Geostatistics, a tool for risk analysis in orebody evaluation and ore-related mine decisions

Precise ore estimates are required both in mining project feasibility and in evaluating alternative operating strategies. The actual or perceived degree of risk or uncertainty attached to the estimate are key elements in ore-related decisions. These decisions may have greater economic impact when the mineralization is erratic, interspersed with waste materials or the grade distribution is strongly skewed. Using geostatistics to produce ore estimates, depending on the quality of information available, provides both a qualitative estimate of the orebody and a quantitative measure of the reliability of these estimates.; This study compares various methods of providing an estimate that best represents the physical characteristics of the ore deposit at the local level. This is achieved by presenting a case study of a gold deposit from which both exploratory drill hole and mining blasthole grade information were available. The actual ore is compared to estimates produced by a conventional method, ordinary kriging, indicator and probability kriging, and by sequential indicator simulation. The implications of differences in these approaches are also discussed.; The study provided an illustration of a methodology of data analysis for better understanding the mineralized deposit and improving the reliability of the estimates. In that context, the variogram proved to be a powerful tool for describing the structures and interpreting or validating the geology as well as for playing some decision making role. The non-parametric geostatistical models fit better for dealing with ore estimation and the risk involved. Particularly, the probability kriging gave the best estimates of the recoverable reserves required for feasibility, grade control, and mine planning. In addition, non-parametric geostatistics provide quantification of the risk involved.; Computer software was developed and used through the course of this study. It applies colour graphics to aid in the analysis of data and estimates. The software is portable between PC and UNIX work-stations to make it more accessible to various end users, especially small companies and developing countries.