Northern Prairie Wildlife Research Center
Each of four groups of streams produced by CA were considered unique for the analysis of species associations. From the original Quattro Pro data base created during this study which contained all sites and species information for all years, four new arrays were created based on stream clusters A-D. Each array contained binary (presence or absence) information for species present at >4% of sites sampled on streams of its corresponding cluster. The four arrays produced were of sizes 633 (sites) x 58 (species), 207x33, 122x27, and 55x18. To determine any degree of similarity between species, the four arrays were used separately to calculate simple matching coefficients (SM) (Kovach 1993) for each species pair (A and B), where SM=(a+d)/(a+b+c+d), with a being the number of sites that species A and B where found together, b the number of sites where species A was present but species B was absent, c the number of sites where species B was present but species A was absent, and d the number of sites where neither species A or species B was present. The SM ranged from near 0 (for a species pair rarely if ever occurring together) to near 1 (for a species pair often occurring together).
To place species into meaningful assemblages, the similarities for all combinations of species pairs were summarized into four arrays or similarity matrices of sizes 58x58, 33x33, 27x27, and 18x18 which corresponded to stream clusters A-D. Four separate cluster analyses (weighted centroid) using MVSP (Kovach 1993) were conducted to produce dendrograms which grouped fish species. A minimum SM of 0.1 was used to define fish assemblages. As with the cluster analysis of streams, other measures of similarity among species were attempted along with several different methods of clustering. All techniques provided similar results, with the SM and weighted centroid clustering making the most sense ecologically.