Spatial continuity of fish assemblages and environmental factors on the Ohio River

Distributions of fish assemblages, habitats, and water quality in large rivers are neither random nor uniform. The interplay between environmental constraints and spatial heterogeneity is a common theme of terrestrial and stream research. However, it is infrequently considered in large rivers. We determined the spatial autocorrelation (the decrease in site similarity with increasing site separation) of data from the Ohio River. Spatial autocorrelation was calculated using ArcGIS Spatial Analyst (version 9.3). Correlations with river-mile (if present) were removed to reveal relationships based on site proximity, not site location. Spatially autocorrelated metrics included the extent of forest and developed areas in the main-channel riparian and local upriver catchments, dissolved oxygen and chlorophyll concentrations, sediment loading, the density of upriver NPDES discharges, and water temperature. Most fish assemblage metrics (e.g. richness, CPUE, proportional, biomass) metrics, physical habitat metrics (e.g. littoral substrate slope, cover, turbidity) and nutrient concentrations were not spatially autocorrelated. Because a probability sampling design was used, correlations should not reflect data density. Patchiness likely is the response to relatively large-scale drivers operating from outside the river channel. Homogeneous or apparently random distributions may reflect subtle in-channel physical processes (e.g. sediment sorting) or biological processes (e.g. nutrient uptake) that were not measured by our approach.