Publications
Conceptual representations of fracture networks and their effects on predicting groundwater transport in crystalline rocks
Understanding subsurface fracture network properties at the field scale is important for a number of environmental and economic problems, including siting of spent nuclear fuel repositories, geothermal exploration, and many others. This typically encompasses large volumes of fractured rocks with the properties inferred from the observations at rock outcrops and, if available, from the measurements in exploratory boreholes, quarries, and tunnels. These data are inherently spatially limited and a stochastic model is required to extrapolate the fracture properties over the large volumes of rocks. This study (1) describes three different methods of generating fracture networks developed for use in the fractured continuum model (FCM) and (2) provides a few examples of how these methods impact the predictions of simulated groundwater transport. A detailed analysis of the transport simulations using FCM is provided in the separate paper by the same authors (to be presented at IHLRWM 2017 conference). FCM is based on the effective continuum approaches modified to represent fractures. The permeability of discrete fractures is mapped onto a regular three-dimensional grid. The x-, y-, and z effective permeability values of a grid block are calculated from the tensor. The tensor parameters are fracture aperture, dip, strike, and number of fractures in the grid block (spacing). All three methods use the fracture properties listed above to generate corresponding permeability fields. However, the assumptions and conceptual representation of fracture network from which these properties are derived are very different. The Sequential Gaussian Simulation (SGSim) method does not require an assumption regarding the fracture shape. Fracture aperture, spacing, and orientation are defined based on the field observations. Spatially correlated features (continuation of fracture in the direction of the orientation) are created using spatially correlated random numbers generated with SGSIM code. With this method an exact number of fractures cannot be generated. The Ellipsim method assumes that the fractures are two-dimensional elliptical shapes that can be described with radius and aspect ratio. The knowledge of the fracture (ellipse) radius probability distribution is required. The fracture aperture is calculated from the ellipse radius. For this option an exact number of fractures can be generated. The fracture networks generated with SGSim and Ellipsim are not necessarily connected. The connectivity is achieved indirectly via matrix permeability that can be viewed as the permeability of much smaller fractions. The discrete fracture network (DFN) generator assumes elliptical fracture shapes and requires the same parameters as Ellipsim. The principal difference is in connectivity. The DFN method creates the fracture network connectivity via an iterative process in which not connected clusters of fractures are removed. The permeability fields were generated with FCM using three different methods and the same fracture data set loosely based on the data from an existing site in granite rocks. A few examples of transport simulations are provided to demonstrate the major findings of the comparison.