Anya M. Reading

Geological mapping using remote sensing data: A comparison of five machine learning algorithms, their response to variations in the spatial distribution of training data and the use of explicit spatial information

Machine learning algorithms (MLAs) are a powerful group of data-driven inference tools that offer an automated means of recognizing patterns in high-dimensional data. Hence, there is much scope for the application of MLAs to the rapidly increasing volumes of remotely sensed geophysical data for geological mapping problems. We carry out a rigorous comparison of five MLAs: Naive Bayes, k-Nearest Neighbors, Random Forests, Support Vector Machines, and Artificial Neural Networks, in the context of a supervised lithology classification task using widely available and spatially constrained remotely sensed geophysical data. We make a further comparison of MLAs based on their sensitivity to variations in the degree of spatial clustering of training data, and their response to the inclusion of explicit spatial information (spatial coordinates). Our work identifies Random Forests as a good first choice algorithm for the supervised classification of lithology using remotely sensed geophysical data. Random Forests is straightforward to train, computationally efficient, highly stable with respect to variations in classification model parameter values, and as accurate as, or substantially more accurate than the other MLAs trialed. The results of our study indicate that as training data becomes increasingly dispersed across the region under investigation, MLA predictive accuracy improves dramatically. The use of explicit spatial information generates accurate lithology predictions but should be used in conjunction with geophysical data in order to generate geologically plausible predictions. MLAs, such as Random Forests, are valuable tools for generating reliable first-pass predictions for practical geological mapping applications that combine widely available geophysical data.

Lithospheric structure of Tasmania from a novel form of teleseismic tomography

In 2001 and 2002, a temporary array of 72 seismic recorders was deployed across northern Tasmania (SE Australia), with the aim of imaging the underlying crust and upper mantle using three-dimensional (3-D) teleseismic tomography. Using a recently developed adaptive stacking technique, 6520 relative P wave arrival time residuals have been picked from 101 distant earthquake records spanning a 5 month period. A novel iterative nonlinear tomographic procedure based on a subspace inversion scheme and the fast marching method, a grid-based eikonal solver, is used to map the residual patterns as P wave velocity anomalies. The new scheme proves to be both fast and robust, making it well suited to large data sets and the reconstruction of complex anomalies. The resultant tomographic images of Tasmania exhibit significant lateral perturbations in P wave velocity structure (5%) from a 1-D reference model. A marked transition from higher velocities in the east to lower velocities in the west strongly supports the idea that eastern Tasmania is underlain by dense rocks with an oceanic crustal affinity, contrasting with the continentally derived siliciclastic core of western Tasmania. Significantly, the Tamar Fracture System does not overlie the narrow transition from relatively fast to slow velocities, which suggests that it may be a near-surface feature rather than a manifestation of deeper crustal-scale suturing as previously thought. Farther west, an easterly dipping zone of relatively high velocity material beneath the Rocky Cape Group and Arthur Lineament may be related to remnant subduction of oceanic lithosphere associated with the mid-Cambrian Delamerian Orogeny.

Improved inversion for seismic structure using transformed, S-wavevector receiver functions: Removing the effect of the free surface

The determination of structure from the inversion of teleseismic receiver functions may be improved by removing the contribution of the free surface. The free surface interaction gives rise to the highest amplitude signal on standard receiver functions and yet this initial pulse tells us little about the receiver structure below the surface layer. We apply a transformation to P, S and Horizontal wavevector components, which removes the free surface response and hence the initial P-pulse. A pure receiver function is calculated by deconvolving the S-wavevector component with the P-wavevector. In general, converted phase amplitudes within the receiver function waveform are better matched by the inverse algorithm, resulting in an improved estimation of seismic structure. In particular, low amplitude receiver function waveforms, often associated with poorly constrained structure, now yield a successful inverse.

Crustal architecture of the Capricorn Orogen, Western Australia and associated metallogeny

A 581 km vibroseis-source, deep seismic reflection survey was acquired through the Capricorn Orogen of Western Australia and, for the first time, provides an unprecedented view of the deep crustal architecture of the West Australian Craton. The survey has imaged three principal suture zones, as well as several other lithospheric-scale faults. The suture zones separate four seismically distinct tectonic blocks, which include the Pilbara Craton, the Bandee Seismic Province (a previously unrecognised tectonic block), the Glenburgh Terrane of the Gascoyne Province and the Narryer Terrane of the Yilgarn Craton. In the upper crust, the survey imaged numerous Proterozoic granite batholiths as well as the architecture of the Mesoproterozoic Edmund and Collier basins. These features were formed during the punctuated reworking of the craton by the reactivation of the major crustal structures. The location and setting of gold, base metal and rare earth element deposits across the orogen are closely linked to the major lithospheric-scale structures, highlighting their importance to fluid flow within mineral systems by the transport of fluid and energy direct from the mantle into the upper crust.

Contrasts in mantle structure beneath Australia: Relation to Tasman Lines?

Surface‐wave tomography for the Australian region, using data mostly from portable seismic recorders, reveals a very strong contrast in seismic shear‐wave speed beneath central and western Australia and the east of the continent. Shear‐wave speeds faster than the continental average extend to at least 200 km depth in the cratonic zone to the west of 140°E. Along an approximately north‐south line there is then an eastward step to thinner lithosphere (∼150 km thick) but still with fast shear‐wave speeds. A further more irregular transition to the east marks the transition to lowered shear‐wave speeds. The eastern transition at 75 km depth is in close agreement with the original Tasman Line, whereas the two more westerly transitions do not bear a simple relation to the more recent group of Tasman Lines defined from crustal information (outcrop and inferred geophysical trends). The westward transition to the thickest coherent lithosphere (near 140°E) may well mark the edge of the ancient core of the continent, but the current mantle structures must bear the scars of the breakups and reassembly that have created the current Australian continent.

Seismic structure of the Yilgarn Craton, Western Australia

The deep crustal and upper mantle structure of the Yilgarn Craton is investigated in this study using receiver‐function analysis of teleseismic earthquake records from temporary stations. Two lines of stations were deployed, the main transect ran between Perth and Kalgoorlie, and a second line of stations ran across the east Yilgarn Craton 200 km to the north of Kalgoorlie. The broadband instrumentation records high‐fidelity waveform data allowing the signal from the near‐receiver structure to be separated from the influence of the earthquake source. The nature of the crust and upper mantle structure under each station is determined from seismic‐velocity models that match the observed receiver‐function waveforms and the resulting coarse‐scale transect provides new, independent controls on the structure of the lithosphere. Mechanisms for the evolution of the Yilgarn Craton, previously put forward to explain surface geological and geochemical observations, and seismic velocity structure from reflection and refraction studies, may be classified as favouring: (i) predominantly accretionary lithospheric evolution; (ii) mixed accretion and other influences; or (iii) no accretionary‐style influence. Characteristics of the deep seismic structure enable the evolutionary mechanism to be inferred. From the teleseismic data, we find that the seismic Moho is sharp in character under stations in the middle of the proposed terranes and more gradational near the proposed terrane boundaries. The Moho dips gently eastward and the seismic velocity of the upper mantle increases moving from west to east across the whole craton. An anomalous region exists under the Southwest terrane that shows a thick high‐velocity gradient zone at the base of the crust and a Moho dipping to the west. The nature of the lateral heterogeneity in structure and its correspondence with proposed terrane boundaries suggest that accretionary processes are significant in the evolution of the Yilgarn Craton.

Lithospheric structure of the Pilbara Craton, Capricorn Orogen and northern Yilgarn Craton, Western Australia, from teleseismic receiver functions

The upper lithospheric structure from the Pilbara Craton, across the Capricorn Orogen, to the northern Yilgarn Craton is determined from high‐fidelity broadband seismic data. Three‐component temporary stations were deployed in a line running southwards from Marble Bar, Western Australia, at approximately 118°E longitude. They were in position between July and October 2000. Receiver functions are calculated from the recorded teleseismic earthquakes stacked to improve the signal to noise ratio and, finally, the stacked waveforms used to model the seismic velocity profile under each recording station. Data from the permanent Global Seismic Network station, MBWA, at Marble Bar, which was installed in August 2001, are also used. The method provides a means of probing the deep structure of the Earths crust and the nature of the seismic Moho at an intermediate scale between that of detailed active‐source refraction techniques and the regional structure obtained from earthquake seismic tomography. The receiver functions and modelled seismic velocity structures along the north‐south profile show the crust‐mantle boundary under the Pilbara Craton to be shallow, at 30 km (±2 km), with a sharp Moho and high‐velocity crust beneath the exposed Pilbara granite‐greenstone terrane. The sharp Moho extends under the metasediments of the Hamersley Basin. Beneath the Capricorn Orogen the Moho is barely discernible, showing a small seismic velocity contrast and a broad zone of high‐velocity gradient. The northern Yilgarn Craton, which extends beneath the basins exposed on the surface, is deeper, at 40 km (±2 km), and again shows a sharply defined Moho.

Dominant seismic noise sources in the Southern Ocean and West Pacific, 2000–2012, recorded at the Warramunga Seismic Array, Australia

Seismic noise is important in determining Earth structure and also provides an insight into ocean wave patterns and long-term trends in storm activity and global climate. We present a long-duration study of seismic noise focused on the Southern Ocean using recordings from the Warramunga Seismic Array, Northern Territory, Australia. Using high-resolution analysis, we determine the seismic slowness and back azimuth of observed seismic noise, microseisms, at hourly intervals through over a decade (2000–2012). We identify three dominant sources of body wave ( P ) noise in the Southern Ocean which we interpret to originate from a South Atlantic source propagating as PP waves, and Kerguelen Island and Philippine Sea sources propagating as P waves. We also identify surface waves from around the Australian coast. All sources show distinct seasonality and a low, but discernable, interannual variability.

New constraints on the seismic structure of West Australia: Evidence for terrane stabilization prior to the assembly of an ancient continent?

We present a new, near-comprehensive survey of the variations in seismic structure across the West Australian craton at the scale of the main terrane groups. Analyzing data from distant earthquakes recorded at temporary and permanent stations located across the region, we found the best-fi tting structure by modeling the conversions from P- to S-wave motion (the receiver function) that take place as the seismic energy travels upward through the lithosphere. Such methods can be used to delineate the extent of cratonic and orogenic terranes in regions where geological exposure of the surface is limited, and they provide an effective alternative to active-source seismic techniques for deep crustal targets. The seismic structure is consistent within several of the individual Archean terranes, most notably the Pilbara, Murchison, and Southern Cross. These terranes are underlain by lower crust of low seismic velocity and show a sharp seismic Moho. The structure shows signifi cant contrasts between neighboring terranes; thus, major tectonic units have a velocity profi le that is a signature of that terrane or terrane group. We infer that the seismic structure of the Archean crust and upper mantle was fixed before craton assembly and preserved through the subsequent collision and accretion of the tectonic units that formed the West Australian craton.

Structure of the Tasmanian lithosphere from 3D seismic tomography

Seismic data from three separate experiments, a marine active source survey with land-based stations, and two teleseismic arrays deployed to record distant earthquakes, are combined in a joint inversion for the 3D seismic structure of the Tasmanian lithosphere. In total, travel-time information from nearly 14 000 source–receiver paths are used to constrain a detailed model of crustal velocity, Moho geometry and upper mantle velocity beneath the entire island. Synthetic reconstruction tests show good resolution beneath most of Tasmania with the exception of the southwest, where data coverage is sparse. The final model exhibits a number of well-constrained features that have important ramifications for the interpretation of Tasmanian tectonic history. The most prominent of these is a marked easterly transition from lower velocity crust to higher velocity crust which extends from the north coast, northeast of the Tamar River, down to the east coast. Other significant anomalies include elevated crustal velocities beneath the Mt Read Volcanics and Forth Metamorphic Complex; thickened crust beneath the Port Sorell and Badger Head Blocks in central northern Tasmania; and distinctly thinner, higher velocity crust beneath the Rocky Cape Block in northwest Tasmania. Combined with existing evidence from field mapping, potential-field surveys and geochemical data, the new results support the contention that east and west Tasmania were once passively joined as far back as the Ordovician, with the transition from lithosphere of Proterozoic continental origin to Phanerozoic oceanic origin occurring some 50 km east of the Tamar River; that the southeast margin of the Rocky Cape Block may have been a former site of subduction in the Cambrian; and that the Badger Head and Port Sorell Blocks were considerably shortened and thickened during the Cambrian Tyennan and Middle Devonian Tabberabberan Orogenies.