Fractures in rock masses strongly influence the underground rock mass mechanical and hydrogeological behavior. However, how these fracture systems spatial structure, and its associated geometrical properties, really contributes to control the rock properties is still an open issue. In such context, it is also crucial to characterize the fracturing properties by using sound DFN models and accurately defined mean estimates. Fracturing properties display multi-scale ranges of fracture sizes together with sharp variations of densities and orientation organization, which prevents any simple characterization. In this paper we focus on the notion of variability: how it can be assessed, what is its importance and how it is integrated in the DFN modeling, from local scale to the largest site scale. The DFN characterization is only based on depth core logging data and the DFN models relate to densities and orientation distributions. We describe a method recently developed, called SFD (Statistical Fracture Domain). It is used to first appraise the uncertainty of density estimates due to spatial variability then to quantify and compute differences between any number of dataset density estimates and finally to group datasets into classes of compatible statistical properties. The method is applied to some data from the SKB Forsmark site.
Fracture systems are ubiquitous in crystalline rock masses. They are a key component for the understanding of their mechanical and hydrogeological characteristics. They also are complex structures which involve both fracture size distributions spread over several orders of magnitude and sharp spatial variations of fracture orientation distri-bution and densities. This makes any detailed geometrical characterization and sound DFN (Discrete Fracture Network) modeling a difficult task. The variability internal to any given domain – the local scale - often arises simply from the random and Poissonian processes of the stochastic modeling.