Citation:

Herlihy, A., J. Sifneos, R. Hughes, D. Peck, AND R. Mitchell. Lotic Fish Assemblage Clusters Across the Conterminous USA and Their Associations with Environmental Variables. Chapter 18, R.M. Hughes, D. M. Infante, L. Wang, K. Chen, and B.F. Terra (ed.), Advances in Understanding Landscape Influences on Freshwater Habitats and Biological Assemblages. American Fisheries Society, Bethesda, MD, , 385-408, (2019).

Impact/Purpose:

Aquatic ecosystem classification is a very important part of how we study and manage streams and rivers. The NRSA data provide a unique opportunity to investigate landscape patterns in lotic fish assemblages at a continental scale using information from a consistent data source. Our previous effort at clustering fish across the U.S. (Herlihy et al. 2006) was limited by the unknowns involved with compiling fish data from many different sources collected with varying sampling protocols. It was also severely limited by the lack of consistent environmental data across those surveys to the extent that we could only look at landscape environmental data. As the natural variability in aquatic biota is quite high, reporting and analyzing the results of biomonitoring requires a classification framework that minimizes this variability so that expectations of least-disturbed condition and the effects of anthropogenic disturbance can be more clearly defined. Our results showed that biologically derived classes of sites have higher classification strength than either basins or ecoregions. Thus, for the purposes of biomonitoring, it could be advantageous to replace physical regions with a classification based on aquatic biota. We used cluster analysis to derive fish assemblage clusters. Assemblage clusters were then related to the local and catchment scale environmental data to assess the primary drivers of fish assemblage structure and to predict cluster membership. We use NRSA data to develop fish clusters across the U.S. using cluster analysis. The relationship of the clusters to local and catchment scale environmental variables are assessed by a variety of methods including ordination and indicator species analysis. We also predict cluster membership from the environmental data, assess the classification strength of various landscape classifications versus that of the fish clusters, and quantify the repeat visit reproducibility of the fish cluster analysis. The results from our study show that whereas variability in fish assemblages is large over the range of stream/river sizes across the conterminous US, it is possible to divide sites into six clusters that could be defined by specific indicator species and predictable from environmental data using both classification tree analysis and discriminant function analysis. Ordination identified three environmental gradients as the primary drivers of the biological clusters; stream size, temperature, and streamwater ionic strength. The biological classification was very reproducible within a sample year based on repeat sample visits to the same sites. Reproducibility was lower based on samples taken four-six years apart. The Office of Water, EPA Regions and State environmental program managers will be interested in the results for informing their management and policy actions.This research contributes to States and EPA being able to better assess the extent to which waters of the US meeting the fishable-swimmable goals of the Clean Water Act.

Description:

Between 2008 and 2014, the first two phases of the National River and Stream Assessment (NRSA) sampled fish assemblages in 2554 stream and river sites across the conterminous United States. Associated physical habitat, water chemistry, and landscape data were also collected. We used cluster analysis to derive fish assemblage clusters. Assemblage clusters were then related to the local and catchment scale environmental data to assess the primary drivers of fish assemblage structure and to predict cluster membership. The results from our study show that whereas variability in fish assemblages is large over the range of stream/river sizes across the conterminous US, it is possible to divide sites into six clusters that could be defined by specific indicator species and predictable from environmental data using both classification tree analysis and discriminant function analysis. Ordination identified three environmental gradients as the primary drivers of the biological clusters; stream size, temperature, and streamwater ionic strength. The biological classification was very reproducible within a sample year based on repeat sample visits to the same sites. Reproducibility was lower based on samples taken four-six years apart. Biological monitoring is essential for the complete assessment of aquatic ecosystem condition. As the natural variability in aquatic biota is quite high, reporting and analyzing the results of biomonitoring requires a classification framework that minimizes this variability so that expectations of least-disturbed condition and the effects of anthropogenic disturbance can be more clearly defined. Our results showed that biologically derived classes of sites have higher classification strength than either basins or ecoregions. Thus, for the purposes of biomonitoring, it could be advantageous to replace physical regions with a classification based on aquatic biota.

Record Details: