Environmental Modeling and Uncertainty

Abstract: Environmental fate models are used to evaluate the fate and effects of chemicals for risk assessment. Fate models may be effective and low-cost substitutes for field measurements and are helpful to project future scenarios. Environmental models describe processes for chemical fate largely determined by environmental and chemical-specific parameters. There is uncertainty in such input parameters arising from lack of knowledge and inherent variability in environmental processes.     The objectives of this thesis are to demonstrate and evaluate ways to quantify, implement, and reduce uncertainty in chemical-specific input parameters and  in the process to improve the overall treatment of uncertainty in environmental modeling. The methods to treat uncertainty were combinations of multimedia environmental modeling, the use of testing and non-testing information in risk assessment, and probabilistic and non-probabilistic measures of uncertainty.    This thesis contains six case studies related to chemical safety assessment which illustrate different aspects of treatment of uncertainty in a regulatory context. Dependent on nature of uncertainty and the available information, the approaches to treat uncertainty were probabilistic, non-probabilistic or combinations of these. Some case studies were put into the perspective to support chemical regulation under REACH. In three studies, the contribution of uncertainty in input parameters was evaluated on characteristics of uncertainty in assessed persistence, long-range transport potential and comparative toxicity potentials of chemicals in the environment. In other studies, the focus was on decision making such as prioritization of chemicals for risk assessment and the need for further testing to reduce input uncertainty.     The main contributions are useful applications of a broader treatment of uncertainty in environmental modeling that address gaps and quality in available background knowledge.  Epistemic uncertainty is treated by filling knowledge gaps using non-testing information from QSARs. Uncertainty in non-testing information is given a probabilistic treatment based on statistical principles of inference. Poor quality of background knowledge, such as sparse data or low confidence in individual QSAR predictions, is treated by non-probabilistic measures. Finally, the suggested treatments of uncertainty are implemented and evaluated in the context of chemical risk assessment.