2.5 Airborne Electromagnetic Inversion:A review of the benefits of moving to a higher dimension

Rod Patterson – Intrepid Geophysics

Introduction

Intrepid Geophysics began operating a 2.5D Airborne Electromagnetic (AEM) inversion service in early 2016, following two years of software development by Intrepid Geophysics and Jovan Silic, the primary EM software modelling and inversion algorithms developer. The software was developed to facilitate targeting of ore bodies, mapping of geology and geological structures, and the location of water aquifers for exploration and development purposes.

The 2.5D AEM inversion software makes use of Intrepid’s 3D GeoModeller as a user interface, a visualisation tool and for creating forward models, but runs as a separate MPI application using control and batch files. The inversion results are loaded into GeoModeller for visualisation and QC. In addition, the results add value to 3D GeoModeller’s implicit geological modelling package, which already contains a stochastic plugin for magnetic and gravity inversion. As such, resistivity/conductivity and chargeability join magnetic susceptibility and density as physical rock properties that can be modelled, conforming with Intrepid’s overall philosophy of deriving rational geology from geophysics.

The 2.5D software is a substantial rewrite and parallelisation of the original CSIRO/AMIRA project P223, ArjunAir code, and is now named Moksha. The rewrite implemented a new adaptive solver and forward model (Silic et al. 2015). For reference, AMIRA P223 ran with strong industry support from 1981 until 2008, a period of 27 years. The partially finished software was released into the public domain in 2010.

Advantages of the 2.5D application compared to the 1D are that it can model topography and irregular subsurface structures where the structure along strike is assumed to be constant for a geo-electric distance greater than the AEM 3D source footprint. The computation is based on the response of a 2D model to a 3D source (hence the 2.5D descriptor) and can be applied to 3D structures whose conductivity precludes the spread of the source wave beyond the 2D region during the time range of the data (Raiche et al. 2008).

While the 2.5D module has been fully operational since 2016, the software is being continually refined to deliver improvements in resolution, performance and to add new features. For example, the recent introduction of a variable finite element mesh resolution in the X direction has allowed better definition of narrow conductors and improved productivity for surveys requiring higher mesh resolution.

Forward modelling and joint inversion of induced polarisation (IP) has also been added in an effort to manage commonly encountered IP effects in surveys flown with the more powerful suspended loop systems such as VTEM, SkyTEM and Xcite.

The software’s advantages over industry standard Conductivity Depth Imaging (CDI) and 1D inversions are:

1. The ability to handle topography and thereby remove topographic artefacts e.g. resistive anomalies on hills.
2. The ability to model resistivity contrasts up to 1 million to 1.
3. The elimination of non-geological “pant-leg” conductors by being able to handle strong vertical/lateral discontinuities. Pant-legs are typical of 2D effects in a 1D inversion. For example, a resistor is imaged beneath a conductive “pant-leg” centred above a vertical conductor. The 1D assumption is inadequate for imaging the 3D structure as the horizontal-dipole behaviour cannot be explained by a 1D conductivity structure (Oldenburg et al. 2019).
4. Prediction of geologically reasonable fold structures that are accurate for dips greater than 20°. 1D is only reasonably accurate for dips up to about 20° over extensive (large conductors) without significant lateral conductivity variations. 1D inversions over an approximate 1D horizontal layer with significant lateral conductivity variations will not only produce artefacts, but can place conductive features at depth where there are none.
5. Joint inversion of X and Z components linked by a full vector treatment of Maxwell’s equations.
6. Joint inductive and IP (chargeability, time constant and frequency effect) inversion.
7. The ability to constrain inversions using a resistivity reference model based on known or hypothesised geology.

In a non 1D geological scenario, it would be preferable to invert the data in 3D and remove the assumption of geological strike with respect to flight direction. However, this is not always a practical option for the following reasons:

1. 3D at a survey scale will not be valid at wider line spacing when lateral continuity is not assured.
2. 3D is invariably an under-determined inversion problem with more unknowns than data points. Thus, it requires the imposition of extra conditions or assumptions on the inverted model.
3. 3D solves for millions of unknowns requiring large compute resources so that inverting at high spatial resolutions becomes very expensive in both time and cost.

2.5D is a good compromise as it uses a numerical implementation of Maxwell’s equation in an over-determined system. This is in contrast to the 3D system of equations being under-determined as described in the second point above. The 1D class of solvers largely ignores the horizontal variations.

Over the past five years, 2.5D inversion service work was completed by Intrepid on surveys using all major AEM systems (TEM and FDEM), see Figure 1, and spread across most mineral exploration regions, see Figure 2. The bulk of these inversion projects was undertaken for conductive targeting and geological/ structural mapping purposes, with the split roughly equal. Approximately 10% of projects were related to groundwater mapping.

Figure 1. 2.5D inversion service by AEM system

Figure 2. 2.5D Inversion service by region.

Examples of 2.5D inversion and forward modelling

Data from Quamby/Dugald River (Queensland, Australia) and Kevitsa (Finland) are used to illustrate the advantages of 2.5D inversions over the more commonly provided and utilised CDIs and 1D inversions when the geology and targets sought do not fit the 1D assumption.

Data from Elura (New South Wales, Australia) are used to illustrate the application of the Moksha software in tandem with a full 3D geology modelling package to forward model the performance of different AEM systems over a realistic exploration target. The detectability of deposit styles in different weathering regimes using different systems can be predetermined, resulting in a more cost-effective survey design.

Quamby/Dugald River, Queensland, Australia

The data used in this example came from the East Isa VTEMTMPlus AEM survey. The survey was flown in 2016, and was funded under the Queensland Government’s Future Resources (Mount Isa Geophysics) Initiative and managed by Geoscience Australia on behalf of the Geological Survey of Queensland. The Initiative aims to attract explorers into ‘greenfield’ terrains and to contribute to the discovery of the next generation of major mineral and energy deposits under shallow sedimentary cover.

This example demonstrates the ability of 2.5D inversion technology to reliably image steeply-dipping and folded geology, and to present exploration targets ready for testing.

Z component-only inversions were performed on eight lines of VTEMTMPlus that were 2 km apart over an 8 x 15 km block in the Quamby/Dugald River mineral district, east of Mt. Isa. A number of discrete exploration targets were defined, and some of these have a close spatial relationship to the shale hosted Pb/Zn mineralisation at Dugald River.

Figure 3 shows the survey area and the location of the Dugald River mine relative to the VTEMTMPlus flight lines on a Google Earth backdrop. The mapped synclinal features are highlighted with white symbols, and inclined drill holes are in magenta. Inversion results are presented for lines 15701 and 16101, which are denoted in blue. Figure 4 shows the geology of the survey area with the mapped synclinal features highlighted by red symbols.

Figure 3. Dugald River area showing the inverted VTEMTMPlus flight lines relative to the mine site. Mapped synclinal features are highlighted by white symbols and inclined drill holes are in magenta.

Figure 4. Quamby 1:100 K geology (Wyborn 1997) showing the conductive shale horizons. Mapped synclinal features are highlighted by red symbols and inverted VTEMTMPlus lines are in blue.

The 2.5D Z component-only inversion results in Figures 5 and 6 highlight the synclinally folded shale package (grey/grey-pink striped colours in Figure 4), which is host to the Dugald River deposit. These shales wrap around the Knapdale Quartzite core.

Figure 5. Line 15701 2.5D inversion, Z component-only, (top) and CDI (bottom). The colour stretches are roughly equivalent; the CDI units are conductivity in Siemens/m and the 2.5D inversion units are log conductivity in milli-Seimens/m.

Figure 6. Line 16101 2.5D inversion, Z component-only, (top) and CDI (bottom) showing the location and geometry of conductive anomalies. The colour stretches are roughly equivalent; the CDI units are conductivity in Siemens/m and the 2.5D inversion units are log conductivity in milli-Seimens/m.

Comparison of AEM inversion results

The log conductivity results for 2.5D, Z component-only, and CDI inversions for Lines 15701 and 16101 across the Quamby/Dugald River subset are shown Figures 5 and 6.

The 2.5D inversion results show synclinal features with dips that closely match the known geology from drilling and surface mapping. Conversely, the CDIs show anticlinal patterns caused by difficulties handling strong lateral resistivity contrasts and a breakdown of the 1D assumption in the presence of steeply dipping complex geology.

Interactive views of channel data misfits, the noise model and conductivity section for line 16101 (Figure 7) enable the user to validate the inversion results in this folded synclinal slate horizon example. Observed profiles are in colours and the predicted profiles are in black. The noise model panel (bright green and blue) shown below the three profile panels in Figure 7 is a by-channel (Y axis) map of the noise estimates used in the inversion. Channel values in blue are below the chosen noise threshold and channel values in red are negative transients assigned as IP effects. The former are down-weighted and the latter are ignored during the inversion.

Figure 7. Line 16101 2.5D inversion profile misfits (top three panels), noise map and conductivity section (bottom panels). Note the colour inverted profiles match the observed data closely such that the observed profiles (black) are not visible at print resolution.

Managing IP effects in 2.5D inversions

The eastern end of line 15701 displays a CDI artefact, which is interpreted as being caused by near surface IP effects in this area (Figure 5). These IP effects manifest as negative late time transients as seen in the profile displays and highlighted in red in the noise panel in Figure 8. The source of the IP effects has not been investigated in the field, but these effects are often associated with near surface clays.

Figure 8. Line 157101 2.5D standard Z component inversion (left) and joint IP/Z component inversion (right).

A joint Z component-only inductive and IP inversion was run over this line, and the results are presented in Figure 8, where it is compared with a Z component-only inductive inversion. The red zones in the accompanying noise map highlight the time gates and areas with negative transients.

The inductive-only inversion in the left side panel does not fit the negative late time IP pull down, (evident in Channels 31-45 in Figure 8), but does in the joint inversion on the right (as denoted by the red arrows). A chargeability section is also generated from the joint inversion as shown in the panel below the noise map.

To summarise, a 3D perspective view of the 2.5D inversion results for the full set of eight lines from this area is shown in Figure 9.

Figure 9. Quamby/Dugald River 3D perspective view of AEM 2.5D inversion results.

Quamby/Dugald River takeaway

The distinct anomalies apparent in the 2.5D sections provide clear drilling targets for rapid strategic decision making. Whilst the VTEMTMPlus survey was conducted at a 2 km line spacing, conductivity imaging by the 2.5D inversion software demonstrates an ability to correctly identify the geometry (dips) of structurally complex exploration targets.

Kevitsa, Finland

The Kevitsa VTEM survey data was provided to Intrepid by First Quantum Minerals Ltd. (FQML) prior to the Kevitsa mine being sold to Boliden in June 2016.

This example illustrates the ability of the 2.5D inversion process to generate 2D depth slices or 2D level plans at constant elevation to highlight geological structure. These products enable direct comparison of conductivity with 2D maps of other geophysical data such as magnetics and gravity, as well as maps of the surface geology. The inversion resolves resistivity and conductivity contrasts very well, and can produce these enhancements accurately when flight lines are spaced close enough to adequately map across line continuity.

The geology of the Kevitsa mine area, as it was known in 2009 (FQML 2009), is shown in Figure 10. The extent of the VTEM survey is shown over a 2020 Google Earth image in Figure 11. The survey was flown in 2009 prior to mine construction.

Figure 10. Solid geology map of the Kevitsa mine area (FQML 2009).

Figure 11. Extent of the Kevitsa 2009 VTEM survey shown over a 2020 Google Earth image. The survey was flown before mining commenced.

A 50 m conductivity depth slice generated from 2.5D inversion of the VTEM survey lines flown from East to West is shown in Figure 12. The correlation of the depth slice with the geological map is very good, and the inversion clearly maps the phyllitic rocks within the mafic intrusive complex. The correlation is emphasised by overlaying the geological structure boundaries on the inversion depth slice, see Figure 13. There is also a significant correlation between the airborne magnetics reduced to pole (RTP) Total Magnetic Intensity (TMI) and the geological structure, see Figures 14 and 15. The 2.5D inversion depth slice more clearly defines the geological structure in this instance, with the TMI adding some detail in the more magnetics units as might be expected. The correlation with the geological map is not surprising since there is little surface exposure in this area, and the map had been interpreted from the existing geophysics and some drilling.

Figure 12. Kevitsa VTEM 2.5D inversion 50 m depth slice.

Figure 13. Kevitsa VTEM 2.5D inversion 50 m depth slice with geological structure overlay.

Figure 14. Kevitsa VTEM TMI RTP (colour) and 1VD drape (grey scale).

Figure 15. Kevitsa VTEM TMI RTP and 1VD drape with geological structure overlay.

Figure 16. Kevitsa: Comparison of the log conductivity 2.5D (top) and CDI (bottom) 300 m depth slice

Figure 17. Kevista: Cross sections along Line 30415 and 30295 of the log conductivity 2.5D (top) and CDI (bottom) 300 m depth slices

This type of enhancement is not achievable for a 1D inversion when there are strong lateral discontinuities in the geo-electrical section. For example, pant-leg artefacts, expected in geological scenarios such as Kevitsa, could create false structural features which would mar the interpretability of the depth slice. 2.5D and CDI 300 m depth slices are shown in Figures 16. Cross sections along lines 30415 and 30295 of the 2.5D and CDI 300 m depth slices are shown in Figures 17. A consistent linear stretch (log conductivity, 1 to 3 mS/m) has been applied to all images. The pant-leg artefacts at the edges of conductive features on the CDI sections and some deeper conductors produce quite coherent conductivity trends in the CDI depth slice that could easily be mistaken for geological structure.

Kevitsa takeaway

Depth slices generated by 2D gridding of the 2.5D inversion log conductivity cross sections are better than vertical sections for defining geology and geological structures (formation boundaries and faults) mapped on a horizontal plane. These products also facilitate the integrated interpretation of AEM, magnetics, gravity and surface geology.

In addition, multiple depth slices can be used to build a more complete 3D picture of the geology. The 2D gridding methodology is faster and easier to control than full 3D gridding, which can be plagued by base level shifts between sections when the inversion nears the depth of investigation. The 2D depth slices are easily de-corrugated to remove this effect.

This type of enhancement is not achievable for a 1D inversion when there are strong lateral discontinuities in the geo-electrical section. For example, pant-leg artefacts, expected in geological scenarios such as Kevitsa, could create false structural features.

Elura, New South Wales, Australia

Forward modelling can be used to test the ability of AEM systems to detect ore bodies of interest beneath conductive cover and, further, to inform choices about the “best” and/or most cost-effective survey system for solving a particular exploration problem.

In this example the Moksha software was used to test whether the NRG Xcite system (or similar helicopter AEM systems) could have detected the Elura Cu-Pb-Zn-Ag massive sulphide orebody.

The Elura orebody was discovered in 1973 by the Electrolytic Zinc Company of Australasia Ltd prior to the availability of modern lower frequency AEM systems. The discovery was made by geochemical follow up of a bullseye magnetic anomaly. The first massive sulphide drill hole intersection occurred in 1974, and mining commenced in 1983, (Schmidt 1989). The Elura Mine was purchased by CBH Resources Ltd in 2003, and renamed the Endeavor Mine. The mine is currently in Care and Maintenance.

The modelling work flow was as follows:

1. Published material from the “Proceedings of the Elura Symposium, Sydney 1980” (Emerson 1980), was used to build a 3D model of the massive sulphide orebody. The papers in the proceedings provide a well described set of mineralised rock units and their resistivities on a series of cross sections and level plans that allowed an accurate model of the upper 500 m of the deposit to be generated.
2. The 3D model was built in GeoModeller, which provides part of the user interface and visualisation engine for Moksha.
3. The model units and assigned properties were exported to a 2D section mesh which was used to generate the finite element mesh used in the 2.5D AEM modelling process.
4. The mesh resolution was optimised for the size and scale of the problem to ensure that an accurate electrical model response could be calculated.
5. The mesh was of variable resolution and was adapted to accurately define the geometry of the smallest features to be resolved.

The main massive sulphide lens at Elura is in the form of a steeply dipping pipe with X, Y, Z dimensions of ∼60 x 150 x 500 m. This pipe lies beneath a layer of conductive regolith ∼100 m thick. This regolith is what limits AEM detection. Weathering of the orebody has resulted in the formation of a gossan, which has a small surface outcrop (Section 5730N, Figure 18).

Figure 18. Ore lens vertical geometry, sections 5730, 5750 and 5800N (see Figure 20 for legend).

The sections and plans in Figures 18 and 19 show the general geometry and mineral zonation of the orebody. The core of the orebody consists of massive pyrite and pyrrhotite that is highly conductive. The ore units, resistivities and forward model mesh dimensions are summarised in the forward model mesh and ore type property legend, Figure 20.

Figure 19. Ore lenses horizontal geometry, plan sections +150 to -300 m (see Figure 20 for legend).

Figure 20. 2.5D Forward model resistivity and mesh geometry and ore type property legend.

The 2.5D forward model has been calculated for both X and Z components using system noise estimated from the recent NRG Xcite Cobar regional survey flown by the Geological Survey of New South Wales in collaboration with Geoscience Australia. The survey was designed to map geology, minerals and groundwater in the Greater Cobar area and was flown as part of the MinEx Cooperative Research Centre’s (MinEx CRC) National Drilling Initiative. The NRG Xcite survey could not be flown over the known Elura deposit because the cultural disturbance at the mine is too extensive to obtain noise-free data. The nearest line is ∼1 km north of the orebody extremities.

The Elura orebody is clearly visible in the last 10 channels of the Z component forward modelled data, and is well above noise levels. The X component is more heavily affected by noise, but the orebody response is visible at early to mid-time, see Figure 21. With the addition of motion noise to the X component, which is not included in these estimates, this response may be difficult to identify.

Figure 21. 2.5D Xcite forward model; section 5750N, flown from left to right.

Elura takeaway

We can conclude from the forward model of section 5750N that the Elura orebody would be recognisable in the Z component of the Xcite data if a survey line had crossed over the main massive sulphide lens, which has a strike length of ∼150 m. It is unlikely that a response would have been seen in a line 100 m to the south, as the mineralisation cuts out quickly in that direction.

Conversely, had a survey line been flown 200 m further to the north, where the top of mineralisation is ∼320 m below surface and the massive sulphide lenses are smaller, the survey would have been unlikely to detect the conductor. Note, it was beyond the scope of the modelling exercise to forward model heliborne systems with larger dipole moments (and therefore broader footprint), although this may be attempted at a later date.

The conductive cover in the northern part of the Cobar Basin, where Elura is located, is generally thicker than further to the south. Clearly thinner regolith would enhance detectability. Nevertheless, to ensure detection of this deposit style (very steeply plunging, short strike length sulphides) a close line spacing (<200 m) is probably required.

Operational challenges

AEM data processing, including 2.5D inversions comes with a number of challenges. These challenges can be broadly grouped as relating to system complexity and geological noise.

System complexity

A high level of AEM processing experience is often necessary to resolve the complexity of some of the problems encountered when processing data from both old and modern AEM survey systems. A lack of experience can lead to poor outcomes. In particular, considerable care is necessary when reviewing the system setup, as recorded system parameters can vary from survey to survey. The more modern helicopter suspended loop systems have the ability to vary transmitter waveforms, receiver filters and time gate positions depending on survey conditions. Errors sometimes occur in documentation, and detecting these errors requires a combination of experience and multiple test runs prior to commencing full survey inversions. Well-designed work flows that formalise a series of checks are essential for a good outcome.

Geological noise

Conductive cover

Conductive cover can pose serious limitations on an AEM system’s ability to detect a buried conductor. In simple terms, a thicker and/or more conductive surficial cover (regolith) limits the detection depths of an AEM system. It is important to understand this attribute of a survey environment before planning and conducting an AEM survey, particularly in Australia. Existing AEM surveys can be informative, and forward modelling particular geological targets within known regolith and geological environments, such as demonstrated by the Elura example, can be very useful, saving time and money.

IP effects

IP effects can be serious problem in AEM surveys, and in some cases can completely mask an AEM system’s ability to detect a late time conductive response. This is often caused by the response of near surface clays to the AEM system’s transmitted signal. A more powerful transmitted signal can increase the IP response. The IP response is opposite in sign to the inductive response and causes pulldown (a decrease in the measured response) at mid to late decay times where the positive response of a deep conductive target is expected. Hence, it can completely overpower the inductive response and in this case, there is no reliable way to recover the target. A typical IP decay is shown in Figure 22.

Figure 22. A typical IP decay curve.

It is important to try to identify whether this might be a problem in a survey area. The 2.5D inversion software is capable of inverting for the inductive and IP responses jointly, which may be helpful in separating an inductive from an IP response. The presence of IP effects commonly results in poor misfits in an inductive inversion, and is a pointer to their presence. The 2.5D software can also forward model a complex inductive and IP response, which may be helpful in understanding the geo-electric section geometry in such a situation.

SPM effects

Super paramagnetic (SPM) effects are much less common than IP effects. They manifest as late time positive anomalies with a characteristic slow decay rate (1/time). Due to their slow late-time decay, SPM responses can be confused with the responses of deep conductors and vice versa. SPM effects are best recognised by their decay rate, and also by their fast falloff with increased survey elevation. The latter attribute means they are less commonly observed in fixed wing AEM surveys. Typical SPM decays are shown in red, green and blue, and compared with more shallow inductive decays in yellow and pink, in Figure 23.

Figure 23. Typical SPM decays in red, green and blue

SPM effects are usually caused by very fine magnetite accumulations in surface soils or rocks (magnetite deposits). The 2.5D inversion software does not fit this type of anomaly due its slow decay rate. Care needs to be taken that the misfit failure is not caused by the inversion being run at too low a resolution.

Correlated noise

Correlated noise at late time can be a serious problem for 2.5D inversions. The 2.5D inversion sees relatively long wavelength (300 to 500 m) correlated noise at late time as signal, and fits it accordingly. This can produce a series of conductive blobs instead of either a deeper flat lying conductor or, alternatively, no conductor at all. This late time noise can have amplitudes well above noise levels, and has a negative impact on inversion quality. It can be removed by lateral smoothing, but this runs the risk of removing real late time anomalies of similar wavelength. This has been seen in some older surveys where it has been removed in 1D inversions by strong lateral smoothing. The source of the noise appears to be related to loop motion or swing. An example of this type of noise before and after lateral smoothing appears in Figure 24.

Figure 24. Late time correlated noise before (top) and after (bottom) lateral smoothing.

Summary

CDI or 1D inversions are well-established processing strategies for AEM data. Experienced users have learned how to recognise and manage some of the shortcomings of these inversions, and appear to be reluctant to change their practice. However, CDIs or 1D inversions can mislead geologists and/or less experienced users and, as a consequence, result in poor exploration outcomes.

The 2.5D AEM inversion technology developed by Intrepid produces very clean and spatially accurate images of subsurface conductivity in both cross section and in plan (with some post processing) that are mostly free from the problems often seen in CDI and 1D inversions – particularly where 1D assumptions are not met.

2.5D inversions are much less computationally expensive than 3D inversions, and not limited by line spacing.

2.5D inversions can be performed on data from all of the common AEM systems and at survey scale on lines of >100 km in length. Constrained inversions are also an option if there is adequate information about the geo-electrical section.

All of the 2.5D inversion products generated by the Intrepid software come with information on reliability through the delivery of survey and predicted profile misfits at survey resolution.

The Quamby/Dugald River and Kevitsa examples demonstrate that 2.5D inversion products can be used confidently by geologists and geophysicists for orebody targeting and for geological and structural mapping in plan as well as in cross section. These products also facilitate the integrated interpretation of AEM, magnetics, gravity and surface geology.

The Elura example demonstrates that 2.5D forward modelling is an effective tool that can assist with the analysis of target detectability when the explorer is in the AEM survey planning phase.

Older AEM survey data and/or data acquired in areas with conductive cover or other forms of geo-electrical noise (e.g. IP effects) can, however, present problems that require experience and appropriate analytical tools to manage.

Raising awareness of new technologies and encouraging their acceptance can be a challenge for geophysicists working in exploration. Hopefully this article has gone some way towards promoting the use of 2.5D AEM inversion technology and inversion products for orebody targeting and geological mapping.

References

• Emerson, D.W. editor, 1980, The Geophysics of the Elura Orebody, Cobar, NSW., The Proceedings of the Elura Symposium, Sydney, 1980, Exploration Geophysics , 11:4. [Google Scholar]
• Oldenburg, D.W. , Heagy, L.J , Kang, S. , and Cockett, R. , 2020. 3D electromagnetic modelling and inversion: a case for open source, Exploration Geophysics , 51:1, 25–37, doi: 10.1080/08123985.2019.1580118 [Taylor & Francis Online], [Google Scholar]
• Paterson, R. , Silic, J. , FitzGerald, D. and Jakica, S. , 2017. High Accuracy 2.5D AEM Inversion Method for Banded Iron-Formation (BIF) and Other Geological Settings, AusIMM Iron Ore 2017, Paper Number: 85 [Google Scholar]
• Schneider, M.N. and Emerson, D.W. , 1980. Physical Properties of the Elura Prospect, Cobar, NSW, Exploration Geophysics , 11:4, 42–43, doi: 10.1071/EG9804042 [Taylor & Francis Online], [Google Scholar]
• RaicheA., SugengF. , and others, 2008. Developing the Modelling & Inversion Environment, Electromagnetic Modelling Group, CSIRO Exploration & Mining, AMIRA Project 223, Report. [Google Scholar]
• Silic, J. , Paterson, R. , FitzGerald, D. and Archer, T. , 2015. Comparing 1D and 2.5D AEM inversions in 3D geological mapping using a new adaptive inversion solver, 14th International Congress of the Brazilian Geophysical Society, Brazil. [Google Scholar]
• Silic, J. , Paterson, R. , and FitzGerald, D. , 2016. Improved Structural Mapping and Conductive Targeting Delivered by a new 2.5D AEM Inversion Solver, ASEG-PESA-AIG 25th Geophysical Conference and Exhibition Abstracts. [Google Scholar]
• Schmidt, D.L. , 1989. Elura Zinc-Lead-Silver Mine Cobar. Geol. Soc. Aust. Cobar Field Meeting Oct 6-8, 1989. [Google Scholar]
• Wilson, G.A. , Raiche, A.P. and Sugeng, F. , 2006. 2.5D inversion of airborne electromagnetic data. Exploration Geophysics, 37, 363–371. doi: 10.1071/EG06363 [Taylor & Francis Online], [Web of Science ®], [Google Scholar]
• Wyborn, L.A.I. , 1997. Quamby 1:100 000 digital geology, part of Mount Isa geological digital data set,Geoscience Australia. [Google Scholar]