SPATIAL PROBABILITY MODELING AS AN EXPLORATION TOOL

By intersecting spatially referenced resource locations with geological, geophysical, and geochemical maps in a GIS, exploration geologists can build predictive modeling tools to help define prospective areas for fieldwork. A variety of techniques will be briefly reviewed in a practical exploration context. In addition, a simple and effective technique, the Density Function (DF) method, is described. The DF method merges some aspects of weights-of-evidence and fuzzy logic modeling, and can be easily applied using two ratios in a spreadsheet. An example is given wherein the performance of the DF method is compared to weights-of-evidence.

Many methods are derivations of the general Bayesian technique. Some of these have been expanded upon by Dempster-Schafer and highly optimized models can be performed with neural networks. These methods all have their advantages and disadvantages. Pure probability models employ categorical methods that require the simplification of data sets in a process that often concerns explorationists. Fuzzy logic methods use continuous data functions, but they can produce less quantitative and potentially biased assessments. Neural Networks are highly optimized, but they do not always allow dissection of the model in ways that permit geologists to find out why a certain area is ranked the way it is.

The DF Method merges categorical probability models with fuzzy logic and contains advantages of both methods. A required assumption is that the resource targets are spatially very small relative to the entire study area. This assumption puts the method outside of pure probability modeling, but the assumption has been found valid for virtually all geological exploration targets and in such circumstances the DF method becomes mathematically equivalent to a pure probability model.

Simple and accessible tools are important for a working exploration environment. Because of their simplicity, tools such as the DF method can be useful to indicate prospective areas. These tools do not choose drill targets, but they do force a more methodical approach to exploration that is repeatable. In addition these tools can provide a less biased way to evaluate layers of geologic data and the enhancement techniques applied to them.