We examined the distribution and abundance patterns of 69 fish species that commonly occur in the rivers of Michigan's lower peninsula to develop a simple, empirically based model for describing fish assemblages. We used cluster analysis to group fishes that shared similar abundance patterns at 226 stream sites. The 17 clusters we identified explained about 39% of the variation in species abundances among the stream sites, providing a reasonable, albeit simplified, picture of general associations of fishes in lower Michigan streams. Known ecological differences among species and further analyses suggested that a measure of cluster abundance should not be used to predict the abundances of its constituent species. We selected catchment area (CA) and low-flow yield (LFY; 90% exceedence flow divided by catchment area) as axes for plotting fish distributions and rivers because these variables link catchment-scale features of the landscape to multiple, site-scale characteristics of stream habitat (e.g., temperature, velocity, and depth) important to fishes. As a measure of groundwater loading to streams, LFY, which integrates the geological, landform, and soil characteristics of catchments, reached its highest values in basins predominated by highly permeable soils and relatively steep topography. Plots of fish clusters and species abundances on LFY-CA axes provided insight into the structure of fish assemblages in lower Michigan streams. When plotted on LFY−CA axes, the 17 fish clusters were distributed in a meaningful pattern that reflected stream size and temperature preferences of constituent species. The LFY-CA axes provided an empirically derived framework for comparing Michigan streams and for assessing the physical and biological potential of different river reaches. This has allowed fishery managers to better explain, justify, and build public support for river management plans and actions. Although the relationships among LFY, CA, and fish abundances we describe are specific to lower Michigan streams, our approach could be used to develop similar models specific to other regions.