Fractures in low-permeability rock play a critical role in the contamination of groundwater resources by providing conduits for the transport of toxic contaminants. The development of conceptual models to describe the hydrogeology of sparsely fractured systems requires the characterization of the physical properties of discrete fractures at the field scale. A tracer experiment was conducted at a site instrumented with 27 boreholes within an 35 x 40 m area. Inflatable packers were used to isolate a horizontal fracture 10.5 m below ground surface in a shale-limestone bedrock. The results of hydraulic tests indicate that the fracture has a spatially variable aperture with a mean of 126 m and standard deviation of 95 m. The tracer experiment was initiated by injecting a radial disk of conservative tracer into the fracture plane. The appearance of tracer down-gradient was monitored using a sampling packer which minimizes the volume of groundwater withdrawn. Tracer breakthrough was observed in 8 boreholes at distances ranging from 11 to 41 m from the source borehole. The experiment was interpreted using a two-dimensional finite element transport model (LTGPLAN) which incorporates diffusion into the rock matrix, longitudinal and transverse dispersion and constant or spatially variable fracture aperture in the plane of the fracture. Results show that breakthrough curves interpreted using a constant aperture require large values of transverse dispersivity to account for deviations in groundwater velocity and the mean flow direction. At this scale there is no evidence for increased longitudinal dispersivity with distance. Simulations were also conducted using spatially correlated aperture distributions having the mean and standard deviation of the natural fracture.
Fluid flow in discrete fractures has traditionally been described using the parallel plate model where flow velocity is proportional to the square of the fracture aperture. Theoretical and experimental studies of both artificial and natural rock fractures at various scales and levels of normal stress have shown that the parallel plate approximation is sometimes inadequate in representing flow through natural fractures (eg. Tsang, 1984; Raven and Gale, 1985). Physical factors which lead to departure from the parallel-plate model include the roughness of the fracture surfaces, tortuosity of the flow paths and the presence of areas where the fracture surfaces are in contact and closed to flow. In addition to the advective and dispersive processes within the fracture plane, solute transport is also influenced by diffusion into the surrounding unfractured rock (Novakowski and Lapcevic, 1994). In the development of alternative conceptual models, it is generally assumed that fracture aperture can be characterized by a statistical distribution (normal, lognormal, gamma or power) and that the aperture is correlated spatially with the degree of correlation related to the scale of measurement (eg. Brown et al., 1986; Piggott, 1990; Brown, 1995). However, determining the spatial distribution and correlation of fracture aperture at the field scale (tens to hundreds of metres), is difficult if not impossible, due to the high cost required to obtain the data necessary for meaningful geostatistical interpretations.
