Channel Network Modeling of Flow and Transport in Fractured Rock at the Äspö HRL: Data-Worth Analysis for Model Development, Calibration and Prediction

Performance assessment of nuclear waste disposal in deep crystalline bedrock demands a thorough understanding of the related flow and transport processes. Uncertainties may arise both from the selection of the conceptual model as well as the estimation of the related model parameters. Discrete fracture network (DFN) models are widely used for such modeling while channel network models (CNM) provide an alternative representation, the latter focusing on the fact that flow and transport in deep fractured media often are dominated by a small number of long preferential flow paths. This study applies the principle of channel networks, implemented in the Pychan3d simulator, to analyze the hydraulic and tracer transport behavior in a 450-m-deep fractured granite system at the Äspö Hard Rock Laboratory in Sweden, where extensive site characterization data, including hydraulic and tracer test data are available. Semi-automated calibration of channel conductances to field characterization data (flow rates, drawdowns, and tracer recoveries) is performed using PEST algorithm. It was observed that an optimal CNM connectivity map for channel conductance calibration can only be developed by jointly fitting flow rates, drawdowns and tracer mass recovery values. Results from data-calibrated CNM when compared to a corresponding calibrated DFN model shows that the CNM calibrates and adapts better than a DFN model with uniform fracture surfaces. This comparative study shows the differences and uncertainties between two models as well as examines the implications of using them for long term model predictions.