| Abstract |
In New Zealand’s high temperature geothermal systems, fluids flow dominantly through fractured rocks with low matrix permeability. Although surface fault mapping and borehole image log data show numerous structures, it remains challenging to efficiently target permeable fractures. A common assumption is that the intersection of several fracture populations provides high connectivity, resulting in high permeabilities and increased dispersion of tracer particles. To test this assumption, we present fluid flow calculations in several distinct discrete fracture networks, each of which is broadly consistent with the fracture density and dip angle interpreted from borehole image logs in the Rotokawa Geothermal Field, a New Zealand andesite-hosted reservoir. There, fractures are dominantly steeply dipping (75°-90°) but with about 25% of fractures dipping between 30° and 70°. The population models we use for dip angle include single and paired normal distributions with different standard deviations, a mixed model with normal and uniform distributions, and an empirical distribution based directly on dip angle measurements. In all population models fracture lengths are calculated from a truncated power-law distribution, and aperture is proportional to fracture length. Most models show pervasive connected fractures, with anisotropic flow and tracer transport dominantly parallel to the mean fracture dip. Permeability anisotropy increases as the standard deviation of the dip distribution decreases, with flow (and tracers particles) becoming increasingly channelled, until channels become separated and permeability drops. Models composed of 2 or 3 orientation sub-populations do not provide the expected increase in permeability compared to single-population models. The spread in orientation in the Rotokawa data, rather than well-defined individual fracture sets, is consistent with them being pervasively fractured lavas under tectonic stresses and allows a higher tracer dispersion than single tight populations. Strikingly, less than 4% of all fractures actively contribute to flow at the length scale of our model, 350 m, with the remainder occurring in isolated smaller groups which may store, rather than transport, fluids. In pervasively fractured rocks, the presence of multiple sub-populations of fracture orientation may thus not necessarily enhance permeability at reservoir length scales. |