Distribution of Fishes in the Red River of the North Basin on Multivariate Environmental Gradients

Species Associations of Stream Fishes

Stream similarity

Cluster analysis based on the presence or absence of stream fish species provided evidence of likeness of fish assemblages among 27 streams in the Red River basin (Figure 7). The stream pair Buffalo River and Clearwater River had the highest JI of 0.76, followed by stream pairs Snake and Tamarac, Sheyenne and Wild Rice, MN, Otter Tail and Red Lake, Pembina and Forest, and Goose and Sandhill, with JI values of 0.71, 0.68, 0.67, 0.64, and 0.61, respectively. Other streams clustered at lower Jaccard's values, and using a minimum JI of 0.00 for defining clusters, the analysis produced four distinct groups of streams (Clusters A-D). Cluster A included the Red, Sheyenne, Otter Tail, Red Lake, Pelican, Buffalo, and Clearwater Rivers and the Wild Rice River in Minnesota. Cluster B included the Pembina, Forest, Park, Turtle, Goose, Sandhill, Two, and Roseau Rivers. Cluster C included the Elm, Rush, Maple, Bois de Sioux, Mustinka, and Rabbit Rivers and the Wild Rice River in North Dakota. Cluster D included the Tongue, Middle, Snake, and Tamarac Rivers. Average species richness of cluster A, B, C, and D streams was 52, 32, 21, and 20 species, respectively.

Figure 7. Cluster analysis of 27 streams in the Red River basin based upon the presence or absence of 76 naturally reproducing stream fish species. The letters (A-D) indicate clusters defined by using a minimum Jaccard's Index of 0.0 (dashed line).

The first two principal components (PCI and PCII) from the PCA accounted for 39% of variance in the binary species data. A scatterplot of streams on PCI and PCII suggested results similar to the CA (Figure 8); and decomposition of these axes provided insight into the species most responsible in defining groups of streams, as the loadings (=correlations) of the individual species (variables) on a PCA axis are an indication of the contribution of each original variable to the new PCA axis (Table 5). Species with relatively high loadings (>0.20) on PCI were the silver redhorse, golden shiner, blackchin shiner, spottail shiner, and bluegill. These species were descriptive of streams with high PCI scores, such as the Otter Tail, Red Lake, Buffalo, and Sheyenne Rivers (Table 6), all of which are members of cluster A in the CA (Figure 7). Streams with relatively high negative scores on PCI included the Elm, Rabbit, and Mustinka Rivers, all of which are members of cluster C. There were no species with loadings <-0.20 on PCI (Table 5). Species with loadings >0.20 on PCII were the channel catfish, river shiner, spotfin shiner, bigmouth buffalo, and sauger. These species were descriptive of streams with high PCII scores, such as the Red, Maple, Sheyenne, and Bois de Sioux Rivers (Table 6), members of clusters A and C (Figure 7). Species with loadings <-0.20 on PCII included the central mudminnow, pearl dace, and finescale dace (Table 5). These species were descriptive of streams with high negative PCII scores, such as the Middle, Clearwater, Pelican, and Snake Rivers (Table 6), members of clusters A and D (Figure 7).

Figure 8. Twenty-seven streams in the Red River basin plotted on principal component axes 1 and 2 (PC1 and PC2). Letters represent clusters formed by cluster analysis (Figure 7). Axes are linear combinations of fish species.

Significance of clusters A-D was determined using PCA scores from the first five axes for all streams. The MRPP caused rejection of the null hypothesis of "no difference" among clusters (P=0.241x10^-9), so in terms of fishes present, the streams within each cluster were considered unique from all others for analyses of species associations which follow.

Association analysis

The four distinct clusters of streams produced by CA were used to place stream fish species into assemblages which represented common associations of fishes found in streams of each cluster. Fifty-eight species occurred at >4% of 633 sampling sites in streams of cluster A (Figure 7) during 1962-1994. Species paired by CA which had simple matching coefficients >0.90 included the northern hogsucker and central stoneroller, goldeye and mooneye, silver chub and sauger, bigmouth buffalo and greater redhorse, mimic shiner and banded killifish, stonecat and burbot, emerald shiner and rosyface shiner, and weed shiner and largemouth bass (Figure 9). Using a minimum simple matching coefficient of 0.1 as an arbitrary cutoff point for defining clusters, seven fish assemblages were produced by CA, including large turbid river species (A1), clear water riffle and run species (A2), clear water, weedy pool species (A3), large stream and pool species (A4), and common species in pools and runs (A5-A7).

Figure 9. Species association analysis for fishes of the Red, Sheyenne, Otter Tail, Red Lake, Pelican, Buffalo, and Clearwater Rivers and the Wild Rice River, Minnesota (cluster A of Figure 7). Species assemblages (A1-A7) were defined using a simple matching coefficient of 0.1 (dashed line).

Thirty-three species occurred at >4% of 207 sampling sites in streams of cluster B (Figure 7) during 1962-1994. Species paired by CA which had SM >0.90 included the chestnut lamprey and burbot, spotfin shiner and channel catfish, pearl dace and northern redbelly dace, brown bullhead and black crappie, and sauger and freshwater drum (Figure 10). Using a minimum SM of 0.1 as an arbitrary cutoff point for defining clusters, five fish assemblages were produced by CA, including deep-pool species (B1), species found in pools often with weeds and mud bottom (B2), shallow pool and riffle species (B3), and common species in pools and runs (B4 and B5).

Figure 10. Species association analysis for fishes of the Pembina, Forest, Park, Turtle, Goose, Sandhill, Two, and Roseau Rivers (cluster B of Figure 7). Species assemblages (B1-B5) were defined using a simple matching coefficient of 0.1 (dashed line).

Twenty-seven species occurred at >4% of 122 sampling sites in streams of cluster C (Figure 7) during 1962-1994. Species paired by CA which had SM >0.90 included the silver redhorse and freshwater drum, emerald shiner and bigmouth shiner, shorthead redhorse and channel catfish, quillback carpsucker and walleye, northern redbelly dace and trout-perch, white crappie and yellow perch, and creek chub and blackside darter (Figure 11). Using a minimum SM of 0.1 as an arbitrary cutoff point for defining clusters, four fish assemblages were produced by CA, including deep pool and run species (C1), shallow pool species (C2), weedy pool and run species (C3), and common pool and run species (C4).

Figure 11. Species association analysis for fishes of the Elm, Rush, Maple, Bois de Sioux, Mustinka, and Rabbit Rivers and the Wild Rice River in North Dakota (cluster C of Figure 7). Species assemblages (C1-C4) were defined using a simple matching coefficient of 0.1 (dashed line).

Eighteen species occurred at >4% of 55 sampling sites in streams of cluster D (Figure 7) during 1962-1994. Species paired by CA which had simple matching coefficients >0.80 included the brassy minnow and yellow perch, common carp and northern pike, and pearl dace and northern redbelly dace (Figure 12). Using a minimum SM of 0.1 as an arbitrary cutoff point for defining clusters, three fish assemblages were produced by CA, including medium pool and slow riffle species (D1), medium and deep pool species (D2), and shallow, weedy pool species (D3).

Figure 12. Species association analysis for fishes of the Tongue, Middle, Snake, and Tamarac Rivers (cluster D of Figure 7). Species assemblages (D1-D3) were defined using a simple matching coefficient of 0.1 (dashed line).

Previous Section -- The Distribution of Stream Fish Species
Return to Contents
Next Section -- The Species-environment Relationship