Abstract
In this study, linear regression models were used to determine the ability of twelve common water quality metrics to predict known levels of habitat disturbance in 87 4 to 10 m wide streams throughout Minnesota based on their adult caddisfly fauna. While most of the models were statistically significant, those generated using species richness, percent of filtering collectors, genus richness, and family richness appeared to exhibit the highest biological significance. Multiple regression analysis of all metrics indicated that only species richness and percent filtering collectors were significant predictors of disturbed habitat; the combination of these two variables explained 86% of the variation in the level of disturbed habitat.