The relative importance of physicochemical stressors in urbanizing streams: Evidence from model selection and multi-model inference

Few studies have simultaneously evaluated the relative importance of physicochemical stressors to biological impoverishment in urban streams, especially across different environmental settings.  We used data collected in 20-30 streams in each of six metropolitan areas (MAs) to evaluate the relative importance of 11 physicochemical stressors on the condition of algal, macroinvertebrate, and fish communities.  Separate linear regression models with 1or 2 stressors as predictors were developed for each MA and biological community.  Model parsimony was evaluated based on AIC and Akaike weights, and variable importance was quantified by summing the Akaike weights across models containing each stressor variable.  We found little evidence that stressors co-varied within MAs.  Stressor variables explained 20-79% of variance in biological condition.  A majority (28 of 44) of the most parsimonious (AIC<2) models contained two predictors, indicating that more variance could be explained by the additive effects of two stressors than by any single stressor alone.  Hydrophobics and hydrological variables were somewhat important to variation in all biological communities.  In addition, chloride, nutrients, and water temperature were often important to algal community condition, while pesticides appeared to be important to some invertebrate communities.  For fish communities, chloride, nitrogen, and water temperature were important stressors.  Our results suggest that the suite of stressors affecting urban streams varies across broad geographic areas and among basins within metropolitan areas.