Seismic structure in the mantle beneath Australia

3 downloads 66628 Views 53MB Size Report
C. D. N. Collins, B. J. Drummond and M. G. Nicoll: Crustal thickness patterns in the ...... sity is best in central and eastern Australia and the 3-D model is less well ...
2003 GEOLOGICAL SOCIETY OF AUSTRALIA INCORPORATED This volume is published in the Special Publication Series of the Geological Society of Australia REFERENCE GENERAL REFERENCE: HILLIS R. R. & MÜLLER R. D. (Editors) 2003. Evolution and Dynamics of the Australian Plate, Geological Society of Australia Special Publication 22 and Geological Society of America Special Paper 372.

TWO FORMS OF REFERENCE TO SPECIFIC PAPERS ARE POSSIBLE: CRAWFORD A. J., MEFFRE S. & SYMONDS P. A. 2003. 120 to 0 Ma tectonic evolution of the southwest Pacific and analogous geological evolution of the 600 to 220 Ma Tasman Fold Belt System. Geological Society of Australia Special Publication 22 and Geological Society of America Special Paper 372, pp 383–404. CRAWFORD A. J., MEFFRE S. & SYMONDS P. A. 2003. 120 to 0 Ma tectonic evolution of the southwest Pacific and analogous geological evolution of the 600 to 220 Ma Tasman Fold Belt System. In: Hillis R. R. & Müller R. D. eds. Evolution and Dynamics of the Australian Plate, pp 383–404 Geological Society of Australia Special Publication 22 and Geological Society of America Special Paper 372.

First published in 2003 by the Geological Society of Australia Incorporated Typeset by Chapter 8 Pty Ltd, Sydney Cover artwork produced by Lagorce Advertising Printed and bound by ATC Printing This book is copyright. Apart from any fair dealing for the purpose of private study, research, criticism or review, as permitted under the Copyright Act, no part may be reproduced by any process without the prior permission of the copyright owner. © 2003 Geological Society of Australia Incorporated ISSN 0072-1085

INQUIRIES AND ORDERS SHOULD BE DIRECTED TO: The Business Manager Geological Society of Australia 706 Thakral House, 301 George Street NSW 2000 Australia Or direct through our website www.gsa.org.au

EVOLUTION AND DYNAMICS OF THE AUSTRALIAN PLATE Bibliography Includes index ISBN 1 876125 32 2 1. Plate tectonics - Australia. 2. Geophysics - Australia. 3. Geology - Australia. 4. Dynamics. I. Hillis R.R. II Muller R.D. III. Geological Society of Australia. (Series: Special publication (Geological Society of Australia); 22). Series: Special paper (Geological Society of America); 372). 551.0994

Edited by R. R. Hillis and R. D. Müller

Geological Society of Australia Special Publication 22 Geological Society of America Special Paper 372

ACKNOWLEDGEMENT

Evolution and Dynamics of the Australian Plate is published jointly by the Geological Society of Australia and the Geological Society of America.

Industry support for the publication of this volume is gratefully acknowledged. Without such support it would not be possible to enhance this work with so many colour figures.

We are particularly grateful to Santos and Woodside for sponsoring the volume and thereby making a major contribution towards the success of the publication.

Front cover figure:

Back cover figure:

Age of the ocean floor around Australia (red/yellow, Tertiary; green, Cretaceous; blue, Jurassic). Grey-shading based on the GTOPO30 digital elevation model.

The Milendella Fault near Cambrai, eastern Mt Lofty Ranges, South Australia. See Sandiford (this volume) for discussion of this outcrop and other evidence of the extensive, and sometimes neglected, neotectonic record of Australia.

(http://edcdaac.usgs.gov/gtopo30/.html)

CONTENTS

R. R. Hillis and R. D. Müller: Introduction. Evolution and dynamics of the Australian Plate

1

B. L. N. Kennett: Seismic structure in the mantle beneath Australia

7

E. Debayle and B. L. N. Kennett: Surface-wave studies of the Australian region

25

P. Tregoning: Is the Australian Plate deforming? A space geodetic perspective

41

R. R. Hillis & S. D. Reynolds: In situ stress field of Australia

49

S. D. Reynolds, D. D. Coblentz & R. R. Hillis: Influences of plate-boundary forces on the regional intraplate stress field of continental Australia

59

S. Zhao and R. D. Müller: Three-dimensional finite-element modelling of the tectonic stress field in continental Australia

71

D. Clark and M. Leonard: Principal stress orientations from multiple focal-plane solutions: new insight into the Australian intraplate stress field

91

M. Sandiford: Neotectonics of southeastern Australia: linking the Quaternary faulting record with seismicity and in situ stress

107

C. D. N. Collins, B. J. Drummond and M. G. Nicoll: Crustal thickness patterns in the Australian continent

121

P. R. Milligan, P. Petkovic and B. J. Drummond: Potential-field datasets for the Australian region: their significance in mapping basement architecture

129

F. E. M. Lilley, L. J. Wang, F. H. Chamalaun and I. J. Ferguson: Carpentaria Electrical Conductivity Anomaly, Queensland, as a major structure in the Australian Plate

141

S. McLaren, M. Sandiford, M. Hand, N. Neumann, L. Wyborn and I. Bastrakova: The hot southern continent: heat flow and heat production in Australian Proterozoic terranes

157

C. O’Neill, L. Moresi, A. Lenardic and C. M. Cooper: Inferences on Australia’s heat flow and thermal structure from mantle convection modelling results

169

O. F. Gaul, S. Y. O’Reilly and W. L. Griffin: Lithosphere structure and evolution in southeastern Australia

185

F. L. Sutherland: ‘Boomerang’ migratory intraplate Cenozoic volcanism, eastern Australian rift margins and the Indian–Pacific mantle boundary

203

B. J. Brown, R. D. Müller, C. Gaina, H. I. M. Struckmeyer, H. M. J. Stagg and P. A. Symonds: Formation and evolution of Australian passive margins: implications for locating the boundary between continental and oceanic crust

223

S. Baldwin, N. White and R. D. Müller: Resolving multiple rift phases by strain-rate inversion in the Petrel Sub-basin, northwest Australia

245

K. C. Hill and R. Hall: Mesozoic–Cenozoic evolution of Australia’s New Guinea margin in a west Pacific context

265

R. H. Findlay: Collision tectonics of northern Papua New Guinea: key field relationships demand a new model

291

M. Keep, I. Longley and R. Jones: Sumba and its effect on Australia’s northwestern margin

309

S. J. D. Cox: Earth-science datasets for the Australian continent online: the AGCRC map-maker, OpenGIS and beyond

319

R. J. Musgrave: Early to Middle Miocene Pacific–Australia plate boundary in New Zealand: an alternative transcurrent-fault system

333

M. Sdrolias, R. D. Müller and C. Gaina: Tectonic evolution of the southwest Pacific using constraints from backarc basins

343

R. Hall and W. Spakman: Mantle structure and tectonic evolution of the region north and east of Australia

361

A. J. Crawford, S. Meffre and P. A. Symonds: 120 to 0 Ma tectonic evolution of the southwest Pacific and analogous geological evolution of the 600 to 220 Ma Tasman Fold Belt System

383

C. Gaina, R. D. Müller, B. J. Brown and T. Ishihara: Microcontinent formation around Australia

405

M. Gurnis and R. D. Müller: Origin of the Australian–Antarctic Discordance from an ancient slab and mantle wedge

417

Author Index

431

Subject Index

432

Geol. Soc. Australia Spec. Publ. 22, and Geol. Soc. America Spec. Pap. 372 (2003), 1–5

Evolution and dynamics of the Australian Plate

INTRODUCTION A session on ‘The Australian Plate’ was held at the 15th Australian Geological Convention in Sydney in July 2000. At that session it was clear that the 1990s had seen an explosion of new geophysical, geochemical and geological data for the Australian continent, and indeed for the wider plate. Further it was clear that the application of new interpretation techniques in the earth sciences, especially numerical modelling, to these improved datasets was yielding a vastly improved level of understanding of the evolution and dynamics of the Australian Plate. This volume stemmed from the realisation that combining these new data and interpretations into a single volume would provide both a much-needed description of the current ‘state-ofthe-science’ and also help reveal the new directions required to further illuminate the evolution and dynamics of the Australian Plate. We sought additional papers beyond those presented at the original conference session in order to widen the scope and maximise the impact of the volume. Of the 27 papers in the volume, 12 can be traced back to the Australian Geological Convention session and 15 additional papers have been contributed. The strength of the volume lies, of course, in the quality of the individual contributions and there are a number of reasons why so many high-quality contributions were received. Authors clearly recognised the significance of a multidisciplinary volume on the Australian Plate at this time. Furthermore, the sponsorship of Santos and Woodside enabled the volume to be fullcolour and hardbound, and also to be priced so as to be accessible not only to researchers, but also to senior undergraduate and postgraduate students. These were all important factors in ensuring the quality of contributions, as was the decision to make this volume the first joint publication of the Geological Society of Australia and the Geological Society of America, in the form of a special publication of the former and a special paper of the latter. Although all the papers in this volume deal with the Australian Plate, the processes controlling the evolution and dynamics of the Australian Plate are of global significance, as indeed are many of the problems specific to the Australian Plate. The volume is multidisciplinary, with papers across the breadth of subdisciplines of the earth sciences and across the spectrum from data-rich to interpretation-rich. Many papers utilise multiple datasets and present new data and interpretations. Any attempt to categorise the papers would be misleading and we have not divided the volume into subsections on different topics. Indeed we believe that such subdivision would be counter-productive with many future developments in our understanding of the evolution and

dynamics of the Australian Plate lying in the integration of results from the different disciplines and approaches covered herein.

Overview of the papers The first two papers in the volume investigate the seismic structure of the Australian lithosphere. Kennett combines recordings from the rather limited number of permanent seismic stations across Australia with several experiments involving portable deployments in order to obtain P- and Swavespeed and attenuation structure. The most striking feature of the seismic structure is that relatively fast wavespeeds associated with the lithosphere extend to around 210 km beneath the older, Archaean and Proterozoic regions of western and central Australia, whereas beneath the Phanerozoic regions of eastern Australia the seismic wavespeeds suggest that the lithosphere is less than 140 km thick. This transition does not coincide exactly with that of the conventionally recognised Tasman Line (the surface projection of the eastern limit of the Australian Proterozoic craton along which Neoproterozoic breakup and subsequent Phanerozoic collision took place). However, the Tasman ‘Line’ is likely to be a broad zone of deformation and may at depth dip away from its surface position. Debayle and Kennett use surface wave data from the same seismic deployments to reveal S-wave anisotropy structure. Two layers of anisotropy are distinguished: an upper layer to approximately 150 km with varying directions of anisotropy and a deeper layer to 200–250 km with dominantly north–south-oriented anisotropy. Anisotropy in the deep lower layer is likely to reflect present-day deformation due to the northward passage of the Australian Plate over the asthenosphere. Debayle and Kennett suggest anisotropy in the upper layer reflects ‘frozen’ deformation (i.e. tectonic evolution). In contrast, Hillis and Reynolds suggest there are significant similarities between the directions of S-wave anisotropy in the upper layer and the maximum horizontal stress directions in the Australian Stress Map, possibly due to the influence of stress-aligned cracks. The following six papers in the volume focus on the strain and stress fields of Australia and their neotectonic expression. Tregoning’s GPS measurements indicate that over approximately the last 10 years there has been no strain within the Australian Plate above the noise level of GPS analysis (i.e. above 2 mm/y). All intraplate stations are moving north-northwest at around 5 cm/y. In contrast, the northern margin of the plate in Papua New Guinea does show motion with respect to the rigid plate. Hence the intraplate neotectonic activity in southeast Australia described by Sandiford is associated with a significantly

2

R. R. Hillis and R. D. Müller

lower strain rate than activity at the plate boundary. It is interesting to speculate whether longer term measurements will witness the significant neotectonic deformation within Australia and whether the differences between plate boundary and intraplate strain rates may be revealed. Hillis and Reynolds describe the in situ stress field of Australia based on earthquake focal-mechanism solutions, wellbore deformation in oil exploration wells and shallower engineering measurements. The Australian Plate has a uniquely complex pattern of maximum horizontal stress orientations. Most stable continental areas such as North and South America and Western Europe are characterised by broad regions where maximum horizontal stress orientation is consistent and parallel to the direction of absolute plate motion. Thus the forces driving and resisting plate motion are inferred to control the intraplate stress field. Stress orientations over much of Australia are variable and in many places at a high angle to plate motion direction. This observation begs the question as to whether the Australian intraplate stress field is subject to different controls than those of other continental areas. Stress modelling papers by Reynolds et al. and Zhao and Müller address the origin of the stress field using finite-element modelling. Reynolds et al. demonstrate that, to a first-order, the intraplate stress field of Australia is indeed, like that of other continents, consistent with control by plate-boundary forces, provided the complex nature of the northeastern convergent boundary of the Australian Plate (from Himalayas to New Zealand) is recognised. Zhao and Müller explore the effect of variation of the mechanical strength of the lithosphere, which is generally ignored in plate-scale models, on the predicted stress field. They also conclude that the first-order pattern of the stress field is controlled by plate-driving forces but demonstrate that maximum horizontal stress orientations and the pattern of stress concentration manifested by seismicity are modulated by the variation in mechanical strength of local/regional geological structures. Clark and Leonard determine the stress tensor for seven domains within Australia based on inversion of multiple earthquake focal mechanisms in each domain. In general there is good agreement between their determinations and those from shallower depth presented by Hillis and Reynolds. The exceptions are the Sydney Basin and Flinders Ranges where localised sources of stress such as those associated with shallow topography may be contaminating the regional tectonic signal in the shallower stress field. The deeply held notion of Australia as a stable continent dominated by ancient land surfaces is challenged by Sandiford who describes some of the rich neotectonic record of southeast Australia (Figure 1). The faults controlling the range-bounding scarps in the Mt Lofty and Flinders Ranges in South Australia are characterised by reversesense Quaternary slip rates in the range of 20–150 m/106 y. Sandiford’s proposal of the term ‘Sprigg Orogeny’ for the ongoing processes that have built the Flinders and Mt Lofty Ranges both appropriately accords priority to Sprigg’s early recognition of Pliocene–Pleistocene faulting in the area and challenges us to reappraise our understanding of the significance of this neotectonic activity. Sandiford argues for a terminal Miocene onset for the present neotectonic regime of southeast Australia that is controlled by interactions

between the Pacific and Australian Plates. We believe the systematic mapping of neotectonic activity and development of a ‘Neotectonic Map of Australia’ would significantly aid our understanding of the Miocene–Holocene evolution and dynamics of the Australian Plate. Collins et al. summarise crustal thickness in Australia using seismic-refraction, seismic-reflection and receiverfunction data. Moho depth varies from 24 to 56 km, with the thinnest crust in the Archaean cratons of Western Australia and in Tasmania and the thickest in Proterozoic north and central Australia and Phanerozoic southeastern Australia. There is no good correlation between crustal thickness and the megatectonic units that comprise the Australian continent, and only occasionally do changes in crustal thickness correspond to the boundaries of these units. We note that the crustal thickness of many of these megatectonic units have been modified by processes subsequent to their formation such as rifting. Australia has world-class coverage by and compilations of magnetic and gravity potential-field data. Surveys undertaken by Geoscience Australia and its predecessors (the Bureau of Mineral Resources and the Australian Geological Survey Organisation), the State and Territory geological surveys, and exploration companies have been compiled yielding 17 million line-kilometres of magnetic data and 900 000 gravity observations. Milligan et al. describe these data and the resulting image compilations and enhancements that provide a major input towards unravelling the tectonic evolution of the continent. Australia’s excellent coverage by potential-field data is a direct consequence of its geological past, and specifically the absence of Quaternary glaciation that removed cover in large parts of the northern hemisphere continents. As a result potentialfield data have long been routinely used in Australia in resource exploration beneath cover, in conventional geological mapping, and also in the recognition of the major tectonic elements and their boundaries. Milligan et al. discuss the position of the Tasman Line beneath cover in the vicinity of Broken Hill based on interpretation of the potential-field data. They also use the data to illustrate how ancient basement controls have influenced the style of subsequent continental breakup. The Carpentaria conductivity anomaly in Queensland (Figure 1), which extends to tens of kilometres depth, is described by Lilley et al. on the basis of magnetotelluric survey data. Lilley et al. argue that it represents a major tectonic boundary. Its location is broadly consistent with that of the major transition in the velocity structure of the mantle lithosphere in the area described by Kennett and, as noted by Lilley et al., does not coincide with the traditional placement of the Tasman Line in north Queensland, being further west. Lilley et al. argue that the Carpentaria conductivity anomaly represents a Proterozoic suture. The following two papers investigate heat-flow patterns in Australia. Australia can be broadly divided into three heat-flow provinces largely consistent with the major geological provinces. Phanerozoic eastern Australia, though comprising primarily Palaeozoic fold belts, is tectonothermally young and exhibits high variations in heat flow probably related to Cenozoic magmatic and tectonic events. McLaren et al. (Figure 1) describe the high heat flow of the central Australian Proterozoic terranes (83 ±18 mWm–2)

Introduction

3

Figure 1 Location of areas discussed in papers in this volume. The papers by Kennett, Debayle and Kennett, Tregoning, Hillis and Reynolds, Clark and Leonard, Reynolds et al., Zhao and Müller, Collins et al., Milligan et al., O’Neill et al. and Cox, all of which essentially cover the entire Australian continent, are not shown.

and argue that this anomalous heat flow is due to anomalous heat production (4.6 µWm–3) in the Proterozoic rocks. Given that at the time of magmatism heat-production rates would have been even higher, this extraordinary enrichment of Australian Proterozoic crust has played a key role in its tectonothermal evolution. The western part of Australia is characterised by the low heat-flow values typical of Archaean terranes worldwide. O’Neill et al. discuss the application of mantle convection modelling to problems of continental heat flow. Convection simulations can provide insights into the thermal coupling of crust and mantle, and the relative contributions of both to continental heat flow. For example, O’Neill et al. argue that high heat flow in Proterozoic central Australia is unlikely to be due to the mantle component of heat flow because Australia has moved over a major past subduction zone (the surface expression of which is now the Australian–Antarctic Discordance discussed further below) and this evolution is inconsistent with large-scale advection of warm mantle material into central Australia. This supports McLaren et al.’s argument that the high heat flow in the Proterozoic terranes of central Australia is due to high heat production in these terranes. Based on a transect crossing the Tasman Line, Gaul et al. (Figure 1) show that geochemical lithospheric tomography sections, in terms of the magnesium number of olivine, can be constructed from mantle-derived pyrope garnet xenocrysts. As both seismic P- and S-wave velocities in mantle rocks are dependent on the magnesium number, this opens up interesting new avenues for interpreting mantle P- and S-wave anomalies in different tectonic domains.

Sutherland (Figure 1) explores the origin of the unusual small-scale geometry of plume related volcanism in eastern Australia, with an east–west offset between central volcanoes and subsequent less voluminous plume activity. The following two papers deal with the tectonic evolution of offshore sedimentary basins. Brown et al. (Figure 1) focus on the passive margins of the North West Shelf and the Great Australian Bight and Baldwin et al. investigate the largely Carboniferous–Permian Petrel Sub-basin (Figure 1), an offshore basin the development of which pre-dates development of the overlying passive margin sequence of the North West Shelf. The preserved sedimentary record of basins provides vital data for elucidation of the tectonic evolution of the lithosphere and both papers determine the stretching () factors based on subsidence history and also the associated strain rates. The Petrel Sub-basin is a deeply subsided (>25 km thick), highly extended (=2–6) basin. The high stretching factors inferred from the subsidence history of the basin are consistent with estimates of whole crustal thinning based on gravity modelling and the results suggest that the uniform stretching model can account for the large-scale evolution of the Petrel Sub-basin. Brown et al. show that the largely volcanic North West Shelf and non-volcanic inner Great Australian Bight margin have been subjected to much lower stretching factors (1000 km sinistral displacement. The Tasman Line extends across New Guinea and the associated change in lithospheric character (thick and strong to the west, thin and weak to the east) was also an important influence on Mesozoic–Cenozoic evolution of the region. The Finisterre, Sarawaget and Adelbert Mountains of northern Papua New Guinea form a thrust complex in the modern, broad, transpressive transform-fault system between the New Guinea Trench and the New Britain Trench. Based on field relationships, Findlay argues that the Finisterre Volcanics formed as an allochthonous plateau in the backarc basin or intra-arc rift-basin of the Sepik Arc which collided with Australia in Oligocene times. Keep et al. (Figure 1) discuss the evolution of the tectonically enigmatic island of Sumba which is located in a forearc position in the Australian/Indonesian convergent-margin system. Sumba lies at the major transition between subduction of oceanic Australian lithosphere beneath the Indonesian arc (to the west) and collision between continental Australian lithosphere and the Indonesian arc (to the east). Keep et al. argue that the origin of Sumba lies in the collision of a promontory of Australian continental lithosphere with proto-Sumba at ca 8 Ma. Significantly for exploration, the proposed 8 Ma age of collision correlates with the age of faulting in the Browse Basin and Timor Sea on the adjacent and prospective Australian North West Shelf. Structures created and reactivated at this time bound distinct deformational provinces on the North West Shelf and had a major influence on the fill–spill history of hydrocarbon traps. A key outcome of the volume, and one for which we strived, was to combine new data and new interpretations for the Australian Plate. Further elucidation of the dynamics and evolution of the Australian Plate will be facilitated by easy access to the key datasets. Hence we believe that Cox’s web-based collection of datasets which is easily accessible using the Web Map Server protocol is a fundamental tool for future research. The final six papers in the volume all consider the tectonic evolution of areas of the Australian Plate beyond continental Australia. The tectonic evolution of the eastern margin of the Australian Plate is still subject to enormous uncertainties, largely because it remains a frontier area for exploration and data coverage is thus poor. Unresolved questions include the time of opening of basins such as the New Caledonia and Norfolk Basins, the origin of the Three Kings Ridge, the time of inception of the Tonga–Kermadec subduction system as we know it today, the ‘tectonic ratio-

nale’ for the existence of the Cook Fracture Zone, links between subduction rate, the age of the slab, subductionhinge rollback, basin formation, dynamic topography and mantle-wedge dynamics, and many other enigmas. This group of papers is opened by a contribution by Musgrave in which he addresses misfits in New Zealand plate reconstructions that arise from assuming that the Alpine Fault has been the location of the boundary between the Pacific and Australian Plates for tens of million of years (Figure 1). Musgrave proposes that a different transcurrent fault system was active during the Early to Middle Miocene, running west of the Marlborough Sounds Block and east of the Fiordland Block. Based on a compilation of all available marine magnetic anomaly data in backarc basins east of Australia, satellite gravity data and a regional plate-tectonic model for the Southeast Indian and South Pacific oceans, Sdrolias et al. map backarc basin formation through time, and derive a revised kinematic model for the region (Figure 1). Hall and Spakman (Figure 1) combine a plate kinematic model with detailed mantle tomography images to reconcile the plate kinematic history with fast mantle velocity anomalies due to subducted slabs. This approach is an extremely valuable complement to kinematic modelling, as it helps to constrain tectonic models in convergent areas where any reconstruction for times before the Miocene is subject to many uncertainties. The eastern portion of the Australian Plate hosts a surprising number of microcontinents—continental ribbons that rift away from plate margins and either remain passive tectonic passengers embedded in oceanic crust or are accreted again close by or after traveling large distances. In a geological tour de force, Crawford et al. (Figure 1) review the evolution of the eastern Australian margin for the entire Phanerozoic and unravel analogies between the evolution of Tertiary New Caledonia and the Palaeozoic Tasman Fold Belt System. Gaina et al. utilise new geophysical data in the Enderby Basin in the southern Indian Ocean, collected by the Geological Survey of Japan, as well as regional review of plate kinematics around Australia to investigate the mechanisms of microcontinent formation (Figure 1). They find that a young mid-ocean ridge close to a mantle plume underneath a continental margin commonly ‘jumps’ onto a weak zone along the inner continental shelf, thereby isolating a continental margin segment. Like Crawford et al., they find Palaeozoic analogies to recent microcontinents, in the Neoproterozoic–Early Cambrian Gnalta and Victorian microcontinents, associated with hot-spot volcanism (Mt Arrowsmith Volcanics). The volume concludes with a contribution by Gurnis and Müller, in which they use a plate kinematic model to drive subduction and mantle convection in a geodynamic simulation. In a previously published model, they suggested that geochemical, thermal and depth anomalies in the southeast Indian Ocean, associated with the Australian – Antarctic Discordance (Figure 1), can be traced back to Cretaceous subduction east of Gondwanaland. In this volume, they present a more sophisticated model in which the Australian–Antarctic Discordance results from the sampling of both an ancient subduction-related mantle wedge, depleted by prolonged melting, and mantle cooled by the subduction system. The papers in this volume demonstrate an unprece-

Introduction

dented explosion of interdisciplinary approaches to understanding how the earth works. This includes combining geochemistry with seismology, plate kinematics with mantle tomography, and geodynamic modelling, treating the mantle and crust as one system, constrained by a variety of geological and geophysical observations. The data-driven aspect is now starting to be underpinned by user-friendly, standardised GISs. This collection of papers yields a glimpse of a future, where industry explorationists and research scientists alike will have access to surface observations, seismic tomographic, crustal geochemistry and plate kinematic models through a ‘dynamic’ GIS. Through this GIS the user will not only be able to reconstruct data through geological time, but also carry out spatial–temporal modelling based on adjacency associations (e.g. how long was this basin close to a subduction zone?), and constrain geodynamic basin models that would run on the user’s desktop workstation or a central server. Judging by the current pace of innovation, we hope to see data analysis and modelling technology evolve to this level before 2010.

Acknowledgements We are particularly grateful to Santos and Woodside for sponsoring the volume and thereby making a major contribution to its success. This sponsorship reflects an understanding that the search for mineral and petroleum resources requires to be underpinned by fundamental knowledge of the tectonic evolution of the crust. The editorin-chief of the Australian Journal of Earth Sciences, Tony Cockbain, is thanked for his enormous contribution to the

5

volume—for advising us on issues too numerous to mention and for picking up many errors missed by the scientific editors. Thanks to Misha Frankel for looking after the business side of the volume including arrangements between the Geological Societies of Australia and America. Thanks are also due to Abhijit Basu, the Geological Society of America’s Science Editor Books, for his input and for formally accepting all papers in the volume on behalf of the Geological Society of America on 19 November 2002. We wish especially to thank the reviewers of papers in the volume, namely: Rich Albert, Stephanie Baldwin, Kevin Burke, Steven Cande, David Christie, David Coblentz, Clive Collins, Stuart Crampin, Tony Crawford, Keith Crook, Jim Cull, Geoff Davies, David Denham, Mike Dentith, Will Featherstone, Robert Findlay, Thomas Flöttmann, Lisa Gahagan, Carmen Gaina, Stephen Gardoll, Anthony Gartrell, Stewart Greenhalgh, Olafur Gudmundsson, Michael Gurnis, Robert Hall, Graham Heinson, Greg Houseman, Gary Ingram, Myra Keep, Brian Kennett, Anthony Koppers, Karen Marks, Sebastian Meffre, Paul Meijer, Peter Morgan, Bob Musgrave, Martin Norvick, Anya Reading, Walter Roest, Mike Sandiford, Roger Scrutton, Heike Struckmeyer, Jon Teasdale, Anahita Tikku, John Veevers, Malcolm Wallace, Antony White, Jon Woodhead and Yanhua Zhang. R. R. HILLIS1 R. D. MÜLLER2 1 National Centre for Petroleum Geology and Geophysics, University of Adelaide, SA 5005, Australia 2 School of Geosciences, University of Sydney, NSW 2006, Australia

Geol. Soc. Australia Spec. Publ. 22, and Geol. Soc. America Spec. Pap. 372 (2003), 7–23

Seismic structure in the mantle beneath Australia B. L. N. KENNETT Research School of Earth Sciences, Australian National University, ACT 0200, Australia ([email protected]). The configuration of earthquake belts around Australia provides a wealth of events at suitable distances to be used as probes into the seismic structure of the upper mantle. The limited number of permanent high-fidelity seismic stations on the continent has been supplemented with extensive deployments of portable broadband stations for periods of a few months at each site. The combination of a long term of recording at the permanent stations and the broad spatial coverage of the portable stations provides an excellent resource for studies of the mantle. A wide range of techniques can be used to gain information on the three-dimensional structure in the mantle, which exploit different aspects of seismograms. The large-amplitude surface waves in the later part of the seismogram travel nearly horizontally and can be used in a tomographic inversion to determine the 3-D variations in shear-wave speed. This approach relies on matching the waveforms on individual paths and then mapping of the path-specific constraints on shear structure into a 3-D model. In contrast, the higher frequency body-wave arrivals are refracted back from the variations in structure in the mantle and are particularly sensitive to discontinuities in structure. Observations out to 3000 km provide coverage of the structures down through the transition zone and for the region below northern Australia, the combination of short-period and broadband observations has provided detailed information on both P and S wave speeds and attenuation structure. Further information on lateral variations in structure can be extracted from the patterns of travel-time residuals. The combination of the different classes of results reveal a complex pattern of 3-D structure beneath the Australian region. The cratonic region in the centre and west of Australia is underlain by a thick mantle lithosphere extending to around 210 km depth with fast wave speeds (especially for S waves). In the asthenosphere below, the S wave speeds diminish and there is significant attenuation and also some level of seismic anisotropy. Beneath the eastern zone with Phanerozoic outcrop the lithosphere is generally thinner (less than 140 km) and the asthenosphere has a pronounced low-velocity zone for S, again with high attenuation. The variations in seismic-wave speeds extend through the upper mantle with noticeable differences in the transition zone. There is also evidence for pervasive small-scale heterogeneity (scale lengths of 100–200 km) superimposed on the broader scale variations that can be imaged using tomographic methods. KEY WORDS: asthenosphere, lithosphere, seismic tomography, upper mantle.

INTRODUCTION The exposed geology of the Australian continent is composed of an assemblage of crustal blocks that can be broadly grouped into the Precambrian western and central cratons and the Phanerozoic eastern province (Figure 1). Structural differences between the Precambrian shield and eastern Australia are inferred from surface-wave dispersion (Muirhead & Drummond 1991; Denham 1991) and teleseismic travel-time residuals (Drummond et al. 1991) whose origin is due to structures that certainly extend below 100 km depth. The extensive seismic activity in the earthquake belt that runs through Indonesia, New Guinea and its offshore islands, Vanuatu, Fiji and the Tonga–Kermadec zone provides an abundant source of seismic probes for the structure in the lithosphere and the upper mantle beneath. A wide range of studies has exploited the different aspects of the seismic wave train from the P and S body waves refracted back from the velocity structure in the upper mantle through to the large amplitude surface-wave trains which travel nearly horizontally on their path from source to receiver.

The Australian continent itself has a fairly low level of seismicity and only a few high-quality seismographic stations. The study of structure beneath the continent has therefore depended on the deployment of portable seismic instrumentation. Up to 1993 the experiments involved the deployment of vertical-component seismometers (with natural frequency around 1 Hz), with firstly analogue recording, and later low-power digital reel-to-reel tape recorders (Figure 2a). The first portable broadband instruments were used in 1992, and from 1993 the emphasis has been on continent-wide coverage using a mobile network of stations (Figure 2b). In the SKIPPY experiment from 1993 to 1996 (van der Hilst et al. 1994), stations were deployed across the whole continent at approximately 400 km spacing. Because only a limited number of broadband seismometers and highfidelity recorders were available, the continental coverage was achieved using between 8 and 12 stations at a time, installed for about 5 months in each location. This period is sufficient to get good coverage of the regional seismicity but is a little short for some classes of studies that depend on less-frequent teleseismic events at specific distance

8

B. L. N. Kennett

Figure 1 Summary of surface geology and crustal units of Australia.

ranges. In 1997 and 1998 a denser array of broadband instruments (KIMBA) was deployed in the remote Kimberley region of northwestern Australia to follow up indications of contrasts in cratonic structure. In 1999 additional stations were deployed in southeastern Australia (QUOLL) to provide additional information on the transition between Phanerozoic and Proterozoic structure. Western Australia was revisited in 2000–2001 with a broadly spread array designed to supplement the SKIPPY stations, which had had a number of technical problems. Sets of more closely space instruments are intended to provide more detailed information on the substructure of the Archaean blocks and the links between them. In less than 10 years therefore there has been a very thorough coverage of the Australian continent with broadband seismic stations. These stations provide information for many different styles of analysis for the 3-D seismic wave-speed distribution, attenuation and anisotropy. The experiments with the short-period recorders provided valuable information on the nature of the radial distribution of seismic-wave speeds in the mantle, and also indicated the presence of lateral heterogeneity in seismic wave speeds in the upper mantle on a variety of scales. The depth profiles for different azimuthal sectors with sampling regions differing by 1000 km show significant differences. The character of the wavefield variations across the recording arrays indicate pervasive medium-scale heterogeneity

with a scale length of 200 km or less. The high-frequency body waves travelling through the cratonic lithosphere have an extended coda indicating very strong scattering due to small-scale variability with very little attenuation. The impetus for moving to broadband studies using three-component recording came from the disappearance of S waves in the short-period vertical records from northern Australia for epicentral distances beyond 18° (Dey et al. 1993). With a broader range of frequency content it became apparent that once the S waves penetrated below the cratonic lithosphere (about 200 km thick) they encountered a zone of slightly lowered seismic-wave speeds and much higher attenuation. S waves returned from below this asthenospheric layer were of sufficiently low frequency to be poorly recorded on short-period instruments (Gudmundsson et al. 1994). The broadband seismometers provide good recordings of the surface waves of earthquakes from the neighbouring belts of seismicity with surface-wave magnitude (Ms) greater than 5.5. For these events we can exploit the source mechanisms in the Centroid Moment Tensor catalogue produced by Harvard University (Dziewonski et al. 1999) and use the waveforms in inversion schemes designed to extract 3-D shear-wave structure. The very large number of surface wave paths crossing the continent provide good azimuthal control, particularly in central and eastern Australia, and so it is possible to extract information on azimuthal anisotropy

Mantle beneath Australia

9

Figure 2 Deployments of portable seismic recorders in Australia by the Australian National University. (a) Short-period recorders, 1975–1993. (b) Broadband recorders 1992–2000. The different colours indicate separate instrument deployments.

10

B. L. N. Kennett

in seismic-wave speeds. With three-component recording we can study both the Rayleigh waves with polarisation in the vertical plane and Love waves with purely horizontal polarisation. The differences in the structures inferred by analysis of the different wavetypes requires polarisation anisotropy.

far-field of the source (Kennett 1995). The fundamental Rayleigh mode places the strongest constraints on the region above 250 km, and the higher modes add information at depth. The excitation of higher modes is enhanced as

SURFACE-WAVE STUDIES The most prominent feature of seismograms from shallow earthquakes are large-amplitude surface-wave trains. With portable broadband instruments it is possible to achieve good recordings of the surface waves to periods around 100 s, and at typical distances from the source of 2000 km this means that it is possible to sample well the zone down to 400 km. Much of the analysis of surface waves has concentrated on the Rayleigh waves recorded on the vertical component. The horizontal components of ground motion tend to have higher levels of ambient noise than the vertical component, but for larger events it is possible to exploit the Love waves, with horizontal polarisation, to provide additional information on the upper part of the mantle. The traditional tool for the analysis of surface waves is the determination of surface-wave dispersion and the mapping of this information into a depth profile of shear-wave speeds. However, the limited number of permanent stations in Australia meant that only a few great-circle paths were available (Denham 1991) and these clearly indicated substantial differences between central and eastern Australia. The dispersion results for the east coast of Australia indicate the presence of a low-velocity zone for shear-wave speed around 150 km depth (Goncz & Cleary 1976), but this is not present in central Australia. The SKIPPY deployments of broadband instruments were designed to exploit advances in the analysis of surface waves via waveform inversion (Nolet 1990; Zielhuis & Nolet 1994). This ‘Partitioned Waveform Inversion’ (PWI) is a two-stage procedure. The first step generates a shearwave speed model for each path and the second combines the path-dependent information into a 3-D model. The first stage of this procedure is based on the matching of observed and calculated seismograms for the surface waves for each source–receiver path, which requires a knowledge of the source mechanism. The inversion for each path generates a radial profile of shear-wave speed which is interpreted as the average structure along the great-circle path between source and receiver. The dependence of the waveforms on shear-wave speed is non-linear and the results are sensitive to the starting shear structure assumed in the inversion. Synthetic tests demonstrate that the influence of fixing the P wave-speed distribution is small. The modelling procedure assumes the independent propagation of individual modes along a great-circle path and thus a smooth change in wave-speed structure. For Rayleigh waves two different group velocity windows are analysed: for a window around the fundamental mode a bandpass filter from 0.010 to 0.025 Hz is applied and for the earlier window covering higher modes the filter extends from 0.008 to 0.050 Hz. These frequency bands are chosen to minimise the influence of deviations from the great-circle path and to ensure that the seismograms are recorded in the

Figure 3 Coverage from Rayleigh waves for surface-wave tomography. (a) Paths for which both the fundamental mode and overtones could be analysed. (b) Paths for which only the overtones were used (usually from deep events). (c) Paths for which just the fundamental mode was analysed (mostly rather shallow events with poor excitation of the overtones).

Mantle beneath Australia

the source depth increases, whilst the fundamental mode diminishes. The combination of fundamental mode and higher mode information in the waveform fitting helps to constrain the different aspects of the velocity profile. However, sometimes the effect of heterogeneity in the Earth is such that it is not possible to fit both of the waveform windows with the same wave-speed profile. An example is for paths crossing the southern Tasman Sea, where there is a very large drop in seismic-wave speed at shallow depth. The presence of the subduction zones around Australia provides a valuable source of intermediate and deep events, but is concentrated to the north and east. The coverage at depth is therefore somewhat different from that provided by the shallower events which occur to the south as well (Figure 3). The second stage of the PWI scheme is a linear inversion to generate a 3-D shear-wave speed model. With the interpretation of the models for each path as the average structure, the path information provides linear constraints on the 3-D structure. The second-stage inversion then reconciles the different constraints to produce the 3-D model. Large deviations from the reference model can be accommodated because of the linear dependence of the ‘data’ (the pathaveraged models) on the 3-D model parameters. Both model norm and gradient damping are used to achieve a balance between data fit and smoothness of the model. The reliability of the resulting model is highest in those regions where there are multiple crossing paths. At this time the path density is best in central and eastern Australia and the 3-D model is less well constrained to the west of 120°. The application of the PWI technique to the SKIPPY dataset proceeded as the data were being collected. Zielhuis and van der Hilst (1996) presented the first model based on the analysis of the Rayleigh wave data from the stations in eastern Australia. This already indicated the presence of a very major contrast in the shear-wave speed in the mantle component of the lithosphere between central Australia and eastern Australia. The 3-D model indicated the presence of lowered shear-wave speeds beneath the east coast of Australia at depths around 140 km and confirmed the inferences from surface-wave dispersion. The location of the zones of low wave speed has a strong correlation with Neogene volcanism. The 3-D models have evolved as more data have been incorporated. Using the PWI technique with vertical-component waveforms, results have been presented by van der Hilst et al. (1998) and Simons et al. (1999) based on a cellular representation of the model with blocks of about 1 x 1°. Figure 4 shows the shear-wave speed distribution for SV waves at 80 and 200 km depth derived from a PWI inversion using approximately 2000 Rayleigh wave paths, and indicates the large deviations in seismic-wave speed from the reference value of 4.5 km/s. The sampling by surfacewave paths is lower in the western part of Australia and features to the west of the 120° meridian should be treated with caution. The areas of exposed Precambrian rocks largely correspond to high seismic-wave speeds but there are zones of enhanced wave speed extending to 150 km or more lying to the east of the conventional Tasman line (Figure 1). The Precambrian regions show the presence of significant internal structure with an indication of the separation of the major cratonic blocks, especially at shallower depths.

11

The contrasts in shear-wave velocity are more pronounced at 80 km depth, but even at 200 km depth we see variations of the order of 8% in shear-wave speed. The major feature beneath the continent is the lowered wave speeds along the eastern margin as compared with the raised wave speeds in the centre and west. The variations in wave speed are such that a purely thermal interpretation is difficult to sustain (see the Conclusions section). There is not a simple relation between the structures in the mantle and the conventional Tasman line (which represents the eastern limit of surface exposure of Precambrian material). Rather the major change in mantle structure occurs on an approximately north–south trend close to 140°E. This is marked by a zone of very high horizontal velocity gradient extending through Cape York and central Queensland to depths below 150 km, which links to slightly weaker features in southern Australia. To the west of this line the higher velocities extend coherently to 200 km or below, whilst to the east although there is an area of elevated wave speed it does not extend to the same depth. The Phanerozoic belt in eastern Australia has a thin zone of high wave speeds in the lithosphere extending to about 100 km, and beneath this there is a zone of lowered wave speed that extends along most of the east coast of Australia. The low seismic-wave speeds extend to the east beneath the Tasman Sea where sea-floor spreading ceased at about 80 Ma. Subsequently the Tasmantid seamounts have been emplaced with a progression of volcanic edifices as Australia has moved towards the north over a presumed mantle plume: the predicted position of the plume based on current plate motions corresponds to a minor sea-floor feature and is a centre of earthquake activity. The PWI inversions have provided significant insights into the nature of the mantle lithosphere. The upper part (80 km depth image) shows clear signs of influence from tectonic events: e.g. the lowered wave speeds in the southern part of central Australia in the region affected by the Alice Springs Orogeny at about 300 Ma. Regions such as the Kimberley Block in northwestern Australia appear to maintain a distinct character compared with their surroundings to significant depth. This suggests that the lithosphere is able to retain its character over very long periods of time and that there is the potential for using the ‘geology’ of the mantle revealed in the seismic images to try to track back into the assemblage of the lithosphere. In central and western Australia the high seismic-wave speeds extend to about 200 km depth and the eastern boundary of this high wave-speed zone lies close to 140°E, and does not follow the configuration of the Tasman line. The abruptness of the transition from high wave speeds to lower values is sufficiently rapid to test the assumptions made in the inversion. The rapid change has the potential of introducing mode coupling and is also such as to produce some deviations of surface-wave propagation paths from the great circle from source to receiver, especially for paths travelling along the transition zone between higher and lower wave speeds. The next generation of inversion schemes will need to take account of such effects and include an iterative development for the construction of the 3-D model. Marquering et al. (1996) have shown how an approximate treatment of mode coupling can improve the

12

B. L. N. Kennett

Figure 4 Map views of the 3-D shear-wave model derived from partitioned waveform inversion using the permanent stations and the deployments of broadband stations: (a) 80 km depth; (b) 200 km depth. The red lines indicate the plate boundaries along which the majority of earthquakes used are sited.

Mantle beneath Australia

treatment of higher modes at shorter periods by providing a better representation of S body-wave propagation. An alternative approach to the development of 3-D models has been pursued by Debayle (1999) and Debayle and Kennett (2000a, b) again using a two-step inversion, but with a different style of waveform inversion for each path based on the work of Cara and Lévêque (1987) and a second-stage inversion using the continuous regionalisation method of Montagner (1986). The second-stage inversion uses the path information to extract not only the variations in seismic-wave speed but also the azimuthal anisotropy in the SV wave speed using the approach of Lévêque et al. (1998). In a weakly anisotropic medium the variation of Rayleigh wave speed with azimuth  is expected to have cos 2, sin 2 dependence. The local direction of fast shear-wave speed can be extracted in regions where there are sufficient crossing paths. The wavespeed inversion imposes an intrinsic smoothing based on an imposed Gaussian correlation function: a width of 500 km has been found to be satisfactory (for further details see Debayle & Kennett 2002). Surface-wave analysis requires the passage of seismic waves across and into the region of interest and can be carried out in most parts of the world, but Australia is one of the few places where it is possible to also use body-wave information as a complement and check on the results of surface-wave tomography.

BODY-WAVE STUDIES The configuration of seismic sources around Australia allows the use of body-wave analysis using waves refracted back from the structure in the mantle out to distances of 3000 km from the source. These waves sample through the whole upper mantle and transition zone, and have their maximum sensitivity to velocity structure near their turning point (close to the midpoint of the path from source to receiver for shallow sources). Unfortunately, the separation between the natural sources to the north of Australia and stations on the continent is such that it is difficult to use refracted wave arrivals (for either P or S) to constrain shallower structure and resolution of structure is best below 150 km depth. The pattern of velocity gradients and discontinuities in the mantle imposes a complex structure on the expected form of seismograms for even a 1-D wave-speed profile (Figure 5). The situation is complicated further by the 3-D structure imaged in the surface-wave work and local smallscale heterogeneity. Earlier studies concentrated on the P wave-speed distribution using arrays of short-period portable instruments on a variety of scales. Subsequently with the advent of high-fidelity broadband recording it became possible to work with S waves as well.

Short-period studies Initial efforts were made to delineate the major features of the mantle velocity profile by using arrays of short-period stations. In a pioneering study Hales et al. (1980) used travel-time analysis of record sections from Indonesian earthquakes at a variety of depths. The span of the recording arrays was such that only part of the wavefield was captured

13

for each earthquake and the velocity profile was pieced together using the information from multiple events. The resulting model, for upper mantle structure under the northwestern part of the Australian margin, is rather complex with many small discontinuities and low-velocity zones, which may reflect the mapping of 3-D structure into a 1-D profile. A subsequent reinterpretation of this data by Leven (1985), using comparisons between observed and synthetic seismograms, leads to a somewhat simplified structure but retains a prominent velocity contrast near 210 km depth. A number of deployments of short-period vertical seismometers were designed to exploit the natural seismicity in the Indonesia/New Guinea region in studies of mantle structure. The distance span of the deployments covers only a limited portion of the triplications produced by the upper mantle discontinuities. Thus a useful tool in such studies is the assembly of seismograms from many events and stations to produce a composite record section in which the phase branches for the different mantle arrivals can be tracked (Figure 5); all arrivals in a 10 km epicentral distance range are combined into a single trace. The influence of variable source-time functions is reduced by stacking the envelope of the seismograms (Bowman & Kennett 1990). This procedure minimises the influence of local heterogeneity and can produce striking results for P waves with clear delineation of phase branches as illustrated in Figure 5a. The patterns of arrivals can then be in terms of the velocity distribution with depth. The shallow structure has to be inferred, but the P velocity structure is well constrained from above the base of the lithosphere near 210 km down to below the 410 km discontinuity. The shortperiod results require a P velocity contrast near 210 km depth to explain multiple arrivals with a few seconds separation around 1400 km from the source. However, for S waves the corresponding record sections only show a clear arrival associated with the lithosphere which cannot easily be traced beyond 2000 km, and no branches associated with greater depth. Synthetic seismograms from a reference model for northern Australia are included in Figure 5b as an aid to identifying the arrivals; note that this model is not designed to provide a direct match to the observations. Although the nature of the phase branches in the composite sections can be well summarised by a single 1-D wave-speed profile, there is commonly substantial variability between events. Bowman and Kennett (1990) have presented examples of individual events in the Flores arc with very strong arrivals associated with the 410 km discontinuity, even though other events from nearby locations show much more subdued arrivals. This indicates the presence of significant focusing and defocusing of short-period seismic energy due to heterogeneity. Using a variety of different sources of information, including the nature of signals on arrays of different aperture, Kennett and Bowman (1990) have suggested the presence of pervasive velocity heterogeneity in the upper mantle on scales of 100–200 km with an amplitude of about 1%. This small-scale heterogeneity is superimposed on the larger scale variations that have been imaged in the surface-wave studies and which is also evident in the nature of the body-wave arrivals. Dey et al. (1993) have summarised much of the shortperiod work in northern Australia and presented composite record sections of upper mantle arrivals that show marked

14

B. L. N. Kennett

Figure 5 (a) Composite record section of short-period seismograms at stations in the Northern Territory from shallow events in the Flores arc. (b) Synthetic seismogram calculations for a reference model for northern Australia indicating the principal seismic phases which can be seen in the observations. The travel-time branches are denoted by the discontinuity with which they are associated (r, refraction; R, retrograde reflection).

variation in P wave velocity structure between paths for events along the Flores arc, and paths to events in New Guinea. The turning points in the mantle for the two profiles differ by nearly 1000 km and indicate larger contrasts at the 410 km discontinuity for the paths to New Guinea and differences in transition zone structure.

Within the Australian continent most of the information on P velocity structure down to 200 km has come from various experiments with large explosive sources (Muirhead & Drummond 1991). The infrequent natural events have also been exploited. Bowman and Kennett (1991) used the aftershocks of the 1988 Tennant Creek earthquakes to

Mantle beneath Australia

15

investigate regional S wave propagation in central Australia and were also able to infer the velocity profile in the crust and uppermost mantle. These aftershocks were also used by Bowman and Kennett (1993) to develop a set of travel times for P and S waves travelling in the shield structures of western Australia, and to find a compatible velocity model for the lithosphere.

Broadband studies Modern broadband seismometers provide a faithful rendition of ground motion over a wide range of frequencies and allow the full exploitation of both P and S body waves. Once such broadband data became available, the reason for the absence of S waves at distances beyond 2000 km was clear. The S waves within the lithosphere are high frequency, but once they penetrate into the asthenosphere beneath there is a dramatic drop in frequency content. This effect implies a major contrast in the attenuation properties of S waves. There is very little attenuation of S in the lithosphere, but strong scattering leading to a complex coda. However, the asthenosphere beneath the cratonic lithosphere has both a slightly lower wave speed than the lithosphere above and much higher attenuation. The initial results were obtained from a broadband sensor operated by the Research School of Earth Sciences (Australian National University) at the Warramunga array in northern Australia since late in 1988 (station code WRA: Figure 2). Over a period of years it was been possible to build up record sections covering the range of interest for the upper mantle by using events in the Indonesia/New Guinea earthquake belt for both P and S waves (Gudmundsson et al. 1994; Kennett et al. 1994). The surface conditions in northern Australia make it possible to exploit both horizontal components of S waves after rotation to the great-circle path. Thus both the transverse SH component and the vertically polarised SV waves on the radial component records can be used: the high surface velocities lead to little contamination by converted P waves.

WAVE-SPEED VARIATION The WRA observations have been used to generate composite record sections from many events for P and S waves and, in association with information from the short-period studies, used to build up velocity profiles for both P and S (Kennett et al. 1994). The analysis includes comparison of observed and synthetic seismograms with inclusion of attenuation. Because the P and S wave speeds are determined from the same events, the P/S velocity ratio can be well constrained and is similar to that for the shield areas of North America derived by combining separate P and S models (LeFevre & Helmberger 1989; Grand & Helmberger 1984). Additional information on upper mantle structure can be obtained from the broadband stations deployed in SKIPPY and later experiments. The 400 km spacing for the SKIPPY stations means that the constraints on mantle structure for an individual event are low, but good coverage can be achieved by suitable combinations of data from many events. Kaiho and Kennett (2000) have exploited such composite record sections to obtain a suite of P and S velocity

Figure 6 Configuration of shallow events indicated by red diamonds recorded at the KIMBA stations marked by triangles. The turning points for the available paths are indicated by small grey triangles.

profiles for 16 azimuthal corridors across the Australian continent. Each 10°-wide corridor includes all source–receiver pairs with path azimuth with ±10° of the orientation of the corridor, and 0.5° bins were used in constructing composite record sections using a variety of stacking techniques. A similar approach can be used for regional deployments such as the KIMBA experiments, which lie in a convenient position relative to the seismicity to the north. This allows sampling of the continental mantle beneath the northward extension of the Australian Shield with a nearly east–west orientation (Figure 6). A single event can only be observed over a 5° distance span, but coverage of the full upper mantle profile can be achieved by using data from many events. Composite P and SH record sections for the KIMBA profile for events to the east are displayed in Figure 7, with event stacking in 0.3° bins using an envelope stack. The events were corrected to a common source depth of 25 km before stacking. The travel-time curves for the ak135 reference model (Kennett et al. 1995) for a 25 km-deep source are superimposed on the sections. As can be readily seen, both the P and S arrivals for distances less than 18° arrive before the expected times for the reference model, indicating the need for higher lithospheric wave speeds than for the ak135 model. The discrepancy is greater for S than P. At 13° the onset of S is about 17 s early compared to the ak135 times, requiring an increase in wave speed of about 5% along the path. The P and S sections show complexity in the character of the lithospheric arrivals out to 17–18° suggesting both the need for structure near 200 km depth and some degree of lateral variability in wave speed. The P arrivals remain slightly early compared to the reference times out to 26°. Beyond 20° the S wave onsets are both significantly longer period and quite close to those predicted by ak135. This behaviour can be achieved by introducing a modest reduction in S wave speed below 200 km to create a lowvelocity zone, which also has enhanced attenuation. The record sections in Figure 7 are characteristic of paths sampling the cratonic regions of northern Australia, but very different from paths to stations in eastern Australia from events to the northeast. Kaiho and Kennett (2000 figure 11) presented a summary of travel-time perturbations for epicentral distances out to 20°, which shows clearly the

16

B. L. N. Kennett

Figure 7 Composite record sections obtained by stacking rotated seismograms at the KIMBA arrays for the events in Figure 6. The records were adjusted to a common depth of 25 km before stacking using 0.3° bins. (a) P wave section for stacked vertical component records. (b) S wave section for stacked tangential component records. The travel times for a 25 km-deep event in the ak135 model are displayed with the seismogram to provide a reference for the phase branches.

transition from fast to slower paths across Cape York in the position predicted by the 3-D models derived from surfacewave tomography. As in Figure 7, the relative travel-time anomalies are much larger for S than P, implying a relative deviation of about 3 to 1 from the ak135 reference model. A selection of P and S wave-speed profiles are plotted in Figure 8 to illustrate the variety of behaviour encountered beneath the Australian region. The shield models (1–4) show a very fast S wave speeds in the lithosphere underlain by a slight low-velocity zone. The corresponding P profiles show only slightly higher wave speeds than the ak135 reference model (6). For paths in the mantle beneath the Phanerozoic zone (5) the thickness of the lithosphere is substantially reduced, and has lower wave speed. The velocity models determined from the composite record sections are derived from single-ended profiles and should be interpreted as representing the wave speeds at the position corresponding to the turning point for the arrival rather than as a vertical profile. The model can therefore be

associated with the horizontal position of the appropriate turning points. Kaiho and Kennett (2000) have exploited this interpretation to combine the 1-D models for 16 different azimuthal corridors into a pseudo-3D velocity model (Figure 9). Because of the use of the turning-point property there is no mapping of portions of velocity models in low-velocity zones. However, we still see the clear contrasts in lithospheric properties in Figure 9 which match well with the surface-wave results. Considerable variation in seismic-wave speeds are found through the asthenosphere and mantle transition zone (Figure 9 260–310 km, 360–410 km).

ATTENUATION STRUCTURE The relative frequency content of P and S waves provides a visual guide to the relative attenuation of P and S. This approach can be quantified by working with the slope of the spectral ratio between S and P windows on the same

Mantle beneath Australia

17

Figure 8 Comparison of 1-D velocity profiles for P and S waves in the Australian region. 1, azns, north–south sampling of mantle below shield (Kaiho & Kennett 2000); 2, kmba, east–west sampling of mantle below shield from KIMBA stations; 3, waus, from travel times for Tennant Creek earthquake aftershocks recorded in western Australia (Bowman & Kennett 1993); 4, njpb, from analysis of records at WRA (Gudmundsson et al. 1994); 5, azne, sampling of mantle beneath eastern Australia (Kaiho & Kennett 2000); 6, ak135, reference model (Kennett et al. 1995)

seismogram to extract a quantitative estimate of differential attenuation (t*). Gudmundsson et al. (1994) used the broadband observations at WRA to construct the variations of t* with epicentral distance. With a knowledge of the velocity structure it is possible to invert these t* observations for a simple layered Q model with depth beneath the shield region. The lithospheric attenuation down to 210 km is very low and Q, for S waves, exceeds 1000. In the asthenosphere beneath, significant attenuation is required with Q around 120, if the attenuation is spread over the whole zone between 210 and 410 km depth. Even lower Q values would be needed if the attenuation were concentrated in a thin layer. The Q value recovers to around 500 in the transition zone and appears to be higher again in the lower mantle below 660 km.

The style of analysis employed by Gudmundsson et al. (1994) for 40 events at WRA has now been extended to nearly 2000 paths covering the continent using the broadband records from the SKIPPY and KIMBA experiments (Cheng & Kennett 2001). Analysis of the spectral ratio between S and P waves for frequencies (ƒ) up to 6 Hz yields the differential attenuation (t*) for a low frequency band around 0.6 Hz, and an estimate g of the averaged frequency exponent along the path (corresponding to frequency dependence Q = Qoƒ. There is a very substantial geographic variation in t*, with a strong correlation between large differential attenuation and enhanced frequency variation. The cratonic lithosphere with fast wavespeeds has very low t* with almost no frequency dependence. Whereas the paths sampling the asthenosphere have much

18

B. L. N. Kennett

Figure 9 Estimates of 3-D wave-speed structure for P and S waves derived from body-wave analysis (Kaiho & Kennett 2000): 165–210 km; 260–310 km; 360–410 km.

enhanced t* and significant frequency dependence ( around 0.5). With the aid of the 1-D velocity models of Kaiho and Kennett (2000) for the different azimuthal corridors, the differential attenuation as a function of both epicentral distance and frequency has been inverted using the Neighbourhood Algorithm of Sambridge (1999) to produce a set of profiles of Q and the local frequency exponent a, as a function of depth. The apparent Q is an integrated property along the path and, with the aid of the vertical and horizontal ray-path density, the results for the different azimuthal corridors have been combined into an estimate of the 3-D structure in both Q and its frequency dependence (Cheng & Kennett 2001). This model is based on a five-layer representation of the attenuation structure. Figure 10 illustrates the geographic variations in attenuation at depths around 150 and 300 km. Beneath the shield the behaviour matches well with the results for WRA. However, here is a profound contrast between the fast lithosphere of west-central Australia and the elevated asthenosphere beneath the eastern part at 150 km, compared with a more modest variation at 300 km depth. As

noted above there is a strong correlation between regions of lowered Q and enhanced frequency variation.

REFRACTED-WAVE ANISOTROPY For many paths sampling the upper mantle transition zone there is an indication of shear-wave splitting in refracted S waves. Tong et al. (1994) have analysed data for events in a broad azimuth range recorded at WRA. For arrivals from the upper mantle discontinuities the S wave is systematically earlier, by more than 1 s, on the tangential (SH) component than the radial (SV) component. This information on the seismic anisotropy for the refracted S wave arrivals has been analysed using correlation analysis for a set of time windows on the seismogram associated with the different phase branches to determine the direction of fast propagation and the time shift between the S components for that branch. The fast directions for groups of events to the west and east of WRA are oriented approximately transverse to each set of paths, so that they are incompatible with material anisotropy confined to the lithosphere

Mantle beneath Australia

19

Figure 10 Map view of sections through (a) 3-D models of Q and (b) its frequency exponent a at depths around 150 and 300 km.

beneath the WRA station. A level of anisotropy of the order of 1% in both the lithosphere and the asthenosphere beneath would explain the data quite well.

Travel-time tomography A further source of information on velocity structure in the Australian region comes from travel-time tomography studies using sources both in the region and from around the globe. This technique exploits the deviations in travel times from those for a reference model and an inversion is undertaken to determine the 3-D distribution of wave speed compatible with the observations. This approach was pioneered for the Australian region by Widiyantoro and van der Hilst (1996) who used the travel-time residuals for P and pP for teleseismic paths, as well as paths to Australian stations (both permanent and portable) in a tomographic inversion for structure in a zone covering the southern Philippines, Malaysia, Indonesia, Papua New Guinea and northern

Australia. In order to minimise the influence of structure external to the region, their inversion for P wave velocity followed the approach introduced by Inoue et al. (1990), in which a detailed grid is used for the region of interest and a coarser grid for the region outside. The resolution attained was of the order of 100 km, both horizontally and in depth, so that the details of the high-velocity subducted slabs can be recognised. A major effort has been made to relocate seismic events across the globe using the ak135 reference model (Kennett et al. 1995) and then reassociate catalogue readings to provide a comprehensive set of information on the travel times of P, S and many other phases (Engdahl et al. 1998). This provides a high-quality database from which to conduct regional delay-time tomography for both P and S waves. The reading of S arrivals is more difficult than for P, at the onset of the seismogram, and coverage is somewhat less than for P, but still sufficient to provide satisfactory images of structure in the subduction zones (Gorbatov & Kennett 2001)

20

B. L. N. Kennett

Figure 11 Tomographic images derived from inversion of the arrival times of both P and S waves from layers around 135 and 245 km depth. A common scale of perturbation is applied for each panel. The red lines indicate the plate boundaries.

Figure 11 shows map views of the P and S wave structure for Australia and its immediate surroundings at depths around 135 and 240 km. The tomography is carried out using a cellular representation of the wave-speed model with 0.5 x 0.5° cells from the surface to 210 km and 1 x 1° cells beneath. The thickness of each of the layers is around 50 km. Resolution is good in the neighbourhood of the subduction zones because of the concentration of sources. Beneath the continent much of the information is imparted by nearly vertically travelling rays from very distant earthquakes and so sampling is concentrated around the stations. The P and S wave images are plotted using the same scale of perturbation and we can see the high relative variations associated with S waves. The main subduction zones are clear in both the P and S images at 135 km depth, and the fast wave speeds in the Australian craton show up in patches reflecting the available path distribution. At 245 km the amplitude of heterogeneity is much reduced beneath the continent indicating that the base of the lithosphere has been reached; note that there will inevitably be some level of vertical smearing of anomalies. However, there is a fast S wave-speed structure in the oceanic region

in front of the Flores arc in Indonesia which does not have an equivalent in the P wave-speed map at 245 km. This suggest some oceanic–continental dichotomy in the nature of the lithosphere.

CONCLUSIONS Over the last 15 years extensive progress has been made in understanding of the nature of the distribution of seismic structure in the mantle beneath Australia. The nature of the regional earthquake sources means that it has been possible to undertake a wide range of studies and to produce an integrated result in a way that has not so far been accomplished for any other continent. Surface-wave studies of the Australasian region have provided broad-scale coverage of 3-D shear-wave speed structure with good resolution currently available to the east of the 120° meridian. Beneath the cratonic regions high seismic-wave speeds extend to depths around 220 km and have a relatively abrupt eastern margin that does not correlate simply with the Tasman line marking the eastern extent of current Precambrian outcrop. The surface-wave

Mantle beneath Australia

21

Figure 12 Map view of the 3-D shear-wave speed structure at 140 km plotted with the neutral tone shifted to 4.6 km/s (a +2.5% velocity perturbation from the reference). The darkening red tones indicate the increasing influence of thermal effects and the patterning the portion of the heterogeneity which may well have a chemical component. The red lines indicate the plate boundaries.

results also indicate the presence of substructure in the mantle within the high wave-speed zones at least to depths of 100 km. The eastern seaboard of Australia and the Tasman have markedly lower shear-wave speeds than the centre and west of the continent, with a marked zone of lowered seismic velocities at depths around 140 km. Body-wave travel times provide an independent check on the nature of lithospheric structure and indicate the need for much larger variability in S wave speed than for P waves. The northern Australian craton shows very high S wave speeds with a rather sharp transition to slower paths occurring in the Cape York peninsula. The full set of refracted body waves from the events in the earthquake belts to the north and east of Australia can be exploited to provide an independent estimate of 3-D wave speed structure for both P and S waves. The results tie well with the surface-wave results for S and the concordance of the different styles of analysis means that we can have considerable confidence in the main features of the structures in the mantle. The regions of high seismic-wave speed beneath the cratons show very little seismic attenuation, but there needs to be a sharp change at depth since waves which penetrate into the asthenosphere rapidly lose high-frequency energy. The cratonic lithosphere is thus underlain by a zone of

slightly lower S wave speeds and significant attenuation. The zones of lower seismic-wave speed in the east of Australia are accompanied by enhanced attenuation with a noticeable frequency dependence. The levels of variation of S wave speed encountered across the Australian region are large. At 150 km deep the S wave speed varies from about 4.9 km/s in the faster parts of the cratonic lithosphere to close to 4.1 km/s in the southern Tasman Sea. Relative to a reference velocity of 4.5 km/s this represents variations from +8% to -8%. Experiments on shear-wave speeds at seismic frequencies indicate that it is possible to achieve substantial reductions in seismic-wave speed and enhanced attenuation as synthetic materials with mantle compositions approach the solidus (Jackson 2000). The wave-speed variation with temperature is non-linear with rapid variation at high temperatures, and would be enhanced by the presence of volatiles. However, at low temperatures the behaviour is close to linear with only a modest temperature coefficient. It would therefore appear that a substantial fraction of the variations in seismic-wave speed will have a thermal origin, which fits with the association of anomalies in the mantle with the regions which have undergone Neogene volcanism. Goes et al. (2000) have sought an entirely thermal explanation for the somewhat milder variations in

22

B. L. N. Kennett

shear-wave speed in Europe. The exceptionally fast shearwave speeds in the mantle beneath the Australian Shield are difficult to explain with any plausible thermal model and suggest that some degree of chemical variability is required. As an indication of the possible segregation of compositional and thermal effects, the shear-wave speed at 140 km depth (from the study depicted in Figure 4) is plotted in Figure 12, with the neutral tone set at 4.6 km/s (a 2.5% increase on Figure 4). The neutral zone then separates the very high velocities with a potential chemical heterogeneity (indicated by patterning) from the lower velocities (in red tones) for which thermal effects are likely to be the dominant influence. The intensity of tone gives an impression of the influence of temperature. We can hope to get further constraints on the nature of the seismic structure when we can obtain comparable quality images of P and S wave speed. This is not easy because of the absence of any P wave analogue of surface waves and will be a goal for future studies.

ACKNOWLEDGEMENTS The study of mantle structure beneath Australia has involved many members of the Seismology Group at the Research School of Earth Sciences, Australian National University, Canberra, both in the field and in subsequent analysis. I would particularly like to thank Doug Christie, John Grant, Armando Arcidiaco, Tony Percival, Jan Hulse and Steve Sirotjuk for their efforts in the field in often trying and uncomfortable conditions. Roger Bowman, Phil Cummins, Subash Dey, Oli Gudmundsson, Cheng Tong, Rob van der Hilst, Alet Zielhuis, Eric Debayle, Sri Widiyantoro, Alexei Gorbatov, Haixu Cheng, Yoshihisa Hiyoshi, Katrina Marson-Pidgeon and Jennifer Eccles have all been involved with different aspects of the structural analysis. This work is dedicated to the memory of Gus Angus whose fieldwork skills made the short-period work possible.

REFERENCES BOWMAN J. R. & KENNETT B. L. N. 1990. An investigation of the upper mantle beneath northwestern Australia using a hybrid seismic array. Geophysical Journal International 101, 411–424. BOWMAN J. R. & KENNETT B. L. N. 1991. Propagation of Lg waves in the North Australian craton: influence of crustal velocity gradients. Bulletin of the Seismological Society of America 81, 592–610. BOWMAN J. R. & KENNETT B. L. N. 1993. The velocity structure of the Australian shield from seismic travel times. Bulletin of the Seismological Society of America 83, 25–37. CARA M. & LÉVÊQUE J. J. 1987. Waveform inversion using secondary observables. Geophysical Research Letters 14, 1046–1049. CHENG H. X. & KENNETT B. L. N. 2001. Attenuation of seismic body waves in the Australian region. Annual Report of Research School of Earth Sciences, ANU, for 2000, 133–135. DEBAYLE E. 1999. SV-wave azimuthal anisotropy in the Australian upper mantle: preliminary results from automated Rayleigh waveform inversion. Geophysical Journal International 137, 747–754. DEBAYLE E. & KENNETT B. L. N. 2000a. The Australian upper mantle: structure & deformation inferred from surface waves. Journal of Geophysical Research 105, 25243–25450. DEBAYLE E. & KENNETT B. L. N. 2000b. Anisotropy in the Australian upper mantle from Love and Rayleigh waveform inversion. Earth

and Planetary Science Letters 184, 339–351. DEBAYLE E. & KENNETT B. L. N. 2002. Surface wave studies of the Australian region. Geological Society of Australia Special Publicatin 22 and Geological Society of America Special Paper 372, 25–40. DENHAM D. 1991. Shear wave crustal models for the Australian continent. In: Drummond B. J. ed. The Australian Lithosphere, pp. 59–66. Geological Society of Australia Special Publication 17. DEY S. C., KENNETT B. L. N., BOWMAN J. R. & GOODY A. 1993. Variations in upper mantle structure under northern Australia. Geophysical Journal International 114, 304–310. DRUMMOND B. J., MUIRHEAD K. J., WELLMAN P. & WRIGHT C. 1991. A teleseismic travel-time residual map of the Australian continent. BMR Journal of Australian Geology & Geophysics 11, 101–105. DZIEWONSKI A. M., EKSTRÖM G. & MATERNOVSKAYA N. N. 1999. Centroid-moment tensors for July–September, 1998. Physics of the Earth and Planetary Interiors 114, 99–107. ENGDAHL E. R., VAN DER HILST R. D. & BULAND R. 1998. Global teleseismic earthquake relocation with improved travel times and procedures for depth determination. Bulletin of the Seismological Society of America 88, 722–743. GORBATOV A. & KENNETT B. L. N. 2001.P wave tomographic imaging of the southwest Pacific. Annual Report of Research School of Earth Sciences, ANU, for 2000, 140–141. GOES S., GOVERS R. & VACHER P. 2000. Shallow mantle temperatures under Europe from P and S wave tomography. Journal of Geophysical Research 105, 11153–11169. GONCZ J. H. & CLEARY J. R. 1976. Variations in the structure of the upper mantle beneath Australia from Rayleigh wave observations. Geophysical Journal of the Royal Astronomical Society 63, 659–670. GRAND S. & HELMBERGER D. V. 1984. Upper mantle shear structure of North America. Geophysical Journal of the Royal Astronomical Society 76, 399–438. GUDMUNDSSON O., KENNETT B. L. N. & GOODY A. 1994. Broadband observations of upper mantle seismic phases in northern Australia and the attenuation structure in the upper mantle. Physics of the Earth and Planetary Interiors 84, 207–226. HALES A. L., MUIRHEAD K. J. & RYNN J. W. 1980. A compressional velocity distribution for the upper mantle. Tectonophysics 63, 309–348. INOUE H., FUKAO Y., TANABE K. & OGATA Y. 1990. Whole mantle Pwave mantle tomography. Physics of the Earth and Planetary Interiors 59, 294–328. JACKSON I. 2000. Laboratory measurement of seismic wave dispersion and attenuation: recent progress. In: Karato S. I., Forte A. M., Liebermann R. C., Masters G. & Stixrude L. eds. Earth’s Deep Interior: Mineral Physics and Tomography from the Atomic to the Global Scale, pp. 265–289. American Geophysical Union Geophysical Monograph Series 117. KAIHO Y. & KENNETT B. L. N. 2000. Three dimensional structure beneath the Australasian region from refracted wave observations. Geophysical Journal International 142, 651–668. KENNETT B. L. N. 1995. Approximations for surface wave propagation in laterally varying media. Geophysical Journal International 122, 470–478. KENNETT B. L. N. & BOWMAN J. R. 1990. The velocity structure and heterogeneity of the upper mantle. Physics of the Earth and Planetary Interiors 59, 134–144. KENNETT B. L. N., ENGDAHL E. R. & BULAND R. 1995. Constraints on the velocity structure in the Earth from travel times. Geophysical Journal International 122, 108–124. KENNETT B. L. N., GUDMUNDSSON O. & TONG C. 1994. The upper-mantle S and P velocity structure beneath northern Australia from broadband observations. Physics of the Earth and Planetary Interiors 86, 85–98. LEFEVRE L. V. & HELMBERGER D. V. 1989. Upper mantle P velocity structure of the Canadian shield. Journal of Geophysical Research 94, 17749–17765. LEVEN J. H. 1985. The application of synthetic seismograms in the interpretation of the upper mantle P-wave velocity structure in northern Australia. Physics of the Earth and Planetary Interiors 38, 9–27. LÉVÊQUE J., DEBAYLE E. & MAUPIN V. 1998. Anisotropy in the Indian Ocean upper mantle from Rayleigh- and Love-waveform inversion. Geophysical Journal International 133, 529–540.

Mantle beneath Australia MARQUERING H., SNIEDER R. & NOLET G. 1996. Waveform inversions and the significance of surface-wave mode coupling. Geophysical Journal International 124, 258–278. MONTAGNER J. P. 1986. Regional three-dimensional structures using long-period surface waves. Annales de Geophysique 4, 283–294. MUIRHEAD K. J. & DRUMMOND B. J. 1991. The seismic structure of the lithosphere under Australia and its implications for continental plate tectonics. In: Drummond B. J. ed. The Australian Lithosphere, pp. 23–40. Geological Society of Australia Special Publication 17. NOLET G. 1990. Partitioned waveform inversion and two-dimensional structure under the Network of Autonomously Recording Seismographs. Journal of Geophysical Research 95, 8499–8512. SAMBRIDGE M. S. 1999. Geophysical inversion with a neighbourhood algorithm—I. Searching a parameter space. Geophysical Journal International 138, 479–494. SIMONS F. J., ZIELHUIS A. & VAN DER HILST R. D. 1999. The deep structure of the Australian continent from surface wave tomography. Lithos 48, 17–43. TONG C., GUDMUNDSSON O. & KENNETT B. L. N. 1994. Shear wave splitting in refracted waves returned from the upper mantle transition zone beneath northern Australia. Journal of Geophysical

23

Research 99, 15783–15797. HILST R. D., KENNETT B. L. N. & SHIBUTANI T. 1998. Upper mantle structure beneath Australia from portable array deployments. In: Braun J., Dooley J., Goleby B., van der Hilst R. & Klootwijk C. eds. Structure and Evolution of the Australian Continent, pp. 39–58. American Geophysical Union Geodynamics Monographs 26. VAN DER HILST R., KENNETT B., CHRISTIE D. & GRANT J. 1994. Project Skippy explores the mantle and lithosphere beneath Australia. EOS 75, 177–181. WIDIYANTORO S. & VAN DER HILST R. 1996. Structure and evolution of lithospheric slab beneath the Sunda arc, Indonesia. Science 271, 1566–1570. ZIELHUIS A. & NOLET G. 1994. Shear wave velocity variations in the upper mantle below Central Europe. Geophysical Journal International 117, 695–715. ZIELHUIS A. & VAN DER HILST R. 1996. Upper mantle shear velocity beneath eastern Australia from inversion of waveforms from Skippy portable arrays. Geophysical Journal International 127, 1–16. VAN DER

Received 27 June 2001; accepted 18 December 2001

Geol. Soc. Australia Spec. Publ. 22, and Geol. Soc. America Spec. Pap. 372 (2003), 25–40

Surface-wave studies of the Australian region E. DEBAYLE1* AND B. L. N. KENNETT2 1

IPGS — Ecole et Observatoire des Sciences de la Terre, CNRS and Université Louis Pasteur, 5 rue René Descartes, 67084 Strasbourg, France. 2 Research School of Earth Sciences, Australian National University, ACT 0200, Australia. The deployment of portable stations over the Australian continent since the beginning of the 1990s has allowed the collection of a unique dataset for surface waves at regional distances. Surface wave tomographic inversions now exploit the excellent azimuthal ray coverage available for central and eastern Australia, so that S wave tomographic models for the upper mantle can be built with a lateral resolution of few hundred kilometres and a vertical resolution of a few tens of kilometres. Our tomographic models include an anisotropic component in addition to the distribution of Swave heterogeneities. When only Rayleigh waves are considered in the inversion, this anisotropic component is represented at each depth in the mantle by the direction of fast horizontally propagating SV waves. When Love and Rayleigh waves are inverted simultaneously, the anisotropic component reflects the difference in wave speed between S waves polarised in horizontal and vertical directions. Our results for the simultaneous inversion of Love and Rayleigh waves agree with previous studies to locate the anisotropy in the uppermost 200–250 km of the mantle. However, there are significant differences between the different models deduced from the Rayleigh wave inversions. We show that the most likely cause of these differences is the frequency band used in the analysis of the seismograms. The application of path averages directly to shear-wave slowness is a reasonable assumption for the recovered models. The shear-wave speed models are found to be robust, especially when removing paths likely to have experienced a complex propagation. In the light of these new inquiries, we attempt to extract the well-resolved part of the surface-wave inversion and to see how far it can be reconciled with other results obtained from body-waves studies. It appears that, due to the horizontal smoothing imposed by surface waves and the difficulty of estimating the best choice for the frequency band used in analysis, the details of the anisotropic directions in the upper layer should be interpreted with caution. However, the existence of at least two layers of anisotropy is well constrained. In the upper layer, the complex anisotropy would reflect ‘frozen’ deformation in the lithosphere while in the lower layer the smoother pattern is more likely to reflect present-day deformation due to the northward motion of the Australian Plate. The observation that anisotropy is not vertically coherent with depth is more easy to reconcile with other anisotropic measurements inferred from body waves. KEY WORDS: anisotropy, Australia, surface waves, tomography, upper mantle.

INTRODUCTION The subduction zones to the north and east of Australia provide frequent events over a broad range of depths which can be exploited in surface-wave tomography of the Australian region. The mid-ocean ridge to the south of the continent provides a less frequent, but very useful, set of shallow events. There are only a limited number of permanent seismic stations with high-fidelity broadband recording in the region and these have been supplemented by extensive deployments of portable broadband instruments. The SKIPPY experiment (van der Hilst et al. 1994) occupied 60 sites across the continent in a series of deployments from 1993 to 1996 and subsequent experiments have improved coverage in southeastern Australia and different parts of Western Australia (Kennett 2003). Fundamental and higher mode Rayleigh waves recorded on vertical component seismograms have been exploited in a number of tomographic studies of the shear-wave speed distribution beneath the Australian region. Zielhuis and van der Hilst (1996) analysed the data from the SKIPPY stations in eastern Australia using the Partitioned

Waveform Inversion technique of Nolet (1990). They produced the first clear evidence of a substantial difference in the structure in the mantle beneath the Precambrian regions of the centre and west of the continent and beneath the Phanerozoic belt in the east. The path coverage was extended to nearly the whole continent by van der Hilst et al. (1998). This work revealed a consistent pattern of a mantle lithosphere extending to around 250 km depth beneath the shield regions, contrasted with a thinner lithosphere (about 100 km thick) below the eastern part of Australia, underlain by a significant low velocity zone for S waves. Subsequently the number of available data have been augmented in various ways. Simons et al. (1999) have included data from the Australian Geological Survey Organisation stations and undertaken a study of regionalisation based on the geologic provinces in the continent. Debayle and Kennett (2000a) have exploited almost 2200 paths using an automated waveform analysis procedure (Debayle 1999) based on the work of Cara and Lévêque * Corresponding author: [email protected]

26

E. Debayle and B. L. N. Kennett

Figure 1 Unfiltered three-component seismograms at station SC08 in the Northern Territory for a shallow event near Tonga at 34°S, showing well-developed surface waves. The fundamental Rayleigh and Love modes constitute the prominent part of the surface waves for this shallow earthquake. Higher modes are less energetic and appear at the head of the surface wavetrain.

(1987), with a continuous regionalisation procedure for extracting a 3-D shear-wave speed model (Montagner 1986). In addition to the shear-wave speed distribution Debayle and Kennett (2000a) have examined the azimuthal anisotropy of the Rayleigh waves as a function of position and depth. Beneath the centre of the Australian continent there is a clear shift in the direction of fast polarisation from nearly east–west at 100 km to close to north–south at 200 km depth. The orientation at greater depth is roughly aligned with the direction of absolute plate motion. Three-component recording of the seismic wavefield allows also the exploitation of horizontal component records. Debayle and Kennett (2000b) have undertaken simultaneous inversion of the Rayleigh and Love wave records from 792 event-station pairs. The two types of surface waves have different polarisation and sampling of the structure. By working with both it is possible to constrain the relation of the velocities of vertically and horizontally polarised S waves. The ‘polarisation anisotropy’ revealed by this work lies dominantly in the uppermost 200–250 km of the mantle beneath the central and western Precambrian cratons where the Rayleigh wave studies indicate high shear-wave speed, but significant ‘polarisation anisotropy’ is also found below eastern Australia, on the edge of the craton. These recent studies indicate the important interaction between anisotropy and heterogeneity, which can have considerable significance for the geodynamic interpretation of the heterogeneity. Because there are some significant differences in the various models proposed for the Australian region, we present here an investigation of the sensitivity of the 3-D models of shear-wave speed, derived from Rayleigh waves, to the frequency content employed in the analysis. For this purpose, we have built a new SV wave tomographic model, by reprocessing all our Rayleigh waveforms in a different frequency band. We also check the robustness of our model, by removing a subset of the data related to paths with complex propagation, and the applicability of the path average hypothesis directly to the shear slowness. Finally, in the light of these new constraints, we provide a synthe-

sis of the various constraints on anisotropy from the surface-wave studies and independent work on SKS splitting (Clitheroe & van der Hilst 1998; Girardin & Farra 1998; Ozalabey & Chen 1999).

SURFACE-WAVE TOMOGRAPHY For earthquakes at depths less than 100 km the surface waves normally form the most prominent part of a seismogram, as illustrated in Figure 1 for a shallow event in the Tonga region recorded in the Northern Territory. The horizontal component seismograms have been rotated to produce traces along the great circle to the source (R) and transverse to the path (T). The Rayleigh waves with polarisation in a vertical plane appear on the vertical (Z) and radial (R) components, while the horizontally polarised Love waves arrive on the tangential (T) component. For a shallow event, as in Figure 1, the records are dominated by the fundamental mode surface waves, but for deeper sources there is also significant excitation of higher mode surface waves which have greater penetration into the Earth. The apparent frequency of the surface waves changes along the wavetrain, this dispersion arises because the waves have an interaction with the structure which depends on frequency. At high frequencies the surface waves are confined close to the surface, but at lower frequencies they penetrate to greater depth and are influenced by the higher seismic-wave speeds at depth. The effect of a change in shear-wave speed structure at a given depth on the fundamental and first four higher Rayleigh modes displacement field associated with Rayleigh waves is illustrated in Figure 2 for different periods. The fundamental mode decays fairly rapidly with depth but the higher modes show significant sensitivity to shear-wave structure in the transition zone. By working with the fundamental and higher modes we have a probe into the Earth in which the character of the wavetrains is dictated by the structures encountered in the outer layers of the Earth and where the dependence on frequency provides information over a range of depths. In a

Surface-wave studies, Australian region

27

Figure 2 Sensitivity kernels for the fundamental and first four higher modes of Rayleigh waves at periods of 50 s (dash-dot line) and 160 s for the fundamental mode, and 50 s (dash-dot line) and 80 s for the higher modes (continuous line). The curves represent the relative partial derivatives of the phase velocity according to the shear-wave velocity [(Vsv /C).(∂C/∂Vsv)]. These partial derivatives are shown for a 1 km thick layer.

smoothly varying medium the phase contribution for an individual mode represents an average of the phase slownesses encountered along the propagation path. This is the property which is exploited in surface-wave tomography. When observations are available from a large number of paths crossing a region, the path-specific models can be combined to produce a 3-D image of shear-wave structure. The most effective means of extracting multimode information is to match the waveforms of observed and calculated seismograms based on a model of the source process. We employ the method of Cara and Lévêque (1987) which is able to achieve good results for the recovery of shear-wave speed profiles for significant deviations from a reference model (up to ±10% perturbations). The broad range of application is achieved by working with secondary variables based on the correlation of full seismograms with individual modal contributions. These secondary variables have a close to linear dependence on the perturbations in shear-wave speed from the reference model. In order to achieve an effective application of the waveform inversion procedure, we have to ensure that we treat the portion of the seismogram that conforms to the underlying assumptions of independent mode propagation and

that the receivers lie in the far-field of the source. Kennett (1995) has examined the nature of these approximations in laterally varying media. For paths longer than 2000 km, typical of the ranges used in this study of the Australian region, the fundamental mode should be useable in the period range from 160 to 40 s and the higher modes from 100 to 25 s period. At higher frequencies the modal contributions become susceptible to the influence of strong heterogeneity at shallow depth, and the propagation paths can deviate significantly from the great circle. When there are sharp variations in seismic properties in the mantle itself the assumption of smoothness, which is employed in the representation of the seismograms, may be violated and we then need to be cautious in the interpretation of the results. The shear-wave speed models derived from the inversion of the waveforms of the surface wavetrains can be regarded as a summary of the multimode dispersion characteristics for the individual paths. When the deviations of the true 3D model from the reference model are not too large, it is a reasonable approximation to transfer the path-average property of phase to the shear slowness. Under this assumption, the set of 1-D path-specific models derived from the waveform inversion can be regarded as linear constraints on the

28

E. Debayle and B. L. N. Kennett

Figure 3 Phase velocity maps for Rayleigh waves at 60 and 100 s periods. (a, c) Obtained by applying the phase-velocity distribution directly to the phase slowness. (b, d) Predicted from the 3-D SV velocity model of Debayle and Kennett (2000a).

3-D shear-wave speed variations and a linear inverse problem can be posed for the recovery of the 3-D model. This is the approach employed by Debayle and Kennett (2000a, b). However, where there are large contrasts in seismic-wave speeds, the path-average interpretation of the shear-wave speed models may not be adequate (Kennett & Yoshizawa 2002). In this case it is necessary to work through the intermediary of multimode phase slowness maps as a function of frequency, employing the path-average property of phase directly. The 3-D shear-wave speed can then be recovered from local inversions of the dispersion properties. In the following section, we test the validity of the path-average approximation directly to shear slowness, for the level of heterogeneity expected in Australia.

A TEST OF THE PATH-AVERAGE APPROXIMATION The development of a 3-D wave-speed model in a two-stage process requires that the contrasts in seismic-wave speeds

are not too large. This restriction is required to make the identification of the path-specific 1-D models as the average of the shear slowness along the path. Kennett and Yoshizawa (2002) have pointed out that the 1-D model derived for a path is best regarded as a summary of multimode dispersion and that this interpretation can be employed even when the shear-slowness profile cannot be regarded as a path average. Since the contrasts in our models are quite large we need to examine the influence of the assumptions we have made in the shear-wave speed inversion. We can do this by comparing the predicted phase-velocity distribution derived from the 3-D model and a phase-velocity map produced by applying the path-average approximation directly to phase slowness, as in the original development of Woodhouse (1974). We have used the results of Debayle and Kennett (2000a) using waves up to 40 s period and combine the phase-velocity information at 60 s and 100 s for the 2194 paths. The inversions for the 3-D wave-speed model and for the geographic distribution of phase speed have been car-

Surface-wave studies, Australian region

29

Figure 4 (a, c) SV wave heterogeneities and azimuthal anisotropy for an inversion with waveforms analysed for the frequency range 40–160 s. (b, d) Inversion for the frequency band 50–160 s. Tones represent velocity contrasts and bars display azimuthal anisotropy, the length of the bars indicating the strength of anisotropy.

ried out with the same continuous regionalisation technique, with the same assumed Gaussian spread about each path (with width 200 km) representing the model covariance. The agreement between the direct phase-velocity maps and those reconstructed from the 3-D shear-wave speed is good (Figure 3) with a very similar configuration of heterogeneity, although there are some slightly larger contrasts in the direct phase-speed maps. We note that for the upper layers of the model, some offsets in reference velocities are needed to bring the images into the closest correspondence. We believe that this is a symptom of a slight bias towards higher wave speeds in the 3-D model arising from the limitations of the two-step inversion. The comparison of the direct calculations of the phasevelocity maps and those reconstructed from the model is very encouraging. We note though that the use of the phase

average along each path rests on the assumption of smooth variations in wave speed on a scale long compared with the wavelength. The level of variation seen in the phase velocity maps in Figure 3a is such that we are at the limit of this assumption. For the fundamental mode at 100 s period, the wavelength in the mantle is around 420 km and is approaching the scale of the medium wavelength heterogeneity. To extract any more information from the seismograms we will need to adopt a description of the propagation process with a full allowance for 3-D effects, including mode-coupling and refraction. The theoretical treatment of Kennett (1998) indicates how this can be accomplished, but is too computationally intensive for use with any current inversion schemes. Suitable restricted approximations need to be found to allow effective waveform modelling for the propagation of surface waves in 3-D heterogeneous models.

30

E. Debayle and B. L. N. Kennett

Figure 5 Location of events and stations used in the Rayleigh wave studies presented in this paper. The large diamonds indicate permanent stations from the IRIS and GEOSCOPE networks, whilst the smaller diamonds are the locations of portable broadband stations. Events are represented by open circles. The main geological outcrop zones are indicated in the background.

SV WAVE AZIMUTHAL ANISOTROPY FROM RAYLEIGH WAVE INVERSION For Australia, even with the substantial level of heterogeneity present in the region, the application of path-averaging directly to shear-wave slowness is not likely to be a significant limitation if we restrict attention to periods between 160 and 40 s. We can therefore restrict our tomographic approach to a two-step process. In the first step, a waveform analysis provides path-average constraints on the shear-wave speed structure. In the second step, once path-average constraints have been obtained for a large number of paths with different azimuths, it is possible to extract from the tomographic inversion the SV wave anisotropic directions in addition to the shear-wave speed heterogeneities. The tomographic inversion is based on the work by Smith and Dahlen (1973) who demonstrated the azimuthal dependence of Rayleigh wave phase and group velocities in a slightly anisotropic medium. Providing the anisotropy is weak, such ordinary perturbation methods are justified for the period range used in the analysis, where the intrinsic coupling between Love and Rayleigh waves is expected to be weak (Maupin 1989). Montagner and Nataf (1986) then showed that the phase and group velocity azimuthal terms can be inverted for some combinations of

the elastic coefficients at depth. For waveform fitting, which provides a direct estimate of 1-D path average SV wave models, the procedure needed to retrieve the distribution of heterogeneities and anisotropy at depth is described in detail by Lévêque et al. (1998). In particular, these authors showed that in a long period approximation in a full but weakly anisotropic medium, the path-average SV velocity models depend on the combination of elastic parameters best resolved by Rayleigh waves, which control the velocity of SV waves propagating horizontally at azimuth . The anisotropic directions presented in Figures 4, 7, 9 and 10 can be interpreted as the directions of fast SV wave propagation. Assuming that the main cause of anisotropy in upper mantle materials is the preferred orientation of olivine crystals, these anisotropic directions would point toward the projection of the fast a-axis of the crystals on the horizontal plane. To investigate possible effects due to the frequency filtering of the seismograms, we present in Figure 4 the results of two inversions where waveform analysis has been performed in different frequency bands. Compared to the previous work by Debayle and Kennett (2000a), we use a more restrictive geographical distribution of events. We use only earthquakes that originate within the Indo-Australian Plate or along its major boundaries. We exclude events from the

Surface-wave studies, Australian region

Figure 6 Path coverage for Rayleigh waves. (a) Data for which the fundamental modes have been taken into account in the inversion. (b) Data for which the higher modes have been taken into account.

trenches bounding the Philippine Sea that were previously used by Debayle and Kennett (2000a). This allows us to estimate the possible bias that may have been introduced previously by paths related to events with more complex source mechanisms or crossing complex structures such as the northern subduction zones. Two maps (Figure 4a, c) have been obtained from an inversion in the period range 40–160 s, similar to Debayle and Kennett (2000a) using the same process of automated analysis of the individual paths, but with the removal of events in the vicinity of the Philippine Sea. Of the 2194 paths analysed by these authors, 1820 were retained. The distribution of events is shown in Figure 5 while the corresponding ray coverage in fundamental and higher modes is presented in Figure 6. The tomographic images obtained are very close to those presented by Debayle and Kennett (2000a) and demonstrate that no significant bias is introduced by paths coming from the Philippine Sea region. Two maps (Figure 4b, d) have been obtained by reprocessing all the waveforms in the period range 50–160 s

31

instead of 40–160 s. A total of 2482 waveforms were successfully matched with the automated process of Debayle (1999) and again whether or not we included paths related to events in the vicinity of the Philippine Sea did not affect the results. Significant differences are observed in the uppermost layers, especially near 100 km depth, between the two inversion at 40 s and 50 s. These differences are more pronounced for short-wavelength heterogeneities, the long-wavelength component of the S wave structure being better preserved. For layers above 150 km depth, the choice of the cut-off period has a direct effect on the wave-speed model from the inversion because the S wave structure is mostly constrained by the fundamental Rayleigh mode between 40 and 100 s period (Figure 2). At greater depths, where 40 and 50 s period fundamental Rayleigh waves have low sensitivity, the differences between the wave-speed distributions for the two inversions diminish, and there is a good agreement between the two maps at 200 km depth (Figure 4c, d). To explain the differences observed near 100 km depth (Figure 4a, b), several effects can be invoked. First, the horizontal resolution of our dataset increases when shorter periods are included in the inversion and it is likely that part of the observed difference between the 40 s and 50 s inversion reflects an actual difference in resolution. However, effects due to the crustal structure or produced by possible off-great-circle propagation are likely to be more important at shorter periods. Therefore, they may contribute to part of the difference observed between the 40 s and the 50 s inversion. Crustal effects probably do not contribute significantly to the observed differences. Both inversions are performed using the same 3SMAC crustal structure (Nataf & Ricard 1995), which has an average crustal thickness of about 40 km for Australia. The maximum sensitivity of the Rayleigh waves at 40 and 50 s period is located below the crust at around 50 and 75 km depth, which indicates that in both cases we use a dataset primarily sensitive to the upper mantle. For the fundamental mode near 40 s period, which provides strong constraints on the shallow structure of the model, the great-circle approximation should still be just valid for epicentre–station distances typical of this study (Kennett 1995). However, ray tracing in our tomographic model (Yoshizawa & Kennett 2002) shows that the strong gradient in shear velocity near the edge of the craton can produce significant deviations from the great circle at 40 s period for the fundamental and even the first higher mode. Debayle and Kennett (2000a) noticed that the strongest effects are expected to be associated with north–south paths grazing the continent edge, but this combination of source–receiver pairs was not actually present in the dataset used in their inversion. However, the differences observed at 100 km depth near the edge of the craton in Figure 4 suggest that off-great-circle effects at 40 s periods may contribute to bias in the 40 s inversion in this part of the model. We also note that even when great-circle deviations remain small at 40 s period, they increase rapidly when the period decreases and by 20 s period become important for the first few higher modes. It is thus rather difficult to estimate which part of the differences in the 3-D shear-wave speed models presented in

32

E. Debayle and B. L. N. Kennett

Figure 4 should be attributed to a real difference in resolution or to the limitation in our approximations to the nature of the surface-wave propagation processes. We believe that the differences obtained with the inversions with upper limits of 40 and 50 s period reflect the level of confidence we should give to this class of tomographic inversion. Even if

variations in seismic heterogeneities and anisotropy with wavelength smaller than 500 km should be interpreted with caution, the major changes are well constrained and provide useful indications of the deep structure of the Australian Plate. Indeed, whether or not the inversion is pushed to include analysis at 40 s period, the main features of our tomographic inversions persist. As revealed in the previous inversion by Debayle and Kennett (2000a), a clear change in the organisation of seismic anisotropy occurs between the uppermost 100 km of the mantle and the deeper structure. For the uppermost layer, down to 150 km depth, Figure 4 suggests that we should be cautious when interpreting the details of the complex anisotropic pattern. This is supported by a synthetic experiment where we tried to retrieve a complex pattern of anisotropy with 90° changes in anisotropic directions superimposed to the 3SMAC model (Figure 9): although the general anisotropic pattern is well retrieved east of 125°E, smoothing effects can locally lead to wrong anisotropic estimations in the vicinity of abrupt changes in anisotropic directions. For the lowermost layer, below 150 km depth, a smoother pattern of anisotropy is retrieved for both the inversions (with either a 40 s or 50 s period upper bound). This confirms that present-day shearing occurs at the bottom of the Australian mechanical lithosphere, but still within the zone of elevated shear-wave speeds. Finally, comparing our results with those previously obtained by van der Hilst et al. (1998) and Simons et al. (1999), we find a better agreement in the uppermost 200 km, between the results of our 50 s inversion and their published models. However, at depths greater than 200 km (Figures 7, 8), where the structure is mostly constrained by the higher modes, significant differences persist. The main differences from the models presented by van der Hilst et al. (1998) and more recently by Simons et al. (1999) may therefore be related to the treatment of higher mode information. In their use of the partitioned waveform inversion method (Nolet 1990; Zielhuis & Nolet 1994), van der Hilst et al. (1998) and Simons et al. (1999) push the higher mode analysis to periods up to 20 s. At 20 s period, off-great-circle effects become important and the size of the wavespeed variations suggest that mode coupling may need to be taken into account [cf. Marquering et al. (1996) in a study of structure in Europe].

LOVE/RAYLEIGH SIMULTANEOUS INVERSION

Figure 7 Slices at 150, 250 and 300 km depth in the SV velocity model from the Rayleigh wave inversion in the period range 50–160 s. As in Figure 4, tones represent velocity contrasts and bars display azimuthal anisotropy.

Rayleigh waves can be used to extract anisotropic directions efficiently because their azimuthal variation depends mostly on terms depending on cos 2, sin 2, where  is the azimuth of the path. These dependencies can be recovered from a relatively modest number of crossing paths. In a cell defined by the horizontal resolution of the tomographic model, a minimum of three paths with different azimuths is necessary to recover the 2 variation. However, with Rayleigh waves alone, it is often difficult to demonstrate the presence of anisotropy without ambiguity. This is because the azimuthal variation of Rayleigh waves generally contributes to explain only a small part of the misfit between

Surface-wave studies, Australian region

33

Figure 8 Vertical cross sections at different latitudes in the SV velocity model after the Rayleigh wave inversion in the period range 50–160 s.

34

E. Debayle and B. L. N. Kennett

Figure 9 Synthetic experiment. (a) The initial model is the isotropic shear velocity distribution at 50 km depth provided by the 3SMAC model of Nataf & Ricard (1995) with a pattern of azimuthal anisotropy superimposed. (b) The result of the synthetic inversion is shown at 100 km depth.

observed and predicted seismograms. Many isotropic shear-velocity models have been found to give a satisfactory fit to Rayleigh waves, which suggests that in most situations it would be acceptable to suppress the azimuthal anisotropy. A simultaneous inversion of the horizontal and vertical components of the surface wavefield is able to provide additional constraints on the presence, relative amplitude and depth location of seismic anisotropy. It is often difficult to explain the dispersion of both Love and Rayleigh waves with a single isotropic model (McEvilly 1964). Generally, the dispersion of the two classes of waves can be reconciled by adding a single anisotropic parameter which allows SH and SV waves to propagate at different velocities. The need to introduce this ‘polarisation’ or ‘radial’ anisotropy to reconcile the Love and Rayleigh wave components is considered as strong support for the presence of anisotropy in the upper mantle. For the upper mantle beneath Australia, surface-wave polarisation anisotropy studies have been conducted by Gaherty and Jordan (1995) and Debayle and Kennett (2000b). Gaherty and Jordan (1995) used body and surface waves to build a depth-dependent 1D elastic model along a single corridor crossing Precambrian Australia. Their inversion revealed the presence of significant polarisation anisotropy down to 250 km in the upper mantle. Debayle and Kennett (2000b) chose a different strategy. They applied the automated procedure developed by Debayle (1999) for the simultaneous analysis of both Love and Rayleigh waves and analysed a larger dataset related to paths crisscrossing the Australian continent. A total of 792 pairs of Rayleigh and Love wave seismograms was used to retrieve the distribution of SV and SH velocity in the upper mantle. Each Love and Rayleigh pair is related to a single

epicentre–station path and inverted for a radially anisotropic model compatible with the waveforms of both Love and Rayleigh waves. 1584 waveforms are used in the inversion. The number of paths is reduced compared to the Rayleigh-only inversion because the Love wave records are generally noisier. We present in Figure 10 three slices through the SH and SV velocity distribution found by Debayle and Kennett (2000b). For SV waves, the cos 2, sin 2 azimuthal variation could be recovered by constraining the horizontal smoothness of the model using Gaussian functions with correlation lengths of 500 km. This large horizontal degree of smoothing provides a smoother model compared to the Rayleigh-only inversion but confirms the presence of two anisotropic layers, the lower layer presenting a nearly north–south component in good agreement with the present-day motion of the Australian Plate. The upper layer is a smooth image of the Rayleigh-only inversion at 100 km depth, where east–west anisotropic directions dominate in central Australia. For SH waves, the azimuthal variation is faster and dominated by terms in cos 4, sin 4 which could not be recovered with the available Love/Rayleigh coverage, and Debayle and Kennett (2000b) preferred to average out the SH azimuthal variation by using the same horizontal degree of smoothing in the inversion as for the SV waves, so that the maps could be compared. Figure 10 shows that the SV and SH heterogeneity patterns are relatively close if an increase of about 4% is used for the SH reference velocity in the uppermost 200 km of the mantle. On average, the amplitude of the SH/SV difference would be about 4–5% but locally there are larger differences. This can be seen in Figure 11 where the parameter  which represents the radial anisotropy

Surface-wave studies, Australian region

35

Figure 10 Left, SV wave heterogeneities and azimuthal anisotropy; right, SH wave heterogeneities. Tones represent velocity contrasts and bars display azimuthal anisotropy, the length of the bars indicating the strength of anisotropy. Note that the reference velocity used for SH is about 4% higher in the uppermost 200 km only.

36

E. Debayle and B. L. N. Kennett

Figure 11 Polarisation anisotropy model at 125 km depth with a vertical cross-section at 25°S. SH waves have been found to propagate significantly faster than SV waves in the uppermost 200–250 km of the mantle for most of Australia.

[=(Vsh/Vsv)2] is plotted. Values of  up to 1.19 (9% of polarisation anisotropy) have been obtained at 150 km for eastern Australia. This locally large anisotropy is difficult to reconcile with current mineralogical models and Debayle and Kennett (2000b) invoked the presence of strong lateral heterogeneities along the path and/or the effects introduced by the simplifying assumption of transverse isotropy in the

starting Preliminary Reference Earth Model (PREM: Dziewonski & Anderson 1981) to explain the discrepancy. The amplitude of polarisation anisotropy, therefore, may not be determined accurately, but there is evidence (Lévêque & Cara 1985; Gaherty & Jordan 1995) which suggests that the location of polarisation anisotropy is correctly retrieved at depths within the current assumptions.

Surface-wave studies, Australian region

Figure 11 shows that, for Australia, significant polarisation anisotropy with SH faster than SV is present down to 200–250 km, in good agreement with the 1-D model of Gaherty and Jordan (1995). At larger depths, Debayle and Kennett (2000b) concluded that the SH/SV differences fall below the measurement errors and so are not required by the data. This layer where SH is faster than SV would correspond to regions where horizontal flow dominates in the upper mantle and where significant azimuthal anisotropy is expected (see Montagner 1994). The results presented here suggest that this layer would be about 250 km thick beneath Australia and somewhat thinner (about 100–150 km thick) in the adjacent oceanic areas (Figures 4, 11).

WHAT HAVE WE LEARNED ABOUT SEISMIC ANISOTROPY BENEATH AUSTRALIA? The different studies that have addressed the question of upper mantle anisotropy beneath Australia have employed different techniques involving both body and surface waves. These techniques have yielded different answers, suggesting that the extraction of reliable information on the deep anisotropic structure of Australia requires a consideration of the limitations of each technique. It is clear that the strongest limitation of classical SKS/SKKS analysis is lack of vertical resolution provided by the core shear phase; there is nothing in the surface observations to isolate the location of the anisotropy. Most of the previous splitting measurements of teleseismic shear waves beneath Australia have been based on the hypothesis that the anisotropic region can be approximated by a single homogeneous layer. Core shear phase-splitting measurements (mostly from SKS and SKKS) have been recently obtained by Clitheroe and van der Hilst (1998), who synthesised measurements from permanent and portable stations, and Ozalabey and Chen (1999), who revisited the data from the permanent stations across Australia. Both studies reached the conclusion of a weak or null splitting beneath Precambrian Australia. For the eastern Phanerozoic margin of the continent, significant splitting compatible with an upper mantle contribution to seismic anisotropy, has been observed by Clitheroe and van der Hilst (1998). The alignment of fast axis directions could either be explained by ‘fossil’ anisotropy frozen in the lithosphere (Silver & Chan 1991) or by present-day anisotropic flow (Vinnik et al. 1992). Recently, Saltzer et al. (2000) have re-examined the validity of the assumption of a single homogeneous anisotropic layer when a vertically varying medium affects the shear-wave splitting. They observed a tendency of the measurements to mimic the anisotropy at the top part of the medium. In addition, for stronger heterogeneities, multiple scattering may produce a null result even for waves travelling in a strongly anisotropic medium. This may explain why no significant splitting has been observed beneath Precambrian Australia. Another possible explanation for central Australia is the presence of two anisotropic layers with orthogonal directions. If each layer produces a similar delay time between the fast and slow polarised wave, a resulting null splitting is expected for a vertically travelling wave which successively crosses the two layers.

37

More sophisticated analysis is thus needed to interpret shear-wave splitting results beneath Australia. Girardin and Farra (1999) used vertically travelling body waves converted at the different discontinuities to constrain the depth location of anisotropy beneath southeast Australia. They found a two-layered anisotropic model in the upper mantle beneath Canberra which would be in qualitative agreement with the surface-wave results presented here. For northern Australia Tong et al. (1994) used refracted shear waves returned from the transition zone and found a few percent of polarisation anisotropy in the upper mantle. They attributed this polarisation anisotropy to the depth interval between 210 and 410 km where a low Q has also been observed (Gudmundsson et al. 1994). A thinner asthenospheric layer located at the bottom of the cratonic lithosphere could also explain the observations but the dataset did not allow examination of the anisotropic properties of the lithosphere. Horizontally propagating surface waves present the advantage of allowing the characterisation of seismic anisotropy as a function of depth, and until now the different studies for Australia have provided consistent results regarding the depth extent of polarisation anisotropy. Both the work by Gaherty and Jordan (1995) and the results presented here or in Debayle and Kennett (2000b) suggest that there is significant polarisation anisotropy down to 200–250 km depth within the Australian upper mantle. However, polarisation anisotropy results inferred from the discrepancy between Love and Rayleigh waves requires additional constraints on the anisotropic direction to allow unambiguous interpretation. Gaherty and Jordan (1995) used a dataset restricted to a single seismic corridor crossing Precambrian Australia, which did not allow them to extract any information regarding anisotropic directions. They interpreted their polarisation anisotropy measurements as reflecting a deformation mostly frozen in the continental lithosphere. The simultaneous inversion of a large Rayleigh wave dataset with a good azimuthal ray coverage has ensured the extraction of the anisotropic directions as a function of depth, with a lateral resolution of a few hundred kilometres and a vertical resolution of a few tens of kilometres. This is sufficient to detect a change in the organisation of seismic anisotropy between the upper 150 km of the Earth and deeper mantle structure, which does not support the hypothesis of Gaherty and Jordan (1995) that the anisotropy is frozen throughout the lithosphere. A change of anisotropic properties with depth is also more easy to reconcile with indications from body waves for polarisation anisotropy in the asthenosphere (Tong et al. 1994) and with the observation of weak SKS splitting beneath Australia. If anisotropy was vertically coherent over a single lithospheric layer, significant SKS splitting would be expected. However, we need to keep in mind that even with a fairly dense azimuthal distribution of rays our surface-wave analysis is based on a number of assumptions regarding the propagation of surface waves in anisotropic media. These limitations of the current theory restrict the frequency range and the number of modes for which surface waves can be safely processed, although the restrictions are relatively well known (Maupin 1989; Kennett 1995) and a careful analysis should allow us to obtain pertinent results on the

38

E. Debayle and B. L. N. Kennett

Figure 12 Top: the main Precambrian tectonic units of Australia (after Myers et al. 1996)—1830 Ma cratons; 1900–1300 Ma Central Australian Terranes; and the 1300 Ma Albany–Fraser Orogen. The Central Australian Terranes were accreted onto the North Australian Craton prior to its 1300 Ma suturing with the West and South Australian Cratons. Bottom: SV wave velocity pattern at 100 km depth derived from the Rayleigh wave inversion pushed to 40 s (a) or 50 s (b) period.

Earth’s upper mantle. The horizontal resolution is limited to a few hundred kilometres and care needs to be taken with the interpretation of complex anisotropic patterns. Figure 9 shows that the horizontal smoothing may locally produce wrong anisotropic results in regions where the anisotropic directions change abruptly. In general, if the anisotropic directions in the upper layer of the mantle change with a scale-length smaller than the horizontal resolution, smoothing effects are likely to reproduce the complex pattern of anisotropy, but with directions that are locally wrong. In addition, we have shown in this paper that the

short wavelength component of heterogeneity and anisotropy present in the upper layer of the model is affected by the cut-off frequency chosen when analysing the data. This is because shorter period surface waves are better for picking up the details of the shallow structures, but are also more likely to be affected by errors in the theory used for the analysis. Choosing a reasonable cut-off frequency when analysing the data is difficult because it depends of the level of heterogeneity present in the actual Earth, which is one of the answers that we are trying to address in this class

Surface-wave studies, Australian region

of studies. To illustrate this difficulty, we present in Figure 12 a comparison between a reasonable a priori guess on the shape of the craton, derived from the work of Myers et al. (1996) and our S wave velocity maps from the 40 and 50 s inversions. A 40 s inversion seems to better underline the shape of the northern, southern and western Precambrian cratons. However, if the 40 s inversion provides a better horizontal resolution, it is also more likely to be sensitive to inaccuracy in the assumptions. A good crosscheck of our results has recently been provided through an independent study of upper mantle anisotropy from electromagnetic data (Simpson 2001). The anisotropic directions obtained in this study are inferred to reflect the orientation of the fast a-axis of olivine crystals at depth greater than 150 km, so that they can be directly compared with our anisotropic maps. A good agreement between the electromagnetic strikes and our anisotropic map at 200 km depth has been observed (Simpson 2001) and suggests that despite the approximations underlying the methods, both techniques provide significant constraints on the Earth’s upper mantle. It is also reassuring to see that a closer correspondence with the results of Simons et al. (1999) can be obtained, at least for the upper 200 km of the mantle, by working in the same frequency range for the fundamental mode. This indicates that the differences which remain at 200 km depth and below are probably related to the treatment of higher mode information. It is likely that the strong level of heterogeneity present in the Australian upper mantle renders analysis of the higher modes dangerous at periods near 20 s, where problems related to the neglect of mode coupling and off-great-circle propagation are likely to become important.

CONCLUSIONS The horizontal smoothing present in surface-wave inversion and the frequency dependence of the surface waves results obtained at 100 km depth for Australia suggest that the detail of the short wavelength variation in anisotropic directions should be interpreted with caution. However, the tomographic models are robust, especially when paths likely to experience complex propagation are removed. The application of the path-average approximation directly to the shear slowness is a reasonable approximation for current models of Australia. Surface waves clearly demonstrate that anisotropy is present down to 200–250 km depth beneath Australia with an organisation which is not coherent with depth. At least two anisotropic layers have been identified, the complex anisotropy in the uppermost layer being likely to reflect frozen deformation in the lithosphere. At depths greater than 150 km, the smoother anisotropic pattern with a dominant north–south component is more likely to reflect present-day deformation due to shearing at the bottom of the northward-moving Australian Plate. The depth of 150 km therefore marks an upper bound for the thickness of the ‘mechanical’ lithosphere, defined as the layer of the mantle that presents coherent horizontal motion. This is especially true for central Proterozoic Australia which is well sampled in the current surfacewaves studies.

39

ACKNOWLEDGEMENTS Part of this work was done while Brian Kennett was a visiting professor at University Louis Pasteur in Strasbourg. Our studies of the reliability of surface-wave images were stimulated by comments by Frederic Simons and Rob van der Hilst. Supercomputer facilities were provided by the French Institute for Development and Resources in Intensive Scientific Computing (IDRIS) and the Australian National University (ANU) in Canberra. The maps and the crosssection presented in this paper have been made using the GMT software (Wessel & Smith 1995). The IRIS and GEOSCOPE teams provided seismological data at the permanent and PASSCAL temporary stations. The management of the data from the portable experiments was made possible by the system established by Rob van der Hilst. Finally, special thanks are addressed to the Research School of Earth Sciences (ANU) personnel who collected the SKIPPY and KIMBA data in the field.

REFERENCES CARA M. & LÉVÊQUE J. 1987. Waveform inversion using secondary observables. Geophysical Research Letters 14, 1046–1049. CLITHEROE G. & VAN DER HILST R. D. 1998. Complex anisotropy in the Australian lithosphere from shear-wave splitting in broad-band SKS records. In: Braun J., Dooley J., Goleby B., van der Hilst R. & Klootwijk C. eds. Structure and Evolution of the Australian Continent, pp. 73–78. American Geophysical Union Geodynamics Monograph 26. DEBAYLE E. 1999. SV-wave azimuthal anisotropy in the Australian upper-mantle: preliminary results from automated Rayleigh waveform inversion. Geophysical Journal International 137, 747–754. DEBAYLE E. & KENNETT B. L. N. 2000a. The Australian continental upper mantle: structure and deformation inferred from surface waves. Journal of Geophysical Research 105, 25423–25450. DEBAYLE E. & KENNETT B. L. N. 2000b. Anisotropy in the Australasian upper mantle from Love and Rayleigh waveform inversion. Earth and Planetary Science Letters 184, 339–351. DZIEWONSKI A. M. & ANDERSON D. L. 1981. Preliminary reference Earth model. Physics of the Earth and Planetary Interiors 25, 297–356. GAHERTY J. & JORDAN T. H. 1995. Lehmann discontinuity as the base of anisotropic layer beneath continents. Science 268, 1468–1471. GIRARDIN N. & FARRA V. 1998. Azimuthal anisotropy in the upper mantle from observation of P-to-S converted phases: application to southeast Australia. Geophysical Journal International 133, 615–629. GUDMUNDSSON O., KENNETT B. L. N. & GOODY A. 1994. Broadband observations of upper-mantle seismic phases in northern Australia and the attenuation structure in the upper mantle. Physics of the Earth and Planetary Interiors 84, 207–226. KENNETT B. L. N. 1995. Approximations for surface-wave propagation in laterally varying media. Geophysical Journal International 122, 470–478. KENNETT B. L. N. 1998. Guided waves in 3-dimensional structures. Geophysical Journal International 133, 159–174. KENNETT B. L. N. 2003. Seismic structure in the mantle beneath Australia. Geological Society of Australia Special Publicatin 22 and Geological Society of America Special Paper 372, 7–23. KENNETT B. L. N. & YOSHIZAWA K. 2002. A reappraisal of regional surface wave tomography. Geophysical Journal International 150, 37–44. LÉVÊQUE J. & CARA M. 1985. Inversion of multimode surface wave data: evidence for sub-lithospheric anisotropy. Geophysical Journal of the Royal Astronomical Society 83, 753–773. LÉVÊQUE J., DEBAYLE E. & MAUPIN V. 1998. Anisotropy in the Indian Ocean upper mantle from Rayleigh- and Love-waveform inversion. Geophysical Journal International 133, 529–540. MAUPIN V. 1989. Surface waves in weakly anisotropic structures: on

40

E. Debayle and B. L. N. Kennett

the use of ordinary or quasi-degenerate perturbation methods. Geophysical Journal International 98, 553–563. MARQUERING H., SNIEDER R. & NOLET G. 1996. Waveform inversion and the significance of surface-wave mode coupling. Geophysical Journal International 124, 258–278. McEVILLY T. 1964. Central US crust upper-mantle structure from Love and Rayleigh wave phase velocity inversion. Bulletin of the Seismological Society of America 54, 1997–2016. MONTAGNER J. 1986. Regional three-dimensional structures using long-period surface waves. Annales Geophysicae 4, 283–294. MONTAGNER J. 1994. Can seismology tell us anything about convection in the mantle? Review of Geophysics 32, 115–137. MONTAGNER J. & NATAF H. C. 1986. A simple method for inverting the azimuthal anisotropy of surface waves. Journal of Geophysical Research 91, 511–520. MYERS J., SHAW R. & TYLER I. 1996. Tectonic evolution of Proterozoic Australia. Tectonics 15, 1431–1446. NATAF H. C. & RICARD Y. 1995. 3SMAC: an a priori tomographic model of the upper mantle based on geophysical modelling. Physics of the Earth and Planetary Interiors 95, 101–122. NOLET G. 1990. Partitioned waveform inversion and two dimensional structure under the network of autonomously recording seismographs. Journal of Geophysical Research 95, 8499–8512. OZALAYBEY S. & CHEN W-P. 1999. Frequency dependent analysis of SKS/SKKS waveforms observed in Australia: evidence for null birefringence. Physics of the Earth and Planetary Interiors 114, 197–210. SALTZER R., GAHERTY J. & JORDAN T. H. 2000. How are vertical shear wave splitting measurements affected by variations in the orientation of azimuthal anisotropy with depth? Geophysical Journal International 141, 374–390. SILVER P. & CHAN W. 1991. Shear wave splitting and subcontinental mantle deformation. Journal of Geophysical Research 141, 16429–16454. SIMONS F., ZIELHUIS A. & VAN DER HILST R. D. 1999. The deep structure of the Australian continent from surface wave tomography. Lithos 48, 17–43. SIMPSON F. 2001. Resistance to mantle flow inferred from the electro-

magnetic strike of the Australian upper mantle. Nature 412, 632–635. SMITH M. & DAHLEN F. 1973. The azimuthal dependence of Love and Rayleigh wave propagation in a slightly anisotropic medium. Journal of Geophysical Research 78, 3321–3333. TONG C., GUDMUNDSSON O. & KENNETT B. 1994. Shear wave splitting in refracted waves returned from the upper mantle transition zone beneath northern Australia. Journal of Geophysical Research 99, 15783–15797. VAN DER HILST R., KENNETT B., CHRISTIE D. & GRANT J. 1994. Skippy: mobile broad-band arrays to study the seismic structure of the lithosphere and mantle beneath Australia. EOS 75, 177–181. VAN DER HILST R. D., KENNETT B. L. N. & SHIBUTANI T. 1998. Upper mantle structure beneath Australia from portable array deployments. In: Braun J., Dooley J., Goleby B., van der Hilst R. & Klootwijk C. eds. Structure and Evolution of the Australian Continent, pp. 39–58. American Geophysical Union Geodynamics Monograph 26. VINNIK L., MAKEYEVA L., MILEV A. & USENKO A. Y. 1992. Global patterns of azimuthal anisotropy and deformations in the continental mantle. Geophysical Journal International 111, 433–447. WESSEL P. & SMITH W. H. F. 1995. New version of the Generic Mapping Tools released. EOS 76, 329. WOODHOUSE J. H. 1974. Surface waves in a laterally varying layered structure. Geophysical Journal of the Royal Astronomical Society 37, 461–490. YOSHIZAWA K. & KENNETT B. L. N. 2002. Determination of the influence zone for surface wave paths. Geophysical Journal International 149, 440–453. ZIELHUIS A. & NOLET G. 1994. Shear-wave velocity variations in the upper mantle beneath central Europe. Geophysical Journal International 117, 695–715. ZIELHUIS A. & VAN DER HILST R. D. 1996. Upper-mantle shear velocity beneath eastern Australia from inversion of waveforms from Skippy portable arrays. Geophysical Journal International 127, 1–16. Received 4 July 2001; accepted 8 January 2002

Geol. Soc. Australia Spec. Publ. 22, and Geol. Soc. America Spec. Pap. 372 (2003), 41–48

Is the Australian Plate deforming? A space geodetic perspective P. TREGONING Research School of Earth Sciences, Australian National University, ACT 0200, Australia ([email protected]). Measurements at discrete points spanning the Australian Plate have been made for over a decade using the Global Positioning System (GPS) space geodetic technique. These measurements show that, to within the resolution of the technique (~2 mm/y at 95% confidence level), there are no significant changes in the dimensions of the Australian Plate across the Australian continent. That is, no changes in baseline lengths are evident between any sites located in Australia when taking into account the measurement and modelling errors. However, during the past two decades, several significant earthquakes have occurred within the Australian Plate indicating that at times stress failure levels are reached, resulting in failure within the crust. Therefore, the rate at which stress accumulates must be slower than is visible in the geodetic measurements. With the exception of two sites affected by earthquake co-seismic displacement and equipment failure, all time series of sites in the interior of the Australian Plate are linear and site velocities are not significantly different from the predicted motion of the ‘rigid’ Australian Plate. However, the northern margin of the plate in Papua New Guinea is undergoing regional deformation. This is probably a result of the interaction with neighbouring plates and the proximity of the GPS site to nearby plate-boundary zones. KEY WORDS: Australian Plate, GPS, intraplate deformation.

INTRODUCTION The Australian Plate is one of 14 major tectonic plates covering the surface of the Earth (DeMets et al. 1990) and shares tectonic boundaries with seven other macro- and micro-plates (Figure 1). An active spreading centre separates the southern margin of the Australian Plate from the Antarctic Plate, while along the eastern margin subduction of the Pacific Plate occurs on several active subduction trenches (Figure 1). The Australian Plate subducts beneath the Solomon Islands at the San Cristobal Trench, separates from the Woodlark Plate along the Woodlark Basin Spreading Centre and collides with the South Bismarck Plate along the Ramu–Markham Fault in Papua New Guinea (Figure 1). The northwestern margin of the Australian Plate subducts beneath the Sundaland Block along the Java Trench while convergence is partitioned between oblique subduction beneath Sumatra and rightlateral slip on the Great Sumatra Fault. The western boundary of the Australian Plate is much more poorly defined. Early global plate-motion models (Le Pichon 1968; Minster et al. 1974) postulated that the IndoAustralian Plate was one rigid entity that stretched west to the boundary with the African Plate. Minster and Jordan (1978) showed that the geophysical data of transform fault azimuths, spreading rates etc. were better fitted if the Indian and Australian Plates were considered to be separate entities. Subsequent global plate models (DeMets et al. 1990, 1994a; Argus & Gordon 1991) considered the two plates as separate, rigid plates although did not define a discrete boundary between them. More recent analyses of gravity data and spreading rates from aeromagnetic surveys suggest that the relative motion between the Indian and

Australian Plates is taken up on many fracture zones and large regions of distributed deformation have been identified (DeMets et al. 1994b; Royer et al. 1997; Gordon et al. 1998) (see Figure 1). Since 1976, 19 earthquakes with magnitude Mw >5.0 have occurred within the interior of the Australian Plate and may be considered to be ‘intraplate’ earthquakes (Figure 1). This indicates that stresses do accumulate within the Australian Plate and, from time to time, large readjustments of the stress field occur (Wdowinski 1998; Hillis & Reynolds 2000) Since the 1970s space geodetic techniques have been used to measure the precise coordinates of discrete sites spanning the Australian Plate. Very Long Baseline Interferometry (VLBI) observations of baselines in Australia began in 1982 (Harvey 1985), while Satellite Laser Ranging (SLR) measurements commenced in Canberra and Yaragadee in the 1970s. Monitoring of the Australian continent using GPS began in the early 1990s and the network of permanently operating sites continues to increase in density. In this paper the motion of 12 sites estimated from GPS observations are presented. The data span at least three years and in some cases 10 years and the GPS sites cover the entire Australian Plate. The other two space geodetic techniques are limited in their spatial distribution, with SLR observations at only Yaragadee and Canberra and VLBI observations at Hobart, Canberra and some at Parkes and Narrabri. Therefore, the spatial resolution is considerably better from the GPS network and only GPS velocity estimates will be considered below. With only a couple of exceptions, the estimated site velocities are consistent with a rigid Australian Plate.

42

P. Tregoning

Figure 1 Tectonic map showing the boundaries of the Australian Plate. Bathymetry and topography is taken from the ETOPO5 dataset (ETOPO5 1986). GPS sites on the Australian Plate and their ITRF97 velocities are plotted along with the locations of ‘intraplate’ earthquakes (black circles) which have occurred since 1976 within the geographical coverage of the GPS network. The areas bounded in pink are diffuse plate boundaries between the Australian, Indian (IND) and Capricorn Plates (CAP) (Gordon et al. 1998). 1, Southeast Indian Ridge; 2, Kermadec Trench; 3, New Hebrides Trench; 4, San Cristobal Trench; 5, Woodlark Basin Spreading Centre; 6, Java Trench; 7, Ninety East Ridge; 8, Carlsberg Ridge; 9, Central Indian Ridge.

GPS NETWORK AND DATA ANALYSIS The Australian Surveying and Land Information Group (AUSLIG) has established a permanent network of continuously operating GPS stations across Australia (Figures 1, 2) and data from this network forms the basis of this analysis. Observations made at these sites prior to their continuous operation (Tregoning 1996) have also been included. In addition, campaign-style observations have been made in Cape York in 1996 and 1999 and data were used from two sites located in Auckland and Noumea, which are freely available from the International GPS Service (IGS) (Beutler et al. 1993) and one site in Port Moresby (operated by the Papua New Guinea National Mapping Bureau), to extend the span of the network out to the most northern and eastern extremities of the Australian Plate. The GPS data have been analysed using the GAMIT/GLOBK software (King & Bock 2000; Herring 2000) following the procedures explained in detail in Feigl et al. (1993) and Dong et al. (1998). In brief, data from sites in the Australasian region were combined with up to 70 sites

from the global IGS network to derive daily estimates of the site positions. Satellite orbits, earth orientation parameters, tropospheric delay parameters and site positions are all estimated simultaneously using a least-squares process. Two different solutions were generated for the analysis presented below. In the first solution the coordinates and velocities of 45 core IGS sites were constrained to their values in the International Terrestrial Reference Frame 1997 (ITRF97) (Boucher et al. 1999) (but did not constrain any sites located on the Australian Plate). A Kalman back-filter solution was performed with stochastic variation permitted on the sites located on the Australian Plate to generate daily estimates of the positions of all such sites. In the second solution a 7-parameter Helmert transformation of the velocities of the 12 sites located on the Australian Plate was performed with respect to a stationary reference frame (i.e. relative to the Australian Plate itself as defined by the velocities of these sites) (Table 1). Any sites with velocities which are significantly different from zero (all statistical tests are two-tailed, 2 confidence interval, normal distribution tests unless otherwise specified) in

Is the Australian Plate deforming? Table 1 Site velocities (and 1 uncertainties) with respect to a rigid Australian Plate defined by a best-fitting Euler vector using the velocities of the sites listed below (except MORE). Site

Code

Alice Springs Auckland Cocos Islandsa Darwin Hobart Karratha Noumea Perth Tidbinbilla Townsville Yaragadee Port Moresby Mt Stromlo

ALIC AUCK COCO DARW HOB2 KARR NOUM PERT TIDB TOW2 YAR1 MORE STR1

Velocity with respect to Australian Plate (mm/y) Vn n Ve e 0.1 –0.9 –0.0 –0.6 0.1 0.2 0.2 0.5 0.5 –0.7 –0.7 0.3 1.2

0.6 0.5 0.6 0.5 0.5 0.5 0.7 0.4 0.5 0.6 0.4 0.5 0.7

0.0 –0.9 –0.7 –0.3 –0.5 0.6 –0.6 0.7 0.2 –0.2 0.6 –3.8 0.6

0.6 0.5 0.6 0.6 0.5 0.5 0.7 0.5 0.5 0.5 0.5 0.5 0.8

a

Velocity estimate for the Cocos Islands excludes data subsequent to the 16 June 2000 earthquake.

such a reference frame indicate that these sites are moving relative to the Australian Plate. With the exception of Port Moresby (see discussion below), none of the residuals is significantly different from zero at even the 1 level (66% confidence interval) (Figure 2; Table 1), indicating that the GPS analysis is not able to detect any significant shortening or extension across the Australian Plate from Cocos Island in the west to Noumea and Auckland in the east. Morgan et al. (1996) suggested that GPS velocities relative to the ‘NUVEL model’ [I presume here that this referred to the No-Net-Rotation NUVEL-1 model (DeMets et al. 1990; Argus & Gordon 1991)] showed evidence for east–west compression of the Australian Plate. Using a more comprehensive dataset, I find no such evidence at the sub-2 mm level. These results confirm the earlier analyses of Tregoning (1996) who concluded that no sig-

43

nificant deformation could be detected within the Australian Plate. The site at Port Moresby (MORE) shows relative motion with respect to the Australian Plate of ~4 ±1 mm/y with an azimuth of 274 ±15° (Figure 2). Previous estimates of the velocity of MORE have found no significant motion relative to the Australian Plate (Tregoning et al. 1998). The estimate here is derived from data with a considerably longer temporal scale (~11 years rather than ~6 years), with most of the observations occurring after the establishment of the IGS global tracking network. The enhanced tracking network after this time improves the accuracy of the the orbit estimations; hence the site coordinates and velocity estimates are more accurate. The uncertainties in these velocity estimate are considerably smaller than those of Tregoning et al. (1998), making the motion relative to the Australian Plate statistically significant.

TEMPORAL ANALYSIS Figure 3 shows the time series of daily coordinates estimates for the sites at Alice Springs (ALIC), Auckland (AUCK), Karratha (KARR) and Townsville (TOW2). These time series are indicative of the eight other sites estimated but not shown here and show clearly that, to within the level of uncertainty of the GPS measurements (~5 mm for each individual daily estimate), each site is moving with linear horizontal motion. There are two sites which do not show consistent linear motion: Cocos Islands (COCO) and Mt Stromlo (STR1). The time series of COCO has a discontinuity on 16 June 2000 (Figure 4) that corresponds to the timing of a Mw =7.8 earthquake which occurred ~150 km to the southeast of the GPS site (Robinson et al. 2001). The offsets of ~ –7 mm in north and ~+34 mm in east are a measure of the co-seismic displacement which occurred at the Cocos Islands as a result of the seismic event. This event has been interpreted as an intraplate earthquake, and Robinson et al. (2001) suggested that the diffuse boundary between the Australian

Figure 2 Site velocities (and 1 error ellipses) with respect to a rigid Australian Plate. Most of the vectors are so small they are barely visible. Relative motion at MORE is clearly evident.

44

P. Tregoning

Figure 3 GPS time series from 1994 to 2001 for four typical GPS sites: Alice Springs (ALIC), Auckland (AUCK), Karratha (KARR) and Townsville (TOW2). nrms, normalised root mean square; wrms, weighted root mean square using the daily formal GPS position uncertainties.

Is the Australian Plate deforming?

45

Figure 4 GPS time series at Cocos Islands (COCO). The time of the Mw =7.8 earthquake is indicated by a vertical line, with subsequent site estimates plotted in red. The pre-earthquake linear velocities are shown for each component.

and Indian Plate may be localising with time into a narrower north–south region of deformation. The correlation between the timing of the discontinuity in the time series at COCO and the physical occurrence of the earthquake shows the small levels of discontinuous motion which can be detected with GPS. It also adds strength to the arguments that, where no motion is detected between other sites, no significant motion has occurred. The second non-linear site, Mt Stromlo, is more difficult to explain. The site has been observing continuously since May 1998, although there are discontinuities of up to 100 mm evident during 1998/99 (Figure 5). Some of these jumps coincide with changes of equipment hardware, but not all. The GPS receiver is co-located with the Satellite Laser Ranging (SLR) system in Canberra. The SLR time series at Mt Stromlo does not show the discontinuities visible in the GPS time series (Figure 5). Between September 1998 and May 1999 the GPS antenna used at STR1 was an Ashtech GPS/GLONASS model (IGS code ASH701073.1), whereas before and after this period the antenna was a Turborogue Dorne Margolin (IGS code AOAD/M_T). Since September 1998, the same GPS receiver has been used (Turborogue: IGS code AOA ICS-4000z); thus, the anomalous record cannot be considered to be caused by receiver malfunctions. It is clear in the time series that the observations before September 1998 and after May 1999 show linear motion of the site, while all

observations between these times are not part of the same population. I suspect that there was either incompatibility between the Turborogue receiver and Ashtech antenna or that the Ashtech antenna failed slowly, causing a horizontal drift in the position estimates. Once this antenna was removed, the position estimates appear to once more fit a pattern of linear motion. This effectively rules out electrical interference from other equipment at the Mt Stromlo site as the cause of the anomalous motion. The antenna which may have been faulty has since been returned to the manufacturer and replaced (R. Govind pers. comm. 2001) and hence it is not possible to perform specific tests to demonstrate conclusively that the explanation above is correct. However, given that since May 1999 the time series at STR1 is essentially linear and that the SLR time series of the co-located SLR tracking station shows no such discontinuities during 1998 and 1999, it is reasonable to assume that the non-linear motion at Mt Stromlo is not related to tectonic movement within the Australian Plate.

DEFORMATION IN PAPUA NEW GUINEA The GPS site at Port Moresby located on the southwestern coast of the Papuan Peninsula is the only site which shows systematic, significant relative motion with respect to the Australian Plate. Previous authors have assumed that it lies

46

P. Tregoning

Figure 5 Time series of position estimates at Mt Stromlo (STR1) from GPS and SLR observations. The GPS observations that occurred during the time of suspected faulty equipment are shown in red. These data were excluded from the calculation of the linear velocity plotted in the figure. SLR estimates are derived from SINEX files provided by AUSLIG.

Is the Australian Plate deforming?

on the rigid Australian Plate (Tregoning et al. 1998). This site lies ~100 km west of the likely boundary between the Australian and Woodlark Plates where the predicted relative motion between these two plates is ~10 mm/y (Tregoning et al. 1998). Assuming that the plate boundary here is locked, one could expect to see interseismic strain accumulation in regions close to the locked boundary. The actual location of the boundary is not known (Tregoning et al. 1998). Modelling the interseismic strain [using a screw dislocation (Okada 1985) and assuming that the plate boundary runs through the middle of the Papuan Peninsula and is locked to 20 km depth] results in predicted relative motion at MORE of v), thrust with strike-slip component, strike-slip (H > v > h), normal with strike-slip component, or normal (v > H > h), where v, H and h are the vertical, maximum horizontal and minimum horizontal stresses respectively (Zoback 1992). Hydraulic fracturing and overcoring tests yield absolute values of stress magnitudes (Hillis et al. 1999) and these have also been incorporated in the determinations of average stress regime in each of the provinces (Table 2). Borehole breakouts and drilling-induced tensile fractures, which make up the 50%

of the Australian stress map database, do not yield information on stress magnitudes.

Stress trajectories Stress-trajectory determination provides a technique for smoothing and interpolating unevenly distributed stress data and thereby clarifying regional trends. A stress-trajectory map (Figure 6) has been calculated from the Australian stress map database following the technique of Hansen and Mount (1990). Stress trajectories determined following this technique indicate the orientation of the maximum horizontal stress at each point along the trajectory. However, they do not imply information about magnitudes (the spacing between the trajectories does not have any significance).

In situ stress field of Australia

The technique applies a statistical smoothing algorithm to create an estimated stress field at each observed data location. In the technique, fidelity to the raw data must be balanced against the degree of smoothing. If there is too much fidelity to the raw data, the calculated stress trajectories simply reflect the stress data, and if there is too much smoothing, variations are obscured. In determining the estimated stress field, three weighting systems were applied to the raw data. First, data were weighted according to their proximity to the observed data point being smoothed. Second, relative weightings of 4 were given to A-quality data, 3 to B-quality data, 2 to Cquality data and 1 to D-quality data. Third, a robustness weight was applied to eliminate the presence of anomalous data that differed substantially from other observed data in the region. The smoothed stress field, which was calculated at each of the observed data points, was then used to calculate the stress trajectories as outlined by Hansen and Mount (1990).

PATTERN OF REGIONAL STRESS ORIENTATION IN THE AUSTRALIAN CONTINENT Both techniques used to clarify regional trends in stress orientation across the Australian continent show consistent patterns (Figures 4, 6). The western part of the Australian continent is characterised by broadly east–west-oriented maximum horizontal stress (Carnarvon Basin and Perth provinces). The east–west maximum horizontal stress orientation in the western part of the Australian continent rotates to northeast–southwest along the northern Australian margin (Canning Basin, Northern and Southern Bonaparte Basin, Irian Jaya and New Guinea provinces). This swing in stress trajectories along the northern Australian margin is broadly paralleled to the south where east–west-oriented maximum horizontal stress in the Perth province rotates to north-northeast–south-southwest in the Amadeus Basin (central Australia). The Bowen Basin also exhibits north-northeast–south-southwest-oriented maximum horizontal stress. Moving west to east in the southern part of the continent, maximum horizontal stress rotates from east–west in the Perth Basin to northwest–southeast in southeastern Australia (Otway Basin and Gippsland Basin provinces). The area of divergence between northnortheast–south-southwest and northwest–southeast maximum horizontal stress trajectories in central eastern Australia is characterised by east–west or poorly defined (low horizontal stress anisotropy) regional stress trends (Cooper Basin, Flinders Ranges and Sydney Basin provinces). Both the stress trajectories and stress provinces identified a regional northwest–southeast stress trend throughout southeastern Australia, although sections of onshore southeastern Australia are dominated by scattered focal mechanisms and engineering-type measurements indicating an approximate east–west stress direction. The scattered nature of these stress indicators suggest that local stress sources strongly influence the stress field and hence we believe these indicators are not a reliable guide to the nature of the regional stress field. It is immediately apparent that, unlike most other continental areas, stress orientations in the Australian conti-

55

nent as a whole are variable and do not parallel the north to north-northeast absolute motion direction of the IndoAustralian Plate. South America, western Europe and midplate North America are all characterised by broad regions where maximum horizontal stress orientation is consistent and parallel to the direction of plate motion (Zoback et al. 1989; Zoback 1992; Richardson 1992). From this it is inferred that the forces driving and/or resisting plate motion are responsible for regional stress orientations in those continental areas. The absence of correlation between stress orientation and absolute plate motion direction in the Australian continent begs the question as to whether the in situ stress field of the Australian continent is subject to different controls than that of other continents. Regional stress orientations in the Australian continent are not obviously affected either by tectonic province, regional structural trends, geological age or by the depth at which stress orientations are sampled. This is perhaps best witnessed by the Perth province. In the eastern part of the Perth province data from focal mechanisms, overcoring and hydraulic fracturing all indicate broadly east–west oriented maximum horizontal stress (Figure 3). These data come from the Precambrian Yilgarn Craton which is separated from the Phanerozoic Perth Basin to the west by an approximately north–south-trending crustal-scale fault (Darling Fault). Breakout data from the Perth Basin similarly indicate that maximum horizontal stress is broadly oriented east–west (although locally anomalous stress orientations are observed in the vicinity of known faults). Hence in the Perth province stress orientations are consistent from nearsurface to several kilometres depth and across a major tectonic boundary. Again although some locally anomalous stress orientations are observed, especially in the vicinity of faults, the northeast–southwest regional maximum horizontal stress orientation along much of the northern Australian margin is also unaffected by regional structural trends, geological age or by the depth at which stress orientations are sampled. The Southern Bonaparte Basin and onshore Canning Basin are Palaeozoic basins characterised by northwest–southeast structural trends. In the latter basin, focal mechanism and breakout data both indicate northeast–southwest-oriented maximum horizontal stress. The Northern Bonaparte Basin is dominated by younger, Mesozoic northeast–southwest trends associated with the formation of the present passive margin. Structural trends in New Guinea are northwest–southeast and associated with Cenozoic collision in the area. All of these provinces, with different age and orientation of dominant structure, display a northeast–southwest regional maximum horizontal stress orientation. The swing in maximum horizontal stress orientation from east–west in western Australia to northwest–southeast along the northern Australian margin is part of a plate-wide anticlockwise rotation in stress orientation from broadly north–south in India to northwest–southeast in the vicinity of the Ninety East Ridge to east–west in western Australia. This broad rotation can be accounted for by focusing of stresses orthogonal to the Himalayan and New Guinea continental collision segments of the northeastern boundary of the Indo-Australian Plate. We believe that the first-order pattern of stresses in continental Australia can indeed be accounted for by plate-boundary forces, if the complex

56

R. R. Hillis and S. D. Reynolds

Figure 7 Comparison of mean maximum horizontal stress directions in the 16 stress provinces defined herein with SV-wave azimuthal anisotropy from Rayleigh wave inversion (waveforms analysed for the frequency range 40–160 s). SV-wave azimuthal anisotropy is from Debayle & Kennett (2003 figure 4a). SV-wave: shear wave polarised such that motion is in the vertical plane which also contains direction of wave propagation.

nature of the northeastern boundary of the Indo-Australian Plate is recognised. In an accompanying paper, Reynolds et al. (2003) have considerably improved the fit of plateboundary force-induced stresses to the observed stress data from that of previous modelling studies (Coblentz et al. 1998), especially in eastern Australia, by considering a very large number of permutations of plate-boundary forces. Stress orientations in Australia contrast with those of other continents in not being parallel to the direction of absolute plate motion, not because the first-order control on stress orientations in Australia differs, but because its plateboundary configuration is more complex than that of most other continents. Continental areas that exhibit consistent maximum horizontal stress orientations that are parallel to the direction of absolute plate motion, such as western Europe, South America and mid-plate North America, are surrounded by much simpler plate-boundary configurations. For example, northwest-oriented maximum horizontal stress in western Europe is a consequence of compressional forces generated orthogonal to the Mid-Atlantic Ridge and the Alpine collisional belt (Gölke & Coblentz 1996). Throughout continental Australia locally anomalous second-order stress variations occur that are not controlled by plate-boundary forces. For example fault-parallel stress rotation perpendicular to the regional stress field occurs in the Perth Basin (Reynolds & Hillis 2000). Such secondorder stress variations are likely caused by structural, topographic and density variations within the lithosphere. Whilst these variations are important on a local scale, their effects are not clearly witnessed in regional stress provinces and stress trajectories determined herein. Determining the source of local stress variations is often problematic due to the lack of detail in the stress/structural datasets. We acknowledge that excluding geology to model the Australia stress field is an oversimplification, although it is justifiable for the modelling of first-order tectonic stresses.

COMPARISON OF REGIONAL STRESS ORIENTATION IN THE AUSTRALIAN CONTINENT WITH OTHER GEOPHYSICAL DATABASES Several geophysical databases for the Australian continent are discussed in this volume. These databases hold the key to our understanding of the evolution and dynamics of the Australian continent. However, in order to elucidate the evolution and dynamics of the Australian continent, it may be possible, indeed necessary, to categorise such databases into those whose dominant response reflects the evolution (geological history) of the Australian continent and those whose dominant response reflects its (present-day) dynamics. The magnetic and gravity anomaly images of Australia can be divided into geophysical domains that, where basement is exposed, correspond with geologically mapped cratons and blocks (Wellman 1998). These domains comprise large contiguous areas with a common geological history. The clear expression of the Tasman Line (the eastern boundary of the Australian Proterozoic craton along which Late Neoproterozoic rifting and breakup and subsequent Phanerozoic collision occurred) in the potential field images (Milligan et al. 2003) clearly illustrates that the response of these datasets is controlled by the structural evolution of the Australian continent. Although long-wavelength components of both the magnetic and gravity anomaly images of Australia do reflect present-day dynamics (isostasy and depth to Curie temperature in the case of the gravity and magnetic images respectively), the dominant response reflects the structural evolution of the different geological domains of Australia, and the variation in physical properties ‘frozen into’ these different domains as a result of their different geological histories. The in situ stress data do not, at least at the first-order, reflect the different geological domains of the Australian continent. Indeed, as discussed above, stress orientations

In situ stress field of Australia

are insensitive to major tectonic boundaries such as that between the Yilgarn Craton and the Perth Basin, and to variation in regional structural trends or geological age along the northern Australian margin. Nor is there a first-order control on stress orientations in the eastern-third of Australia by the dominantly north–south trends of the Tasman fold-belt system. Reynolds et al. (2003) show that the first-order variation of in situ stress orientations in the Australian continent can be successfully accounted for by forces acting on the present-day boundaries of the plate. Hence, as in other continental areas (Richardson 1992), the in situ stress field of Australia is controlled by the present-day dynamics rather than the structural evolution of the lithosphere. The pattern of stress orientation in the Australian continent shows distinct similarities to the pattern of shear (SV-) wave azimuthal anisotropy from Rayleigh wave inversion (Debayle & Kennett 2003). The lateral resolution of the determinations of seismic anisotropy is of the order of a few hundred kilometres and is thus similar to that of the stress provinces defined herein. The fit between stress orientations and SV-wave azimuthal anisotropy is good in the Bowen, Gippsland, Otway and Cooper Basins of eastern Australia (Figure 7). The fit is poor in the Amadeus Basin. However, the broad variation in direction of seismic anisotropy from essentially north–south in the Bowen Basin to east–west in the Cooper Basin to north–south west of the Amadeus Basin is very similar to the variation exhibited by the stress data. The fit between stress orientations and SV-wave azimuthal anisotropy is also good in the Canning Basin and Northern and Southern Bonaparte Basins (Figure 7). The fit is poor in western Australia, where stress directions are east–west and the seismic anisotropy poorly developed, but essentially north–south (Figure 7). However, due to the distribution of earthquakes along the plate boundaries surrounding Australia, seismic anisotropy is relatively poorly constrained west of 125°E (Debayle & Kennett 2003). Two distinct layers of seismic anisotropy are observed in the Australian continent (Debayle & Kennett 2003). The upper layer is characterised by varying directions of anisotropy as shown in Figure 7. However, at depths greater than 150 km the pattern of anisotropy is smoother with the dominant north–south direction subparallel to the absolute direction of plate motion (Debayle & Kennett 2003). Debayle and Kennett (2003) suggest that the anisotropy in the upper layer reflects ‘frozen’ deformation in the lithosphere, while in the lower layer the smoother pattern is more likely to reflect present-day deformation due to shearing at the bottom of the northward moving Australian Plate. We agree that the anisotropy in the lower layer probably reflects processes in the asthenosphere associated with northward motion of the Australian Plate. However, given the correspondence between the anisotropy in the upper layer and in situ stress directions, we believe that the seismic anisotropy in the upper layer reflects in situ stresses associated with present-day plate dynamics, possibly due to the influence of stress-aligned fluid-saturated microcracks (cf. extensive dilatancy anisotropy: Crampin et al. 1984). Both the upper and lower layers of seismic anisotropy in the Australian continent may thus reflect present-day plate dynamics, with the difference between the two layers witnessing decoupling between lithospheric and asthenospheric stresses.

57

Continental areas that exhibit consistent maximum horizontal stress orientations that are parallel to the direction of absolute plate motion, such as western Europe, South America and mid-plate North America, may provide a test of the processes controlling the upper layer, lithospheric seismic anisotropy. If present-day plate dynamics/in situ stress control both lithospheric and asthenospheric seismic anisotropy (as is suggested above for Australia) then in these continental areas both lithospheric and asthenospheric seismic anisotropy would be expected to be consistent and parallel to the direction of plate motion. If lithospheric seismic anisotropy is controlled by pre-existing structure then these continental areas would be expected to show variable directions of seismic anisotropy in the lithosphere, as is seen in Australia. In order to elucidate the evolution and dynamics of the Australian Plate it is necessary to distinguish datasets controlled by present-day dynamics from those such as the magnetic and gravity anomaly images of Australia that are dominantly controlled by the structural history of the crust.

SUMMARY AND CONCLUSIONS (1) The Australian stress map comprises 549 stress indicators of which 331 yield reliable, A–C-quality information on the orientation of horizontal, tectonic stresses. Compared to the World stress map database (Mueller et al. 2000), data based on earthquake focal mechanisms are relatively underrepresented in the Australian stress map, due largely to the relatively low levels of seismicity in Australia. Data based on borehole breakouts and drilling-induced tensile fracture in hydrocarbon exploration wells are relatively over-represented. Engineeringtype measurements are also well represented, especially in eastern Australia. There are no reliable stress indicators based on young geological features, but we consider that the increased recognition of neotectonic activity in the Australian continent will lead to the future incorporation of such data into the Australian stress map. (2) Sixteen stress provinces have been defined within the Australian continent. Fifteen of the 16 provinces show statistically significant mean stress orientations at the 95% or greater confidence level. The consistency of stress orientations in individual provinces indicates that statistically significant regional orientations are being resolved. However, both the stress provinces and stress trajectory mapping reveal systematic, continental-scale rotations in stress orientation. Hence, unlike most other continental areas, stress orientations in the Australian continent are variable and do not parallel the north to north-northeast absolute motion direction of the Indo-Australian Plate. (3) Regional stress orientations in the Australian continent are not affected to a first-order by either tectonic province, regional structural trends, geological age, or by the depth at which orientations are sampled. A number of locally anomalous stress orientations appear influenced by second-order sources of stress such as structure, topography and density heterogeneities. Despite the absence of parallelism between absolute plate motion and stress orientations, the regional pattern of stress orientation in the Australian continent is consistent with control by plate-

58

R. R. Hillis and S. D. Reynolds

boundary forces, if the complex nature of the northeastern boundary of the Indo-Australian Plate, and stress focusing by collisional segments of the boundary, is recognised (see also Reynolds et al. 2003). (4) In situ stress orientations show a strong correlation with the direction of seismic anisotropy in the lithosphere. It is suggested that both datasets are, to a first-order, controlled by present-day plate dynamics.

ACKNOWLEDGEMENTS The Australian stress map is an ongoing project originally funded by the Australian Research Council (1996–1998). For updates on the Australian stress map see . David Denham and Chris Windsor are thanked for their support of the project. Jim Enever is thanked for providing an extensive engineering-based stress database from eastern Australia, and for collaborative work on that data. Much of the data on the Australian North West Shelf was compiled as part of a PhD project undertaken by Scott Mildren and funded by Geoscience Australia. CSIRO’s Division of Petroleum Resources is thanked for its collaboration with the project. The following companies are thanked for providing data: Ampol, Apache, BHPP, Boral, BP, Cultus, Magellan, MIM, Norcen, Oil Company of Australia, Petroz, Phillips, Santos, TCPL, WMC and Woodside. We are also grateful for data provided by the South Australian, Western Australian and Northern Territory Departments of Mines and Energy and the Australian Bureau of Resource Sciences.

REFERENCES BRUDY M. & ZOBACK M. D. 1999. Drilling-induced tensile wall-fractures: implications for determination of in-situ stress orientation and magnitude. International Journal of Rock Mechanics and Mining Sciences 36, 191–215. COBLENTZ D. D. & RICHARDSON R. M. 1995. Statistical trends in the intraplate stress field. Journal of Geophysical Research 100, 20245–20255. COBLENTZ D. D., ZHOU S., HILLIS R. R., RICHARDSON R. M. & SANDIFORD M. 1998. Topography boundary forces, and the Indo-Australian intraplate stress field. Journal of Geophysical Research 103, 919–931. CRAMPIN S., EVANS R. & ATKINSON B. K. 1984. Earthquake prediction: a new physical basis. Geophysical Journal of the Royal Astronomical Society 76, 147–156. DEBAYLE E. & KENNETT B. L. N. 2003. Surface-wave studies of the Australian region. Geological Society of Australia Special Publication 22 and Geological Society of America Special Paper 372, 25–40. DENHAM D., ALEXANDER L. G. & WOROTNICKI G. 1979. Stresses in the Australian crust: evidence from earthquake and in-situ stress measurements. BMR Journal of Australian Geology & Geophysics 4, 289–295. GÖLKE M. & COBLENTZ D. 1996. Origins of the European regional stress field. Tectonophysics 266, 11–24. GREENHALGH S. A., LOVE D., MALPAS K. & MCDOUGALL R. 1994. South Australian earthquakes, 1980–92. Australian Journal of Earth Sciences 41, 483–495. HANSEN K. M. & MOUNT V. S. 1990. Smoothing and extrapolation of

crustal stress orientation measurements. Journal of Geophysical Research 95, 1155–1165. HILLIS R. R. 1998. Mechanisms of dynamic seal failure in the Timor Sea and Central North Sea. In: Purcell P. G. & Purcell R. R. eds. The Sedimentary Basins of Western Australia 2, pp. 313–324. Proceedings of Petroleum Exploration Society of Australia Symposium, Perth. HILLIS R. R., ENEVER J. R. & REYNOLDS S. D. 1999. In situ stress field of eastern Australia. Australian Journal of Earth Sciences 46, 813–825. HILLIS R. R., MONTE S. A., TAN C. P. & WILLOUGHBY D. R. 1995. The contemporary stress field of the Otway Basin, South Australia: implications for hydrocarbon exploration and production. The APEA Journal 35, 494–506. LAMBECK K., MCQUEEN H. W. S., STEPHENSON R. A. & DENHAM D. 1984. The state of stress within the Australian continent. Annales Geophysicae 2, 723–742. LOVE D. N., PREISS W. V. & BELPERIO A. P., 1995. Seismicity, neotectonics and earthquake risk. In: Drexel J. F. & Preiss W. V. eds. The Geology of South Australia Volume 2 The Phanerozoic, pp. 268–273. Geological Survey of South Australia Bulletin 54. MCCUE K. 1996. Australian Seismological Report 1994. Australian Geological Survey Organisation Record 1996/019. MARDIA K. V. 1972. Statistics of Directional Data. Academic Press, New York. MASTIN L. 1988. Effect of borehole deviation on breakout orientations. Journal of Geophysical Research 93, 9187–9195. MILLIGAN P. R., PETKOVIC P. & DRUMMOND B. J. 2003. Potential-field datasets for the Australian region: their significance in mapping basement architecture. Geological Society of Australia Special Publication 22 and Geological Society of America Special Paper 372. MUELLER B., REINECKER J. & FUCHS K. 2000. The 2000 release of the World stress map. . REYNOLDS S. D., COBLENTZ D. D. & HILLIS R. R. 2003. Influences of plate-boundary forces on the regional intraplate stress field of continental Australia. Geological Society of Australia Special Publication 22 and Geological Society of America Special Paper 372, 59–70. REYNOLDS S. D. & HILLIS R. R. 2000. The in situ stress field of the Perth Basin, Australia. Geophysical Research Letters 27, 3421–3424. RICHARDSON R. M. 1992. Ridge forces, absolute plate motions, and the intraplate stress field. Journal of Geophysical Research 97, 11739–11748. WELLMAN P. 1998. Mapping of geophysical domains in the Australian continental crust using gravity and magnetic anomalies. In: Braun J., Dooley J., Goleby B., van der Hilst R. & Klootwijk C. eds. Structure and Evolution of the Australian Continent, pp. 59–71. American Geophysical Union Geodynamics Series 26. ZAJAC B. J. & STOCK J. M. 1997. Using borehole breakouts to constrain the complete stress tensor: results from the Sijan Deep Drilling Project and offshore Santa Maria Basin, California. Journal of Geophysical Research 102, 10083–10100. ZOBACK M. D. & ZOBACK M. L. 1991. Tectonic stress field of North America and relative plate motion. In: Slemmons D. B., Engdahl E. R., Zoback M. D. & Blackwell D. D. eds. Neotectonics of North America, pp. 339–366. Geological Society of America, The Geology of North America Decade Map Volume 1. ZOBACK M. L. 1992. First- and second-order patterns of stress in the lithosphere: the World stress map project. Journal of Geophysical Research 97, 11703–11728. ZOBACK M. L., ZOBACK M. D., ADAMS J., ASSUMPCAO M., BELL S., BERGMAN E. A., BLUMLING P., BRERETON N. R., DENHAM D., DING J., FUCHS K., GAY N., GREGERSEN S., GUPTA H. K., GVISHIANI A., JACOB K., KLEIN R., KNOLL P., MAGEE M., MERCIER J. L., MULLER B. C., PAQUIN C., RAJENDRAN K., STEPHANSSON O., SUAREZ G., SUTER M., UDIAS A., XU Z. H. & ZHIZHIN M. 1989. Global patterns of tectonic stress. Nature 341, 291–298. Received 8 August 2001; accepted 21 August 2002

Geol. Soc. Australia Spec. Publ. 22, and Geol. Soc. America Spec. Pap. 372 (2003), 59–70

Influences of plate-boundary forces on the regional intraplate stress field of continental Australia S. D. REYNOLDS1*, D. D. COBLENTZ2 AND R. R. HILLIS1 1 2

National Centre for Petroleum Geology and Geophysics, University of Adelaide, SA 5005, Australia Department of Geological Sciences, University of Oregon, Eugene OR 97403, USA. The Indo-Australian Plate is unique compared to other plates in that the maximum horizontal stress (Hmax) orientation is not uniform and does not parallel the direction of absolute plate velocity. Consequently, the Australian continent provides an ideal setting in which to study the interaction between tectonic forces and the intraplate stress field. Finite-element modelling of the intraplate stress field of the Indo-Australian Plate using a new ‘basis-set’ approach enables the evaluation of a very large number (several million) of boundary and potential energy force combinations acting on the plate. Constraint for the modelling is provided by an ‘observed’ regional stress field based on observations in 12 stress provinces, which is greatly enhanced from the observed data used in previous modelling efforts. The results of the present study indicate that modelling of the Australian intraplate stress field is inherently non-unique in that a large number of different boundary-force combinations can produce similar predicted stress fields. Nevertheless a number of fundamental conclusions may be drawn about the tectonic forces acting along the principal plate-boundary segments: (i) the Himalayan and New Guinean boundaries exert a compressional force on the IndoAustralian Plate producing a stress focusing normal to the boundaries and rotating between them; (ii) fitting the stress field in the Bowen Basin requires compressional boundary forces along the Solomon and New Hebrides subduction zones directed towards the interior of the Indo-Australian Plate; (iii) east–west compression in eastern Australia requires only a small compressional force along the Tonga–Kermadec subduction zone; and (iv) fitting the stress field in southeastern Australia (Otway Basin and Gippsland Basin stress provinces) requires compressional forces along the New Zealand and south of New Zealand boundary segments. The orientation of the modelled stress field over most of Australia is robust with northeastern and southern Australia the most sensitive to variations in the plate-boundary force combination. Furthermore, the modelling suggests that large sections of eastern Australia exhibit a relatively isotropic stress field compared to the rest of Australia. KEY WORDS: Australia, modelling, plate dynamics, stress.

INTRODUCTION Maximum horizontal stress (Hmax) orientations in continental Australia are variable and do not parallel the northnortheast direction of absolute plate motion (Zoback et al. 1989; Richardson 1992; Coblentz et al. 1995; Hillis & Reynolds 2000). In contrast, other continental areas such as Western Europe, South America and stable North America have relatively uniform Hmax orientations that do parallel the direction of absolute plate motion (Zoback et al. 1989; Richardson 1992; Gölke & Coblentz 1996). In these continental areas the alignment of Hmax orientations and the direction of absolute plate motion has led to the conclusion that plate-boundary forces (ridge push and basal shear) are the principal control on the character of the first-order intraplate stress field (Zoback et al. 1989; Richardson 1992; Coblentz & Richardson 1996; Gölke & Coblentz 1996). The lack of parallelism between Hmax orientations and the direction of absolute plate motion in Australia might lead to the conclusion that the plateboundary forces do not exert the same first-order control on the intraplate stress field. However, results from finite-element studies have shown that the Australian intraplate stress field can indeed be broadly accounted for in terms of

plate-boundary forces if the heterogeneous convergent northeastern boundary of the Indo-Australian Plate is recognised (Cloetingh & Wortel 1986; Coblentz et al. 1995, 1998). These studies focused on evaluating the relative contribution of topographic and plate-boundary forces on the intraplate stress field for a very limited number of models and used a qualitative comparison between the predicted and observed stresses. Furthermore, these studies failed to develop a satisfactory model of the entire intraplate stress field of continental Australia. The present study builds on these previous modelling studies by conducting a detailed investigation into the nature of the boundary forces along the northeastern margin of the Indo-Australian Plate. In the present study we use a new ‘basis-set’ approach to computing the predicted stress field. This approach allows the plate-boundary forces to vary over a range of forces and the intraplate stress fields predicted by a large combination of boundary forces can be assessed. Each predicted stress field is quantitatively compared with the observed stress field allowing determination of the optimal boundary force combination. The observed stress field used in this study is significantly refined from * Corresponding author: [email protected]

60

S. D. Reynolds, D. D. Coblentz and R. R. Hillis

Figure 1 The 12 stress provinces used to constrain the predicted stress field. The length of each bar is proportional to the weight used in the calculation of the misfit between modelled and observed stress orientations. The four stress provinces in New Guinea are shown for comparison, but were not used in testing the fit of modelled to observed data.

that used in previous studies, with 12 stress provinces used to constrain the predicted stress field. A detailed description of the current Australian stress data is presented by Hillis and Reynolds (2003). At the outset we emphasise that the primary aim of this study is to fit the regional intraplate stress field of continental Australia using the primary tectonic forces (e.g. boundary and topographic) acting on the plate. We acknowledge that the conclusions drawn from our modelling results, like those of the previous investigations, are limited by the inherent non-uniqueness of the problem in that a large number of different boundary-force combinations can produce a very similar predicted stress fields. Given this caveat, we feel that our refined approach to modelling the intraplate stresses has allowed us to make an important advance in understanding the complex nature of the Australian intraplate stress field.

REGIONAL AUSTRALIAN INTRAPLATE STRESS FIELD Inspection of the World stress map (Zoback 1992) clearly shows considerable scatter in Hmax orientation of the Australian intraplate stress field, particularly in comparison to other stable intraplate regions. This observed scatter has been reduced by the increase of A–C-quality data from 95 in the World stress map compilation to 331 in the current Australian stress map database (Hillis & Reynolds 2003). In addition the systematic variation of the stress field over the Australian continent has been further substantiated. A detailed description of the Australian stress field can be found in Hillis and Reynolds (2003). The current modelling study uses the mean Hmax orientation from 12 stress provinces in the Australian continent, as defined by Hillis and Reynolds (2003), to constrain the modelled stress field. The stress provinces define signif-

icant regions of the earth’s subsurface where the stress orientations are broadly consistent. Only 12 of the 16 stress provinces defined in Hillis and Reynolds (2003) are used to compare the modelled stress field (Figure 1). The four provinces that are not used are located at the collisional plate boundary in New Guinea and are characterised by Hmax orientations that are perpendicular to the plate boundary. Hence, Hmax orientations from these provinces are parallel to the force applied at the New Guinea boundary and thus, would provide only minor additional constraint on the modelled stress field. Information relating to stress magnitudes can be obtained from shallow hydraulic fracturing and overcoring data and from deeper focal mechanism solutions, but not generally from the breakouts and drilling-induced tensile fractures in petroleum wells that constitute 50% of the Australia stress map database (Hillis & Reynolds 2000). As a result there is significantly less information on stress magnitudes in the Australian stress map database than on stress orientations, with approximately half of the A–C-quality Hmax orientations having information on the stress regime. Of the data available, 60% indicate a thrust-fault stress regime, 27% indicate a strike-slip fault stress regime, 7% indicate a normal-fault stress regime and 6% indicate a mixed fault stress regime. Magnitude data are available for only around half of the stress provinces, and given the paucity of these data we used the available magnitudes only as a guide during the modelling. Furthermore, stress magnitudes (unlike stress orientations) appear to be sensitive to stress indicator type/depth (Hillis & Reynolds 2000) and thus exhibit a depth control that cannot be removed in the plate-scale modelling when the derived stresses are lithospheric averages. The reliability of the mean Hmax orientation in the 12 stress provinces varies depending on the number of measurements within the province and their standard deviation. As a consequence, each of the 12 observed stress orientations was given a weighting between 1 (maximum) and 0 (minimum) (Table 1). This weighting was used in calculating the fit of each model to the observed data and is a Table 1 Stress province data used to constrain the predicted stress field. Province

Latitude

Longitude

Hmax

Weight

orientation Amadeus Basin Northern Bonaparte Basin Southern Bonaparte Basin Bowen Basin Canning Basin Carnarvon Basin Cooper Basin Flinders Ranges Gippsland Basin Otway Basin Perth Sydney Basin

–23.50 –11.25

132.25 127.00

013 048

0.278 1.000

–14.75

129.00

057

0.350

–22.75 –17.50 –20.25 –27.75 –32.25 –38.50 –37.75 –31.00 –33.25

149.50 124.00 115.00 140.25 138.50 148.00 139.75 115.50 151.25

014 053 101 102 083 130 134 096 054

0.827 0.650 0.271 0.759 0.000 0.339 0.416 0.527 0.108

A weight of between 1 (highest) and 0 (lowest) was given to each province.

Modelling the Australian stress field

61

Figure 2 Indo-Australian Plate with the boundaries and forces used in the modelling. Orange arrows indicate mid-ocean ridge force, red and green arrows indicate plate-boundary forces determined from case 2 and 3 respectively and open arrows represent the force associated with continental margins. Note that during case 2 and 3 plate-boundary forces were allowed to apply compressional and tensional forces to the plate (only resultant forces are displayed). Also note that force arrows are not drawn to scale. Solid triangles indicate the direction of the subducting plate and open triangles indicates the Banda Arc. H, Himalaya; S, Sumatra Trench; J, Java Trench; B, Banda Arc; NG, New Guinea; SM, Solomon Trench; NH, New Hebrides; TK, Tonga–Kermadec Trench; NZ, New Zealand; SNZ, south of New Zealand; MOR, midocean ridge; cb, collisional boundary; sz, subduction zone; ia, island arc.

direct measure of how consistent the stress orientations are within each stress province. Each weight was determined – by computing the R value for the province and then sub– tracting the R value from the 90% confidence level used in the Rayleigh test (Davis 1986; Coblentz & Richardson 1995). The weights were then scaled so the maximum – weight applied was equal to 1. Any province with an R value less than the 90% confidence level was automatically given a weight of 0, because the data within that stress province was considered random. Thus the weights applied ranged between 1 and 0. Figure 1 displays the mean stress orientation in each province, with the length of the bar indicating the weighting applied. The Flinders Ranges stress province has been allocated a weight of 0, as it fails the Rayleigh test at the 90% confidence level and is represented by a circle on Figure 1.

MODELLING METHOD Predictions about the magnitude and orientation of the tectonic stresses in the Indo-Australian Plate were made using a two-dimensional elastic finite-element analysis. The finiteelement grid represents a portion of a spherical surface consisting of 2527 constant-strain triangular elements composed of a network of 1374 nodes. This grid provides a spatial resolution of about 2° in both latitude and longitude. The sensitivity of the modelled stresses was therefore limited to large-scale tectonic features with wavelengths of a few hundred kilometres. Stress magnitudes were calculated for a lithosphere of constant thickness, assumed to be 100 km, and stress concentrations may occur where variations in the lithospheric thickness are present. In addition, we ignored any bending-moment stresses associated with the subduct-

ing plates, and the predicted stresses cannot be regarded as significant for locations within the flexural wavelength of such boundaries. All elements were assigned a Poisson’s Ratio value of 0.25 (reflecting typical earth materials). In an effort to reflect the greater deformability of young oceanic lithosphere, the Young’s Modulus for ridge elements younger than about 70 Ma were assigned a value about half (3.5 x 1010 Nm–2) that of continental material (7.0 x 1010 Nm–2). Static equilibrium is assumed based on the fact that the plate is not accelerating. The use of an elastic rheology to model whole-plate deformation is an oversimplification, although it is justifiable for the modelling of first-order tectonic stresses. Alternative rheologies, such as viscoelastic, are useful for studying how tectonic stresses relax over time. A more detailed description of the finite-element modelling technique can be found in Richardson et al. (1979), Richardson and Reding (1991) and Coblentz et al. (1995). We have used three principal tectonic processes to model the forces acting on the Indo-Australian Plate: (i) ridge push; (ii) boundary tractions (representing both compressive collisional and slab pull forces); and (iii) buoyancy forces resulting from lithospheric density variations associated with continental margins, elevated continental crust and Lord Howe Rise (see discussions by Cloetingh & Wortel 1986; Richardson & Reding 1991). The various tectonic forces acting on the Indo-Australian Plate are illustrated in Figure 2. In order to ensure mechanical equilibrium, basal drag was applied only as needed to balance the net torque acting on the plate (as outlined by Richardson et al. 1979). In addition to the basic tectonic forces used in previous modelling studies (Cloetingh & Wortel 1986; Coblentz et al. 1995, 1998), we include the buoyancy force resulting from the lithospheric density variation associated with the Lord

62

S. D. Reynolds, D. D. Coblentz and R. R. Hillis

Table 2 Parameters used in case 2, the initial coarse case 3 run A and the second detailed case 3 run B. Boundary Himalaya New Guinea Sumatra Java Banda Solomon New Hebrides Tonga–Kermadec New Zealand South of New Zealand

Min

Case 2 Max

Step

Min

1 1 0 0 0 0 0 0 1 0

5 5 0 0 0 0 0 0 5 0

1 1 0 0 0 0 0 0 1 0

0 0 –6 –6 –6 –6 –6 –6 0 –6

Case 3 Run A Max Step 6 6 6 6 6 6 6 6 6 6

2 2 3 3 3 3 3 3 2 3

Min

Case 3 Run B Max Step

1 1 –3 –4 1 –3 –3 –4 1 0

5 5 3 0 3 3 3 4 4 4

1 1 1 1 1 1 1 1 1 1

Each boundary was allowed to vary between the maximum and minimum force magnitude. All forces are x1012 Nm–1.

Howe Rise and the force associated with the plate boundary south of New Zealand. Lord Howe Rise is a large segment of continental crust that separated from the eastern side of continental Australia at ca 95–83 Ma as a result of the opening of the Tasman Sea (Gaina et al. 1998). The plate boundary south of New Zealand is particularly complex consisting of the Puysegur Trench and the Macquarie Ridge Complex. These two additional forces have been included primarily to achieve a better fit to the observed stress field in eastern Australia. Here we have extended the finite-element modelling approach used in previous intraplate stress modelling studies (Richardson et al. 1979; Cloetingh & Wortel 1986; Richardson & Reding 1991; Coblentz et al. 1994, 1995, 1998) through the use of a ‘basis-set’ approach. The basisset approach works by exploiting the linearity of the purely elastic model used for the analysis. This method allows a way to expand the utility of the elastic approach to the problem by providing a means to explore the nature of the solution space in the vicinity of a ‘best-fitting’ model in a forward sense. In the basis-set approach, we calculated the predicted stresses and resistive basal drag required to balance the torque exerted on the plate for a force of magnitude 1.0 x 1012 Nm–1 acting along the individual boundaries (Table 3). The predicted stress field for various combinations of the tectonic forces is a weighted linear sum of this basis-set. Thus, various superpositions of the boundary forces acting on the plate can be evaluated in a matter of seconds rather than minutes as required by a new run of the finite-element model. This approach greatly increases the speed of the analysis and allows a detailed search for the best solution. For example, consideration of only three values for each of 20 tectonic forces would require approximately 109 finite-element runs, which in turn would require a prohibitive amount of computing time, despite the availability of fast computing platforms. An algorithm that computes the predicted stress through the addition of weighted linear combinations reduces the computation time required by several orders of magnitude. Evaluation of each model combination involves the determination of a misfit value for the predicted stress fields by averaging the difference between the 12 observed stress orientations and the closest modelled stress orientations. The difference between the observed and modelled stress orientations is multiplied by the weight given to the observed data, as

described above. Consequently, the models can be ranked according to individual misfit values. Topographic forces such as ridge push, elevated topography, continental margins and other lithospheric density anomalies remain unchanged throughout the model runs because they can be determined with relative confidence. The topographic forces range in magnitude from 2–3 x 1012 Nm–1 for ridge push, 1–2 x 1012 Nm–1 for continental margins and 6 x 1012 Nm–1 for continental lithosphere with an elevation of 5000 m (Lister 1975; Parsons & Richter 1980). Only plate-boundary forces at convergent plate margins and drag at the base of the plate were varied over the model runs. The magnitude of collisional forces at convergent plate boundaries and basal drag forces are poorly constrained with estimates varying by orders of magnitude (Solomon & Sleep 1974; Forsyth & Uyeda, 1975; Solomon et al. 1975). Hence, the collisional forces were allowed to vary between specified limits. In this study we use three different cases to demonstrate the fit that can be achieved by varying the tectonic forces along the northeastern margin of the Indo-Australian Plate. The first case uses solely the ridge push force with mechanical equilibrium achieved by homogeneously fixing the entire northeastern margin of the plate. This first case is equivalent to model 1 from Coblentz et al. (1995) and was used in this study as the basic model upon which the other two cases build. The second case consisted of the three collisional boundaries (Himalaya, New Guinea and New Zealand) applying an active force on the margin of the plate. The new basis-set approach was used in the second case to optimise the fit of the modelled stress field. The force on the collisional boundaries was allowed to vary between +1 and +5 x 1012 Nm–1. The third case in this study allowed all the boundaries surrounding the Indo-Australian Plate to apply a force. This case also used the new basis-set approach and required two separate sequences of model runs due to the large number of combinations. An initial coarse run (case 3 run A: Table 2) of 5 million models was undertaken to constrain a more detailed run (case 3 run B: Table 2). In the coarse run, forces on the 10 plate boundaries were allowed to vary between –6 and +6 x 1012 Nm–1, except for the three collisional boundaries, Himalaya, New Guinea and New Zealand, which were allowed to vary between 0 and +6 x 1012 Nm–1 (Table 2). This enabled the boundaries to range over values thought to be geologically plausible and also included values used in

Modelling the Australian stress field

63

Figure 3 Model 1 showing the predicted stress field from case 1 using the force from only the mid-ocean ridge with the northeastern plate boundary fixed. Table 3 Tectonic force magnitudes and relative torque contributions. Force

Magnitude x1012 Nm–1

Himalaya New Guinea Sumatra Java Banda Solomon New Hebrides Tonga–Kermadec New Zealand South of New Zealand Ridge Push Continental margins Elevated continent Lord Howe Rise

Total torque x1025 Nm

Latitude

Longitude

1 1 1 1 1 1 1 1 1 1

1.59 0.94 0.93 1.04 1.02 0.77 0.49 1.46 0.79 1.12

–10.4 –84.3 –19.0 –18.7 –15.7 79.9 73.0 –64.1 –48.0 –42.8

167.1 –37.3 –170.4 –156.3 –141.8 159.5 166.5 9.8 –6.8 –13.8

– –

8.30 2.19

40.7 –22.5

31.7 –179.2



3.66

–7.6

176.5



0.12

–24.1

14.2

boundary forces (1 x 1012 Nm–1, directed towards the plate interior) are listed in Table 3.

MODELLING RESULTS

Note that for the purposes of creating the basis set all boundary forces are directed towards the interior of the plate.

previous modelling studies (Cloetingh & Wortel 1986; Coblentz et al. 1995; Coblentz et al. 1998). The coarse run was required to reduce the number of boundary force combinations, so a more detailed evaluation of boundary forces could be conducted. The more detailed run (case 3 run B) of approximately 23 million models was based on the average obtained from the best 10% of models in the coarse run. The constraints placed on each boundary for the more detailed run can be seen in Table 2. The torque magnitudes and torque pole locations used for the basis-set of tectonic

Case 1: fixed northern and eastern boundaries In case 1, only the force from the mid-ocean ridge was applied, with mechanical equilibrium achieved by homogeneously fixing the entire northern and eastern margins of the plate. The force generated from the mid-ocean ridge was distributed throughout the oceanic lithosphere. Only one model was calculated since the ridge force can be determined with relative confidence. The modelled Hmax orientations calculated from the application of ridge push have a consistent north-northeast–south-southwest orientation and are shown in Figure 3. This model is termed model 1 for this study and is the same as that produced by Coblentz et al. (1995, 1998). In general, the model produces an extremely poor fit over most of the Australian continent with the exception of the Amadeus and Bowen Basin stress provinces. The predicted Hmax orientations parallel the direction of absolute plate velocity, but not the observed Hmax orientations in continental Australia. The plate-boundary force configuration in other continental areas such as Western Europe, stable North America and South America, where Hmax orientations parallel the direction of absolute plate velocity, are simpler than those acting on the Indo-Australian Plate and are somewhat similar to the situation modelled here.

Case 2: applied collisional zone forces In case 2, the basis-set approach was used, allowing the collisional boundaries of Himalaya, New Guinea and New Zealand to vary in force (Table 2). Displacements along all

64

S. D. Reynolds, D. D. Coblentz and R. R. Hillis

Table 4 Force magnitudes applied to the different boundaries for the best fitting model of case 2 (model 2) and case 3 (model 3). Boundary Himalaya New Guinea Sumatra Java Banda Solomon New Hebrides Tonga–Kermadec New Zealand South of New Zealand

Model 2

Model 3

Total torque x1025 Nm

Average of top 100

Stdev of top 100

Max of top 100

Min of top 100

2.0 2.0 0.0 0.0 0.0 0.0 0.0 0.0 5.0 0.0

2.2 1.6 1.6 –0.6 1.6 3.0 2.8 1.0 1.4 3.0

3.49 1.50 1.48 0.62 1.63 2.32 1.39 1.46 1.10 3.35

2.04 1.46 1.18 –1.29 1.37 2.96 2.54 0.98 1.64 3.39

0.942 0.558 1.598 1.200 0.485 0.197 0.642 0.141 0.718 0.764

1 1 –3 –4 1 2 1 0 1 1

5 3 3 0 2 3 3 1 4 4

Total torque magnitudes listed for each boundary using the forces from model 3. Also listed, are the average, standard deviation (stdev) and the maximum and minimum values from the top 100 best fitting models from case 3 run B.

other convergent plate boundaries were fixed. The boundary force combination with the smallest misfit between the predicted and observed stress field is shown in Figure 4. The force applied at each boundary is listed in Table 4. This model is considered the ‘best-fitting’ collisional boundary model and is termed model 2 for this study. The fit between the observed and modelled stress fields is very close in the western half of Australia. Also, the fit in the southeastern corner of Australia (Otway and Gippsland Basins) is good. However, the modelled stress field does not match the observed stress field in eastern Australia, particularly in the Cooper and Bowen Basins. This model is similar to model 3 of Coblentz et al. (1995). However, the basis-set approach has allowed its optimisation with respect to the fit to the observed data. It is clear that, as stated by Coblentz et al. (1995, 1998), much of the variation of the stress field of Australia can be explained by this model. Hence, by solely focusing the ridge-push force against the collisional boundaries requires no need to appeal to poorly constrained forces at the subduction zone boundaries. However, since the modelled stress field does not fit the observed stress field well in eastern Australia we undertook a third case of models, which allowed the forces along all components of the northeastern convergent boundary of the Indo-Australian Plate to vary.

Case 3: all boundary forces active Case 3 comprised an initial coarse run (A) to reduce the boundary force combinations so a more detailed run (B) could be made (Table 2). Examination of the top 10% of models from the coarse model run revealed a number of boundaries had distinct magnitude ranges. The best fit to the observed data was obtained if the Banda and south of New Zealand boundary segments applied a compressional force on the Indo-Australian Plate. Hence these boundaries were restricted to positive force magnitudes for the detailed model run. A tensional force applied at the Java subduction zone provided a good fit to the observed data. Consequently the Java boundary segment was restricted to tensional (negative) force magnitudes for the detailed model run. Results from the coarse run were unable to provide further restrictions on the forces acting at the other subduction zones, Sumatra, Solomon, New Hebrides and

Tonga–Kermadec. Thus, in the detailed model run these boundaries were allowed to range over both positive and negative force magnitudes. Himalaya, New Guinea and New Zealand collisional boundaries were all restricted to positive force values in the coarse model run. Consequently their force range was only refined by a small amount for the detail model run (Table 2). Results from case 3 run B indicated that averaging the force combination from a number of the best-fitting models significantly lowered the misfit value. The misfit value was minimised when the forces applied at each boundary from the five best-fitting models were averaged. Also, this boundary force ensemble is considered to be more tectonically plausible than the forces associated with the absolute lowest misfit value, since it reduces the forces acting along a number of boundaries to more closely match expected values. The average of the top five models is thus regarded as the ‘best-fitting’ model from the permutations considered in case 3 and is termed model 3 for this study (Table 4; Figure 5). This best-fitting model, while non-unique, can be considered to be a plausible representation of the tectonic forces acting on the Indo-Australian Plate, and the one that results in the intraplate stress field most consistent with the observed stress field shown in Figure 1. The torques resulting from the force on each boundary used in model 3 are listed in Table 4. The boundaries at Himalaya and south of New Zealand produce significantly more torque on the Indo-Australian Plate than the other boundaries. However, the combined torque of both these boundaries is less than that produced by ridge-push alone. Thus, the results from this modelling study are consistent with previous studies in that ridge push is the principal contributor to the intraplate stress field of the Indo-Australian Plate (Coblentz et al. 1995, 1998). The combination of forces from model 3 exerts a net torque 4.9 x 1025 Nm about a pole located at 47°S, 169°E. In order to ensure mechanical equilibrium of the plate, this torque has been balanced by a basal drag force acting along the base of the plate generated by plate motion about a pole located at 39.0°N, 140.9°E which yields a balancing drag torque of 4.9 x 1025 Nm about a pole located at 47°N, 11°W. This drag torque was calculated using a drag coefficient that generates 1 MPa of shear stress on an element for a linear velocity of 1 cm/y at the equator and a plate velocity of 2.90 cm/y. The balancing drag torque is roughly parallel to the

Modelling the Australian stress field

65

Figure 4 Model 2 showing the predicted stress field of the best-fitting model from case 2, where only the collisional boundaries were allowed to vary.

observed plate velocity direction in the western region of the plate and decreases to very small values beneath continental Australia. While we acknowledge the magnitude of the balancing drag torque is non-trivial (about half that generated by the ridge-push force), we interpret the balancing drag torque to be explainable in terms of a complicated flow pattern beneath the plate and do not venture a tectonic interpretation of its magnitude or orientation. This treatment of the basal drag force acting on the plate is justifiable given our modelling assumption that mechanical equilibrium can be ensured through the use of a balancing basal drag force. Information on the stress regime (stress magnitudes) was not included in the calculation of the misfit in this analysis. This was due to the lack of stress regime data available for all the stress provinces. However, in model 3, and most of the top models from case 3 run B, all stress provinces are characterised by thrust faulting conditions (Hmax>Hmin>v), which is broadly consistent with the limited stress regime information available. Consequently, including information on the stress regime would only have a very minor effect on the modelling. The stress field produced from the top 100 best-fitting models from case 3 run B predicts relatively low tectonic stress magnitudes for the Australian continent of between 20 and 45 MPa averaged over the thickness of the lithosphere (100 km). In contrast, modelling by Cloetingh and Wortel (1986) calculated tectonic stress magnitudes of over 100 MPa for the Australian continent. The lower tectonic stress magnitudes predicted here principally result from the inclusion of topographic forces. Stress trajectory mapping is a technique for identifying the regional stress field (Hansen & Mount 1990; Lee & Angelier 1994). Hillis and Reynolds (2003) have applied the stress trajectory mapping technique to the same stress data from which the stress provinces were calculated. It is not possible to use the stress trajectory map to constrain the modelling, although a comparison can be made with

the modelling results. The best-fitting model from case 3 and the stress trajectory map from Hillis and Reynolds (2003) are consistent over much of Australia (Figure 6). The stress trajectory map corresponds with the modelled stress field better than the stress province data used to constrain the model in a number of areas such as the Amadeus and Sydney Basins. However, some areas at the edges of the stress trajectory map fit poorly with the modelled stress field. This is due to the lack of stress data constraining the stress trajectories at the edges of the map.

MISFIT BETWEEN THE OBSERVED AND MODELLED STRESS ORIENTATIONS Western Australia The western part of Australia is characterised by an east–west Hmax orientation in the Carnarvon Basin and Perth region and a northeast–southwest Hmax orientation in the Canning and Bonaparte Basins. These Hmax orientations cannot be accounted for by ridge push alone as calculated in model 1 (Figure 3). The large misfit between the modelled and observed Hmax orientations for model 1 is shown in Figure 7. The regional observed stress field in the Perth region has a particularly poor match with the stress field produced by the ridge-push model, being almost perpendicular to the predicted stress orientation (Figure 3). Application of a compressional force on the Himalayan and New Guinean boundary segments in model 2 results in a close match between the observed and modelled Hmax orientations in the western part of Australia (Figure 4). The misfit in all five stress provinces has been significantly reduced from that of model 1 (Figure 7). Further reduction in the misfit is achieved when all of the boundaries actively apply a force in model 3 (Figure 5; Figure 7). However, the reduction

66

S. D. Reynolds, D. D. Coblentz and R. R. Hillis

Figure 5 Model 3 showing the predicted stress field of the best-fitting model from case 3 run B, where all boundaries were allowed to apply a force to the Indo-Australian Plate.

Figure 6 Model 3 overlain by the stress trajectory map of Hillis and Reynolds (2003).

in the misfit between model 2 and 3 for the western section of Australia is not considered significant. In the case of the south Bonaparte Basin stress province, the misfit actually increased slightly between models 2 and 3 (Figure 7). The results indicate that the inclusion of a tensional force along the Java subduction zone in model 3 only has a

minimal influence on the predicted sHmax orientations in the region. The subduction zone at Sumatra does not influence the stress field in the western part of Australia as witnessed by the large range of forces applied to the boundary in case 3 run B (Table 4). Furthermore, the forces applied along the subduction zone at Sumatra do not influence the

Modelling the Australian stress field

67

Figure 7 Misfit between the observed and predicted stress field from the bestfitting models of the three cases used in this study. Lines indicate average misfit for case 1, ridge-push model (orange); case 2, best-fitting collisional boundaries model (red); and case 3, best-fitting all boundaries model (green).

stress field predicted anywhere in Australia due to Australia’s position relative to this subduction zone. However, it should be noted that the subduction zone at Sumatra might still have a significant influence on the stress field in the Indian Ocean and continental India. For the purpose of this study comparisons were only made with observed stress data for Australia. Overall, the principle features of the observed stress field in the western part of Australia can be accounted for by application of a compressional force on the Himalayan and New Guinea boundary segments and without the requirement of forces along the subduction zone segments.

Central Australia The stress field in central Australia is characterised by two contrasting orientations. The Amadeus Basin exhibits a north–south Hmax orientation while the Cooper Basin exhibits a strongly developed east–west Hmax orientation (Figure 1). The predicted Hmax orientations from model 1 closely match the observed Hmax orientation in the Amadeus Basin, but are approximately perpendicular to the observed Hmax orientation in the Cooper Basin (Figure 3). Application of tectonic forces at only the collisional boundaries on the Indo-Australian Plate in model 2 cannot produce the north–south to east–west change in Hmax orientation seen in central Australia (Figure 4). Collisional boundary models can fit the observed stress field in the Amadeus Basin but not in the Cooper Basin. However, it should be noted that the Cooper Basin stress province has a more reliable regional Hmax orientation than the Amadeus Basin based on the consistency of observed stress data in the stress province. The predicted stress field in the Cooper Basin can be modelled to fit the observed if subduction zone boundaries are allowed to exert a force on the Indo-Australian Plate. However, the fit between modelled and observed stress fields in the Amadeus Basin must be compromised in order to fit the Cooper Basin (Figure 5). In order to fit the observed stress field in the Cooper Basin, the modelled Hmax orientation in the Amadeus Basin is rotated in an easterly direction. No plate boundary force model can successfully match the rotation between north–south Hmax in

the Amadeus Basin and east–west Hmax in the Cooper Basin which occurs over a short distance compared to plate boundary force induced rotations. Nevertheless, in models where the predicted Cooper Basin stress orientation matches the observed, the predicted northeast–southwest stress orientation in the Amadeus Basin corresponds very well with the stress trajectories for the region (Figure 6). Observed stress measurements in the Amadeus Basin may be influenced by local structure in the two thrust-related anticlines (Mereenie and Palm Valley Fields) from which much of the observed data for this region are sourced. In order to generate an east–west stress field in the Cooper Basin the Tonga–Kermadec subduction zone was required to exert a slight compressional force on the IndoAustralian Plate. Previous modelling by Cloetingh and Wortel (1986) applied a substantial tensional force along the Tonga–Kermadec subduction zone. However, this produces a north–south Hmax orientation through central and also eastern Australia, which is not consistent with the observed stress data. The results from the top 100 best-fitting models from case 3 run B show a consistent compression of around +1 x 1012 Nm–1 is required from the Tonga–Kermadec subduction zone to match the observed stress data (Table 4). Both the Solomon and New Hebrides subduction zones apply a substantial compressional force to the Indo-Australian Plate in the top 100 best-fitting models from case 3 run B. This force also constitutes towards an east–west Hmax orientation in the Cooper Basin by balancing the compression exerted from the southeastern boundaries (New Zealand and south of New Zealand).

Eastern Australia The stress field of eastern Australia is characterised by two main trends. Northeastern Australia exhibits of a strong north-northeast–south-southwest Hmax orientation in the Bowen Basin, while southeastern Australia exhibits a consistent northwest–southeast Hmax orientation in the Otway and Gippsland Basins. The Sydney Basin, located midway between these two areas, is characterised by northwest–southeast Hmax orientation. However, the Sydney Basin stress province was given a very low weighting for the

68

S. D. Reynolds, D. D. Coblentz and R. R. Hillis

Figure 8 Circular standard deviation of the top 100 best-fitting models from case 3 run B.

Figure 9 Average differential horizontal stress field (Hmax magnitude minus Hmin magnitude) of the top 100 best-fitting models from case 3 run B.

misfit calculations due to the highly scattered nature of the stress field (Hillis et al. 1999). The consistent north-northeast–south-southwest stress field predicted by model 1 corresponds poorly to the highly variable observed stress field in eastern Australia (Figure 3). Only the observed stress field in the Bowen Basin provides an adequate match for the predicted stress field from model 1. In model 2 the application of a force at the collisional boundaries also cannot match the observed stress field throughout eastern Australia (Figure 4). If compressional forces are applied at the collisional boundaries, predicted stresses match the observed northwest–southeast Hmax orientation in both the Otway and Gippsland Basins. This is a result of the New Zealand boundary segment exerting a substantial compressional force on the Indo-Australian Plate (Table 4). Nonetheless, the modelled stress field in the Sydney and Bowen Basins does not match the observed. In the Sydney Basin the modelled Hmax orientation using only collisional boundary forces is approximately perpendicular to the observed Hmax orientation determined using the stress province data. However, the observed stress data in the Sydney Basin are scattered and stress trajectories for the region are approximately perpendicular to the stress province orientation (Figure 6). The modelled stress field using only collisional boundaries provides a good fit with the stress trajectories for the Sydney Basin, and the misfit to the stress province data for the Sydney Basin is not seen as significant. The modelled and observed Hmax orientations in the Bowen Basin differ by approximately 45° when applying forces only at the collisional boundaries (Figure 4). This is a result of the Hmax orientation rotating between the New Zealand and New Guinea boundaries. Thus, the observed stress field in the Bowen Basin cannot be modelled using forces applied at only the collisional boundaries. The best-fitting model from case 3 provides an adequate fit over much of eastern Australia (Figure 5). To successfully

model the observed stress field in the Bowen Basin requires the Solomon and New Hebrides subduction zones to exert a compressional force on the Indo-Australian Plate. This compression has the effect of rotating the Hmax orientation in the Bowen Basin in an easterly direction from that generated by the collisional boundaries in case 2. Results from the top 100 best-fitting models from case 3 run B indicate that both the Solomon and New Hebrides subduction zones are required to consistently exert a substantial compressional force on the Indo-Australian Plate (Table 4). The observed northwest–southeast Hmax orientation in both the Otway and Gippsland Basins is successfully modelled by exerting a compressional force along the New Zealand and south of New Zealand boundary segments (Figure 4; Table 4). A substantial force was required on the boundary south of New Zealand in order to counteract the force applied by the northern boundaries of Solomon and New Hebrides. The modelled northwest–southeast Hmax orientation in the Sydney Basin from the best-fitting model in case 3 was the same orientation as that produced using just the collisional boundaries in case 2 and hence does not match the stress province data, but does match the stress trajectory data.

VARIABILITY IN THE MODELLED STRESS FIELD Despite using a greatly improved observed-stress dataset, the results of the modelling herein demonstrate that a significant amount of ambiguity remains when modelling the plate-boundary forces responsible for the stress field of continental Australia. The top 100 best-fitting models from case 3 run B predict stress orientations across the continent that are visually similar, but result from substantially different forces on certain boundaries (Table 4). Thus, absolute force magnitudes acting on particular boundary segments cannot be constrained. Nonetheless, broad conclusions

Modelling the Australian stress field

have still been determined about the nature of the force at particular boundary segments from this study. The circular standard deviation was calculated for the top 100 best-fitting models from case 3 run B in order to quantify the variation in predicted stress orientations (Figure 8). Predicted stress orientations over much of the Australian continent have a very low circular standard deviation of less than 10°. Nevertheless, the circular standard deviation has highlighted two areas where the predicted stress orientations are extremely variable (areas A and B in Figure 8). These areas of high circular standard deviation are a result of the low differential horizontal stress magnitude predicted in both areas (Figure 9). Hence, the predicted tectonic stress field in both these areas is relatively isotropic and is highly sensitive to changes in the forces applied to the plate boundaries. If the model predictions are correct and these areas exhibit low tectonic stress anisotropy, observed stress orientations in these areas are likely to be dominated by local stress sources and thus highly variable. However, this statement cannot be assessed due to the lack of observed stress data in these areas. The northeastern area (A) holds the greatest potential to further constrain the plate-boundary forces along the northeastern boundary. The stress field throughout the western part of Australia is consistent in the top 100 best-fitting models from case 3 run B (Figure 8). The general consistency between the observed and modelled stress fields for the western region indicates the robustness of the tectonic models as a result of the compressional forces applied at the Himalayan and New Guinean boundaries throughout all models. The tectonic forces generate a stress field with high horizontal anisotropy in the western region (Figure 9). Thus, local sources of stress do not apparently influence the horizontal stress orientations. This explains the consistent observed Hmax orientations in a number of stress provinces in the western region (Figure 1). The relatively small variation in modelled Hmax orientation for the Sydney Basin for the top 100 models from case 3 run B suggests that the modelled orientation of the tectonic stress field is robust in the area (Figure 8). The average difference between the two tectonic horizontal stress magnitudes is approximately 7 MPa (Figure 9). Thus, even though the modelled stress orientations are consistent in the Sydney Basin, the small difference between the two tectonic horizontal stress magnitudes allows local sources of stress to play a critical role in the province. As previously mentioned, the observed stress field in the Sydney Basin is reasonably scattered, suggesting that local sources of stress such as those due to topography may dominate the stress field of the Sydney Basin. The fact that the predicted stress orientation in the Sydney Basin is consistently modelled perpendicular to the stress province direction but parallel to stress trajectories further suggests that local sources of stress strongly influence stress orientations in the Sydney Basin.

NATURE OF PLATE-BOUNDARY FORCES Results from this study indicate a number of interesting conclusions can be made about the nature of the plateboundary forces despite significant ambiguity still existing

69

regarding the absolute magnitude of forces along most plate boundaries. For example the subduction geometry along the Java, Sumatra, Solomon and New Hebrides trenches (with the Indo-Australian Plate subducting under the adjacent overriding plate) might suggest that tensional slab-pull forces act on the Indo-Australian Plate along these boundaries. However, our modelling results predict that tensional slab-pull forces exist only along the Java subduction zone. Along the Tonga–Kermadec subduction zone the Indo-Australian Plate is overriding the Pacific plate with backarc spreading occurring on the IndoAustralian Plate side, suggesting the existence of tensional forces related to the Tonga–Kermadec subduction zone. Since the modelling results predict a moderate compressional force along this boundary, we conclude that tensional forces associated with backarc spreading are not transmitted into the plate interior. Thus, the nature of the forces predicted along the subduction zones surrounding the Indo-Australian Plate suggest that the net slab-pull force is being reduced by other forces such as slab resistance and as consequence slab-pull exerts only a secondorder control on the stress field of continental Australia. The boundary south of New Zealand, which consists of the Puysegur Trench and the Macquarie Ridge Complex (Sutherland 1995) was required to exert a substantially greater force on the Indo-Australian Plate than was initially expected. The total torque produced by this boundary on the Indo-Australian Plate was approximately the same magnitude as the total torque produced by the Himalayan boundary (Table 4). A large compressional force on the boundary south of New Zealand is supported by other geophysical data, with a number of studies proposing that the area is experiencing regional deformation (DeMets et al. 1988; Valenzuela & Wysession 1993; Spitzak & DeMets 1996). The southern Tasman Sea, an area to the west of the Macquarie Ridge Complex, appears to be undergoing significant regional deformation as suggested by the diffuse intraplate seismicity and the presence of three parallel southeast-trending gravity undulations (Valenzuela & Wysession 1993). The tectonics of the area has been likened to the internally deforming area of the Indian Ocean (Valenzuela & Wysession 1993). Thus, we feel that a large force on the boundary south of New Zealand is tectonically plausible given the agreement between the predicted and observed stress data used in this study as well as other geophysical data for the region.

CONCLUSIONS The observed regional stress field in continental Australia, as defined by 12 stress provinces, cannot be modelled using the force generated from the mid-ocean ridge alone. However, applying a force on the collisional boundaries of Himalaya, New Guinea and New Zealand produces a modelled stress field that corresponds well to all but two of the stress provinces (Cooper and Bowen Basins). Further refinement of the fit between the modelled and observed stress field is achieved by using a combination of principal tectonic forces from all plate boundary segments along the Indo-Australian Plate.

70

S. D. Reynolds, D. D. Coblentz and R. R. Hillis

This study finds, as discussed in previous investigations, that a number of different boundary force combinations can produce similar predicted stress fields and thus it is not possible to uniquely define the plate-boundary forces acting on the Indo-Australian Plate by modelling the intraplate stress field. Despite this limitation, results indicate the following key points: (i) compressional forces act along the Himalaya and New Guinea boundaries to produce stress-focusing normal to those boundaries and stress rotation between them; (ii) the observed stress field in the Bowen Basin requires compressional forces along the Solomon and New Hebrides subduction zones; (iii) east–west compression in areas of central Australia requires moderate compression acting along the Tonga–Kermadec subduction zone: the fit between the predicted and observed stress field degrades with the application of extensional force along this boundary; (iv) modelling stresses in southeastern Australia (specifically in the Otway and Gippsland Basin stress provinces) requires compressional forces along the New Zealand and south of New Zealand boundary segments; and (v) extensional forces act along the Java subduction zone: the Australian stress field is not sensitive to the forces acting along the Sumatran boundary segment. The orientation of the modelled stress field over most of Australia is robust for a number of the best-fitting models, with predicted stresses in northeastern and southern Australia the most sensitive to variations in the plateboundary force combination. Additional data in northeastern Australia has the greatest potential to further constrain the plate-boundary forces acting along the northeastern plate boundary.

ACKNOWLEDGEMENTS We wish to thank P. Meijer and R. Albert for their critical reviews and comments on the manuscript.

REFERENCES CLOETINGH S. & WORTEL R. 1986. Stress in the Indo-Australia plate. Tectonophysics 132, 49–67. COBLENTZ D. D. & RICHARDSON R. M. 1995. Statistical trends in the intraplate stress field. Journal of Geophysical Research 100, 20245–20255. COBLENTZ D. D. & RICHARDSON R. M. 1996. Analysis of the South American intraplate stress field. Journal of Geophysical Research 101, 8643–8657. COBLENTZ D. D., RICHARDSON R. M. & SANDIFORD M. 1994. On the gravitational potential of the Earth’s lithosphere. Tectonics 13, 929–945. COBLENTZ D. D., SANDIFORD M., RICHARDSON R. M., ZHOU S. & HILLIS R. R. 1995. The origins of the intraplate stress field in continental Australia. Earth and Planetary Science Letters 133, 299–309. COBLENTZ D. D., ZHOU S., HILLIS R. R., RICHARDSON R. M. & SANDIFORD M. 1998. Topography, boundary forces, and the Indo-Australian intraplate stress field. Journal of Geophysical Research 103, 919–931.

DAVIS J. C. 1986. Statistics and Data Analysis in Geology. John Wiley, New York. DEMETS C., GORDON R. G. & ARGUS D. F. 1988. Intraplate deformation and closure of the Australia–Antarctica–Africa plate circuit. Journal of Geophysical Research 93, 11877–11897. FORSYTH D. & UYEDA S. 1975. On the relative importance of the driving forces of plate motion. Geophysical Journal of the Royal Astronomical Society 43, 163–200. GAINA C., MÜLLER R. D., ROYER J-Y., STOCK J., HARDEBECK J. & SYMONDS P. 1998. The tectonic history of the Tasman Sea: a puzzle with 13 pieces. Journal of Geophysical Research 103, 12413–12433. GÖLKE M. & COBLENTZ D. 1996. Origins of the European regional stress field. Tectonophysics 266, 11–24. HANSEN K. M. & MOUNT V. S. 1990. Smoothing and extrapolation of crustal stress orientation measurements. Journal of Geophysical Research 95, 1155–1165. HILLIS R. R., ENEVER J. R. & REYNOLDS S. D. 1999. in situ stress field of eastern Australia. Australian Journal of Earth Sciences 46, 813–825. HILLIS R. R. & REYNOLDS S. D. 2000. The Australian Stress Map. Journal of the Geological Society of London 157, 915–921. HILLIS R. R. & REYNOLDS S. D. 2003. in situ stress field of Australia. Geological Society of Australia Special Publication 22 and Geological Society of America Special Paper 372, 49–58. LEE J. & ANGELIER J. 1994. Paleostress trajectory maps based on the results of local determinations: the ‘Lissage’ program. Computers and Geoscience 20, 161–191. LISTER C. R. 1975. Gravitational drive on oceanic plates caused by thermal contraction. Nature 257, 663–665. PARSONS B. & RICHTER F. M. 1980. A relation between driving force and geoid anomaly associated with the mid-ocean ridges. Earth and Planetary Science Letters 51, 445–450. RICHARDSON R. M. 1992. Ridge forces, absolute plate motions, and the intraplate stress field. Journal of Geophysical Research 97, 11739–11748. RICHARDSON R. M. & Reding L. M. 1991. North American plate dynamics. Journal of Geophysical Research 96, 12201–12223. RICHARDSON R. M., SOLOMON S. C. & SLEEP N. H. 1979. Tectonic stress in the plates. Reviews of Geophysics and Space Physics 17, 981–1019. SOLOMON S. C. & SLEEP N. H. 1974. Some simple physical models for absolute plate motions. Journal of Geophysical Research 79, 2557–2567. SOLOMON S. C., SLEEP N. H. & RICHARDSON R. M. 1975. On the forces driving plate tectonics: inferences from absolute plate velocities and intraplate stresses. Geophysical Journal of the Royal Astronomical Society 42, 769–801. SPITZAK S. & DEMETS C. 1996. Constraints on present-day motions south of 30°S from satellite altimetry. Tectonophysics 253, 167–208. SUTHERLAND R. 1995. The Australia–Pacific boundary and Cenozoic plate motions in the SW Pacific: some constraints from Geosat data. Tectonics 14, 819–831. VALENZUELA R. W. & WYSESSION M. E. 1993. Intraplate earthquakes in the southwest Pacific Ocean Basin and the seismotectonics of the southern Tasman Sea. Geophysical Research Letters 20, 2467–2470. ZOBACK M. L. 1992. First- and second-order patterns of stress in the lithosphere: the World stress map project. Journal of Geophysical Research 97, 11703–11728. ZOBACK M. L., ZOBACK M. D., ADAMS J., ASSUMPCAO M., BELL S., BERGMAN E. A., BLUMLING P., BRERETON N. R., DENHAM D., DING J., FUCHS K., GAY N., GREGERSEN S., GUPTA H. K., GVISHIANI A., JACOB K., KLEIN R., KNOLL P., MAGEE M., MERCIER J. L., MULLER B. C., PAQUIN C., RAJENDRAN K., STEPHANSSON O., SUAREZ G., SUTER M., UDIAS A., XU Z. H. & ZHIZHIN M. 1989. Global patterns of tectonic stress. Nature 341, 291–298. Received 27 August 2001; accepted 2 May 2002

Geol. Soc. Australia Spec. Publ. 22, and Geol. Soc. America Spec. Pap. 372 (2003), 71–89

Three-dimensional finite-element modelling of the tectonic stress field in continental Australia S. ZHAO1, 2 AND R. D. MÜLLER1* 1 2

School of Geosciences, University of Sydney, NSW 2006, Australia. Japan Marine Science and Technology Center, Yokosuka 237-0061, Japan. Traditionally, intraplate stress orientations have been modelled using an isotropic elastic plate. For the Australian Plate this method has been applied successfully to model the first-order pattern of stress orientations. However, the distribution of intraplate earthquakes and the juxtaposition of strong, cold with hotter, younger lithosphere in many areas suggest that the spatial variation in mechanical strength of the plate may result in substantial regional anomalies in stress orientations and magnitudes. We explore this idea with a three-dimensional finite-element model to investigate the regional response of the Australian continent to tectonic forces. The model covers the area of –40 to –10° (S) and 111 to 155° (E) with a spatial resolution of 90 x 90 x 50 km. The relative magnitudes of the ridge-push and boundary forces, which act on the Australian continent, are estimated through an inversion analysis of in situ stress data. The differences between modelled and observed stress orientations are minimised in a least-squares sense. Major tectonic blocks and the differences in their elastic strength are included in the model, and the initial estimates of Young’s modulus for the tectonic blocks are adapted from a published coherence analysis of gravity and topographic data. The values of Young’s modulus are adjusted in the inversion analysis to best fit the stress orientations observed on the Australian continent. The inversion analysis of rheological parameters is most efficient for estimating Young’s modulus for the Northern Lachlan Fold Belt, the New England Fold Belt, and the Southern Lachlan Fold Belt. The adjusted values for the flexural rigidity are 0.040 x 1025 Nm for the Northern Lachlan Fold Belt, 0.037 x 1025 Nm for the New England Fold Belt, and 0.040 x 1025 Nm for the Southern Lachlan Fold Belt, which correspond to an effective elastic thickness of about 30 km. Based on the optimised body and boundary forces acting on the plate, a map of maximum principal-stress distribution is constructed so that variations of the relative magnitude of tectonic stresses can be assessed. We find a good match between predicted zones of stress concentration and the distribution of major belts of seismicity in Australia. The results show that while the overall pattern of stress orientations in the Australian continent is controlled by the forces which drive the Indo-Australian Plate, the maximum horizontal stress orientations and the pattern of the stress concentration manifested by seismicity are modulated by local/regional geological structures. KEY WORDS: Australia, crustal plates, in situ stress, numerical modelling, seismicity.

INTRODUCTION Studies of intraplate stresses show that most intraplate regions are characterised by compressive stress regimes and crustal seismic activity (Zoback et al. 1989) and that the inferred maximum horizontal principal-stress orientations are roughly parallel to the directions of ridge-push forces. On the basis of the analysis of earthquake focal mechanisms, in situ stress measurements and surface deformation, Denham et al. (1979) have shown that the Australian continent is in a state of substantial horizontal compression. However, the maximum horizontal compression measured in situ and inferred from seismic source mechanism solutions is approximately normal to the eastern passive margin (Denham et al. 1979). This is not in agreement with the absolute velocity trajectories of the Australian Plate (Richardson 1992; Zoback 1992). Numerical modelling of the tectonic forces applied to the Indo-Australian Plate (Figure 1) has been performed with the finite-element method. Cloetingh and Wortel (1986) investigated the tectonic stress field in the Indo-

Australian Plate with a two-dimensional finite-element model. Five types of tectonic forces were included in their analysis: slab pull, ridge push, resistant force, trench suction force and drag force. The combination of the forces in their model resulted in a concentration of the compressive stresses of the order of 300–500 MPa in some parts of the plate (e.g. the Ninetyeast Ridge). Using a similar 2D finiteelement model, Coblentz et al. (1998) reinvestigated the tectonic forces acting on the Indo-Australian Plate. The stress indicators from the World Stress Map Project (Zoback 1992) were used to constrain their numerical models. They found that: (i) the ridge-push force is likely the primary force in controlling the first-order stress pattern in the Indo-Australian Plate; (ii) if imposing resistance along the Himalaya, Papua New Guinea and New Zealand collisional boundaries to balance the ridge-push force, then many of the first-order stress patterns of the observed stress field can be explained without including either subduction or basal-drag forces; and (iii) the observed maximum hori* Corresponding author: [email protected]

72

S. Zhao and R. D. Müller

Figure 1 Map showing the location of the Australian continent and the geometry of the Indo-Australian Plate boundaries (inset). MOR, Mid-Ocean Ridge; J, Java Trench; B, Banda Arc; PNG, Papua New Guinea; SM, Solomon Trench. Arrows indicate the directions of the major tectonic forces (not to scale). FL, the boundary force from the west; FR, the boundary force from the east; FC, the collision and subduction-related forces from the north; FP, the ridge-push force from the south. Inset (bold lines are plate boundaries): EU, Eurasian Plate; PA, Pacific Plate; S, Sumatra Trench; NH, New Hebrides; TK, Tonga–Kermadec Trench; NZ, New Zealand.

zontal-stress orientations and stress-regime information in the Indo-Australian Plate can also be explained with the models predicting low tectonic-stress magnitudes (e.g. tens of megapascals averaged over the thickness of the lithosphere). This implies that the large stress magnitude (hundreds of megapascals) inferred by Cloetingh and Wortel (1986) for some parts of the Indo-Australian Plate is not required to explain the observed stress orientations and regime information. Coblentz et al. (1995) also discussed the origins of the intraplate stress field in continental Australia and suggested that stress focusing effects along the heterogeneous convergent boundaries (implemented by fixing the boundaries) are necessary to produce the significant compression within the continent. In addition, Zhang et al. (1996) constructed a 2D finite-element model for part of the east Australian passive margin, although lateral stress changes could not be fully investigated because of the nature of the 2D elastic model, and it could not be used to interpret the stress orientations observed in the Australian continent. The main tectonic forces acting on the Australian continent that have been identified in previous studies are shown in Figure 1. The ridge-push force from the mid-oceanic ridge was inferred to be the dominant force of driving the IndoAustralian Plate northwards (Coblentz et al. 1998). In the east, the Australian continent may be affected by the subduction of the Pacific Plate near New Zealand. The force transferred from the subduction zone, which is over 3000 km away from the Australian continent, was considered to have only a secondary effect (Coblentz et al. 1998). In the north, the boundary between the Indo-Australian, Eurasian and Pacific Plates is very complex. At the Java Trench, the Australian Plate is subducting beneath the Eurasian Plate, while the Eurasian Plate is subducting under the Australian

Plate at the Banda Arc. At the Solomon Trench, the Australian Plate is subducting beneath the Pacific Plate, while the Pacific Plate is subducting beneath the Australian Plate at Papua New Guinea. Although the results from the previous studies have greatly improved our understanding of the relative magnitudes of the tectonic forces and their role in controlling the tectonic stress field in the Australian continent, there are several important factors that could not be addressed by previously applied methodologies, (1) The Australian continent is assumed to be homogeneously rigid in most of the previous models. Recent studies on lithospheric structures from a seismic tomographic analysis of the Australian continent reveal that the lithosphere is not homogeneous in elastic strength, but heterogeneous (Simons et al. 1999; Simons & van der Hilst 2002). Many surface geological structures (e.g. cratons and basins) in the Australian continent have extensions in the upper mantle in terms of seismic velocity anomalies (Kennett 1997). The formation of the elastically/seismically inhomogeneous structures is closely associated with the stress-evolution process in the lithosphere. Since some of the geological structures in the Australian continent are quite large, up to a width of 800 km for some cratons, their effect on spatial changes in stress orientations and magnitude is probably significant. These elastically inhomogeneous structures were not considered in any previous stress modelling for the Australian continent. (2) Two-dimensional finite-element modelling was employed in all of the previous models. In the 2D models, vertical stresses have been ignored due to their plane stress approximation used in these studies. (3) Crustal seismicity has occurred throughout the Australian continent, and the origin of these intraplate earthquakes are still puzzling (Denham 1988; Denham &

3D Modelling of Australian stress field

73

Figure 2 Map showing the main geological blocks in continental Australia. YB, Yilgarn Block; MUB, Musgrave Block; GB, Gawler Block; ARB, Arunta Block; MB, Mt. Isa Block; Hodg. F. B., Hodgkinson Fold Belt; N. LA. F. B., Northern Lachlan Fold Belt; New Eng. F. B., New England Fold Belt; S. LA. F. B., Southern Lachlan Fold Belt; Adel. F. B., Adelaide Fold Belt. B5, B11, B9 and B13 are isolated blocks (see Table 1).

Windsor 1991). Due to the limitations inherent in the previous 2D finite-element models, the spatial distribution of the tectonic stress in the Australian continent, as well as the origins of the intraplate seismicity, have not been investigated. (4) A trial-and-error method was used in the previous modelling studies and the calculated stress orientations were visually compared with the observed stress orientations, which is not technically efficient. The studies were mainly qualitative or semiquantitative so that a formal fit between the observed and modelled stress orientations could not be achieved. While recent improvement has been made through some refined strategies in forward stress modelling (Reynolds et al. 2003), an inverse approach for estimating model parameters from observed stress orientations would improve our ability to find the best-fit model. The present study is an extension of previous modelling efforts of the tectonic stress field in continental Australia (Coblentz et al. 1995). Our work differs from the previous studies mainly in the following aspects: (i) it is focused on continental Australia, and the area covered by the model is about 45 x 31° (in longitude and latitude) with a spatial resolution of about 90 x 90 x 50 km; (ii) heterogeneous features of the Australian continent are included in our analysis, which are associated with the differences in the elastic strength for different geological domains, such as major tectonic provinces and fold belts; and (iii) since a wide range of boundary conditions can be configured to match the observed intraplate stress field, the non-uniqueness of the problem is investigated. In this study, an inversion method is used to estimate the relative magnitudes and directions of the tectonic forces associated with the Australian continent from stress orientation data. Previous studies demonstrated that the gravitational potential energy difference across the boundary between continental and oceanic crust may significantly affect the regional stress field at a plate scale (Coblentz et al. 1994;

Sandiford et al. 1995). In this study, the contribution of topography and gravity potential energy differences to a local/regional stress field is not simulated, partly because a quantitative simulation requires a detailed crustal/lithospheric (density) structure model, which is not presently available. Ignoring the effect of the gravity potential energy differences in the Australian continent will affect our results, especially at the continental margin, and will be discussed later.

TECTONIC BLOCKS AND FLEXURAL RIGIDITY OF THE AUSTRALIAN CONTINENT The Australian continent, which is geologically and tectonically complex, can be divided into several crustal blocks. Each block has its own distinctive tectonic style and represents a significant stage in the evolution of the Australian continent (Plumb 1979a, b). Figure 2 shows the main tectonic units in the Australian continent. The yellow areas are the cratons, which are geologically stable and the dark green areas are fold belts. The blank area in the continent is composed of basins and smaller blocks, which will be treated indiscriminately as continental crust/lithosphere, and assumed to have a mechanical strength less than that of the cratons and larger than that of the fold belts for our modelling. The flexural rigidity of the tectonic blocks in the Australian continent has been investigated by Zuber et al. (1989) and Simons et al. (2000) on the basis of the analysis of Bouguer gravity and topography data. The flexural rigidity values estimated by Zuber et al. (1989) for the New England Fold Belt and Southern Lachlan Fold Belt (~ 1022 N m) are about three orders of magnitude lower than those (~ 1025 N m) of cratons. A revised estimate of the effective elastic thickness for central Australia is about a factor of two less than that of Zuber et al. (1989) (Simons et al. 2000). This suggests that there are large uncertainties

74

S. Zhao and R. D. Müller

Table 1 Flexural rigidity and Young’s Modulus for the major tectonic blocks in the Australian continent. No. 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17

Flexural rigiditya (x 1025 Nm) Basins – Adelaide Fold Belt – Yilgarn Block 2.0 Pilbara Block 2.0 Block 5 (B5) 2.1 Arunta Block 0.61 Musgrave Block 0.61 Gawler Block 0.69 Block 9 (B9) 2.1 Mt. Isa Block 2.1 Block 11 (B11) 0.69 Southern Lachlan Fold Belt 0.0011 Block 13 2.0 Hodgkinson Fold Belt & 0.0016 Northern Lachlan Fold Belt New England Fold Belt 0.0036 Oceanic crust – Continental shelf – Type

Scalingb – – 0.95 0.95 1.0 0.29 0.29 0.33 1.0 1.0 0.33 0.00052 0.95 0.00076 0.0017 – –

Young’s modulus (x 1010 Pa) Estimatedc Adjustedd 3.0 – 0.113 – 5.7 – 5.7 – 6.0 6.656 1.74 1.970 1.74 – 1.98 – 6.0 – 6.0 – 1.98 – 0.0031 0.113 5.7 – 0.0046 0.113 0.01 5.7 0.21

0.106 – –

Refined rigidityd (x 1025 Nm) – 0.040 – – 2.330 0.690 – – – – – 0.040 – 0.040 0.037 – –

a

From Zuber et al. 1989. Obtained after dividing the flexural rigidity values by the maximum flexural rigidity value (2.1 x 1025 Nm). c Assign a Young’s modulus value of 6 x 1010 Pa for Blocks 5, 9 and 10, which have the maximum flexural rigidity and then estimate the Young’s modulus values of other blocks based on the scaling factor. d Estimated from the inversion analysis. b

in the estimated effective elastic thickness and the flexural rigidity of the tectonic blocks in Australia, depending on the method used. In this study, we are mainly concerned with the relative values of rigidity among the tectonic blocks; uncertainties in the absolute values of the flexural rigidity will not affect our results in terms of stress orientations. We use the relative values of the rigidity of the tectonic blocks estimated by Zuber et al. (1989) to scale the rheological parameters associated with the elastic strength and as initial input for our model. In addition, the values of rheological parameters will be adjusted in the inversion analysis of the observed stress orientations. The age and estimated flexural rigidity for the main tectonic blocks used here are outlined below.

Yilgarn and Pilbara Blocks The Yilgarn and Pilbara Blocks in Western Australia were formed about 3500–3100 Ma. This region is geologically stable (Plumb 1979a). The largest Moho depth here is estimated to be ~50 km (Clitheroe et al. 2000). The flexural rigidity for this area was estimated to be 2.0 x 1025 N m (Zuber et al. 1989).

Arunta and Musgrave Blocks The Arunta and Musgrave Blocks are located in central Australia, which consists of heavily faulted Proterozoic blocks and basins. Moho offsets with amplitude variations more than 20 km have been inferred from gravity modelling (Lambeck 1983a) and the analysis of seismic travel time anomalies (Lambeck 1983b; Lambeck & Penney 1984). The flexural rigidity for this region is estimated to be 6.1 x 1024 N m (Zuber et al. 1989).

Gawler Block The Gawler Block in South Australia was formed during the Late Archaean to Palaeoproterozoic (Plumb 1979a). The flexural rigidity for the block is estimated to be 6.9 x 1024 N m (Zuber et al. 1989) and the Moho depth is about 40 km (Clitheroe et al. 2000).

Mt Isa Block and Northern Craton To the north of the Arunta Block, the entire region (including B5 and B9 in Figure 2), containing the Mt. Isa Block, is simply called the Northern Craton (Zuber et al. 1989; Plumb 1979b) and consists of Palaeoproterozoic blocks bounded by Mesoproterozoic orogenic belts. The flexural rigidity for this region is estimated to be 2.1 x 1025 N m (Zuber et al. 1989) and the Moho depth is estimated to be ~40 km (Clitheroe et al. 2000).

Eastern Highlands The Eastern Highlands consist of several Palaeozoic fold belts along the Australian coast: the Hodgkinson Fold Belt and Northern Lachlan Fold Belt in the northeast, and the New England Fold Belt and Southern Lachlan Fold Belt in the southeast. Southeastern Australia is characterised by the highest seismicity on the continent (Lambeck et al. 1984), anomalously high heat flow (Cull 1991), and high mantle conductivity (Lilley et al. 1981). The Moho depth in the Eastern Highlands varies from about 32 km in the north to 52 km in the south (Clitheroe et al. 2000). The flexural rigidity is estimated to be 1.6 x 1023 N m for the Hodgkinson Fold Belt and the Northern Lachlan Fold Belt, 3.6 x 1022 N m for the New England Fold Belt, and 4.4 x 1022 N m for the Southern Lachlan Fold Belt (Zuber et al. 1989).

3D Modelling of Australian stress field

75

Figure 3 Stress orientations (solid lines for the maximum horizontal compressive stress directions) at 386 points in the Australian continent, compiled from the Australian Stress Map Project (Hillis & Reynolds 2000; Mueller et al. 2000). Also shown are the major tectonic blocks in the Australian continent (see also Figure 2). NW, northwest Australia; BONA, Northern Bonaparte Basin; WA, western Australia; CA, central Australia; SA, south Australia; EA, eastern Australia.

Figure 4 Stress orientations at the 163 sample points which are used in the inversion analysis. The solid lines denote the observed maximum horizontal compressive stress orientations, which is a subset of the stress orientations shown in Figure 3. Also shown are the major tectonic blocks in the Australian continent (see also Figure 2).

STRESS ORIENTATION DATA ON THE AUSTRALIAN CONTINENT The stress data used in this study are from the World stress map website (Mueller et al. 2000). The dataset was compiled by the Australian Stress Map Project (Hillis et al. 1998, 1999; Hillis & Reynolds 2000) from borehole breakouts, hydraulic fracturing measurements, earthquake focal mechanisms, drilling-induced fracturing, and fault orientations. A total of 386 orientations (the maximum horizontal compressive-stress directions) are plotted in Figure 3. The stress orientations on the Australian continent do not form a uniform direction, although there is some noticeable uni-

formity in several regions (abbreviations in round brackets refer to Figure 3): (i) in northwest Australia there are two dominant stress orientations: 140–150°N, 80-90°N (NW1) and 30–50°N (NW2 and BONA); (ii) in west Australia (WA), there is a trend for the 130–140°N stress orientation but east–west orientations also appear and the scatter in the stress directions is apparent; (iii) in central Australia, the north–south stress orientations are dominant (CA1), and east–west orientations are also present at some points (CA2); (iv) in northeast Australia, north-northeast stress orientations dominate (EA1); (v) in southeast Australia (EA2), there are two dominant stress orientations, 130–140°N in the southernmost part and north–south

76

S. Zhao and R. D. Müller

stress orientations are displayed at some points in the northern part of the region, although the stress orientations in almost every direction were observed in this area; and (vi) in south Australia, the dominant stress orientation is about 130–140°N near the coast (SA2) and further inland near the Adelaide fold belt it is quite scattered, from 130–140°N to east–west (SA1). In this study, we use the stress orientations with quality level A–B exclusively, because of their relatively high reliability (Zoback & Zoback 1991; Zoback 1992). There are a total of 163 stress orientations with a quality level A–B (Figure 4). Comparing Figures 3 and 4, we can see that the main pattern of the stress orientations is still maintained in the dataset containing only level A–B stress indicators.

NUMERICAL MODELLING Stress changes and related phenomena, such as fault activity or earthquakes, are associated with the action of various tectonic forces, which are usually not directly observable. Therefore, the relationship between stress observations and tectonic forces is mainly investigated through modelling studies. The tectonic forces acting on the Australian continent, model assumptions and strategy used in this study are discussed below.

Major tectonic forces acting on the Australian continent RIDGE-PUSH FORCE FROM THE MID-OCEAN RIDGE

Mechanically, the ridge-push force results in a slide force along the ridge-spreading direction (Lister 1975). The horizontal component of the slide force can be viewed as a component of the gravitational force, which is subparallel to the topographic slope. Therefore, the slide force, when averaged over the plate thickness H, can be expressed as (Lister 1975): fR = g    T /(vH)

(1)

where g is the gravitational acceleration,  is the density,  is the thermal expansion coefficient,  is thermal diffusivity, and v is the average half-spreading rate of the ridge. T is the temperature at which the mantle material becomes sufficiently weak such that the lithosphere is decoupled from the asthenosphere. T is usually assumed to be about 900–1000°C for olivine rheology (Goetze & Evans 1979). The Australian continent is located over 3000 km north of the mid-ocean ridge (Figure 1), therefore the horizontal component of the slide force along the ridge-spreading direction (south–north) can be taken as constant throughout the entire continent. The average half-spreading rate for the mid-ocean ridge is about 30 mm/y (Müller et al. 1997). Using typical rock property values  = 4 x 10–5°C–1,  = 3300 kg m–3,  = 8 x 10–7 m2 s–1 and T = 950°C, we have fRH = 1.03 MPa. For the interior of the Australian Plate, which is about 3000 km from the ridge, this force integrates to 3.03 x 1012 N/m. We use a body force (slide force) to represent the ridge push in 3D, and this value is adjusted in the inversion analysis. As defined by a 950°C isotherm, the

thickness of the oceanic lithosphere increases from nearly zero at the crest to about 30 km at the age of ca 10 Ma. The slide force is estimated to be FR = 51 N/m3 for oceanic lithosphere with an average thickness of 20 km near the ridge. This body force is considered to act uniformly throughout the Australian continent and drives the continent northwards. BOUNDARY FORCES

In the northern part of the Australian continent, as a result of the interaction among the Indo-Australian, Eurasian and Pacific Plates, the geometry of boundaries between the plates is very complex. At the Java Trench, the IndoAustralian Plate is subducting beneath the Eurasian Plate. At the Banda Arc and Papua New Guinea, the Eurasian and Pacific Plates are subducting beneath the Australian Plate, and at the Solomon Trench, the Australian Plate is subducting beneath the Pacific Plate (Figure 1). Since the properties of the forces acting at the northern boundary of the Australian Plate are not clear, we assume that their combined effect is to exert a resistant force (relative to the ridge-push force) along the plate boundary near northern Australia (FC in Figure 1). Along the eastern boundary of the Australian Plate, there might be a possible boundary force (FR in Figure 1) transferred from the (New Zealand) subduction zone where the Pacific Plate is subducting beneath the Australian Plate. In west Australia, it is not clear whether there is an equivalent boundary force transmitted from the west (see later). In previous models (Coblentz et al. 1995), the magnitude of the boundary forces for both the northern and eastern boundaries of the Indo-Australian Plate was taken to be ~6 x 1012 Nm–1. DRAG FORCE

Basal drag is the shear traction that the asthenosphere applies to the base of the lithosphere. The direction of the basal drag is usually assumed to be opposite to the direction of the absolute plate motion, but the actual direction is difficult to assess. The magnitude of the basal drag is likely small and of the order of 10–2–10–1 MPa (Richardson 1992). Coblentz et al. (1995) estimated a torque value for the basal drag that is about 50% of the torque of the ridge-push force for the Indo-Australian Plate. In our study, we choose to ignore the effect of the basal drag on stress modelling because: (i) its magnitude is probably small and its other properties (such as its distribution over the base of the lithosphere) are unknown; and (ii) the direction of the drag force, if resisting the plate motion of the Australian Plate, is oriented north–south, which is the same (opposite in direction) as that of the ridge-push force. Therefore, it is not possible to distinguish its presence or absence from the stress orientation data used in this study.

Finite-element model We used a two-layer elastic model to simulate the response of the Australian continent to various tectonic forces (Figure 5). The use of the elastic rheology in the stress analysis for the Australian continent is an oversimplifica-

3D Modelling of Australian stress field

77

is indirectly included in our analysis by considering the differences in their elastic strength, as determined from the coherence of Bouguer gravity anomalies and topography (Zuber et al. 1989).

Boundary conditions

Figure 5 The finite-element grid used in this study.

tion, as recognised by Cloetingh and Wortel (1986) and Coblentz et al. (1995; 1998). An elastic rheology is a justifiable approach for investigating the first-order tectonic stresses in continental Australia when the data are insufficiently restricting the use of a more complicated rheology. A viscoelastic model would be more suitable for investigating stress-relaxation processes in the crust/lithosphere (Lambeck 1983a; Stephenson & Lambeck 1985). The finite-element model (Figure 5) contains a total of 4185 nodes and 2640 brick elements, and its spatial resolution is about 90 x 90 x 50 km (the dimension of each element). We use the Lambert azimuth projection to transform geographical coordinates into Cartesian coordinates, in which the finite-element method computation is carried out. Since the errors in delineating the boundaries of tectonic blocks is up to ~100 km, only tectonic blocks wider than 100 km are included in the analysis, and small structures, such as faults, are ignored in our continental-scale model. Figure 6 shows the distribution of the finite-element nodes and the major tectonic blocks in the Australian continent. The approximation of the geometry of the tectonic blocks is achieved by representing the irregular boundaries with rectangles (bricks in 3D) (Figure 6). The differences in the effective elastic thickness (and the flexural rigidity) of the blocks reflect their differences in elastic strength, used here to constrain the rheological parameters of our numerical model. The geological/rheological provinces with different elastic strength are incorporated into the upper layer with a thickness of 50 km (Figure 5), which is close to the maximum Moho depth in continental Australia (Clitheroe et al. 2000). A bottom layer of 50 km is introduced as a reference layer with a Young’s modulus of 14 x 1010 Pa (Turcotte & Schubert 1982) to reflect the fact that the elastic strength of the mantle (bottom layer) is higher than the crust (upper layer). A change of the thickness of each layer affects the magnitude of the stresses, but it does not affect the relative magnitudes and pattern (including the orientations) of the calculated stresses. Our main objective is to estimate the relative magnitudes and the pattern of tectonic stresses, rather than estimating the absolute magnitude of the tectonic stresses in the Australian continent. The effect of variations in the equivalent elastic thickness of the lithosphere

We assume that the nodes at the bottom (depth = 100 km) of the tectonic block are fixed (Figure 5), which implies that the motion/deformation of the top layer (lithosphere), if any, is relative to the fixed bottom layer. Other boundaries are free. In one of our models, the western boundary of the block, corresponding to the west Australian coast, is assumed fixed, and its effect on modelling results is examined in an inversion analysis. In all other models, boundary forces are imposed at the western, eastern and northern boundaries. Their typical values from previous studies (Coblentz et al. 1995) are used and adjusted in the inversion analysis.

Rheological parameters The major geological structures considered in this study correspond to those investigated by Zuber et al. (1989), except for the Adelaide Fold Belt, which was not included in their study. We assume that the rheological contrast in the Australian continent can be represented by 17 groups of material in terms of their differences in elastic strength (Table 1). We adopt a constant value 0.25 for the Poisson ratio throughout the investigated area and use different values of Young’s modulus to represent the difference in the elastic strength for different geological structures. The estimates for the rigidity of the geological structures in the Australian continent are taken from the flexural analysis based on the gravity and topographic data by Zuber et al. (1989) (Table 1). The initial values of Young’s modulus for the tectonic blocks are estimated on the basis of the relative magnitude of the flexural rigidity (Table 1). The scaled values of the flexural rigidity in Table 1 are obtained by dividing by the maximum value of the originally estimated flexural rigidity (2.1 x 1025 N m). The Young’s modulus values of the blocks are estimated on the basis of their scaling factors. Young’s modulus is taken to be 3.0 x 1010 Pa for the basins (the blank area in Figure 2) and 5.7 x 1010 Pa for the oceanic crust/lithosphere (the area outside the continent). In Table 1, a Young’s modulus value of 0.21 x 1010 Pa is used for the continental shelf, which is assumed to have a lower strength than the continental crust. Inclusion of the continental shelf in the modelling analysis did not improve the quantitative analysis of the stress orientations, implying that these stress orientations are not sensitive to the rheological parameters of the continental shelf. A possible reason is that we used a single value to describe the strength of the continental shelf throughout the continent. Actually, the mechanical properties of the continental slope could be different from region to region (e.g. from the west to south Australia), but the stress orientation data in the continent could not resolve such differences. The flexural rigidity values of fold belts (Table 1) estimated by Zuber et al. (1989) are about three orders of magnitude lower than those of cratons. The errors in estimating

78

S. Zhao and R. D. Müller

the flexural rigidity of the fold belts from the gravity and topographic data (Zuber et al. 1989) may be large compared with those of cratons because of their relatively small dimensions (Simons et al. 2000). These errors in the estimated flexural rigidity are transferred to the Young’s modulus values used in this study. Therefore, the values of the Young’s modulus for the fold belts are adjusted (re-estimated) in an inversion analysis of the stress orientation data. The values for the adjusted Young’s modulus, and the values for refined rigidity in Table 1 are the resulting flexural rigidities for some of the tectonic blocks obtained from the inversion analysis of the stress orientation data and will be discussed later.

Inversion analysis Numerical modelling with the finite-element method can be classified into two types: forward and inverse analyses. Forward modelling calculates the stresses (orientations) in the continent from known tectonic forces and rheological parameters. Inverse modelling estimates some unknown forces and rheological parameters from observations (e.g. orientations) and some known tectonic forces and rheological parameters. Suppose that the stress orientations (Y) in the Australian continent are a function of the tectonic forces (Fx) and rheological parameters (R): Y = f(Fx, R)

(2)

where f is an operator which expresses the relationship between Y and (Fx, R). The description of the forward problem is that we want to estimate Y, given Fx and R. Unfortunately, we only have very limited knowledge of the tectonic forces (Fx) and rheological properties of the continent. In previous studies, the rheological parameter R was taken as a constant, and only the parameter Fx was adjusted by visually comparing calculated stress patterns with observed stress orientations. If there are relatively ample stress observations and we wish to use the observations as quantitative constraints, the problem has to be considered in a reverse way—the inverse problem—which can be described as wanting to estimate Fx and/or R, given Y (to estimate tectonic forces and rheological parameters from observed stress orientations): (Fx, R) = f–1 (Y)





|| Y – Y ( R ) || = min b1 ≤ R ≤ b2

( 6) ( 7)

where –Y is the observational vector of the stress orientations, Y ( R ) is the modelled stress orientation (vector), which now depends on the rheological parameters (R), and b1 and b2 are the lower and upper limits of the rheological parameters. The constraints for the rheological parameters are relatively easy to obtain. For example, we know that R >0, and from the results of laboratory experiments we can determine approximate upper limits for the rheological parameters. Methods for solving geophysical inverse problems have been presented by Menke (1984) and Tarantola (1987), and will not be discussed here. The inversion analysis is conducted in a stepwise fashion. We first take the Young’s modulus scaled from the flexural rigidity analysis as initial values, and then investigate the response of tectonic blocks to the assigned tectonic forces. Different combinations of the tectonic forces are examined and adjusted to fit the stress orientations in the Australian continent. We then take the estimated tectonic forces as known parameters and refine the estimates of the Young’s modulus to fit the stress orientations. These procedures are repeated until the squared residuals between the observed and modelled stress orientations reach a minimum, in a least-squares sense. The final estimates of the tectonic forces and the Young’s moduli can be viewed as global least-squares estimates.

(3)

where f–1 is an inverse operator which expresses the inverse relationship between Y and (Fx, R). Since the stress dataset on the Australian continent is not (mathematically) complete the solutions of the associated inverse problems are not unique. We need to introduce a priori information into the modelling analysis. The known geometry of geological elements and basic information on the directions and/or magnitude of tectonic forces from the previous studies are taken as a priori information in our model. The inverse problem for estimating tectonic forces from stress orientations, can then be expressed as: || Y – Y (Fx ) || = min a1 ≤ Fx ≤ a2



where Y is the observational stress orientations (vector), Y is the modelled stress orientations (vector), Fx is the parameter vector of the tectonic forces to be estimated; a1 and a2 are coefficients that are the lower and upper limits of the parameter (Fx). The rheological parameters in model (4) are assumed to be known. In addition, the constraints for the tectonic forces are easily obtained based on information from plate tectonics or previous forward models. For example, for the ridgepush force associated with the Australian continent, its direction is approximately northward (along the Y-axis in our model), so we have a1 >0 or Fx >0. We can also infer a rough value (a2) for the upper limit of the ridge-push force. Likewise, the inverse problem of estimating the rheological parameters from stress orientations can be expressed as:

(4) (5)

RESULTS Ridge-push force An initial value of 6.0 N/m3 was assigned to the magnitude of the ridge-push force in the Australian continent. The refined value (FP in Table 2; Figure 1) after the inverse analysis is 55.8 N/m3 (0°N), which is close to the value 51 N/m3, calculated independently from Equation (1) based on parameters for the Southeast Indian Ridge. To examine the possible east–west component of the slide force, the tectonic block is divided into four sections (Figure 6), and for each of the sections, an eastward ridge-push force component is added: the resultant changes in the stress orientations are not significant (Table 2).

3D Modelling of Australian stress field

79

Figure 6 Map showing the distribution of the finite-element nodes in plan view, and the approximation for the geometry of the major geological structures in the Australian continent with the brick elements (50 km thick each). The areas coloured red are cratons; fold belts are green. The yellow curve denotes the boundary of the continental shelf. The numbers represent the estimates of the effective elastic thickness in kilometres (Zuber et al. 1989). Arrows denote the force vectors (not to scale) considered in the modelling analysis.

Boundary forces In the eastern boundary (Figure 1), there may be a force (FR) resulting from the Pacific Plate, which is subducting beneath the Australian Plate. Similar to the value (1012–1013 Pa/m) used in previous studies (Coblentz et al. 1995), we assume an initial value of 2.0 x 1012 Pa/m for the magnitude of the boundary force at the eastern boundary, and the search for the optimal value gives an estimate of > 5.99 ±5.2 x 1012 Pa/m: that is, the upper limit of the magnitude is uncertain. The error in this estimate is significant and we interpret this as implying that the dataset is not sensitive to the magnitude of boundary force from the east. For the boundary force (FC) associated with the northern boundary, we used the same initial value as that used for the eastern boundary. The estimated value is 11.8 ±3.2 x 1012 Pa/m. The absolute values of the forces are difficult to evaluate from the stress orientation data used in this study. The estimated values of the forces only have relative importance, and they are model dependent. In other words, only the relative magnitudes of the tectonic forces can be constrained by the stress-orientation data.

In order to explore the possible range of boundary forces at the western boundary of the Australian continent, we tested three models: (i) an east–west boundary force of 1.0 x 1012 Pa/m; (ii) a free boundary; and (iii) a fixed boundary. We found that none of the three models produces a significant improvement in the residuals of the stress orientations. Since the western boundary of the Australian continent is in the interior of the IndoAustralian Plate (Figure 1), the tectonic deformation in western Australia associated with plate-boundary forces is much smaller than that in northern Australia, which supports the use of a fixed boundary. In addition, the fixed western boundary in the third model serves to resist any plate-tectonic forces from deflecting the western boundary, which seems reasonable. Therefore, we adopt the fixed western boundary assumption. A further test of the boundary force vectors acting on four sub-segments of the western boundary of the block (Figure 6) did not produce any substantial improvements on the misfit, which might suggest that the stress observations (WA in Figure 7) discussed here cannot be fully accounted for by the continental-scale model. Possible mechanisms for the local

80

S. Zhao and R. D. Müller

Figure 7 Observed (black lines) and estimated (red lines, Model A1) maximum horizontal compressive stress orientations. NW, northwest Australia; BONA, Northern Bonaparte Basin; WA, western Australia; CA, central Australia; SA, south Australia; EA, eastern Australia. Table 2 Estimated magnitude of the ridge push and boundary forces. Type of force Ridge push (slide force) (FP) Boundary force (FR) Boundary force (FC)

Component Northward Eastward Northward Eastward Northward Eastward

Initial N/m3

Final

6.0 – – 2.00 x 1012 Pa/m 2.00 x 1012 Pa/m –

stress field in west Australia will be discussed later. To test the effect of the possible oblique-type forces at the northern boundary of the Australian continent, the northern segment of the tectonic block is divided into four segments (Figure 6). For each segment, in addition to the north–south component of the boundary force (along the Yaxis), an east–west component (along the X-axis) is also assumed to be unknown in the inversion analysis. A combination of the two components (X and Y) constitutes a force vector. However, the inversion fails to give a significant estimate for the east–west component of the collision force for all segments. This suggests that detailed characteristics of the tectonic-force vectors cannot be resolved from the available stress-orientation data. This may indicate that the variations of the boundary forces along the northern Australian plate boundary have little effect on the stress orientations observed within the Australian continent. This supports the interpretation that a large amount of the energy associated with subduction zones may be dissipated by resistance to subduction, and therefore a surface plate may not experience substantial slab pull (Richardson 1992; Hillis et al. 1997). However, this does not mean that the effect of the forces acting at the plate boundary on the intraplate tectonic stress field can be dismissed. Forces originating at the plate boundary, other than the boundary forces considered here, could have some significant effects

55.8 ±6.1 N/m3 not significant not significant FC >5.99 ±5.2 x 1012 Pa/m 11.8 ±3.2 x 1012 Pa/m not significant

on the intraplate stress field, such as the body forces transmitted from the plate boundaries, but the properties of such forces are not very clear. The maximum horizontal stress orientations estimated from the inversion analysis (Model A1) are shown in Figure 7. The estimated stress orientations (red lines) are fairly consistent with the observations (black lines) in the western part of northwest Australia (NW1), Northern Bonaparte Basin (BONA), central part of Australia (CA1) and northeast Australia (EA1). While the inclusion of the geological structures in our analysis has provided a reasonable fit to some of the observed stress orientations, there are deviations between the observed and modelled stress orientations, e.g. in the southern part of west Australia (WA), the eastern part of northwest Australia (NW2), central Australia (CA2), eastern Australia (EA2) and south Australia (SA2).

Refined rheological parameters The stress orientations for the points near or inside a geological structure are associated with the material contrast between the structure and its surrounding area. Therefore, the rheological parameters for some of the tectonic blocks are adjusted to obtain their optimal values by inversion analysis of the stress-orientation data.

3D Modelling of Australian stress field

81

Figure 8 Observed (black lines) and modelled (red lines, Model A2) maximum horizontal principal stress orientations (after adjustment of the rheological parameters of the tectonic blocks). NW, northwest Australia; BONA, Northern Bonaparte Basin; WA, western Australia; CA, central Australia; SA, south Australia; EA, eastern Australia.

Figure 9 Observed (black lines) and modelled (red lines, Model A3) maximum horizontal principal stress orientations (after inclusion of the effect of local stress fields). The shaded bars denote the orientations of the introduced local stress fields. NW, northwest Australia; BONA, Northern Bonaparte Basin; WA, western Australia; CA, central Australia; SA, south Australia; EA, eastern Australia.

INLAND BASINS

FOLD BELTS

1010

The Young’s modulus value is taken to be 3.0 x Pa for the basins in the Australian continent (Figure 2). A change in the value of the Young’s modulus of the basins has little effect on the overall residuals between the observed and modelled stress orientations, suggesting that stress orientations are not sensitive to the change of this parameter. A possible reason could be that we used a single value to represent the elastic strength for all of the basins, so that the difference among the basins, which (if any) affects the local stress distributions, could not be distinguished in the continental-scale model.

The inversion analysis indicates that the stress orientations are sensitive to the Young’s modulus values of fold belts. Comparing the Young’s modulus values (~1010 Nm) of the cratons, the re-estimated values for the fold belts are about one to two orders of magnitude lower (Table 1). The re-estimated value of the Young’s modulus is 0.113 x 1010 Pa for the Northern Lachlan Fold Belt, 0.106 x 1010 Pa for the New England Fold Belt and 0.113 x 1010 Pa for the Southern Lachlan Fold Belt (Table 1). The adjusted flexural rigidity value is 0.040 x 1025 Nm for the Northern Lachlan Fold Belt, 0.037 x 1025 Nm for the New England Fold Belt,

82

S. Zhao and R. D. Müller

and 0.040 x 1025 Nm for the Southern Lachlan Fold Belt (Table 1). The estimates correspond to an effective elastic thickness of about 30 km.

COMPARING THE OBSERVED AND MODELLED STRESS ORIENTATIONS Figure 8 shows the observed (black lines) and estimated (red lines) maximum horizontal stress orientations as obtained after the adjustment of the rheological parameters (Model A2). The standard deviation of the residuals is ±45.3° and ±44.5° for Models A1 and A2, respectively. Since the standard error in the observed stress orientations could be about ±15° (Zoback 1992), the difference between the two models is not statistically significant. However, considering that there are many non-statistical uncertainties in the quantitative analysis of observed stress orientations as well as the associated tectonic forces, the results obtained in this study should be viewed as semiquantitative. Comparing the observed (black lines) and modelled (red lines) stress orientations in Figure 8, the general pattern of the observed stress orientations has been reconstructed by the numerical model, although deviations still exist at some sites. In eastern Australia, the fold belts are simulated as ‘weak zones’ in the numerical analysis, and substantial rotations in the stress orientations occurred near these ‘weak zones’. The variations in the stress orientations reflect the combined effect of the tectonic forces and the contrast in the elastic strength of tectonic elements on producing the intraplate stresses. For the northern part of eastern Australia (EA1), Model A2 (Figure 8) predicts two types of stress orientations: northeast and north-northeast. The modelled stress orientations are generally consistent with observed stress orientations. The stress orientations predicted for the southern part of eastern Australia (EA2) are also of two types, northwest and northeast, and apparent deviations exist between the predicted north-northeast and observed northeast orientations. For the stress indicators around the Southern Lachlan Fold Belt (Figures 2, 8), the orientations predicted by the model are largely of two types, northwest and north-northeast, and significant deviations exist between observations and predictions. Model A2 could not reflect the rotations of the observed stress orientations from northeast in the north (EA2) to northwest in the southernmost part of the region. Stephenson and Lambeck (1985) constructed an erosion-rebound model for southeastern Australia to explain the geomorphological and geological observations for the uplift that has occurred since Early Cenozoic time, and predicted a tensile stress field for southeastern Australia (with northwest orientations: Stephenson & Lambeck 1985 figure 12, p. 50). Since only the regional trends of the stress orientations related to the continental-scale tectonic forces, as well as the contrast in the elastic strength among major tectonic blocks, are simulated in our model, the local stress changes caused by different kinematic/dynamic mechanisms, such as the erosion-rebound effect discussed by Stephenson and Lambeck (1985), can not be directly accounted for. However, after superimposing a local tensile stress field (103°N: green bar near EA2 in Figure 9) estimated by Stephenson and Lambeck (1985) onto the

Figure 10 Distribution of the residuals (between the observed and modelled stress orientations) for Models A1, A2 and A3. The vertical lines indicate the one standard deviation.

regional stress field predicted by Model A2, a hybrid model (Model A3) is obtained, and the stress orientations predicted by the model are shown in Figure 9. There are some significant improvements on the fit between the observations and predictions. For the points around the Southern Lachlan Fold Belt (Figures 2, 9), the stress orientations are now consistent with the observations. In northwest Australia, the east–west and northeast stress orientations (NW1 and NW2 in Figure 8) are not fitted by the model. As mentioned before, adjusting the distribution of the collision forces at the northern boundary failed to reduce the misfit. One possible cause for this misfit might be the effect of some local geological structures in the region. The borehole breakout data in this area are from the Canning Basin, which is bounded by Fitzroy Trough in the north. The sediments in the Fitzroy Trough are about 14 km thick (Borissova & Symonds 1997); its length is about 700 km and its width is about 100 km. The presence of the Fitzroy Trough could have some effect on the local stress field, but it is difficult to include this effect into our analysis due to the narrowness of the trough, for which a model with

3D Modelling of Australian stress field

83

Figure 11 Principal-stress distribution in continental Australia (in units of 100 MPa). Also shown are the boundary of the major geological structures (yellow), the boundary of the continental shelf (green), and epicentres of the earthquakes with magnitudes of M ≥3.0 (triangles) and M ≥5.0 (stars). The blank areas represent the zones of least compression (with a compressive stress value ≤0 MPa).

a resolution of at least 50 km is required. In addition, a tensile stress regime has been reported in northwest Australia (Coblentz et al. 1995), which might be related to the effect of the continental shelf and deep basins in the region. We therefore tentatively introduce a local tensile field with its direction perpendicular to the coast (140°N: green bar near NW2 in Figure 9). After inclusion of the local stress field (Model A3, Figure 9), we see that the predicted stress orientations are now consistent with the observations. In west Australia (WA), for the eight stress indicators (Model A3) used in the analysis, the average deviation between the observed and modelled orientations is about 46°, which is larger than the standard deviation of the model (±37.6°). In addition, a test for inclusion of the boundary-force vector with different magnitudes and orientations on four sub-segments along the western boundary failed to improve the fit (Figure 6). This suggests that a further improvement on the fit between the observed and modelled stress orientations with the present model is difficult. In previous studies, two mechanisms were proposed for the rotation of the local stress field in west Australia. Cloetingh and Wortel (1986) suggested that the

state of compression in the western part of central Australia is induced by the action of resistant forces at the Himalayan and Banda arc collision zones (Figure 1 inset). However, as demonstrated by Coblentz et al. (1995, 1998), the collision forces produce stress focusing only near the boundaries, and their effects on the orientations of the stresses within the plate are secondary. Since most of the stress orientations in central and northwest Australia have been fitted by the present model (Figure 9), a causative mechanism for the local variations of the stress orientation in the interior of the Australian continent due to regional- or plate-scale forces is not likely. A local mechanism for the stress changes and seismicity in west Australia has been proposed by Lambeck et al. (1984). They suggested that: (i) there might be a local/regional stress field resulting from the interaction between the Yilgarn Block (YB in Figure 2) and the nearby Darling Fault; and (ii) the stress regime of the local stress field could be tensile. The north–south-oriented Darling Fault (Borissova & Symonds 1997) is more than 800 km long, but its width is less than 50 km. Therefore, the effect of the fault and its interaction with the Yilgarn Block as

84

S. Zhao and R. D. Müller

Figure 12 Shear-wave speed anomalies (depth = 80 km) for the upper mantle of the Australian continent (modified from Kennett 1997, 2003). The letters mark the major zones of shear-wave speed anomaly in southern Australia (A), central Australia (C and C1), eastern Australia (E1 and E2), northern Australia (N1) and western Australia (W1) (see text for discussion).

well as their combined effects on the local/regional stress field could not be simulated here, given the resolution of the present model. Figure 10 shows the distribution of the residuals for Models A1, A2, and A3. The numerical model has statistically fit the observed stress orientations to ±37.6° (Model A3). More than 45% of the observed stress orientations have been fitted by our model within ±25°. Overall, the numerical model provides a reasonable interpretation of the observed stress orientations in the Australian continent.

PRINCIPAL-STRESS DISTRIBUTION AND SEISMICITY IN CONTINENTAL AUSTRALIA Figure 11 shows the principal-stress distribution predicted in this study with seismicity in continental Australia superimposed. Seismicity in the Australian continent is concentrated in several zones (Figure 11).

(1) In Western Australia, earthquakes are observed mostly in the southern part of the Yilgarn Block (also see Figure 2) and near the North West Shelf (northwest Australia). Fault plane solutions for two earthquakes (M = 6.8, 1968; 5.9, 1970) in Western Australia indicate thrust faulting (Fitch et al. 1973). The source mechanisms for the earthquakes in the North West Shelf are not well determined, as most of them occurred along the continental shelf. (2) In South Australia, seismicity is largely confined to the Adelaide Fold Belt and the adjacent gulf graben regions. An average depth of about 10 km is estimated for the earthquakes recorded during 1976–77 by McCue and Sutton (1979). In addition, the source mechanism solutions of the events indicate failure by strike-slip faulting. (3) In central Australia, seismicity is relatively diffuse. A concentration of seismicity is observed in the Gawler Block and near the Arunta Block (Figure 2). Source mechanism solutions of two earthquakes (M = 6.2, 1972; 4.7, 1978) in the Simpson Desert show failure by compression.

3D Modelling of Australian stress field

85

Figure 13 Principal-stress distribution in continental Australia (in units of 100 MPa) computed after including the rheological information from seismic tomography. Also shown are the boundary of the major geological structures (yellow), the boundary of continental shelf (green), and epicentres of the earthquakes with the magnitude of M ≥3.0 (triangles) and M ≥5.0 (stars). The blank areas represent the zones of least compression (with a compressive principal stress value ≤0 MPa).

(4) In east Australia, seismicity is mostly concentrated around the Southern Lachlan Fold Belt along the coast adjacent to the Northern Lachlan Fold Belt and the New England Fold Belt (Figure 2). Most of the earthquakes in southeast Australia indicate horizontal compressive failure. Earthquakes are indicative of where stress is concentrated so that the brittle failure limit of the crust has been exceeded. Therefore, a correlation between seismicity and the predicted stress distribution is expected. Comparing the distribution of seismicity and the pattern of the stress predicted in this study, we see that such a correspondence does exist: the seismicity in northwest and southeast Australia falls into two bands where stress concentration is predicted. Nevertheless, there are still several zones where the predicted stress concentration is not compatible with seismicity observed in the continent. In southwest Australia, a zone of intense seismicity in the Yilgarn Block (Figure 2) does not correspond to any concentration of stress predicted by our model. Along the Great Australian Bight coastline (south Australia), a large zone of stress concentration is predicted, which is unsupported by observations.

Only the eastern part of this stress concentration zone corresponds to the seismicity near or around the Adelaide Fold Belt. The western part of the zone does not correspond to any recorded seismicity. In central Australia, the diffuse seismicity is not accounted for by the stress concentration predicted in this model. To further interpret the seismicity in continental Australia, it is necessary to include additional information on the contrast in elastic strength of the tectonic elements, such as results from seismic tomography (Kennett 1997, 2003; Simons et al. 1999) (Figure 12). Shear-velocity anomalies reveal the relative contrast in elastic strength among the tectonic elements: seismically ‘slow’ (negative anomaly in Figure 12) indicates the material in the area is of ‘lower strength’, and seismically ‘fast’ (positive anomaly) indicates ‘higher strength’. Seismically ‘slowest’ is predicted for the Southern and Northern Lachlan Fold Belts (marked E1 and E2), and ‘fastest’ is predicted for western (marked W1), central (C1) and northern Australia (marked N1). In addition, relatively small, but noticeable, seismically ‘fast’ zones also appear in the western part of South Australia

86

S. Zhao and R. D. Müller

(marked A) and eastern part of central Australia (marked C). These velocity anomalies may reflect material contrasts between the cratons, basins and fold belts. We investigate this hypothesis by including these contrasts between lower and higher strength lithospheric blocks based on seismic tomography in terms of differences in their Young’s moduli. After inclusion of this additional information, a revised map of the principal stress distribution for continental Australia is constructed (Figure 13). A noticeable feature in the predicted stress pattern is that the areas with the least compression (the blank areas in the continent) are almost seismicity free. The band of stress concentration along the Great Australian Bight, where little seismicity is observed, has disappeared and moved further north where it now matches a belt of seismicity from the Musgrave Block to the Yilgarn Block. The magnitude of the principal stresses is estimated between 10 and 40 MPa, and the deformation style is largely compressive. The predicted area with the least compression (blank zone) in the North West Shelf corresponds approximately to the normal faulting stress regime inferred from the in situ stress data (Hillis 1991; Hillis & Williams 1992, 1993a, b). The stress concentration zones predicted by the model now correspond quite well with the areas where concentrated seismicity is observed. The improvements in model prediction by including lateral variations in lithospheric rheology based on seismic tomography illustrate the shortcomings of relying on surface geology and gravity/topography coherence results for estimating the spatial variation in lithospheric rigidity. However, the inclusion of information from seismic tomography in our model does not contribute to further improvement of the fit between the observed and modelled stress orientations on the Australian continent. Features revealed by seismic tomographic analysis have a larger length-scale and poorer resolution than those from surface geological investigations. As discussed before, we need to include more information on local and small-scale stress sources into a higher resolution model in order to interpret the variations of stress orientations in some regions. The extra information obtained from the seismic tomographic analysis at the current resolution does not contribute significantly to a better and quantitative interpretation of the stress orientations observed on the Australian continent, as compared to the rheological model based on the coherence of Bouguer gravity and topography.

DISCUSSION A three-dimensional finite-element model has been constructed and used to investigate the pattern and orientations of the tectonic stresses in continental Australia. The model, which consists of two layers (Figure 5), provides a spatial resolution of about 90 x 90 x 50 km. The major geological structures such as cratons and fold belts are included in the analysis. The difference in the elastic strength of the tectonic structures are initially estimated on the basis of their rigidity values inferred from the coherence of Bouguer gravity and topography (Zuber et al. 1989). The major tectonic forces which act on the Australian continent (such as ridge-push and plate-boundary forces) are investigated in the analysis. An inversion approach is used to estimate the relative magnitude of tectonic forces from the

observed stress orientations (equations 4 and 5). In addition, an approach for estimating the main rheological parameter (Young’s modulus) from the inversion analysis of the observed stress orientations is also developed (equations 6 and 7) and used to estimate the values of the Young’s modulus for some of the geological structures. Our results suggest that the slide force associated with ridge push is the dominant force that controls the magnitude and orientations of the stress field in the Australian continent, confirming the results of Coblentz et al. (1995, 1998). The magnitude of the slide force is estimated to be 55.8 N/m3, and the magnitude of the forces at the eastern and northern boundaries is estimated to be >5.99 x 1012 Pa/m, and 11.8 x 1012 Pa/m, respectively (Table 2). The estimates for the magnitude of the forces are model dependent and subject to many uncertainties (e.g. the assumed rheological parameters and geometry of tectonic blocks). Therefore, they may be interpreted only as semiquantitative estimates. The boundary forces acting on the northern and eastern boundaries of the Australian continent only have a secondary effect on the overall stress pattern, and they do not significantly affect the pattern of the stress in the interior of the continent. The presence of tectonic domains with different rigidities has a significant influence on the pattern of the estimated regional and local stresses. After combining the tectonic forces, major geological structures, and the effect of the local stress fields in the numerical model, a reasonable fit has been achieved between the observed and modelled stress orientations (Figure 9). The in situ stress orientations can be statistically fitted within ±37.6° by the numerical model. The inversion analysis of rheological parameters is useful for estimating the Young’s moduli for the Northern Lachlan Fold Belt, the New England Fold Belt, and the Southern Lachlan Fold Belt. The adjusted values for the flexural rigidity are 0.040 x 1025 Nm for the Northern Lachlan Fold Belt, 0.037 x 1025 Nm for the New England Fold Belt, and 0.040 x 1025 Nm for the Southern Lachlan Fold Belt (Table 1), which correspond to an effective elastic thickness of about 30 km. These estimates are about two orders of magnitude lower than those for the cratons (~1025 Nm). The original estimates (~1022 Nm) for the fold belts from Zuber et al. (1989), which are about three orders of magnitude lower than those for the cratons (Table 1), may have been underestimated (Simon et al. 2000). It appears that the re-estimated values of the rigidity for the fold belts from this study, which are between the maximum and minimum of the flexural rigidity estimated by Zuber et al. (1989) and constrained by the stress-orientation data, are more geologically plausible. Therefore, we have provided an indirect estimate for the flexural rigidity of the fold belts in continental Australia. Another significant result from this study is the estimated distribution of the principal stress in the Australian continent (Figure 13). We predict stress concentration in northwest Australia, south Australia, and southeast Australia. In addition, several zones with least compression are also identified in the continent. Although the predicted deformation style in the Australian continent by our model is of compression and strike-slip faulting, it is plausible to infer that normal faults are most likely to develop in the

3D Modelling of Australian stress field

areas where the least compression is predicted. It is also noteworthy that the concentration of seismicity is not observed inside the predicted least compression zones, but it is mostly inside the zones of significant compression. Therefore, the principal-stress distribution predicted here has furnished a preliminary interpretation for the seismicity observed in continental Australia. Considering lateral variations in lithospheric strength in the modelling analysis by including results from shear-wave tomography proved to be essential to remove some firstorder artefacts from the initial model, and improve the match of modelled zones of stress concentration with observed belts of seismicity. The results demonstrate that by combining surface geology, lithospheric rigidity estimates from gravity–topography coherence, and seismic tomography, we have assembled a simple rheological model for the Australian Plate that, together with an optimised model for plate-driving forces, accounts for the observed large-scale patterns of intraplate seismicity in Australia. However, like any other numerical analysis (Richardson et al. 1979; Cloetingh & Wortel 1986; Coblentz et al. 1995, 1998), there are many limitations inherent in our model. Although the estimated magnitude of the principal stress between 10 and 40 MPa is compatible with the value (~tens of megapascals, over a 100 km-thick layer) estimated by Coblentz et al. (1998), it is subject at least to the following uncertainties: (i) since the magnitude of the boundary forces is actually unknown, a geologically plausible value has been adopted (typically, a value of ~1012 N/m was used), the absolute value of the forces could not be well determined by the analysis of the stress orientation data alone; (ii) the absolute values of the rheological parameters of the crust/lithosphere are unknown, and a value of ~1010 Pa was used for Young’s modulus; and (iii) the magnitude of the stresses is estimated over a layer of 50 km thickness, and the effect of the rigidity layering as well as any other depth dependent-stress changes have been ignored, which affects the magnitude of the calculated stresses. What we have estimated in this study are the relative magnitude and the pattern of the tectonic stresses, rather than the absolute magnitude of the tectonic stresses in the Australian continent. Our analysis shows that ignoring the effect of the gravity potential energy differences in the Australian continent influences the modelling results. Two additional stress fields required to fit the observed stress orientations in northwest (NW1 and NW2) and southeast (EA2 in Figure 9) Australia may reflect the possible contribution of the topography or gravity potential energy difference at areas near the continental margin. One of the mechanical effects of the gravity potential energy difference at the continental margin is to produce a local stress field. The stress concentration reflected by seismicity near the continental margin predicted in our model indicates that the mechanical strength of the continental shelf is weaker than that of continental crust whose last thinning/reheating event is substantially older (see Fowler & McKenzie 1989). This weakening effect has been incorporated in our model by including the continental shelf as a weak zone. However, the forces arising from the gravity potential energy difference at the continental margin are not directly simulated in our study. Since the crustal structure at the continental

87

margin could vary from place to place, a separate analysis of the local stress field associated with the gravity potential energy difference or gravity instability based on a detailed (density) structure model is required in the future. Moreover, many small- to intermediate-scale geological structures are not included in our study, such as basins and faults. For some of the basins, the depth of the sediment to the basement is more than 10 km (e.g. the Browse Basin in the North West Shelf: Borissora & Symonds 1997), and for some crustal-scale faults, their length scale is up to 500 km (e.g. the Darling Fault in Western Australia). Inclusion of geological structures into a future model with a higher spatial resolution will alter the magnitude as well as the pattern of the calculated stress in the areas around or close to these structures. Further, the present activity or reactivation of faults also influences the pattern of the local/regional stress field (Sandiford & Hand 1998). These could be the objects of future local or regional stress analysis, which may be designed to explore the effects of the local or regional geological structures as well as their activity on the tectonic stress field. These factors discussed above could in part account for the reason that about 20% of the observed stress orientations are not well fitted by our continentalscale model. The interaction between the lithosphere and mantle or the upper and lower crust has not been considered in our analysis. The stress transferred from the lower or the upper mantle into the upper crust or lithosphere associated with pre-existing geological structures could influence the pattern of the local or regional stress field (Kusznir & Bott 1977; Lambeck et al. 1984). However, the magnitude and properties of the transferred stresses, which are model dependent, are very difficult to assess. For instance, a stress difference of 50–200 MPa for eastern Australia is predicted by the erosion–rebound model of Stephenson and Lambeck (1985), which is almost at the same magnitude as that of the predicted regional stress field (Coblentz et al. 1995, 1998). Our study has shown that a combination of local stress-relaxation processes associated with some distinct geological structures with a continental-scale model better accounts for the observed stresses. However, the increasing uncertainties with adding more geodynamic mechanisms into any model will further increase the ambiguity of the results. Therefore, these geodynamic processes are currently modelled separately. The type of deformation and the stress regime inferred from the in situ stress measurements, the seismic source mechanisms, and the numerical model experiments are all depth dependent. The information of the deformation type and faulting style estimated from earthquakes in the upper crust could be different from the earthquakes in the middle crust, or different from those obtained from the in situ stress measurements. Therefore, the available information on the stress regime from the in situ stress measurements, the earthquakes source mechanisms, and the numerical modelling of the Australian continent is incomplete. While the dominant deformation style in the Australian continent inferred from this study is compression, which is consistent with that from the in situ stress measurements; caution should be taken when extrapolating the results to the state of the stress at depth because of the inherent uncertainties stated above.

88

S. Zhao and R. D. Müller

SUMMARY

REFERENCES

The main results from the 3D stress analysis with the finiteelement method for the Australian continent are as follows. (1) The ridge-push force is the dominant force which controls the magnitude and pattern of the first-order stresses in the Australian continent. The effect of the boundary forces are secondary and they mostly influence the pattern of the stress in areas near the boundaries. There is no need to invoke the drag force to explain the first-order stress pattern in the Australian continent partly because of our poor understanding of the properties of the drag force and the insensitivity of the stress orientation data to the drag force. These results are consistent with those obtained by Coblentz et al. (1995). (2) Geological structures significantly affect the magnitude and pattern of modelled stresses. Combining spatial variations in rigidity between major geological structures (cratons and fold belts) and a tectonic-force model, by simultaneously inverting for stress orientations and tectonic-force vectors, a fairly good fit has been achieved between the observed and modelled stress orientations. The model can explain statistically about 45% of the observed stress orientations within ±25°, and about 62% within ±40°. (3) The model also provides an indirect estimate of the flexural rigidity for the Northern Lachlan Fold Belt (0.040 x 1025 Nm), the New England Fold Belt (0.037 x 1025 Nm) and the Southern Lachlan Fold Belt (0.040 x 1025 Nm). These estimates correspond to an effective elastic thickness of about 30 km. (4) A preliminary map of principal-stress distribution (Figure 13) is constructed for continental Australia, in which the relative magnitude of the principal stress over the continent can be assessed. The predicted stress-concentration zones in general correspond to the areas where intensive seismicity is observed. In addition, the least compression is predicted in several zones where earthquakes are relatively sparse, and it is also inferred that normal faults would mostly likely develop in these zones. (5) While the model from this study provides a reasonable interpretation for the stress orientations and seismicity observed in the Australian continent, about 20% of the observed stress orientations are not well-fitted by the model. The main reason for this could be that the disturbances in the stress field associated with some local or regional geological structures (and their present activity) cannot be simulated in our present continental-scale model.

B ORISSOVA I. & S YMOND , P. A. 1997. Basins of Australia (1:6 000 000 scale map) (1st Edition). Australian Geological Survey Organisation, Canberra. CLITHEROE G., GUDMUNDSSON O. & KENNETT B. L. N. 2000. The crustal thickness of Australia. Journal of Geophysical Research 105, 13697–13713. CLOETINGH S. & WORTEL R. 1986. Stress in the Indo-Australian plate. Tectonophysics 132, 49–67. COBLENTZ D. D., RICHARDSON R. M. & SANDIFORD M. 1994. On the gravitational potential of the Earth’s lithosphere. Tectonics 13, 929–945. COBLENTZ D. D., SANDIFORD M., RICHARDSON R. M., ZHOU S. & HILLIS R. 1995. The origins of the intraplate stress field in continental Australia. Earth and Planetary Science Letters 133, 299–309. COBLENTZ D. D., ZHOU S., HILLIS R. R., RICHARDSON R. M. & SANDIFORD M. 1998. Topography, boundary forces, and the Indo-Australian intraplate stress field. Journal of Geophysical Research 103, 919–931. CULL J. 1991. Heat flow and regional geophysics in Australia. In: Cermak V. & Rybach L. eds. Terrestrial Heat Flow and the Lithosphere Structure, pp. 486–500. Springer-Verlag, New York. DENHAM D. 1988. Australian seismicity: the puzzle of the not so stable continent. Seismological Research Letters 49, 289–295. DENHAM D., ALEXANDER L. G. & WOROTNICKI, G. 1979. Stress in the Australian crust: evidence from earthquakes and in situ stress measurements. BMR Journal of Australian Geology & Geophysics 4, 289–295. DENHAM D. & WINDSOR C. R. 1991. The crustal stress pattern in Australia continent. Exploration Geophysics 22, 101–105. FOWLER S. & MCKENZIE D. 1989. Gravity studies of the Rockall and Exmouth Plateaux using Seasat altimetry. Basin Research 2, 27–34. FITCH T. J., WORTHINGTON M. H. & EVERINGHAM I. B. 1973. Mechanisms of Australian earthquakes and contemporary stress in the Indian ocean plate. Earth and Planetary Science Letters 18, 345–356. GOETZE C. & EVANS B. 1979. Stress and temperature in the bending lithosphere as constrained by experimental rock mechanics. Geophysical Journal of the Royal Astronomical Society 59, 463–478. HILLIS R. R. 1991. Australia–Banda collision and in situ stress in the Vulcan sub-basin (Timor Sea) as revealed by borehole breakout data. Exploration Geophysics 22, 189–194. HILLIS R. R., ENEVER J. R. & REYNOLDS S. D. 1999. in situ stress field of eastern Australia. Australian Journal of Earth Sciences 46, 813–825. HILLIS R. R., MEYER J. J. & REYNOLDS S. D. 1998. The Australian stress map. Exploration Geophysics 29, 420–427. HILLIS R. R. & REYNOLDS S. D. 2000. The Australian stress map, Journal of the Geological Society of London 157, 915–921. HILLIS R. R., SANDIFORD M., COBLENTZ D. D. & ZHOU S. 1997. Modelling the contemporary stress field and its implications for hydrocarbon exploration. Exploration Geophysics 28, 88–93. HILLIS R. R. & WILLIAMS A. F. 1992. Borehole breakouts and stress analysis in the Timor Sea. Geological Society of London Special Publication 66, 157–168. HILLIS R. R. & WILLIAMS A. F. 1993a. The contemporary stress of the Barrow–Dampier Sub-basin and its implications for horizontal drilling. Exploration Geophysics 24, 567–576. HILLIS R. R. & WILLIAMS A. F. 1993b. The stress field of the North West Shelf and wellbore stability. APEA Journal 33, 373–385. KENNETT B. L. N. 1997. The mantle beneath Australia. AGSO Journal of Australian Geology & Geophysics 17, 49–54. KENNETT B. L. N. 2003. Seismic structure in the mantle beneath Australia. Geological Society of Australia Special Publication 22 and Geological Society of America Special Paper xy. KUSZNIR N. J. & BOTT M. H. P. 1977. Stress concentration in the upper lithosphere caused by underlying visco-elastic creep. Tectonophysics 43, 247–256. LAMBECK K. 1983a. Structure and evolution of the intracraton basins of central Australia. Geophysical Journal of the Royal Astronomical Society 74, 843–886. LAMBECK K. 1983b. Teleseismic travel-time anomalies and deep crustal structure in central Australia. Geophysical Journal of the

ACKNOWLEDGEMENTS We wish to thank D. Coblentz, R. Hillis and M. Sandiford for their constructive reviews, which improved this manuscript substantially, B. L. N. Kennett for kindly providing us his latest shear-wave model of the Australian lithosphere, and G. Clitheroe for providing some of the data used in this study. This research is supported by an ARC SPIRT grant and industry sponsorship by BHP, Santos, Shell and Woodside.

3D Modelling of Australian stress field Royal Astronomical Society 94, 105–124. LAMBECK K. & PENNEY C. 1984. Teleseismic travel time anomalies and crustal structure in central Australia. Physics of the Earth and Planetary Interiors 34, 46–56. LAMBECK K., MCQUEEN H. W. S., STEPHENSON R. A. & DENHAM D. 1984. The state of stress within the Australian continent. Annales Geophysicae 2, 723–742. LILLEY F. E. M., WOODS D. V. & SLOANE M. N. 1981. Electrical conductivity profiles and implications for the absence or presence of partial melting beneath central and southeast Australia. Physics of the Earth and Planetary Interiors 25, 202–209. LISTER C. R. B. 1975. Gravitational drive on oceanic plates caused by thermal contraction. Nature 257, 663–665. MCCUE K. F. & SUTTON D. J. 1979. South Australian earthquakes during 1976 and 1977. Journal of Geological Society of Australia 26, 231–236. MENKE W. 1984. Geophysical Data Analysis: Discrete Inverse Theory. Academic Press, Orlando. MÜLLER R. D., ROEST W. R., ROYER J-Y., GAHAGAN L. M. & SCLATER J. G. 1997. Digital isochrons of the world’s ocean floor. Journal of Geophysical Research 102, 3211–3214. MUELLER B., REINECKER J. & FUCHS K. 2000. The 2000 release of the World Stress Map. . PLUMB K. A. 1979a. The tectonic evolution of Australia. Earth-Science Reviews 14, 205–249. PLUMB K. A. 1979b. Structure and tectonic style of the Precambrian shields and platforms of northern Australia. Tectonophysics 58, 291–325. REYNOLDS S. D., COBLENTZ D. D. & HILLIS R. R. 2003. Influences of plate-boundary forces on the regional intraplate stress field of continental Australia. Geological Society of Australia Special Publication 22 and Geological Society of America Special Paper 372, 59–70. RICHARDSON R. M. 1992. Ridge forces, absolute plate motions, and the intraplate stress field. Journal of Geophysical Research 97, 11739–11748. RICHARDSON R. M., SOLOMON S. C. & SLEEP N. H. 1979. Tectonic stress in the plates. Reviews of Geophysics 17, 981–1019. SANDIFORD M., COBLENTZ D. D. & RICHARDSON R. M. 1995. Ridge torques and continental collision in the Indian-Australian plate. Geology 23, 653–656. SANDIFORD M. & HAND M. 1998. Controls on the locus of intraplate

89

deformation in central Australia. Earth and Planetary Science Letters 162, 97–110. SIMONS F. J. & VAN DER HILST R. D. 2002. Age-dependent seismic thickness and mechanical strength of the Australian lithosphere. Geophysical Research Letters 29, 1029–1033. SIMONS F. J., ZIELHUIS A. & VAN DER HILST R. D. 1999. The deep structure of the Australian continent from surface-wave tomography. Lithos 48, 17–43. SIMONS F. J., ZUBER M. T. & KORENAGA J. 2000. Isostatic response of the Australian lithosphere: estimation of effective elastic thickness and anisotropy using multitaper spectral analysis. Journal of Geophysical Research 105, 19163–19184. STEPHENSON R. & LAMBECK K. 1985. Erosion-isostatic rebound models for uplift: an application to southeastern Australia. Geophysical Journal of the Royal Astronomical Society 82, 31–55. TARANTOLA A. 1987. Inverse Problem Theory, Methods for Data Fitting and Model Parameter Estimation. Elsevier, Amsterdam. TURCOTTE D. L. & SCHUBERT G. 1982. Geodynamics: Applications of Continuum Physics to Geological Problems. John Wiley & Sons Inc., New York. ZHANG Y., SCHEIBNER E., ORD A. & HOBBS B. E. 1996. Numerical modelling of crustal stresses in the eastern Australian passive margin, Australian Journal of Earth Sciences 43, 161–175. ZOBACK M. D. & ZOBACK M. L. 1991. Tectonic stress field of North America and relative plate motion. In: Slemmons D. L., Engdahl E. R., Zoback M. D. & Blackwell M. L. eds. Neotectonics of North America, pp. 339–366. Geological Society of America, Boulder. ZOBACK M. L. 1992. First- and second- order patterns of stress in the lithosphere: the World Stress Map Project. Journal of Geophysical Research 97, 11703–11728. ZOBACK M. L., ZOBACK M. D., ADAMS J., ASSUMPCAO M., BELL S., BERGMAN E. A., BLUMLING P., BRERETON N. R., DENHAM D., DING J., FUCHS K., GAY N., GREGERSEN S., GUPTA H. K., GVISHIANI A., JACOB K., KLEIN R., KNOLL P., MAGEE M., MERCIER J. L., MULLER B. C., PAQUIN C., RAJENDRAN K., STEPHANSSON O., SUAREZ G., SUTER M., UDIAS A., XU Z. H. & ZHIZHIN M. 1989. Global patterns of tectonic stress. Nature 341, 291–298. ZUBER M. T., BECHTEL T. D. & FORSYTH D. W. 1989. Effective elastic thickness of the lithosphere and mechanisms of isostatic compensation in Australia. Journal of Geophysical Research 94, 9353–9367. Received 23 July 2001; accepted 29 August 2002

Geol. Soc. Australia Spec. Publ. 22, and Geol. Soc. America Spec. Pap. 372 (2003), 91–105

Principal stress orientations from multiple focal-plane solutions: new insight into the Australian intraplate stress field* D. CLARK† AND M. LEONARD Geoscience Australia, PO Box 378 Canberra ACT 2601, Australia. Stress tensor reconstructions are presented for seven domains within the Australian crust based on the formal inversion of four or more earthquake focal mechanisms in close geographic proximity. The data for inversion was sourced from a set of 70 independent quality-ranked focal mechanisms forming part of the recently compiled Geoscience Australia focal mechanism database. When analysed in conjunction with in situ stress data held by the Australian Stress Map Project, the new data make possible for the first time a rigourous comparison of the Australian continental stress field at near-surface and seismogenic depths. A more complete picture of the character of the Australian intraplate stress field is thereby made available. The tensor data agrees well with in situ determinations in western, northern and far-southeastern Australia suggesting that the continental stress field is homogeneous between shallow and seismogenic depth in these areas. Plate-boundary forces are considered to be the dominant source of stress. In contrast, the results for the Sydney Basin and Flinders Ranges imply significant heterogeneity and influence by more localised sources of stress. An apparent persistent northeast stress orientation in the Sydney Basin contrasts with the variable orientations displayed by in situ data, suggesting that the shallow stress field is dominated by near-surface effects, such as those generated by deeply incised topography. Uniformity in orientation of the seismogenic stress field is interpreted in terms of the influence of linear topographic sources of stress (the continental margin and the elevated topography of the Great Dividing Range) superimposed onto a regional stress field of low horizontal anisotropy. In the Flinders Ranges, an anomalous tensor result is interpreted in terms of perturbation of the regional stress field by a locally enhanced geotherm. Key words: earthquake focal mechanism, intraplate stress field, stress field, stress tensor.

INTRODUCTION Australian seismicity and the continental stress field In apparent contradiction to the plate-tectonics paradigm, the occurrence of intraplate earthquakes testifies to continuing internal deformation of the plates. Within the Australian Plate, earthquakes of magnitude 6.0 or greater occur on average once every five years (McCue 1990). Earthquakes of lesser magnitude, but still quite capable of causing significant social and economic impact, are far more numerous (Figure 1). The set of conditions that lead to large intraplate earthquakes in Australia are poorly understood, to the extent that almost half of the events of magnitude 5.0 or greater in the last 10 years have occurred in areas designated as of low risk on the Earthquake Hazard Map developed for the Australian Building Code (AS 1170.41993). There is no obvious one-to-one correlation between Australian earthquake epicentre locations and the crustal building blocks that form the continent (McCue 1990) (Figure 1). Except for a handful of historic fault scarps, there are no conclusive links to structure of a finer scale either.

As all seismicity stems from the action of stress upon a given mass of rock, a good knowledge of the stress distribution within the Australian Plate is fundamental to developing a successful predictive model. Significant steps have been made towards characterising the Australian continental stress field since the first release of the World stress map (Zoback et al. 1989; Zoback 1992). The Australian stress map database (Hillis & Reynolds 2000; ) currently holds over 320 stress determinations from which the maximum horizontal component of the stress field may be reliably determined. Over 70% of the data are derived from shallow in situ testing methods, which typically sample the stress field to a depth significantly less than 1 km. Stress information derived from earthquake events therefore has the potential to add to the completeness of the dataset by providing information from seismogenic depth (typically 2–15 km) free of near-surface interferences, such as those which may accompany topographic expression. † Corresponding author: [email protected] * Appendix 1 [indicated by an asterisk (*) in the text and listed at the end of the paper] is a Supplementary Paper; copies may be obtained from the Geological Society of Australia website .

92

D. Clark and M. Leonard

Figure 1 Plot of Australian earthquake epicentres 1841–2001 for events with magnitude ≥ 4.0. The dataset is complete for magnitude 6 and above from 1920 and for 5 and above from 1959. Earthquakes relating to the northern margin of the Australian Plate have been removed. The base map shows mega-element boundaries of the Australian Crustal Elements map (after Shaw et al. 1995).

Deriving stress data from earthquakes By studying the direction of movement, or polarity, of the first seismic waves from an earthquake arriving at a variety of seismograph stations distributed over the Earth’s surface, the type of earthquake (thrust, strike-slip or normal) and the geometry of the fault plane can usually be determined. The most common method of achieving this is by plotting the azimuth of each station from the epicentre, together with the angle of departure from the vertical of the direct ray from the source to the station, onto an equalarea stereographic projection (for a detailed treatment of this process see Fowler 1990 pp. 97–104). Each point so plotted is differentiated according to its polarity. Regions of positive and negative polarity are then separated into quadrants by two mutually orthogonal nodal planes, one of which represents the fault plane and the other a construct denoted the auxiliary plane. The resulting diagram is known as a focal-plane solution, or focal mechanism (Figure 2). The line bisecting the dilative dihedron of a focal mechanism is denoted the P-kinematic axis and the line bisecting the compressive dihedron the T-kinematic axis. These axes, and the line of intersection of the nodal planes (B axis), correspond very crudely to the 1, 3 and 2 principal stress axes, respectively (Ramsay & Huber 1983), and thereby provide stress information. However, a

one-to-one correlation is not possible as intact rock does not fracture at 45° to the principle stresses (for typical coefficients of internal friction the angle is between 20 and 40°), and movement along a pre-existing fault would be influenced by static friction. In general, the greater the number of stations that record an event, the more precise the resulting focal mechanism will be, and the better the stress data which may be derived from it. This places major constraints on which events can be used to construct mechanisms (cf. Figures 1 & 3). For example, in areas of sparse seismograph coverage only very large events are recorded with sufficient amplitude to be used. There are few regions in Australia where focal mechanisms may be constructed for earthquakes of a magnitude of less than about 4.0 (the Flinders Ranges in South Australia and southeastern Victoria are exceptions). Consequently, earthquakes with magnitudes greater than 4.0 comprise over 70% of compiled mechanisms. The magnitude distribution also varies with time, with larger events featuring more prominently in the early days when the seismographic network was not as extensive as today. McKenzie (1969) showed that the maximum compressive stress could have an orientation anywhere within the dilatational quadrants of a focal mechanism, as slip can occur on pre-existing planes of weakness. Hence, the principal stress directions are poorly constrained by a single

Stress from multiple focal-plane solutions

Figure 2 Basic principle of the Right Dihedron method of focal mechanism combination (after Angelier & Mechler 1977). Compressive dihedra in dark-grey, dilative dihedra in light-grey; incompatibility zones left white. Slip vectors are marked on the fault plane. P and T kinematic axes bisect the dilative and compressive dihedra and are marked as black and white dots, respectively. The line of intersection of the nodal planes is the B kinematic axis (grey dots).

fault-plane solution, and more importantly, by any number of P and T axis orientations. However, if more than four different focal mechanisms occur within a region of uniform stress, then both the principal stress directions (1>2>3 with compression positive) and a measure of the relative stress magnitudes [R = (2–3)/(1–3)] may be determined (Bott 1959; Angelier 1979; Célérier 1988). Standard procedures are now well established for stress tensor reconstruction from fault-slip data using both numerical techniques (Angelier 1984, 1989, 1991, 1994; Etchecopar et al. 1981; CareyGailhardis & Mercier 1987; Michael 1987; Gephart 1990; Hardcastle & Hills 1991; Delvaux 1993) and/or graphical techniques (Angelier & Mechler 1977; Reches 1987; Angelier 1984; Lisle 1987; Delvaux 1993). The fundamental assumption underpinning all these methods is the premise that in a body of rock under stress, slip on a plane occurs in the direction of maximum resolved shear stress (Wallace 1951; Bott 1959). Further assumptions are that stress is homogeneous and faults do not interact mechanically. The method of focal mechanism inversion is being used increasingly more often around the world to better define the crustal stress field at seismogenic depth and thereby gain an improved understanding deformation patterns and hence seismic risk (Petit et al. 1996; Baroux et al. 2001).

METHODOLOGY: OBTAINING STRESS DATA FROM THE FOCAL-MECHANISM DATABASE Data quality control This paper is accompanied by a comprehensive database of 70 independent focal-plane solutions compiled for Australian intraplate earthquakes (Appendix 1*; summarised in Table 1). The quality of each focal mechanism has been checked using the original P-wave arrival data and centroid moment tensor solution (see Dziewonski & Woodhouse 1983 for explanation) where available, and each solution given a rank according to two quality ranking systems. Source references are included in Appendix 1*. The first system is intended as a measure of the quality of the earthquake focal mechanism determination itself, and provides an estimate of the precision of the fault plane determination (Leonard et al. 2002). The data are ranked

93

from 1 to 5 according to the following criteria: 1. Tight; 2. Some uncertainty as solution too tight with overlapping polarities, non-orthogonal; 3. Uncertainty of N (B) axis less than about 15 x 15°, well constrained, or some additional information; Centroid Moment Tensor solutions; 4. Uncertainty of N axis less than about 30 x 30°, or some additional information; 5. Uncertainty of N axis greater than about 30 x 30°; X. Not rated (arrival data not available or two very different solutions possible). A measure of the quality of the tectonic stress information that can be obtained from an individual focal mechanism is then obtained by applying the World stress map ranking scheme (Zoback 1992). The data are ranked from A (best) to E (worst): A. Average P axis or formal inversion of four or more single event solutions in close geographic proximity (at least one event M≥4.0, other events M≥3.0); B. Well-constrained single event solution (M≥4.5) or average of two well-constrained single event solutions (M≥3.5) determined from first motions and other methods (e.g. moment tensor waveform modelling or inversion); C. Single event solution constrained by first motions only, often based on author’s quality assignment; M≥2.5; or Average of several well-constrained composites (M>2.0), and Moment Tensor solutions; D. Single composite solution; or poorly constrained single event solution; or single event solution for M2>3 indicates that 1 is the maximum principal compressive stress and 3 is the minimum principal compressive stress (after Zoback 1992). DALTON–GUNNING ZONE

The Dalton–Gunning Zone contains the greatest concentration of earthquake events for which focal mechanisms have been determined on the Australian continent. Nine focal mechanisms occur within 45 km of each other (Table 2). The nodal planes are distinguished from the auxiliary planes in this instance by a common southeast-plunging slip vector. This analysis is supported by linear patterns of aftershocks obtained for the 1971 and March 1974 Dalton–Gunning earthquakes (Denham et al. 1981). Graphical analysis of the data suggests that there are well-constrained common areas of compression and dilatation (Figure 4a), supporting the assumption that the mechanisms sample a region where the stress field is homogeneous. However, numerical analysis of the group resulted in the rejection of the Bowning A mechanism on the basis that the angle between the slip and resolved shear vectors was in the order of 45° (it is generally accepted that the slip deviation must be less than 30° in order for the active fault to be compatible with the calculated stress tensor). Manipulation of the fault planes within their uncertainties, and choice of the other nodal plane for the Bowning A mechanism, could not overcome the problem. In order to address the possibility that the results may be non-unique depending on the choice of mechanisms for inversion we removed arbitrary mechanisms from the entire set to test if the Bowning A result could be made compatible. The results of this exercise show that the Bowning mechanism remains incompatible, for the same reason of slip misfit found in the original calculation, if any of the other mechanisms are removed from the set. An explanation for the incompatibility of the mechanism is not readily apparent. Apart from two incompatible

97

polarities (of 25) the focal mechanism is reasonably well constrained. It is possible that the incompatibility has a geological basis, with mechanical interaction occurring between the rupturing fault and other faults. Alternatively, a magnitude 3.2 earthquake which occurred 40 km away from the Bowning A event a couple of months previously may have generated a transient perturbation of the regional stress-field orientation. The stress tensor derived from the remaining eight mechanisms is reasonably precise (±6.0%) and indicates horizontal northwest–southeast-trending compression. The very close correlation between the common dilative and compressional regions determined graphically and the numerical result based on only eight mechanisms, suggests that the stress determination is robust and not greatly influenced by the omission of the Bowning A mechanism. SOUTHWEST SEISMIC ZONE

Five focal mechanisms have been constructed for the Southwest Seismic Zone, with a maximum separation of 110 km. There is common consensus that the 1968 Meckering and 1970 Calingiri earthquakes occurred on subvertical northwest-trending structures relating to the Cape Riche–Yandanooka gravity lineament (Everingham 1966; Gordon & Wellman 1971; Fitch et al. 1973; Gordon & Lewis 1980). The appropriate fault plane on the Meckering centroid moment tensor solution and the Calingiri focal mechanism were thus determined. Fault planes on the remaining three mechanisms were chosen such that the slip vectors aligned most closely with the known two. Because of the close agreement in the orientations of the kinematic axes from all five events, the graphical combination resulted in broad bands of compatibility being identified (Figure 4b). Numerical analysis resulted in the reconstruction of a precise stress tensor (±3.8%) indicating horizontal east–west-trending pure compression. FLINDERS RANGES

Ten focal mechanisms have been compiled for the Flinders–Mt Lofty Ranges region of South Australia, reflecting events ranging from a preferred magnitude of 3.6 to 5.0. The 1997 Burra earthquake is excluded as it lies 150 km south of the nearest other mechanism. The remaining mechanisms are within 190 km of each other. In such a large and geologically complex area it is unreasonable to assume without supporting evidence that the mechanisms sample a homogeneous stress field. The population was therefore divided into northern and southern sub-populations for purposes of comparison with the whole. Five focal mechanisms occur in reasonably close proximity in the southern Flinders Ranges (Table 2), four of which record almost pure strike-slip movement. The other mechanism, constructed for the Yalpere event, suggests pure thrust faulting. Reference to the 1:250 000 scale geological maps of the area (e.g. Orroroo SI 54-1, Parachilna SH54-13) reveals complex structural geology with fault orientations closely matching both the nodal plane orientations of the solutions. Comparison of slip vectors reveals no consistent trend, possibly reflecting the quality of the focal mecha-

6

4

4

8

8/9

4

5

5

All Flinders

North Sydney Basin

South Sydney Basin

Sydney Basin (all)

Dalton– Gunning

Snowy Mountains

SW seismic Zone

Tennant Creek

30

110

85

45

200

50

140

190

150

30

190

310

500

255

11650

1170

1490

13430

4080



69

84

69

62

66

73

57

22

7

78

Plunge

273

146

2

126

266

199

273

348

341

151

Az.

Sig3

18

4

20

5

19

14

27

67

78

12

Plunge

127

11

198

26

126

343

132

148

216

335

Az.

11

4

5

27

15

10

18

7

10

1

Plunge

33

281

106

293

31

75

33

255

72

245

Az.

Graphical results Sig2 Sig1

0.20

0.20

0.50

0.44

0.75

0.75

0.50

0.67

1.00

0.00

R

11

16.4

13.7

36.1

14.6

22.2

27.7

17.2

13.8

2.55

SD

79

70

76

81

72

87

68

1

1

24

238

7

55

87

217

215

224

172

170

159

Plunge Az.

Sig3

5

20

14

4

1

0

1

65

63

66

Plunge

122

185

224

200

310

306

133

264

262

343

Az.

10

0

3

8

18

4

22

25

27

1

Plunge

R

0.23

0.63

0.21

0.64

0.52

31

0.13

275 0.47

315 0.52

290 0.29

40

35

43

82

79

250 0.03

Az.

Tensor reconstruction Sig2 Sig1

6.6

3.8

3.2

6.1

6.4

9.3

5.7

6.2

2.7

0.65

SD

S1 > S2 > S3 where compressive stress is positive in sign. R = (S2-S1) / (S3-S1). SD = the standard deviation of counting deviation values (counting deviation expresses the divergence of individual data relative to the general dataset). Max. dist. Is a measure of the distance between the furthest separated FPS (focal plane solution) in the set. Area is a measure of the area enclosed/sampled by the set of FPS. ASM = Australian stress map (Hillis & Reynolds 2000). Stress regime classification after Zoback (1992).

4

South Flinders

1) 2) 3) 4) 5) 6) 7)

2

No. of Max. Area FPS dist. (km2) (km)

North Flinders

Location

Table 2 Results of graphical and numerical combination.

A

A

A

A

A

A

A

A

A

B

strike-slip compressive (excludes 620610 and 620611)

pure compressive

pure compressive

pure compressive (tensor excludes Bowning 1977a)

strike-slip compressive

pure compressive

strike-slip compressive

pure strike-slip (excludes Moralana, Yalpere & Wilpena)

pure strike-slip (excludes Yalpere)

compressive strike-slip (excludes Moralana and Wilpena)

ASM Stress regime rank (Comments)

98 D. Clark and M. Leonard

Stress from multiple focal-plane solutions

99

Table 3 Dalton–Gunning Zone focal-mechanism data. Name of event

Date

Dalton–Gunning Dalton–Gunning Dalton–Gunning Dalton–Gunning Dalton–Gunning Dalton–Gunning Bowning A Bowning B Oolong

03/11/1971 08/01/1973 22/03/1974 06/05/1974 04/08/1975 05/04/1977 30/06/1977 04/07/1977 09/08/1984

Fault Plane Azimuth Dip 312 303 73 322 104 79 119 102 346

50 80 58 55 70 50 50 64 45

nisms. Fault planes were therefore discriminated in this instance with the TENSOR program using the previously described misfit function. While graphical combination of the five mechanisms did produce a reasonably well-constrained result (indicating a strike slip stress regime with an east–west compression direction), the Yalpere mechanism was incompatible with the strike-slip mechanisms when subjected to numerical analysis. Both nodal planes of the Yalpere mechanism produced excessive slip deviations, requiring the mechanism to be discarded. The uniqueness of this result was tested in a similar fashion as was the Dalton–Gunning set, with the same results. The Yalpare mechanism could not be made compatible by the arbitrary removal of any other mechanism in the set. Graphical combination of the four remaining strike-slip results defined very well-constrained regions of common compression and dilatation (Figure 4c). Numerical analysis produced a very precise stress tensor (±2.7%) suggestive of pure strike-slip motion with an east-northeast compression axis. Implicit in this result is that the maximum and minimum compression axes (1 and 3) lie within the horizontal plane, in contrast to the other results presented herein. In the northern Flinders Ranges four focal mechanisms occur in reasonably close proximity (Table 2). Two populations with incompatible stress regimes are evident, as was the case in the south: a strike-slip population (Blinman and Beltana) and a pure thrust population (Moralana and Wilpena). The Wilpena mechanism was discarded from the thrust population on the basis of excessive uncertainty in the orientation of the nodal planes (cf. Greenhalgh et al. 1994 figure 7b). The graphical combination of the two strike-slip mechanisms provides little constraint on the plunge of the 1 and 3 axes but does tightly constrain their azimuths, which are consistent with those determined for the southern Flinders Ranges (cf. Figures 4c and d). The population contains too few data to perform a stress inversion as at least four differently oriented fault planes are require to solve for the four unknowns in the stress tensor (stress axes and stress ratio). The orthogonalised results of the Right Dihedron analysis, which are a rough approximation of the principal stress axes, are therefore presented in Figure 4d. The coincidence of 1 and 3 axes azimuth for the northern and southern Flinders Ranges strike-slip subsets supports the possibility that a homogeneous stress field

Auxiliary Plane Azimuth Dip 204 212 180 217 212 187 210 205 200

70 82 65 70 50 70 88 64 50

Fault Pick Aftershocks Slip vector Aftershocks Slip vector Slip vector Slip vector Slip vector Slip vector Slip vector

acts over the entire area sampled by the mechanisms. Combination of the sets yields well-constrained areas of compatibility and a reasonably precise stress tensor (±7.8%) indicative of pure strike-slip motion (Table 2; Figure 5a). The azimuth of the compression axis of this result is very similar to the northern and southern sets (within 12° and 3° respectively). Because the data for the northern Flinders Ranges are too few to define an individual stress province (cf. Hillis & Reynolds 2000), the combination for the entire Flinders Ranges is the preferred result (Figure 6). Thrust mechanisms in the Flinders Ranges are too few and too widely spaced to reconstruct a reliable stress tensor. The significance of this subordinate class of mechanism is considered in the discussion. TENNANT CREEK

Seven focal mechanisms within 30 km of each other were compiled for earthquakes in the Tennant Creek area between 1988 and 1999. Five of the mechanisms reflect reverse faulting with a varying strike slip component and roughly northeast–southwest compression axes. The remaining two mechanisms, both relating to 1991 normalfaulting events, have extension axes closely coincident with the compression axes of the other five mechanisms. The general geometry of the Tennant Creek fault scarp is suggestive of a northwest–southeast-trending, southwestdipping failure surface upon which north-northeast thrust movement occurred (Bowman et al. 1990; Crone et al. 1992). Fault planes of the strike-slip/thrust mechanisms were consequently determined based upon the closest fit to these criteria. Both nodal planes were tested as fault planes for the normal-faulting mechanisms. Graphical combination of the entire set failed to find areas of common dilatation and compression. Similarly, numerical analysis resulted in the normal-faulting mechanisms being discarded on the basis of excessive slip misfit (i.e. an incompatible faulting regime). With one exception, arbitrary removal of other mechanisms from the complete set did not result in the normal-faulting mechanisms becoming compatible. Removing the 1999 mechanism resulted in the two normal-faulting mechanisms becoming numerically compatible. However, the 1 axis of the tensor result obtained has a plunge of 35°, which is geologically unrealistic given the

100

D. Clark and M. Leonard

shallow depth of the earthquakes (Anderson 1951; Zoback 1992). This result is therefore not preferred. Graphical analysis of the set of five thrust/strike-slip solutions resulted in the definition of very well-constrained regions of compatibility (Figure 5b). Subsequent numerical analysis produced a reasonably precise stress tensor (±6.6%) suggestive of northeast–southwest compression with an element of strike-slip. Interestingly, the maximum horizontal compression direction of the previously mentioned suspect result was within 10° of the preferred result, suggesting that the maximum horizontal compression direction is relatively insensitive to choice of mechanism. SYDNEY BASIN

Eight focal mechanisms occur within the geographic bounds of the Sydney Basin, within a maximum distance of about 200 km of each other. In situ stress measurements suggest that the stress field within the Sydney Basin is highly variable in orientation and shows no consistent trends at a scale of greater than ~100 km (Hillis et al. 1999). The population of mechanisms was therefore divided into two subsets of four mechanisms each: one forming a rather loose group in the northern part of the basin and the other a good concentration in the south. The northern subset comprises focal-plane solutions located within 140 km of each other, occupying a broad band between Newcastle and Lithgow. The Lithgow mechanism differs from the others in that it records a mildly dilational strike-slip failure. Second-order topographic stresses (Lambeck et al. 1984) could be locally important in this area given that the epicentre is in the middle of the Great Dividing Range. The mechanism, while questionable, is however graphically compatible with the other three and so was included in stress calculations. The discrimination of fault planes from this population was complicated by a lack of consensus as to whether seismicity occurring within the geographic bounds of the basin originates from the basin itself or from the Lachlan Fold Belt rocks beneath. Even for the relatively well-studied Newcastle 1989 earthquake, uncertainty persists as to which nodal plane of the mechanism represents the fault plane (Huftile et al. 1999). In the absence of unequivocal geological or geophysical evidence, fault-plane discrimination in our analysis is based upon a common northeast–southwest alignment of slip vectors. Graphical analysis resulted in quite tightly constrained regions of compatibility (Figure 5c), which are strongly controlled by the Lithgow mechanism. Numerical analysis produced a relatively precise stress tensor (±5.7%) indicative of northeast–southwest compression with a component of strike-slip motion. If the Lithgow mechanism is removed from the set, Right Dihedron Analysis suggests a consistent, but slightly more northerly maximum compression direction (~16° cf. 43°). The southern subset is much more tightly clustered than the solutions in the north, with events being located within 50 km of each other. A linear pattern of aftershocks following the 1961 Robertson earthquake reveals the steeply dipping, northwest-trending nodal plane to be the fault (Denham 1980). Master event analysis conducted on the Appin 1981 event supports the east-dipping nodal plane

(Denham et al. 1982). Both planes have east-plunging slip vectors. Fault planes were discriminated on the remaining two solutions by a coincidence of slip vectors. Graphical analysis of the set only loosely constrains the locations of the principal axes due to the close coincidence of P and T kinematic axes (Figure 5d). The results of the stress inversion reflect this, with a relatively large imprecision of ±9.3%. The results are consistent with pure northeast–southwest compression. SNOWY MOUNTAINS

Four focal mechanisms occur within 85 km of each other in the Snowy Mountains region. There is good evidence from patterns of aftershocks that the 1971 Middlingbank earthquake resulted from rupture of an east–west-trending subvertical fault (Bock & Denham 1983). Holding the corresponding plane in this focal mechanism fixed, the faults for the remaining mechanisms were chosen numerically to minimise the misfit of their slip vectors. Northeast-trending planes were chosen in each case. Graphical combination of the solutions resulted in a fairly loosely constrained result (Figure 5e), reflecting the combination of thrust and strike-slip mechanisms. In contrast, numerical analysis produced a very precise result (±3.8%) suggestive of pure northwest–southeast compression.

Quality and uniqueness of the tensor results The average standard deviation between the stress tensors reported here and the individual data points is about 5%, which is considerably less than the uncertainties of the focal mechanism determinations. Therefore, in each modelled area, significant stress inhomogeneities are not required by the data, i.e. the assumption of uniform stress within each stress domain is supported. According to the quality ranking system used by the World and Australian stress map projects (Zoback 1992; Hillis & Reynolds 2000), the fault-slip inversion data presented here should be assigned an A ranking. The orthogonalised Right Dihedron result presented for the northern Flinders Ranges is indicative only and should be assigned a B rank according to the above system. It should be qualified that the above precision was attained by removing incompatible mechanisms from several of the studied sets. In order to address the possibility that the results may be non-unique depending on the choice of mechanisms for inversion we graphically and numerically tested a great number of combinations for each case where a potential incompatibility was identified (e.g. Dalton–Gunning, Flinders Ranges, Tennant Creek). While a rigourous statistical analysis was not performed, the results set out in the previous section plainly indicate that removing arbitrary mechanisms from a given set did not in general increase the compatibility of mechanisms identified as incompatible. Further tests showed that removing mechanisms from a compatible set has little apparent effect on the orientation of the maximum horizontal compression direction (the change is certainly within the error of the determinations). These results suggest that the compatibilities/incompatibilities identified in the graphical and statistical analyses are likely

Stress from multiple focal-plane solutions

101

to be real. We therefore propose a geologic explanation for each omission, as discussed in the following section.

Comparison with existing (mainly in situ) stress orientation data: tectonic and continental stress field implications

DISCUSSION

Average maximum horizontal stress trajectory lines compiled using A–D quality data from the Australian stress map database (Hillis & Reynolds 2000) are plotted onto Figure 6 along with our new data, and the P-axis orientations from isolated focal mechanisms not falling within a stress domain. Because 76% of the stress data used to construct the trajectory lines is derived from depths of 5 km or less, and focal-mechanism data (24%) are typically accorded very low weighting, the lines can be treated as essentially representing the near-surface stress field. Our new data therefore have the potential to reveal variations in the stress field between the near-surface and at seismogenic depth (generally 2–15 km), and also to check the veracity of numerically modelled stress fields based on plate boundary forces (Cloentingh & Wortel 1986; Coblentz et al. 1995, 1998). There is generally very good agreement between the trajectory lines and tensor results in western, southern and northern Australia. This is an important result when it is considered that the in situ data used to construct the trajectories in these areas were derived mainly from shallow sedimentary basins (Hillis & Reynolds 2000). In contrast, the tensor reconstructions were derived from data sources within the crystalline rocks forming the basement to the basins. This not only implies a uniform stress field between shallow and seismogenic depths, but that the basins in which the in situ data were collected do not have a measurable influence on the first-order stress field. As noted by Hillis and Reynolds (2000) and Reynolds and Hillis (2000), the major crustal element boundaries (Figure 3) also appear to have no significant influence on the first-order stress field. The implication is therefore that the major crustal element boundaries, in addition to the basins, do not represent zones of relative weakness or strength (McKinnon & Garrido de la Barra 1998) at the scale of investigation. The reason for this is not obvious, especially considering that some large-scale structures, the Darling Fault for example (Figure 3), are in a state of obvious gravitational inequilibrium (Lambeck et al. 1984). The orientation of the stress field in western and northern Australia is well accounted for by modelling of plateboundary forces (Coblentz et al. 1995, 1998). Ridge push from the southwestern margin of the Australian Plate is dominant in western Australia and produces the roughly east–west compression seen in the Southwest Seismic Zone. Further to the north, plate motion resistive stress arising at the Papua New Guinea collisional boundary becomes more important, resulting in a counterclockwise rotation to a more north–south orientation. The success of the modelling, which assumes a purely elastic rheology (constant Young’s modulus and Poisson’s ratio) and an absence of defects over the entire Australian Plate, provides independent evidence that the major crustal elements and basins are ‘invisible’ to the continental stress field in this area. The Dalton–Gunning Zone and Snowy Mountains tensor results show a continuation of the northwest–southeast trend seen in the Otway and Cooper Basins (Figure 6). The

Heterogeneity in the focal-mechanism population within stress domains Of the seven stress domains considered in this study, two show a strongly bimodal distribution of focal mechanisms—the Flinders Ranges and Tennant Creek domains. The Flinders Ranges mechanisms are an incompatible combination of strike-slip and pure thrust motion, which is suggestive of an approximate equivalence of 3 and 2 magnitudes in this case. Indeed, if all mechanisms are subjected to Right Dihedron analysis, a stress ratio [R = (2-3)/(1-3)] of 0.07 obtains. The strike-slip set, when considered in isolation, is more stable, with an R value of 0.60 (Table 2). This perhaps explains the dominance of this type of mechanism within the domain. This domain is unique in that it is the only one where almost pure strike-slip movement is implied by a significant proportion of focal mechanisms studied (two thirds of the population). It also contains some of the most youthful topography in Australia (Tokarev et al. 1998; Sandiford 2002), and is underlain by crystalline basement of unusually high heat production (Sandiford et al. 1998; Neumann et al. 2000). Exploration of the possibility that the observed stress regime (and topography) is influenced by the local anomalous thermal structure of the crust is beyond the scope of this contribution. However, considerable stress refraction might be expected due to thermally related mechanical contrasts in and around the Flinders Ranges (M. Sandiford pers. comm. 2002). The Tennant Creek mechanisms also reveal a quite complex stress regime within the 190 km2 region sampled. The five mechanisms used to reconstruct the stress tensor range from strongly strike-slip to moderately thrust geometry. This distribution is reflected in the stress ratio of 0.13 (Table 2), which in a similar fashion to the Flinders Ranges results, indicates that the 3 and 2 axes are roughly equivalent in magnitude and may exchange. As previously mentioned, the two Tennant Creek normal-faulting mechanisms suggest a strong component of normal faulting almost coincident with the compression axis of the dominant set for which the stress tensor was constructed. However, they are compatible with the preferred geological model; that of a northwest-trending southwestdipping fault zone upon which roughly north-northeast thrusting occurred. As the main fault system and scarp are not perpendicular to the calculated maximum horizontal compression direction, a component of dextral lateral movement is introduced into the motion. In this geologic context, the two mechanisms could relate to earthquakes on an east–west-oriented releasing bend in the main fault zone. An alternative, and perhaps less likely, explanation for the extensional events is that they reflect failure within a perturbed transient stress field arising from the three main shocks three years previously. If this was the case, extension appears to have been short-lived with a re-establishment of the compressive stress field within 10 years.

102

D. Clark and M. Leonard

Figure 6 Plot of stress orientation at the seven locations for which reduced stress tensors were determined. The stress tensors are displayed in map view by symbols representing the orientation and magnitude of the horizontal and vertical stress axes (Guiraud et al. 1989). The heavy black arrows represent the maximum horizontal stress direction (Shmax). P-axis orientations from isolated focal mechanisms not falling within the stress domains are also shown. Dashed lines are stress trajectories calculated by Hillis & Reynolds (2000) using the Australian stress map database. Red arrows are the mean Shmax directions for stress domains from the same database.

trend is further continued off the coast to the east in the Gippsland Basin and into the Tasman Sea (i.e. P-axis orientations of individual focal mechanisms are consistent). This arrangement demonstrates that stresses associated with the topographic relief of the Australian alps, as predicted by Lambeck et al. (1984), and any margin-parallel continental margin effects of the type described by Bott and Dean (1972) and Stein et al. (1989), have little influence on the regional trend. As previously noted by Hillis et al. (1999) on the basis of in situ data alone, this situation suggests that the stress field in southeastern Australia is controlled by the action of plate-boundary forces, most likely those arising from the New Zealand collisional boundary (Coblentz et al. 1995, 1998). The orientation of the principal horizontal stress and the relationship between stress and crustal structure is not so straightforward in South Australia and eastern Australia. There is significant deviation from the mean trajectory lines in Hillis and Reynolds’ (2000) analysis, especially in the Sydney Basin. The modelling of Coblentz et al. (1995, 1998) is similarly afflicted. These more complex areas form the basis for the following discussion.

SOUTH AUSTRALIA

In the Flinders Ranges, the stress data consist almost entirely of isolated individual focal mechanisms and our focal mechanism inversion (Table 2). Neotectonic geologic indicators (Sandiford 2002) also provide indirect evidence for the stress field orientation. Both instrumental and geologic datasets are consistent with a roughly east–west maximum horizontal compression direction. It is contended on the basis of our analyses that the bimodal population of focal mechanisms in the Flinders area is indicative of a near equivalence in magnitude of vertical and minimum horizontal compressive stresses (2 and 3 respectively). We find no indication in the focal mechanism data to support the assertion of Hillis and Reynolds (2000) that the heterogeneity observed in the Flinders Ranges data is due to relatively low horizontal force anisotropy. Indeed, the tensor result lends weight to the opposite postulate: that there is significant horizontal force anisotropy. The maximum horizontal compression direction implied by the tensor result is broadly consistent with in situ deter-

Stress from multiple focal-plane solutions

minations from the nearby Otway and Cooper Basins (Figure 6). However, the regional significance of the Flinders data is questionable given the possibility that the stress field in this domain is influenced by local geologic conditions (as discussed previously). Additional stress data from the surrounding area would be required in order to clarify this uncertainty.

103

sive in the south of the Sydney Basin to strike-slip compressive in the north (Table 2) may be an artefact of the small sample space. However, it is consistent with the above hypothesis, as it requires that the regional stress field becomes more isotropic as the centre of the area of divergence is approached. East-oriented dilative topographic forces will therefore have a greater influence in the north relative to the south of the Sydney Basin.

EASTERN AUSTRALIA

The area of transition between northerly and southeasterly stress trajectories in eastern Australia has been termed the ‘area of divergence’ by Hillis and Reynolds (2000). Close to the vertex of this area lies the Sydney Basin, which is characterised by a significantly more variable in situ stress distribution than the rest of Australia (Zoback et al. 1989; Zoback 1992; Zhang et al. 1996; Hillis et al. 1999; Hillis & Reynolds 2000). The variability has been explained in terms of local sources of stress being important in an area where the force balance acting along the plate boundary results in relatively low horizontal force anisotropy. The modelling of Coblentz et al. (1998) suggested that low horizontal anisotropy may also be accompanied by a reduction in the absolute magnitude of the stress. On the basis of measurements collected in the upper 1 km of the crust, Hillis and Reynolds (2000) concluded that while in some areas of the Sydney Basin there is a significant unimodal mean stress orientation at the 100 kmscale, there is no consistency at a scale greater than 100 km. In contrast to these findings, our focal-mechanism data do not require significant heterogeneity in the maximum compressive stress direction at seismogenic depth. The tensors are consistent with northeast–southwest maximum horizontal compression throughout the basin, differing in azimuth by only 8° from north to south. That being said, there are many independent in situ measurements from various locations in the Sydney Basin and only eight focal mechanisms. It is therefore possible that the apparent uniformity in the stress tensor results is a sampling artefact. As this possibility cannot be tested until more moderate to large earthquakes occur in the Sydney Basin, we will explore the implications of assuming our tensor result to be valid. In this case, the need to appeal to a multitude of local stresses, such as those associated with deeply incised topography or basin structure, in order to explain variations in trend (Zhang et al. 1996; Hillis & Reynolds 2000) is avoided. If the postulates of low regional stress anisotropy (Hillis & Reynolds 2000) and magnitude (Coblentz et al. 1998) are accepted, then a local source of stress that is uniform in orientation over the entire Sydney Basin must be appealed to in order to obtain the observed effect. The most likely linear sources are the Australian alps and the continental margin. The east–west extensional stresses associated with the topographic relief of the alps (Lambeck et al. 1984) will tend to reduce the magnitude of the collinear component of the regional stress field, while margin-parallel continental margin effects (Bott & Dean 1972; Stein et al. 1989) will reinforce the northerly component. The observed stresses are therefore well accounted for by this combination. The minor change in stress regime from pure compres-

CONCLUSIONS (1) Seventy independent focal-plane solutions are presented in a database providing information on the continental stress field between ~2 and 15 km depth. The original P-wave arrival data and the focal mechanisms are currently in the process of being digitised and incorporated into an on-line version of the database. It is anticipated that the on-line database will be available by early 2003. (2) Seven stress provinces are defined on the basis of compatible focal mechanisms. The provinces comprise a minimum of four independent focal mechanisms within a distinct geographic region and within 200 km of each other. (3) The calculated maximum horizontal stress directions for each domain are broadly consistent with those derived mainly from near-surface in situ data in the Australian stress map (Hillis & Reynolds 2000). An important area of difference is in the Sydney Basin. Where the shallow in situ data show significant scatter, our seismogenic data are consistent with a northeast-trending maximum compressive stress orientation at depth. (4) In contrast to in situ determinations, we find that at seismogenic depth there is an apparent northeast–southwest stress trend throughout the entire Sydney Basin area. The trend is explained in terms of local sources of stress induced by adjacent high relief and the continental margin overriding a locally low-magnitude and relatively isotropic regional field. (5) Our data for the Flinders Ranges suggest a nearequivalence of vertical and minimum horizontal compressive stresses (2 and 3 respectively). Implicit in this relationship is that the maximum horizontal compressive stress is well determined by the data. However, further work is required to determine whether the result represents the regional stress trend, or is influenced by local sources of stress.

ACKNOWLEDGEMENTS We would like to thank Kevin McCue and John Schneider for their constructive comments on early versions of the manuscript. Li Yu and Ian Ripper are thanked for their work in tracking down the focal mechanisms used to populate the database and checking their veracity. Thanks also to Damien Delvaux for updating and allowing us the use of his TENSOR program for calculation of the stress tensors. The manuscript benefited significantly from the constructive comments of Mike Sandiford and an anonymous reviewer. Thanks to them. This contribution is published with the permission of the Chief Executive Officer of Geoscience Australia.

104

D. Clark and M. Leonard

REFERENCES ANDERSON E. 1951. The Dynamics of Faulting and Dyke Formation with applications to Britain (2nd edition). Oliver and Boyd, Edinburgh. ANGELIER J. & MECHLER P. 1977. Sur une methode graphique de recherche des constraintes principales egalament utilisable en tectonique et en seismologie : la methode des diedres droits. Bulletin of the Geological Society of France 7, 1309–1318. ANGELIER J. 1979. Determination of the mean principal directions of stresses for a given fault population. Tectonophysics 56, T17–T26. ANGELIER J. 1984. Tectonic analysis of fault slip data sets. Journal of Coastal Research 89, 5835–5848. ANGELIER J. 1989. From orientation to magnitudes in palaeostress determinations using fault slip data. Journal of Structural Geology 11, 37–50. ANGELIER J. 1991. Inversion directe et recherché 4-D: comparaison physique et mathematique de deux methodes de determination des tenseurs des paleocontraintes en tectoniquede failles. Comptes Rendus de l’ Academie des Sciences, Paris 312, 1213–1218. ANGELIER J. 1994. Fault slip analysis and paleostress reconstruction. In: Hancock P. L. ed. Continental Deformation, pp. 101–120. Pergamon, Oxford. BAROUX E., BETHOUX N. & BELIER O. 2001. Analyses of the stress field in southeastern France from earthquake focal mechanisms. Geophysical Journal International 145, 336–348. BOCK G. & DENHAM D. 1983. Recent earthquake activity in the Snowy Mountains region and its relationship to major faults. Journal of the Geological Society of Australia 30, 423–429. BOTT M. H. P. & DEAN D. S. 1972. Stress systems at young continental margins. Nature 235, 23–25. BOTT M. H. P. 1959. The mechanics of oblique slip faulting. Geological Magazine 96, 109–117. BOWMAN J. R., GIBSON G. & JONES T. 1990. Aftershocks of the 1988 January 22 Tennant Creek, Australia intraplate earthquakes: evidence for a complex thrust-fault geometry. International Geophysical Journal 100, 87–97. CAREY-GAILHARDIS E. & MERCIER J. L. 1987. A numerical method for determining the state of stress using focal mechanisms of earthquake populations: application to Tibetan teleseisms and microseismicity of southern Peru. Earth and Planetary Science Letters 82, 165–179. CÉLÉRIER B. 1988. How much does slip on a reactivated fault plane constrain the stress tensor? Tectonics 7, 1257–1278. CLOETINGH S. & WORTEL R. 1986. Stress in the Indo-Australian plate. Tectonophysics 132, 49–67. COBLENTZ D. D., SANDIFORD M., RICHARDSON R. M., SHAOHUA Z. & HILLIS R. 1995. The origins of the intraplate stress field in continental Australia. Earth and Planetary Science Letters 133, 299–309. COBLENTZ D. D., ZHOU S., HILLIS R. R., RICHARDSON R. M. & SANDIFORD M. 1998. Topography boundary forces and the Indo-Australian intraplate stress field. Journal of Geophysical Research 103, 919–938. CRONE A. J., MACHETTE M. N. & BOWMAN J. R. 1992. Geologic investigations of the 1988 Tennant Creek, Australia, earthquakes— implications for paleoseismicity in stable continental regions. US Geological Survey Bulletin 2032-A. DELVAUX D. 1993. The TENSOR program for paleostress reconstruction; examples from the East African and the Baikal rift zones. Terra Abstracts 5, 216. DENHAM D. 1980. The 1961 Robertson earthquake—more evidence for compressive stress in southeast Australia. BMR Journal of Australian Geology & Geophysics 5, 153–156. DENHAM D., BOCK G. & SMITH R. S. 1982. The Appin (New South Wales) earthquake of 15 November 1981. BMR Journal of Australian Geology & Geophysics 7, 219–223. DENHAM D., WEEKES J. & KRAYSHEK C. 1981. Earthquake evidence for compressive stress in the southeast Australian crust. Journal of the Geological Society of Australia 28, 323–332. DZIEWONSKI A. & WOODHOUSE J. 1983. An experiment in the systematic study of global seismicity: centroid-moment tensor solutions for 201 moderate and large earthquakes of 1981, Journal of Geophysical Research 88, 3247–3271. ETCHECOPAR A., VASSEUR G. & DAIGNIERES M. 1981. An inverse problem

in microtectonics for determination of stress tensors from fault striation analysis. Journal of Structural Geology 3, 51–65. EVERINGHAM I. 1966. The seismicity of Western Australia. Bureau of Mineral Resources Record 1966/127. FITCH T. J., WORTHINGTON M. H. & EVERINGHAM I. B. 1973. Mechanisms of Australian earthquakes and contemporary stress in the Indian Ocean plate. Earth and Planetary Science Letters 18, 345–356. FOWLER C. M. R. 1990. The Solid Earth: an Introduction to Global Geophysics. Cambridge University Press, Cambridge. GEPHART J. W. 1985. Principal stress directions and the ambiguity in fault plane identification from focal mechanisms. Bulletin of the Seismological Society of America 75, 621–625. GEPHART J. W. 1990. Stress and the direction of slip on fault planes. Tectonics 9, 845–858. GORDON F. R. & LEWIS J. D. 1980. The Meckering and Calingiri earthquakes October 1968 and March 1970. Geological Survey of Western Australia Bulletin 126. GORDON F. R. & WELLMAN H. W. 1971. A mechanism for the Meckering earthquake. Royal Society of New Zealand Bulletin 9, 95–96. GREENHALGH S. A., LOVE D., MALPAS K. & MCDOUGAL R. 1994. South Australian Earthquakes, 1980–92. Australian Journal of Earth Sciences 41, 483–495. GUIRAUD M., LABORDE O. & PHILLIP H. 1989. Characterisation of various types of deformation and their corresponding deviatoric stress tensor using microfault analysis. Tectonophysics 170, 289–316. HARDCASTLE C. H. & HILLS L. S. 1991. Brute3 and Select: QuickBasic 4 programs for determination of stress tensor configurations and separation of heterogeneous populations of fault-slip data. Computers & Geosciences 17, 23–43. HILLIS R. R. & REYNOLDS S. D. 2000. The Australian stress map. Journal of the Geological Society of London 157, 915–921. HILLIS R. R., ENEVER J. R. & REYNOLDS S. D. 1999. in situ stress field of eastern Australia. Australian Journal of Earth Sciences 46, 813–825. HUFTILE G. J., VAN ARSDALE R. & BOYD R. 1999. Pleistocene faulting offshore Newcastle, New South Wales. Proceedings of the 1999 Conference, Sydney, paper no. 17. Australian Earthquake Engineering Society, Sydney. JUPE A. 1993. How well are principal stress axes constrained by nonunique fault plane solution data? Geophysical Journal International 115, 319–322. LAMBECK K., MCQUEEN H. W. S., STEPHENSON R. A. & DENHAM D. 1984. The state of stress within the Australian continent. Annales Geophysicae 2, 723–741. LEONARD M., RIPPER I. D. & LI Y. 2002. Australian earthquake fault plane solutions. Geoscience Australia Record 2002/19. LISLE R. J. 1987. Principal stress orientations from faults: an additional constraint. Annales Tectonicae 1, 155–158. MCCUE K. 1990. Australia’s large earthquakes and Recent fault scarps. Journal of Structural Geology 12, 761–766. MCKENZIE D. P. 1969. The relationship between fault plane solutions for earthquakes and the directions of the principal stresses. Bulletin of the Seismological Society of America 59, 591–601. MCKINNON S. D. & GARRIDO DE LA BARRA I. 1998. Fracture initiation, growth and effect on stress field: a numerical investigation. Journal of Structural Geology 20, 1673–1689. MICHAEL A. J. 1987. Use of focal mechanisms to determine stress: a control study. Journal of Geophysical Research 92, 357–368. NEUMANN N., SANDIFORD M. & FODEN J. 2000. Regional geochemistry and continental heat flow: implications for the origin of the South Australian heat flow anomaly. Earth and Planetary Science Letters 183, 107–120. PETIT J. P., DEVERCHERE J., HOUDRY-LEMONT F., SANKOV V. A., MELNIKOVA V. I. & DELVAUX D. 1996. Present-day stress field changes along the Baikal Rift and tectonic implications. Tectonics 15, 1171–1191. RAMSAY J. G. & HUBER M. I. 1983. The Techniques of Modern Structural Geology, Volume 2. Academic Press Inc., London. RECHES Z. 1987. Determination of the tectonic stress tensor from slip along faults that obey the coulomb yield condition. Tectonics 6, 849–861. REYNOLDS S. D. & HILLIS R. R. 2000. The in situ stress field of the Perth Basin, Australia. Geophysical Research Letters 27, 3421–3424. SANDIFORD M. 2002. Neotectonics of southeastern Australia: linking the Quaternary faulting record with seismicity and in situ stress.

Stress from multiple focal-plane solutions Geological Society of Australia Special Publication 22 and Geological Society of America Special Paper 372, 107–119. SANDIFORD M., PAUL E. & FLÖTTMANN T. 1998. Sedimentary thickness variations and deformation intensity during basin inversion in the Flinders Ranges, South Australia. Journal of Structural Geology 20, 1721–1731. SBAR M. L. 1983. An explanation for contradictory geodetic strain and fault plane solution data in western North America. Geophysical Research Letters 10, 177–180. SHAW R. D., WELLMAN P., GUNN P., WHITAKER A. J., TARLOWSKI C. & MORSE M. 1995. ‘Australian crustal elements’ map; a geophysical model for the tectonic framework of the continent. AGSO Research Newsletter 23, 1–3. STEIN S. S., CLOETINGH S., SLEEP N. H. & WORTEL R. 1989. Passive margin earthquakes, stresses and rheology. In: Gregersen S. & Basham P. eds. Earthquakes at North Atlantic Passive Margins: Neotectonics and Postglacial Rebound, pp. 231–259. NATO Advanced Study Institute Series C: Mathematical and Physical Sciences 266. TOKAREV V., DANDIFORD M. & GOSTIN V. 1998. Landscape evolution in the Mount Lofty Ranges: implications for regolith development. In: Taylor G. & Pain C. eds. New Approaches to an Old Continent, 3rd Australian Regolith Conference Proceedings, Regolith 98, pp. 127–134. Cooperative Research Centre for Landscape Evolution and Mineral Exploration, Perth. WALLACE R. E. 1951. Geometry of shearing stress and relation to faulting. Journal of Geology 59, 118–130.

105

ZHANG Y., SCHEIBNER E., Ord A. & Hobbs B. E. 1996. Numerical modelling of crustal stresses in the eastern Australian passive margin. Australian Journal of Earth Science 43, 161–175. ZOBACK M. D. 1992. First- and second-order patterns of stress in the lithosphere: the world stress map project. Journal of Geophysical Research 97, 11703–11728. ZOBACK M. L. & ZOBACK M. D. 1980. State of stress in the conterminous United States. Journal of Geophysical Research B85, 6113–6156. ZOBACK M. L., ZOBACK M. D., ADAMS J., ASSUMPCAO M., BELL S., BERGMAN E. A., BLUEMLING P., BRERETON N. R., DENHAM D., DING J., FUCHS K., GAY N., GREGERSEN S., GUPTA H. K., GVISHIANI A., JACOB K., KLEIN R., KNOLL P., MAGEE M., MERCIER J. L., MUELLER B. C., PAQUIN C., RAJENDRAN K., STEPHANSSON O., SUAREZ G., SUTER M., UDIAS A., XU Z. H. & ZHIZIN M. 1989. Global patterns of tectonic stress. Nature 341, 291–298. Received 6 July 2001; accepted 4 August 2002

SUPPLEMENTARY PAPER Appendix 1: Geoscience Australia Australian focal-plane database.

Geol. Soc. Australia Spec. Publ. 22, and Geol. Soc. America Spec. Pap. 372 (2003), 107–119

Neotectonics of southeastern Australia: linking the Quaternary faulting record with seismicity and in situ stress M. SANDIFORD School of Earth Sciences, University of Melbourne, Vic. 3010, Australia ([email protected]). The ‘snapshot’ of active deformation provided by the historical record of seismicity in southeastern Australia correlates with the distribution of faults with significant demonstrable Quaternary displacement. This is most evident in the Mt Lofty and Flinders Ranges, a region of relatively high seismic activity and fault density, as well as the southern Victorian uplands. In the Mt Lofty and Flinders Ranges, faults associated with prominent range-bounding scarps are characterised by reverse-sense Quaternary slip rates in the range 20–150 m/106 y. In comparison, in the Murray Basin, a region of low seismic activity and low fault density, the largest faults have slip rates of less than 15 m/106 y, averaged over the last 5 million years. The modern neotectonic regime can be traced back at least until the terminal Miocene, where it is marked by regional unconformities between Upper Miocene and Pliocene sequences. A terminal Miocene onset for the modern neotectonic regime implies an important role played by Pacific–Australian plate-boundary forces in defining the unusual pattern of in situ stress in southeastern Australia characterised by east–west to southeast–northwest Hmax. KEY WORDS: Australia, faults, in situ stress, neotectonics, Quaternary, seismicity.

INTRODUCTION The historical record of seismicity within continental interiors provides an intriguing ‘snapshot’ of tectonic activity although it is not always clear how this relates to activity at the longer geological time-scale. For a continental interior remote from plate boundaries, southeastern Australia shows substantial seismic activity (Figure 1). However, there is little understanding of the relationship between the seismicity and the indicators of tectonic activity at geological time-scales (neotectonic structures). This is partly due to an entrenched belief that the Australian continent is tectonically inert. Nevertheless, there is a surprisingly rich neotectonic record in southeastern Australia. The most dramatic evidence is found in the Mt Lofty and Flinders Ranges in South Australia (Figures 2–4), which is also one of the most seismically active parts of the continent. The geomorphology of this region testifies to the profound role of faulting in shaping the landscape, more so than in any other part of the continent, and evidence for the role played by active faults in shaping this region has long been recognised (Sprigg 1945, 1946). To quote Sprigg (1946 p. 341) on the Mt Lofty Ranges: ‘At approximately the end of Miocene time the instability of the landmass was again apparent…. Faulting continued actively throughout the Pliocene and Pleistocene times’. More recently, a number of workers have documented compelling evidence for dramatic Quaternary faulting (Williams 1973; Bourman & Lindsay 1989). Much of the literature documenting this neotectonic activity is old or in obscure journals, with the consequence that it has been largely overlooked by the wider geological community. Moreover, the evidence for neotectonic activity has been further obscured by a tendency for much of the more recent

geomorphological literature to emphasise the antiquity of the Australian landscape at the expense of more youthful features (Ollier 1978; Twidale & Bourne 1975). Wittingly or unwittingly, this has contributed to the perception of an ancient landscape largely unaffected by tectonic processes

Figure 1 Distribution of seismicity in the southeastern part of the Australian continent, showing distinct concentrations in seismic activity in the Mt Lofty–Flinders Ranges–eastern Gawler Craton region of South Australia, and in a belt trending from the west coast of Tasmania, through south-central Victoria (in the vicinity of the southern uplands), northeast through the eastern highlands to southern New South Wales. Primary data are from Geoscience Australia.

108

M. Sandiford

Figure 2 Shaded relief image of the southeastern mainland Australia, showing the location of the main physiographic features and localities referred to in the text. Image derived from the AUSLIG/AGSO 9 s DEM. Numbers refer to Figures of associated localities.

at least since the fragmentation of Gondwana. Therefore, the prime objective of this paper is to draw attention to some of the most emphatic indicators of active faulting, highlighting a few key localities that bear testimony to the rich neotectonic record. The focus is on the distribution, extent and timing of faulting in the Mt Lofty and Flinders Ranges, with a brief excursion across the Murray Basin into the southern Victorian uplands (Figure 2).

IN SITU STRESS AND SEISMICITY IN SOUTHEASTERN AUSTRALIA The geodynamic framework for understanding the neotectonic evolution of southeastern Australia is provided by the historical record of seismicity and the in situ stress field. In terms of seismicity, the southeast is one of the most active parts of the Australian continent (Figure 1). Records extending back ~150 years show a widespread distribution of earthquakes up to Richter Magnitude (ML) ~6.4 across a zone ~1000 km in width from the eastern seaboard to the Gawler Craton in the west. ML 6 earthquakes are estimated to have a return period of ~29 years (McCue et al. 1990) although the largest 20th century earthquake was only ML 5.6 implying a marked temporal clustering of events. Distinct concentrations in seismic activity are recorded in the Mt Lofty – Flinders Ranges – eastern Gawler Craton region of South Australia, and in the belt trending from the

west coast of Tasmania, through south-central Victoria (in the vicinity of Port Phillip Bay and the Strzelecki Ranges), northeast through the eastern highlands to southern New South Wales (Figures 1, 2). The intensity of seismic activity in these zones contrasts considerably with the intervening Murray Basin and the cratons to the west. The in situ stress field in the Australian continent (Figure 5) is constrained by both earthquake focal mechanisms and borehole breakouts (Denham et al. 1979; Lambeck et al. 1984; Denham & Windsor 1991; Hillis et al. 1999; Hillis & Reynolds 2000). Greenhalgh et al. (1986, 1994) have calculated focal mechanisms for seven earthquakes in the Flinders Ranges, four of which are inferred to have had strike-slip fault mechanisms, and three reverse-fault mechanisms. These focal mechanisms define a principal horizontal compression (Hmax) of 83 ±30° (Hillis & Reynolds 2000). In the eastern highlands of Victoria, reverse-fault mechanisms have been resolved for a number of seismic events, and define a southeast–northwest azimuth for Hmax (Gibson et al. 1981). Hillis and Reynolds (2000) summarised borehole breakout data from two basins along the southeastern margin where the data are considered sufficient to define a statistically significant trend. In the Otway Basin, along the Victorian – South Australian border, the azimuth of Hmax derived from breakouts is 136 ±15°, while in the Gippsland Basin near the southeastern corner of the continent breakouts yields a Hmax of 130 ±20°. In summary, the southeastern part of the continent is

Neotectonics of SE Australia

109

Figure 3 Shaded relief image of the Flinders Ranges, Mt Lofty Ranges and western Murray Basin showing the location of the main physiographic features and localities referred to in the text. Image derived from the AUSLIG/AGSO 9 s DEM.

characterised by a broad zone of east–west to southeast–northwest compression. The ‘snapshot’ of active deformation provided by the historical record of seismicity is heterogeneous, with marked concentrations in the Mt Lofty and Flinders Ranges and in the eastern half of Victoria. Notwithstanding the fact that the seismic moment in all these regions is relatively low, it seems likely that if the regional variation in seismic activity is representative of the seismic moment distribution at geological time-scales it should be reflected in the neotectonic record. The following sections review some of the most dramatic evidence for Quaternary faulting in the southeastern Australia.

EVIDENCE FOR QUATERNARY FAULTING IN THE MT LOFTY AND FLINDERS RANGES The Mt Lofty and Flinders Ranges form an upland system extending some 800 km inland from the southern coast in the vicinity of the Adelaide (Figure 3). The Flinders Ranges (max-

imum elevation of ~1200 m) comprise the central and northern parts, north of about 33°S, while the Mt Lofty Ranges (maximum elevation of ~700 m) constitute the southernmost part. The ranges are almost entirely bordered by anomalously low regions with elevations typically less than 50 m asl (above sea level) and frequently below sea level. These lowlands and basins include (to the west) St Vincents Gulf, Spencer Gulf and Torrens Basin, (to the north) the Eyre Basin, (to the northeast) the Frome Embayment and (to the southeast) the Murray Basin (Figure 3). To the east the ranges merge with a broad, low upland system (200–450 m asl) through the Olary district that connects with the Barrier Ranges in western New South Wales (Figure 3). The morphology of the ranges and the surrounding lowlands suggests a regional-scale flexural compensation, as does the fact that the Bouguer gravity field is generally higher in the ranges than in the surrounding lowlands (Wellman & Greenhalgh 1988). The Mt Lofty Ranges are bounded by a set of discrete, curvilinear scarps defining the most dramatic fault-bound landscapes anywhere in the Australian continent (Figure

110

M. Sandiford

Figure 4 Shaded relief image of the Mt Lofty Ranges and Fleurieu Peninsula, South Australia, showing the main Quaternary faults referred to in the text. Image derived from a 100 m resolution digital elevation data provided by Department of Environment and Natural Resources, South Australia.

4). The main scarps are associated with the Para, Eden–Burnside, Ochre Cove (or Clarendon) and Willunga Faults on the western range front, and the Bremer, Palmer and Milendella Faults on the eastern side of the ranges. The lack of degradation of these scarps implies the profound role that faulting has played in the landscape evolution of this Mt Lofty Ranges (Sprigg 1945; Tokarev et al. 1999). The highest topography occurs along the western range front, markedly offset from the main drainage divide (Figure 4). The western scarps are deeply and steeply incised, particularly where the west-flowing rivers cross the axis of maximum topography, such as in the Torrens and Onkaparinga Gorges. Nick points in most valleys remain close to their generative scarps, again testifying to the youthful nature of the drainage systems along the western range front. The bounding faults associated with the main scarps in the Mt Lofty and Flinders Ranges are largely buried beneath extensive alluvial fans and associated pediments. These fans accumulated episodically, largely during the Quaternary, but are currently being dissected (Williams 1973), resulting in the exposure of the fault planes at a few localities, some of which are described below.

Milendella Fault at Cambrai, eastern Mt Lofty Ranges The Milendella Fault (Figure 4) is exposed in two ~8 m-high, undercut creek-bank sections near Cambrai (Bourman & Lindsay 1989) (Figure 6a, b). The fault is defined by a west-

dipping thrust at the foot of the Milendella scarp, which has a total topographic relief of ~250–300 m from ~380–400 m asl at its crest to 160 m asl at the exposed fault trace to ~80–100 m at the base of the footwall pediments (Figure 7b). The fault juxtaposes metamorphosed Cambrian rocks of the Kanmantoo Group in the hangingwall above a footwall comprising a Lower Miocene limestone (the Mannum Limestone) and a Quaternary sequence comprising mottled ferruginous clay interbedded with coarse conglomerate. The Mannum Limestone, which outcrops as disrupted, rotated and locally overturned lenses, consists largely of bryozoan fragments with a minor (~5 vol%) clastic component. The Quaternary sequence includes a distinctive mottled clay resembling the Ochre Cove Formation (Ward 1966) that elsewhere has been shown to contain the Brunhes–Matuyama palaeomagnetic reversal at ca 780 ka BP, interbedded with angular conglomerates. The total postEarly Miocene throw on the Milendella Fault is at least ~60–90 m, based on the differential displacement of Mannum Limestone in the footwall sequence between the scarp and exposures of stratigraphic equivalents along the Murray River. A slightly greater minimum displacement is indicated further south, in the Bremer valley, where Middle Miocene (ca 16 Ma) Lepidocyclina-bearing Mannum Limestone is reported at ~170 m asl, about 160 m above the elevation of the nearest equivalent exposures some 23 km to the northeast in the Murray Basin (Lindsay 1986). Siliciclastic debris in the Mannum Limestone adjacent to the scarp suggests that deposition occurred close to the

Neotectonics of SE Australia

111

Figure 5 Synthesis of the in situ stress field for the Australian region (adapted after Coblentz et al. 1998).

Miocene shoreline, and thus some of ~300 m present-day relief across the scarp is likely to have pre-dated the Early Miocene. At Cambrai, the Quaternary sequence is reported to have a thickness of ~30 m (Mills 1965). The exposure shows that the fault places basement above the entire Quaternary sequence, implying a minimum displacement over the last ~1 million years of at least 30 m.

Sellicks Beach, Fleurieu Peninsula The Willunga scarp is exposed in spectacular profile at Sellicks Beach (Figure 4), where marine processes have eroded sea cliffs up to 50 m into the alluvial fan, with a number of deep canyons penetrating through the footwall across the fault trace (May & Bourman 1984; Lemon & McGowran 1989). The canyon exposures include a nearcomplete ~60 m-thick section through the alluvial fan which comprises conglomerate and ferruginous clay equivalent in age to those at Cambrai. These sequences unconformably overly the Oligocene–Lower Miocene Port Willunga Formation (equivalent of the Mannum Limestone), locally with very high-angle discordance (Figure 8a). A near-vertical fault contact between weakly metamorphosed Cambrian sedimentary rocks and the Quaternary has an exposed relief of ~50 m. Reverse-fault motion is indicated by steep east-dipping fault traces in the hangingwall sequence within metres of the main fault trace (Figure 8b). A prominent wave-cut bench in the footwall limestone sequence ~4–5 m asl (Figure 8a) is attributed to the ca 120 ka BP last interglacial high sea stand (May & Bourman 1984). Further south, correlative interglacial benches in the hangingwall of the Willunga Fault are up to 12 m asl (Bourman et al. 1998), implying a time averaged vertical displacement ~50–70 m/106 y. This estimate is independently corroborated by reported elevation differences of ~130 m in the Lower Pleistocene (ca 1.7 Ma) Burnham Limestone along the eastern margin of the St Vincents Basin (Belperio 1995; Bourman et al. 1998). The

differential elevation of the Port Willunga Formation between the Myponga Basin (on the hangingwall block: Figure 6a) and the St Vincents Basin (the footwall block) suggests a post-Early Miocene displacement of ~240 m (Tokarev et al. 1999). The presence of locally abundant basement clasts in the Port Willunga Formation fringing the Willunga scarp suggests some relief must have existed during the Oligocene and/or Early Miocene. Consequently, the present-day ~360–400 m scarp relief (Figure 7a) most probably exceeds the total post-Early Miocene displacement.

Wilkatana Fan, western Flinders Ranges The western bounding fault of the central Flinders Ranges is exposed in Wilkatana Creek where it exits the ranges and incises the Wilkatana Fan (Figure 3) to depths of 15 m (Williams 1973). At this locality, a steep east-dipping fault plane at the foot of the bounding escarpment places a hangingwall comprising Neoproterozoic quartzite above Quaternary outwash gravels of the Wilkatana Fan (Figure 7c). Fault surfaces with 2–3 m reverse-sense offsets occur within the fanglomerate, while Williams (1973) estimated the total Quaternary movement along the bounding fault to be ~50 m. Williams (1973) attributed the major aggradation of Wilkatana fanglomerates of the Pooraka Formation to the interval 35–30 ka BP based on 14C dates. However, this is near the limit of 14C dating and should be considered a minimum age. More recent estimates place the age of the Pooraka Formation in the range ca 130–120 ka BP (Bourman et al. 1998). Long-term slip rates along the western bounding fault of the central Flinders Ranges of the order of ~20–30 m/106 y are indicated by these constraints.

Paralana Fault, northern Flinders Ranges A number of fault exposures are known from the northern Flinders Ranges (Figure 3) along Paralana scarp, which separates the northern Flinders Ranges from the Frome

112

M. Sandiford Figure 6 Photograph (a) and outcrop interpretation (b) of the Milendella Fault exposure near Cambrai, in the eastern Mt Lofty Ranges (AMG 338800E 6165900N Zone 54), first described by Bourman & Lindsay (1988). The view is looking south along the north–south-trending fault. See text for further discussion. Figure for scale.

Embayment to the east. This scarp face is characterised by some of the steepest, most deeply dissected relief in southern Australia, with an antecedent drainage system forming Yudnamutana Gorge incised some 600 m beneath the 800–850 m asl ‘Freeling Heights’ surface (Figure 7d). Exposures of the Paralana Fault near the foot of the bounding scarp show low-angle (