MULTIDIMENSIONAL STATISTICAL TECHNIQUE FOR DETECTION OF LOW CONTRAST GEOCHEMICAL ANOMALIES

Abstract

In this article we propose an advanced technique for detecting low contrast geochemical anomalies using a set of features. There are three principal elements in this technique: (1) a statistical measure of the contrast of the anomaly, denoted as τ; (2) selection of a background population; and (3) reduction of the dimensionality of the feature space. In the frame of the model, which describes the statistical distribution of geochemical background as a multidimensional normal distribution of logarithms of concentrations, the index, τ, is a powerful test statistic for the hypothesis of abnormality of an observation. Maps of τ anomalies can be rigorously interpreted on the basis of statistical inferences. Under all equal conditions this technique allows the detection of geochemical anomalies with at least the same contrast (if the chemical elements in a background population are correlated, then even the better) as using selective extractions of metals from soil or other techniques for data processing. The advantages of the proposed technique are demonstrated both theoretically and on examples of rare-metal and copper-nickel mineral deposits.