Environment and predation govern fish community assembly in temperate streams

Aim: The elucidation of patterns and drivers of community assembly remains a fundamental issue in ecology. Past studies have focused on a limited number of communities at local or regional scales, thus precluding a comprehensive examination of assembly rules. We addressed this challenge by examining stream fish community assembly within numerous independent watersheds spanning a broad environmental gradient. We aimed to answer the following questions: (1) are fish communities structured non-randomly, and (2) what is the relative importance of environmental filtering, predator–prey interactions and interspecific competition in driving species associations? Location: The conterminous USA. Methods: We used null models to analyse species associations in streams. Non-random communities were defined as those where the summed number of segregated and aggregated species pairs exceeded the number expected by chance. We used species traits to characterize species dissimilarity in environmental requirements (ENV), identify potential predator-prey interactions (PRED) and estimate likely degree of competition based on species similarity in body size, feeding strategies and phylogeny (COMP). To evaluate the effect of environmental filtering, predation and competition on species associations, we related ENV, PRED and COMP to the degree of species segregation. Results: The majority (75-85%) of watersheds had non-random fish communities. Species segregation increased with species dissimilarity in environmental requirements (ENV). An increase in competition strength (COMP) did not appear to increase segregation. Species pairs engaging in predator–prey interactions (PRED) were more segregated than non-predator–prey pairs. ENV was more predictive of the degree of species segregation than PRED. Main conclusions: We provide compelling evidence for widespread nonrandom structure in US stream fish communities. Community assembly is governed largely by environmental filtering, followed by predator–prey interactions, whereas the influence of interspecific competition appears minimal. Applying a traits-based approach to continent-wide datasets provides a powerful approach for examining the existence of assembly rules in nature.