Predicting the occurrence of chemicals of emerging concern in surface water and sediment across the US portion of the Great Lakes Basin

Chemicals of emerging concern (CECs) are introduced into the aquatic environment via various sources, posing a potential risk to aquatic organisms. Previous studies have identified relationships between the presence of CECs in water and broad-scale watershed characteristics. However, relationships between the presence of CECs and source-related watershed characteristics have not been explored across the Great Lakes basin. Boosted regression tree (BRT) analyses were used to develop predictive models of CEC occurrence in water and sediment throughout 24 U.S. tributaries to the Great Lakes. Models were based on the distribution of both broad-scale and source-related watershed characteristics. Twenty-one upstream watershed characteristics, including land cover, number of permitted point sources, and distance to point sources were used to develop models predicting the probability of CEC occurrence in surface water and bottom sediment. Total accuracy of BRT models ranged from 66% to 94% for both matrices. All 21 watershed characteristics were important predictor variables in at least one surface-water model; twenty were important in at least one bottom-sediment model. Among the model variables, developed land use and distance to point sources were important predictors of the presence of CEC classes in both water and sediment. Although limitations exist, BRT models are one tool available for assessing vulnerability of fisheries and aquatic resources to CEC occurrences. Published by Elsevier B.V.