Comparative Utility of Selected Frameworks for Regionalizing Fish‐Based Bioassessments across the United States

Abstract

AbstractRegional frameworks for bioassessment are necessary to stratify natural geographic variation in biotic assemblages and to calibrate bioassessment metrics and indices. In the United States, alternative frameworks have not been evaluated over large geographic extents (nationwide or continent‐wide) or compared with neutral models to document the relative utility of existing frameworks. We used the U.S. Geological Survey's National Water Quality Assessment fish assemblage data from 1,140 fluvial sites to evaluate the utility of physiographic regions, U.S. Environmental Protection Agency ecoregions, aquatic zoogeographic regions, and hydrologic landscape regions (HLRs). A zoogeographic—physiographic (Z‐P) region combination was tested along with nesting of HLR within all other frameworks. All frameworks were compared with a hierarchical grid that represented a neutral spatial framework and enabled us to examine effects of spatial autocorrelation. Classification strengths were inferred from the intraclass correlation coefficients (ICCs) of random‐effects analysis of variance models. Seventy percent of the variation in eight commonly used functional bioassessment metrics was explained by two principal components (PCs), which were used to characterize the underlying structure of current bioassessment metrics and geographic variation in fish assemblages. The order of classification strengths of the frameworks tested was Z‐P > ecoregion > zoogeography > physiography > HLR. The highest estimated ICC was 30% of the variance in the first PC of metrics explained by Z‐P regions. As stand‐alone frameworks, all frameworks except HLR could explain some variance in functional metrics above that attributable to spatial autocorrelation at the national scale. The HLR framework was the only spatially noncontiguous framework, and its weak classification strength reflects the importance of spatial autocorrelation in national‐scale regionalizations. Whenever nested in other frameworks, HLR explained significant additional variation. A refinement of the Z‐P framework and further integration with other frameworks such as HLR could lead to an improved, generally applicable framework for fish‐based bioassessments.