Leaf trait area- and mass-proportionality between and within tropical tree species, across and within forest canopies

Background/Question/Methods

Quantifying leaf trait relationships across tree species and within forest canopies is necessary for accurately modeling forest carbon cycling.  It has been shown recently that across species in the global flora, leaf traits (photosynthetic capacity, dark respiration rate, leaf lifespan and concentrations of nitrogen and phosphorus) vary largely in proportion to leaf area instead of leaf mass.  This nonintuitively makes relationships between leaf traits normalized by leaf mass appear strong but largely precludes the utility of mass-based relationships for biological inference.  Within the forest canopy light gradient, however, leaf traits within tree species may be largely mass-proportional.  We quantify area- and mass-proportionality of leaf traits between and within species by fitting statistical models to a dataset of traits of leaves collected at different canopy positions from canopy cranes at one wet and one dry tropical forest in Panama.  The models also allow us to quantify trait relationships within and across canopy strata and sites independent of area- or mass-normalization.  We evaluate the generality of the Panama results for tropical forests by fitting our models to tropical tree data from a global leaf trait dataset and, for nitrogen, to canopy and understory data from the Yasuni forest plot in Ecuador.    

Results/Conclusions

Across species within a canopy stratum (canopy or understory) in the Panama data, traits are largely area-proportional.  Within species across canopy strata, however, traits are primarily mass-proportional, which is consistent with some temperate examples of intraspecific trait variation within individual tree or forest canopies.  This general pattern is responsible for strong mass-based trait correlations within each stratum and strong area-based correlations within sites.  Normalization-independent trait relationships within strata and sites are relatively weak but are qualitatively similar to those in other studies.  Few significant differences exist between strata or between sites in the slopes of normalization-independent bivariate relationships, and those that do involve leaf lifespan.  The dry site is home to many deciduous species and consequently has considerably lower leaf lifespans than the wet site, which is home to many evergreen species.  This leaf lifespan disparity between sites creates some polarization of the sites along bivariate normalization-independent axes, with higher trait values at the dry site.  This site disparity is reduced by including leaf lifespan in our models.  Results from the global trait dataset and the Yasuni plot are remarkably consistent with the Panama results, suggesting that our results have broad applicability to analyses of global tropical forest trait data.