Stochastic analysis of fluid flow and tracer pathways in crystalline fracture networks

Abstract: Understanding groundwater flow systems and how these control transport is an essential part in assessing the suitability of subsurface environments as hosts for storage of toxic waste. Therefore it is important to be able to integrate knowledge obtained from field characterisation of the subsurface with methods which can be used to evaluate and predict possible impact on surrounding environments.In this thesis I investigate the characteristics of flow and transport in discrete fracture networks by analysing Eulerian and Lagrangian descriptions within a stochastic framework. The analysis is conducted through numerical flow and transport simulations configured according to available field data, combined with independent theoretical analytic and semi-analytic methods which are able to reveal insight to relevant constitutive properties. It is shown that numerical simulations conducted with the discrete fracture network approach can be both conditioned and confirmed against field measurable quantities, and the developed theoretical methods are evaluated against results obtained from simulation. Thereby, a methodology which can provide links between field measurable quantities and tracer discharge is presented, developed and evaluated. It is shown to be robust with respect to underlying assumptions used for flow configurations.In particular, a specific sampling algorithm for obtaining a Lagrangian description of transport based on a Eulerian description of flow is proposed, evaluated and shown to be robust for the cases considered, providing accurate replications. Also a generalisation of both the advection-dispersion solution and the one-sided stable distribution is shown to be able to evaluate advective transport quantities, and combined with a Lagrangian retention model it is shown to be a fairly accurate and robust method for upscaling distributions, enabling predictions of transport in terms of tracer discharge. Evaluation of transport is also conducted against the advective-dispersion assumption, where results indicate advective transport is generally non-Fickian for the fracture networks and domain scales considered, but not necessarily anomalous. Additionally, the impact certain model assumptions have on tracer discharge are analysed. For example, transport is evaluated for assumptions regarding injection mode, fracture network heterogeneity, relationship between aperture and transmissivity, relationship between transmissivity and size, as well as scale and modelling dimension. In relation to hydraulic testing and flow analysis, a method for conditioning fracture transmissivity from field measurements of flow by simulation is developed and evaluated against homogenisation assumptions commonly used in field applications. Results indicate the homogenisation assumption generally fails for current interpretations of field data.