Exploration Geophysics - Volume 31, Issue 1-2, 2000

Magnetic anomalies in the Palaeoproterozoic Willyama Supergroup in the Broken Hill region have a number of different sources, including stratiform magnetite, structurally controlled magnetite and magnetite in igneous rocks. Magnetic sedimentary units are relatively uncommon; most metasedimentary rocks are chemically reduced and host few stratiform anomalies. Exceptions include units within the Paragon Group and volumetrically minor banded iron formation. Structurally controlled anomalies are common across the region and are caused by magnetite in hightemperature shear zones, tectonic fabrics, or in high-strain zones developed along the contact between rocks of contrasting competency. Magnetite formation occurred more than once during the deformational history, during both high-grade and lower-grade metamorphism. Many of the observed linear anomalies coincide with the regional S3 fabric, which formed during upper greenschist to amphibolite-facies metamorphism. Retrograde shear zones are generally magnetite-destructive and form linear zones of low magnetic intensity. Amphibolite has a generally low magnetic susceptibility, although certain units have a high susceptibility, possibly due to alteration as a result of fluid flow. The abundance of structurally controlled anomalies and the relative paucity of stratigraphic anomalies hamper the detailed extrapolation of lithological units beneath cover and means that a ‘magnetic stratigraphy’ is difficult to construct for the Willyama Supergroup. Rather, aeromagnetic data are more useful in the delineation of high-strain zones that have acted as channels for fluid flow, resulting in the formation or destruction of magnetite. Tracing these zones of fluid flow has potential benefits for the reconstruction of mineralising systems and magnetite formed in these zones may have acted as a chemical trap for metals such as gold.

In interpretation of gravity surveys, it is essential to exploit all the information available from drill holes in order to reduce the ambiguity. Accordingly, a new modelling and inversion methodology has been developed to expedite joint geological/geophysical interpretation of gravity data. The key features of the approach are the enforcement of drill constraints (pierce points) and the imposition of density bounds on geological formations and basement. 3D density models are constructed from close-packed vertical rectangular prisms with internal contacts. Prism tops honour topography, so that terrain effects are modelled, not ‘corrected’. Detailed local models can be embedded in regional models to permit fitting of full free-air data, not residual gravity. The geological sense of models is preserved during inversion: the shape and density of homogeneous geological units are adjusted iteratively, subject to the drilling and density constraints.

The methodology is illustrated using data from an advanced exploration project in South Australia. Integrated interpretation of a drilled area has been undertaken in four stages. The first stage entailed construction of a ‘regional’ density model, satisfying gridded gravity data on a coarse mesh over a large area centred on the drill grid. Next, a local density model was created on a fine mesh for the drill grid area, based on drill intercepts and density logs. Thirdly, the detailed density model was inserted into the regional model. Finally, constrained inversion was performed, to adjust the local starting model until a fit to the free-air gravity data was achieved.

Analyses of aeromagnetic and gravity data in the Cooper Basin area of South Australia, reveals varied basement lithologies and structure. Most of the observed magnetic responses arise from sources beneath the Cooper Basin sequence. A prominent northeast-trending structural basement grain is evident. Emplacement of a large igneous mass into the upper crust is interpreted beneath the Patchawarra Trough. Interpreted faulting along the southeastern edge of the igneous body resulted in up to 3.5 km of block uplift, coincident with the Gidgealpa and Merrimelia Ridges. A broad and prominent gravity low in the Nappamerri Trough region is attributed to Late Carboniferous Big Lake Suite granodiorite masses, which have intruded Warburton Basin strata. Intense gravity lows within the broader gravity low are attributed to granodiorite cupolas, which became non-tectonic palaeo-topographic highs during Cooper Basin deposition. Another broad gravity low in the Tenappera Trough region is due to a belt of Early Devonian granitoids that have intruded Proterozoic basement. A number of small magnetic anomalies, in the south of the survey, are interpreted as sourced from skarns above and adjacent to the granitoid intrusions. The only magnetic units above Proterozoic basement in the region are the basalts in the Warburton Basin sequence, which coincide with small magnetic anomalies. Depth to magnetic source modelling correlates well with the known depths of the basalts.

The crustal structure beneath the eastern part of the boundary between the Hamersley and Ashburton Basins near Paraburdoo in the northwest of Western Australia has been investigated previously using seismic refraction and gravity data obtained in the late 1970s. The availability of recently acquired gravity data over the area of the WYLOO 1:250,000 map sheet and a newly established density database of the Hamersley Basin have prompted this study of the western part of the boundary north of the Wyloo Dome.

On WYLOO the Hamersley Basin is characterised by short wavelength and high-amplitude magnetic anomalies produced by banded iron formations. These are in striking contrast to the short to medium wavelength and moderate-amplitude magnetic anomalies that occur over the Ashburton Basin. A well-defined positive gravity anomaly along the contact zone between the two basins can be explained with a model of a southwesterly dipping contact between the higher density rocks of the Hamersley Basin and the lower density rocks in the Ashburton Basin.

Whereas the interpreted supracrustal geology and basin structure largely explain the aeromagnetic and shorter-wavelength gravity responses on WYLOO, they do not account for the regional decrease in the Bouguer gravity values, from +100 gu in the Ashburton Basin to about -600 gu in the Hamersley Basin. Using constraints imposed by the earlier seismic investigations, the observed gravity response has been simulated with a model of a two-layer crust that thins southwards from 30-34 km beneath the Hamersley Basin to 27-30 km below the western part of the Ashburton Basin.

This is in contrast to the southwards-thickening model that was derived from the earlier seismic refraction and gravity investigations to the east. It suggests that there may be a significantly different crustal structure beneath the western and eastern parts of the boundary between the Hamersley and Ashburton Basins.

Aeromagnetic surveys are commonly flown at a constant height above the terrain to minimise the magnetic effects of variable terrain clearance. This is known as drape flying. However, in mountainous regions it is often not operationally feasible to perform a drape survey. Instead, the survey is flown at a constant barometric height and the draped magnetic data are calculated numerically using a level-to-drape continuation operator. Existing techniques for this calculation include the chessboard and Taylorseries methods. An alternative method described here, based on the wavelet transform, approaches the problem by representing the continuation integral using a family of wavelet basis-functions localised in both space and frequency. This allows the generation of a set of coefficients that can be efficiently applied to the wavelet transform of the signal. The wavelet approach can be used for both 1D and 2D signals. If the drape surface is closer to the ground than the barometric survey height, a major difficulty in the drape correction is the control of noise. This is achieved in the wavelet domain by using a locally-adaptive, exponential noise-reduction filter which can be designed based on the wavelet coefficients. The method can be extended in some cases to generate draped images below the ground surface that can be used to sharpen images of magnetic basement in sedimentary basins. The wavelet method is compared with conventional techniques using data from the Edge Hills region in Canada and the Browse Basin in Western Australia. In this study, the wavelet approach combined with the exponential smoothing filter produces sharper images than either the chessboard or Taylor-series methods.

Older airborne gamma-ray spectrometric survey data were often presented in units of counts per second. These data values are dependent on survey and equipment parameters such as detector volume, survey height, and the window energy limits used to measure the gamma radiation. Thus, data values on adjacent surveys may not be directly comparable. This is a critical problem for the interpretation of data over large areas, as it is difficult to reliably relate radiometric signatures in one survey area to those in another.

The conventional solution to this problem is to ‘back-calibrate’ the data from older surveys to ground level concentrations of the radio-elements. This requires a linear transformation of the data that incorporates both a scaling (to accommodate detector volume, flying height and window widths) and a base-level shift (to accommodate inadequate background removal). The usual method of determining these scale and shift parameters is through field measurements using a calibrated portable spectrometer. However, because field work is expensive, an alternative solution is sought.

An alternative approach is to use the differences between gridded survey data values in those areas where the surveys overlap to automatically estimate a base-level shift and scaling factor which, when applied to one of the surveys, minimises the differences in the overlap region. In this way an un-calibrated survey can be merged with a survey already calibrated to elemental concentrations of the radio-elements. Surveys can then be sequentially merged (or ‘levelled’) to produce regional compilations of the data. A similar approach has been used for many years to level aeromagnetic data. However, this sequential approach tends to propagate joining errors and introduce regional trends into the merged data.

A new method overcomes this problem by considering the levelling of all of the grids in the regional compilation as a single inverse problem. Using a two-stage approach, the best base-level shift and scale for each survey grid is estimated. The first stage estimates the best relative shift and scale for each overlapping grid pair which, when applied to one of the pair, gives a least-squares fit to the grid values in the overlap area. The second stage uses the relative shift and scale parameters to estimate an absolute shift and scale for each grid that both honours the relative shift and scale parameters (in a least-squares sense) and brings each grid to the same absolute level as one or more base grids. The method works well as long as the data in the overlapping survey areas has a reasonable dynamic range.

In the CICADA97 experiment a line of simultaneously recording stationary vector magnetometers was deployed from inland NSW, across the east Australian coast, and into the Tasman Sea. The purpose of the experiment was to investigate the effect of electrical conductivity structure near a coastline on natural timevariations in Earth’s magnetic field. Aeromagnetic surveys regularly take place in such coastal areas, and removal of time variations of the magnetic field is a prime task of data reduction. CICADA97 data show that long-period variations of the total field are systematically enhanced near the NSW coast, while spatial patterns of short-period variations (such as pulsations) may be strongly influenced by electrical conductivity structures on a smaller scale, such as bays and inlets.

Textural-based processing of airborne magnetic data is becoming a recognised tool for image enhancement. The variation method of fractal dimension (FD) estimation is a measure of texture that can resolve subtle textural contrasts as well as edge features that are otherwise difficult to discern.

Application of the variation method to synthetic fractal datasets highlighted its ability to distinguish textural contrasts when using local estimates of FD. The variation method was also able to enhance thin linear anomalies, linear ramp anomalies and sinusoidal ramp anomalies contained in synthetic datasets. The variation method clearly resolved these features even in the presence of Gaussian noise. The results demonstrated that smaller window sizes will more effectively discriminate and enhance edges.

Application of the variation method on the Ghanzi-Chobe aeromagnetic dataset highlighted features that were not resolved in the greyscale total magnetic intensity image. Comparison of the variation method with standard derivative-based enhancements showed that the variation method enhanced similar trends but with greater clarity. It also enhanced features that were not revealed by horizontal- and vertical-derivative images. Whilst the variation method needs to be applied to data from a wider variety of magnetic regimes, the results suggest that the technique can provide useful information not available from conventional enhancement processes.

Airborne and ground magnetic surveys are a primary exploration tool in the search for kimberlites. Interpretation of magnetic data provides direct information on possible pipe magnetic signatures and indirect information about the structural setting of the area.

Magnetic anomaly screening in an area of highly active magnetic relief and magnetic texture such as an Archean shield area or volcanic terrain is very difficult. Conventional techniques such as stacked profile interpretation and modelling/inversion are almost impossible, as there is so much interference and overlap between adjacent anomalies. There is also the additional problem of distinguishing possible kimberlite sources from other sources with very similar magnetic signatures. However, the magnetic signature of kimberlite pipes is band limited so enhancing wavelengths of interest plays an important part in interpretation.

Semi-automatic interpretation, using a combination of specialised filters and analytical techniques can provide objective information on anomaly attributes. These can be used in conjunction with the geological setting of the area to prioritise anomalies for follow-up. The application of separation or layer filters allows deconvolution of the causative sources around a particular mean depth. Plotting of gradient maxima/strike symbols maps circular or elliptical anomalies as closed clusters. 3D Euler deconvolution provides first-pass depth estimates and an indication of the nature of the source. Finally, matched filtering using a cylinder model can identify roughly circular anomalies over a broad range of amplitudes and wavelengths.

Anomalies selected in the first phase are refined by automatic profile analysis and the total anomaly attributes compiled into a database. Finally targets are assessed in their geological context, using available geology and the magneto-tectonic interpretation of the area.

Low-noise images of the U/Th ratio can be obtained from aerial gamma-ray survey data by using the maximum noise fraction (MNF) method, which applies statistical linear algebra operators to most of the 256 channel raw spectral data of a survey. With reduced noise, U/Th data may be better used to map geological variations and explore for minerals such as U, Th and Sn. Data from a small (120 line-km) survey in north Queensland are used to demonstrate the method and confirm the reliability of the detail in the U/Th map. The map is compared with those derived using conventional processing methods and using noise adjusted singular value decomposition (NASVD). Statistical tests of selected areas and profile displays show all methods produce similar U/Th ratios on a broad scale, but only the MNF method enables small-scale features to be identified. A small dam in the study area, which being waterfilled has low count-rates, gives anomalously high U/Th ratios. This dam is seen in the conventionally processed U/Th image as a small area of high ratios, is difficult to see in the NASVD image, but is shown with its correct size and shape in the MNF image.

A method for reducing a grid of total field magnetic data to the pole, where the field and magnetisation directions vary, was published by Arkani-Hamed in 1988. He called the method differential reduction-to-the-pole (DRTP) because the variations in direction are treated as perturbations about mean directions and the problem is solved iteratively. There appear to be advantages in applying this (or a similar method) to large regional magnetic grids, particularly at low magnetic latitudes. Nevertheless, it appears to have been rarely used. There are two problems with DRTP. Firstly, the algorithm calls for 12 two-dimensional arrays (5 real and 7 complex) so an implementation capable of dealing with a 4000 × 4000 data grid on a PC or small workstation requires careful coding. Secondly, as noted by Arkani-Hamed, the method is unstable at low inclinations where, like the standard wavenumberdomain method, it selectively amplifies noise (as well as anomalies) in the magnetic north-south direction. The problem of low inclination stabilisation is addressed using a similar technique to that used in Geosoft’s MAGMAP to stabilise its wavenumberdomain RTP filter. This introduces a pseudo-inclination into the denominator of the filter transfer function, which is increased from the true inclination in order to reduce the amplification of northsouth wavenumbers. It is not obvious that this modification should work with the DRTP algorithm because of the iterative corrections being applied, but tests with synthetic data show that it has a very similar behaviour to the uniform direction algorithm. As an example of its use the low-inclination DRTP is applied to a large regional magnetic dataset from Brazil, where the inclination varies from -40° to +20°.

Aeromagnetic surveys in rugged terrain such as the Turner Syncline in the Hamersley Basin, Western Australia, are often flown as loosely draped surveys within operational safety limits. Loosely draped surveys are a compromise between retaining good spatial resolution of small magnetic sources, maintaining fairly constant terrain clearance, and operating the aircraft in a safe manner. The main problems in these surveys relate to lines flown in opposite directions causing a variable and biased terrain clearance. Over deep narrow valleys, the magnetic source depth increases, anomaly amplitudes are reduced and spatial resolution of small sources lost. Over narrow ridges, the magnetic source depth decreases, anomaly amplitudes are increased and resolution of small sources much improved. The high wavenumber content of lines flown in opposite directions may be very different, resulting in zones of marked texture differences, especially for race-track flight paths.

The ideal drape correction involves projection of the field onto a surface with constant terrain clearance, preserving the full bandwidth of the data. Several grid- and profile-based drapecorrection algorithms have been tested on the Turner Syncline with varying degrees of success. The best results were obtained with a two-pass correction approach. A first order drape correction, using a modified chessboard method was applied to the located data prior to tie-line levelling and microlevelling. The simple ‘slide-rule’ filter of Cordell (1985) was replaced by a technique of continuing different lags of the input data to avoid large downward continuation distances. A second pass of the ‘normal’ chessboard drape correction was then applied to the gridded data to further refine the correction. The results show that the drape corrections have been effective in reconstructing anomaly amplitudes, but less effective in preserving the texture of the data. Conventional approaches to drape correction appear unable to restore the missing high-wavenumber information content caused by extreme height variation.

Analysis of satellite gravity data offers an opportunity to rapidly evaluate the sedimentary structure of frontier basins. The accurate and evenly distributed measurements of the gravity field determined from satellite orbits contain information about the bathymetry, sediment thickness and crustal structure of the world’s oceans. These data have been used by various authors, e.g., Smith and Sandwell (1994), to predict bathymetry for ocean basins where sediment cover is thin and where crustal structure is well known. In these areas, over a short wavelength band (15-160 km), the gravity field is highly correlated with the bathymetry. Along continental margins, however, the gravity field is complicated by sedimentary basins and changes in crustal structure.

The method presented here differs from previous techniques for analysis of satellite gravity data by inverting simultaneously for water, sediment and crustal thickness instead of decomposing spectral bands in the gravity field. The inversion is constrained by interpretations of seismic and other data, and by assumptions about the nature and spatial variability of the interfaces. The seabed, basement and crust/mantle boundaries are defined by a series of triangular facets, whose size varies as the amount of constraining data changes. The sediment/basement boundary and the base of the crust are defined by larger facets than those defining the bathymetry. Conditioning equations smooth the interfaces, make the depths tend towards the input values and ensure that the interfaces do not intersect. In areas of sparse bathymetric and crustal structural control a degree of ambiguity between longwavelength shallow structures and short-wavelength deeper features inevitably remains.

This technique has been used to identify the presence and location of several sedimentary basins in an area 1250 by 1000 km west of New Zealand.

There are a number of estimators of a long-memory process’ long-memory parameter when the parameter is assumed to hold constant over the entire data set, but currently no estimator exists for a time-varying long-memory parameter. In this paper we construct an estimator of the time-varying long-memory parameter that is based on the time-scale properties of the wavelet transform. Because wavelets are localised in time they are able to capture the time-varying statistical properties of a locally stationary longmemory process, and since wavelets are also localised in scale they identify the self-similarity scaling behaviour found in the statistical properties of the process. Together the time and scale properties of the wavelet produce an approximate log-linear relationship between the time-varying variance of the wavelet coefficients and the wavelet scale proportional to the local long-memory parameter. To obtain a least-squares estimate of the local long-memory parameter, we replace the time-varying variance of the wavelet coefficient with the sample variance of the wavelet coefficients computed over the so-called ‘cone of influence.’ That is, we use only those wavelet coefficients whose time index falls within the support of the wavelet basis function in order to compute the local sample wavelet variance. To test the empirical properties of our estimator we perform a number of Monte Carlo experiments. We find the wavelet-based estimator of the local long-memory parameter to have empirical properties similar to other waveletbased estimators of the long-memory parameter for globally stationary long-memory processes. For processes where the longmemory parameter suddenly changes, the wavelet-based estimator again performs well, only exhibiting an elevated positive empirical bias at points in time right before the long-memory parameter increases, and a negative bias immediately after the change. The wavelet-based estimator of the local long-memory parameter is demonstrated using vertical ocean shear data.

The enhancement of magnetic data for qualitative interpretation involves manipulating magnetic relief and magnetic texture. Magnetic relief consists primarily of anomaly amplitude and is relatively objective. Magnetic texture consists of shape, size and continuity of adjacent anomalies and is more subjective. Conventional filtering responds primarily to amplitude variations within the data and high-amplitude anomalies often mask more subtle anomalies of interest. Changes in geology, or specifically rock magnetisation, within a survey area cause changes in both texture and relief. When rock magnetisation is weak, anomalies are subdued and particular filtering and enhancement methods are required. However, the effects of these on areas of greater magnetic relief must also be considered.

Three fundamentally different approaches to enhancing subtle anomalies have been implemented and tested on a wide range of sedimentary basin and low magnetic gradient areas. The first approach uses separation or layer filtering to deconvolve the effects of magnetic sources around a mean depth. The second uses texture filtering using grey-level co-occurrence matrix (GLCM) based filters and the third uses the gradient tilt angle. The responses of these filters are illustrated using a regional aeromagnetic dataset from the Proterozoic Arunta block in the Northern Territory of Australia.

Results show that each method has its advantages and limitations. Each shows different amplitude and wavelength responses for a given dataset, but all three are relatively broadband compared with a conventional filter such as a first vertical derivative. No single method performed well on all areas of the test dataset. The GLCM filters were almost pure texture with very limited tonal content. The gradient tilt angle provides good texture content, but retained more tonal content than the GLCM filters. Combining filters with different bandwidth such as the separation filter and tilt angle can be very effective and provides the best results.

Conventional gravity measurements are only of the maximum magnitude of the gravity acceleration vector. This is because the gravimeter is aligned with the gravity vector and not in a selfconsistent reference frame. However, when a high-resolution geoid model is available, vector gravity data can be computed. The geoid model is used to compute the deflections of the gravity vector from the ellipsoidal normal, thus providing direction to the measured magnitude of gravity acceleration. This allows the components of the gravity vector to be computed from the gravity measurements. Since the geoid is mostly generated by deeper Earth structure, the components of the vector gravity anomaly enhance the information on deeper mass variations above that contained in the conventional gravity anomaly. This approach has been applied to the Australian gravity database and shows a number of linear features not clearly evident in conventional Bouguer gravity anomaly maps.

Airborne electromagnetic (AEM) surveying is an important exploration tool because it can map conductivity variations over large areas at a fraction of the cost of ground survey methods. Using rapid but approximate techniques, large volumes of data may be processed to show the variation of conductivity with depth beneath the survey. These approximate methods work well in regions with horizontal layering, but in certain circumstances they can imply the presence of false conductors in the vicinity of 2D and 3D structures. By comparing the AEM response of several 2.5D models, each of which contains a lateral conductivity contrast, we show that artefacts associated with conductivity contrasts can imply the presence of a false conductor when flight direction is towards the area of greater conductivity. When flight direction was away from the area of greater conductivity artefacts associated with the lateral conductivity contrast implied a false resistor. These artefacts were of sufficiently high magnitude that they masked the response of a genuine conductor (1.0 Ω.m) at a depth of 50 m. We show that multiple-component data sets utilising the inherent directional dependence qualities of AEM prospecting systems can be used to minimise interpretational errors in the presence of lateral conductivity contrasts.

Analytical expressions for the electromagnetic field components at an arbitrary location in a half-space are presented for a vertical magnetic dipole source on the surface of the Earth. The Hankel transforms required in the computation are enumerated using expressions based on the modified Bessel functions of the first and second kind. For surface electromagnetic fields, the Bessel function expressions are shown to consistently produce results that are up to four orders of magnitude more accurate than the results commonly obtained using digital filters. In addition, initial speed tests reveal that considerable computational time is saved compared with the use of numerical routines based on digital filters. It is expected that the increased speed and accuracy of the analytical solution will find application in forward modelling and inversion routines where a half-space is considered to be a suitable model and Hankel transforms need to evaluated many times.

Electromagnetic (EM) coupling of frequency-domain induced polarisation (IP) data has been the subject of many studies, and a number of ‘de-coupling’ procedures have been devised. However, there has been far less emphasis on coupling in the time domain, the normal approaches being to wait until late times and assume the EM contribution is insignificant or, less frequently, to invoke a Cole-Cole model to account for the EM-coupling response. A fast and simple procedure has been devised for suppression of EM-coupling effects in time-domain IP data. The essence of the approach is to represent the EM-coupling as a half-space decay. The half-space resistivity (EM apparent resistivity) is adjusted via inversion until the fit to the observed transient voltage decay is optimal in the least squares sense. The EM voltages associated with this best-fitting EM half-space decay are then subtracted from the measured voltages to yield a de-coupled ‘IP transient’. Transients that are well represented by an EM half-space decay are deemed ‘non-responsive’ in the context of IP. Transients that deviate markedly from an EM half-space decay are indicative of high apparent chargeability. The application of the new procedure is illustrated on dipole-dipole IP data from the Yandal greenstone belt of Western Australia.