Comparing 1D and 2.5D AEM inversions in 3D geological mapping using a new adaptive inversion solver

1, 2.5 and/or 3D Inversion of Airborne EM data - options in the search for sediment-hosted base metal mineralisation in the McArthur Basin, Northern Territory

“…A 2.5D inversion of the VTEM data using the Moksha code was undertaken by Intrepid Geophysics. The inversion code used in this work has been described by Paterson et al, (2016), and Silic et al, (2015), and comprises a significantly re-engineered version of ArjunAir (Wilson et al, 2006). Among the changes is a new forward model algorithm (ArjunAir only produced accurate results for layer models), and a new 2.5D inversion solver with adaptive regularisation which allows the incorporation of a misfit to the reference model and the model smoothness function.…”

“…Among the changes is a new forward model algorithm (ArjunAir only produced accurate results for layer models), and a new 2.5D inversion solver with adaptive regularisation which allows the incorporation of a misfit to the reference model and the model smoothness function. The regularisation parameter is chosen automatically and changed adaptively at each iteration, as the model, the sensitivity and the roughness matrices change (Silic et al, 2015). The estimation of regularisation parameter requires calculation of only one forward model and sensitivity matrix at each iteration and is controlled by an easily understood parameter -the Relative Singular Value Truncation (RSVT) parameter.…”

“…However, until more recently the prospect of inverting the large volumes characterized by airborne EM datasets was often considered too time consuming given the processing complexities and computational overheads involved in 2.5 and 3D approaches. These challenges have been addressed through, for example, a new forward model algorithms and a new 2.5D inversion solver with adaptive regularisation (see Silic et al, 2015); the use of a moving footprint to limit the number of data points required in calculating the large sensitivity matrix needed as input to a 3D inversion at any one location (see Cox et al, 2010); and by partitioning the forward problem into multiple meshes, resulting in a forward modelling mesh that has far fewer cells than the full inversion mesh thereby reducing the bottleneck for 3D AEM inversions (see Yang et al, 2014).…”

2.5D vs 1D AEM Forward and Inversion Methods at a Survey Scale : A Case Study

“…To date, the combination of AEM surveying and 1D inversion methods has proven useful in defining the regolith geometry and stratigraphic and structural architecture. Many case studies have now proven that AEM can provide this crucial information for mineral exploration (Fraser, 1981;Silic et al, 2015;Silic et al, 2016). In their recent studies, Neroni, Murray and Kepert (2016) showed that constrained 1D AEM inversion in the Hamersley Province, Western Australia enabled exploration geologists to interpret weathering profiles, shallow dipping stratigraphy and steep structures, all of which are crucial aspects of bedded iron ore deposits models.…”

Improved Structural Mapping and Conductive Targeting Delivered by a new 2.5D AEM Inversion Solver

“…This inversion application includes a new forward model algorithm and a new 2.5D inversion solver with adaptive regularisation which allows the incorporation of a misfit to the reference model and the model smoothness function. The regularisation parameter is chosen automatically and changed adaptively at each iteration, as the model, the sensitivity and the roughness matrices are changing, Silic et al, 2015. We optionally allow a starting or reference geology/resistivity model to influence the inversion instead of a simple halfspace. The inversion application is incorporated within 3D geological modelling software with an intelligent graphical user interface allowing simple integration of this process.…”