Finding Balance Between Quantifying Structural and Parametric Uncertainty in Groundwater Modelling Studies

ABSTRACT: Numerical models and their predictions inherently contain many uncertainties due to the assumptions that are made in conceptualizing (e.g., primary hydrogeologic and transport processes), constructing (e.g., assigning parameter values such as hydraulic conductivity), and calibrating a model (e.g., interpreting conditions between sparse field data with imperfect measurements). These uncertainties influence model predictions, such as the capture zone of a pumping well or the predicted contaminant concentration at a receptor, which in turn can hamper the reliability of predictions. Uncertainty quantification is necessary to understanding the potential variability of predictions arising from numerical groundwater flow and contaminant transport models. These uncertainties have been classed into two types: structural and parametric and uncertainties. Structural uncertainty includes the continuity of aquifer and aquitard units between observation boreholes and the simulated flow and transport processes that is consistent with observation data. Parametric uncertainty refers to the range of parameter values, for the given conceptual model, that are consistent with the observation data. Ideally, uncertainty quantification undertaken in groundwater flow and transport modelling studies should balance the evaluation of both structural and parametric uncertainties on model predictions. While parameter estimation tools such as PEST are ideal for evaluating parametric uncertainty, evaluation of structural uncertainties (e.g., aquitard continuity) cannot be automated. This paper presents findings from case studies where parametric and structural uncertainty analyses were applied. It highlights the approaches undertaken to evaluate uncertainties, and the relative impact of the uncertainties on model predictions.