Abstract

 The risks posed by contaminated sediments are a function of the toxicology of the contaminants present and the processes controlling exposure to these contaminants. Biological tests used in assessments of sediments provide information about the hazard posed by the material, but relatively little information about the nature and extent of exposure to those hazards. Exposure to sediments is controlled by processes operating at both small (e.g., bioavailability) and large (e.g., site use) spatial scales. Assumptions that are commonly made about the spatial and temporal elements of exposure to contaminated sediments can lead to poor management decisions. The authors have developed an approach for addressing some of these elements in sediment exposure assessment. Our approach makes use of a spatially explicit two-dimensional aquatic food web model. A probabilistic adaptation of the Gobas bioaccumulation model is used to account for temporal variation in concentrations of polychlorinated biphenyls (PCBs) in sediment and water. The spatial submodel is used to account for spatial and temporal characteristics of fish exposure. We applied different assumptions concerning the geographic distribution of several offshore fish subpopulations based on species biology, including such variables as foraging area, size of the site, migration characteristics of the species, and hotspot area. We incorporated these variables together with an estimate of differential attraction to a management site within a spatially explicit model to assess the sensitivity of fish tissue PCB concentrations to spatial exposure parameters. Key input parameters are defined as "uncertain" or "variable" and described by probability distributions in a two-dimensional Monte Carlo simulation model. The selected fish species include mummichog, summer flounder and winter flounder. The model provides more realistic estimates of exposure than is typically assumed. Risk estimates in the cases we've evaluated to date using such a modeling approach are as much as one or more orders of magnitude lower than estimates made using more simplified, assumption-laden approaches.