2019 journal article

Assessing potential anthropogenic drivers of ecological health in Piedmont streams through hierarchical modeling

FRESHWATER SCIENCE, 38(4), 771–789.

co-author countries: United States of America πŸ‡ΊπŸ‡Έ
author keywords: biological assessment; anthropogenic stressors; watershed development; stream restoration; stream buffers; mixed-effects models; macroinvertebrates
Source: Web Of Science
Added: December 9, 2019

Urban streams consistently have poorer ecological condition than natural streams. Poor ecological condition is caused by a myriad of anthropogenic impacts that alter hydrology and increase pollutant concentrations. Urban streams are monitored frequently, but viable management options for improving stream condition are ill-defined. A more complete understanding of the factors that influence biological condition, as well the ability to identify sites that deviate from expected condition, would help managers develop more efficient stream restoration strategies. Here, we use a hierarchical (multilevel) framework to model >3000 macroinvertebrate samples from the North Carolina Piedmont region, identify important natural gradients and anthropogenic factors that relate to stream condition, and demonstrate how hierarchical modeling can help identify potential restoration sites. In addition, we explore spatial (e.g., watershed versus stream buffer) and temporal (e.g., age of construction) aspects of land cover development. We found that watershed impervious cover (IC) is the best predictor of biotic index (BI) values. Additional factors significantly related to BI include age of watershed IC, canopy loss in stream buffers, reservoirs, wastewater treatment plants, antecedent precipitation, and geologic soil types. Synthesizing these factors in a hierarchical multiple linear regression model explained 76% of the variability (R2) in the BI, relative to 65% with only watershed IC. Of the remaining variability in the observations (24%), most was accounted for by site-specific random effects (16%), which characterize the deviation between predicted and actual biological condition. The model also suggests that newer development (post-1980) degrades stream health 30% less than older development. Additionally, canopy removal in stream buffers had 2 to 9Γ— the effect on BI relative to the addition of IC in upstream watersheds on a per hectare basis.