Limiting the propagation of error in probabilistic risk assessment modeling methylmercury as an example

摘要

Probabilistic risk assessment has as its goal the delineation of the distribution of risks likely to result from process or action. Such distributions not only display the expected value of risk (i.e., the most probable outcome given a model, a set of conditions, and the distributions of input parameters), but also the expected range of outcomes. In this paper, we compare risks estimated by using the 95th percentiles of each parameter. In general, we find that the frequency distribution of model outputs is controlled by the relative width of the distribution of the most widely dispersed of the inputs and that adding additional parameters that are less widely dispersed has little or no effect on the relative dispersion of outcomes. We use the health risks of methylmercury (MeHg) resulting from coal combustion as a case in point.