Validation of hydraulic models in fractured rocks. Application to the Forsmark site (Sweden)

Philippe Davy (1) , Romain Le Goc (2) , Caroline Darcel (2) , Jan-Olof Selroos (3) , Diane Doolaeghe (2)

1 GR - Géosciences Rennes
2 Itasca Consultants
3 Swedish Nuclear Fuel and Waste Management Company

The validation of hydrogeological models is a multi-dimensional problem including theoretical, data-related, and numerical aspects. The exercise is particularly challenging for highly heterogeneous media such as fractured rocks, where the difficulty of acquiring data is compounded by a high spatial variability that raises the question of both the representativeness of the data and the complexity of models. We discuss a validation method for fractured rocks that is based on a sensitivity analysis of models, a calibration with data, and the rejection of models unable to reproduce fundamental flow characteristics derived from observations. The latter step is dependent on the type of predictions that the models are supposed to achieve. The exercise is performed to assess the risk of storing toxic wastes in a fractured underground. The site characterisation consists of a series of 20 1000m long cored wells with a 1m resolution flow logging. The statistical distribution of fractures in terms of sizes, orientations and intensities has been assessed from outcrops and well mapping. The models are required to reproduce both the flow intensity and the basic elements of the flow structure (e.g., channelling and scaling). A basic analysis of data shows that the connectivity of fracture networks is scale dependent mostly due to the wide range of fracture sizes and to the large amount of fracture clogging (up to 80% of the total surface). The flow paths are close to the percolation threshold for scales of interest for the models (1-100 m), which eliminates models that rely on property homogenization. The validation process is based on the DFN – Discrete Fracture Network – approach, which combines a geometrical fracture model of the fracture network and its transmissivity structure. Each model relies on a set of geometrical and hydraulic assumptions. The sensitivity analysis, calibration and rejection process emphasizes the importance of (i) the geometrical fracture model in particular the size distribution, (ii) the correlation between transmissivity T, orientation (or the stress dependency of T) and fracture size, (iii) the clogging distribution model for the most critical parameters.

Domaines: Hydrologie

Philippe Davy, Romain Le Goc, Caroline Darcel, Jan-Olof Selroos, Diane Doolaeghe. Validation of hydraulic models in fractured rocks. Application to the Forsmark site (Sweden). American Geophysical Union Fall Meeting 2019, Dec 2019, San Francisco, United States. , pp.H41H-1760, 2019. ⟨insu-02402624⟩