We often develop and apply models to estimate contaminant concentrations in environmental media as part of exposure assessment. Models can be used to supplement measurements, or to predict impacts when measurements cannot be made. For example, since emissions from a proposed source cannot be measured, modeling is necessary to determine whether (and under what circumstances) the source is likely to meet air quality standards. For existing sources, the expense of monitoring or physical barriers may make it impractical to measure concentrations at all of the times and locations of interest, and also introduces the problem of separating source-related contributions from background due to other sources.
We employ both standard and special purpose models as appropriate to project needs. In some fields, like air dispersion, we use established models such as the AERMOD air pollutant dispersion model, applied according to guidance to gain regulatory acceptance and credibility. Sometimes simpler models may be used — for example, in air dispersion modeling box models are often useful for estimating air pollutant concentrations near sources, and their simplicity facilitates the communication of model results. Where necessary, we use more complex models; or complex models may arise from simple ideas — for example, emission of PCBs from layers of contaminated concrete can be modeled using simple diffusion, but the resultant mathematics is quite complex. Fusion of models and empirical data can be useful in establishing robustness and exploring sensitivity and uncertainty. For example, application of the Johnson and Ettinger algorithms to model vapor intrusion can be applied to different groundwater, soil-gas, and soil data to provide insights on indoor air quality predictions, and more detailed models can be applied to estimate the complexities of vapor transport in the vadose zone. Critical evaluation and (to the extent possible) verification of model predictions is essential. We assess the theoretical underpinnings of standard models to ensure they are used appropriately, and use first-order principles to develop special-use models as needed. Model predictions are examined with mass balance checks and other tests to check for spurious results.
Mercury from cement kiln emissions
We developed a multi-pathway risk assessment to assess the consequences of air toxics emissions from a Portland cement manufacturing facility in Indiana. The risk assessment was required to demonstrate that there were no significant risks associated with the facility's use of hazardous waste fuel. The U.S. EPA's Human Health Risk Assessment Protocol methodology was applied to estimate the environmental concentrations of various chemicals potentially emitted to the atmosphere. A preliminary assessment indicated that mercury levels in fish could be increased to unacceptably high levels if the facility released mercury at its permitted rate. The projected mercury levels in fish, however, seemed unrealistically high compared with typical measured levels. Consequently, a critical evaluation was undertaken of the series of models used to relate mercury levels in fish to levels released to the atmosphere. By systematically imparting biases to consistently overpredict mercury transfer and accumulation, the multiplicative sequence of models introduced a substantial margin of likely overprediction. Imparting best estimates in deposition model calculations and relying on watershed-specific measurements of mercury water-to-fish bioaccumulation, while still maintaining conservative biases in mercury loading models, resulted in plausible estimates of increases of mercury levels in fish.
Modeling lead at a firing range
We constructed a food chain accumulation model to assess potential risks associated with consuming game from the vicinity of a firing range contaminated by lead shot. Spatially explicit modeling was used to account for the typical ranges of deer and turkey, assumed to obtain a fraction of their food supply from vegetation on the site. Models were used to estimate the uptake of lead into vegetation and the dietary biotransfer rate of lead to deer and turkey. Predicted levels of lead in deer and turkey were used in conjunction with high-end game consumption rates to estimate lead intake by people, which in turn were used to estimate potential increases in fetal blood lead concentrations using the U.S. EPA's Adult Lead Methodology.
Hazardous waste site models
We typically use models to supplement measured contaminant levels at hazardous waste disposal sites in order to derive appropriate exposure-point concentrations. For example, for sites under the auspices of the Massachusetts Contingency Plan, a simple distance-dependent dilution model is used in conjunction with measured contaminant levels in groundwater to estimate contaminant concentrations in surface water at the point of potential discharge to surface water. Surface water concentrations can then be compared to ecologically-based screening criteria. As another example, mass transfer models are used to estimate concentrations of vapors in open and confined excavations based on contaminant levels measured in subsurface soil, groundwater, and soil-gas to assess potential hazards to construction and utility workers who might someday work in contaminated site areas.