Geovariances: Where no one has gone before

 News from G²DC - Nov 2012

Domaining in mining industry is one of the preliminary steps for resource estimation. Unfortunately, this step is performed independently from the geostatistical modeling and, for this reason, provides one solution without any attached uncertainty. The consequence of the deterministic and graphical approaches, found in many different software packages, is to make very difficult the modifications and updates of the geological models.

Geovariances and the Centre de Geosciences of Mines ParisTech propose to examine different geostatistical techniques to analyze and/or to interpolate the domains from drillholes and to assess the uncertainty on the volumes using stochastic simulations.

This international project is open to all companies. Today, five leading mining companies are supporting our program of focused research and software development:
 Anglogold Ashanti
 Areva
 BHP Billiton
 Eramet
 Vale

For more information on the project, please contact info@geovariances.com.

General context

• Current Practice
  
  The domaining and the estimation of grades are generally achieved in two independent steps. In practice the domains are mainly defined graphically using CAD systems. It is often a time-consuming process and the modeled domains are difficult to update when new drillholes become available. Moreover when the deterministic mapping of domains boundaries is uncertain, it is difficult to assess the uncertainty on the total ore tonnages. An additional point lies in the difficulty of characterizing the domains, while having to combine an increasing number of parameters (lithology, alteration, grades…).

• Geostatistical Approaches
  
  The idea is to use the flexibility of geostatistics to estimate and simulate the boundaries of the domains, possibly starting from a deterministic geological model. The only well known workable case is when the volume is defined by means of hanging wall - footwall surfaces, which can be estimated and simulated by classical geostatistical methods in 2D, to derive the estimation and uncertainty of the volume. However new approaches are being developed and proposed in the literature, that could lead to a modern, user-friendly and geostatistically robust domaining.
  
  Methods proposed in the literature include:
  
  
  • Distance based methods
  • Implicit methods based on radial basic functions
  • Potential-field method
  • Truncated Gaussian
  • Indicator based methods
  • Indicator P-Field
    
    When several parameters are to be taken into account together, statistical classification methods could be applied to help characterize the domains.

Content of the work

The general objective of the consortium is to provide geostatistically based tools that will help domaining, through the characterization, estimation, simulation and management of the domains.

The geostatistical methods mentioned above will be reviewed from the theoretical point of view, data requirements and practicability. A selection of the best methods will be made and applied to synthetic and real examples provided by partners.

The selected methods will be developed using a new software platform that will be in the future the basis of the new Mining Package developed and commercialized by Geovariances. This platform will support full 3D interactive editing.

A very first module of this new Mining Package will start with drillhole analysis and display tools to define contacts between different intersected domains. This tool will be fully interactive to allow rapid checking of different hypotheses. A new modeling geostatistical module will allow defining, from the drillholes contacts, the 3D envelope of the domain. The uncertainty on the related volume will be quantified through stochastic simulations. This will be illustrated by examples and case-studies on datasets provided by partners.

Tools for the management of domains will also be developed. It will be possible to store or import envelopes from other packages (e.g. as 3D wireframes) and new techniques will be developed to qualify and possibly calculate uncertainty on these imported, deterministic domains.

Proposed Program Year 1

During the first year, the basic elements for domain management and display will be implemented and tests made to select the most appropriate methods to model geostatistically domain boundaries. Work concerning the characterization of domains when several parameters are present will commence.

The orientation of the work, within the general lines of the objective of the consortium, will be examined every six months by the Steering Committee.

Bibliography

• Aug C., Chilès J-P., Courrioux G., Lajaunie C. 3D Geological Modelling and Uncertainty: The Potential-field Method, In: O. Leuangthong and C.V. Deutch (eds.), "Geostatistics Banff 2004", Springer, 2005, p. 145-154
• Srivastava Mohan, Probabilistic Modeling of Ore Lens Geometry: An Alternative to DeterministicWireframes, Mathematical Geology, Vol. 37, No. 5, July 2005
• Cowan E.J. and al. Rapid geological modelling, Applied Structural Geology for Mineral Exploration and Mining, International Symposium, Kalgoorlie, September 2002.
• Machuca-Mory D., Tonnage Uncertainty Assessment of Vein-Type Deposits Using Distance Functions and Location-Dependent Correlograms, APCOM 2009, Vancouver, pp 115-122.