Distinguishing the signatures of local environmental filtering and regional trait range limits in the study of trait–environment relationships

Abstract

Understanding the imprint of environmental filtering on community assembly along environmental gradients is a key objective of trait‐gradient analyses. Depending on local constraints, this filtering generally entails that species departing from an optimum trait value have lower abundances in the community. The community‐weighted mean (CWM) and variance (CWV) of trait values are then expected to depict the optimum and intensity of filtering, respectively. However, the trait distribution within the regional species pool and its limits can also affect local CWM and CWV values apart from the effect of environmental filtering. The regional trait range limits are more likely to be reached in communities at the extremes of environmental gradients. Analogous to the mid‐domain effect in biogeography, decreasing CWV values in extreme environments can then represent the influence of regional trait range limits rather than stronger filtering in the local environment. We name this effect the ‘trait‐gradient boundary effect’ (TGBE). First, we use a community assembly framework to build simulated communities along a gradient from a species pool and environmental filtering with either constant or varying intensity while accounting for immigration processes. We demonstrate the significant influence of TGBE, in parallel to environmental filtering, on CWM and CWV at the extremes of the environmental gradient. We provide a statistical tool based on Approximate Bayesian Computation to decipher the respective influence of local environmental filtering and regional trait range limits. Second, as a case study, we reanalyze the functional composition of alpine plant communities distributed along a gradient of snow cover duration. We show that leaf trait convergence found in communities at the extremes of the gradient reflect an influence of trait range limits rather than stronger environmental filtering. These findings challenge correlative trait–environment relationships and call for more explicitly identifying the mechanisms responsible of trait convergence/divergence along environmental gradients.