Spatial Prediction of Fracture Intensity Based on Drill Hole Observations

"The objective of this work is to validate a methodology for spatially predicting the fracture intensity, defined as the average fracture area per unit volume. First, using the observations from drill holes, the fracture intensity is directly inferred at the support of drill-hole composites based on a Terzaghi correction. Subsequently, the composite values are interpolated to larger blocks. The methodology is tested on sets of simulated fracture networks, in which each fracture consists of a disc with a random diameter, random position and random orientation. For each of these simulated networks, one can calculate the actual fracture intensity within any region of space, based on geometric considerations. A regular sampling mesh is then defined to emulate a set of vertical drill holes and the proposed methodology is applied to predict the fracture intensity, both globally and locally. The actual and predicted fracture intensities are finally compared, showing that predictions are unbiased, insofar as, on average, the prediction error is always close to zero. Also, the error dispersion increases when the support of the prediction decreases, which is consistent with the well-known support effect. In practice, the proposed methodology allows predicting fracture intensity based exclusively on drill-hole information, without the need for simulating fracture networks in space, hence its interest. INTRODUCTION The characterization of fractures in rock masses is an important part in the evaluation of mechanical and hydraulic properties of rocks. Fractures have a large impact in disciplines such as reservoir engineering, safety of underground storage of CO2 or of nuclear waste, assessment of high-enthalpy geothermal installations, safety of mining exploitations, efficiency of in-situ leaching operations or of extraction by block-caving. As a rock mass property, fractures must be characterized in the three-dimensional space, whereas observations are limited to 1D (drill holes) and 2D (outcrops, drift walls), moreover very often to rather short stations, so that the available data are subject to geometric bias, truncation and censoring. Fracture characteristics includes the number of fracture sets and, for each set, the number, orientation, spacing, location, shape and size of the fractures must be inferred from data sampling. Some important parameters that are considered nearly in all models are fracture intensity and fracture size."