Abstract
One of the fundamental challenges in ecology is to identify the signature of assembly mechanisms resultingfrom patterns of community composition. For this purpose, the trait-based approach has promotedthe analysis of functional trait distributions within communities. Until now, much attention has beenpaid to the design of appropriate null models and the definition of relevant functional metrics for inferringcommunity assembly mechanisms from trait distributions. However, less consideration were givento the set of methodological choices preceding the statistical analysis – i.e. from designing a samplingscheme to measuring traits – and how likely they influence the conclusions drawn, as this may subjectthe analysis to methodological biases. In this regard, a comprehensive perspective on how the overall setof methodological choices influence the inference of community assembly mechanisms is needed.We extensively reviewed recent studies that have addressed animal and plant community assemblyby applying a trait-based null model approach. We analyzed how the set of methodological choicesin these studies depended on the mechanisms of interest and how they could influence the conclusionsdrawn by the authors. We found that methodological choices only weakly depended on the hypothesizedassembly mechanisms studied by the authors, especially because the two main assembly mechanismshypothesized, i.e. environmental filtering and limiting similarity, were often tested based on a commonexperimental design. In contrast, the detection of assembly mechanisms was strongly dependent on thesampling scale, the type of data, the origin of trait values and the delineation of the reference speciespool, while less affected by the null model and the functional metrics chosen. These results underlineplausible methodological biases in favor of the detection of certain mechanisms to the detriment of others.In addition, there was a significant relationship between the predominant mechanisms concluded by theauthors and the type of organism used as biological model, suggesting either that the methodologicalchoices depend on a common strategy used by the different authors when studying similar biologicalmodels, or that the methodological choices depended on certain particular properties of the organisms.From this extensive review, we highlight major conceptual and methodological issues that need to beaddressed in trait-based null model approaches. We synthesize the methodological choices relevant tostudy several assembly mechanisms while minimizing methodological biases. We then derive practicalguidelines and emphasize the importance of spatial structure in sampling strategy and null model design,because of the scale-dependent signatures of ecological processes.