Decision making using geostatistical models of uncertainty

The selection of dig limits and well locations in the mining and petroleum industries have enormous economic consequences. Sometimes these decisions are easy, but usually the decision making process is obscured by uncertainty and an abundance of plausible alternative decisions. Geostatistical tools can be employed to account for uncertainty, but incorporating a model of uncertainty into a decision making framework adds another layer of complexity. Common practice for decision making under uncertainty is to construct a global objective function that quantifies loss and to use an optimization algorithm to find the decision that minimizes loss. Many optimization algorithms require that the global objective function meet certain criteria for use. The dig limit and well location selection problems cannot be set up in a way that satisfies the constraints for many optimization algorithms. The algorithm known as Simulated Annealing has very few constraints and can be used to select dig limits and well locations that minimize loss, given a well constructed global objective function.; This dissertation develops techniques that semiautomatically select dig limits and well locations that account for subsurface uncertainty. The dig limit selection technique has the additional feature of selecting dig limits that account for the limitations of the mining equipment. The techniques are referred to as semiautomatic because the user must select seed dig limits or well locations.; For dig limit selection, applications to hypothetical and real mines are explored. As well, hand drawn dig limits and semiautomatically selected dig limits are compared. In experiments, semiautomatic dig limits always outperform the hand drawn dig limits. When the semiautomatic dig limit selection technique was applied at the Bingham Canyon mine improvements of up to 1.5% are observed. The semiautomatic well location selection technique is applied to hypothetical reservoirs and a real data set: the Smiley Buffalo waterflood project. The semiautomatic well location selection outperformed well locations selected by the Asset Team by more than 19%.