b'Education matters event can provide significant improvementwhilst the receiver term provides the bulkis estimated to be ~3150m (within 4% in signal-to-noise. We have used stackedamplitude. These model results, consistentof ice-core data). A spiking Wiener filter nodal records to infill between recordingswith theory, suggest that the sourcedesigned on the ice-model inversion is made at permanent seismographs.term should be a more useful indicator ofused to attenuate ice reverberations on Relative travel-time residuals betweenlateral variations in geology. stacked real-data PRFs. Lag-times are stations in Queensland show variationspicked for Sp from stacked SRF data and of order 1-2 s. Teleseisms arriving fromLinear inversion has been applied to aPs from ice-layer filtered PRF data. Lag different azimuths produce conflictingproduction Vibroseis data set that hadtimes for both P and S are consistent with relative-residual patterns. However, ray- previously been studied using a non- a crustal thickness of 43-44 km.path back projection to upper-mantlelinear inversion method. We obtain depths reveals coherent delay patterns.similar results to the previous study.Callum Kowalski, University of The observations are consistent withHowever, for both inversions, the sourceQueensland: Experiments in geophysical velocity variations of order 10% occurringand receiver profiles are quite similar,compressive sensingin the depth zone 200 - 300 km. Thesein apparent conflict with theoretical results are consistent with earlier work onmodelling. This dilemma has prompted the upper mantle in eastern Queensland. alternative parameterisations in the inversion process. Of the alternative Harrison Button, University ofparameterisations studied, it is found that Queensland: Linear inversion of seismic- simply taking the average of amplitudes refraction amplitudes for near-surfaceat each source location, or receiver velocity control location, provides a solution that is similar to both the non-linear and linear inversion results. Further real-data trials are needed to clarify the applicability of this tool for near-surface velocity control.Dale Harpley, University of Queensland: Crustal structure under Concordia, Antarctica from teleseismic P and S phasesCompressive sensing is a relatively recent development in data acquisition which can improve the economics and logistics of geophysical surveys by relaxing traditional sampling constraints. This work provides simple examples of key compressive sensing concepts, aimed at improving understanding of the topic. Fourier-domain experiments This thesis examines the linear inversion ofon stationary time series demonstrate seismic refraction amplitudes to constrainthe key requirements for signal recovery seismic velocities in the near-surface.from sub-Nyquist sampling, namely Seismic refraction amplitude is relatedrandom sampling and transform-domain to the head-wave coefficient, whichsparsity. Extension to non-stationary can, in theory, be parameterised as thegeophysical time series is more complex. product of a source term, a receiver term,P and S teleseismic data are analysedVariations of the Fourier transform handle and an offset term. Previous approacheswith the Receiver Function to estimaterandom sampling, and reduce transform have included convolutional stackingice and crustal thickness at Concordia,domain noise. The wavelet transform is and non-linear inversion. Here we useAntarctica. Standard Receiver Functionexamined as an alternative to the Fourier a linear-inversion process to extractprocessing of P-wave data on ice sheetstransform to demonstrate that recovery surface-consistent source and receiveris contaminated by reverberations inis possible via various domains. The terms from observed amplitudes. Linearthe ice sheet, concealing the Ps Mohoiterative Projection Onto Convex Sets inversion of model data produces similarconversion. The S-wave Receiver Function(POCS) algorithm is shown to be a simple results to non-linear inversion, with less(SRF) can show the Sp Moho conversionbut effective method of optimising signal computational effort. The final outputwhich arrives before the main S phaserecovery. Fourier and wavelet-based from the inversion is a calibrated velocity- and ice reverberations. The strongimplementations are demonstrated ratio profile across the survey. For modelreverberations in the P-wave Receiveron time series. Fourier-based POCS data we observe that for realistic velocity- Function (PRF) are used to measureimplemented in the time-wavenumber ratios, the source term provides the bulkice-layer parameters via a Generaliseddomain allows efficient recovery from shape of the head-wave coefficient,Linear Inversion (GLI). The ice thicknessspatially-subsampled seismic shot records.39 PREVIEW DECEMBER 2020'