Deriving Plant Functional Types for Amazonian forests for use in vegetation dynamics models

Recent advances in our understanding of the linkages between plant physiological and morphological traits suggest a new means by which to define Plant Functional Types (Φ) for use in conceptual and mathematical models of vegetation dynamics. In this study we used data from the RAINFOR-network database, aiming to numerically derive Φ for tropical forest trees by jointly analysing an Amazon-wide dataset of (409) species abundance, species functional traits (10) and site edaphic and climatic conditions across 53 plots. We followed a stepwise procedure of numerical Φ definition with increasing complexity, starting from a simple PCA on species functional traits. We subsequently applied a three-table (RLQ) multivariate ordination method in two ways: with and without spatial autocorrelation between plots being taken into account. In all cases the environmental contribution to trait variation had been partialled out. Thus our results link species-specific “inherent” trait values with associated species abundances along environmental gradients. Our final classification of Amazonian tree species based on foliar dry leaf mass per area (MA), leaf concentrations of C, N, P, Ca, K, Mg, carbon isotopic discrimination (Δ), branch xylem density (ρX) and maximum tree height (Hmax) yielded four discrete Φ. These Φ were found to represent distinct life-history strategies and can be aligned with previous empirical definitions of tropical tree guilds. In particular, two ecological dimensions are identified: (1) a leaf deployment dimension which co-varies with soil fertility and (2) a stem deployment dimension which co-varies with soil texture. By analysing diameter growth rates of the same trees used to define the four Φ we found each Φ to have a different overall growth pattern. Furthermore, from a Basin-wide forest survey, differences in the relative abundance of the four Φ were related to stand level basal area growth and/or turnover rate variations. These new derived Φ should enhance our ability to better understand and model the dynamics of the Amazon forest, with the general procedure for plant functional trait definition described here potentially applicable to many other ecosystems.