The Gas Migration Test (GMT) at the Grimsel Test Site underground laboratory in central Switzerland was designed to investigate gas migration through an engineered barrier system (EBS). The EBS consists of a concrete silo embedded in a sand/bentonite buffer emplaced in a silo cavern that intersects a shear zone in the surrounding granite host rock. The experiment was performed in a series of stages: (a) excavation of the access drift and silo cavern, (b) construction and instrumentation, (c) saturation of the EBS, (d) water tests, (e) long-term gas injection at different rates, (f) postgas water testing, (g) gas injection with a "cocktail" of gas tracers, and (h) depressurization and dismantling. A numerical model was developed for the design and analysis of the different stages and to describe the relevant phenomena associated with gas migration from a potential repository for transuranic waste.

A numerical model of the GMT was implemented with the two-phase-flow code TOUGH2, representing the GMT silo with a multilayered radially symmetric mesh and the surrounding water-conducting granite shear zone with a two-dimensional vertical feature. The different stages of the experiment were simulated in sequence using the results of the previous stage as initial conditions for the subsequent stage. Two-phase-flow parameters for the EBS were derived from laboratory experiments on core samples of the different materials that comprise the EBS, while hydraulic properties of the sand/bentonite and of relevant interface zones were calibrated to the pressure responses in the silo and selected piezometers in the sand/bentonite. The results of the numerical modeling of the GMT experiment show that the main features and processes of the different stages of the experiment could be reasonably well reproduced. Following the initial calibration of effective properties from the water test response during stage 4, property changes during the subsequent test phase were calibrated as stress- or pressure-dependent permeability changes. During the gas injection phases, the pressure-dependent permeability change could be related to the minimum effective stresses along interfaces. The inferred coupled hydromechanical phenomena were implemented using pressure-dependent permeability relationships on interfaces at the top of the silo and between the sand/bentonite and the granite host rock. During the recovery sequence following the first injection gas phase, the calibrated decrease in permeability of the sand/bentonite above the silo was related to the pressure decline in the upper cavern, but there was no apparent stress change. However, the calibrated permeability reduction in the sand/bentonite was in the range of values measured during the EBS excavation. In addition, time-dependent permeability relationships were calibrated for the tunnel seal to account for the gradual decrease in water inflow from the upper cavern into the access tunnel and the drift.