Postaudit evaluation of conceptual model uncertainty for a glacial aquifer groundwater flow and contaminant transport modelEvaluation de l’incertitude d’un modèle conceptual des écouments d’un aquifère de sédiments glaciaires et d’un modèle de transport de contaminant suite à un auditEvaluación postauditoría de la incerteza de un modelo conceptual para un modelo de flujo de agua subterránea y transporte de contaminante de un acuífero glacial冰川堆积含水层地下水流和污染物运移模拟中概念模型不确定性的后评价Avaliação a posteriori da incerteza do modelo conceptual para o fluxo de água subterrânea e do modelo de transporte de contaminantes num aquífero glacial

Numerical groundwater flow and contaminant transport modeling incorporating three alternative conceptual models was conducted in 2005 to assess remedial actions and predict contaminant concentrations in an unconfined glacial aquifer located in Milford, Michigan, USA. Three alternative conceptual models were constructed and independently calibrated to evaluate uncertainty in the geometry of an aquitard underlying the aquifer and the extent to which infiltration from two manmade surface water bodies influenced the groundwater flow field. Contaminant transport for benzene, cis-DCE, and MTBE was modeled for a 5-year period that included a 2-year history match from July 2003 to May 2005 and predictions for a 3-year period ending in July 2008. A postaudit of model performance indicates that predictions for pumping wells, which integrated the transport signal across multiple model layers, were reliable but unable to differentiate between alternative conceptual model responses. In contrast, predictions for individual monitoring wells with limited screened intervals were less consistent, but held promise for evaluating alternative hydrogeologic models. Results of this study suggest that model conceptualization can have important practical implications for the delineation of contaminant transport pathways using monitoring wells, but may exert less influence on integrated predictions for pumping wells screened over multiple numerical model layers.