Analysis of Estuarine Sediment Contaminant and Toxicity Data for Eliciting Responses

John F. Paul 1 and Thomas P. O'Connor 2

1 U.S. Environmental Protection Agency, NHEERL (B205-01), Research Triangle Park, NC 27711
2 NOAA/NOS, National Centers for Coastal Ocean Science, Silver Spring, MD 20910

An important aspect of criteria development is understanding how well we can predict biological response. Field et al. (2002) have developed logistic regression models for predicting likelihood of sediment toxicity using bulk sediment chemical concentrations. We used the EMAP-Estuaries Virginian Province 1990-1993 database to test the applicability of these models. A probability based sampling design was used to collect these data, so comparison can be made with the likelihood of observing biological response. This comparison could only be done approximately because Field et al. defined toxic as less than 90% amphipod survival in 10-day laboratory exposures. This led to the seemingly anomalous conclusion that approximately 22% of the entire estuarine area in the Virginian Province had toxic sediments. The conventional definition of toxicity based on 80% survival lowered this to less than 10%. The small estuarine systems across the province had the highest proportion of area with observed toxicity. Overall comparison of the Field et al. predictions of likelihood of observing toxicity against actual distributions of observations (based either on the 90% or 80% definition of toxicity) indicates that the predictions overestimate the actual occurrence of toxicity. Details of the comparison will be discussed.

Keywords: sediment toxicity, estuaries, statistical models, sediment contaminants

Analysis of Estuarine Sediment Contaminant and Toxicity Data for Eliciting Responses

Local Navigation