Modeling fracture flow with a stochastic discrete fracture network: calibration and validation: 1. The flow model

TLDR: In this paper, a large-scale investigation of fracture flow was conducted in a granite uranium mine at Fanay-Augeres, France, and four types of data were collected: (1) geometry of the fracture network; (2) local hydraulic properties measured by injection tests in boreholes; (3) global hydraulic behavior from flow rate and piezometric head distribution at a 106 m3 scale; and (4) tracer tests performed at a scale of up to 40 m.

Abstract: A large-scale investigation of fracture flow was recently conducted in a granite uranium mine at Fanay-Augeres, France. Its aim was to develop a methodology for the investigation of possible nuclear waste repository sites in crystalline environments, and thus to determine what measurements to make and what models to use in order to predict the flow and transport properties of the medium, i.e., their average behaviors and spatial variabilities at different scales. Four types of data were collected: (1) geometry of the fracture network; (2) local hydraulic properties measured by injection tests in boreholes; (3) global hydraulic behavior from flow rate and piezometric head distribution at a 106 m3 scale; and (4) tracer tests performed at a scale of up to 40 m. A stochastic fracture network model assuming negligible matrix permeability was developed and calibrated essentially on data 1 and 2 above; this was then used to predict data 3 and 4 in an attempt to validate both the parameters and the structure of the model. In this first part, only the flow problem (data 1) is discussed.