ASEG Extended Abstracts - ASEG2013 - 23rd Geophysical Conference, 2013

This paper describes a field experiment designed to test various forms of marine broadband acquisition and their appropriate processing techniques. These methods are compared against a traditional “shallow tow” (8 m) cable depth and de-phasing processing flow. We take particular interest in a slanted cable acquisition technique and other ways of creating “notch diversity” with a marine streamer. We analyse the resultant data with displays, spectra and frequency split displays.

Overall the experiment shows that we can model the cable ghost response of the system well enough to understand the response of real data. This allows the application of deterministic inverse operators to not only “de-phase” and re-datum the data but also frequency shaping the output data closer to the spectrum of the input pulse. This is particularly important when the cable is slanted in an oblique manner relative to the sea surface. However, noise at frequency extremes will limit this process and other shaping methods. The final results show remarkably similar end results despite quite different acquisition methods.

In the Southeast Asia offshore exploration, shallow reefs and channels, widely spread volcanic rocks and basement fracture system are some of the major challenges in seismic imaging. Geo-mechanical modeling, reflectivity inversion and TTI/HTI modeling in basement are introduced in this area. Together with the conventional reflection tomography, they generate high resolution velocity model for PSDM thus provide much needed imaging uplift.

A mine is a difficult geophysical environment to work in because of the presence of large amounts of noise. For seismic techniques in particular, the presence of drilling, blasting, shearers cutting rock, pumping and other activities leads to high levels of background seismic noise. But this seismic noise interacts with the rock in the same way as energy from an active seismic source would do, so it could, in principle, be used to image the rock.

Most sources of noise -- drills, shearers, pumps -- are continuous, so there is no well defined “shot time” and no well-defined direct or reflected arrivals in measured traces. However, a process of coherence-weighted cross- correlation across an array of sensors can produce a set of relative travel times, which can be used for tomographic imaging.

Experiments have been done in a number of coal mines, using a coal shearer as a source of energy, and recording the signal on arrays of geophones, installed either in roadways underground, or on the ground surface above the mine. This data has been processed using various techniques to try to extract velocity and attenuation information. Results indicate that the technique is successful in extracting relative arrival times across an array from continuous noise.

Conventional IP is not the only technique that is sensitive to chargeable material. Any electromagnetic method applied in the presence of chargeable material will be affected. Unfortunately, the effects are often hard to recognize in the data. For the particular case of coincident loop time-domain EM data, negative transients - soundings with a reversal in sign of the received fields - are diagnostic of chargeable materials. This property can also be extended to center loop systems, including many airborne systems. Negative transients are commonly observed in airborne TEM systems, such as Fugro’s AeroTEM system or Geotech’s VTEM system.

We develop an inversion methodology to attempt to recover a three dimensional distribution of chargeability from observations of negative transients in airborne time- domain electromagnetic data. Forward modeling of chargeable targets is performed directly in the time domain, and the sensitivity of these data to the presence of chargeable material is derived. The methodology is applied to a synthetic data set. Areas of future work and potential problems are discussed.

The southern extension of the Thomson Orogen near the New South Wales border in southern Queensland is an underexplored greenfields area with mineral potential. The uncertainty surrounding cover thicknesses in the area leads to increased exploration risk. Calculation of a depth to basement surface lowers risk for greenfields explorers and is a valuable contribution to regional geological understanding.

Available drillholes show basement depths vary from about 100 m to over 3 km in the study area but data is too sparse to create a reliable surface. A combination of different automatic depth to basement techniques, including Euler deconvolution and Naudy, were used to create a depth to basement surface. These techniques were preferred due to a regionally extensive high frequency, low amplitude magnetic signature attributed to shallow sources in one of the cover sequences in the study region.

A combination of Geosoft’s Located Euler and standard Euler Deconvolution were used to calculate depth to basement solutions. Naudy depths were also compared to the Euler solutions, particularly in areas where the drill holes indicated the basement was relatively shallow (where smoothing associated with gridding can cause overestimation of source depths) or dyke-like features were present.

Seismic data available in the north of the study area was interpreted and used as a secondary quality control check (along with the drill hole data) on the basement surface. The final depth to basement surface defines an area of shallow basement in the south-west of the study area.

Voxel inversion is a well-established method for constructing a physical property model from geophysical data. However, a limitation on Cartesian voxel inversion is that the voxel earth model is restricted to prism shaped elements which do not conform to geology and can lead to numerical artefacts. Octree and unstructured meshes have been used to overcome this limitation, but both add significantly to either the number of voxel elements, or to the complexity of the representation. We propose an alternative method which maintains much of the simplicity of the regular Cartesian voxel model while allowing very accurate geometric representation of geological surfaces: the Cartesian cut cell (CCC) method.

A significant geologic surface in most voxel inversions is the topography, which is very poorly represented by voxels. There are many other common geologic sub- surfaces including faults, contacts, unconformities, mineralized zones, etc. We demonstrate the necessity of improving the conventional voxel representation to ensure accurate geophysical modelling and how this is achieved with the CCC method. We also show the value of the CCC method in constrained inversion. These examples demonstrate that the CCC method allows accurate representation of geologic surfaces with a minimal extension to the simplicity of a conventional voxel model.

Shale plays have revolutionised the oil and gas industry in North America and exploitation of these kinds of plays is steadily gathering pace in other parts of the world. Because hydrocarbon bearing shales usually have insufficient permeability to allow significant flow to a well, production from these unconventional reservoirs comes with unique challenges. Optimizing recoverable reserves from shales requires strategic placement of horizontal wells: placing the well in the best areas, drilling the lateral in the proper direction and keeping the lateral portion of the wellbore in the optimum layer. It further requires production stimulation by hydraulic fracturing (fracking) of the rocks to connect the natural fractures with induced near-well fractures.

In this paper, we present a methodology to identify these optimum areas and layers in the shales using a seismic characterisation workflow where well and seismic data are rigorously integrated. The first part of the approach is well data analysis to extract petrophysical, rock physics and mechanical information. Shale formations have a complex mineralogy requiring a sophisticated petrophysical analysis. Then a seismic inversion is performed to predict rock properties, which characterise the shale reservoirs and importantly allow us to predict how the rocks will respond to fracking. The final part of the methodology is an interpretation of multiple rock property models in terms of defined shale facies. A Bayesian approach was adopted to generate shale facies models that describe the thickness and complex architecture of shale reservoirs. These facies models can be used to significantly reduce the risk of poorly performing wells and improve asset performance.

A SkyTEM™ airborne electromagnetic dataset was inverted using a 1D reversible jump Markov chain Monte Carlo algorithm. The inversion of each dual-moment sounding generates an ensemble of 300,000 models that fit the data. The algorithm automatically varies the number of layers in the large range of models that are tested.

Analysis of the statistical properties of the ensemble yields a wealth of information on the probable conductivity distribution plus the mean, mode, median and most likely summary models. Robust information on the non-uniqueness and uncertainty of the results is also afforded by the ensemble. These are conveyed on conductivity map and section products. Estimates of the probable depths to interfaces are a further outcome. These depth estimates show great potential as an aid for mapping geological surfaces.

The resulting conductivity maps and sections are coherent and appear to be geologically realistic on face value. However it is demonstrated with 3D modelling that a plausible hydrogeological interpretation on the sections is likely to be an artefact of 1D inversion of a 3D geological scenario.

Unconformity-type uranium deposits are high-grade and constitute over a third of the world’s uranium resources. The Cariewerloo Basin, South Australia, is a region of high prospectivity for unconformity-related uranium. An airborne electromagnetic (AEM) survey was flown in 2010 using the Fugro TEMPEST® system to delineate the unconformity surface at the base of the Pandurra Formation. However highly-conductive regolith attenuated the signal in the northern and eastern regions, requiring application of deeper geophysical methods. In 2012 a magnetotelluric (MT) survey was conducted along a 110 km transect of the north-south trending AEM line.

MT data were collected at 29 stations and successfully imaged the depth to basement, and additionally providing evidence for deeper fluid pathways. The AEM data were integrated into the regularisation mesh as a-priori information generating an AEM constrained resistivity model and also correcting for static shift. The AEM constrained resistivity model best resolved resistive structures, allowing strong contrast with conductive zones.

Data from 68 broadband magnetotelluric stations were inverted to obtain a 2D electrical resistivity model beneath the Delamerian Orogen in southeast Australia along a 150 km east-west transect. Station spacing of 5 km in the west and ~2 km in the east resolved structure with changes in resistivity from 10-10,000 ?m occurring laterally over several kilometres. To the west, the crust is generally resistive, with a more complex structure to the east involving narrow paths of low resistivity (10-300 ?m). These conductive regions extend from Moho depths up to the surface and align with fault structures. The narrow conductive pathways possibly track mineral alterations from reactions with mantle fluids moving upwards late in the Delamerian Orogeny.

A reversible-jump Markov chain Monte Carlo inversion is used to generate an ensemble of millions of models that fit the forward response of a geoelectric target. Statistical properties of the ensemble are then used to assess the resolving power of the AEM system.

Since its inception in 2000, TEMPEST, a fixed wing time domain electromagnetic (TDEM) system, has been used in mineral, environmental and groundwater exploration and regolith and salt mapping. The versatility of the system can be attributed to its broad operational bandwidth, multifaceted software approach and distinctive calibration technique which allows both early and late time ground response to be imaged.

In the last decade the system has undergone a range of hardware and software developments with the aim of satisfying the changing landscape of geophysical exploration. Recent developments have resulted in improved system compensation, spatial resolution and noise levels which have helped further expand the system’s applications.

Forward and inverse modelling results and survey data are used to demonstrate the impact of recent developments on enhancing the system’s ability in a range of geophysical environments and exploration targets.

To properly construct broad features of lithospheric magnetization structure derived through inversion of the regional magnetic anomaly data, the curvature of the Earth should be taken into consideration. A method for inverting large-scale magnetic anomaly data in spherical coordinates has been developed. The recovered model is in the form of a 3D spherical orthogonal mesh. The inversion uses a classical Tikhonov regularization approach by incorporating a specially formed model objective function in spherical coordinates. We illustrate the algorithm using a synthetic example and then apply it to the satellite magnetic data over Australia. The results from the synthetic data example show that the inversion method in spherical coordinates is able to recover large- scale magnetization distributions. The inverted model in the field example is consistent with independent geological, heat flow, and Curie depth features published in the literature.

We constrain the depth and seismic structure of stiff sediment cover overlying a prospective basement terrane using a passive seismic technique which uses surface wave energy from microtremor (also known as ambient seismic energy or seismic noise). This may be applied to mineral exploration under cover to decrease the inherent ambiguity in modelling potential field data for exploration targeting.

Data from arrays of portable broadband seismometers are used to produce vertical profiles of seismic velocity structure using Both the Multimode Spatially Averaged Coherency (MMSPAC) method which measures the azimuthal average of the coherency between sensor pairs with a common separation, and the Horizontal to Vertical Spectral Ratio (HVSR) method, to estimate the seismic velocity structure and cover thickness respectively.

We have developed field protocols to ensure consistent acquisition of high quality data in a variety of ground conditions. A wavefield approaching the theoretical ideal for MMSPAC processing is created by combining the energy content of an off-road vehicle, driven around the seismometer array, and ambient sources. We find that this combination results in significantly higher quality MMSPAC waveforms in comparison to that obtained using ambient energy alone. Under ideal conditions a theoretical maximum depth of investigation of 600 m can be achieved with a hexagonal sensor array with 50 m radius and both MMSPAC and HVSR, although the maximum thickness of sedimentary cover in the study area limits the depth of investigation to approximately 180 m. The modelling procedure we employ is sensitive to layer thicknesses of ± 5%.

Until recently, seismic data acquisition has been fundamentally limited by the requirement that the delay time between one shot and the next be sufficient to avoid significant contamination of data from one shot with energy from another. Acquisition with simultaneous sources drops this requirement, and therefore provides potential for enormous improvements in acquisition rates and source sampling. In order to realize this potential however, the way we acquire and process data must change.

Although simultaneous-source technology is already a commercial reality for both land and marine acquisition, there is considerable scope for further optimization, especially in the marine case. In this work, I address the specific issue of relaxing the constraint that sources fire near-synchronously, such that the options for survey design in particular are greatly increased. I show that separation into individual shot records is possible even with arbitrary shot firing times, provided those times have some randomness to them. I address the major issue of removing strong interference from deep data, and show how this can be significantly improved by supplying prior information about the expected general decay of amplitude with time. Simulated simultaneous- source data, for which the correct answer is known, are used to illustrate the method.

The Wagga-Omeo Belt in NSW has been a significant tin, gold, silver and copper producer. An analysis of the mineral prospectivity of the region has been undertaken by estimating the number of undiscovered deposits, estimating mineral endowment, and by predicting the likely locations of undiscovered deposits. Both spatial prediction (mineral prospectivity mapping - MPM) and quantitative resource assessment (QRA) were undertaken with a focus on tin mineralisation in the region.

Although this study is in progress, preliminary results from the analysis of the spatial pattern of known deposits using the weights of evidence method, indicate likely locations of undiscovered tin deposits.

Geoscience Australia in collaboration with the Geological Survey of Western Australia conducted a seismic testing program on the Eucla Basin carbonate sediments during May 2012. These data were collected as part of the Albany-Fraser Orogen Seismic Survey that consists of four lines in south-east Western Australia with a total length of 671 km. The major aim of this survey was to image the basement relationship between the Yilgarn craton, the Albany-Fraser zone, and basement rocks further east. Much of this eastern area is covered by the limestones of the Eucla Basin, and there has been little seismic data acquired in this area prior to the survey commencing. Tests were designed to optimise the acquisition parameters for collecting deep seismic data beneath the limestones through the region.

Several sets of recording parameters were tested, including 10 Hz geophones compared to lower frequency 4.5 Hz geophones as parallel spreads, and linear versus non-linear logarithmic sweeps.

Self-demagnetisation can significantly reduce the amplitude and modify the shape of the magnetic response from highly magnetic bodies. For quasi-planar bodies, only the transverse component of magnetisation is reduced, with the result that the direction of magnetisation rotates towards the plane of the body. Furthermore, when highly magnetic bodies are in close proximity, the assumption of uniform inducing field is violated. Rather, highly magnetic bodies can modify the local magnetic field appreciably, with the result that the magnetisation induced in one body is affected by the magnetisations induced in all the others. It is important to take such interactions between highly magnetic bodies into account.

Potential field modelling and inversion software “VPmg” has been upgraded to account for self demagnetisation and interaction between magnetic bodies. The algorithm computes H-field perturbations at the model cell centres in two stages: initialisation and optimisation. During initialisation, a demagnetisation tensor is estimated for each cell, from which a first estimate for the H-field perturbation is derived. During optimisation, the H-field field estimate is refined iteratively via an inversion procedure. Remanence can be taken into account.

The algorithm has been validated for homogeneous spheres, spheroids, slabs, and cylinders. It has also reproduced magnetic interactions between two horizontal cylinders for the case published by Hjelt (1973). Explicit verification for complex heterogeneous bodies requires a suitable independent algorithm for benchmarking.

The application to inversion in highly magnetic environments is illustrated on field data examples.

The Petrel Sub-basin Marine Survey was undertaken in May 2012 by Geoscience Australia and the Australian Institute of Marine Science to support assessment of CO2 storage potential in the Bonaparte Basin. The aim of sub bottom profiling was high resolution data to investigate regional seal breaches and potential fluid pathways.

The sub bottom profiler data were acquired aboard the AIMS RV Solander, a total of 51 lines and 654 line km. Acquisition employed a Squid 2000 sparker and a 24 channel GeoEel streamer. Group interval of 3.125 m and shot interval of 6.25 m resulted in 6 fold stacked data. Record length was 500 ms, sampled every 0.25 ms. Rough sea conditions during the trade winds resulted in obvious relative motion between source and streamer.

Multichannel seismic reflection processing compensated for most of the limitations of sparker acquisition. Front end mute and band pass filter removed low frequency noise. Non surface consistent trim statics corrected for the relative motion of sparker and streamer, aligning reflections pre stack and improving signal to noise. Post stack minimum entropy deconvolution both suppressed ghosting and enhanced high frequencies (>1000 Hz). Vertical resolution of better than 1 m allowed delineation of multiple episodes of channelling in the top 100 m of sediment. Imaging of small channels was improved by collapsing diffractions with finite difference migration.