Comparison of observed and model-predicted fish assemblages in the St. Lawrence River valley

St. Lawrence Valley lotic habitats harbor a variety of fishes within an array of relatively small tributaries draining into North America’s Second largest river. Fish distributions and abundances in those habitats range from widely common to some of the rarest in the Great Lakes Basin and field samples are limited. Species-habitat models provide a means of predicting the occurrence (and abundance) of any given species or species assemblage within a given habitat and have been used to predict St. Lawrence fish assemblage structure. To help evaluate model validity, we determined the similarity of predictions from neural network species-habitat models and field-collected species assemblages from St. Lawrence habitats. Assemblage diversity of the 125 field samples varied widely (0.0 – 1.9). Neural network models performed well (R²>0.9), but preliminary results indicate generally low similarity among assemblages.  Model predictions appear to representative of beta diversity, rather than the alpha diversity observed in the field. Detailed analyses will examine species composition and relative abundances to explain similarity patterns among observed and predicted fish assemblages. These models have the potential to be valuable aids in assessment and prioritization of streams for restoration and conservation action.