Fracture width and spacing distributions from borehole televiewer logs and cores in the Rotokawa Geothermal Field, New Zealand

Abstract

The successful targeting of permeable fractures in geothermal fields is aided by understanding the spatial and geometric characteristics of fracture populations. Studies of numerous outcrop, and a limited number of geothermal reservoirs using cores and borehole logs, indicate that fracture frequency and width most commonly follow power-law distributions, with exponential, lognormal, gamma, and power-exponential distributions also reported. This paper presents the first statistical analysis of fracture width and spacing in the high-temperature Rotokawa Geothermal Field, Taupo Volcanic Zone, New Zealand. The fracture dataset comprises: (1) c. 3.6 km of acoustic borehole televiewer (BHTV) logs from three wells and, (2) c. 33 m of core. Statistical distributions have been fitted to the BHTV data using a maximum likelihood estimation method and statistical models selected using the Schwarz Bayesian Criterion. Fracture widths observed on BHTV logs range between c. 1 105 mm. Image resolution and sampling bias reduce the useable range of fracture width to less than one order of magnitude (c. 8 50 mm). Over this range, considering the sampling effects and core observations, the fracture width is best modelled by an exponential distribution with coefficients between 0.13±0.01 and 0.29±0.02, which should be treated as a lower bound. Analysis of fracture spacing of the four fracture sets identified on BHTV logs indicates that the dominant set (striking NE SW) is best modelled by a log-normal distribution, while power-law, power-exponential and gamma are also possible for individual wells. These spacing distributions indicate the presence of a characteristic scale which has not been observed in other geothermal reservoirs hosted in crystalline formations. The characteristic scale may be associated with mechanical interfaces associated with stratigraphic layering, faults, or cooling joints and/or sub-horizontal flow-banding in andesitic formations. Stratigraphic layering can consist of a succession of lava flows with intercalated breccia layers in the andesites, welding variations in tuffs and sedimentary layering in the sedimentary formations sampled by the BHTV logs. The subordinate fracture set striking N S is best modelled by a pareto (power-law) distribution which suggests that the spacing is more likely to be controlled by tectonic processes than by layering. This N S fracture set is predominant in only one of the wells studied which may indicate a structural control on their occurrence in the vicinity of this well. Low fracture spacing (<0.5 5 m) is best modelled by an exponential distribution and higher spacing by lognormal or pareto (power-law) distributions, except for the N – S striking dataset and the NE – SW striking fracture set in well RK32. The change of distribution model at different scales may be linked to the threshold at which fractures start interacting with each other. This work to date underlines the need to combine data spanning a broad range of length scales to conduct a sound statistical analysis of fracture populations and highlights the control on fracture formation by a combination of processes including tectonics, lava cooling and stress perturbations associated with stratigraphic anisotropy. The resulting distributions provide a basis for simulating and calibrating fracture models of geothermal reservoirs beyond those areas directly sampled with BHTV logs or cores and will integrate variations observed over a range of scales between the study wells.