RESUMO
A primary component to human health risk assessments required by the U.S. Environmental Protection Agency in the registration of pesticides is an estimation of concentrations in surface drinking water predicted by environmental models. The assumptions used in the current regulatory modeling approach are designed to be "conservative", resulting in higher predicted pesticide concentrations than would actually occur in the environment. This paper compiles previously reported modeling and monitoring comparisons and shows that current regulatory modeling methods result in predictions that universally exceed observed concentrations from the upper end of their distributions. In 50% of the modeling/monitoring comparisons, model predictions were more than 229 times greater than the observations, while, in 25% of the comparisons, model predictions were more than 4500 times greater than the observations. The causes for these overpredictions are identified, followed by suggestions for alternative modeling approaches that would result in predictions of pesticide concentrations closer to those observed.