Premium
Quantifying multimodal trait distributions improves trait‐based predictions of species abundances and functional diversity
Author(s) -
Laughlin Daniel C.,
Joshi Chaitanya,
Richardson Sarah J.,
Peltzer Duane A.,
Mason Norman W.H.,
Wardle David A.
Publication year - 2015
Publication title -
journal of vegetation science
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.1
H-Index - 115
eISSN - 1654-1103
pISSN - 1100-9233
DOI - 10.1111/jvs.12219
Subject(s) - chronosequence , trait , ecology , species richness , niche , specific leaf area , biology , species diversity , biodiversity , plant community , ecosystem , botany , computer science , photosynthesis , programming language
Question Niche differentiation results in functionally diverse communities that are often composed of dominant species with contrasting trait values. However, many predictive trait‐based models that emphasize environmental filtering have implicitly assumed that traits exhibit unimodal distributions among individuals within communities centred on an optimal trait value. Does accounting for more complex, multimodal trait distributions among individuals in a community improve predictions of species abundances and functional diversity along environmental gradients? Location Franz Josef soil chronosequence, central Westland, New Zealand. Methods Leaf nitrogen (N) and phosphorus (P) concentrations from 23 woody plant species were modelled as functions of soil total N and P from eight sites of declining soil P. We compared predictions to observations of species abundances and functional diversity along the soil chronosequence using two modelling approaches: (i) the standard application of the hierarchical Bayesian Traitspace model that assumes unimodally distributed traits at each point along the gradient, and (ii) a modified application of the model that accounts for multimodal trait distributions within each community. Results Soil P was the strongest predictor of traits and species abundances. The strength of the environmental filter of leaf traits changed along this gradient, as evidenced by highly constrained variances and low modality of the trait distribution at low soil P, and high variance and multimodality at high soil P. Both modelling approaches predicted species abundances that were significantly correlated with observations, but the multimodal approach significantly improved predictions of species abundances and functional diversity. Conclusions Our results indicate that predictive models that emphasize environmental filtering over niche differentiation by assuming unimodal trait distributions can be more parsimonious than more complex approaches, especially when predicting species abundances along strong environmental gradients. However, models need to account for trait multimodality if they are to accurately replicate spatial patterns in functional diversity. This is important since functional diversity may be a key predictor of ecosystem function and resilience to global change.