Experimentally disentangling the influence of dispersal and habitat filtering on benthic invertebrate community structure

Abstract

Abstract Ecological communities are structured by a combination of local processes like habitat filtering and species interactions, and regional forces driven by the dispersal of organisms between localities on a landscape. Previous studies suggest that the position of local communities within a dispersal network can greatly influence the relative influence of these two sets of processes on community assembly. However, the majority of previous investigations have used models or inferences based on observational data to investigate these hypotheses, while experiments directly addressing this question have been rare. We experimentally investigated the relative influence of local and regional processes in structuring benthic invertebrate communities using artificial streams. We manipulated three factors—source pool for the macroinvertebrate community (headwater vs. mainstem) as a surrogate of network location, habitat complexity (high vs. low) in the flume, and dispersal (high vs. low)—and followed changes in macroinvertebrate community structure for 8 weeks. Previous research suggests that because headwater (HW) streams are isolated within river networks, HWs are less influenced by regional processes relative to more well‐connected mainstems (MSs). We therefore predicted (i) that flumes colonised from a HW source community would respond more strongly to our dispersal treatment than those colonised by MS communities because MS were already largely structured through dispersal‐driven processes, and (ii) that both HW and MS communities would respond to manipulations of local habitat, indicating that responses to the dispersal treatment were a direct result of dispersal driven dynamics rather than specific affinity for conditions in the flumes. Both of our predictions were strongly supported by the results of the experiment. For flumes with HW source pools, the high dispersal treatment had significantly higher diversity than low dispersal flumes. However, this difference only occurred in flumes with HW source pools and did not occur in flumes with MS sources. There was also strong evidence of community composition in HW flumes shifting significantly towards the macroinvertebrate composition in our experimental dispersal treatment. The major effect of experimental dispersal was to introduce new species in fairly low abundances as would be expected from dispersal via drift over a relatively short time. Both MS and HW colonised flumes showed highly significant signals of habitat filtering, though the influence of specific habitat differed between the source types. These results support the hypothesis that dispersal driven processes are a more important structuring force in well‐connected areas of networks by experimentally demonstrating the responsiveness of previously isolated communities to experimentally induced dispersal. They also demonstrate that this responsiveness is not due to an inherent difference in habitat affinity since source communities from both HWs and MSs responded to manipulation of habitat variables. This experiment only simulated one type of dispersal process in streams—drifting—and did not include simulated dispersal from other sources, nor did it include population dynamics given the relatively short duration of the experiment. Nevertheless, the sensitivity of previously isolated communities to one type of simulated dispersal is a powerful indication of the mechanisms that structure these systems.