transpiration angiosperms in the tropics not only likely altered regional climate but created conditions that increased the spatial extent of their dominance, with consequential effects on C, nutrient cycling and fire regimes (Feild et al. 2011). Given the close coupling of carbon uptake and water loss (see earlier in the review), it is likely that trait means at community scale generally have parallel impacts on water cycling as on carbon cycling. However, the relationship of traits to water cycling has received much less attention than relationships of traits to carbon and nutrient cycling – this represents a major opportunity for future work.

A logical first connection between plant economic traits and below-ground processes is through litter decomposition, that is, a test for afterlife effects. Considerable evidence shows strong impacts of a variety of litter traits associated with the LES on decomposition (e.g. Cornelissen 1996; Cornelissen & Thompson 1997; Santiago 2007; Hobbie 2008). In a synthesis of data for 818 species in 66 litter decomposition experiments across six continents, Cornwell et al. (2008) showed that the magnitude of species-driven differences was larger than climate-driven variation. Moreover, litter decomposition was faster for species with higher green leaf [N] and SLA, or higher litter [N] and lower lignin. In a separate meta-study, the N-release patterns of decomposing litter strongly influence litter N mineralization and are jointly regulated by the initial chemical composition of the litter and the stoichiometric requirements of the decomposers (Manzoni et al. 2008). Freschet et al. 2013 (Fig. 17) found a striking correspondence between decomposition of leaves, fine stems and fine roots across > 100 species from 13 ecosystems, driven both by shared traits and by climate variation.

Figure 17.

Relationships between decomposition constant k of leaves, fine roots and fine stems across numerous species and studies (redrawn from Freschet et al. 2013). Each point represents one species. Ellipses show the 95% confidence intervals for the data. The slope of standardized major axis regressions across all data is shown in the dotted lines.

Several studies show that plant traits influence soil microbial communities at a range of scales, with likely consequences for biogeochemical cycling. de Vries et al. (2012) identified a combination of abiotic and biotic predictors of landscape-scale soil microbial community composition. In addition to the usual suspects – climatic factors and soil chemical and physical properties – community-weighted plant traits also explained variation in soil microbial community composition. Based on studies of leaf, fine-stem and fine-root litter decomposition rates in terrestrial, riparian and freshwater systems, Freschet, Aerts and Cornelissen (2012) found that plant traits regulate litter decomposition through both direct litter effects and indirect effects (mediated by regulation of heterotroph community composition). In a study of temperate grasslands, Grigulis et al. (2013) found coupled control by plant and microbial functional traits of biomass production and soil C and N cycling. Plant species with the ‘fast’ strategy (high SLA, high Nmass) were linked to fast cycling and low retention of C and N. Given that decomposition rates are controlled jointly by environment (temperature, moisture) and tissue quality (Cornwell et al. 2008; Manzoni et al. 2008; Freschet, Aerts & Cornelissen 2012), continuous quantitative traits provide an avenue towards mechanistic modelling of decomposition.

Several studies have linked mycorrhizal status with biogeochemical processes at local and biogeographical scales (Cornelissen et al. 2001). As ectomycorrhizal association (and especially for ericoids) per se has been considered a functional trait, linked to the slow strategy, this is relevant to the theme of this review. Assuming ectomycorrhizal species are better able to acquire and use organic nutrients (Read, Leake & Perez-Moreno 2004, Wurzburger & Hendrick 2009; Orwin et al. 2011; Phillips, Midgley & Brozstek 2013), their organic nutrient uptake should slow soil C cycling through changes in the form and quantity of C exudates and detritis (H4, H6). Organic nutrient uptake may result in high soil C in ectomycorrhizal and especially ericoid-dominated habitats (Read & Perez-Moreno 2003; Read, Leake & Perez-Moreno 2004; Wurzburger & Hendrick 2009). The role of mycorrhizas in acquiring organic N and influencing C and N cycling as a result may be more widespread however. Hodge and Fitter (2010) and Whiteside et al. 2012 similarly found AM fungi to be able to acquire organic N in both controlled settings and in a boreal forest, respectively.

Phillips, Midgley and Brozstek (2013) looked further at whether ECM vs. AM trees differ in other aspects of their biogeochemical nutrient economy. As found elsewhere (Cornelissen et al. 2001; Hobbie et al. 2006), decomposition rates were higher for foliage of AM than ECM species. Net N mineralization rates did not differ between AM- and ECM-dominated plots, but the ammonium concentration was greater in AM-dominated plots and likely promoted greater nitrification rates there. Phillips, Midgley and Brozstek (2013) hypothesize that trait-integrated biogeochemical syndromes exist in AM and ECM forests due to differences in their ‘nutrient economies’ (i.e. the primary forms of nutrients utilized by plants and microbes). They suggest that this is a useful framework to characterize the biogeochemical attributes of AM- and ECM-dominated temperate forests.

Studies in 32-year-old monoculture experiment with 14 temperate tree species in Poland also examined consequences of plant economic traits for soil biology and geochemistry (H4, H6). Leaf litter chemical traits were strongly correlated with green leaf traits (Reich et al. 2005), but idiosyncratically correlated with root traits (Reich et al. 2005; Withington et al. 2006; Hobbie et al. 2010). Differences in litter calcium concentrations between tree species result from intrinsic differences in species physiology – the faster-growing species had higher tissue [Ca] (Dauer et al. 2007) and caused profound changes in soil acidity and fertility (Fig. 18) (Reich et al. 2005). Calcium-rich species had higher soil pH, exchangeable calcium, percentage base saturation and forest floor turnover rate (Reich et al. 2005; Hobbie et al. 2006) (H4, H6). Species with Ca-rich tissues and high soil pH also had greater SOM decomposition and microbial biomass in the mineral horizon. However, species drove soil net N mineralization and nitrification rates largely via differences in tissue [N] (Hobbie et al. 2007), as has been seen in other ecosystems (Fig. 18) (e.g. Reich et al. 2001; Orwin et al. 2010; Laughlin 2011). In one such study, nitrification potential was more strongly linked to dominant leaf traits than to functional diversity (Laughlin 2011), consistent with the mass ratio hypothesis (Grime 1998).

Figure 18.

Soil net N mineralization rate in relation to mean litter N across 20 oak savanna and woodland stands of differing fire frequency in eastern Minnesota, USA (left) and exchangeable soil Ca (0–40 cm mineral soil horizon) in relation to litter Ca across replicated monocultures of 14 tree species in western Poland (right). Redrawn from data of Reich et al. (2001, 2005).

Similar results come from a 7-year, 9-species monoculture grassland experiment that found support for some parts of the ‘fast traits–soil process’ hypothesis (Orwin et al. 2010). Species with high relative growth rate (RGR) had leaf and litter with high [N] and low toughness, an elevated bacteria : fungi biomass ratio in soil, high rates of soil N mineralization and concentrations of extractable inorganic N, and to some extent higher available phosphorus pools. However, fast above-ground traits did not match fast soil C cycling (soil respiration nor decomposition). The authors concluded that it may be more complex and difficult to use plant traits to predict processes that influence soil C cycling than to predict processes that influence N and P cycling.

The links noted above between plant traits and biogeochemical processes (decomposition, net N mineralization) are crucial for understanding vegetation–soil feedbacks and for improving models of global C and nutrient cycles. Clearly, plant traits influence soil processes in a manner generally consistent with a ‘fast–slow’ framework and can feedback to influence performance of competing species (Tilman & Wedin 1991; Berendse 1994; Dybzinski & Tilman 2007; McCarthy-Neumann & Kobe 2010a,b). Beyond direct evidence of such feedbacks is circumstantial evidence from literally hundreds of studies showing that plants modify soils in ways likely to benefit themselves and their offspring (e.g. acidiphile species tend to reduce soil pH and nutrient availability).

Economic Trait-Based Approaches to Modelling Ecosystems and Beyond

That plant traits influence ecosystem-scale properties and processes (Lavorel & Grigulis 2012) is a cornerstone of many mechanistically oriented ecosystem models, dynamic global vegetation models and land surface models (terrestrial biosphere models hereafter) (H4–H5). Many traits, most economic in nature, are included in model algorithms and are used to estimate ecosystem properties (e.g. Brovkin et al. 2012) and/or to drive calculated process rates (e.g. leaf life span, root life span, leaf, stem or root [N] and/or respiration rates, several photosynthetic traits, hydraulics). In most such models, traits have been parameterized using plant functional types (5–10 or so depending on the model), each of which is assigned a set of traits, usually based on empirical data. The use of plant functional types has persisted because of the challenge of developing continuous mapped trait surfaces. Improving characterization of trait variability in models is an active area at present; both trait–trait and trait–environment approaches may be fruitful (Swenson et al. 2012a,b; van Bodegom et al. 2012). Additionally, consideration of realistic trait values and correlations can uncover problems with process models. Herein, I describe three examples of attempts to take a continuous trait-based approach to terrestrial biosphere models.

Bonan et al. (2012) attempted to reconcile the problem of a mis-match between the maximum carboxylation velocity (which strongly correlates with Amax) needed to accurately predict gross primary productivity (GPP) with the Community Land Model (v4) and the maximum carboxylation velocity derived from a synthesis of empirical observations (Kattge et al. 2009). A set of exploratory simulations identified previously unrecognized deficiencies in the Community Land Model (version 4) parameterization of the canopy as a sunlit and a shaded ‘big-leaf’. This demonstrates that working towards accurate prediction of outcomes (e.g. GPP) simultaneously with accurate characterization of trait-based vegetation physiology and chemistry can help identify ‘off-setting’ model errors that provide the ‘right’ answer in some contexts but with the wrong parameter values.

Another example comes from Reich et al. (2014) who incorporated biogeographical patterns of evergreen boreal needle longevity and [N] (H8) into a land surface model, Australian Community Atmosphere Biosphere Land Exchange (CABLE), to assess their impacts on C cycling processes (H4–6). Incorporating realistic parameterization of these variables improved predictions of canopy LAI and GPP as compared to observations from flux sites. However, this was a ‘one step back, two steps forward process’, because at first the model performed much more poorly with the new needle trait parameterization. Exploratory simulations identified problems with how needle life span influenced canopy LAI in CABLE. When more realistic biomass distribution algorithms were also incorporated, CABLE did a better job of predicting LAI and GPP than previously. A third trait-based approach was taken by Wang et al. (2012) who incorporated multi-leaf trait covariance ([N], LMA, leaf life span) into CABLE. Constraining trait correlations to those observed for the three parameters (H1) did not alter the mean but reduced the variance of modelled GPP (H5) by 28% and resulted in fewer extremely high or extremely low (and unlikely) GPP predictions. The results suggest that correlations between plant traits, and relationships with environmental drivers, offer promise as constraints on the estimates of model parameters or predictions by those models.


The works reviewed in this article describe the nature, causes and consequences of functional trait spectra at organ and whole-plant scales, within and among species, communities, lineages and biomes. It is fair to ask just how much do traits (and in particular those directly involved in resource economics) help us to better understand, quantify and model key ecological processes at tissue to organism to ecosystem to global scales. In essence, do such traits provide enough meaningful characterization, explanation and quantification of the nature of key ecological relationships to be of broad use? Although below I use the percentage of variance explained as a metric, and some might complain that is just a statistical correlation, it is hoped that the literature synthesized above makes it clear that there are functional, causal links at work.

As an illustration of whether the proverbial glass is half full, or half empty, I consider the role of traits in explaining the strong trade-off between growth and survival (Fig. 8) among tropical tree species (Wright et al. 2010). This narrative reproduces a set of conversations (while writing the paper) among co-authors who came at this work with different perspectives. In that study, four traits were available to be tested as potential predictors of the growth–survival trade-off. Two were significant (wood density and LMA) and explained ≈40% of the variance in growth and survival. Is this a useful amount of explanation? One co-author initially thought not, because, after all, 40% is far less than a full explanation. However, the four traits available did not include other traits likely equally or more influential to the processes in question: traits such as leaf hydraulic conductance or stem hydraulic conductivity, photosynthetic capacity, leaf dark respiration rate or leaf nutrient concentration, or anything about roots. Thus, other authors thought that if the discipline can move 40% down the path to a full understanding of a process or system using only a few traits, and not even those one would choose from a full menu that suggests considerable potential for a trait-based approach.

The weight of evidence reviewed in this document conveys a similar message. The key traits involved in carbon, nutrient and water economics vary in coordinated ways both within and among the leaf, stem and root systems of higher plants (H1). Traits vary with environment (H7–8) and evolutionary history (H9) and influence whole-plant performance (H2), community assembly and ecosystem/landscape function (H3–5), and soil feedbacks (H6, H8). Thus, I conclude that a trait-based ecology, and in particular, one based on the plant economic spectrum, has already half-filled ecology's glass, because where the traits of species fall on that spectrum tells us much about their ecology and that of the community and ecosystem they comprise.


I thank two anonymous referees and a large number of colleagues who shared papers, data, and ideas with me as I struggled to develop this manuscript. In particular, I would like to thank J. Baltzer, T. Brodribb, B. Choat, H. Cornelissen, G. Freschet, T. Givnish, T. Kursar, H. Lambers, L. Markesteijn, Meinzer, L.Sack, L. Santiago and L Williams. This work was supported by the U.S. National Science Foundation Long-Term Ecological Research program (DEB-1234162) and the Institute on the Environment, University of Minnesota.