USING OBJECTIVE AND SUBJECTIVE INFORMATION TO DEVELOP DISTRIBUTIONS FOR PROBABILISTIC EXPOSURE ASSESSMENT

The propagation of variance through complex exposure models has been simplified by simulation software running on desktop computers; however, the appropriate representation of variability in, and uncertainty about, model inputs remains a challenge. The U.S. Environmental Protection Agency (EPA) has indicated that the Monte Carlo simulation approach to exposure assessment is acceptable as long as input distributions are credible, i.e., rooted in fact (EPA, 1992). The use of disrributions based on empirical evidence is certainly desirable; however, it is not always possible to obtain measurements for all exposure model inputs under the conditions of interest. Where objective evidence is scarce or unavailable, subjective judgments are necessary. In this work, the application of both statistical and subjective approaches to the development of distributions for exposure model inputs is explored. The consequences of one's approach to distribution development, on the interpretation and applicability of the output, are also considered.