Background and objective
Understanding the in-situ stress field in the Forsmark tectonic lens will lay the fundamental basis to characterise the prevailing mechanical and hydraulic behaviour of the rock mass. Only with a good understanding of the geomechanical setting including the stress field an analysis of future rock mass behaviour under the evolution of conditions in a repository for spent nuclear fuel is possible. Based on the existing stress measurements several stress models have been published. The stress model frequently used by SKB (Martin, 2007), mostly ignores hydraulic measurements, and proposes a reverse faulting regime. An alternative stress model (Ask et al. 2007) heavily relies on the hydraulic data and suggests a strike slip stress model with smaller stress magnitudes. A recent review by Gipper et al. (2015), incorporating most measurements, suggests a model in the hybrid regime. In order to get a better understanding of the stress situation of the Forsmark area and to gain more confidence in the analyses, a comprehensive stress modelling and simulation study was performed on the Forsmark lens taking into account existing stress measurements and large brittle structures.
The numerical model developed is the result of an integrated study based on the elaboration of geomechanical concepts applied to a structurally complex rock mass. The input data for the model included the geometrical elements that characterise the region of interest at the km scale (i.e. the fault network), measurement data collected over various campaigns for the determination of the in-situ state of stress, experimentally derived values for the main geomechanical properties of the rock mass as well as the interpretations of said measured and derived parameters formulated at different stages of the site characterisation process.
The complexity of the structural elements and the high variability of many of the input parameters and data collected in the area generated uncertainties in the resulting numerical model. While it is difficult to quantify the accumulated uncertainty in the material parameters, generated by all the mentioned uncertainties, it is reasonable to suggest that the overall heterogeneity of the rock volume could be at least partially underestimated. As mentioned, local (at times considerable) heterogeneities are recognised within and around deformation zones; given the complexity of the tectonometamorphic evolution of the area, as well as the relative lithological variations within the region of interest, further sources of uncertainty could have been overlooked and not properly sampled.
As a result, the proposed model is characterised by an orthotropic material behaviour for the rock volume, where the increase of the Young’s modulus with depth is different in the directions parallel and perpendicular to the major principal stress. While the testing conducted on intact rock core samples does not support this hypothesis, strongly converging SH and Sh gradients with depth as proposed by Martin (2007), cannot be reproduced for a completely isotropic body. The choice is further justified by the mentioned sampling bias, which could have overlooked the effect of preferentially oriented sets of fractures and other structures on the overall material behaviour of the rock volume as a whole.
However, in the absence of better quality (and less variable) stress measurements, our modelling cannot reduce the risk that none of the originally proposed models for the Forsmark site (neither Martin, 2007 nor Ask et al., 2007) is in fact a realistic representation of the actual in-situ stress in the area.
Discrete faults seem to play a minor role in the current stress field in contrast to the spatial and heterogeneous distribution of rock properties. Research should be conducted on the spatial distribution and the range of rock properties so that calibrated distribution functions can be derived to populate the subsurface not with constant rock property values but with distribution functions that are based on spatial information or calibration. This might better address the uncertainties in the measured stresses than trying to approximate homogenized gradients not accurately describing the scatter of measured values. The performed modeling highlights the influence of elastic parameters on the stress field and shows how an assumption of a heterogeneous Young’s modulus and Poisson’s ratio of ±10% could describe the spread in the range of the most-likely stress field.
Hence, upcoming work could improve the capabilities of the model by deriving spatial distribution function of elastic properties that are calibrated against the stress field measurements.