Geostatistical Analysis for Evaluating Spatial Dependence in Fracture Set Characteristics ? Introduction

Analysis of discontinuous rock masses for engineering design purposes requires information about fracture characteristics. For such studies, fractures or other elements of the rock fabric (bedding, foliation, etc.) should be considered in a statistical manner because they generally have high frequencies of occurrence and are discontinuous over the design area. A practical design procedure includes defining structural domains as those areas which contain fractures with similar orientation patterns. Within a structural domain, the statistical distributions of fracture set characteristics (spacing, fracture density, dip, dip direction, etc.) may be computed from mapped data and then used to quantitatively describe the rock fabric. These distributions can then be randomly sampled (using Monte Carlo techniques) to simulate the rock fabric and to help define potential failure surfaces along fractures. 1 The prediction of failure geometries is an essential part of stability analyses used in designing rock slopes or underground openings. In current probabilistic design procedures, Monte Carlo techniques that rely upon the iterative generation and analysis of deterministic instability models do not incorporate spatial aspects of the random variation in fracture set characteristics, 1,2 but consider the characteristics to be spatially independent.