b'David Annetts best of Exploration GeophysicsFeatureFast AEM data processing and inversionJames Macnae The resulting AEM data have typically been presented as Andrew King stacked profiles for manual- or computer-assisted interpretation. Ned Stolz In areas with surficial glacial sediments or in regolith dominated Alex Osmakoff terrain, there is usually a variable background response with Andrej Blahaone or more local anomalies per line-kilometre. Conductivity-*CRC AMET, School of Earth Sciences, Macquarie University,depth sections have proven to be a useful tool to map this NSW2109 variable regolith background.The second stage in mineral exploration with AEM is to Abstract determine which of the many local anomalies has the greatest When airborne electromagnetic (AEM) data is acquired aschance of being an economic mineral deposit. Because a streamed or time-series data set, the great redundancy ineconomic mineral deposits come in many shapes and sizes the data favours compression as a first step in processing.depending on their origin and tectonic history, and can have a Traditional data compression schemes are time windowingwide range of possible conductivities, it is desirable to model and spatial averaging. An alternative, more efficient dataor invert each of these local anomalies to determine the compression scheme is to transform time or frequency domainlikelihood of correspondence to a range of physical property data to time-constant tau space, which has the effect of(and by inference geological) models. As with any interpretation removing the waveform dependence of the AEM response. technique, inferences made will rely on correspondence between physical properties and geology, and conductors When there are many local anomalies and a variablemay not necessarily be ore itself, but indicative of nearby background, the next stage of rapid processing is to transformmineralisation.the response to a conductivity-depth image (CDI) to facilitate geological interpretation of the background response. Use ofSystem-independent EM representationthe full time range of recorded data, particularly the inclusion of on-time data, improves the stability of the CDI process. Stolz and Macnae (1998) described a method by which arbitrary waveform data can be transformed into a form that is The final AEM data processing step for mineral exploration isindependent of the EM system. We consider this to be a key step to assess the likelihood that any local anomaly corresponds toin the development of fast processing and interpretation tools, a desired economic target. This step involves the extraction ofas generic rather than system waveform-specific algorithms target geometry and conductivity information from the AEMcan be implemented in any later processing steps. There is a data. The only economically feasible route at the present timetrade-off in the time taken to transform the data, a requirement is to parameterise both the data (using inductive and resistivethat the transformation adequately and stably represents the limits) and the model to allow inversion of the local anomalies. Aoriginal data, and in practice there are problems accurately fit to one or two platelike conductors can be achieved in seconds;defining the actual system waveform and transformation fits to a blocklike body take minutes on a fast PC. A significantparameters.research challenge remains to speed up and stabilise this process.In time domain, this procedure consists of preconvolving a set Introduction of exponential decays with the repetitive system waveform (Appendix A), and fitting a linear combination of these to the A recent trend in airborne electromagnetic (AEM) data is theobserved data. In the frequency domain the procedure involves acquisition of data using continuous analogue-to-digitalfitting a linear sum of mathematically equivalent single-pole, conversion and recording. Sampling of the transmittersingle-zero responses (Grant and West 1965) to the observed and receiver components occurs on the order of every 10data. Following the fitting procedure, the time or frequency microseconds, and if all sampled data are recorded, then datadata is reduced to a set of amplitudes with an associated time volumes are of the order of gigabytes per day of flying. There isconstant, and is independent of system waveform. Differences generally far too much data to comprehend or interpret in thisbetween individual systems or different waveforms of a given raw form, and data compression is obviously required. Becausesystem lie in the range of time constants (tau values) that can of the nature of the EM diffusive process, the vast majority ofbe resolved and the noise levels in the data which carry through the data is actually redundant, and stacking and windowinginto the tau domain.methods have long been used to improve signal-to-noise ratios and reduce data volumes to manageable size. Both the physicsConductivity-Depth-Imaging (CDI)of diffusion (Stolz and Macnae 1998) and principal component analysis (Green 1998) suggest that there are no more that aboutOne method of converting AEM data into a conductivity-depth-4 or 5 independent piece of information per decade of systemsection is to invert the data on a point-by-point basis to a set bandwidth. With (say) 256 equally spaced waveform samplesof 1D models, and then stitch these together to form a pseudo covering just over two decades in time, there will be no more2D section. This method requires relatively powerful computing than about 10 (spaced logarithmically in time) independentresources, but is in routine use by World Geoscience on parameters required to completely characterise the EMQuestem data (Sattel 1998). However, inversion using a 3-layer response, although a much larger number may be required tomodel is considerably slower than data acquisition.define the noise characteristics. There are a number of methods in the literature for deriving fast conductivity-depth sections from EM data. One group of *Note, these contact details have almost certainly changed methods is based on the Maxwell receding image concept, 45 PREVIEW OCTOBER 2020'