Probabilistic models for ore body recognition

Four statistical models have been developed to aid in exploration drilling for ore deposits. The Austinville (Virginia) deposit, which is a Mississippi Valley-type deposit with 31 million metric tons of ore in 17 ore bodies, provides the data base. Because most ore bodies are not solid ore, holes drilled through them will intersect sub-ore mineralization as well as ore. The statistical models were designed to show whether drilling is through an ore body or not, depending on the mineralization intersected by holes. The models are based on a threefold classification of mineralization penetrated by surface drill holes and were developed for cases of fixed budgets with specified number of drill holes and floating budgets with variable number of drill holes. Two classifications were used: (1) binomial ore/non-ore classification and (2) trinomial ore/mineralized/barren. For the fixed budget model with the ore/non-ore classification, at least five holes need to be drilled to decide whether or not an ore body has been penetrated. For the fixed budget model with the ore/mineralized/barren classification, and for both floating budget models, at least three holes need to be drilled. All decisions are based on a 90% confidence level.