Abstract. Pore pressures below hydrostatic (underpressures) have been measured at the Wellenberg site in Switzerland, that Nagra (Swiss National Cooperative for the Disposal of Radioactive Waste) has selected as a potential repository for short-lived radioactive waste. The rock consists of overconsolidated Cretaceous marls and Tertiary shales, characterized by low to very low hydraulic conductivity. The maximum measured underpressures are between 4.8 and 7.1 MPa below hydrostatic and occur at depths of-800 m below ground. Results from this investigation show that mechanical unloading is capable of producing the observed underpressures. Unloading can be produced in conjunction with denudation (uniform erosion), differential erosion, and glaciation. The effect of these processes, of relevant material parameters, and of boundary conditions are investigated in two phases. In the first phase, simple one-dimensional and twodimensional "column" models are used under uniform erosion (denudation) conditions. These investigations show that within the range of the investigated parameters and physical processes, the hydraulic conductivity, the stiffness, and the drainage boundary conditions are more important than erosion rate, horizontal stresses, and nonlinear material behavior. However, although underpressures are produced with parameters and conditions corresponding reasonably well to observations, uniform rates of denudation alone cannot account for the actually encountered magnitude of underpressure. The second phase in which differential erosion and, particularly, the effect of glaciation are considered produces very realistic results. This is established by examining a number of combinations of denudation, differential erosion, and glacial loading and unloading.

No abstract

The Wellenberg K-model describes the distribution of the hydraulic conductivity of the host rock at Nagra's proposed site for a LLW/ILW repository. The description takes the form of the expected value of hydraulic conductivity at a location and the uncertainty around this expected value. The development of the K-model is based on geostatistical concepts and, in particular, on kriging and conditional simulation methods. The model is built of cubes with sides of 100 m. Upscaling of the transmissivity profiles at the borehole locations to effective K-values on the 100 m scale is realised by fracture network modelling. The K-model accounts explicitly for the hydraulic conductivities estimated at each borehole. It is consistent with the hydrogeological conceptualization of the host rock and generates realistic hydraulic conductivity distributions with a degree of variability similar to that observed in the data.

Wellenberg is the proposed candidate site for LLW/ILW repository in Switzerland (Figure 1). The performance assessment of the planned repository calls for a thorough hydrogeological characterization of the site. The spatial distribution of the hydraulic conductivity within the host rock is a key input for the evaluation of the host rock performances. The spatially variable conductivity field is modelled geostatistically in 3 dimensions using a conditional simulation method. In order to account for the possible range of spatial variability, several simulations of the conductivity are generated. The conditional simulation of the conductivity - representing possible versions of the unknown reality - are used as input for numerical modelling. For each simulated conductivity field, the groundwater flow is numerically modelled using finite elements. This Monte Carlo approach allows the propagation of the input uncertainty due to spatial variability of the conductivity onto the predictions delivered by the flow model. With the help of this approach several specified performance criteria - together with their uncertainty - were estimated for the evaluation of storage capabilities of the Wellenberg host rock.

This paper gives an overview of the numerical models employed for the simulation of ground-water flow at the Wellenberg site, the runs carried out and the kind of products which were specified as results of the modelling calculations. With respect to more conventional modelling approaches, based mainly on fully deterministic and steady-state schemes, this new approach integrates both deterministic and stochastic aspects (hydraulic conductivity distribution, geometry of the fracture network, etc.) in three different groundwater models, accounting for different scale levels. Moreover, this new approach is as close as possible to a realistic representation of the in situ conditions, since it is based on plausible distributions of transmissivity and hydraulic conductivity at all scales and takes the transient nature of flow fully into account. This latter aspect is fundamental since, at Wellenberg, both the underground facilities and a natural head anomaly in the host rock cause pressure disturbance in the flow field. Finally, and because boundary conditions (especially topographic surface and overburden) vary over time, model predictions based on current topography and climatic conditions are of limited duration and hence a validity duration has to be defined.