Towards understanding the influence of data-richness on interpretational confidence in image interpretation

Abstract

Geological interpretations of aeromagnetic and gravity images are highly subjective but are rarely accompanied by a quantitative confidence assessment, which is a key limitation on the usefulness of the results. This paper outlines a method with which the relative level of data richness can be assessed quantitatively, leading to an improved understanding of spatial variations in interpretational confidence. Simple rules were used to quantify the likely influence of several major sources of uncertainty. These were: 1) the level of geological constraint, using the local abundance of outcropping rock and the quality of geological mapping; 2) the interpretability of the data, considering the strength of edge-like features and the degree of directionality of these features, a proxy for structural complexity; 3) data collection and processing errors, including gridding errors, derived from the statistical error returned during kriging, and the influence of anisotropic line data collection on the detection of gradients. From these individual sources of uncertainty an overall data richness map was generated through a weighted summation of these grids. Weightings were assigned so as to best match the result to the interpreter's perception of interpretational confidence. This method produced a map of data richness, which reflects the opportunity that the data provided to the interpreter to make a correct interpretation. An example from central Australia indicated that the data influences were preserved over a moderate range of weighting factors, and that strong bias was required to override these. In addition to providing a confidence assessment, this method also provides a way to test the potential benefits of additional data collection.