Stochastic Inversion of Three?Dimensional Discrete Fracture Network Structure With Hydraulic Tomography

We introduce an approach for the stochastic characterization of geometric and hydraulic parameters a three-dimensional (3D) discrete fracture network (DFN) estimating their uncertainty based on data from tomography experiments. The inversion relies Bayesian framework resulting posterior distribution is characterized by generating samples Markov chain Monte Carlo (MCMC) methods. method evaluated four synthetic test cases related to Grimsel site in Switzerland. Comparison original reconstructed DFN models shows that presented suitable identifying variable locations orientations. This especially case those fractures represent preferential flow paths simulated It also revealed useful discriminate reliability inversion, which illustrated probability maps taken as sections through studied rock mass. Moreover, it demonstrated apertures can be calibrated together with geometries. A premise applicability practice, however, measurements are complemented additional information sufficiently constrain value ranges inverted together. work expands previously promising two-dimensional procedure transdimensional field-based 3D problems. theoretical findings here open door highly flexible structural practice tomography, well alternative or complementary tomographic