Watersheds have historically been used as the appropriate spatial classification unit for managing water resources. However, geology, soil type, predominant vegetation, and climate have obvious influences on water quality and are not constrained by watercourses or political boundaries. This concept has evolved for several decades and developed the concept of ecoregions and other spatial schemes. While this approach to water resource management has considered the interaction between water quality and biological integrity (aquatic community structure and assemblage), it has not been applied in the context of predicting aquatic toxicity. As such, a previously published study providing a chemical and toxicological data set consisting of 24 sampling sites in South Carolina, USA, and was used to develop empirical models for predicting acute copper (Cu) toxicity to larval fathead minnows (Pimephales promelas). Moreover, numerous spatial classifications (hydrologic units, ecoregions, stream order, adjacent land use, and proximity to certain land uses) and seasonality were used to delineate sites and develop empirical models based on these different classifications. An independent sampling and testing regime was implemented to determine the performance of the empirical models and whether certain classifications could be used to extrapolate toxicity data across spatial landscapes. Additionally, a computational model (biotic ligand model [BLM]) for deriving site-specific water quality criteria for Cu also was used as a reference for current regulatory application. Empirical models based on delineations of stream order, hydrologic unit, and downstream distance to urbanization accurately predicted at least 60% of the observed Cu toxicity values within the supplemental data set. Delineations based on adjacent land use, ecoregions, and seasons were not as useful for predicting acute Cu toxicity but demonstrated better performance than the BLM.
Integrated Environmental Assessment and Management
Vol. 4 • No. 2
Vol. 4 • No. 2
biotic ligand model