LETTER A trait-based approach to community assembly: partitioning of species trait values into within- and among-community components D. D. Ackerly1* and W. K. Cornwell2 1 Department of Integrative Biology, University of California, Berkeley, CA 94720, USA 2 Department of Biological Sciences, Stanford University, Stanford, CA 94305, USA *Correspondence: E-mail: dackerly@berkeley.edu Abstract Plant functional traits vary both along environmental gradients and among species occupying similar conditions, creating a challenge for the synthesis of functional and community ecology. We present a trait-based approach that provides an additive decomposition of species�� trait values into alpha and beta components: beta values refer to a species�� position along a gradient defined by community-level mean trait values alpha values are the difference between a species�� trait values and the mean of co-occurring taxa. In woody plant communities of coastal California, beta trait values for specific leaf area, leaf size, wood density and maximum height all covary strongly, reflecting species distributions across a gradient of soil moisture availability. Alpha values, on the other hand, are generally not significantly correlated, suggesting several independent axes of differentiation within communities. This trait-based framework provides a novel approach to integrate functional ecology and gradient analysis with community ecology and coexistence theory. Keywords Community assembly, functional diversity, height, leaf size, niche breadth, plant strategies, plasticity, seed size, specific leaf area, wood density. Ecology Letters (2007) 10: 135���145 IN T R OD U C TI ON The objective of this paper is to develop a quantitative framework integrating two traditions in plant ecology. The first is the study of plant form and function along climatic and edaphic gradients, which has its roots in plant geography (Schimper 1903) and played a central role in the development of plant functional and community ecology (Mooney & Dunn 1970). The second is the development of niche theory (Hutchinson 1957 MacArthur & Levins 1967), and the study of demographic and functional differences among co-occurring species in relation to mechanisms of coexistence (Pacala & Tilman 1994 Tilman 1994 Chesson 2000). Trait-based approaches to community ecology, linking ecological strategies, com- munity assembly theory and functional diversity, have the potential to unify these contrasting viewpoints (Grime 2006 McGill et al. 2006 Westoby & Wright 2006). Global syntheses of plant trait diversity have highlighted two prevalent patterns. On the one hand, mean values of key plant traits exhibit significant shifts across climatic gradients, at both global and local scales (Bailey & Sinnott 1916 Baker 1972 Dolph & Dilcher 1980 Wright et al. 2005 Moles et al. 2007). On the other hand, studies at all scales reveal that high levels of trait disparity are observed within communities. For example, 35% of the global variation in specific leaf area (leaf area/mass, SLA) is found within sites, compared with the variation among sites that may reflect large-scale climatic factors (Wright et al. 2004). This parti- tioning of plant trait diversity into within- and among-site components may be termed alpha and beta trait diversity, by analogy with Whittaker���s (1975) distinction between alpha and beta patterns in species diversity. In this paper, we present a new method for the analysis of trait variation at the community level along environ- mental gradients, based on a modification of Finlay & Wilkinson���s (1963) approach to the analysis of adaptation in plant breeding programmes (also see Garbutt & Zangerl 1983). In this approach, the mean trait value of the species co-occurring in a local community is used to array Ecology Letters, (2007) 10: 135���145 doi: 10.1111/j.1461-0248.2006.01006.x �� 2007 Blackwell Publishing Ltd/CNRS

communities along a one-dimensional gradient. This gradi- ent reflects the integrated effects of multiple environmental factors, as well as dispersal limitation or other historical factors that may shape the species composition of the communities in question. It also incorporates the contribu- tions of ecotypic variation and phenotypic plasticity to the trait values of each species, and hence to the mean trait value of the community. A key to the subsequent analysis is that the values along the trait-gradient (plot means) are in the same units as the trait values for the individual species within the communities. Analysis of species distributions and trait variation along this gradient provides an integrated approach to quantify intraspecific variation, niche breadth of individual species, and a novel partitioning of species mean trait values into alpha and beta components. The beta component refers to a species�� mean location along the gradient (i.e. a measure of niche position), while the alpha component is the difference between a species�� trait value and its beta value, i.e. a measure of how the traits of each species differ from those of co-occurring taxa. The partitioning of trait values into alpha and beta components builds on recent discus- sions of the alpha and beta niche as species characteristics associated with diversity within (alpha) and across (beta) habitats or communities (Ackerly et al. 2006 Silvertown et al. 2006a,b). The alpha niche refers to those attributes that differentiate a species from co-occurring taxa, and therefore may contribute to non-neutral maintenance of species diversity. The beta niche, on the other hand, refers to species distributions across habitat or geographic gradients beta-niche characteristics will tend to be shared among co-occurring species. There are two important points about these definitions to note at the outset. First, like all ideas in community ecology, they are intrinsically scale-dependent. The partitioning of trait values into alpha vs. beta components will depend on the scale used to define a community. In the effort to better link functional ecology to both gradient analysis and coexistence theories, one advantage of this scale dependence is that the traits can be partitioned at the same scales that are considered important for community processes. Second, the concept of beta diversity (Whittaker 1975) has two compo- nents: (i) turnover of species associated with different environmental conditions and (ii) turnover of ecologically similar species in different geographic areas. Cody (1993), and those who have followed his usage, restricted the term beta diversity to the first component, using gamma for the second. Others, following Whittaker, have used beta for both components, at times differentiating the environmental vs. spatial components (Couteron & Pelissier 2004), and used gamma for the total amount of sampled diversity (see Crist et al. 2003). In any case, ecologically similar species in similar environments are expected to have similar traits, so there would be no phenotypic signature associated with contrast- ing distributions. Thus, the concepts of beta niche (Ackerly et al. 2006 Silvertown et al. 2006a,b) and species-level beta trait values, discussed here, specifically refer to the first component of beta diversity: the distribution of species in communities that occupy distinct environmental conditions. O B J EC T I V E S We present the method of trait-gradient analysis to partition species traits into alpha (within-community) and beta (among-community) components, and to quantify species niche breadth and the magnitude of intraspecific variation along environmental gradients. Trait-gradient analysis uses a simple data set with four pieces of information: a list of plots or communities, a list of species sampled in each plot, the value for a particular trait measured for each species in each plot, and a measure of relative abundance of each species (or simply presence/absence, if abundance data are not available). We introduce trait-gradient analysis with an empirical data set on five functional traits in woody plant communities in coastal California: SLA, leaf size, wood density, maximum height and seed size (Cornwell 2006). In this landscape, gradients in plot-level means for the first four traits are known to be strongly correlated with underlying gradients in topographic position and soil moisture (Ackerly et al. 2002 Cornwell 2006). Here, we develop a quantitative framework to address the following questions: (1) How is variation in species trait values partitioned into alpha vs. beta components? (2) What is the magnitude of intraspecific variation, relative to the overall shift in trait values across communities? (3) Are the beta and alpha components of different traits correlated with each other? We use a simple null model of community structure to determine if observed patterns reflect non-random aspects of community assembly, vs. intrinsic results of our methodology. In the Discussion, we address the significance of these patterns in relation to ecological strategies and potential mechanisms of coexistence across these commu- nities. The study system All woody plants were sampled in 44 20 �� 20 m plots at the Jasper Ridge Biological Preserve, Stanford University, San Mateo Co., CA, USA (Cornwell et al. 2006 Cornwell 2006). The plots were randomly located across the range of communities present, from riparian deciduous woodland to sclerophyll chaparral shrubland. A total of 54 native species were sampled, with alpha diversity ranging from three to 18 species per plot (Table A2 nomenclature follows Hickman 1993). There were a total of 471 species-plot observations that make up our data set for the analyses presented here. 136 D.D. Ackerly and W.K. Cornwell Letter �� 2007 Blackwell Publishing Ltd/CNRS