Revealing the functional traits that are linked to hidden environmental factors in community assembly

Abstract

AimTo identify functional traits that best predict community assembly without knowing the driving environmental factors. MethodsWe propose a new method that is based on the correlation r(XY) between two matrices of potential community composition: matrix X is fuzzy-weighted by trait similarities of species, and matrix Y is derived by Beals smoothing using the probabilities of species co-occurrences. Since matrix X is based on one or more traits, r(XY) measures how well the traits used for fuzzy-weighting reflect the observed co-occurrence patterns. We developed an optimization algorithm that identifies those traits that maximize this correlation, together with an appropriate permutational test for significance. Using metacommunity data generated by a stochastic, individual-based, spatially explicit model, we assessed the type I error and the power of our method across different simulation scenarios, varying environmental filtering parameters, number of traits and trait correlation structures. We then applied the method to real-world community and trait data of dry calcareous grassland communities across Germany to identify, out of 49 traits, the combination of traits that maximizes r(XY). ResultsThe method correctly identified the relevant traits involved in the community assembly mechanisms specified in simulations. It had high power and accurate type I error and was robust against confounding aspects related to interactions between environmental factors, strength of limiting factors, and correlation among traits. In the grassland dataset, the method identified five traits that best explained community assembly. These traits reflected the size and the leaf economics spectrum, which are related to succession and resource supply, factors that may not be always measured in real-world situations. ConclusionsOur method successfully identified the relevant traits mediating community assembly driven by environmental factors which may be hidden for not being measured or accessible at the spatial or temporal scale of the study.