Building a staircase of confidence in groundwater modeling: a summary of ten years data collection and model development

Building reliable models is an iterative process whereby uncertainties in the initial model are reduced progressively through data collection and sensitivity and uncertainly analysis. In the north of Belgium such an iterative research cycle was started some ten years ago for a surface repository for low-level radioactive waste. Within this period, two site characterizations and two groundwater models were preformed aiming at reducing uncertainty in the groundwater flow solution (i.e. heads) and contaminant transport solution (i.e. concentrations) in a systematic way guided by model sensitivity analyses. One-percent sensitivity maps identified the areas which would benefit most from additional head observations. The transport model prediction scaled sensitivities identified the parameters with the largest leverage. The sensitivity and uncertainty analyses informed the next data collection. Analysis of the data resulted in a refined stratigraphy for the entire model domain and improved the concept of a semi-permeable aquitard that separates the two major aquifers. With an improved 3D model structure and more reliable hydraulic properties, the updated model was shown to perform better, had less conceptual uncertainty about the aquitard and represents an arguably better tool to predict contaminant transport.