Local and Landscape Determinants of Smallmouth Bass Habitat Suitability in Michigan Streams

Black bass Micropterus spp. are among the most popular sportfish species in North America, however, managers currently have insufficient information about patterns and drivers of their distribution. Tradeoffs exist with the spatial scale at which to collect information associated with the distribution of species. Local-scale data are expensive to collect, but are likely directly associated with occupancy of a particular site. Landscape-scale data are increasingly available, are easily scaled and linked to site occupancy, but management often does not occur at this scale. We utilized site-specific habitat data collected as part of a statewide stream sampling program from 2002 to 2013 to generate presence-absence data for Smallmouth Bass M. dolomieu. We used site specific local-scale, landscape-scale, and a combination of local- and landscape-scale data to create random forest classification tree models of Smallmouth Bass presence and absence. All three models had similar total error rates ranging from 13.9% to 16.6%. Overall rates of success and error did not differ among the three models (P = 0.95). The model using only landscape-scale data were used to predict the presence or absence of Smallmouth Bass for all Michigan stream segments and had an error rate of 17% based on an independent data validation. These data suggest that our approach has utility for predicting Smallmouth Bass occurrence in Michigan streams by identifying important habitat features associated with Smallmouth Bass occurrence. This approach could be extended to understand the distribution of other black bass species in locations where distributional data may be more limited.