AEM system target resolvability analysis using a Monte Carlo inversion algorithm

Abstract

There are several factors that must be considered when choosing an airborne electromagnetic (AEM) system for a specific survey task. These factors always include cost, availability and logistics. However, the most important consideration is the ability of an AEM system to resolve the target(s) to be mapped. To date Geoscience Australia has tackled this later consideration in terms of 'detectability' rather than the 'resolvability', which are two distinct concepts. We say that a target is detectable if the difference between the AEM response of the target and the background is sufficiently greater than the AEM system's noise levels. Resolvability not only requires that the target's data anomaly be detectable, but that we can also estimate, with sufficient confidence, via an inversion procedure, the cause of the anomaly. Geoscience Australia is now addressing the resolvability question though a reversible jump Markov Chain Monte Carlo (rj-McMC) inversion algorithm. A 1D forward model code generates synthetic data for each AEM system under consideration, for a suite of type-model scenarios that represent the expected range of situations to be mapped, which may include actual downhole conductivity logs. The data are then inverted using the rj-McMC inversion, which, importantly, uses the expected AEM system noise levels. The rj-McMC algorithm samples millions of models, possibly on independent parallel Markov Chains, that fit the data to within the AEM system's expected noise levels. Analysis of the ensemble of models yields a robust estimate of the uncertainty of resolving the model at any particular depth. It is a simple matter to then compare and contrast the results for each AEM system under consideration. We also show how the method can be used to provide depth of investigation estimates.