On the use of uncertainty analyses to test hypotheses regarding deterministic model predictions of environmental processes

Abstract

This paper illustrates the use of Monte Carlo parameter uncertainty and sensitivity analyses to test hypotheses regarding predictions of deterministic models of environmental transport, dose, risk and other phenomena. The methodology is illustrated by testing whether 238Pu is transferred more readily than 239 + 240 Pu from the gastrointestinal (GI) tract of cattle to their tissues (muscle, liver and blood). This illustration is based on a study wherein beef-cattle grazed for up to 1064 days on a fenced plutonium (Pu)-contaminated arid site in Area 13 near the Nevada Test Site in the United States. Periodically, cattle were sacrificed and their tissues analyzed for Pu and other radionuclides. Monte Carlo parameter uncertainty analyses were conducted using the Pu data in a deterministic model of GI-to-tissue transfer to obtain a subjective probability density function (pdf) of the ratio of predicted 238Pu over 239 + 240Pu fractional transfers. This subjective pdf was used to test, conditional on the assumed deterministic model and parameter uncertainties, the hypothesis that 238Pu was transferred more readily from the GI tract to tissues than 239 + 240Pu. Differences in fractional transfers for 238Pu and 239 + 240Pu were indicated using this methodology. Conditional sensitivity analyses of the model predictions were also conducted. These analyses indicated that Pu cattle tissue concentrations had the largest impact of any model parameter on the pdf of predicted Pu fractional transfers. Issues that arise in conducting uncertainty and sensitivity analyses of deterministic models are discussed. This research was supported by the Basic Environmental Compliance and Monitoring Program (BECAMP), US Department of Energy, Nevada Field Office, Las Vegas, Nevada.