Origin of the dust bunny distribution in ecological community data

Abstract

The distribution of sample units in multivariate species space typically departs strongly from the multivariate normal distribution. Instead of forming a hyperellipse in species space, the sample points tend to lie along high-dimensional edges of the space. This dust bunny distribution is seen in most ecological community datasets. The practical consequences of the distribution to the analysis of community data are well known and severe, but no one has demonstrated how population processes generate these problems. We evaluate potential causes of dust bunny distributions by simulating a large number of non-equilibrial communities under varying conditions, verifying that they resemble real data, then analyzing the relationship between the intensity of the dust bunny distribution in these datasets and the population and environmental parameters that gave rise to them. All community datasets, both simulated and real, departed strongly from multivariate normal and lognormal distributions. Four parameters influenced intensity of dust bunnies: time since community-replacing disturbance, number of environmental factors, dispersal limitation, and niche width. Samples measured soon after community-replacing disturbance had strong dust bunny distributions. Near-equilibrial communities sampled from a narrow range in environments lead to only weak dust bunnies. Community samples taken across multiple simultaneous strong environmental gradients are likely to show strong dust bunnies, regardless of the successional state, niche width of the component species, and degree of dispersal limitation. Dust bunny intensity depends not only on population processes and disturbance, but also on the properties of the sample, such as sample unit area or volume.