• habitat size;
  • isolation;
  • local processes;
  • regional processes;
  • stream network


1. Both local and regional processes simultaneously control species assemblages depending on spatial habitat configuration. In dendritic networks like streams, the unique spatial arrangement of habitats produces various combinations of local habitat size and isolation. Stream invertebrate assemblages could therefore be controlled by different combinations of local and regional processes, depending on their location in the network.

2. Using quantile regression, we investigated how local habitat size, local environmental conditions and spatial isolation influenced variation in assemblage composition. Adult Trichoptera and benthic macroinvertebrate assemblages were represented by non-metric multidimensional scaling (NMDS) ordination scores, as were local environmental conditions, in four headwater stream networks in New Zealand.

3. With increasing local habitat size, there was a decrease in variation in assemblage composition (NMDS scores) of both adult Trichoptera and benthic macroinvertebrates. This relationship between habitat size and assemblage variation was related to local habitat conditions at the upper limit of assemblage variability and spatial isolation at the lower limit of assemblage variability, for both adult Trichoptera and benthic assemblages, indicating joint local and regional controls on stream invertebrate assemblages.

4. The relationships between local assemblages and their neighbours, based on community similarity scores, differed between benthic macroinvertebrates and adult Trichoptera. For benthic assemblages, the larger the stream, the more similar assemblages were to neighbouring assemblages, whereas there was no consistent relationship between assemblage similarity and stream size for adult Trichoptera. This difference in structuring could be attributed to contrasting spatial influences linked to the different dispersal modes of adults and larvae. However, because adult and benthic assemblages are not independent, the influence of life stage on spatial distribution is difficult to determine (i.e. it is essentially a ‘chicken and egg’ argument).

5. Overall, our approach using quantile regression to evaluate limit responses, rather than regressions on means, has highlighted the joint importance of local habitat and spatial processes in structuring stream invertebrate assemblages. Furthermore, we have provided evidence for the importance of the spatial network arrangement and interactions between life stages and dispersal processes, in structuring stream assemblages.