Predictive abilities of numerical sediment quality guidelines in Sydney Harbour, Australia, and vicinity.

Matching chemical and toxicological data of surficial sediments from Sydney Harbour, Australia, and vicinity, were collected to evaluate predictive abilities of Effects Range-Low (ERL), Effects Range-Median (ERM) and other sediment quality guidelines (SQGs). Samples (n=103) containing a wide range of chemicals and concentrations were subjected to a battery of 4-6 toxicity tests. ERLs and functionally equivalent low-range SQGs were highly predictive of non-toxicity when not exceeded, as incidences of toxicity were 0-8%. ERMs and other mid-range SQGs were predictive of toxicity in combined test data with > or = 80% of toxic samples with one or more SQG exceeded. Predictive abilities of mid-range SQGs were low for amphipod survival tests (generally < 20% toxic samples), but the incidence of toxicity increased with increasing numbers of SQGs exceeded and increasing mean SQG quotients. Predictive abilities of SQGs generally matched their narrative intent when outcomes of multiple toxicity tests were considered, and were consistent with North American data. Functionally equivalent SQGs gave comparable results (except where developed for single chemical classes), indicating that predictive abilities were primarily influenced by sensitivities of test species and/or the bioavailability of contaminants.