Technischer Bericht NTB 96-01

Geosynthese Wellenberg 1996 Ergebnisse der Untersuchungsphasen I und II

This report synthesises the results of all the investigations performed at the Wellenberg site to date. The investigation programme consisted of seven deep boreholes, several reflection and refraction seismic campaigns, piezometer and shallow exploratory boreholes and a series of geological field studies.

The aim of the investigations was to obtain a sufficiently extensive geological database to allow a thorough evaluation of the site suitability. The most important questions in this respect relate to the long-term safety of a planned repository. Consequently, the concrete investigation objectives were directed primarily at fulfilling performance assessment needs. Additional objectives included demonstrating engineering feasibility and providing a reliable geological database for planning future underground investigations.

Completion of the surface-based investigations represents an important milestone for the geological site characterisation. A significant improvement on the current level of understanding of the site is possible only through the planned exploratory drift which will provide direct access to the disposal zone. The aim of the present report is to document the investigations carried out at Wellenberg to date, to present and discuss evaluations and analyses of the resulting data and to explain the conclusions reached. Considerable importance is attached in this connection to presenting in a transparent and traceable manner the synthesis methodologies used.

Geological data used directly within the scope of performance assessment consist of parameter values (usually recommended values with associated bandwidth) and the most important conceptual assumptions used to derive these values. The term "geo­dataset" (GDS) is used to denote all the geological data which are used directly for performance assessment. In most -if not all-cases, GDS parameter values are derived from model calculations. Models and model representations are therefore pivotal points in the data flow scheme for the geosynthesis. They contain, in condensed form, both information from the field investigations as well as conceptual assumptions, and thus provide a transparent link between the field data on one hand and the parameter values of the geo-dataset on the other.

The geological site model is used to describe the large-scale geometric and structural configuration of the site. This is then used as a boundary condition for all further processing steps. Determination of the spatial boundaries of rock formations is based on results of borehole investigations and on studies of surface exposures, as well as on stratigraphic and tectonic studies. The latter are drawn upon wherever spatial locations cannot be fixed by field observations alone.

The host rock formation was originally restricted to the Palfris Formation and the Vitznau Marls of the Drusberg nappe. In 1994, the host rock was extended to include the interhelvetic melange and the Tertiary shales (Globigerina Marls, Schimberg Shales) of the Axen nappe because the safety-relevant properties of these formations were judged to be equally favourable. Recent investigations in the borehole SB4a/v/s confirmed these favourable properties, thus fully justifying their inclusion. The host rock is bounded in the north by the limestones of the Drusberg nappe and in the south by the limestones and sandstones of the Axen nappe. The base of the host rock -made up of the infrahelvetic melange and the tectonic equivalent of the "Wissberg-Scholle" ­ is around 1000 m below the planned repository level. Based on the most realistic inter­pretation variant, at repository level the host rock body has a N-S extent of more than 1700 m, while the E-W extent is determined by topographie factors. Even in the most pessimistic variant it has a N-S extent of almost 1200 m, which would still be completely sufficient for construction of a repository.

No large-scale fault zones or inclusions of different rock bodies were found in the host rock which, due to their hydraulic properties (T > 10-7 m2/s or K > 10-9 m/s) and their spatial extent, could represent preferential pathways to the biosphere or to an adjacent rock unit with a higher permeability. This could mean that such inhomogeneities (Iayout-determining features) are either absent in the disposal region (preferred hypothesis) or, however, that the exploratory boreholes were not successful in locating such features (alternative hypothesis). The underground investigations planned for the next phase of site exploration will allow this uncertainty to be clarified, in particular with respect to the repository layout.

Quantification of the stress field is based on in situ measurements, which were carried out in various depth intervals in four of the exploratory boreholes. The results give a very consistent picture. The direction of the maximum horizontal principal stress, SH, was found to have a tectonically plausible azimuth of 131° ± 10°, which is almost per­pendicular to the front of the Axen nappe. The rock stress distribution for the whole investigation area (particularly within the disposal zone) was estimated using a threedimensional (distinct element) model based on the results of the borehole measurements.

Although Central Switzerland can be considered as a region with enhanced seismic activity, worldwide experience from underground construction projects and site-specific investigation results indicate that earthquake activity is not expected to represent any real risk for the long-term safety of the repository. The underpressure zone (UPZ) detected in several boreholes, and the old saline formation waters encountered in the extremely low-permeability domain of the host rock, are evidence that the numerous earthquakes from the geological past up to the present day have failed to leave behind any relevant traces (e.g. in the form of new, high-permeability water flowpaths).

Evaluation of the geological long-term scenarios relied to a large extent on the results of regional studies on a much larger scale and empirical data on the links between climate and erosion. It is expected that the long-term evolution of the site will continue to be dominated by the effects of alpine orogeny (uplift and erosion). Even taking the most unfavourable of all climatic scenarios, together with a very conservative parameter selection, the repository will not be directly affected by erosion within a period of 100'000 years. What is expected within this time period, however, is a gradual increase in the hydraulic permeability of the surrounding host rock.

All observations to date indicate that the host rock is to be considered as a fractured medium with an extremely low-permeability matrix. Water flow is associated almost exclusively with structures formed by brittle deformation and with ductile structures which have undergone brittle reactivation (overprinting). Purely ductile structures show no preferential water flow characteristics. In terms of providing a quantitative description of the internal structure of the host rock, and particularly of the network of hydraulically active small-scale structures, only a non-deterministic (statistical) model comes into question due to the otherwise required high data density.

The block model represents a typical "cube" of host rock with a side length of 500 m. lt is based essentially on a statistical and stochastic analysis of geological and hydrogeological observations. Results from the exploratory boreholes confirm that the distri­bution of hydraulically relevant structures on a scale of 500 m can be considered as location-independent and thus representative for the whole host rock. The hydraulic parameters, on the other hand, show a marked depth-dependence which is taken into account by a corresponding (depth-dependent) function for their mean values. It is assumed that the increasing rock decompaction as the surface is approached has led to opening of individual shear surfaces and microfissures and, consequently, to a relative increase in hydraulic permeability.

Combining results from previous analyses, four types of water-conducting features (WCF) are distinguished: cataclastic shear zones (type 1); thin, discrete shear zones . (type 2); calcareous marl/limestone layers with drusy veins within and outside interbedded limestone zones (type 3a/b) and joints in (argillaceous) marls (type 4). Their representation in the block model is based on a statistical description of their geometric properties (extent, frequency, orientation, internal heterogeneity) and a detailed hydraulic characterisation. Results from boreholes and subsequent modelling studies have shown that WCF type 1 (cataclastic shear zones) is by far the most significant for water flow in the host rock and the heterogeneity of transmissivity distribution (channeling) was therefore taken into consideration for large fault zones. The ratio of transmissive to non-transmissive WCFs in the boreholes was used to estimate a chan­neling factor. The greatest uncertainties in conceptualising the WCFs relate to their extent and interconnectedness. These uncertainties were addressed with parameter variations derived partly from studies of surface exposures (fault zones) and partly from correlations between neighbouring boreholes (limestone bed sequences).

Depending on the particular application, quantification of groundwater circulation was carried out on different scales. These scales correspond to two conceptual approaches which were selected for representing the hydraulic properties of the host rock. The block model provides the basis for detailed investigation of flow conditions focused on the immediate vicinity of the caverns (hectometre scale). In this case, the host rock is represented as a network of discrete water-conducting features which are approximated by planar elements (fracture network approach). In cases where a large (kilo­metre)-scale description of flow conditions is required, the fracture network approach is no longer feasible. The host rock and adjacent rock formations are therefore treated on this scale as equivalent porous media (EPM approach).

Whereas each adjacent rock unit is assigned an appropriate, constant hydraulic conductivity, a novel approach is used to describe the conductivity distribution in the host rock. This approach uses a 3D model based on a geostatistical procedure known as kriging. For each point in the host rock, the K-model provides an interpolated hydraulic conductivity value, together with the associated estimated variance. Conditional simu­lations are used to generate K-fields (realisations) which reflect the natural spatial variability of the hydraulic conductivity distribution. The input parameters for the K­-model (and the boundary conditions for the K-field realisations) are derived from the WCF transmissivities determined in the boreholes. In the upper section of the host rock these vary between 10-5 m2/s and 10-8 m2/s and in the lower section between 10-9 m2/s and 10-12 m2/s. The block model plays a central role in the conversion of measured transmissivities into effective conductivities (T→K conversion) and, conversely, in the derivation of WCF transmissivities from the location-dependent hydraulic conductivities in a K-field (K→T conversion). The fact that the database used is common to both procedures ensures internal consistency of the model chain.

The hydraulic heads in the adjacent rock units represent important boundary conditions for the regional-scale hydrodynamic model calculations. This is particularly true of the saturation conditions in the limestone formations of the Drusberg and Axen nappes. Observations in springs and in piezometer boreholes indicate that, in contrast with the host rock, these limestones are not fully saturated. The greatest element of uncertainty, which was taken into account by parameter variations in the hydrodynamic modelling, relates to the karst water level in the Axen nappe.

The hydraulic heads in the host rock range from hydrostatic to artesian in the upper section to very low values immediately below the level of the Engelberg valley. The latter clearly lie below the local and regional exfiltration areas and almost reach sealevel values in boreholes SB1 and SB2. The consistent form of head profiles indicates the presence of an underpressure zone (UPZ) in the central part of the host rock. Special modelling studies have shown that a plausible (and the most conservative) hypothesis for its generation is mechanical unloading of the very low-permeability rock body following the glacial retreat at the end of the last ice age. The UPZ is thus a non­stationary dissipating phenomenon which is taken into account in the hydrodynamic modelling by transient calculations. Initial conditions for modelling are its present form and the expected (conservatively estimated) dissipation rates.

Three different scales were selected for the numerical modelling of groundwater flow conditions, tailored to the different questions to be investigated. Correspondingly, three different but inter-compatible models were also developed. The regional model is used primarily to investigate interactions between the host rock and adjacent rock formations and the influence of the UPZ on the large-scale flow regime. One important result of the modelling calculations is that, as long as the underpressure zone exists (Le. for at least 20'000 years), exfiltration from the repository zone into the biosphere is practically impossible. Once steady-state flow conditions have been reached, the valley of the Engelberger Aa is the only possible exfiltration area. Even under extreme boundary conditions there will be no exfiltration into the slopes of Altzellen or the valley of the Secklis Bach.

The repository model is used primarily to investigate the effects of the planned underground structures on the natural flow-field, such as extent and duration of head perturbation resulting from repository construction and operation, impact of the excavation disturbed zone surrounding the tunnels and caverns, hydraulic effect of seals, etc. The components of an example repository facility are explicitly modelied. In the repository model, with advanced dissipation of the UPZ the topographically determined ground­water divide beneath the Eggeligrat appears as a marked ridge of potential that runs through the centre of the repository. The large-scale flow pattern is hardly affected by the underground structures or by the temporary pressure sink caused by their presence. Drainage of the caverns via the tunnels is effectively prevented by a planned system of seal zones. The hydraulic gradients in the caverns remain very low (< 0.1 m/m) and, even after 20'000 years, the calculated flow rates are an average of onlyapprox. 1 m3/a per cavern.

Unlike the other two models, the cavern-scale model is based on a fracture network approach. It is used to investigate the distribution of groundwater flow among the different water-conducting features in the rock immediately surrounding the disposal caverns. Compared with the results of previous studies, the simulations show a higher number of expected structures per tunnel metre. This is because all water-conducting features are taken into consideration and not only the hydraulically active ones. Under quasi-stationary flow conditions, the total water flux through the repository, on the other hand, is around one order of magnitude lower than estimated in previous studies.

Information on groundwater age (residence time) and flowpaths is derived from the hydrochemical and isotopic characteristics of groundwater sampies collected from the exploratory boreholes. The groundwaters and the extracted porewaters show a depth zonation which is consistent with hydrogeological modelling results, with recent Ca-HCO3 waters near the surface, Na-HCO3 waters at least forty to several thousand years old below this and, from depths of approximately 400 m below ground level, significantly older Na-CI waters that still contain large neoalpine (several million year old) components. Waters with variable mixing ratios of the two end-members Na-HCO3 and Na-CI may occur in the repository region and in the disposal zone. Hydrochemical results were used inter alia to check the plausibility of some key modelling assumptions (K distribution, porosity, storage coefficient). The comparison has shown that the assumptions and host rock parameters used for the hydrodynamic modelling calculations are consistent with hydrochemical observations. A consistency check using other independent evidence (bounding conditions from the preferred UPZ evolution hypothesis) also gave positive results.

After evaluating all available geological site data, the results presented in previous reports can largely be confirmed and refined. Improvements in the database and further developments of modelling and synthesis methodologies have made it possible in some areas to replace overconservative assumptions with more realistic ones. Taken overall, the result is a complete picture of the site which is both internally consistent and plausible.