PATHWAY is a dynamic simulation model of the transport of radionuclides from fallout through the agricultural foodchain. The model was developed to fulfill the need for a realistic, testable, completely time dependent model to estimate radionuclide ingestion by humans exposed to fallout patterns originating at the Nevada Test Site during the 1950's. The model simulates the fate of radionuclides in various management units (pastures, gardens, croplands, etc.) of an agro-ecosystem. The agricultural products currently considered in the model are pasture grass, alfalfa, grains, garden vegetables, milk, eggs, beef, and poultry. Harvested and stored products, such as hay and canned vegetables, are also represented in the model. Animal diets are specified on a daily basis. Thus, the model can be used to simulate the fate of radionuclides from both acute and chronic fallout deposition. The utility and credibility of radionuclide transport simulation models are severely limited without rigorous testing of their predictive capabilities against real observations. This paper reviews exercises testing the overall accuracy and temporal dynamics simulated by the PATHWAY model against 37 sets of observed data. These data include concentrations of 137Cs, 90Sr, 140Ba, and 131I in milk, beef, alfalfa, and pasture grass from several western states following deposition of the radionuclides in fallout. In addition, the model is tested against observations of concentrations of 131I in milk following the accident at the Windscale reactor in England. The statistical tests used to compare the predictions of PATHWAY to the observations include a correlation analysis, a paired t-test, and a binomial test. We use the correlation coefficient between observations and predictions through time to compare the dynamics of the simulated and real world system. Plots of the residuals from regression are then examined for bias between the predictions and observations. The significance of any trends in the residuals is evaluated using a runs test. The paired t-test and the binomial test are used to evaluate the accuracy of PATHWAY's predictions. The hypothesis for the paired t-test is that the ratio of predictions to observations is 1. The paired t-test can be used to test hypotheses about ratios because the distributions of observations and predictions appear to be lognormal. However, the paired t-test does not consider uncertainty in the predictions of the model. We use a binomial test to compare the observed data to an interval estimate from PATHWAY. The interval corresponds to a 95% confidence interval on the prediction, and is derived from uncertainty analyses that have been conducted on PATHWAY. PATHWAY's predictions are significantly correlated with observed levels of 137Cs and 90Sr in pasture and alfalfa. PATHWAY also simulates the dynamics of 131I, 140Ba, and 137Cs in milk well, but fails to predict what appears to be a long term accumulation of 90Sr in the agro-ecosystem. PATHWAY predicts the absolute concentrations of 131I in milk quite well, but tends to predict levels of 140Ba, 90Sr, 137Cs in milk that are different from those observed by factors of 2 to 7. PATHWAY predicts levels of 137Cs and 90Sr in pasture and beef within a factor of 2 of those observed.
Read full abstract