Predictive Expert Models for Mineral Potential Mapping

Background

Mineral deposits are the concentration or existence of one or more useful substances that are for the most part sparsely distributed in the Earth’s crust (Bateman, 1951). Mineralization consists of a set of processes that lead to the formation of mineral deposits. The secondary deposits originate from superficial processes caused by the environment and physical or chemical phenomena thereby causing ore materials to concentrate at the regolith. Physical components include erosion and weathering. To discuss the mineral formation, we recall the theory of Ore genesis which describes its formation in three different components − namely, Source, Transport or Conduit and Trap or Deposit Point. Mineral deposits hardly fit snugly into boxes in which geologist expect them to. Because of the multiple cause of their formation, they are often classified based on their type (Bowden & Jones, 1978; Falconer, 1912).

In mineral prospecting, one of the major goals is discovering new mineral deposits. This can be done by predicting their occurrence using spatial analysis of the distribution of known mineral deposits (Carranza et al., 2003; Bonham Carter et al., 1994). As the concept of mineral potential becomes more established, several methods of predicting hidden mineral deposit through GIS have been developed. At the moment, there is a great paradigm shift towards research in data mining using machine learning. This is motivated by the increase in volume of heterogeneous data and the need to make sense with it. This include the application of machine learning to model geographical and geological data for the purpose of predicting (with some degree of uncertainty) the presence or absence of minerals in a given area.

Research in spatial data analysis has been considerably active over the last two decades. It has helped improve different kinds of computer applications, such as Geographic Information Systems (GIS), Computer Aided Design (CAD), multimedia information systems, data warehousing and earth observation systems (Shekar et al., 2001). Satellite images and digital maps are examples of spatial data because information can be extracted from them by processing the data with respect to a spatial frame of reference relative to the earth’s surface. Computer aided spatial data analysis, mapping and modelling technique have been used in applied geosciences for many years for detecting pattern in the distribution of natural phenomenon (David, 1977).