Classification strengths of ecoregions, catchments, and geographic clusters for aquatic vertebrates in Oregon

The usefulness of ecoregions and catchments (hydrologic units) as bases for classifying aquatic vertebrate assemblages in western Oregon was compared using samples collected by electro-fishing from 137 wadeable stream sites d istributed evenly throughout the region. The classification strengths of th ese regionalizations were also compared with neutral-model classifications that were based either on intersite proximities in geographic distance or o n intersite similarities in the sampled vertebrate assemblages. The strengt h of each classification was assessed by the extent to which average within -class assemblage similarities exceeded the average similarity between clas ses. Mean similarity dendrograms were used as a concise graphical compariso n of between and within-class similarities for alternative classifications, whether they were specified a priori or constructed by clustering. For eac h a priori classification, a permutation test of the no class structure hyp othesis was performed. Classification strengths were assessed using the Sor enson-Dice (presence/absence) and Bray-Curtis (relative abundance) similari ty measures, applied to both species-level and family-level assemblage char acterizations, as well as for the Bray-Curtis measure applied to a set of 5 assemblage metrics that were designed to reflect stream impairment. For al l 5 measures of assemblage similarity, ecoregions had higher classification strengths than did large catchments, and large catchments had about the sa me strength as a stream-order classification A catchment classification wit h 1 of the 3 catchments split into 2 ecoregions separated assemblages as st rongly as ecoregions alone. A neutral-model classification based solely on geographic site proximity classified assemblages with about the same streng th as ecoregions. Another neutral-model classification of sites, based sole ly on their sampled assemblages, was at least twice as strong as any of the geographic classifications. Intermediate strength was seen in site groupin gs derived from a composite measure of between-site assemblage and geograph ic dissimilarities. Our results suggest that ecoregions and large catchment s do indeed have utility for classifying stream vertebrate assemblages. How ever, much of their classification strength may be a result of spatial auto correlation effects, rather than ecological factors that determine their pa rticular boundaries. Our similarity analyses also suggest that geographic p artitions can be expected to account for only a minor portion of the total variation seen in stream vertebrate assemblages across a large region.