Beyond community-weighted means: quantifying trait distributions for detecting community assembly patterns

Abstract

The distributions of ecological traits are commonly used to infer the processess structuring ecological communities, as these processes (either deterministic or stochastic) select and filter species, leading to distint trait patterns across communities. Trait distributions are frequently characterised by their average, the community-weighted mean, but increasing evidence exists for non-gaussian trait distributions in empirical communities. In such situations, community-weighted means are insufficient to capture the patterns of community traits and to infer the implied ecological processes. Here we analyse the empirical distributions of 6 functional traits of butterflies from a natural community and a filtered community from a urban area across a period of six years. First, we show that to adequately describe trait distributions, statistical descriptors beyond community-weighted means are needed, as distributions were in all cases clearly non-gaussian. In particular, besides distribution averages and standard deviations, we compute the skewness, kurtosis, the range of trait values, and a multimodality index. Second, we compare this set of descriptors between our natural and filtered communities. We find clear differences between communities, detected in particular by combining the relationships between the skewness, kurtosis, and range of the distributions. These analyses allow us to infer that the filtered community is mostly shaped by deterministic filtering processes, through a mixture of directional and stabilizing assembly filters. These patterns are furthermore consistent across the six years of data, providing further evidence of deterministic processes shaping the assembly of our urban butterfly community. Overall, we provide evidence of the ubiquity of non-gaussian trait distributions across natural and filtered communities, propose key descriptors for understanding and comparing such distributions, and identify filtering process structuring these communities.