ABSTRACT

This study develops and evaluates variable statistical approaches for propagating conceptual and parametric uncertainty from regional-scale to local-scale groundwater models. Regional models are commonly used to determine boundary conditions (e.g., heads) and calibration targets (e.g., fluxes) along the boundaries of a higher resolution local-scale model developed to focus on a particular area of interest. For illustrative purposes, a three-dimensional hypothetical domain is used as a reference model for evaluating the developed tools. Five Conceptual models with a total of 1500 realizations are developed for the regional scale model. Boundary conditions (e.g., heads) and calibration targets (e.g., fluxes) are mapped from the regional models over to the local models. Different statistical approaches for mapping the regional data over the local models are tested including using all realizations with and without incorporating the weights reflecting the regional model calibration goodness of fit results. In addition, averaging the results of the regional models prior to mapping the data to the local scale model is also considered as an alternative tool. The overarching conclusion is that it is better to incorporate as many conceptually different models as possible, with small number of realizations in each model, than to rely on a single conceptual model no matter how large the number of realizations used for that single model.