Abstract: |
Many geostatistical techniques perform better when anisotropy is considered to be locally varying, but there are few techniques available for inferring the necessary locally varying anisotropy (LVA) field. All geostatistical modeling methodologies assume a form of stationarity, one common assumption is second order stationarity where anisotropy is considered to be globally constant. This assumption is often geologically inappropriate and requires geostatisticians to subdivide the modeling area into stationary domains, thereby increasing the required professional time for modeling and potentially producing disjointed domains that are globally inconsistent. Existing algorithms can be modified to relax the assumption of second order stationarity, for example: multiple point statistics with local pattern reorientation; local kriging search reorientation, and; kriging with LVA. These techniques require an exhaustive model of the local orientation and magnitude of anisotropy, termed an LVA field. In practice there are rarely direct measurements of the LVA field, even local dipmeter data can be unreliable due to the discrepancy between the measurement of small scale anisotropy and the larger scale anisotropy required for modeling.
The focus of this work is to develop a suite of methodologies for LVA field inference from various 2D or 3D data sources. No single methodology is appropriate for all deposits because local geology and data availability vary. Methodologies presented for inference of 2D LVA fields include (1) computer assisted manual generation (2) moment of inertia (3) polylines, and (4) geometric geological interpretation. Methodologies for inference of 3D LVA fields are similar but require additional considerations for visualization and checking. Techniques are demonstrated with real data when available, examples include: a porphyry deposit; a uranium role front deposit; a complex layered reservoir; a massive disseminated deposit, and; a folded gold deposit. Recommendations to aid in technique selection are provided and are largely determined by the nature of the data available and the geometry of the formation under study. |
