Fish communities were surveyed at 20 wadable stream sites during 1993-95 as part of the U.S. Geological Survey's (USGS) National Water- Quality Assessment (NAWQA) Program's assessment of the Western Lake Michigan Drainages. Part of the NAWQA design is to incorporate ecological data into an overall environmental assessment. Collection of fish-community data was part of this ecological assessment.

The Western Lake Michigan Drainages study area is located in eastern Wisconsin and parts of the Upper Peninsula of Michigan. To isolate the effects of individual factors on stream quality, the study area was subdivided into 28 environmental settings, or relatively homogeneous units (RHUs), on the basis of land use/land cover, texture of surficial deposits, and bedrock geology. A fixed monitoring site was established on a wadable stream within 8 of these RHUs to determine the status and trends of water quality in a representative stream. Water-quality characteristics, ecological- community data, and stream-habitat factors were measured at these sites during 1993- 95.

Fish communities were sampled at the 8 wadable fixed sites once a year during 1993-95. At three of these sites, multiple-reach samples were collected in 1994 to determine within-site variation. Fish communities also were sampled at an additional 12 sites, 11 in 1993 and one in 1995, within the 6 largest RHUs. The sites, 1-3 per each of the 6 RHUs, were located on streams with drainage basins of similar size as the fixed sites within the same RHUs.

A total of 44 fish species from 12 families were collected at the 20 sites. The family with the most species represented were the minnows. The number of species per site ranged from one at a small urban site (Lincoln Creek) in 1995 to 21 at an agricultural site (North Branch Milwaukee River) in 1995. The number of individuals collected in one sampling pass ranged from 21 at a stream in the forested northwest part of the study area (Peshekee River) in 1995 to 498 at an agricultural site (East River) in 1995. White sucker (Catostomus commersoni) were collected at 17 sites, the most of any species. Species that are indicative of a coldwater environment were collected at 12 sites.

Detrended correspondence analysis (DCA) of multiple-reach and multiple-year data indicated that species composition at each of these sites were fairly consistent between reaches and years. Thus, for simplicity, most analyses were done using 1993 data only.

Index of Biotic Integrity (IBI) scores on 1993 data ranged from very poor at a channelized urban site to excellent at 3 sites; 2 in primarily agricultural areas and 1 in a forested area. Seven sites each scored good or fair, and two sites scored poor. Sites with multiple-year or multiple-reach data did not vary significantly within the error factor of the IBI.

DCA of fish-community data from 19 sites indicated that coldwater sites were tightly grouped, whereas warmwater sites showed a larger gradient. This was expected, given the potential for greater diversity among warmwater sites. Fixed sites were shown to be representative of the study area as a whole, while specific fish communities could not be attributed to particular RHUs.

Cluster analysis revealed two major groups of sites and two outlier sites. The two groups represented coldwater and warmwater streams, while the outlier sites were the urban site and a species- rich site with high biotic integrity that drains primarily agricultural land.

Canonical correspondence analysis (CCA) revealed that soil credibility was a significant predictor of species composition. Though not statistically significant, land use, soil permeability, and bedrock permeability also were indicated as predictors of fish-species composition by CCA.