Predictive Habitat Models for the Occurrence of Stream Fishes in the Mid-Atlantic Highlands

In most wadeable streams of the mid-Atlantic Highlands region of the eastern USA, physical habitat alteration is the primary stressor for fish. Models that predict the occurrence of stream-fish species based on habitat measures can be useful in management, and predicted probability of occurrence can be a measure of habitat suitability with which to compare alternative habitat management scenarios and assess the effectiveness of stream restoration, We developed such models for each of 13 mid-Atlantic Highlands stream-fish species and species groups by using multiple logistic regression and six instream habitat measures: depth, temperature, substrate, percent riffles, cover, and riparian vegetation. The predictive ability of the models ranged from 61% to 79% in cross-validation and from 38% to 85% on an independent data set. The models predicted well for both the original and test data sets for black bass Micropterus spp., brook trout Salvelinus jantinalis, darters Etheostoma and Percina spp., shiners Notropis spp., and suckers Hypentelium and Moxostoma spp. Suitable habitat for most of the fish species groups was characterized by intermediate depth, a high percentage of cobble and riparian vegetation, and a low percentage of instream cover. The relatively high predictive ability and reasonable responses to habitat measures indicated that these models could be useful for management. However, the models were more sensitive to depth and temperature than to measures that are more commonly affected by restoration activities, such as cover and riparian vegetation.