Predicting biological condition in southern California streams

► We developed a boosted regression tree model for predicting stream biological condition at unsampled sites. ► The two most important variables were population density and development in the riparian buffer. ► The response of stream biological condition to population density was large, rapid and nonlinear. ► The response to development in the riparian buffer was approximately linear. ► Predictive models are useful for understanding how continued urban development may affect stream biological condition.