ASEG Extended Abstracts - 24th International Geophysical Conference and Exhibition – Geophysics and Geology Together for Discovery, 2015

Slimline geophysical logs are frequently used worldwide in iron ore exploration because they provide key data for ore evaluation. Application can be limited by the fact that many formations associated with iron ore deposits are friable, increasing the occurrence of borehole collapse before geophysical logs can be obtained. A cased-hole correction scheme for density logs based on an existing technique developed for oil and gas (C-thru) has been developed. The technique enables accurate and reliable near-spaced density measurements in a cased-hole (or through-rods) environment by recharacterising the response equations of the density tool to account for the casing or rods. The method effectively treats the casing or rods as part of a “modified” density tool. The method means that it is possible to obtain quantitative data when the logging tools are run inside the drilling rods. The application of this technique minimizes the risks associated with logging unstable open holes in iron ores, and can reduce costs and operation times.

The revolution in computational power and integrated geoscience modelling approaches over the past decade is set to accelerate. Geoscientists (collectively; geologists, geometallurgists, mineralogists, geochemists and importantly geophysicists) will lead a transformation in the way the mining value chain can be conceived, evaluated and operated. The emerging capability to process large numbers of stochastic images of the mineralised system - each characterised by rich multivariate information - will allow better decision-making about alternative value chain configurations in the face of uncertainty. While this decision making has obvious implications for capital decisions in project evaluation, it has equally dramatic possibilities for real-time optimisation of existing operations. The advent of more flexible, highly configurable and in many instances automated and intelligent approaches to mining and mineral processing is perfectly timed to enable these inputs to deliver step-changes in value.

We present an overview of the Adrok radar scanning technology and describe experimental results that suggest that ground penetrating radar can be utilized to much greater depths in selected environments than commonly assumed. High frequencies were found to penetrate very little, but the low frequency component had very low losses. Results were analysed to estimate the skin depth and interpreted in terms of a constitutive model incorporating Maxwell’s equations with conductivity and polarization losses. To explain these results we hypothesize that moisture penetrates limestone only relatively superficially and once the outer wet layer is penetrated the conductivity and therefore the losses are greatly reduced. In a second experiment we successfully detected the reflection of the radar pulse from a body of water through 350m of rock. A numerical simulation of the model confirmed that these results do not contradict theoretical expectations for dry limestone.

Mineral deposits frequently have complex structures that can only be resolved by 3-D inversion of resistivity and I.P. data. A nonlinear optimisation routine is commonly used to create a 3-D model from the measured apparent resistivity and I.P. data. It is particularly important to be able to assess the reliability of the anomalies seen in the inversion model before further tests are conducted. In this paper, we examine the model resolution (MR) and volume of investigation (VOI) approaches in determining model reliability. The MR method produces sections that are easier to interpret but more computationally intensive that puts practical limitations for models with more than 50000 cells. The VOI method can be used for any data set where an inversion can be carried out, but produces sections with more complex patterns and prone to local artefacts. Either method should be used for any interpretation to discern anomalies that are likely to be supported by the data.

The Darling Fault Zone (DFZ) is one of the largest lineaments in the world, mapped over approximately 1000km. It is a long-lived feature with imprints of multiple deformation phases with multiple orientations since the Archean. Although it is still topography forming in some areas, and therefore must have recent activity, its seismic quiescence reduces the perceived need for scientific investigation into the extent and physical properties of this crustal scale fault. Seismic activity is common in the south west of Western Australia and evidence suggests these are located on faults that communicate with the DFZ. It is therefore paramount to have more detailed understanding on the fault architecture and the role of fluids in lubricating aseismic slip.

Magnetotelluric (MT) data are acquired along transects across the Perth Basin and the western margin of the Yilgarn Craton providing deep, high resolution data about the electrical structure of the DFZ. In this contribution we focus on the interpretation of the data, details on the acquisition and analysis are presented in the poster session.

Using impedance tensor analysis and 2D modelling techniques, we map the DFZ to the base of the crust, confirming it as a lithospheric feature. We reveal a complex pattern of deep-seated conductivity associated with the foot wall of the DFZ that persists to depth. Resistivity models are used to estimate porosity on the DFZ, identifying a more complex internal structure for the DFZ than generally considered, with important implications for fluid transport in the basin.

The global mineral exploration industry is currently perhaps a decade or two into the most important transition in its several thousand year history - the move from a world where discovery was primarily about surface prospecting (in various form) to a world where important future discoveries will be blind, with little or no surface expression. This transition is, and will continue, profoundly influencing all aspects of our industry, from financing and government policy through to targeting methods and detection technologies. Geophysics will play an increasingly central role in the exploration industry as this transition progresses, as it did in the analogous transition in the history of the petroleum exploration industry about a century ago. However, this future for geophysics will not simply be doing more of the same - the relationship of geophysics to the exploration industry will need to evolve significantly to enable cost-effective exploration performance in this future world. Some of the required key areas of development include; a) better characterisation of mineral systems at multiple scales from the continental to the deposit, b) improved integration between geological and multi-parametric geophysical observations at multiple scales, c) improved capabilities to image critical deep-seated ore-controlling structures and perhaps metal-enriched deep source regions which are cryptic in near-surface data and d) more specific rather than just more sensitive detection technologies, which reduce the usually high false-positive rate of geophysical targets. An important strategic enabler for these required advances will be ever increasing access to supercomputing capability. However, the potential of entirely new physical techniques cannot be overlooked either, with Muon tomography having recently been applied to mineral exploration for the first time.

We present a case study whereby unconstrained magnetic modelling accurately defined the altered host lithology of the Kintyre Uranium deposits, verified by detailed geological modelling. The Kintyre Uranium deposits are hosted by a sequence of iron and carbonate rich meta-pelites, which makes it an ideal target for magnetic prospecting. As part of the resource definition, magnetic modelling and geological modelling were performed over the Kintyre deposit independently. In the process of further refinement of the magnetic model through incorporation of geological constraints it was determined the two models were already highly complementary and further modelling was not warranted, particularly at the resolution of the magnetic data. This case study demonstrates that in some geological environments, unconstrained geophysical models can adequately map stratigraphy & structure for drillhole target generation.

A long period magnetotelluric (MT) survey, comprising 39 sites over an area of 270 by 150 km, has identified partial melt within the thinned lithosphere of Quaternary Newer Volcanics Province (NVP) in southeast Australia. MT inversion models reveal several important tectonic features and unravel critical information about the tectonics of the area. The models have imaged a conductive anomaly beneath the NVP at -40-80 km depth, which is consistent with the presence of 1.5-4% partial-melt in the lithosphere. The conductive zone is located within thin juvenile oceanic lithospheric mantle, which was accreted onto thicker Proterozoic continental lithospheric mantle, suggesting that the NVP origin is due to decompression melting within the asthenosphere, promoted by lithospheric thickness variations in conjunction with rapid shear. In addition, inversion modelling shows that there is a conductivity contrast across the Moyston Fault that suggests the transition from Proterozoic continental lithospheric mantle under the Delamerian Orogen to the Phanerozoic lithospheric mantle under the Lachlan Orogen.

ZTEM data acquired across the Humble magnetic anomaly of almost 30,000 nT were analyzed for the presence of a magnetic gradient response and the effects from elevated magnetic susceptibilities.

The response of moving the receiver coil through the magnetic-field gradient peaks at 0.01 Hz and drops off strongly with frequency. Lacking information about the field strength at the base station precludes the comparison of amplitudes between computed gradient responses and the survey data, but the comparison of response shapes suggests that the gradient responses are too small to have a noticeable effect on the survey data.

The 3D inversion of the magnetic survey data indicates magnetic susceptibility values as high as 2.0 (SI). Forward-modeling the ZTEM response for these K-values combined with resistive half-spaces indicates that the response amplitudes and shapes strongly depend on the background resistivities. Ignoring the elevated K-values during an inversion can result in the underestimation of conductivities and other artifacts, such as the mapping of patterns that resemble crop circles. For an environment such as Humble, with deep-seated zones of elevated K-values, the shallow inverted conductivity structure appears to be reliable, but the deeper structure should be interpreted with caution.

Jointly inverting different data sets can greatly improve model results, provided that the data sets are sensitive to similar features. Such a joint inversion requires assumed connections between the different geophysical data sets, which can either be of analytical or structural nature. Classically, the joint problem is expressed as a scalar objective function that combines the misfit functions of all involved data sets and a joint term accounting for the assumed connection. This approach has two major disadvantages: Firstly, by aggregating all misfit terms a weighting of the data sets is enforced, and secondly, false models are produced, if the connection between data sets differs from the assumed one. We present a Pareto efficient multi-objective evolutionary algorithm, which treats each data set as a separate objective, avoiding forced weighting. The algorithm jointly inverts one-dimensional datasets from different electromagnetic techniques and also treats any additional information as separate objectives, rather than imposing them as a fixed constraint. Additional information can include, for example a priori models, seismic constraints, or well log data. Statistical analysis of the final solution ensemble yields an average one-dimensional model with associated uncertainties. Furthermore, the shape and evolution of the Pareto fronts is analysed to evaluate dataset compatibility and to judge if the assumed connection between datasets was valid.

Density is one of the fundamental physical properties required in a mining operation, underpinning the calculation of ore tonnages and thus metal produced. The resource density model captures this information, but is often based on a relatively sparse collection of density measurements. Gravity data are a direct reflection of the true distribution of subsurface density, and can be used to improve the resource model. The example of the Ravensthorpe nickel laterite mine illustrates the improvement in the resource density model that results from combining high resolution surface gravity with the set of borehole logged density readings.

Muon geotomography is a new geophysical imaging technology that creates 3D images of subsurface density distributions. Similar in concept to computed tomography scanning, muon geotomography uses naturally occurring cosmic radiation that gets attenuated when traversing matter. Cosmic ray muon data were acquired in the Pend Oreille Zn-Pb mine in Metaline Falls, Washington State, USA without prior knowledge of the presence or absence of ore bodies. The resulting 3D density distribution indicated a substantial volume of rock with higher density than the host stratigraphy above the survey location. Subsequently, a model of existing ore shells based on drill core data was provided and a simulation of the expected muon tomography data was found to be consistent with the muon geotomography measurements. This is the first blind test demonstration of muon geotomography applied to mineral exploration.

MT and ZTEM data were inverted with a number of 2D and 3D algorithms to recover the subsurface conductivity structure of an area of interest. A 2D inversion algorithm was used to model the magnetotelluric TM and TE mode impedances and the ZTEM tipper data, separately. The derived conductivity-depth sections don’t show much agreement, possibly indicating the conductivity structure of the area to be highly three-dimensional.

A 3D inversion algorithm was used to invert the MT and ZTEM data, separately and jointly. Overall, there is good agreement between the derived conductivity structures. This suggests that a joint inversion can extract successfully the combined subsurface conductivity information from the two data sets.

We have developed and tested code to optimise electromagnetic (EM) sensors to improve performance of the ARMIT B field induction coil sensor at desired frequencies. We aim to use the optimised parameters to develop a compact air core transmitter, which will form the basis for developing a compact ferromagnetic core transmitter. Techniques for optimising induction coil sensors are well established in literature and use analytical equations for the objective and constraint functions. Alternatives for EM sensor design are also well documented. In contrast, the design of compact transmitter systems needed for portability or in boreholes have limited discussion in the literature and have many more design constraints than sensors. Our ultimate intention is to use established sensor optimisation techniques to build a compact transmitter with sufficient magnetic dipole moment.

To optimise an ARMIT induction current sensor we develop the algebraic expression for the total internal sensor noise to use as a constraint function. The objective function is the weight of the sensor. We aim to achieve noise goals of 1 pT / and 1 fT / frequencies of 1Hz and 2 kHz, respectively. 1 Hz was chosen because that is a common base-frequency for conductive sulphide exploration and 2 kHz was chosen as being appropriate for nuclear magnetic resonance investigations. We use numerical non-linear constraint optimization techniques to predict a target noise level of 1 pT at 1 Hz. , At this stage we predict the best 2 kHz sensor to have 4 fT noise at 2 kHz. This was based on existing dimensional and weight constraints on the induction coil sensor. We introduce an analogous method of transmitter optimisation using transmitter dipole moment as the objective function.

As a consequence of diminishing shallow mineral resources, the exploration industry has turned its focus to deeper targets. For this reason, the magnetotelluric (MT) method has gained much attention due to its unique penetration in regions of thick cover sequences. As the setting and geometries of mineral deposits are often complex, 3D models are required for their interpretation.

However, there has been little critical analysis of the ability of 3D MT surveys to recover structural geometry. A comparison of synthetic model responses demonstrate that while MT is greatly sensitive to conductive and symmetrical bodies at depth, its resolution for detecting finite 3D bodies is significantly reduced under conductive regolith cover. Although 2D inversions can recover the geometry of finite conductive bodies, it is possible to successfully interpret 2D survey data using 3D inversion algorithms. Utilising all components of the impedance tensor, off-profile 3D conductive structure can be obtained from 2D survey data alone.

A manual litho-structural interpretation of airborne magnetic, radiometric and digital elevation model (DEM) data over the Nolans Bore rare earth element (REE) deposit, in northern Australia, has identified additional REE targets in the area. These targets were compared to automated targets generated using a Levenberg-Marquardt neural network (LMNN) analysis of the data.

A number of different quantitative analyses were performed: one with only the geophysical data as an input and one that included the structural interpretation. The geophysical data needed modification for use in the LMNN algorithm, as the known deposit is anomalous in its absence of magnetite and a corresponding magnetic anomaly. Therefore, a shallow depth slice was extracted from a 3D inversion model of the magnetic data and the magnetic susceptibility values were inverted to form anomalies over non-magnetic regions. A Th/U ratio was calculated from the measured single radioelement responses. The structural interpretation was modified to incorporate only those faults oriented in the direction of the controlling structures at the known deposit. DEM data were used as a mask to ensure that targets were found in low-lying areas where the radiometric data is influenced less by topography or masks the underlying bedrock.

It was found that the interpreted targets closely match the predicted targets, but that the predicted targets yielded smaller, more specific locations for follow-up work.

There is a strong need to develop real-time imaging technologies to enable the driller to ‘see’ the subsurface structures ahead of the drill-bit and around the borehole during borehole drilling. One of the ways to realise such imaging while drilling is to use borehole radar (BHR) techniques. In this paper, a conventional non-directional mono-static BHR will be evaluated for its forward-looking capability by using the data collected at an abandoned mine site at Brukunga, South Australia. Here we demonstrate that the conventional BHR can be electrically coupled on to a conductive wire or drill-rod whilst a guided wave is induced along the axial wire or drill string making it possible for imaging ahead of the drill-bit by integrating the BHR with the steel drill string. The drill-rod ahead of the BHR acts as a forward-looking antenna. When the guided wave travels to the end of the drill-bit, part of the energy is reflected by the drill-bit and the remaining energy radiates in front of the drill-bit, and is reflected by the geological/electrical discontinuities, recorded by the BHR. The forward-looking capability of the BHR is about 2-6m in the tested borehole section.

All EM transmitters, sensors and data acquisition systems have bandwidth limitations. Transmitters have upper bandwidth limitations due to finite slew rate issues, and systems have a lower bandwidth set by the base frequency used. Receivers and data acquisition systems ideally should have a flat bandwidth response that spans the transmitted signal bandwidth. The data acquisition system should sample fast enough to capture the highest frequencies of interest, with anti-alias filters to prevent data contamination from unwanted signals. Sensors however may have physical bandwidth limitations, for example fluxgates and feedback MT sensors may have an upper corner frequency of a few kHz, and ARMIT and feedback MT sensors have lower corner frequencies in the sub 1 Hz range. In many cases, the sensor corner frequency can be mathematically described as a single- or multi-pole response. In this case, it is possible to exactly deconvolve the data to exactly correct for the sensor imperfection. A limitation of this process is that noise as well as signal may be amplified in this correction process. Without correction, data may be incorrectly modelled or interpreted. This paper illustrates the correction of fluxgate (mostly a time delay of hundred or more microseconds), ARMIT 2 (where a significant but exact correction is required), ANT23 feedback and 3D3 dBdt data.

The Basil Cu/Co deposit comprises a 26.5Mt JORC-compliant inferred resource of copper and cobalt, grading 0.57% Cu and 0.05% Co. It lies in the Harts Range, central Australia, within the Riddock Amphibolite of the Irindina Province. The deposit coincides with a prominent anomaly in aeromagnetic data. Intersections of mineralisation at depth follow the magnetic anomaly trend. Analysis of drilling within the mineralised zone determined a spatial association between pyrrhotite with high magnetic susceptibility and chalcopyrite, with no other significant magnetic mineralisation present.

A study was commissioned to examine if geophysical inversion could predict the distribution of mineralisation, using pyrrhotite as a proxy for chalcopyrite, from the surface to the drillhole intersections, as well as predicting further mineralisation at depth or in the vicinity of the deposit.Commercial software was chosen for performing geostatistical analysis, 3D geological modelling, forward modelling and stochastic inversion.

Petrophysical data from core and information on mineralisation from drilling were used to constrain 3D geological modelling of mineralisation based on domain kriging of susceptibility data and sulphur assays. Magnetic data was conditioned for inversion in Intrepid software. Sensitivity testing of results to source depth and distribution was performed using 3D forward modelling. Alternative 3D geological models were tested during inversion for their behaviour and adherence to observed drilling data and the limitations imposed by the sulphide distribution in the geostatistics.

The resulting initial geological model had known property voxels from drilling fixed and surrounding property voxels locally interpolated from kriging. Using these as a seed, geophysical inversion was performed alternating property and lithology inversion, until a desired minimum misfit with the observed magnetics signal was reached.

The new predicted mineralisation distribution was compared with estimated mineralisation shells from conventional geostatistical modelling and found to be in good agreement, with reliability increasing closer to the surface.

Predictions of mineralisation at greater depth and beneath weaker anomalies were more diffuse and showed less tendency to change during inversion, and were limited by the flight acquisition specifications. Deep targets have not yet been tested by follow-up drilling.