Fracture formation and evolution in crystalline rocks: Insights from attribute analysis

Abstract

Fractures are ubiquitous in crystalline rocks and control the strength and geophysical and fluid transport characteristics of the Earth's upper crust. A quantitative description of fracture attributes may constrain models of fracture formation and evolution. In this study, fracture attributes collected from one-dimensional samples across exposures of typical crystalline rocks show comparable variability in fracture size and spacing to sedimentary rocks. Vein thickness and fracture aperture data show predominately power-law distributions. Vein and fracture spacing data are best described by exponential distributions with negative slopes and appear to vary with composition in intrusive rocks. The fracture systems exhibit a range of anti-clustered to clustered patterns, and densities are an order of magnitude higher for joints compared to veins. Fracture clustering data can be used in conjunction with the spatial distributions to provide information on the controlling processes of fracture spacing. We suggest that exponential spacing distribution is produced as a sampling effect for both periodic-spaced and clustered fracture sets. In the examples given here, thermal stress-related joint patterns are distinguishable from tectonic-related fractures in plutonic rocks and fracture density and clustering is increased towards a major reactivated basement fault.

Evidence for tectonic, lithologic, and thermal controls on fracture system geometries in an andesitic high‐temperature geothermal field

“…A common way of determining the clustering tendency of a data set is to use the coefficient of variation (Cv) for normal distributions, the ratio of the spacing standard deviation to the mean spacing. Cv > 1 indicates a clustered data set, while Cv = 1 indicates a randomly distributed data set and Cv < 1 an anticlustered data set [ Gillespie et al , ; McCaffrey et al , ]. At Rotokawa, Cv of fracture spacings within truncation limits varies between 1.2 and 1.9, without clear correlation to specific fracture sets, which indicates that the fractures are slightly clustered.…”

“…By contrast, the highest ranking distribution for spacings of sets FS2 (N‐S strike) and FS4 in borehole RK32 (NE‐SW strike, dipping ∼60°) is power law and thus scale independent, with coefficients similar to those reported by Gillespie et al [] for faults. Thus, spacing of these subordinate fracture set is likely controlled by rupture processes driven by anisotropic tectonic forces (Figure b) [ Bonnet et al , ; McCaffrey et al , ]. While only three major subsurface faults have been identified with confidence at Rotokawa, the presence of other subsurface active faults has been inferred from localized S Hmax direction rotations observed in BHTV logs [ McNamara et al , ] and may cause this power law spacing distribution locally.…”

“…Power law coefficients of fracture thickness in BHTV logs at Rotokawa (2.6–3.1) are similar to the coefficients derived from BHTV log data for geothermal fields hosted in faulted crystalline formations [ Barton and Zoback , ; Radilla et al , ], and in andesites based on outcrop scanline measurements [ McCaffrey et al , ]. Fault zone thicknesses (0.01–∼110 m thick) measured in tunnels in the south of the TVZ are also approximated by a power law distribution, with a coefficient of 1.5 and 1.8 in Miocene sandstone and mudstone, respectively [ Seebeck et al , ].…”

“…Fluids tend to flow along these joints and contacts between layers (Figure a). In contrast, the spacing of fractures close to faults are commonly clustered according to power law distributions which are scale independent (Figure b) [see Bonnet et al , , for review; McCaffrey et al , ; Johri et al , ]. Power law distributions of fracture mechanical aperture, length, and density are commonly used to generate reservoir‐scale fracture models [ Bonnet et al , ; Davy et al , ], albeit resulting from observations rarely made in volcanic formations.…”

“…and fractures in volcanic units has been extensively studied in outcrops [e.g., DeGraff and Aydin, 1987;Gudmundsson, 2000;Conway et al, 2015] but can be difficult to quantify in reservoirs due to sparse core and wireline logging measurements. In lava-hosted reservoirs, permeability is strongly influenced by faults and fractures [e.g., Nemčok et al, 2007;Davatzes and Hickman, 2010;McNamara et al, 2015], but the key fracture attributes are often not sufficiently well constrained to make accurate fluid flow predictions [Aprilina et al, 2015;Hernandez et al, 2015;Kissling et al, 2015].Description of the three-dimensional (3-D) geometry of a fracture system is necessary to develop reservoirscale models which can assist resource management [Bonnet et al, 2001;Berkowitz, 2002;Chilès, 2005] and constrain the origin and evolution of the fracture systems [Ackermann et al, 2001;McCaffrey et al, 2003]. Fracture systems can be described in terms of the orientation of different fracture sets, and the probability distribution of fracture attributes: aperture (for open fractures), thickness (for veins), spacing, density, and length [Priest, 1993].…”