SPECIES ASSOCIATIONS AND HABITAT USE OF STREAM FISHES - THE EFFECTS OF UNAGGREGATED-DATA ANALYSIS

Fish species in several families were collected in microhabitat quadra ts during warmer seasons of 1989-1990 in the upper Roanoke River (URR) , a midsized stream of southwestern Virginia. The quadrats were visual ly classified into five mesohabitat types (medium and shallow pools, r uns, and slow and fast riffles) before sampling. Uni-, bi-, and multiv ariate statistical analyses on fish-abundance data were used to establ ish species associations, fish-habitat relations, and habitat-use guil ds. This included examination of coefficients of variation for species densities across quadrats, ANOVAs for species densities across differ ent habitat types, and factor analyses of species associations. Althou gh these analyses were generally concordant, they enabled only crude s egregation of fishes by habitat niche, namely rheophilic, limnophilic, or generalized. These analyses also showed that fish species were lar gely independent of each other in their habitat use at smaller spatial resolutions (quadrat-level) and moderately large spatial extents (str eam reaches of three to five meander sequences). Hence, pooling of qua drat-abundance data for fishes into several mesohabitat types (e.g., s hallow-pool vs. fast-riffle), or consideration of mean habitat use for fish species over all quadrats (e.g., for depth and velocity) will al low more efficient and accurate delineation of habitat-use guilds. The se analyses also suggest that abundant fish species do not differ from rarer species in variability of occurrence across spatiotemporal samp les.