Regression models for estimating mineral resources from geological map data

Abstract

Several regression models can be used for explaining discovered occurrences and estimating undiscovered occurrences of mineral deposits of a given type from geological data quantified for equal-area cells. All the models are based on the assumption that the occurrences of mineral deposits in an area are estimable as a function of the mapable geological attributes in the area. Depending upon further assumptions made with respect to the dependent variable, that is, the occurrence of mineral deposits, different regression techniques such as logistic and Poisson regression analysis can be employed. In this paper, a jackknife method was used to estimate the variances of the regression coefficient and the occurrences. The effect of undiscovered deposits on the regression models was considered by treating the known deposits separately (Separate Event Method). As an illustrative example, the occurrence of volcanogenic massive sulfide deposits has been expressed in terms of the geological framework in parts of the island of Newfoundland. Each of the models provides a similar and relatively stable pattern that explains the occurrence of the known deposits in a study area of 6000 km 2.The methods presented for calculating the variances of the estimated values are meaningful because they can be used to test whether or not an estimated probability or frequency value for any one cell is significantly greater than zero. The logistic model is used to indicate cells outside the study area where undiscovered deposits are likely to occur.