Spatially-explicit exposure models are designed to bridge the disconnect that exists between site wide averages and an assessment that captures exposures based on species-specific habitat preferences. The objective of this project was to test (verify) risk estimates for terrestrial wildlife (small mammals) and fish using the spatially explicit exposure models, SEEM and FishRand, respectively. This study compared model outputs to deterministic risk calculations and blood-lead based risk calculations to determine if spatially explicit exposure models increased the realism of exposure assessment.

Complex interactions among biological entities within the environment occur across varying spatial and temporal scales. Using geographical information systems (GIS) and other exposure models that incorporate spatial considerations, risks can be more accurately and realistically estimated. This project describes a test exposure method using the terrestrially-based SEEM model along with FishRand – a model that estimates body burdens of organic compounds in fish. Both models incorporate habitat suitability and contaminant heterogeneity.

Heterogeneous distribution of compounds at DoD training ranges are often found where wildlife use the environment in biased ways linked to habitat requirements. Exposure is modeled for a generalized single representative of a species where influences of habitat preferences are lost, and although likely protective, this process confounds exposure and risk estimates. SEEM has been tested for songbirds and shown to be predictive within a factor of 3, whereby the conventional deterministic model was 10× to 500× off from field observations. This project further tests and refines these models. Field verification of results was based on blood lead toxicity reference values for mammals and fish concentrations of persistent organic substances for fish.

For small mammals with small foraging areas, SEEM is no more predictive than site-wide average-based risk calculations. Although expected results emphasize that if habitat is not heterogeneous at ecologically-relevant scales, then spatially-explicit exposure models cannot improve risk estimates. SEEM increases the realism of wildlife exposure assessment and improves the analysis of population risk by tracking individual exposure in a local population, rather than for a single representative individual.

FishRand results did not provide a direct, quantitative linkage to GIS files; more information is likely required on fish foraging areas and strategies than would otherwise be developed. FishRand applications were developed for total PCBs, two individual PCB congeners, three homologue groups, DDT, DDE, and DDD. The spatially-explicit model consistently predicts tissue concentrations that closely match both the average and the variability of observed data across contaminants and environments. This probabilistic framework allows direct linkages to ecological assessments of fish population impacts.

There is spatial disconnect between site wide averages and an assessment that captures exposures based on species-specific habitat preferences. A key aim is to overcome this by applying spatially-explicit exposure models. Mistakes might occur by failing to consider properly the spatial aspects of exposure relative to the spatial domain (i.e., the habitat) of the particular population.

Implementing the FishRand model is uncomplicated; however, characterization data quality and quantity is key to successful implementation. In practice, data availability is a limiting factor with respect to model implementation. At many sites, data are not necessarily representative in time or space; similarly, often the bioaccumulation model is held accountable for limitations in understanding the sediment-water relationship. For many bioaccumulative contaminants, bulk sediment concentration is assumed to represent the relevant exposure metric with incomplete understanding of 1) how sediment concentrations may change over time (e.g., deposition, erosion, etc.); 2) the sediment-water relationship (e.g., low flow/highly dissolved concentrations at certain times of year, resuspension/contaminant transport events), and 3) potential sources and flux mechanisms (e.g., groundwater recharge, bioturbation, mechanisms for releasing "buried" sediments). A challenge in bioaccumulation modeling has been in understanding what is meant by true exposures. The FishRand model tries to overcome this by: 1) providing a mechanism for characterizing spatial heterogeneity exposure concentrations, and 2) simulating fish movement probabilistically rather than by static site averages, site-use factors, and similar deterministic adjustments.