Collection, Use, and Impact of Site-Specific Distributional Exposure Duration Data As an Alternative to Default Assumptions. R. L. Sielken Jr., Sielken, Inc., 3833 Texas Avenue, Suite 230, Bryan, TX 77802; and B. A. Trenary, Dow Environmental Inc., 10940 NE 33rd Place, Suite 210, Bellevue, WA 98004
Several innovative uses of Monte Carlo and other probabilistic risk analysis techniques emerged during the development of an integrated endangerment assessment/risk characterization for the Rocky Mountain Arsenal near Denver, Colorado. The Monte Carlo methods used to determine site-specific distributional characterizations of exposure durations in different subpopulations and populations from probability distributions describing the relative frequencies of the different subpopulations within different populations of visitors to the site, the number of years of site visitation, the types of visitation activities, the number of days per year of participation in each activity, the frequency of different combinations of multiple activities during a site visit, and the number of hours spent on the site participating in a specified activity are indicated. The implementation of these methods can be greatly facilitated by a new Monte Carlo simulation software system called DistGEN (probability Distribution GENerator). Several ways to find and qualify sources of such site-specific data and how to estimate component distributions based on that data are also discussed. The use of probabilistic exposure analysis for public health risk assessments at the Rocky Mountain Arsenal provides many important lessons including the following: Probability distributions rather than default constants should be used to characterize exposure parameter values for subpopulations and populations in order to estimate the relative likelihood of different exposures and the variability of exposure from individual to individual. The estimated distributions of exposure parameter values should correspond to real populations and subpopulations rather than hypothetical subpopulations of maximally exposed individuals. Exposures should be summed over exposure days rather than assuming that the exposure is the same value on every exposure day. Bounds on exposure should not be determined by simply evaluating an exposure equation or model with each exposure parameter's distribution replaced by a bounding constant. Separate distributional characterizations should be developed and presented for each subpopulation and combined population; so that, the differences between subpopulations and populations are preserved. Furthermore, numerical examples indicate that there can be huge quantitative differences in the exposure characterization between using constants and probability distributions, between the 95th percentile of exposure and the bound on exposure determined by using each parameter at its 95th percentile, between assuming the same exposure duration on every day in which some exposure occurs and summing independent exposure durations over all days in which some exposure occurs, and between subpopulations and populations.