• spatial scale;
  • rivers;
  • site selection;
  • substrate texture


Few studies have tested whether a small-scale, experimental difference in biotic density and diversity detected at one site has any predictive capacity at other locations, nor whether such experimental differences are significant relative to overall spatial variation. In this study, we tested textural differences between substrata– previously shown to be critical to macroinvertebrate density and diversity at one site on the Steavenson River in southeastern Australia–and replicated the experiment at this site and at two others on the same river, plus at three sites on each of two other rivers (the Acheron and Little Rivers) in the same catchment. Nine rough and nine smooth colonization substrata were set out at each site on 19 December 1995, and at the start of the experiment (26 and 27 February 1996), invertebrates were hand-picked from substrata while leaving epilithon intact. We re-collected substrata after 28 days. Substrate texture had a consistent and strong effect upon species richness at all sites, except one, and was also associated with compositional differences at the family level. Textural effects on macroinvertebrate densities were present and varied among sites and rivers somewhat, but only three of nine common species showed significant interactions between textural effects and one of the two spatial scales, and these interactions explained <10% of variance. In most but not all cases, textural effects were the same as that previously observed–higher numbers on rough surfaces compared to smooth. Most variation in species richness and densities was explained by site, which accounted for 30% of variation in species richness, 60% of total numbers of individuals, and 9–75% of variation in abundances in all nine common taxa. Site differences were also associated with large shifts in faunal composition at the family level. Differences between rivers explained 0% of the variance in most cases but it was significant in explaining the abundance of one common species (the beetle Simsonia wilsoni), where it accounted for 59% of the variance. Our results indicate that a small-scale effect, such as textural differences between substrata, can produce a consistent effect on measures of community structure and have some predictive capacity at other locations. However, the large differences between sites, relative to those seen between whole rivers, suggest that efforts must be directed to discovering what factors cause such localized fluctuations. Sites cannot be reasonably used as ‘representative’ of larger spatial units, such as long sections of entire rivers. Additionally, rivers may not necessarily have signature variation that always overrides localized, site-based differences.