A TEST FOR COMMUNITY CHANGE USING A NULL MODEL APPROACH

Abstract

Quantifying patterns of temporal or spatial change in community structure is critical for assessing the impact of disturbances on biological systems and the stability of ecosystems. Detecting change in communities can be problematic, however, because of the inherent variability of systems and limitations of commonly used methods, such as similarity indices and ordination that do not explicitly test a hypothesis. Here we present empirical data to show a strong relationship between species mean abundance and variability across three broad taxonomic groups (plants, zooplankton, and fish). These statistical relationships were then used to construct null models of expected community variability that were used to test against the observed temporal change of these communities. We evaluated the ability of this approach to detect significant temporal change above that associated with random variation with nine communities (three Midwestern stream fish, three north temperate zooplankton, and three tallgrass prairie plant), each having long-term data sets and different expected levels of change. Nonrandom change was detected in 21.3% of samples from the expected low-change communities, 52.6% in moderately changed communities, and 60.4% in high-change communities. Thus, this approach was effective in detecting change over time in those communities expected to change most. By using empirical relationships between species abundance and variability, this null model approach provides ecologists and resource managers an objective tool, which can be used along with existing community indices and statistical techniques to assess the type and magnitude of community change with limited data sets.