Structurally Complex Reservoirs
The Geological Society of London Books Editorial Committee Chief Editor
BOB PANKHURST (UK) Society Books Editors
JOHN GREGORY (UK) JIM GRIFFITHS (UK) JOHN HOWE (UK) PHIL LEAT (UK) NICK ROBINS (UK) JONATHAN TURNER (UK) Society Books Advisors
MIKE BROWN (USA) ERIC BUFFETAUT (FRANCE ) JONATHAN CRAIG (ITALY )
RETO GIERE´ (GERMANY ) TOM MC CANN (GERMANY ) DOUG STEAD (CANADA ) RANDELL STEPHENSON (NETHERLANDS )
Geological Society books refereeing procedures The Society makes every effort to ensure that the scientific and production quality of its books matches that of its journals. Since 1997, all book proposals have been refereed by specialist reviewers as well as by the Society’s Books Editorial Committee. If the referees identify weaknesses in the proposal, these must be addressed before the proposal is accepted. Once the book is accepted, the Society Book Editors ensure that the volume editors follow strict guidelines on refereeing and quality control. We insist that individual papers can only be accepted after satisfactory review by two independent referees. The questions on the review forms are similar to those for Journal of the Geological Society. The referees’ forms and comments must be available to the Society’s Book Editors on request. Although many of the books result from meetings, the editors are expected to commission papers that were not presented at the meeting to ensure that the book provides a balanced coverage of the subject. Being accepted for presentation at the meeting does not guarantee inclusion in the book. More information about submitting a proposal and producing a book for the Society can be found on its web site: www.geolsoc.org.uk.
It is recommended that reference to all or part of this book should be made in one of the following ways: JOLLEY , S. J., BARR , D., WALSH , J. J. & KNIPE , R. J. (eds) 2007. Structurally Complex Reservoirs. Geological Society, London, Special Publications, 292. DEE , S. J., YIELDING , G., FREEMAN , B. & BRETAN , P. 2007. A comparison between deterministic and stochastic fault seal techniques. In: JOLLEY , S. J., BARR , D., WALSH , J. J. & KNIPE , R. J. (eds) Structurally Complex Reservoirs. Geological Society, London, Special Publications, 292, 259–270.
GEOLOGICAL SOCIETY SPECIAL PUBLICATION NO. 292
Structurally Complex Reservoirs
EDITED BY
S. J. JOLLEY Shell UK Limited, Aberdeen, UK
D. BARR BP Exploration, Aberdeen, UK
J. J. WALSH University College Dublin, Ireland and
R. J. KNIPE University of Leeds, UK
2007 Published by The Geological Society London
THE GEOLOGICAL SOCIETY The Geological Society of London (GSL) was founded in 1807. It is the oldest national geological society in the world and the largest in Europe. It was incorporated under Royal Charter in 1825 and is Registered Charity 210161. The Society is the UK national learned and professional society for geology with a worldwide Fellowship (FGS) of over 9000. The Society has the power to confer Chartered status on suitably qualified Fellows, and about 2000 of the Fellowship carry the title (CGeol). Chartered Geologists may also obtain the equivalent European title, European Geologist (EurGeol). One fifth of the Society’s fellowship resides outside the UK. To find out more about the Society, log on to www.geolsoc.org.uk. The Geological Society Publishing House (Bath, UK) produces the Society’s international journals and books, and acts as European distributor for selected publications of the American Association of Petroleum Geologists (AAPG), the Indonesian Petroleum Association (IPA), the Geological Society of America (GSA), the Society for Sedimentary Geology (SEPM) and the Geologists’ Association (GA). Joint marketing agreements ensure that GSL Fellows may purchase these societies’ publications at a discount. The Society’s online bookshop (accessible from www.geolsoc.org.uk) offers secure book purchasing with your credit or debit card. To find out about joining the Society and benefiting from substantial discounts on publications of GSL and other societies worldwide, consult www.geolsoc.org.uk, or contact the Fellowship Department at: The Geological Society, Burlington House, Piccadilly, London W1J 0BG: Tel. þ44 (0)20 7434 9944; Fax þ44 (0)20 7439 8975; E-mail:
[email protected]. For information about the Society’s meetings, consult Events on www.geolsoc.org.uk. To find out more about the Society’s Corporate Affiliates Scheme, write to
[email protected]. Published by The Geological Society from: The Geological Society Publishing House, Unit 7, Brassmill Enterprise Centre, Brassmill Lane, Bath BA1 3JN, UK (Orders: Tel. þ44 (0)1225 445046, Fax þ44 (0)1225 442836) Online bookshop: www.geolsoc.org.uk/bookshop The publishers make no representation, express or implied, with regard to the accuracy of the information contained in this book and cannot accept any legal responsibility for any errors or omissions that may be made. # The Geological Society of London 2007. All rights reserved. No reproduction, copy or transmission of this publication may be made without written permission. No paragraph of this publication may be reproduced, copied or transmitted save with the provisions of the Copyright Licensing Agency, 90 Tottenham Court Road, London W1P 9HE. Users registered with the Copyright Clearance Center, 27 Congress Street, Salem, MA 01970, USA: the item-fee code for this publication is 0305-8719/07/$15.00. British Library Cataloguing in Publication Data A catalogue record for this book is available from the British Library. ISBN 978-1-86239-241-0 Typeset by Techset Composition Ltd., Salisbury, UK Printed by MPG Books Ltd, Bodmin, UK Distributors North America For trade and institutional orders: The Geological Society, c/o AIDC, 82 Winter Sport Lane, Williston, VT 05495, USA Orders: Tel þ1 800-972-9892 Fax þ1 802-864-7626 E-mail
[email protected] For individual and corporate orders: AAPG Bookstore, PO Box 979, Tulsa, OK 74101-0979, USA Orders: Tel þ1 918-584-2555 Fax þ1 918-560-2652 E-mail
[email protected] Website http://bookstore.aapg.org India Affiliated East-West Press Private Ltd, Marketing Division, G-1/16 Ansari Road, Darya Ganj, New Delhi 110 002, India Orders: Tel þ91 11 2327-9113/2326-4180 Fax þ91 11 2326-0538 E-mail
[email protected] Preface This volume was inspired by the Structurally Complex Reservoirs conference held in Burlington House, London between 28th February and 2nd March 2006. We are very grateful to ITF (Industry Technology Facilitator, Aberdeen) and to Duncan Anderson, in particular, for their tremendous efforts in helping us to arrange sponsorship and conference materials for the meeting. The event was sponsored by BG Group, BP, Chevron, ConocoPhillips, Hess, Maersk Oil, Shell, Statoil, Total, and Badleys, Beicip, EarthDecision, RDR, Roxar and Schlumberger: part of their sponsorship also helped to fund the colour production of this volume. We owe a debt of gratitude to all of the authors who wrote and submitted papers to what we think is a unique collection of benchmark papers on the topic of structurally complex reservoirs. We thank them for investing their time and science in this enterprise, and for producing their manuscripts by the deadlines required to guarantee publication in 2007. We also owe a similar debt to all of the referees who devoted their time, effort and expertise to reviewing papers included in this volume: Rolf Ackermann, Mohammed Ameen, Andy Aplin, Wayne Bailey, Tony Batchelor, Mark
Bentley, Stephan Bergbauer, Peter Boult, Peter Bretan, Paul Brockbank, Rob Butler, John Cain, Ben Clennell, Conrad Childs, Gary Couples, Tim Couzens, Steve Dee, David Ferrill, Quentin Fisher, Haakon Fossen, Brett Freeman, Peter Frykman, Alan Gibbs, Richard Gibson, Mark Hempton, Richard Hillis, Bob Holdsworth, Richard Jolly, Greg Jones, Richard Jones, Tor Anders Knai, Hemin Koyi, Bob Krantz, Tron Kristiansen, Gavin Lewis, Richard Lisle, Jingsheng Ma, Laurent Maerten, Gerhard Ma¨kel, Tom Manzocchi, Eddie McAllister, Alan Morris, Steve Naruk, Tim Needham, Jon Olson, Signe Ottesen, David Peacock, Paul Pijush, Peter Rowbotham, Iain Sinclair, Takeo Tanuguchi, Simon Tod, Chris Townsend, Jonathan Turner, Bruno Vendeville, Alastair Welbon, Scott Wilkins, Martha Withjack, Graham Yielding, Bjornar Ystad, Mark Zoback, and other referees who preferred to remain anonymous. We hope readers will find this book both interesting and informative. Steve Jolley, Dave Barr, John Walsh, Rob Knipe, 18 June 2007.
Contents Preface JOLLEY , S. J., BARR , D., WALSH , J. J. & KNIPE , R. J. Structurally complex reservoirs: an introduction DOMI´ NGUEZ , R. Structural evolution of the Penguins Cluster, UK northern North Sea WELBON , A. I. F., BROCKBANK , P., BRUNSDEN , D. & OLSEN , T. S. Characterizing and producing from reservoirs in landslides: challenges and opportunities HOFFMAN , K. S. & NEAVE , J. W. The fused fault block approach to fault network modelling TERTOIS , A. L. & MALLET , J. L. Editing faults within tetrahedral volume models in real time KRE´ ZSEK , C., ADAM , J. & GRUJIC , D. Mechanics of fault and expulsion rollover systems developed on passive margins detached on salt: insights from analogue modelling and optical strain monitoring KENDALL , J.-M., FISHER , Q. J., COVEY CRUMP , S., MADDOCK , J., CARTER , A., HALL , S. A., WOOKEY , J., VALCKE , S. L. A., CASEY , M., LLOYD , G. & BEN ISMAIL , W. Seismic anisotropy as an indicator of reservoir quality in siliciclastic rocks WILKINS , S. J. Fracture intensity from geomechanical models: application to the Blue Forest 3D survey, Green River Basin, Wyoming, USA LEWIS , H., HALL , S. A., GUEST , J. & COUPLES , G. D. Kinematically-equivalent but geomechanically-different simulations of fault evolution: the role of loading configurations HALL , S. A. & LEWIS , H. A damage domain approach to integration of geomechanics and seismic anisotropy for fractured reservoir characterization BERGBAUER , S. Testing the predictive capability of curvature analyses FERRILL , D. A., MORRIS , A. P. & SMART , K. J. Stratigraphic control on extensional fault propagation folding: Big Brushy Canyon monocline, Sierra Del Carmen, Texas FISHER , Q. J. & JOLLEY , S. J. Treatment of faults in production simulation models CHILDS , C., WALSH , J. J., MANZOCCHI , T., STRAND , J., NICOL , A., TOMASSO , M., SCHO¨ PFER , M. P. J. & APLIN , A. C. Definition of a fault permeability predictor from outcrop studies of a faulted turbidite sequence, Taranaki, New Zealand DEE , S. J., YIELDING , G., FREEMAN , B. & BRETAN , P. A comparison between deterministic and stochastic fault seal techniques MYERS , R. D., ALLGOOD , A., HJELLBAKK , A., VROLIJK , P. & BRIEDIS , N. Testing fault transmissibility predictions in a structurally dominated reservoir: Ringhorne field, Norway ZIJLSTRA , E. B., REEMST , P. H. M. & FISHER , Q. J. Incorporation of fault properties into production simulation models of Permian reservoirs from the southern North Sea MANZOCCHI , T., WALSH , J. J., TOMASSO , M., STRAND , J., CHILDS , C. & HAUGHTON , P. D. W. Static and dynamic connectivity in bed-scale models of faulted and unfaulted turbidites MA , J., VASZI , A. Z., COUPLES , G. D. & HARRIS , S. D. The link between a heterogeneous model and its flow response: examples from fault damage zones highlighting issues in domain discretization and flow simulation HARRIS , S. D., VASZI , A. Z. & KNIPE , R. J. Three-dimensional upscaling of fault damage zones for reservoir simulation MA¨ KEL , G. H. The modelling of fractured reservoirs: constraints and potential for fracture network geometry and hydraulics analysis MATTHA¨ I , S. K., GEIGER , S., ROBERTS , S. G., PALUSZNY , A., BELAYNEH , M., BURRI , A., MEZENTSEV , A., LU , H., COUMOU , D., DRIESNER , T. & HEINRICH , C. A. Numerical simulation of multi-phase fluid flow in structurally complex reservoirs BARR , D. Conductive faults and sealing fractures in the West Sole gas fields, southern North Sea
vii 1 25 49 75 89 103
123
137 159 173 185 203 219 235
259 271 295 309 337
353 375 405
431
vi
CONTENTS
ZHANG , X., KOUTSABELOULIS , N. C., HEFFER , K. J., MAIN , I. G. & LI , L. Coupled geomechanics –flow modelling at and below a critical stress-state used to investigate common statistical properties of field production data MAIN , I. G., LI , L., HEFFER , K. J., PAPASOULIOTIS , O., LEONARD , T., KOUTSABELOULIS , N. C. & ZHANG , X. The Statistical Reservoir Model: calibrating faults and fractures, and predicting reservoir response to water flood Index
453
469
483
Structurally complex reservoirs: an introduction S. J. JOLLEY1, D. BARR2, J. J. WALSH3 & R. J. KNIPE4 1
Shell UK Limited, Altens Farm Road, Nigg, Aberdeen AB12 3FY, UK (e-mail:
[email protected])
2
BP Exploration, Burnside Drive, Farburn Industrial Estate, Dyce, Aberdeen AB21 7PB, UK 3
Fault Analysis Group, School of Geological Sciences, University College Dublin, Belfield, Dublin 4, Ireland
4
Rock Deformation Research Ltd, School of Earth Sciences, University of Leeds, Leeds LS2 9JT, UK Abstract: Structurally complex reservoirs form a distinct class of reservoir, in which fault arrays and fracture networks, in particular, exert an over-riding control on petroleum trapping and production behaviour. With modern exploration and production portfolios commonly held in geologically complex settings, there is an increasing technical challenge to find new prospects and to extract remaining hydrocarbons from these more structurally complex reservoirs. Improved analytical and modelling techniques will enhance our ability to locate connected hydrocarbon volumes and unswept sections of reservoir, and thus help optimize field development, production rates and ultimate recovery. This volume reviews our current understanding and ability to model the complex distribution and behaviour of fault and fracture networks, highlighting their fluid compartmentalizing effects and storage-transmissivity characteristics, and outlining approaches for predicting the dynamic fluid flow and geomechanical behaviour of structurally complex reservoirs. This introductory paper provides an overview of the research status on structurally complex reservoirs and aims to create a context for the collection of papers presented in this volume and, in doing so, an entry point for the reader into the subject. We have focused on the recent progress and outstanding issues in the areas of: (i) structural complexity and fault geometry; (ii) the detection and prediction of faults and fractures; (iii) the compartmentalizing effects of fault systems and complex siliciclastic reservoirs; and (iv) the critical controls that affect fractured reservoirs.
Structurally complex reservoirs form a distinct class of reservoir in which fault arrays and fracture networks, in particular, exert an over-riding control on petroleum trapping and production behaviour (Møller-Pedersen & Koestler 1997; Coward et al. 1998; Jones et al. 1998; McClay 2004; Swennen et al. 2004; Sorkhabi & Tsuji 2005; Lonergan et al. 2007), (e.g. Fig. 1). With ‘easy oil’ becoming scarce, modern exploration and production portfolios are commonly held within geologically complex settings, in which reservoirs of this type are the common form. This means that there is an increasing technical challenge to find new prospects and to extract remaining hydrocarbons from structurally complex reservoirs in mature provinces such as the North Sea. New technologies developed in recent years permit exploration in increasingly hostile environments and economic development and production from some structurally complex discoveries that were previously ‘parked’ decades ago for technology catch-up. Our understanding, detection and ability to model and predict the complex distribution of faults, fracture networks, and other reservoir heterogeneities and their fluid compartmentalizing
effects and storage-transmissivity characteristics, is a critical element in predicting the dynamic fluid flow and geomechanical behaviour of these fields under production conditions. Improved analytical and modelling techniques enhance our ability to locate connected hydrocarbon volumes and unswept sections of reservoir, and ultimately help optimize field development, production rates and ultimate recovery. Geoscientists and engineers are addressing these issues within research institutions and operating asset environments around the world. Although research initiatives on structurally complex reservoirs vary considerably in scope, size and content, their ultimate goal from a practical perspective is to optimize the production of hydrocarbons from reservoirs. The research programmes brokered by the Industry Technology Facilitator (ITF) in the UK are a good example of such initiatives. These were 3-year thematic research collaborations between nine oil companies and over 25 academic and related research institutions in Europe, the USA and Australasia. The research programmes were specifically designed to improve our understanding
From: JOLLEY , S. J., BARR , D., WALSH , J. J. & KNIPE , R. J. (eds) Structurally Complex Reservoirs. Geological Society, London, Special Publications, 292, 1–24. DOI: 10.1144/SP292.1 0305-8719/07/$15.00 # The Geological Society of London 2007.
2
S. J. JOLLEY ET AL.
Fig. 1. Examples of structurally complex reservoirs. (a) Geo-cellular model of an intensely faulted field from the North Sea, in which shallow marine reservoirs are disaggregated into a patchwork of fault blocks. These fault blocks are mostly compartmentalized by sealing faults, and an intra-reservoir shale formation. Most of the fault blocks therefore require a dedicated producer well, and a water injector well to give pressure support to the producer as the block depletes. (b) Models of the fractured Clair reservoir, UKCS (fig. 12 of Barr et al. 2007). On the left are multiple stochastic realizations of a discrete fracture (conductive fault) network. One realization is shown in black and nine others generated using the same method, but with different seed numbers, in cyan. The red realization was made using a different method but with the same seed number as the black realization. On the right is a geo-cellular model showing upscaled effective fracture permeability from one realization.
STRUCTURALLY COMPLEX RESERVOIRS
of the broad range of geoscience and engineering issues associated with production from these complex reservoirs. The results were the catalyst for an international conference (held 28 February – 2 March 2006) and this Special Publication, which brings together a critical mass of papers on related topics. Together they provide the reader with an overview of global research on structurally complex reservoirs. This introductory paper aims to provide context to this collected work (papers cited in bold are those presented in this book) and an entry point into the subject. We have selected a series of sub-themes for this overview, which include: (i) structural complexity and fault geometry; (ii) the detection and prediction of faults and fractures; (iii) the compartmentalizing effects of fault systems and complex siliciclastic reservoirs; and (iv) the critical controls that affect fractured reservoirs. A final section gives a brief comment on future directions and priority areas that emerge from the collected papers.
Structural complexity The past couple of decades have seen the emergence of a variety of significant innovations in the analysis and modelling of structurally complex reservoirs. These developments have been driven mainly by the increased demands of the oil and gas industry and its reservoirs, but have also, crucially, been underpinned by vast improvements in both the quantity and quality of available data. 3D seismic datasets are now common and have been supplemented by a multitude of processing techniques and attribute analyses that facilitate the definition and mapping of structures. Just as 2D seismic has given way to 3D, recent developments in the technology of 4D seismic provide strong indicators that direct imaging of the impact of structures on flow will eventually become an essential tool in optimizing production from many complex reservoirs. Well production data have also increased in both quality and quantity, providing more refined indicators of reservoir production flow and pressure, with the improved constraints arising from horizontal wells and inclined well trajectories, and the general increase in the number of wells available from mature fields in particular. In addition, improved core recovery and improvements in the geological and fluid flow data from reservoirs have been matched by developments in both the capacity and functionality of existing modelling approaches. All these developments have enabled and stimulated both fundamental and applied research on the many technical issues related to the study of structurally complex reservoirs.
3
This volume presents a series of papers on the full range of these technical issues. The complete workflow of the reservoir structural geologist is well represented, with papers extending from the detection, mapping and prediction of faults through to fault property modelling and flow modelling of reservoir production. The running order of papers generally tracks this workflow, with the downstream side of it being essentially subdivided into faulted siliciclastic reservoirs and fractured reservoirs. This basic distinction is important, since it focuses the emphasis of studies which address the impacts of structural complexity on reservoir fluid flow. In many siliciclastic reservoirs, the reservoir units which host the hydrocarbons have higher porosities and permeabilities than the faults that transect them. In these circumstances, the faults are detrimental to flow, acting as baffles or barriers within a generally more permeable host rock sedimentary sequence (e.g. Knipe 1993; Downey 1994; Knipe et al. 1997, Yielding et al. 1997; Jones et al. 1998; Manzocchi et al. 1999, 2002; Hesthammer & Fossen 2000; Fisher et al. 2001; Brown 2003; James et al. 2004). Issues such as the primary connectivity of reservoir units will also impact the behaviour of the flow system. In fractured reservoirs, the host rocks (which include limestone, chalks or granite/basement as well as siliciclastic rocks) generally have lower permeabilities than the faults and fractures that transect them (Reiss 1980; Plumb 1994; Nelson 2001; Lonergan et al. 2007). Fluid flow in such reservoirs incorporates the combined effects of pervasive fracture systems, including joints, combined with faults. In these circumstances, the faults will often represent the main flow pathways that tap into the host rock volume which provides the hydrocarbon storage capacity. In a low porosity rock such as granite, storage is primarily in the fracture system but the storage and flow domains may be separate (e.g. with most storage in joints but most flow in conductive faults). In an impermeable but porous rock like some chalks, storage is primarily in the matrix (host rock) but flow is in the fractures. Some siliciclastic fractured reservoirs have a similar split between the flow and storage domains, but in others the matrix permeability is high enough to provide a significant flow contribution, leading to particularly complex behaviour during hydrocarbon recovery. Some fractured limestone reservoirs similarly have zones of high matrix porosity, or of leaching (e.g. palaeokarsts) or secondary porosity (e.g. due to dolomitization). This duality of faults and fractures as flow conduits or barriers is a fundamental property and provides the primary distinction between the two reservoir types. Faults within higher porosity-permeability reservoir units can represent baffles/barriers
4
S. J. JOLLEY ET AL.
whereas faults in tighter porosity, lower permeability reservoir units, can act as conduits to fluid flow. This duality of behaviour is not drawn on geographical grounds, but can in fact occur on the same structure intersecting different rock sequences, or form on the same structure at different times, when the host rock properties and deformation conditions have changed over geological time. The convenience of defining end-member faulted reservoirs, nevertheless, provides a useful conceptual framework, even if individual reservoirs or faults may sometimes incorporate both types of flow behaviour. For example, faults within fractured reservoirs may act as both seals and conduits over different parts of the fault surface or even at different times during the production history of the field.
Fault geometry Characterization and modelling of individual structurally complex reservoirs, typically begins with the 3D seismic definition and mapping of faults and other structures. Using newly developed modelling tools it is now possible to generate high quality 3D structural ‘framework’ models which can cope with structural complexities, such as intersecting and mutually cross-cutting faults, but also provide a means of examining displacement variations and cross-fault juxtapositions (Badley et al. 1990; Needham et al. 1996; Rutten & Verschuren 2003). These analyses are central elements in many of the studies presented in this volume and are now relatively routinely performed on reservoirs characterized by normal faulting. Geometrical complexities, including those involving cross-cutting and antithetic faults (and related branch-lines), and rapid changes in fault system polarity, present challenges to existing modelling techniques and have led developers to research alternatives which can better represent geological ‘reality’ (e.g. Hoffman & Neave). However, geological modelling is often conducted with the aim of producing geo-cellular models for use in reservoir flow simulation. Thus, it is necessary not just to honour relevant geological complexity at the ‘input’ stage of the 3D structural model, but also at the ‘output’ stage of the simulator grid. The adverse consequences of failing to do so are amply demonstrated in the field examples provided by Fisher & Jolley. Tertois & Mallett show that newly developed methods of tetrahedral volume modelling are capable of modelling such complexly faulted reservoirs, and Mattha¨i et al. describe a hybrid meshing approach which, combined with innovative methods for the discretization of governing
equations, could provide a comprehensive basis for future modelling efforts. Nevertheless, for the foreseeable future, most simulation of structurally complex reservoirs will take place in ‘conventional’ simulators operating on more-or-less regular grids of six-sided cuboid-shaped cells. More sophisticated approaches can and should be used to ‘ground-truth’ such simple models so that we understand the consequences of simplification and where it is or is not acceptable to do so. A recurrent theme in the reservoir modelling of structurally complex reservoirs is that the technical limitations of reservoir modelling packages and the computing hardware which runs them will always be a constraining factor, and that their improvement will always lag behind our technical demands. Interpretation, mapping and visualization tools, by contrast, are now very refined, as are those permitting structural modelling of various types (e.g. displacement analysis, restoration, cross-fault juxtaposition analysis). These tools arise not only from the practical demands of reservoir studies but also from developments arising from research into the geometry and kinematics of faults. Indeed current models for many fundamental aspects of faulting either derive from, or were significantly advanced by, the analysis of seismic data at reservoir scales, including: (i) fault growth models (e.g. Walsh et al. 2002); (ii) polygonal faulting (e.g. Cartwright 1994; Watterson et al. 2000); (iii) salt-related deformation (e.g. Vendeville & Jackson 1992a, b; Jackson 1995); (iv) relays and segment linkage (e.g. Childs et al. 1995); and (v) fault populations (e.g. Yielding et al. 1996; Cowie et al. 1996). There are, of course, many outstanding technical issues relating to the geometry and kinematics of faults, some of which are considered in this volume. Some reservoirs are characterized by very complex fault geometries with different modes of faulting developed at different times and with varying degrees of reactivation. These types of reservoir present a major challenge because although quantitative constraints on normal faults are relatively good, the characteristics of reverse fault systems and strike-slip fault systems, in particular, are less well understood and therefore much less predictable. Similarly, the nature and controls on the reactivation of earlier faults is not well understood, not just on geometrical grounds, but also in terms of flow. Nevertheless, Domı´nguez provides a comprehensive description of structural complexity arising from the interaction of two different fault trends and later fault reactivation in the Penguins field cluster, North Sea. Although this study shows how structurally complex reservoirs can arise from a relatively simple configuration of deformations, careful analysis is capable
STRUCTURALLY COMPLEX RESERVOIRS
of unravelling the structural evolution of the fault arrays. Barr shows that contractional inversion of earlier rift-related normal faults in Southern North Sea gas fields has implications beyond the purely geometrical, with the breaching of earlier seals and creation of conductive fracture networks sometimes having a profound effect on reservoir flow. Many of the studies in this volume have been conducted on reservoirs which include what might be considered relatively conventional tectonic normal fault systems and existing constraints on their geometry, at least in a generic sense, are generally good. High quality reservoir modelling demands the accurate characterization of fault geometries as a prelude to fault seal prediction (e.g. Dee et al.), juxtaposition analysis (e.g. Myers et al.) and fault property characterization and modelling (Fisher & Jolley; Zijlstra et al.), and therefore benefits from existing geometrical constraints. Similarly our current knowledge of the geometry and growth of gravity-slide normal fault systems, related either to the instability of delta slopes or salt, is now quite refined. The analysis and restoration of related structures of both tectonic and gravity-driven fault systems from 3D seismic data has provided excellent constraints on their kinematics (e.g. Jackson 1995). Physical modelling has made a significant contribution to our knowledge of fault array development (e.g. McClay et al. 2002) and the kinematics of gravity-driven fault systems, in particular (e.g. Vendeville & Jackson 1992a,b). The work of Kre´zsek et al. shows how recent technological innovations in imaging and quantifying deformation are capable of defining refined models for the kinematic evolution of margins characterized by salt-related thin-skinned tectonics. We anticipate that similar types of physical modelling studies will provide excellent constraints on the origin, geometry and growth of these types of fault system. Footwall collapse-related landslides are a type of gravity slide system which is less well understood, despite the fact that it is now well established that they dominate the structure of many reservoirs, including some in the North Sea. Although these reservoirs present challenges that are very different from other types of faulted reservoirs, they have not received a great deal of attention. Welbon et al. provide a comprehensive consideration of landslide structures and outline the challenges and opportunities provided by reservoirs in landslides and a new workflow for their characterization. The foregoing discussion highlights some of the technical issues associated with fault systems that have different geometries, origins and multiple event histories. These issues are generally the subjects of the first phase of reservoir characterization, which is mainly conducted by seismic
5
interpretation, sometimes supplemented by core analysis. The results of this type of analysis provide the essential backdrop for later phases of the structural geology and flow modelling workflow. This workflow involves a variety of technical components, the selection of which depends on the characteristics and type of reservoir concerned.
Detection and prediction of faults and fractures Fault detection Fault and fracture prediction in the modern exploration and production industry typically begins with 2D or 3D seismic interpretation. The issues and pitfalls involved in representing the geometry of discrete, seismically mappable faults have been discussed in the previous section. In principle, every fault that can be identified on 3D seismic can be mapped in three dimensions and characterized for flow simulation purposes. In practice, of course, inherent limitations of seismic data resolution mean that only the largest faults can be mapped as discrete objects. For example, very high quality seismic may permit faults with throws down to c. 5–10 m (and lengths of hundreds of metres) to be mapped, but a decrease in seismic quality could mean that faults with throws of c. 30–40 m may not be resolvable. In a general sense, the fault system can be subdivided into faults that are large enough to be visible and mappable from offset of reflectors, and the smaller faults that have more subtle seismic signatures that can sometimes be mapped from ‘attribute’ lineations on seismic reflectors (bedding horizons) using various amplitude variation and discontinuity detection techniques. Thus, in the absence of clear reflector offsets, the smaller faults are seen where amplitude becomes dimmed due to net destructive interference of diffracted seismic energy at the fault scarps (see Townsend et al. 1998 for discussion), and where horizon dip and dip azimuth changes sharply. Some of the latter attribute types assume that bedding is approximately planar or gently curved and that lines along which reflector dip changes rapidly are the ‘smeared out’ response to a fault too small to resolve as a discrete object. Others use wavelet correlation techniques to detect a change in seismic character, tracking along a single horizon or in time-slices or a 3D volume. In the first case the discontinuity detected is Horizon A – Fault – Horizon A; and in the second case, Horizon A – Horizon B. Most seismic workstations and many geomodelling software packages have tools such as coherency (Bahorich & Farmer 1995), dip,
6
S. J. JOLLEY ET AL.
edge-detection or semblance (Marfurt et al. 1998) available; indeed they are routinely used as interpretation aids in the early stages of defining seismically resolvable faults (e.g. Jones & Knipe 1996; Townsend et al. 1998). Care needs to be taken, however, to filter-out erroneous interpretations where seismic quality is sub-optimal, since some ‘semi-automated’ methods are capable of picking noise in addition to ‘real’ faults in the data (Hesthammer et al. 2001). It is a natural extension of this process to use these techniques to define subseismic lineaments. Curvature of seismically mapped surfaces can also serve as an edge detection tool, with hanging wall and footwall cut-offs corresponding to parallel bands of negative and positive curvature respectively (e.g. Murray 1968; Lisle 1994; Stewart & Podolski 1998). It is advisable to filter the input data spatially, to separate gross structural configuration (long wavelength) from faults (short wavelength), and to enhance the signal:noise ratio and resolution at the target wavelength (Bergbauer et al. 2003). Other techniques to enhance the detection of small-scale faulting include shaded illumination, dip-azimuth and directional curvature (extracted along parallel, vertical observation planes). Some techniques are optimally sensitive to faults with particular azimuthal orientations, and care must be taken to ensure that preconceived ideas about the fault trend do not lead to a failure to detect faults with an unanticipated orientation. Practical limitations on the time allocated for interpretation and the requirement to construct a computationally tractable model commonly lead to simplifications or omissions in the final model. Jolley et al. (2007) and Fisher & Jolley have shown how careful thought about the flow implications of a particular simplification can lead to better designed models, given the same data and model size restrictions. They also show that time spent in getting the starting model right is more than repaid by the reduced reservoir engineering effort required to generate a robust, history-matched flow simulation model.
to one which is kinematically self-consistent, by retro-deforming the current geometry to a plausible initial condition and then replicating the current geometry through a plausible forward deformation path. Software implementations of this approach are available in 2D and 3D modelling packages which link fairly seamlessly with seismic interpretation and simulation grid building packages. Typically, these assume that displacement is concentrated on the mappable faults, with intervening fault blocks deforming passively as they are carried on the faults. They also commonly assume constant displacement and slip directions on each fault segment, with discontinuous changes in displacement taking place at fault linkages or branch lines. This model is, of course, a simplification and an alternative view of the fault linkage in a system, which is supported by 3D seismic data and by closely spaced 2D observations in mines and quarries, is that displacement variation along fault planes is the norm and that displacement transfer can be accommodated by deformation of the intervening rock volume. These alternative models for fault linkage, referred to as hard- and softlinkage, may generate a very different level of fault connectivity (Fig. 2; cf. Walsh & Watterson 1991). These two end members will display very different flow behaviour, with Figure 2a being more connected than Figure 2b if the faults are conductive, but less connected if they are sealing. Although there is inherent complexity in the variation of natural fault displacement, mapping measurable displacement components such as fault throw, onto fault-plane-projection ‘Allan’ diagrams (Allan 1989), can provide constraints and quality control checks on the interpretation (Barnett et al. 1987; Badley et al. 1990). A common assumption is that gradual displacement
Fault network geometry Fields with only 2D seismic or outcrop data require much more interpolation and have increased ambiguities around fault geometry and linkage. Rules of thumb often come into play then, perhaps informed by analogue data from fields with 3D seismic, from well-exposed outcrops or from laboratory-scale models and kinematic or geomechanical modelling studies. The balanced crosssection approach (e.g. Dahlstrom 1969; Gibbs 1983) aims to constrain the structural interpretation
Fig. 2. Alternative map interpretations of sparse fault observations, e.g. on 2D seismic lines. (a) A connected (hard-linked) end-member with displacement changing only at fault intersections and a trellis-like fault network. (b) A disconnected (soft-linked) end-member with displacement varying continuously across the fault planes and decreasing to zero at the fault tips.
STRUCTURALLY COMPLEX RESERVOIRS
variation is expected of a simple fault, but abrupt changes in displacement imply the presence of a fault intersection or branch line, even if there is currently no intersecting fault mapped at that location (Badley et al. 1990; Needham et al. 1996; Nicol et al. 2002). Modern seismic and geological interpretation packages typically have some ability to display Allan planes, contoured with appropriate parameters, but the facility is not as widely used as it might be. Even 3D seismic has hidden connectivity issues which can benefit from the same interpretation approach as 2D data. For example, fault planes are rarely imaged directly on seismic cross-sections and have to be interpolated between bedding offsets; and interpretation typically begins on a grid of seed lines with the major faults being embedded in the model early as explicit, manually connected features. However, faults can be isolated features and die out downwards as well as upwards (e.g. Barnett et al. 1987; Walsh & Watterson 1991; Cartwright 1994), or merge into a de´collement surface or ductile zone such as salt or overpressured shale (Cartwright 1994; Jackson 1995; Watterson et al. 2000). Whatever the origin or nature of fault displacement variations, failure to define the geometry and connectivity of faults properly may in fact be one of the main sources of error in the modelling of structurally complex reservoirs, despite the fact that fault mapping now benefits from a variety of supporting techniques. The possibility that fault mapping could unhinge a significant number of reservoir studies may reflect the inherent pressures on geologists and geophysicists, to create the definitive ‘top reservoir map’ rapidly, when a more measured and discriminating 3D mapping approach would represent the best means of defining the basic geometry of faulted reservoirs. Failure to recognize the importance of basic fault mapping in 3D can introduce spurious connections or offsets of stratigraphic models. Even where the reservoir geometry is mapped accurately it is also possible that it is not represented accurately in the 3D simulation grid. However, techniques are available in 3D modelling packages to reproduce important fault-related geometrical features faithfully, such as the 3D variations in fault displacement, the geometries of fault intersections and the presence of fault discontinuities (e.g. relays). If these are not applied consistently, they may introduce unavoidable computational penalties to the model and may compromise later fault property modelling (e.g. some geometrical solutions to so-called ‘Y-fault’ geometries involve severe discretization and aliasing of faults). Hoffman & Neave discuss the advantages and limitations of some 3D fault modelling approaches in common use today.
7
Subseismic scale faulting As well as being used to define discrete faults, seismic data, particularly 3D seismic, can be used to detect smaller fault and fracture distributions. Simple seismic attributes responsive to surface roughness at a scale too small to be resolved individually include coherency and its relatives (discrete faults give rise to linear features, whereas background faulting or fracturing generates a broad zone of low coherency) and reflection amplitude (small-scale faulting produces waveform interference and scattering of the seismic energy which dims the amplitude). By their nature these attributes are insensitive to whether the fractures are conductive or sealing. More sophisticated attributes are available with multicomponent and multidirectional seismic acquisition, which allows measurements of seismic anisotropy (Verwest 1994; Bouska & Johnston 2005). A directionally anisotropic seismic response is strongly suggestive of open fractures, because the seismic response of a closed fracture or granulation seam is only subtly different from that of matrix, but that of an open fracture is very different, especially if it is filled by a compressible fluid such as oil or gas. The simplest measure is velocity anisotropy, where seismic rays travelling across an aligned fracture set are slower than those travelling parallel to the fractures (Crampin 1981; Hudson 1981; MacBeth 1995). Multidirectional or orthogonal fracture sets can give rise to an isotropic seismic response, but it may still be possible to infer fracture presence if the velocity is anomalously slow in all azimuths. Obviously that requires a meaningful definition of ‘anomalously low’, and may not be possible where there are large matrix velocity variations due to fluid or lithology effects. Fracture predictions based on seismic anisotropy and coherency-like attributes have been presented by, for example Bloch et al. (2003) and Barr et al. (2007). Calibration against core or image log data is advisable (e.g. Smith & McGarrity 2001) and proof-of-concept may be necessary before committing to a 4C OBC survey. More sophisticated approaches involve shear wave splitting (e.g. Winsterstein 1989; Owen et al. 1998; Maultzsch et al. 2003) and seismic amplitude variation with offset and azimuth (AVOA; Lynn & Thomsen 1990; Hall & Kendall 2003). Kendall et al. show that the magnitude of the seismic response in the Clair field, west of Shetland, depends not only on the fracture anisotropy but also on the matrix anisotropy introduced by bedding and mineral grain alignment. The seismic rays at large source:receiver offset travel at an angle to bedding rather than perpendicular, and therefore ‘see’ a combination of fracture and
8
S. J. JOLLEY ET AL.
matrix anisotropy. Kendall et al. therefore devoted considerable effort to matrix characterization, as a necessary precursor to seismic fracture prediction. They find an AVOA response between thin-bedded and massive sandstone units which is different in fractured and unfractured cases, even with the same fracture intensity in each unit, and thus they can distinguish between cases where the upper or lower unit is the more fractured. The two end-member geometries in Figure 2 have very different strain distributions between and around the faults. This strain distribution is important because it provides a potential entrypoint to the geomechanical prediction of subseismic faults and fractures, which can enhance or retard flow depending on whether they are more or less conductive than their host rock. Most faults are surrounded by a ‘damage zone’ tens of metres wide, comprising subsidiary faults and fractures (Antonellini & Aydin 1994; McGrath & Davison 1995; Caine et al. 1996; Knipe et al. 1997; Foxford et al. 1998; Beach et al. 1999; Hesthammer et al. 2000; Billi et al. 2003; Berg & Skar 2005). In this context damage refers to a change in the effective bulk properties of the rock caused by a swarm of small-scale structures and may cause flow enhancement or retardation (a drilling or production engineer will often restrict the term damage to flow retardation). In addition, the volumes between the faults have typically suffered volumetric or shear strain, perhaps related to bending during accommodation to the fault surface (e.g. Nicol et al. 2002; see also papers in McClay 2004) or to satisfy compatibility contraints as displacement dies out in a soft-linked fault array (e.g. Fig. 2b, Barnett et al. 1987; see also Hedland 1997; Soliva & Benedicto 2005). There may also be a systematic distribution of subseismic faults related to the large-scale structures. These can have a simple relationship to the major faults (e.g. antithetic in the hanging wall). If their distribution is unpredictable or poorly understood they can be modelled stochastically by assuming or observing a size, frequency relationship such as a fractal or power-law distribution and extrapolating downscale (Childs et al. 1990; Gauthier & Lake 1993; Ma¨kel). There are potential pitfalls in the analysis of such data (Heffer & Bevan 1990; Cowie et al. 1996; Yielding et al. 1996; Belfield 1998) and it is only meaningful within a genetically related range of structures. Subseismic faults might reasonably be extrapolated downscale from seismic faults but in circumstances where mechanical stratigraphy exercises a strong control on the fracturing process, fault or fracture systems may not have fractal geometries and the notion of extrapolation of fault size distributions could be flawed (see Nicol et al. 1996 and Soliva & Benedicto 2005 for discussion). This scenario almost
certainly applies to stratabound joints even in circumstances where they have a component of shear displacement and may display some fractal characteristics (e.g. Odling et al. 1999; Barr). Finally, the mapped fault tip locations are themselves limited by seismic resolution. Where displacement:distance relationships are well defined (e.g. Walsh & Watterson 1987; Cowie & Scholz 1992) it is possible to extrapolate the fault beyond the seismically resolved location to a predicted tip-line, based on the observed rate of displacement loss where it is still seismically observable (Yielding et al. 1996; Rutten & Verschuren 2003). Similarly, the width and internal geometry of fault damage zones has been well characterized in numerous outcrop analogues (Antonellini & Aydin 1994; Hesthammer et al. 2000; Billi et al. 2003; Odling et al. 2005) and can be predicted by selecting the right analogue and control parameters such as fault displacement and host-rock lithology. Once the geometry and petrophysical properties of the small-scale features are defined (see below), their flow implications can be simulated by explicitly modelling the observed fractures or a stochastic representation of them (Heath et al. 1994; Manzocchi et al. 1998; Walsh et al. 1998; Harris et al. 1999, 2003, this volume).
Strain modelling and fracture prediction Perhaps less widely acknowledged is the fact that two identical mapped fault networks can have different bulk strain distributions—although it has long been taught in structural geology textbooks that there are many potential paths to a particular deformed state (e.g. Hobbs et al. 1976, p. 32). If a record of the strain history during progressive deformation is preserved it may be possible to distinguish between alternative deformation paths (e.g. Ramsay 1967 p. 119–120). Lewis et al. demonstrate with a simple example how very different deformation paths resulting in very different internal strain distributions can produce identicalseeming fault and horizon geometries. Structural modelling software is available and under development which tracks the deformation history of the fault blocks (i.e. ‘kinematic restoration’) during the forward modelling step. Such models are nonunique, but they can yield valuable insights into potential subseismic deformation. In some software packages the non-uniqueness is exposed to the user, in that explicit kinematic choices have to be made about the fault-slip and inter-fault deformation mechanism; in that case it should be obvious that a matched result is a possible but not unique solution. But the more automated the process, developed perhaps in the interest of broadening the pool of potential practitioners beyond structural geology specialists, there is a greater risk of users
STRUCTURALLY COMPLEX RESERVOIRS
assuming that the result must be right ‘because the model says so’. Notwithstanding these limitations, considerable effort is currently devoted to developing techniques to predict fracture and small-scale fault distribution at a scale much less than seismic resolution and comparable to that of core or well logs. Typically these will be predictions of the ‘joint-like’ rather than ‘fault-like’ populations. The joint-like population includes shear fractures or granulation seams as well as tension fractures—essentially those features that are small and dispersed enough to form components of an effective medium at the scale of observation, rather than discrete entities. Long-wavelength curvature can be used as a strain predictor (e.g. Stewart & Podolski 1998) with the expectation that outer-arc extensional strains will be associated with open, tensile fractures. Use of this flexural beam model requires the correct choice of mechanical layering and identification of the neutral surface. Fractures would be predicted in synclines below the neutral surface as well as in anticlines above the neutral surface. Caution is advised in interpreting the output maps, as many existing approaches make potentially unacceptable simplifying assumptions (see Bergbauer & Pollard 2003 for review). Combinations of curvature attributes can be used to define a ‘shape curvature’ (Bergbauer et al. 2003), which classifies a mapped surface (at a particular wavelength) into anticlines, synclines, domes, basins etc. Shape curvature may be a predictor of fracture style rather than fracture intensity (e.g. orthogonal v. conjugate v. unidirectional, or shear v. tensile). Bergbauer describes an outcrop example of a fold where curvature shape and magnitude were poor predictors of fracture orientation and intensity but good predictors of fracture style. In this case, fractures developed within the relatively slab-like limbs are passively rotated and propagate along their axes, whereas additional strains at the curved fold hinge have reactivated fractures in shear. Indeed, Ferrill et al. describe the deformation within a monoclinal fault propagation fold, in which they find flexural shearing of well-bedded stratigraphy in the mid-limb has re-worked earlier formed fractures, such that in this case, fracture style is related to dip domain. Both papers emphasize the role of mechanical stratigraphy in governing the underlying deformation processes. Geomechanical models, typically discretized using finite element or boundary element techniques (Crouch & Starfield 1983), go beyond simple curvature. These use elastic (e.g. Wilkins) or more complex, elastic –plastic, constitutive laws to predict the stress and strain distribution between mapped faults. The situation is at its simplest where the structure in question formed as
9
a result of a single episode of progressive deformation. In that case it may be acceptable to approximate the driving stress or strain state by that required to deform an initial bedding configuration to the current field geometry. Predicted elastic strains are often scaled, to compensate for the fact that deformation was intermittent with some relaxation of elastic strain and accumulation of permanent strain between increments (e.g. Wilkins). More sophisticated constitutive laws such as the elastic – plastic one used by Lewis et al. attempt to model the progressive evolution of permanent strain through a deformation cycle. Refinements used include variable treatment of fault mechanics (purely frictional v. more complex formulations where some faults are effectively given finite strength) and the introduction of compaction, fluid pressure and gravity effects. An alternative approach where the deformation history is too complex to unravel or represent, is simply to model the present day stress state, treating the reservoir as a relatively homogeneous body of rock dissected by weak faults which locally perturb the far-field stress trajectory; an approach similar to that used in earthquake modelling. Stress perturbations and interactions around faults can be used to predict stress trajectories between faults (from which favoured open fracture orientations can be deduced), or to high-grade as potentially conductive those faults which are closest to a frictional failure criterion. Attempts are now being made to track strain evolution through complex deformation episodes (e.g. Dunbar & Cook 2003; Maerten & Maerten 2006), which in principle allows better control of fracture initiation and subsequent modification. The geomechanical models can be used to help define the displacement history (i.e. geomechanically based structural restoration, e.g. Maerten et al. 2006) or the displacement history can be extracted from a kinematic structural restoration and used as boundary condition inputs to a geomechanical model (e.g. Lewis et al. 2004). The end-point of such modelling is rarely the prediction of individual fracture formation, except for those subseismic faults that are large enough to implement in a 3D reservoir simulation model. More typically, dilational strain is taken as an open fracture indicator and compactional strain as a closed fracture indicator. The magnitude and orientation of the principal stress or strain axes are used to constrain fracture orientation, type and intensity (Ma¨kel; Wilkins). The constraint can either be deterministic, via empirically defined lookup tables or correlations to well data; or stochastic where the geomechanical model outputs are used as constraints on a discrete fracture network which is then upscaled to effective flow properties in a reservoir simulator (e.g. Sabathier
10
S. J. JOLLEY ET AL.
et al. 1998). An important modelling decision is whether to calibrate directly to static well and seismic observations, or to bypass those and calibrate directly to dynamic, reservoir engineering observations (e.g. Barr). An argument in favour of the latter approach is that the fracture description is a means to an end and that end is populating a reservoir simulation model. Ultimately, the fracture description has to be upscaled to effective cell properties (e.g. Bourbiaux et al. 1997), a step that introduces its own set of assumptions and uncertainties. In addition, the fractures in a vertical well may be unrepresentative even of the surrounding (say) 100 m 100 m grid cell area. A weakness of direct dynamic calibration is that it introduces an empirical step involving poorly defined or understood processes. If the next well fails to match predictions, there is little chance of understanding why, and it will likely be handled by making another poorly understood empirical adjustment. Hall & Lewis attempt to introduce some rigour into this process, by placing seismic and geomechanical attributes on a common descriptive footing, as an effective medium sampled at the target reservoir simulation scale.
Fault compartmentalization Faults transecting siliciclastic reservoirs in which the reservoir units have high porosities and permeabilities generally act as baffles or barriers to flow. In these circumstances, following the 3D mapping of seismically imaged faults and, sometimes, the prediction of subseismic faults, the next major step is definition of fault properties and their incorporation within reservoir models, either as transmissibility multipliers on individual faults, or as upscaled effective permeabilities, in the case of subseismic faults. Here we describe the technical issues and methodologies associated with this phase of the structural geology workflow.
Fault zone properties Natural fault zones are characterized by threedimensionally complex juxtapositions and displaced lenses of host stratigraphy, and modified porosity-permeability variations within the suite of different fault rock products and fracture arrays that the zones usually contain (e.g. Knipe 1993, Childs et al. 1997, this volume; Knipe et al. 1997; Foxford et al. 1998; Fisher & Knipe 1998, 2001; Gibson 1998; Ingram & Urai 1999; Skerlac 1999; Sperrevik et al. 2000; Aydin & Eyal 2002; Jourde et al. 2002; Berg & Skar 2005; Eichhubl et al. 2005; Shipton et al. 2005; van der Zee & Urai 2005). However, there are significant hurdles
to overcome in predicting the distribution of these lithological fragments and fault rock properties within a fault zone, from the often sparse information available to field appraisal and development teams. It is well known that clay content plays a significant role in reducing permeability in fault rocks developed within siliciclastic rocks (e.g. Fisher & Knipe 1998, 2001; Manzocchi et al. 1999; Sperrevik et al. 2002). Consequently algorithms have been developed which attempt to calculate the distribution of average clay content (and therefore the general permeability distribution) within fault zones—from the reservoir properties, clay content and shale bed distributions within the adjacent stratigraphy—at a similar scale to that of a typical simulation model fault (cf. Yielding et al. 1997; Doughty 2003). As discussed by Fisher & Jolley, at the modelling stage many of the predictive algorithms are based on the input of ‘clay contents’ derived from geophysical well log data—and care is therefore needed to account for uncertainties in petrophysical calculation of the ‘shale’ or ‘clay’ content measures (VShale, VClay), and to ensure that these results are compatible with independent measures of clay content (often taken from core samples). The three most commonly used algorithms, developed over a decade ago, use some of the basic processes which entrain clay minerals and discretely bedded shales into a fault zone (Fig. 3). Thus, Shale Gouge Ratio (SGR, Yielding et al. 1997) is based on mechanical mixture of shaley material within a fault gouge, assuming that the resulting fault rock/gouge clay content approximates to the average collective stratigraphic clay content which has been displaced past any given point on the fault. The Clay Smear Potential (CSP, Bouvier et al. 1989; Fulljames et al. 1997) predicts the length continuity of a shale bed plastically smeared into the fault from the fault throw and source shale bed thickness. The Shale Smear Factor (SSF, Lindsay et al. 1993) focuses on predicting the thickness continuity of smears caused by ductile smearing and abrasion of shales and also as a function of throw and source shale bed thickness. In some situations, for example where shale and/or sand beds are relatively thin, it can be argued that clay smears are accounted for within the SGR algorithm, once stratigraphy is upscaled. Where the shale layers are thicker, the smears can become more robust and continuous, such that they preserve stratigraphic compartmentalization despite the faulting but the detail of this is not captured by SGR. Consequently, several proprietary algorithms have developed within the industry as a spin-off from these basic SGR/CSP/SSF forms, mostly in an attempt to integrate them into a harmonic averaging tool for predicting fault zone clay content. Childs et al. provide an important
STRUCTURALLY COMPLEX RESERVOIRS
(a)
Vcl5, Δz5 Vcl4, Δz4
throw
Vcl3, Δz3 Vcl2, Δz2 Vcl1, Δz1
Shale Gouge Ratio
SGR =
Σ (Vcl.Δz) throw
x 100%
(b)
Δz distance
Clay Smear Potential
CSP =
Σ
thickness distance
2
(c)
Δz
throw
Shale Smear Factor
SSF =
throw thickness
Fig. 3. Fault seal algorithms, commonly applied in low/ mid net-to-gross (mixed sand-shale) reservoir stratigraphies. (a) Shale-Gouge-Ratio (SGR; Yielding et al. 1997), (b) Clay-Smear-Potential (CSP; Bouvier et al. 1989; Fulljames et al. 1997), (c) Shale-SmearFactor (SSF; Lindsay et al. 1993). From Jolley et al. (2007), modified from Yielding et al. (1997).
11
dataset in this regard, which shows that clay smears can become disengaged from their source layers, to form ‘slugs’ within the fault zone. Their outcrop observations on faulted New Zealand turbidites do not support a systematic arrangement of clay smears in relation to their source beds, and so they have developed a stochastic approach to distribute the smears within the fault planes, which they term the probabilistic Shale Smear Factor (PSSF). An interesting outcome of such a model is to provide effective properties at high throw to bed thickness ratios, which are indeed similar to other approaches, such as SGR. This work therefore provides a rationale for the successful application of these approaches, even though the deformation mechanism which it implicitly assumes may not be correct. Similarly, detailed modelling of fluid flow through a realistic representation of fault damage zones by Harris et al. supports the conclusion that the simpler, commonly used approach of harmonically averaging volume-weighted fault-rock permeability, is a useful first approximation in assessments of the flow impact of subseismic faults (Walsh et al. 1998). The prevailing stress and temperature during deformation also controls the degree of grain breakage (cataclasis) and crystallization of some cementation types, and consequent permeability collapse within the fault zone (Knipe 1989; Zhang et al. 1990; Wong et al. 1997; Chester & Chester 1998; Fisher et al. 2000, 2003). Thus, despite the popular view from the usage of clay-content algorithms in fault seal analysis described above (that low clay content faults developed in sand-rich reservoirs do not seal), under the ‘right’ conditions and geohistories, sealing on a production timescale is also possible in low clay content fault rocks. Where there is a strong palaeotemperature gradient across a field or cluster of fields, this can lead to profound differences in fault compartmentalization and consequent production characteristics (e.g. Hesthammer et al. 2002). However, these low clay content cataclastic seal types can become brittle, damaged and leaky if a field is subsequently deformed under different stress and lower temperature conditions (e.g. Leveille et al. 1997). Barr gives a detailed description of the complex distribution of fault seal compartmentalization, and conductive faults and open fracture systems in the sand-rich aeolian reservoirs of the West Sole gas fields. He shows that sealing lithified cataclasites formed in fault zones during an early rifting phase at high pressures and temperatures with some influence from host sediment facies type on crystallization of certain cementation phases; and that open seal-breaching fracture systems developed during later contractional inversion and fault reactivation at lower pressures and temperatures. In general
12
S. J. JOLLEY ET AL.
terms the sealing and open structures tend to be mutually exclusive. Although data resolution and a significant element of clustering or selectivity in the reactivation processes introduce technical difficulties, it has nevertheless been possible to use this model to detect reactivation and therefore predict the general distribution of intact and breached seals and/or open fracture networks within the fields.
Static fault seal prediction Exploration projects commonly use a deterministic approach to ‘map’ fault zone properties, such as SGR, within a fault plane and thereby calculate the predicted hydrocarbon columns that can be accumulated and held by sealing faults over geological time (e.g. Yielding 2002). These methods are based on explicit modelling of reservoir and non-reservoir juxtapositions and the sealing properties of fault rocks. However, James et al. (2004) suggest that static fault seals are controlled exclusively by juxtaposition, and they describe a stochastic approach to fault seal analysis, which models variation of stratigraphic stacking across a fault to assess its hydrocarbon retention capacity. An energetic debate has since developed on the conference circuit between proponents of these two radically differing approaches. Dee et al. use the data presented by James et al. (2004), in order to compare and contrast the results that are achieved by these two radically different methods on the same dataset. They compare the stochastic analysis with a standard SGR-based approach and suggest that despite the conceptual differences between the deterministic and stochastic methods, the results are remarkably similar. This study therefore appears to provide another useful validation of a widely applied ‘rule of thumb’.
generally tested by comparing the match between actual historical production data and the simulated production history, the so called ‘history match’. In detail, the fluid flow between adjacent cells in a standard simulation model is expressed by the cell-cell ‘transmissibility’ (a function of the geometry and permeability of the cells). The reduced permeability of an intervening fault is accounted for by correcting the cell – cell transmissibility with a fault transmissibility multiplier, a function of fault rock thickness and permeability (Fig. 4; e.g. Knai & Knipe 1998). However, although the stratigraphic controls on field compartmentalization are routinely addressed within generally accepted geologically-rationalized tools and workflows, it has been an industry-wide experience that the flow retarding effects of the faults are treated in an ad hoc manner by the reservoir engineer, late in the workflow. It is possible to achieve a history match by using the production data to guide manual application of geologically unrealistic faults and fault properties to steer flow and pressures around the model in this way. However, this is likely to be an artificial compensation for other inadequacies and uncertainties in the model, which then become obscured by this activity (Fisher & Jolley). It follows that the more of these trial-and-error amendments there are in a model, the less likely it is that its simulated flow approximates to reality (despite the history match) and consequently the ‘prediction mode’ becomes unreliable. Such structural uncertainty can seriously impact field development planning and production management (e.g. Corrigan 1993; Lia et al. 1997). However, as described below, data, tools and methods have evolved in recent years to permit more valid, systematic incorporation of fault properties within simulation models.
Production fault seal modelling Under production conditions lower permeability fault zones will generally lead to the compartmentalization of pressure distributions, hydrocarbon saturations and contacts across the faults. Fault property modelling under these conditions therefore attempts to define the rate of fluid flow across the faults in order to quantify the connected petroleum volumes that can be accessed by any given well or group of wells. Numerical flow simulation models including the effects of fault properties are now routinely used to guide field development, production management and well planning decisions (e.g. Dake 2001). The reliability of the ‘prediction mode’ of a model is
Fig. 4. Fault transmissibility multiplier (TM) calculation between faulted cells of a simulation model. These calculations differ in detail between simulation software packages and this cartoon (from Jolley et al. 2007, modified after Manzocchi et al. 1999) ignores cell dip terms and assumes net:gross ratios and intersection areas of 1.0.
STRUCTURALLY COMPLEX RESERVOIRS
Incorporating fault properties in production simulation models The fault seal algorithms described above provide the building blocks for inclusion of fault properties within simulation models. For example, Manzocchi et al. (1999, 2002) provided robust methods with which it is possible to invoke these algorithms, to incorporate the flow-retarding effects of faults systematically. Their methods calculate geologically realistic fault transmissibility multipliers from the upscaled model geometry and geo-cellular properties (e.g. stratigraphic distribution of clay, porosity, permeability). Thus, fault rock clay content prediction methods (e.g. SGR/CSP/SSF and other similar algorithms) can be calculated from the reservoir geometries and properties implicit within the geo-cellular cornerpoint grid of a simulation model. The first-order sensitivity of a simulation to structural influence is caused by juxtaposition of flow units and non-flow units across the faults, since this affects the basic ‘plumbing’ within the model. Hoffman & Neave and Tertois & Mallet discuss some of the pitfalls, procedural limitations and potential solutions to the necessary compromises that are made in constructing a 3D fault model from seismic data and the subsequent simulation grid. Thus, geometric flaws introduced at that stage result in inappropriate layer juxtapositions in the cellular model, and also restrict the geologist’s ability to properly represent fault zone properties. Constraints on simulation cell geometry, driven by a desire for simple flow formulae that can be solved by existing technology in a commercially useful timeframe, can introduce compromises in fault representation since it is typically the faults that contribute most geometrical complexity to the model. However the validity of flow simulation models (and therefore the reliability of their results) can be improved vastly through inclusion of geologically realistic faults, integrated with systematically calculated fault zone properties (e.g. Jolley et al. 2007; Fisher & Jolley). In this approach, SGR and/or an SGR/smear algorithm combination can be calculated and used as a proxy for fault rock clay content and integrated with fault rock permeability data to systematically assign fault transmissibility multipliers within simulation models (sensu Manzocchi et al. 1999). Myers et al. provide a case study from a North Sea fluvial reservoir in which the history matches obtained from simulation models are progressively improved by incorporating increasingly realistic fault geometries and stratigraphic architectures. This incremental approach helped to narrow the uncertainty range on fault properties applied to faults in the simulation. Jolley et al. (2007) found that provided a geologically valid model was
13
transferred to the simulator, the best history matches were then achieved in a fraction of the usual project time by integrating fault rock property data acquired from drill cores obtained within and close to a given reservoir to calculate the multipliers. This was particularly the case where the sampled fault rocks had experienced a similar stress-temperature (burial) history to that of the study reservoir.
Modelling interaction between faults and stratigraphic complexities Fisher & Jolley review the wide-range of uncertainties associated with the data acquisition and processing, interpretation and modelling phases of the fault property modelling workflow. Additionally, limitations within the generally available modelling technology, force a tension between efforts to capture and preserve the geological information which is critical to fluid storage and flow, and efforts to distil the geology down to more basic elements in order to satisfy a simulation model’s computational memory budget. Care therefore needs to be exercised when characterizing and simplifying stratigraphic details to assign average properties to model cells, as this can disengage continuous depositional features and/or introduce erroneous connections between layers in a model (Myers et al.). For example, it is a common experience that actual flow connectivity within a reservoir is less than that implied by a geo-cellular model of the field. As Fisher & Jolley point out, these effects are frequently assumed to be caused by sealing faults, leading to erroneous ad hoc edits being applied to the structure of the model. There is an alternative, entirely logical explanation for these effects—since the averaging of thin shale beds into a net:gross value for each cell, and the stacking of these cells within the model, can overlook the compartmentalizing effects of relatively thin shale layers between sand bodies, unless these shale layers are explicitly modelled. Despite the obvious interdependence between faults and stratigraphic complexities in controlling compartmentalization, there have been few published attempts to characterize the interplay between these elements directly (e.g. Ainsworth 2006). Manzocchi et al. use a comprehensive suite of faulted and unfaulted models of sheet-like turbidite deposits, to examine the interplay between stratigraphic elements, faulting and fault zone properties using the PSSF method developed by Childs et al. Instead of using the more traditional cellular modelling methods and net:gross ratio to build their models of sand and shale distribution, Manzocchi et al. have developed a bed-scale modelling method which explicitly
14
S. J. JOLLEY ET AL.
includes a measure of sand body connectivity, known as the amalgamation ratio. Compared to data collected from outcrops of similar turbidite deposits, this method was found to give a far more realistic stratigraphic architecture and interbed connectivity in the models. Describing their results in terms of percolation theory, they found that high net:gross sheet turbidite sequences can be very poorly amalgamated/connected. In those circumstances, the introduction of arrays of subseismic faults (, c. 5 m throw) only reduces the connectivity within models under a rare combination of circumstances and in some situations, the relative influence of faults and stratigraphic elements on flow connectivity in the models were indistinguishable.
Multi-phase flow properties of faults Given the very small pore throats which characterize many fault rocks, water-wet faults (i.e. those having a water film coating all the grain surfaces in the fault rock and thus impinging on its available pore space) have such high water saturations close to the free water level (FWL) of a reservoir that they may have negligible relative permeability to hydrocarbons. At some distance above the FWL the buoyancy force in the hydrocarbon column may be sufficient to overcome the capillary threshold pressure of the fault rock, giving it a finite relative permeability to hydrocarbons, and thus permitting cross-fault flow of oil or gas (for discussion see Fisher et al. 2001; Fisher & Jolley). Traditionally, the multi-phase flow properties of faults have not been included during production simulation modelling, although several key publications have highlighted their importance (e.g. Manzocchi et al. 1998, 2002; Manzocchi 1999; Fisher & Knipe 2001; Rivenæs & Dart 2002; Al-Busafi et al. 2005). A recent innovation in fault handling within reservoir models is the development by Manzocchi et al. (2002) of a method for the inclusion of the two-phase flow properties of faults. Because two-phase properties, unlike single fluid phase properties, cannot be attached to the face of grid blocks in reservoir simulations, this method derives pseudo-relative permeability functions including the fault rock properties in the upstream grid block for cross-fault flow. This method incorporates the saturation and flow rate dependencies of two-phase flow and is a fairly comprehensive treatment of the problem. Zijlstra et al. present flow data from a number of faulted reservoirs suggesting that two-phase flow properties of the faults are important in controlling compartmentalization of fluid production. They support their conclusions by presenting the results of a method for fault property modelling which is easy to
implement and accounts for some of the effects of fault-related single phase and multiphase flow (as described by Manzocchi et al. 1999, 2002).
Upscaling the flow effects of subseismic faults The recognition that subseismic faults could have an impact on flow within siliciclastic reservoirs, has only recently been matched by the development of methods which provide a basis for their incorporation into flow simulation models. Typical approaches involve definition of the upscaled effective properties of subseismically faulted rock volumes, with their eventual implicit inclusion in reservoir simulations. Early work showed that for typical fault densities and geometries (including connectivities), subseismic fault arrays will generally only begin to have a significant impact on flow within reservoirs (i.e. decreasing effective permeabilities by more than c. 20%) when fault rock permeabilities are at least two orders of magnitude below those of the reservoir host rocks (e.g. Manzocchi et al. 1998; Walsh et al. 1998). For less permeable fault rocks, cross-fault flow decreases rapidly with an increase in flow tortuousity, until flow is dominated by the connectivity of sealing faults (Walsh et al. 1998). Analysis of the impact of damage zones surrounding seismically imaged faults reveals similar features, with newly developed methods being capable of exploring the sensitivity of flow to the full range of geometric and scaling parameters associated with damage zones (see Harris et al. and references therein). Harris et al. (1999, 2005, this volume) extend these sensitivity studies into 3D, and make a strong case for the routine definition of damage zone effective properties and their implicit inclusion in reservoir flow simulations. Ma et al. show the extent to which modelling of small fault arrays is sensitive to the modelling methods used, including such issues as discretization and the preservation of fault connectivity. Other issues need to be resolved, such as the multiscaling properties of faults, associated upscaling challenges and the incorporation and displacement of stratigraphic architectures. In that respect, the bed scale modelling of Manzocchi et al. highlights the complex flow sensitivity of faulted sedimentological sequences, a topic which we anticipate will be the subject of future research. Nevertheless, although adoption of associated modelling and upscaling approaches has, hitherto, been relatively slow (with most attention focused on larger seismically imaged faults), recent advances suggest that the incorporation of subseismic faults and damage zones should become relatively routine within the next decade.
STRUCTURALLY COMPLEX RESERVOIRS
Fractured reservoirs Fractured reservoirs form a special class of structurally complex reservoirs in which hydraulically conductive fractures or faults make a significant contribution to, or dominate, subsurface fluid flow. The interaction between the storage domain (typically dominated by matrix lithologies with relatively high pore volume and relatively low permeability) and the flow domain (typically dominated by fractures with relatively low pore volume and relatively high permeability) leads to complex fluid and pressure behaviour. This makes it difficult to predict field performance, even assuming a ‘perfect’ understanding of the nature and distribution of the fractures. The problem is compounded by the fact that fracture properties are generally more difficult to characterize than matrix, whether from core, well-log or seismic observations. Ma¨kel provides a detailed review of the issues which need to be addressed in describing and modelling fractured reservoirs, focusing on analysis, description and calibration of the fracture network; that is, up to the point at which the fracture model is upscaled to a simulation grid and ‘handed off’ to the reservoir engineer. The fractured reservoir papers in this volume supplement a larger collection of papers on this topic, provided in Lonergan et al. (2007). More general reservoir engineering and geological aspects of fractured reservoirs are also covered by, for example, Aguilera (1995) and Nelson (2001).
Types of fractured reservoir Fractured reservoirs are traditionally classified according to the relative contributions of fracture
15
and matrix permeability (e.g. Reiss 1980; Nelson 2001; Allan & Qing Sun 2003; Ma¨kel; Fig. 5). Because of their high conductivity and low pore volume, fractures typically make a large contribution to the flow domain but a small contribution to the storage domain. For a given well or reservoir, a ‘fracture index’ can be defined to represent the magnitude of the fracture v. matrix contribution. For example, the Fracture Productivity Index of Reiss (1980) ratios the well-test Kh (average permeability times the height of the tested interval) to the matrix Kh; it can be assumed that values much larger than unity reflect a significant fracture contribution. This leads to the interesting observation that ‘how fractured’ a field appears depends not just on the conductivity of its fracture network, but also on that of the matrix properties. For example, the West Sole field (Barr) behaves more like a ‘classical’ fractured reservoir than does the Clair field (Barr et al. 2007), despite having an effective fracture permeability an order of magnitude smaller. That is because it has two or three orders of magnitude less matrix permeability. Fractured reservoirs typically have very heterogeneous porosity and permeability distributions (Mattha¨i et al.), which result in characteristic patterns of well performance with most production coming from the best few wells (e.g. Nelson 2001; Barr). The worst wells have failed to intersect connected, conductive fractures and a large financial benefit would flow from an ability to target only the best well locations. In practice that requires a robust pre-drilling description of the effective fracture network, hence much industry and academic attention is focused on that objective.
Fig. 5. Schematic representation of the common subdivision of fractured reservoir types based primarily on matrix character (Nelson 2001; Allan & Qing Sun 2003; fig. 1 of Ma¨kel (this volume).
16
S. J. JOLLEY ET AL.
Fracture detection and description The recognition of fractures in wells is also dealt with in detail by Ma¨kel. Core is the most definitive source of fracture information and procedures for describing, identifying and classifying them are well established (e.g. Kulander et al. 1990) but it is expensive to acquire and there are some data limitations to consider. Drilling, core recovery and sample preparation and handling can all modify natural fractures or create new fractures which must be screened out of the subsurface description. Purely drilling-induced fractures have characteristic features that make them easy to recognize (Kulander et al. 1990) but more subtle modifications to the geometry or aperture of pre-existing fractures can be harder to recognize. Open fractures are particularly vulnerable to disturbance because they weaken the rock and the most fractured part of a core may be recovered as uninterpretable rubble. Fractures are likely to have larger apertures at surface than in the subsurface, due to the reduction in effective closure stress, and if the fracture faces can be fitted together perfectly they may have had no subsurface aperture. Most fractures described in reservoirs are not simple planar breaks but have irregular faces and are partially propped or bridged by cementing minerals. Such partially cemented fractures may give the best indication of subsurface aperture and if a genetic link can be drawn between cemented and open fractures, and the cement can be shown to have grown in a single phase, vein widths can also provide a useful indicator (e.g. Ma¨kel). In fields produced by depletion drive, the effective stress acting to close a fracture will increase during production and the most hydraulically effective ones may be partially cemented fractures and shear (or shearreactivated) fractures with mismatching walls (e.g. Barr). Borehole image logging tools are available that can detect fractures with varying degrees of success (Ma¨kel). All image logging techniques have their limitations and ultimately benefit from calibration against overlapping core and all suffer to some degree from an inability to detect fractures subparallel to the wellbore (Ma¨kel). For those fractures that can be detected, a sampling correction may be made for the intersection angle between fracture and wellbore (Terzaghi 1965; Ma¨kel). Fractures perpendicular to the well will be overrepresented relative to fractures oblique to the well and this must be corrected in order to model the 3D fracture density in the surrounding reservoir. However, well productivity may be more influenced by the number of fractures intersected (a well drilled perpendicular to fracture strike will be more productive than one drilled parallel to strike)
and uncorrected data may be preferable if that is the objective of the study. Generally, fracture description and prediction or interpolation in the petroleum industry is not carried out in isolation, but combined with integrated reservoir description and simulation modelling. The process of populating and upscaling a fracture model is conceptually similar to that for a matrix model but generally more difficult and less advanced. Techniques include simple interpolation between wells through conventional geostatistical techniques such as kriging (Olarewaju et al. 1997), neural networks (Ouenes & Hartley 2000) and methods which simultaneously incorporate static and dynamic data (Gauthier et al. 2002). Much modern effort is devoted to the construction and analysis of a discrete fracture network (DFN) model, in which a stochastically or (rarely) deterministically generated set of fractures is populated into a map or a 3D volume. Stochastic models are typically conditioned on seismic, geometric or geomechanical inputs (e.g. Bourbiaux et al. 2002; Maerten et al. 2006; Barr et al. 2007). Ma¨kel provides a worked example for a dataset comprising three wells. Traditionally, the fracture model has been built independently of the matrix, which can result in unwanted interactions, e.g. open fractures may be modelled in shales where observation shows they are absent. Most modelling packages now offer some ability to condition the DFN on a layer-cake or geo-cellular matrix model, enabling the modeller to control the effect of mechanical stratigraphy better (the influence of matrix lithology, typically some combination of rock-strength parameters, on fracture initiation and growth). Fracture models are typically built at multiple scales, to represent both discrete conductive faults or ‘fracture corridors’ and small, dispersed fractures or joints. Calibration can be at both the full-field scale and the scale of a well test (e.g. Rawnsley & Wei 2001).
Flow modelling and reservoir simulation The typical situation whereby fractures dominate the flow domain, and matrix the storage domain; means that ideally, fractures and matrix should be kept independent of one another during flow simulation modelling, by use of explicit fracture and matrix cells. The complexity of fracture networks and the sheer number of simulation grid cells required mean that this is rarely done at scales larger than that of a well test, and even then complex models are difficult to represent fully (Basquet et al. 2005). The extreme aspect ratios, low pore volumes and large permeabilities of fractured cells also create computational difficulties for simulators optimized to solve problems
STRUCTURALLY COMPLEX RESERVOIRS
involving more-or-less cuboidal cells having much the same pore volumes and permeabilities in each. In traditional simulators the simplest solution is to upscale the fractures to effective porosity and permeability at the matrix grid-cell size and merge them with the matrix description. That is successful only in the simplest of cases, typically involving single-phase flow in, for example, dry gas fields without an active aquifer. More complex cases are handled by the dual porosity and dual permeability formulations, in which parallel fracture and matrix descriptions are carried in two identical framework grids, with a transfer function controlling flow between the two domains (Warren & Root 1965; Kazemi et al. 1976). The dual porosity approach allows matrix to fracture fluid flow but not the reverse and is suited to cases where matrix permeability is sufficiently low to be neglected. The dual permeability approach allows flow in both directions and is suited to reservoirs with significant matrix permeability, but is much more computationally demanding. The limitations of these approaches are well known (see Mattha¨i et al. for a summary), particularly with respect to multiphase flow, and research is in progress to develop better framework descriptions (unstructured grids based on tetrahedral or polyhedral cells rather than cuboids, e.g. Mattha¨i et al.; Tertois & Mallet) and alternative solver approaches (e.g. Mattha¨i et al.).
Stress-sensitive reservoirs and critically stressed faults Much fractured reservoir description takes a relatively static view of the fracture network and its flowing properties. Geomechanical models are used to predict fracture occurrence much more often than to represent production-induced changes in the fracture network. Where the matrix can be considered unchanging, the elastic response of pre-existing fractures to changing pressure and stress can be measured or modelled (e.g. Jones 1975; Bagheri & Settari 2005). Production-induced fluid pressure changes can do more than just change the aperture of existing fractures by elastic or plastic opening or closing. They can also reactivate preexisting faults and fractures, create new fractures and deform the rock matrix. An extreme example of the latter effect is seen in the Valhall and Ekofisk fields offshore Norway (Agarwal et al. 1997; Zoback & Zinke 2002; Barkved et al. 2003; Toublanc et al. 2005), where overpressured and undercompacted chalk underwent dramatic production-induced changes in porosity and, particularly, permeability. In such stress-sensitive reservoirs it can be difficult to distinguish matrix
17
from fracture response. Zhang et al. describe a scenario in which they simulate the geomechanical response of a faulted and fractured reservoir to hydrocarbon production and water injection. Many fractured reservoirs are geomechanically insensitive (or at least sufficiently so that it can be treated as a second-order effect) or have a sufficiently homogeneous and reversible response that they can be adequately modelled by introducing pressure-sensitive permeability modifiers to the flow simulation model. Others are geomechanically sensitive and require a coupled simulation modelling approach (e.g. Koutsabeloulis & Hope 1998; Maillot et al. 1999; Settari & Walters 1999; Bagheri & Settari 2005). Acute sensitivity to stress or fluid pressure perturbations which are small relative to the total stress state are a characteristic of critically stressed geological systems. In a reservoir context the critically stressed elements are typically faults that are on the verge of frictional slip or failure (e.g. Zhang et al.). In a broader context the same phenomenon is seen with earthquakes, where small stress perturbations can lock or release previously unstable or stable fault segments (Scholz 1990; Harris 1998). Critically stressed faults can be considered to buffer the subsurface stress state; stress cannot be increased substantially without activating some faults and relieving the stress increase. It is not necessary that every fault is critically stressed, only that the system as a whole is locally near failure. Near-critically stressed faults are likely to have slipped in the recent past in response to tectonic and other stress perturbations; as brittle faulting typically causes dilation, such faults may act as fluid conduits (e.g. Barton et al. 1995; Sanderson & Zhang 2004). That forms the basis for some fracture modelling approaches, where the proximity to failure of each mapped fault, and of smaller features such as subseismic faults or an idealized Andersonian joint set, is used as a proxy for fracture conductivity. Microseismic recording of production or hydraulicfracturing induced earthquakes (e.g. Raleigh et al. 1976; Shapiro et al. 1997, 1999; Segall & Fitzgerald 1998; Rutledge et al. 1998, 2004; Maxwell et al. 2006) suggests that some reservoirs are at least close to being critically stressed, although interpretation may be complicated by poro-elastic effects and/or matrix compaction (e.g. Zoback & Zinke 2002). Where there is observational evidence for fault reactivation, the original proximity of the fault to failure can be inferred if the magnitude of the pore pressure perturbation is known (e.g. Raleigh et al. 1976). In other cases independently determined stress and fluid pressure conditions may lie close to the frictional failure envelope for favourably oriented reservoir faults. A critical
18
S. J. JOLLEY ET AL.
state can be inferred from the stress-buffer argument, although care must be taken to avoid circularity where some of the input parameters were estimated by assuming a critically stressed state. The extent to which reservoirs in general or only certain faults are critically stressed is debated, although it is strongly indicated in cases where in-situ stress or fracture orientations change across them (e.g. Finkbeiner et al. 1997). Heffer et al. (1995) and Heffer (2002) have documented directionality in flow and pressure transmission during water injection in nominally unfractured reservoirs. The direction is consistent with the predicted strike of those faults which were closest to failure, implying that they were already hydraulically conductive or became so as a result of the fluid pressure increase caused by injection. The inclusion or exclusion of such effects in field simulations can have a significant impact on waterflood sweep efficiency; an unrecognized flow directionality will probably mean that most of the injectors are in the wrong place. Critically or near-critically stressed faults can also pose a drilling hazard. Mud losses may occur if the fault is reactivated by drilling mud pressure, which is typically hundreds or thousands of psi higher than formation pressure; and a fault that has previously been reactivated by injection may be associated with a swarm of open fractures which takes losses even if the drilling mud pressure is too low to cause renewed activation. Conversely, such a fault may be responsible for a formation fluid inflow or ‘kick’ if it communicates directly with an active water injector. A bridge between the categorization of structurally complex reservoirs into faulted and fractured types is provided by Main et al. who document the geomechanical response of a conventional faulted reservoir to water injection. Fault behaviour (as barriers or pathways for enhanced flow) changed during field production but in a complex manner related to stress release or transfer along and between faults. It is likely that most of the permeability increases were due to small-scale fracturing, which locally enhanced fluid flow without necessarily forming a widely connected fracture network. In a sense they describe a fractured reservoir, but perhaps one that was not a fractured reservoir prior to water injection and which might never have displayed such behaviour if produced by primary depletion only.
growing understanding and management of structurally complex reservoirs. These include: 1.
2.
3.
4.
5.
6.
7.
Concluding remarks The papers contained in this volume allow the identification of some priority future directions, where continued research and improved application, calibration and validation will add to the
The vital importance of generating robust 3D structural models as a platform for the detailed modelling of complex faulted or fractured reservoirs. Many reservoir studies suffer from the relatively poor quality of basic fault mapping, a shortcoming which is not compensated for by the application of progressively more sophisticated modelling techniques further along the workflow. The importance of applying new methods for the inclusion of fault properties in reservoir models. These methods provide a means of performing geologically refined history matches, and an improved basis for defining fault properties and for production forecasting. The need to develop effective upscaling workflows to ensure inclusion of subseismic structural complexity at the right level of detail in flow simulations. This requirement applies not only to the inclusion of subseismic faults and damage zones, but also to the preservation of features that may be below the resolution of a simulation grid (e.g. relays). The requirement to progress beyond the notion of a single deterministic model, to incorporate the broad range of fault- and structure-related uncertainties. New methods can be used to incorporate uncertainties, but their widespread application may require a change in culture, together with the general acceptance that high quality geologically-refined production forecasting takes time! The importance of improving the links between flow and mechanical feedback processes for the complex stress paths, reactivation of faults and other dynamic fracture damage experienced by reservoirs. Algorithmic improvements in geomechanical simulation (and reductions in computer costs) would extend the range of fields where coupled flow simulation modelling is routinely applied. The introduction of geomechanical rigour into kinematic structural restoration and of kinematic constraints into geomechanical simulations would yield a coupled approach which simplifies both sets of constraints simultaneously with consequential benefits for flow prediction, particularly in fractured reservoirs. Understanding the role of structure in reservoir and fluid behaviour on timescales much longer than production, will yield important insights, primarily in the context of CO2 sequestration. There are lessons to be learned here from the contrasting behaviour of faults and fractures in conventional reservoirs, as observed when
STRUCTURALLY COMPLEX RESERVOIRS
8.
comparing static (exploration) performance with dynamic (production) performance. Cross-learning is available from the mining, toxic and radioactive waste disposal industries, which have had to make similar judgements about the long-term behaviour of fault or stratigraphic seals and fractures. The importance of improving our understanding and ability to model multiphase fluid flow behaviour in both fault-seal and fractured reservoir environments. This will require not just conceptual modelling and computational advance but also laboratory measurements at the limit of current technology and careful calibration against dynamic oil and gas field data. The introduction of a potentially miscible phase in the form of carbon dioxide introduces additional complexity to tertiary recovery and/ or long-term sequestration plans.
References Full reference details for the papers presented in this book are provided in the prelim pages. A INSWORTH , R. B. 2006. Sequence stratigraphic-based analysis of reservoir connectivity: influence of sealing faults – a case study from a marginal marine depositional setting. Petroleum Geoscience, 12, 127– 141. A GARWAL , B., A LLEN , L. R. & F ARRELL , H. E. 1997. Ekofisk Field reservoir characterization: mapping permeability through facies and fracture intensity. Society of Petroleum Engineers Formation Evaluation, December 1997, 227– 233. A GUILERA , R. 1995. Naturally Fractured Reservoirs. Pennwell Publishing Company, Tulsa, Oklahoma. A L – B USAFI , B., F ISHER , Q. J. & H ARRIS , S. D. 2005. The importance of incorporating the multiphase flow properties of fault rocks into production simulation models. Marine and Petroleum Geology, 22, 365– 374. A LLAN , J. & Q ING S UN , S. 2003. Controls on recovery factor in fractured reservoirs: lessons learned from 100 fractured fields. Society of Petroleum Engineers Paper, 84590. A LLAN , U. S. 1989. Model for hydrocarbon migration and entrapment within faulted structures: American Association of Petroleum Geologists Bulletin, 73, 803–811. A NTONELLINI , A. & A YDIN , A. 1994. Effect of faulting on fluid flow in porous sandstones: petrophysical properties. American Association of Petroleum Geologists Bulletin, 78, 355–377. A YDIN , A. & E YAL , Y. 2002. Anatomy of a normal fault with shale smear: Implications for fault seal. American Association of Petroleum Geologists Bulletin, 86, 1367–1381. B ADLEY , M. E., F REEMAN , B., R OBERTS , A. M., T HATCHER , J. S., W ALSH , J. J., W ATTERSON , J. & Y IELDING , G. 1990. Fault interpretation during seismic interpretation and reservoir evaluation. In: The integration of geology, geophysics, petrophysics
19
and petroleum engineering in reservoir delineation, description and management. Proceedings of the 1st Archie Conference, Houston, Texas. Association of American Petroleum Geologists, 224 –241. B AGHERI , M. & S ETTARI , A. 2005. Modelling of geomechanics in naturally fractured reservoirs. Society of Petroleum Engineers Paper, 93083. B AHORICH , M. S. & F ARMER , S. L. 1995, 3 –D Seismic coherency for faults and stratigraphic features; the coherence cube. The Leading Edge, 14, 1053–1058. B ARKVED , O., H EAVEY , P., K JELSTADLI , R., K LEPPAN , T. & K RISTIANSEN , T. G. 2003. Valhall field – still on plateau after 20 years of production. Society of Petroleum Engineers Paper, 83957. B ARNETT , J. A. M., M ORTIMER , J., R IPPON , J. H., W ALSH , J. J. & W ATTERSON , J. 1987. Displacement geometry in the volume containing a single normal fault. American Association of Petroleum Geologists Bulletin, 71, 925– 937. B ARR , D., S AVORY , K. E., F OWLER , S. R., A RMAN , K. & M C G ARRITY , J. P. 2007. Pre-development fracture modelling in the Clair Field, west of Shetland. In: L ONERGAN , L., J OLLY , R. J. H., R AWNSLEY , K. & S ANDERSON , D. J. (eds) Fractured Reservoirs. Geological Society, London, Special Publications, 270, 205– 225. B ARTON , C. A., Z OBACK , M. D. & M OOS , D. 1995. Fluid flow along potentially active faults in crystalline rock. Geology, 23, 683– 686. B ASQUET , R., C OHEN , C. E. & B OURBIAUX , B. 2005. Fracture flow property identification: an optimized implementation of discrete fracture network models. Society of Petroleum Engineers Paper, 93748. B EACH , A., W ELBON , A. I., B ROCKBANK , P. J. & M C C ALLUM , J. E. 1999. Reservoir damage around faults: outcrop examples from the Suez rift. Petroleum Geology, 5, 109– 116. B ELFIELD , W. C. 1998. Incorporating spatial distribution into stochastic modelling of fractures: multifractals and Le´vy-stable statistics. Journal of Structural Geology, 20, 473– 486. B ERG , S. S. & S KAR , T. 2005. Controls on damage zone asymmetry of a normal fault zone: outcrop analyses of a segment of the Moab fault, SE Utah. Journal of Structural Geology, 27, 1803–1822. B ERGBAUER , S. & P OLLARD , D. D. 2003. How to calculate normal curvatures of sampled geological surfaces. Journal of Structural Geology, 25, 277–289. B ERGBAUER , S., M UKERJI , T. & H ENNINGS , P. 2003. Improving curvature analyses of deformed horizons using scale-dependent filtering techniques. American Association of Petroleum Geologists Bulletin, 87, 1255– 1272. B ILLI , A., S ALVINI , F. & S TORTI , F. 2003. The damage zone-fault core transition in carbonate rocks: implications for fault growth, structure and permeability. Journal of Structural Geology, 25, 1779– 1794. B LOCH , G., E L D EEB , M., B ADAAM , H., C AILLY , F., L ECANTE , G., F ONTA , O. & M EUNIER , A. 2003. Seismic facies analysis for fracture detection: a powerful technique. Society of Petroleum Engineers Paper, 81526. B OUSKA , J. & J OHNSTON , R. 2005. The first 3D/4-C ocean bottom seismic surveys in the Caspian Sea:
20
S. J. JOLLEY ET AL.
Acquisition design and processing strategy. The Leading Edge, 24, 910– 921. B OURBIAUX , B. J., C ACAS , M. -C., S ARDA , S. & S ABATHIER , J. C. 1997. A fast and efficient methodology to convert fractured reservoir images into a dual-porosity model. Society of Petroleum Engineers Paper, 38907. B OURBIAUX , B., B ASQUET , R., C ACAS , M.-C., D ANIEL , J.-M. & S ARDA , S. 2002. An integrated workflow to account for multi-scale fractures in reservoir simulation models: implementation and benefits. Society of Petroleum Engineers Paper, 78489. B OUVIER , J. D., K AARS -S IJPESTEIJN , C. H., K LEUSNER , D. F., O NYEJEKWE , C. C. & VAN DER P AL , R. C. 1989. Three-dimensional seismic interpretation, and fault sealing investigations, Nun River field, Nigeria. American Association of Petroleum Geologists Bulletin, 73, 1397–1414. B ROWN , A. A. 2003. Capillary pressure and fault sealing. American Association of Petroleum Geologists Bulletin, 87, 381 –396. C AINE , J. S., E VANS , J. P. & F ORSTER , C. B. 1996. Fault zone architecture and permeability structure. Geology, 24, 1025–1028. C ARTWRIGHT , J. A. 1994. Episodic basin-wide hydrofracturing of overpressured Early Cenozoic mudrock sequences in the North Sea Basin. Marine and Petroleum Geology, 11, 587 –607. C HESTER , F. M. & C HESTER , J. S. 1998. Ultracataclasite structure and friction processes of the punchbowl fault, San Andreas System, California. Tectonophysics, 295, 199– 221. C HILDS , C., W ALSH , J. J. & W ATTERSON , J. 1990. A method for estimation of the density of fault displacements below the limits of seismic resolution in reservoir formations. In: B ULLER , A. T., B ERG , E., H JELMELAND , O., K LEPPE , J., T ORSÆTER , O. & A ASEN , J. O. (eds) North Sea Oil and Gas Reservoirs II. Graham & Trotman, London, 193 – 203. C HILDS , C., W ATTERSON , J. & W ALSH , J. J. 1995. Fault overlap zones within developing normal fault systems. Journal of the Geological Society, London 152, 535– 549. C HILDS , C., W ALSH , J. J. & W ATTERSON , J. 1997. Complexity in fault zone structure and implications for fault seal prediction. In: M ØLLER -P EDERSEN , P. & K OESTLER , A. G. (eds) Hydrocarbon seals Importance for exploration and production. Norwegian Petroleum Society (NPF), Special Publication 7, 61–72. C ORRIGAN , A. F. 1993. Estimation of recoverable reserves: the geologists job. In: P ARKER , J. R. (ed.) Petroleum Geology of Northwest Europe: Proceedings of the 4th Conference. The Geological Society, 1473–1481. C OWARD , M. P., D ALTABAN , T. S. & J OHNSON , H. (eds) 1998. Structural geology in reservoir characterization. Geological Society, London, Special Publications, 127. C OWIE , P. A. & S CHOLZ , C. H. 1992. Displacement – length scaling relationships for faults: data synthesis and discussion. Journal of Structural Geology, 14, 1149–1156.
C OWIE , P. A., K NIPE , R. J. & M AIN , I. G. 1996. Scaling laws for fault and fracture populations - analyses and applications. Journal of Structural Geology, 18, v– xi. C RAMPIN , S. 1981. A review of wave motion in anisotropic and cracked elastic media. Wave Motion, 3, 343–391. C ROUCH , S. L. & S TARFIELD , A. M. 1983. Boundary element methods in solid mechanics: with applications in rock mechanics and geological engineering. Allen and Unwin, Winchester, Massachusetts. D AHLSTROM , C. D. A. 1969. Balanced cross-sections. Canadian Journal of Earth Sciences, 6, 743– 757. D AKE , L. P. 2001. Practice of Reservoir Engineering. Elsevier, Amsterdam, Development in Petroleum Science, 36, 570pp. D OUGHTY , P. T. 2003. Clay smear seals and fault sealing potential of an exhumed growth fault, Rio Grande rift, New Mexico. American Association of Petroleum Geologists Bulletin, 87, 427– 444. D OWNEY , M. 1994. Hydrocarbon Seal Rocks. In: M AGOON , L. B. & D OW , W. G. (eds) The Petroleum System - From Source to Trap. American Association of Petroleum Geologists, Memoir 60, 159– 164. D UNBAR , J. A. & C OOK , R. W. 2003. Palinspastic reconstruction of structure maps: an automated finite element approach with heterogeneous strain. Journal of Structural Geology, 26, 1021–1036. E ICHHUBL , P., D’O NFRO , P. S., A YDIN , A., W ATERS , J. & M C C ARTY , D. K. 2005. Structure, petrophysics, and diagenesis of shale entrained along a normal fault at Black Diamond Mines, California – Implications for fault seal. American Association of Petroleum Geologists Bulletin, 89, 1113–1137. F INKBEINER , T., B ARTON , C. A. & Z OBACK , M. D. 1997. Relationships among in-situ stress, fractures and faults, and fluid flow: Monterey Formation, Santa Maria Basin, California. American Association of Petroleum Geologists Bulletin, 81, 1975–1999. F ISHER , Q. J. & K NIPE , R. J. 1998. Fault sealing processes in siliciclastic sediments. In: J ONES , G., F ISHER , Q. J. & K NIPE , R. J. (eds) Faulting, Fault Sealing and Fluid Flow in Hydrocarbon Reservoirs. Geological Society, London, Special Publications, 147, 117– 134. F ISHER , Q. J. & K NIPE , R. J. 2001. The permeability of faults within siliciclastic petroleum reservoirs of the North Sea and Norwegian Continental Shelf. Marine and Petroleum Geology, 18, 1063–1081. F ISHER , Q. J., K NIPE , R. J. & W ORDEN , R 2000. The microstructure of deformed and undeformed sandstones from the North Sea: its implications for the origin of quartz cement. In: W ORDEN , R. H. & M ORAD , S. (eds) Quartz Cementation in Sandstones. International Association of Sedimentology, Special Publication, 29, 129–146. F ISHER , Q. J., H ARRIS , S. D., M C A LLISTER , E., K NIPE , R. J. & B OLTON , A. J. 2001. Hydrocarbon flow across sealing faults: theoretical constraints. Marine and Petroleum Geology, 18, 251– 257. F ISHER , Q. J., C ASEY , M., H ARRIS , S. D. & K NIPE , R. J. 2003. The fluid flow properties of faults in sandstone: the importance of temperature history. Geology, 31, 965–968. F OXFORD , K. A., W ALSH , J. J., W ATTERSON , J., G ARDEN , I. R., G USCOTT , S. C. & B URLEY , S. D.
STRUCTURALLY COMPLEX RESERVOIRS 1998. Structure and content of the Moab Fault Zone, Utah, U.S.A., and its implications for fault seal prediction. In: J ONES , G., F ISHER , Q. J. & K NIPE , R. J. (eds) Faulting, Fault Sealing and Fluid Flow in Hydrocarbon Reservoirs. Geological Society, London, Special Publications, 147, 87– 103. F ULLJAMES , J. R., Z IJERVELD , L. J. J. & F RANSSEN , R. C. M. W. 1997. Fault seal process: systematic analysis of fault seals over geological and production time scales. In: M ØLLER -P EDERSEN , P. & K OESTLER , A. G. (eds) Hydrocarbon Seals: Importance for Exploration and Production. Norwegian Petroleum Society, Special Publication 7, 51–60. G AUTHIER , B. D. M. & L AKE , S. D. 1993. Probabilistic modelling of faults below the limit of seismic resolution in Pelican Field, North Sea. American Association of Petroleum Geologists Bulletin, 77, 761–777. G AUTHIER , B. D. M., G ARCIA , M. & D ANIEL , J. -M. 2002. Integrated fractured reservoir characterization: a case study in a North Africa field. Society of Petroleum Engineers Paper, 79105. G IBBS , A. D. 1983. Balanced cross section construction from seismic sections in areas of extensional tectonics. Journal of Structural Geology, 5, 153– 160. G IBSON , R. G. 1998. Physical character and fluid flow properties of sandstone-derived fault zones. In: C OWARD , M. P., D ALTABAN , T. S. & J OHNSON , H. (eds) Structural Geology in Reservoir Characterization. Geological Society, London, Special Publications, 127, 82– 98. H ALL , S. A. & K ENDALL , J. M. 2003. Fracture characterization at Valhall: application of P-wave AVOA analysis to a 3D ocean-bottom data set. Geophysics, 68, 1150–1160. H ARRIS , R. 1998. Introduction to special section: Stress triggers, stress shadows, and implications for seismic hazard: Journal of Geophysical Research, 103, 24347– 24358. H ARRIS , S. D., M C A LLISTER , E., K NIPE , R. J., E LLIOTT , L. & I NGHAM , D. B. 1999. Scaling of fluid flow associated with flow through fault damage zones and networks. In: L IPPARD , S. J., N ÆSS , A. & S INDING -L ARSEN , R. (eds) Proceedings of the 5th Annual Conference, International Association for Mathematical Geology, 711 –716. H ARRIS , S. D., M C A LLISTER , E., K NIPE , R. J. & O DLING , N. E. 2003. Predicting the three-dimensional population characteristics of fault damage zones: a study using stochastic models. Journal of Structural Geology, 25, 1281– 1299. H ARRIS , S. D., F ISHER , Q. J., K ARIMI -F ARD , M., V ASZI , A. Z. & W U , K., 2005. Modelling the effects of faults and fractures on fluid flow in petroleum reservoirs. In: I NGHAM , D. B. & P OP , I. (eds) Transport Phenomena in Porous Medi. Volume III. Elsevier, Netherlands, 441–476. H EATH , A. E., W ALSH , J. J. & W ATTERSON , J. 1994. Estimation of the effects of subseismic sealing faults in effective permeabilities in sandstone reservoirs. In: A ASEN , J. O., B ERG , E., B ULLER , A. T., H JELMELAND , O., H OLT , R. M., K LEPPE , J. & T ORSÆTER , O. (eds) North Sea Oil and Gas Reservoirs – III. Kluwer Academic, Trondheim, Norway, 173–183.
21
H EDLAND , C. A. 1997. Fault-propagation, ductile strain, and displacement-distance relationships. Journal of Structural Geology, 19, 249– 256. H EFFER , K. J. 2002. Geomechanical influences in water injection projects: An overview. Oil & Gas Science and Technology Review. Institut Franc¸ais du Pe´trole, 57, (5), 415–422. H EFFER , K. J. & B EVAN , T. 1990. Scaling relationships in natural fractures – Data, theory and applications. Society of Petroleum Engineers Paper, 20981. H EFFER , K. J., F OX , R. J. & M C G ILL , C. A. 1995. Novel techniques show links between reservoir flow directionality, earth stress, fault structure and geomechanical changes in mature waterfloods. Society of Petroleum Engineers Paper, 30711. H ESTHAMMER , J. & F OSSEN , H. 2000. Uncertainties associated with fault sealing analysis. Petroleum Geoscience, 6, 37–45. H ESTHAMMER , J., J OHANSEN , T. E. S. & W ATTS , L. 2000. Spatial relationships within fault damage zones in sandstone. Marine and Petroleum Geology, 17, 873– 893. H ESTHAMMER , J., L ANDRØ , M. & F OSSEN , H. 2001. Use and abuse of seismic data in reservoir characterization. Marine and Petroleum Geology, 18, 635– 655. H ESTHAMMER , J., B JØRKUM , P. A. & W ATTS , L. 2002. The effect of temperature on sealing capacity of faults in sandstone reservoirs: examples from the Gullfaks and Gullfaks Sør fields, North Sea. American Association of Petroleum Geologists Bulletin, 86, 1733– 1751. H OBBS , B. E., M EANS , W. D. & W ILLIAMS , P. F. 1976. An Outline of Structural Geology. Wiley, New York. H UDSON , J. A. 1981. Wave speeds and attenuation of elastic waves in materials containing cracks. Geophysical Journal of the Royal Astronomical Society, 64, 133– 150. I NGRAM , G. M. & U RAI , J. L. 1999. Top-seal leakage through faults and fractures: the role of mudrock properties. In: A PLIN , A. C., F LEET , A. J. & M ACQUAKER , J. H. S. (eds) Muds and Mudstones: Physical and Fluid Flow Properties. Geological Society, London, Special Publications, 158, 125–135. J ACKSON , M. P. A. 1995. Retrospective salt tectonics. In: J ACKSON , M. P. A., R OBERTS , D. G. & S NELSON , S. (eds) Salt Tectonics: a Global Perspective. American Association of Petroleum Geologists, Memoir, 65, 1– 28. J AMES , W. R., F AIRCHILD , L. H., N AKAYAMA , G. P., H IPPLER , S. J. & V ROLIJK , P. J. 2004. Fault-seal analysis using a stochastic multifault approach. American Association of Petroleum Geologists Bulletin, 88, 885– 904. J OLLEY , S. J., D IJK , H., L AMENS , J. H., F ISHER , Q. J., M ANZOCCHI , T., E IKMANS , H. & H UANG , Y. 2007. Faulting and fault sealing in production simulation models: Brent Province, northern North Sea. Petroleum Geoscience, 13, 321– 340. J ONES , F. O. 1975. A laboratory study of the effects of confining pressure on fracture flow and storage in carbonate rocks. Journal of Petroleum Technology January 1975, 21– 27. Society of Petroleum Engineers Paper, 4569.
22
S. J. JOLLEY ET AL.
J ONES , G. & K NIPE , R. J. 1996. Seismic attribute maps: application to structural interpretation and fault seal analysis in the North Sea basin. First Break, 14, 449– 461. J ONES , G, F ISHER , Q. J. & K NIPE , R. J. (eds) 1998. Faulting, Fault Sealing and Fluid Flow in Hydrocarbon Reservoirs. Geological Society, London, Special Publications 147. J OURDE , H., F LODIN , E. A., A YDIN , A., D URLOFSKY , L. J. & W EN , X. -H. 2002. Computing permeability of fault zones in eolian sandstones from outcrop measurements. American Association of Petroleum Geologists Bulletin, 86, 1187– 1200. K AZEMI , H., M ERRIL , L. S., J R ., P ORTERFIELD , K. L. & Z EMAN , P. R. 1976. Numerical simulation of water– oil flow in naturally fractured reservoirs. Society of Petroleum Engineers Paper, 5719. K OUTSABELOULIS , N. C. & H OPE , S. A. 1998. Coupled stress/fluid-thermal multi-phase reservoir simulation studies incorporating rock mechanics. Society of Petroleum Engineers Paper, 47393. K NAI , T. A. & K NIPE , R. J. 1998. The impact of faults on fluid flow in the Heidrun Field. In: J ONES , G., F ISHER , Q. J. & K NIPE , R. J. (eds) Faulting, Fault Sealing and Fluid Flow in Hydrocarbon Reservoirs. Geological Society, London, Special Publications, 147, 269–282. K NIPE , R. J. 1989. Deformation mechanisms - recognition from natural tectonites. Journal of Structural Geology, 11, 127 –146. K NIPE , R. J. 1993. The influence of fault zone processes and diagenesis on fluid flow. In: H ORBURY , A. D. & R OBINSON , A. D. (eds) Diagenesis and Basin Development. American Association of Petroleum Geologists Studies in Geology, 36, 135–154. K NIPE , R. J., F ISHER , Q. F., J ONES , G. ET AL . 1997. Fault seal analysis: successful methodologies, application and future directions. In: M ØLLER -P EDERSEN , P. & K OESTLER , A. G. (eds.) Hydrocarbon Seals: Importance for Exploration and Production. Norwegian Petroleum Society, Special Publication 7, 15– 40. K ULANDER , B. R., D EAN , S. L. & W ARD , B. J. 1990. Fractured Core Analysis: Interpretation, Logging and Use of Natural and Induced Fractures in Core. American Association of Petroleum Geologists Methods in Exploration Series, 8. L EVEILLE , G. P., K NIPE , R. J., M ORE , C., E LLIS , D., D UDLEY , G., J ONES , G., F ISHER , Q. J. & A LLINSON , G. 1997. Compartmentalisation of Rotliegendes gas reservoirs by sealing faults, Jupiter Fields area, southern North Sea. In: Z IEGLER , K., T URNER , P. & D AINES , S. R. (eds) Petroleum Geology of the Southern North Sea; Future Potential. Geological Society, London, Special Publications, 123, 87–104. L EWIS , H. G UEST , J, H AMMOND , L. & H ALL , S. A. 2004. Kinematics to Geomechanics - an Exercise in Predicting Reservoir Deformation and Flow. American Association of Petroleum Geologists Abstracts with Programmes, Dallas, Texas. http://aapg.confex.com/ aapg/da2004/techprogram/A87944.htm L IA , O., O MRE , H., T JELMELAND , H., H OLDEN , L. & E GELAND , T. 1997. Uncertainties in reservoir production forecasts. American Association of Petroleum Geologists Bulletin, 81, 775–802.
L INDSAY , N. G., M URPHY , F. C., W ALSH , J. J. & W ATTERSON , J. 1993. Outcrop studies of shale smears on fault surfaces. In: F LINT , S. T. & B RYANT , A. D. (eds) The Geological Modelling of Hydrocarbon Reservoirs and Outcrop. International Association of Sedimentology, Special Publication, 15, 113–123. L ISLE , R. J. 1994. Detection of zones of abnormal strains in structures using Gaussian curvature analysis. American Association of Petroleum Geologists Bulletin, 78, 1811– 1819. L ONERGAN , L., J OLLY , R. J. H., R AWNSLEY , K. & S ANDERSON , D. J. (eds) 2007. Fractured Reservoirs. Geological Society, London, Special Publications, 270. L YNN , H. B. & T HOMSEN , L. A. 1990. Reflection shearwave data collected near the principal axes of azimuthal anisotropy, Geophysics, 55, 147– 156. M AC B ETH , C. 1995. How can anisotropy be used for reservoir characterization? First Break, 13, 31– 37. M AERTEN , L. & M AERTEN , F. 2006. Chronologic modelling of faulted and fractured reservoirs using geomechanically based restoration: Technique and industry applications. American Association of Petroleum Geologists Bulletin, 90, 1201–1226. M AERTEN , L., G ILLESPIE , P. & D ANIEL , J -M. 2006. Three-dimensional geomechanical modelling for constraint of subseismic fault simulation. American Association of Petroleum Geologists Bulletin, 90, 1337– 1358. M AILLOT , B., N IELSEN , S. & M AIN , I. 1999. Numerical simulation of seismicity due to fluid injection in a brittle poro-elastic medium, Geophysical Journal International, 139, 263– 272. M ANZOCCHI , T. 1999. Towards an improved fault representation in two-phase reservoir simulation. EAGE 61st conference and technical exhibition, Helsinki, Finland, 7 –11 June 1999; Volume 1; C-05. M ANZOCCHI , T., R INGROSE , P. S. & U NDERHILL , J. R. 1998. Flow through fault systems in high porosity sandstones. In: C OWARD , M. P., J OHNSON , H. & D ALTABAN , T. (eds) Structural Geology in Reservoir Characterization. Geological Society, London, Special Publications, 127, 65– 82. M ANZOCCHI , T., W ALSH , J. J., N ELL , P. & Y IELDING , G. 1999. Fault transmissibility multipliers for flow simulation models. Petroleum Geoscience, 5, 53–63. M ANZOCCHI , T., H EATH , A. E., W ALSH , J. J. & C HILDS , C. 2002. The representation of two phase fault-rock properties in flow simulation models. Petroleum Geoscience, 8, 119–132. M ARFURT , K. J., K IRLINZ , R. L., F ARMER , S. L. & B AHORICH , M. S. 1998. 3-D seismic attributes using a semblance-based coherency algorithm. Geophysics, 63, 1150– 1165. M AULTZSCH , S., C HAPMAN , M., L IU , E. & L I , X-Y. 2003. The potential of measuring fracture sizes with frequency-dependent shear-wave splitting. First Break, 21, 45– 51. M AXWELL , S. C., W ALTMAN , C. K., W ARPINSKI , N. R., M AYERHOFER , M. J. & B OROUMAND , N. 2006. Imaging seismic deformation induced by hydraulic fracture complexity. Society of Petroleum Engineers Paper, 102801.
STRUCTURALLY COMPLEX RESERVOIRS M C C LAY , K. R. (ed.) 2004. Thrust Tectonics and Hydrocarbon Systems. American Association of Petroleum Geologists, Memoir, 82. M C C LAY , K. R., D OOLEY , T., W HITEHOUSE , P. & M ILLS , M. 2002. 4-D Evolution of Rift Systems: Insights from Scaled Physical Models. American Association of Petroleum Geologists Bulletin, 86, 935–960. M C G RATH , A. G. & D AVISON , I. 1995. Damage zone geometry around fault tips. Journal of Structural Geology, 17, 1011– 1024. M ØLLER -P EDERSEN , P. & K OESTLER , A. G. (eds) 1997. Hydrocarbon Seals: Importance for Exploration and Production. Norwegian Petroleum Society, Special Publication 7. M URRAY , G. H. 1968. Quantitative fracture study Spanish Pool, McKenzie County, North Dakota. American Association of Petroleum Geologists Bulletin, 52, 57– 65. N EEDHAM , D. T., Y IELDING , G. & F REEMAN , B. 1996. Analysis of fault geometry and displacement patterns. In: B UCHANAN , P. G. & N IEUWLAND , D. A. (eds) Modern Developments in Structural Interpretation, Validation and Modelling. Geological Society, London, Special Publications, 99, 189–199. N ELSON , R. A. 2001. Geologic Analysis of Naturally Fractured Reservoirs. Gulf Professional Publishing, Boston. N ICOL , A., W ALSH , J. J., W ATTERSON , J. & G ILLESPIE , P. A. 1996. Fault size distributions - are they really power law? Journal of Structural Geology, 18, 191–197. N ICOL , A., G ILLESPIE , P. A., C HILDS , C. & W ALSH , J. J. 2002. Relay zones between mesoscopic thrust faults in layered sedimentary sequences. Journal of Structural Geology, 24, 709–727. O DLING , N. E., G ILLESPIE , P., B OURGINE , B. ET AL . 1999. Variations in fracture system geometry and their implications for fluid flow in fractured hydrocarbon reservoirs. Petroleum Geoscience, 5, 373–384. O DLING , N. E., H ARRIS , S. D., V ASZI , A. Z. & K NIPE , R. J. 2005. Properties of fault damage zones in siliclastic rocks: a modelling approach. In: S HAW , R. P. (ed.) Understanding the Micro to Macro Behaviour of Rock– Fluid Systems. Geological Society, London, Special Publications, 249, 43– 59. O LAREWAJU , J., G HORI , S., F USENI , A. & W AJID , M. 1997. Stochastic Simulation of fracture density for permeability field estimation. Society of Petroleum Engineers Paper, 37962. O UENES , A. & H ARTLEY , L. J. 2000. Integrated fractured reservoir modelling using both discrete and continuum approaches. Society of Petroleum Engineers Paper, 62939. O WEN , R. J., L I , X-Y., M AC B ETH , C. & B OOTH , D. C. 1998. Reservoir characterization: how can seismic anisotropy help? In: C OWARD , M. P., D ALTABAN , T. S. & J OHNSON , H. (eds) Structural Geology in Reservoir Characterization. Geological Society, London, Special Publications, 127, 115 –119. P LUMB , R. A. 1994. Influence of composition and texture on the failure properities of clastic rocks. Society of Petroleum Engineers Paper, 28022.
23
R ALEIGH , C. B., H EALY , J. H & B REDEHOEFT , J. D. 1976. An experiment in earthquake control at Rangely, Colorado. Science, 191, 1230–1237. R AMSAY , J. G. 1967. Folding and Fracturing of Rocks. Wiley, New York. R AWNSLEY , K & W EI , L. 2001. Evaluation of a new method to build geological models of fractured reservoirs calibrated to production data. Petroleum Geoscience, 7, 23–33. R EISS , L. H. 1980. The Reservoir Engineering Aspects of Fractured Formations. Editions Technip Paris. R IVENÆS , J. C. & D ART , C. 2002. Reservoir compartmentalization by water-saturated faults - Is evaluation possible with todays tools? In: K OESTLER , A. G. & H UNSDALE , R. (eds) Hydrocarbon seal Quantification. Norsk Petroleumsforenig, Special Publications, 11, 187–201. R UTLEDGE , J. T., P HILLIPS , W. S. & S CHUESSLER , B. H. 1998. Reservoir characterization using oilproduction-induced microseismicity, Clinton County, Kentucky. Tectonophysics, 289, 129 –1532. R UTLEDGE , J. T., P HILLIPS , W. S. & M AYERHOFER , M. J. 2004. Faulting induced by forced fluid injection and fluid flow forced by faulting: an interpretation of hydraulic-fracture microseismicity, Carthage Cotton Valley Gas Field, Texas. Bulletin of the Seismological Society of America, 94, 1817– 1830. R UTTEN , K. W. & V ERSCHUREN , M. A. J. 2003. Building and unfaulting fault-horizon networks. In: N IEUWLAND , D. A. (ed.) New Insights into Structural Interpretation and Modelling. Geological Society, London, Special Publications, 212, 39– 57. S ABATHIER , J. C., B OURBIAUX , B. J., C ACAS , M. C. & S ARDA , S. 1998. A new approach of fractured reservoirs. Society of Petroleum Engineers Paper, 39825. S ANDERSON , D. J. & Z HANG , X. 2004. Stress-controlled localization of deformation and fluid flow in fractured rocks. In: C OSGROVE , J. W. & E NGELDER , T. (eds) The Initiation, Propagation, and Arrest of Joints and Other Fractures. Geological Society, London, Special Publications, 231, 299– 314. S CHOLZ , C. H. 1990. The Mechanics of Earthquakes and Faulting. Cambridge University Press, New York. S EGALL , P. & F ITZGERALD , S. D. 1998. A note on induced stress changes in hydrocarbon and geothermal reservoirs. Tectonophysics, 289, 117–128. S ETTARI , A. & W ALTERS , D. A. 1999. Advances in coupled geomechanical and reservoir modelling with applications to reservoir compaction. Society of Petroleum Engineers Paper, 51927. S HAPIRO , S. A., H UENGES , E. & B ORM , G. 1997. Estimating the crust permeability from fluidinjected-induced seismic emission at the KTB site. Geophysical Journal International, 131, F15– F18. S HAPIRO , S. A., A UDIGANE , P. & R OYER , J.-J. 1999. Large-scale in-situ permeability tensor of rocks from induced microseismicity. Geophysical Journal International, 137, 207 –213. S HIPTON , Z. K., E VANS , J. P. & T HOMPSON , L. B. 2005. The geometry and thickness of deformation-band fault core and its influence on sealing characteristics of deformation-band fault zones. In: S ORKHABI , R. & T SUJI , Y. (eds) Faults, Fluid Flow and Petroleum Traps. American Association of Petroleum Geologists, Memoir, 85, 181–195.
24
S. J. JOLLEY ET AL.
S KERLAC , G. M. 1999. Evaluating top and fault seals. In: B EAUMONT , E. A & F OSTER , N. H. (eds) Handbook of Petroleum Geology: Exploring for Oil and Gas Traps. American Association of Petroleum Geologists. S MITH , R. L. & M C G ARRITY , J. P. 2001. Cracking the fractures – seismic anisotropy in an offshore reservoir. The Leading Edge, 20, 18– 26. S OLIVA , R. & B ENEDICTO , A. 2005. Geometry, scaling relations and spacing of vertically restricted normal faults. Journal of Structural Geology, 27, 317– 325. S ORKHABI , R. & T SUJI , Y. (eds) 2005. Faults, Fluid Flow and Petroleum Traps. American Association of Petroleum Geologists, Memoir, 85. S PERREVIK , S., F ÆRSETH , R. B. & G ABRIELSEN , R. H. 2000. Experiments on clay smear formation along faults. Petroleum Geoscience, 6, 113– 123. S PERREVIK , S., G ILLESPIE , P. A., F ISHER , Q. J., H ALVORSEN , T. & K NIPE , R. J. 2002. Empirical estimation of fault rock properties. In: K OESTLER , A. G. & H UNSDALE , R. (eds) Hydrocarbon Seal Quantification. Norwegian Petroleum Society, Special Publication 11, 109 –125. S TEWART , S. A & P ODOLSKI , R. 1998. Curvature analysis of gridded geological surfaces. In: C OWARD , M. P., D ALTABAN , T. S. & J OHNSON , H. (eds) Structural Geology in Reservoir Characterization. Geological Society, London, Special Publications, 127, 133–147. S WENNEN , R., R OURE , F. & G RANATH , J. W. (eds) 2004. Deformation, Fluid Flow, and Reservoir Appraisal in Foreland Fold and Thrust Belts. American Association of Petroleum Geologists, Hedberg Series, 1, 1– 2. T ERZAGHI , R. D. 1965. Sources of error in joint surveys. Geotechnique, 15, 287–304. T OWNSEND , C., F IRTH , I. R., W ESTERMAN , R., K IRKEVOLLEN , L., H ARDE , M. & A NDERSEN , T. 1998. Small-scale seismic fault identification and mapping. In: J ONES , G., F ISHER , Q. J. & K NIPE , R. K. (eds) Faulting, Fault Sealing and Fluid Flow in Hydrocarbon Reservoirs. Geological Society, London, Special Publications, 147, 1 –25. T OUBLANC , A., R ENAUD , S., S YLTE , J. E., C LAUSEN , C. K., E IBEN , T. & N A˚ DLAND , G. 2005. Ekofisk Field: fracture permeability evaluation and implementation in the flow model. Petroleum Geoscience, 11, 321– 330. VAN DER Z EE , W. & U RAI , J. L. 2005. Processes of normal fault evolution in a siliciclastic sequence: a case study from Miri, Sarawak, Malaysia. Journal of Structural Geology, 27, 2281–2300. V ENDEVILLE , B. C. & J ACKSON , M. P.A. 1992a. The rise of diapirs during thin-skinned extension. Marine and Petroleum Geology, 9, 331– 353.
V ENDEVILLE , B. C. & J ACKSON , M. P.A. 1992b. The fall of diapirs during thin-skinned extension. Marine and Petroleum Geology, 9, 354–371. V ERWEST , B. J. 1994. Seismic migration in elliptically anisotropic media. Journal of Geophysical Prospecting, 37, 149–166. W ALSH , J. J. & W ATTERSON , J. 1987. Distributions of cumulative displacement and seismic slip on a single normal fault surface. Journal of Structural Geology, 9, 1039–1046. W ALSH , J. J. & W ATTERSON , J. 1991. Geometric and kinematic coherence and scale effects in normal fault systems. In: R OBERTS , A. M., Y IELDING , G. & F REEMAN , B. (eds) The Geometry of Normal Faults. Geological Society, London, Special Publications, 56, 193–203. W ALSH , J. J., W ATTERSON , J., H EATH , A. E. & C HILDS , C. 1998. Representation and scaling of faults in fluid flow models. Petroleum Geoscience, 4, 241–251. W ALSH , J. J., N ICOL , A. & C HILDS , C. 2002. An alternative model for the growth of faults. Journal of Structural Geology, 24, 1669–1675. W ARREN , J. E. & R OOT , P. J. 1963. The behaviour of naturally fractured reservoirs. Society of Petroleum Engineers Journal, September 1963, 245– 255 (SPE426). W ATTERSON , J., W ALSH , J. J., N ICOL , A., N ELL , P. A. R. & B RETAN , P. 2000. Geometry and origin of a polygonal fault system. Journal of the Geological Society, London, 157, 151 –162. W INTERSTEIN , D. F. 1989. Velocity anisotropy terminology for geophysicists. Geophysics, 55, 1070– 1088. W ONG , T.-F., D AVID , C. & Z HU , W. 1997. The transition from brittle faulting to cataclastic flow in porous sandstones: Mechanical deformation. Journal of Geophysical Research, 102,B2, 3009– 3025. Y IELDING , G. 2002. Shale gouge ratio – calibration by geohistory. In: K OESTLER , A. G. & H UNSDALE , R.(eds) Hydrocarbon Seal Quantification. Norwegian Petroleum Society, Special Publication 11, 1–15. Y IELDING , G., N EEDHAM , T. & J ONES , H. 1996. Sampling of fault populations using sub-surface data: a review. Journal of Structural Geology, 18, 135– 146. Y IELDING , G., F REEMAN , B. & N EEDHAM , D. T. 1997. Quantitative fault seal prediction. American Association of Petroleum Geologists Bulletin, 81, 897–917. Z HANG , J., W ONG , T.-F. & D AVIS , D. M. 1990. Micromechanics of pressure-induced grain crushing in porous rocks. Journal of Geophysical Research, 95, 341–352. Z OBACK , M. D. & Z INKE , J. C. 2002. Production-induced normal faulting in the Valhall and Ekofisk oil fields. Pure and Applied Geophysics, 159, 403 –420.
Structural evolution of the Penguins Cluster, UK northern North Sea R. DOMI´NGUEZ Shell UK Ltd, 1 Altens Farm Road, Nigg, Aberdeen AB12 3YF (e-mail:
[email protected]) Abstract: The Penguins Cluster is a group of four oil and gas fields in the northern end of the East Shetland Basin. Its structural complexity is caused by the interaction between two or more fault trend populations, fault reactivation and the impact of faulting on the Brent reservoir architecture. This structural picture is further complicated by a NW–SE trending basement lineament interpreted as a Caledonian shear zone. The present day structural configuration is the result of two Mesozoic rifting episodes and their associated thermal subsidence phases. The Permo-Triassic rifting created a number of north– south-trending tilted fault blocks, and was followed by a period of tectonic quiescence until the Middle Jurassic, when a faulting episode coeval with the Brent Group deposition caused footwall rotation, uplift and erosion of the upper Rannoch Formation prior to the deposition of the Etive Formation across the area. The rifting climaxed in the late Jurassic, when the reactivation of pre-existing faults under oblique-slip conditions in the Penguin C Field created small-scale lozenge-shaped transpressional and transtensional fault blocks. The presence of reverse faults in the area is explained with a continuous kinematic model of structural evolution and oblique-slip fault reactivation rather than positive basin inversion.
The Penguins Cluster is a group of four oil and gas fields located at the northern end of the East Shetland Basin (Fig. 1), in Blocks 211/13a and 211/ 14 of the UK northern North Sea, 65 km north of the Brent Field, to which it is tied back (Fig. 2). It was discovered in 1974 by well 211/13-1, which targeted the north–south-trending Penguin Horst, a major Triassic structural feature around which the Penguins Cluster is structured (Fig. 3). The Penguin A field is located in the western flank of the Penguin Horst, with its play being a stratigraphic trap that consists of an up-dip pinchout of the Upper Jurassic Magnus sandstones (Richards et al. 1993; Partington et al. 1993). The Penguin C, D and E fields run from north to south along the eastern flank of the Penguin Horst and share the Brent Group sandstones (Deegan & Scull 1977; Taylor et al. 2003) as their primary reservoir. Here the structural trap is created by a north–south-trend of easterly-tilted Jurassic fault blocks, also known as the Penguins Ridge. Crustal-scale regional cross-sections that pass near the study area based on Yielding et al. 1992 (Fig. 4) reveal a lower section of Triassic and Jurassic tilted fault blocks at a depth of c. 3.0 km overlain by an upper sequence of mostly unfaulted Cretaceous to Cenozoic post-rift sedimentary rocks. The Penguins Cluster is situated on a structurally complex region influenced by two separate structural styles: the Brent Province to the South, mainly affected by north– south-trending faults
related to the opening of the Viking Graben (Badley et al. 1988; Gabrielsen et al. 1999) and the Magnus Province to the NW (De’Ath & Schuyleman 1981; Shepherd et al. 1991) influenced by the opening of the North Atlantic, and mainly affected by NE–SW-trending faults (see Figs 1 & 2). The Penguins Cluster has been on stream since 2003 and, to date, its development has included the drilling of eight sub-horizontal wells aimed at mitigating reservoir compartmentalization caused by the presence of sealing faults and poor intra-reservoir layer connectivity. These wells have sub-horizontal reservoir sections up to 4700 ft long and typically connect two or more fault block compartments. The structural complexity of the Penguins Cluster reservoirs is caused by a combination of four main factors: i) The heavily faulted nature of the producing reservoirs, with fault spacing of less than 100 m in some cases (Penguin D), and the presence of two or more fault trend populations that create intra-field fault block compartments. ii) The sealing nature of some of these faults. Sealing faults have been recognized in Penguin A (different oil compositions sampled at either side of NW –SE-trending faults), Penguin D (oil-bearing reservoir in the north versus a gas accumulation in the south), and Penguin E (different gas– water contact depths logged across faults).
From: JOLLEY , S. J., BARR , D., WALSH , J. J. & KNIPE , R. J. (eds) Structurally Complex Reservoirs. Geological Society, London, Special Publications, 292, 25– 48. DOI: 10.1144/SP292.2 0305-8719/07/$15.00 # The Geological Society of London 2007.
26
R. DOMI´NGUEZ
iii) The complex internal Brent reservoir architecture, with thickness changes of the Rannoch Formation and a patchy distribution of the Etive and Tarbert Formations caused by Middle Jurassic faulting and erosion. iv) The presence of small-scale transpressional and transtensional fault blocks that may pose a threat to the drilling of long horizontal wells.
A better understanding of the structural evolution of Penguins and how this fits within the large-scale evolution of the northern North Sea is key for the future development of the Penguins reservoirs and for the exploration and appraisal of near-field prospects. The model of structural evolution presented in this paper aids seismic interpretation by presenting a clear model of structural styles. The study also
' '
Fig. 1. Structural elements map of the North Sea showing the position of the study area in the northern end of the East Shetland Basin. Inset box shows the location of Figure 2. Interpretation of seismic lines A–A0 and B–B0 is shown in Figure 4. Modified from Faerseth (1996).
PENGUINS CLUSTER STRUCTURAL EVOLUTION 20 Km
N Magnus Embayment A C Magnus
Penguins Cluster
D E Don
Tern-Eider Ridge
Beta Murchison
Eider
Tern
Statfjord
Dunlin East Shetland Basin Cormorant
Pelican
Brent Hutton
East Shetland Platform
North Viking Graben Heather Ninian
UK Sector
Fig. 2. Schematic map of the East Shetland Basin showing the location of the Penguins Cluster in relation to the main structural elements and hydrocarbon fields (shaded red). For location see Figure 1.
contributes to reducing the drilling hazards in complex structural areas by predicting the potential location of subseismic flower structures, as well as improving the geological models built in the future by applying a better knowledge of the Brent Group internal reservoir architecture.
Stratigraphy The general stratigraphy of the Penguins area is illustrated in Figure 5. Sedimentary rocks penetrated by wells in Penguins range in age from the Triassic Cormorant Formation sandstones to Cenozoic unconsolidated sediments. Although the base of the Cormorant Formation has not been drilled in Penguins, it is assumed to lie unconformably on a basement of crystalline Caledonian rocks, as encountered in well 211/21-2 of the North Cormorant Field. In the Penguin A field the reservoir consists of Upper Jurassic Magnus Sandstone Member turbidites, absent along the Penguin Ridge. The deltaic and shallow marine Brent Group sandstones (Deegan & Scull 1977; Taylor et al. 2003) form the primary reservoir of the Penguin
27
C, D and E fields, targeted by several sub-horizontal wells. The Brent Group rests unconformably over the deep marine Dunlin Group claystones, which form the topseal for the underlying of the Hegre Group. This is made up of the Triassic Cormorant Formation sandstones and the Lower Jurassic Statfjord Formation sandstones, and it constitutes a secondary hydrocarbon reservoir under the Penguin Ridge. The Brent Group is an unconformity-bound sedimentary sequence which varies in thickness between 150 and 300 feet over the Penguins fields. It is divided into four lithostratigraphic formations, with the Ness Formation absent over Penguins. At least one internal unconformity can be identified, at the top of the Rannoch Formation, interpreted as the result of syndepositional fault movements. The base of the Brent Group can be seen to be erosional and unconformable in nature over certain parts of the study area, with its base truncating the underlying Dunlin seismic reflectors. The top Brent is a diachronous unconformity surface, with the overlying Upper Jurassic Heather Formation resting unconformably on top of either the Rannoch, Etive, or Tarbert Formations. The Broom Formation is a non-reservoir quality sandstone package that is 40 feet thick on average. The Rannoch Formation is a shoreface sandstone unit that varies in thickness between 50 feet in 211/14-1s2 and 150 feet thick in 211/14-3. It rests conformably over the Broom Formation, and it can be subdivided into up to eight coarseningupwards shoreface parasequences. The Etive Formation has a patchy distribution over the Penguin C field, only found in the southern and northern ends of the field (PENG-C02 and PENG-C03 wells, respectively), and it is present in all wells drilled to date over the Penguin D and E fields. Interpretation from well data shows that it has a rough tabular shape lying unconformably over the Rannoch Formation, with an average thickness of c. 50 feet and a general trend thickening towards the south, although some degree of erosion at the top of this unit cannot be ruled out. To date, the Ness Formation has not been found in any of the Penguins wells, although it has been identified to the NW of Penguins, in well 211/7-1 of the Magnus area, and is also widespread in the Don Field wells to the south and SW. The Tarbert Formation is a thin veneer of sandstones, up to 15 feet thick, also of patchy distribution encountered when present either at the top of the Rannoch or Etive Formations.
Tectonic history The East Shetland Basin is one of several linked arcuate half-grabens that form the northern North
28
R. DOMI´NGUEZ
Fig. 3. Penguins Cluster Base Cretaceous unconformity depth map in feet. Contour interval is 250 feet. The Base Cretaceous unconformity roughly equates to the top reservoir in the Penguin A (top Magnus Sandstone Member) and C, D & E fields (top Brent Group). The location of seismic lines discussed in the text is shown.
Sea (Lee & Hwang 1993; Odinsen et al. 2000; Coward et al. 2003; Zanella & Coward 2003). Internally, it consists of Jurassic and Triassic tilted fault blocks overlain by Cretaceous and Cenozoic sediments. The East Shetland Basin is an eastfacing rhombic-shaped half graben bounded to the west by the Palaeozoic East Shetland Platform, to the east and south by the Viking Graben, and to the north by the ENE– WSW trending Magnus Embayment (see Figs 1 & 2). The Viking Graben is a NNE– SSW trending Mesozoic rift that represents the northern arm of a Jurassic triple rift system in the North Sea (Yielding 1990).
The northern North Sea is the result of multiple stretching during the Mesozoic, with the two main rifting episodes dated as Permo-Triassic and Jurassic (Badley et al. 1984, 1988; Ziegler 1988; Yielding et al. 1992; Færseth 1996), followed by episodes of passive post-rift thermal subsidence in the early–middle Jurassic, and Cretaceous–Cenozoic, respectively. The underlying basement is Palaeozoic in age and is affected by Caledonian structures formed during the progressive collision of Baltica with Laurentia (Jones et al. 1999). The northern North Sea was affected by east – west (Færseth 1996) or NW–SE (Beach 1987)
PENGUINS CLUSTER STRUCTURAL EVOLUTION
29
Fig. 4. Crustal cross-sections of the East Shetland Basin and North Viking Graben, showing approximate location of the study area in Profile A (After Yielding et al. 1992). Location of the lines is shown in Figure 1.
extension that caused the first rifting episode during Permo-Triassic times, followed by tectonic quiescence and post-rift thermal subsidence during the early and middle Jurassic. This Permo-Triassic rifting created a north–south-trending, 180 km wide sedimentary basin (Færseth 1996). Regional seismic lines through the Viking Graben show the Triassic sediments to thicken towards the east, suggesting a relatively deep Triassic basin that had formed in the centre or further east (Færseth 1996) (see Fig. 4). The Permo-Triassic rifting was characterized by north–south-trending faults that form easterly or westerly tilted half grabens. This phase of rifting has been strongly overprinted by late Jurassic extension. There was very little faulting activity during the early to middle Jurassic in the northern North Sea. During this time the area underwent an episode of passive post-rift thermal subsidence during which the Dunlin and Brent Groups were deposited. The second Mesozoic rifting episode affected the northern North Sea during the Jurassic, with the climax of faulting occurring between MidOxfordian to early Kimmeridgian times, and continuing until the early Cretaceous (Rattey & Hayward 1993). There is evidence that faulting may have begun as early as the Middle Jurassic in the northern North Sea (Roberts et al. 1999). This rifting episode developed as a multiple pulses of faulting separated by intervening stages of relative tectonic quiescence; this pattern was a major influence on the nature and architecture of the sediments in the basin (Ravna˚s et al. 2000). Although most authors agree on a rough east– west extension direction for the Jurassic rifting (e.g. Roberts et al. 1990), this structural picture was probably complicated
by slight changes in the extension direction over time. Thus, Thomas & Coward (1995) identified an episode of Oxfordian to Kimmeridgian NW– SE directed extension, followed by a change to NE –SW directed extension during the late Kimmeridgian to early Cretaceous. Overall, the East Shetland Basin was stretched by about 15% during the Jurassic rifting (Roberts et al. 1993). Some authors (Booth et al. 1992; Thomas & Coward 1995) invoke an episode of basin inversion in the East Shetland Basin during the latest Jurassic–early Cretaceous, with strike-slip reactivation of pre-existing structures under a compressional tectonic regime. The tilted fault blocks formed by the Triassic and Jurassic rifting were subsequently eroded by subaerial and shallow-marine processes during the early Cretaceous, and thermal subsidence allowed onlap of sediments onto these eroded blocks, forming the diachronous base Cretaceous unconformity (Zanella & Coward 2003). Following the Jurassic rifting, during the Cretaceous and the Cenozoic, the East Shetland Basin underwent an episode of passive post-rift thermal subsidence, with accumulations of sedimentary rocks up to 4 km thick in the Penguins area.
Structural styles in the Penguins Cluster For discussion purposes, the Penguins Cluster has been subdivided into the following structural areas with characteristic structural styles: the Penguin Horst, the Penguin Lineament, the Penguin Ridge, and the Penguin Basin north and south, as encountered at either side of the Penguin Lineament (Fig. 6).
30
R. DOMI´NGUEZ
Fig. 5. Generalized lithostratigraphy of the Penguins Area.
As it is to be expected within its northern North Sea context, the main structural fabric in Penguins has a north– south orientation, as interpreted from time slices (Fig. 6) and fault maps. However, the detailed structural picture is, much more complex, with three other main fault orientation populations identified: west–east, NW–SE and, to a minor extent, NE–SW (see Fig. 3). This diversity in fault trends can be interpreted as derived from the influence of underlying basement
fabrics, the rotation in extension direction during the Jurasssic rifting, and the influence of the NE – SW trending structural grain of the Magnus structural province to the NW.
Penguin Horst The structure of the Penguins Cluster is dominated by the north–south-trending Penguin Horst, a major Triassic structural high some 3 km wide
PENGUINS CLUSTER STRUCTURAL EVOLUTION
31
Fig. 6. (a) Uninterpreted time slice taken through the 3D seismic volume at 3.3 seconds two-way time, and (b) structural interpretation of the time slice showing the structural grain and the main structural areas discussed in the text.
32
R. DOMI´NGUEZ
Fig. 7. (a) uninterpreted, and (b) interpreted west to east seismic line across the Penguin A Field, Penguin Horst and Penguin C Field.
PENGUINS CLUSTER STRUCTURAL EVOLUTION
which occurs at about 8500 ft below sea level and contrasts with the neighbouring Penguins Fields structured around it and which occur at depths of around 11000 ft below sea level (see Figs 3 & 6). The Penguin Horst lies on a NE–SW trend of structural highs that can be traced north into the Haltenbanken and Nordland terraces of offshore Norway (Thomas & Coward 1995), and SW into the Tern-Eider Ridge. The Penguin Horst is a striking feature observed in seismic (Fig. 7), which made it the original exploration target for Penguins, discovered by well 211/13-1, drilled on its crest. Figure 7 illustrates that the horst is bounded by normal faults at either side (the West Penguin Fault and East Penguin Fault, respectively), contrasting with previous interpretations of it as either a large-scale positive flower structure (see fig. 12 in Booth et al. 1992) or bounded at either side by compressional positive flower structures (see fig. 14 in Thomas & Coward 1995). The more recent 3D pre-SDM seismic data (1998) was acquired and processed before the field development and has greatly improved the imaging of the Penguins Fields with respect to the previous vintage of 1985, on which the Booth et al. (1992) and Thomas & Coward (1995) interpretations were based. The new level of detail allows for a reinterpretation of previous structural geometries and models, particularly the ‘inverted’ nature of the Penguins Horst. The detailed seismic shown in Figure 7 shows a west –east-trending seismic line and seismic interpretation across the Penguin A field, Penguin Horst, and Penguin C field. This line is similar to those discussed in figure 12 of Booth et al. (1992) and figure 14 of Thomas & Coward (1995). This section is dominated by the presence of the Triassic Penguin Horst which occurs at a depth of c. 2.5 seconds two-way time (TWT), and tagged at a subsea depth of c. 8500 ft by exploration well 211/13-1. The Penguin Horst is a pronounced structural high bounded at either side by steep extensional faults that are poorly imaged deeper than 3.0 seconds two-way time. Internally, it consists of a series of easterly-dipping internal Triassic reflectors that continue across the East Penguin Fault below the Penguin C field. The top of the basement occurs at or around 3.0 seconds TWT, forming a broad antiform of chaotic internal reflectors. The top Triassic, top Dunlin, top Brent and base Cretaceous unconformity seismic reflectors can be picked on the eastern side of the Penguin Horst. The Penguin A field occurs in the hanging wall of the West Penguin Fault, adjacent to the Penguin half graben, which displays a synformal shape. A series of deep reflectors at around 3.5 s TWT and deeper indicate the top of the basement under Penguin A. Above these reflectors the
33
seismic shows an easterly tilt of the Triassic sediments, contrasting with the overlying westerlydipping reflectors at the Upper Jurassic level Humber Group. The sediment thickening against the fault planes of the Humber Group indicates these faults were active at the time of deposition. Seismic and well data indicate the Magnus Sandstone Member to be 600 ft true-vertical thickness (TVT) in the depocentre of the Penguins Basin north (see Fig. 7), thinning into the Penguin A field structural high to about 250 ft TVT (211/ 13-3 well) and eventually pinching out to zero thickness on the edge of the Penguins Horst. These faults also show a much larger throw at top basement level than at Jurassic level, pointing towards a reactivation during the late Jurassic of Triassic structures. The rapid subsidence associated with the late Jurassic fault movement is responsible for the westerly dip observed in the Upper Jurassic seismic reflectors that occur under Penguin A. The Magnus turbiditic sandstones do not occur along the Penguins Ridge, indicating this structural high represented an obstacle to the progradation of turbiditic submarine fans coming from the north and east. The interaction between fault movement and sand deposition caused the thinning and pinchout of the Magnus sands towards the Penguin Horst. Small-scale synthetic faults with small reverse throws on their upper tips can be observed on the hanging wall of larger scale extensional faults off the western flank of the Penguins Horst (see Fig. 7). The reverse nature of these faults can be explained by the rotation of the fault block during the late Jurassic extensional episode and the subsequent steepening of the original normal fault plane. Well and seismic data indicate the base of the Humber Group is unconformable in nature, truncating the Middle Jurassic Brent Group and older Formations. The easterly tilt of the Triassic reflectors, together with the difference in throw observed at top basement level, points to the West Penguin Fault predating in age the East Penguin Fault. The East Penguin Fault probably developed at a later stage to accommodate further extension in the basin. The West Penguin Fault is a major structural feature that can be traced southwards into the Tern-Eider Ridge.
Penguins Lineament The southern end of the Penguin Horst can be illustrated with a west-line through well 211/13a-9s1, at the southern end of the Penguin A field. At this level the Penguin Horst experiences a dramatic reduction in structural height as it encounters a sharp change in structural grain from north–south to NW–SE, as observed mainly in time slices (see Fig. 6) and fault maps (see Fig. 3). At the southern end of the
34
R. DOMI´NGUEZ
Penguins Horst the structural fabric swings to the SE and becomes parallel to a NW –SE fault trend that can be traced between the south of the Penguin A field to the NW and the boundary between the Penguin C and D fields to the SE. This structural grain band is interpreted as the shallower expression of an underlying lineament, possibly a basement shear zone or fault system of Caledonian age reactivated in the subsequent Triassic and Jurassic rifting episodes. This basement lineament is observed in time slices as a c. 2 km wide zone of deformation where the Mesozoic stretching resulted in either reactivated older fault planes propagating upwards into the overburden, or as newly-formed faults with a NW –SE trend and therefore an oblique angle to the main west – east extension direction, contrasting with the main north–south structural grain at either side of the lineament. The basement lineament bounds the Penguin A field to the south and the Penguin C field to the SW and south. Several large-scale Jurassic faults in the Penguin A field have a NW–SE trend that intersect a north–south fault population, creating a number of intra-field fault block compartments. At least one of these NW–SE faults is known to be sealing, as demonstrated from two different oil compositions sampled in wells at either side. The southeastern continuation of the basement lineament becomes harder to interpret. Here the lineament seems to run along the boundary between the Penguin C and D fields, although fault maps show this boundary to have a different orientation, created by a NE –SW-trending fault dipping to the south.
Penguins Ridge The Penguins Ridge represents the southern expression of the Penguin Horst across the Penguins Lineament. The Penguin C, D and E fields can be found along this trend. The Penguins Ridge has the same north–south trend as the Penguins Horst, albeit with a much smaller structural relief (1000 feet v. 3500 feet). Its position is some 3 km further SE than the main horst trend across the Penguins Lineament (Fig. 6). This jog in the main structural grain can be explained by the presence of the underlying Caledonian structural fabric that has offset the main locus of faulting either side of it. South of the Penguins Lineament the stretching associated with the Mesozoic rifting was distributed across a number of north–South-trending extensional faults, rather than focusing on a single fault, causing the ridge to be a less pronounced structural feature than the Penguin Horst to the north. It is bounded to the west by a major westerlydipping, north–south-trending fault. Internal
seismic reflectors within this structural high vary from gently easterly-dipping to horizontal (Fig. 8). As in the case of the Penguin Horst, younger faults bound the Penguin Ridge to the east, particularly along the south of the Penguin D field and along the eastern margin of the Penguin E field.
Penguins Basin The Penguins Basin is located west of the Penguins ridge. North of the Penguins Lineament it forms a gentle syncline west of the Penguin A field (see Fig. 7). South of the lineament the basin widens and is composed of a number of easterly-dipping Triassic and Jurassic fault blocks that display a characteristic domino-style and show some degree of thickening of the Upper Jurassic Humber Group against the fault planes. This thickening is a combination of ‘wedging’ in the hanging wall of the faults due to the syn-rift nature of the Humber Group (for further evidence see Figs 7 and 9) with some degree of erosion in the uplifted footwalls. Figure 8 shows a SW –NE trending seismic line and seismic interpretation some 19 km in length which ties the Penguins Basin and the south of the Penguin C field, illustrating the structural styles observed in the Penguins Basin, characterized by a number of westerly-dipping extensional faults with associated minor antithetic faults developed in the hanging walls. The top of the basement occurs at c. 3.75 s TWT over this area, although it is not possible to trace it with confidence due to the poor seismic resolution at this level. Figure 9 shows a staggered seismic line (B– B0 ) some 13 km in length that goes from the Penguin Basin into the southern part of the Penguin D field (west to east), before heading north into the Penguin C field. This line is tied to four exploration wells: 211/13-7, 211/14-1s2, 211/14-3s1(Z) and 211/14-4RE. This geological cross-section has been used as the basis for a palinspastic restoration used to illustrate the structural evolution of Penguins. In the geological cross-section it can be seen that the Penguin D west-bounding fault has accumulated large amounts of throw, estimated in c. 1000 feet from well data. The poor seismic resolution at deeper levels does not allow estimation of the effects of the faulting on the Triassic and deeper levels with a great degree of certainty, although the mapping of an intra-Cormorant Formation seismic reflector and the nature of the intraTriassic seismic reflectors point to a thickening against the fault planes. This geo-seismic section illustrates the syn-rift nature of the Humber Group, eight times thicker in the hanging wall of the main Penguin D bounding fault. The Heather and Kimmeridge Clay formations can be correlated at either side of the fault, and their condensed
PENGUINS CLUSTER STRUCTURAL EVOLUTION
35
Fig. 8. (a) uninterpreted, and (b) interpreted west– east seismic line through the Penguin Basin and the Penguin C South Field, highlighting the main structural elements of the Penguin Basin, and in particular the ‘domino’ nature of the extensional faults.
thickness nature in the footwall indicates the thickness change was mostly driven by the syndepositional movement of the fault, with some smaller degree of footwall erosion in the early Cretaceous. It is also evident from both seismic and well data
that the Dunlin and Brent Groups thicken slightly across the fault into the main Penguins Basin. Most of this thickness difference can be attributed to more accommodation space available in the Penguins Basin during the early to middle Jurassic
36
R. DOMI´NGUEZ
Fig. 9. (a) uninterpreted, and (b) interpreted seismic line between the Penguin Basin, Penguin D and C fields (Penguins Ridge) and tied to wells 211/13-7, 211/14-1s2, 211/14-3s1(z) and 211/14-4RE (see Fig. 3 for location). This interpreted cross-section has been used as the basis for the palinspastic restoration shown on Figure 17.
thermal subsidence that followed the Triassic rifting episode.
Structural evolution Lower Jurassic The early Jurassic was a time of little faulting activity across the northern North Sea, when the
area underwent an episode of passive thermal subsidence following the Triassic rifting during which the Dunlin deep marine shales where deposited. An isochore map of the Dunlin Group over the Penguins and Don Northeast areas (Fig. 10), shows variations in thickness which reflect the underlying main structural elements, with the Dunlin package thinning to less than 300 feet over the structural highs and thickening in the structural
PENGUINS CLUSTER STRUCTURAL EVOLUTION
37
Fig. 10. Thickness in feet of the Dunlin Group in the Penguins Area. This isochore map has been constructed from the thickness in the wells shown. Contour interval every 50 feet.
lows of the Penguins Basin to over twice that thickness. This map was constructed using data from wells that penetrate a full Dunlin stratigraphic sequence (Fig. 11), with all four internal formations. Therefore the changes of thickness cannot be attributed to post-depositional erosion, but are instead interpreted as the result of changes in accommodation space across faults and the effects of remnant late Triassic– early Jurassic palaeotopography, inherited from the Triassic rifting episode. Seismic lines across some major faults such as those shown in Figures 8 and 9 display a thicker Dunlin package in the hanging wall. The overall geometry does not show typical syn-rift wedge geometries against fault planes, suggesting instead a passive, post-kinematic deposition of the Dunlin Group. Fault activity was nonexistent to minor during this period of time, with fault movements perhaps having accommodated
some of the thermal subsidence that followed the Permo-Triassic rifting.
Middle Jurassic Evidence from Penguin D indicates that the initiatial movements of the Jurassic rifting took place as early as the middle Jurassic, coinciding with the Brent Group sandstones deposition, and eventually climaxing during the late Jurassic. The middle Jurassic fault movements had an impact on the deposition and distribution of the internal Brent Group formations, creating at least one intra-Brent unconformity (top Rannoch) in the Penguins area. Figure 12 shows a south– north seismic line and seismic interpretation along the Penguin D field tied to exploration wells 211/14-1s2 and 211/14-3. This line focuses on the Lower to Middle Jurassic
38
R. DOMI´NGUEZ
Fig. 11. Well correlation of the Dunlin Group formations across the Penguins Basin, Penguin D and C fields. See Figure 10 for location of correlation line. Depths are in feet.
intervals (Dunlin and Brent Groups), and illustrates the impact of small-scale intra-field faulting on the Brent reservoir architecture. The base Cretaceous unconformity is a high amplitude reflector that occurs at a depth of around 3 seconds TWT (black loop in the figure). The top Brent has been picked on the red loop immediately underneath. The Dunlin Group is characterized by a series of subparallel seismic reflectors affected by extensional faults. These normal faults have created a depocentre between 211/14-1s2 and 211/14-3 where the Brent Group achieves its maximum thickness. The Brent Group thins progressively southwards, where it has been uplifted due to footwall rotation. A comparison of throw at the base Cretaceous unconformity versus top Dunlin level reveals that these faults were more active during the Middle Jurassic, dying out progressively during the Late Jurassic, where the climax of the extension switched to the large-scale, fieldbounding faults. Figure 13 is a correlation panel of the Brent Group taken from south to north along the Penguin D field wells. It corresponds roughly to the seismic line shown in Figure 12 and it illustrates the Brent reservoir architecture and the impact that the Jurassic faulting had on it. The correlation panel
has been flattened at the top Brent level (effectively top Etive Formation). The Brent Group shows important thickness changes, varying from as little as 150ft true-vertical thickness (TVT) in southern well 211/14-1s2 to 250 ft TVT in northern well 211/14-3. Wells PENG-D01 and PENG-D02 are sub-horizontal development wells which only penetrate the upper section of the Brent Group. The correlation panel shows that the Etive Formation has a relatively constant thickness of around 50 ft TVT, and that the changes in Brent Group thickness are taken up by variations in the Rannoch Formation thickness and the number of internal parasequences this formation can be split up into. Thus, the well with the thinnest Brent section (211/14-1s2), drilled only through the lowermost three Rannoch parasequences, whereas the thicker Brent section of well 211/14-3 consists of a much thicker Rannoch sequence, split up into eight internal parasequences. These thickness and internal reservoir architecture changes are related to the structural position of the wells within the field, with the thinnest Rannoch sequences occuring in the structural highs created by rotated footwalls. This is interpreted as a pulse of fault activity that took place after the Rannoch deposition, creating the top-Rannoch unconformity and leading to fault
PENGUINS CLUSTER STRUCTURAL EVOLUTION
39
Fig. 12. (a) uninterpreted, and (b) interpreted south– north seismic line along the Penguin D Field. Note the thickening of the Rannoch layers in the downthrown fault blocks. See text for discussion.
block rotation and erosion of the uplifted footwalls before the deposition of a tabular-shaped Etive Formation across the field.
Late Jurassic Although the main extension direction of the Late Jurassic rifting was roughly east –west (Roberts
et al. 1990), alternative orientations, mainly NW– SE and NE –SW, have been proposed by some authors (e.g. Thomas & Coward 1995). The Late Jurassic extension vector must have been directed at an oblique angle to some fault trends, causing the reactivation of these pre-existing fault planes to take place under oblique-slip conditions. Where these faults experienced jogs or bends along their
40
R. DOMI´NGUEZ
Fig. 13. (a) well correlation of Brent Group formations across the Penguins D Field, including the Rannoch Fm shoreface parasequences 1 to 8, flattened at top Brent (top Etive) level. Depths are in feet. (b) Schematic structural interpretation of the log data. See text for discussion.
trend, transpression and transtension took place, as demonstrated by the small-scale flower structures drilled by wells PENG-C01 and PENG-C02. A steep normal fault can be mapped along the main trend of the Penguin C field, running roughly parallel to the top Brent contours, with a NNW–SSE strike orientation (Fig. 14). Detailed mapping of this fault shows two small-scale jogs that occur along its trajectory. A left-stepping jog can be mapped in the central part of the Penguin C field, having created a releasing bend drilled by the sub-horizontal PENG-C01 well, which penetrated a downthrown fault block of Heather shales. The second jog, in the southernmost end of the C field, is a restraining bend that has created a transpressional fault-block of Dunlin shales penetrated by the PENG-C02 well.
Figure 15 shows a NW– SE seismic line and its geological interpretation along the PENG-C01 well trajectory, drilled in the central part of the Penguin C field (see Fig. 14). Whilst drilling a subhorizontal section of Brent Group sandstones, PENG-C01 encountered a c. 300 feet-long section of Upper Jurassic Heather shales, confirmed with biostratigraphic samples, before penetrating again the Brent Group sandstones for the last 1000 feet of the well. The shale section was bounded by faults as confirmed by image logs, and is interpreted as a transtensional ‘pop-down’, or negative flower structure. The small scale of this feature makes it difficult to interpret on seismic, and only one of the faults bounding it can be mapped with confidence. This seismic line also illustrates the unconformable nature of the base of the Brent Group,
PENGUINS CLUSTER STRUCTURAL EVOLUTION
41
Fig. 14. Penguin C Field top Brent depth structure map. Contour interval is 125 feet. The fault highlighted in red shows two jogs of opposing orientations. This fault was reactivated in the late Jurassic under an oblique-slip extensional regime, causing the jogs to generate a ‘pop-down’ drilled by PENG-C01 and a ‘pop-up’ drilled by PENG-C02. Refer to text for details.
42
R. DOMI´NGUEZ
Fig. 15. (a) uninterpreted, and (b) interpreted seismic line along the PENG-C01 well trajectory. This well drilled through a downthrown fault block composed of Heather shales along the wellpath. See text for discussion. A, motion away from page; T, motion towards page.
seen here truncating a series of underlying Dunlin Group seismic reflectors. Figure 16 shows a SW –NE seismic line and seismic interpretation along the PENG-C02 well trajectory, drilled in the southern part of the Penguin C field and illustrating also the impact of oblique-slip fault movements in the structural evolution of the Penguin C field. PENG-C02 drilled through another intra-Brent shale section, this time dated with biostratigraphic samples as Lower Jurassic Dunlin Group, and also bounded by faults. This feature has been interpreted as a
transpressional ‘pop-up’, or positive flower structure. One of the block-bounding faults can be mapped on seismic with relative confidence, and it can be followed northwards where it ties with the fault drilled through by the PENG-C01 well where the Heather ‘pop-down’ occurred.
Discussion Figure 17 summarizes the Mesozoic structural evolution of the Penguins Cluster using the palinspastic
PENGUINS CLUSTER STRUCTURAL EVOLUTION
43
Fig. 16. (a) uninterpreted, and (b) interpreted seismic line along the PENG-C02 well trajectory. This well drilled through an upthrown fault block composed of Dunlin shales along the wellpath. See text for discussion. A, motion away from page; T, motion towards page.
restoration of seismic shown in Figure 9 as its basis. Figure 18 summarizes the timing of the main tectonic events in relation to the lithostratigraphy and the unconformities discussed in the text. The present day structural configuration of the Penguins area is the result of two rifting episodes during the Mesozoic (Permo-Triassic and Jurassic) together with their associated phases of thermal relaxation and subsidence. The rifting initiated in
the Permo-Triassic continued in the Middle Jurassic as intermittent pulses of fault activity contemporaneous with the Brent Group deposition, eventually climaxing in the late Jurassic. The faulting was influenced south of Penguin A and the Penguin Horst by a pre-existing NW– SE structural grain inherited from the Caledonian orogeny. The interaction between two subsequent rifting episodes resulted in some cases in the reactivation of
44
R. DOMI´NGUEZ
Fig. 17. Palinspastic restoration applied to the geological cross-section shown on Figure 9. See text for details.
PENGUINS CLUSTER STRUCTURAL EVOLUTION
45
Fig. 18. Generalized stratigraphy of the Penguins Area showing unconformities in red and the tectonic events discussed in the text.
pre-existing faults, leading to oblique-slip transpressional and transtensional fault block movements in the Penguin C field. The structural evolution of the study area during the Triassic is less well understood than the Jurassic evolution due to the poorer resolution of the seismic data at the Triassic depths and the superimposition of the Jurassic deformation on the original Triassic fabric. Triassic fault activity is evident from large-scale faults such as the Penguin West fault, its southern continuation into the Tern-Eider ridge
area to the SW, and the west-bounding fault of the Penguin D and E fields. These are all west-dipping extensional faults against which thickening of the lower part of the Hegre Group, of Triassic age, can be observed in seismic. The Triassic rifting created a structural picture consisting of large-scale easterly-dipping fault blocks against which thickening of the Cormorant Formation occurs. Deposition of the Upper Cormorant and the lowermost Jurassic Statfjord formations took place mostly under thermal subsidence conditions following the
46
R. DOMI´NGUEZ
Triassic rifting episode. Some degree of uplifted fault-block palaeotopography was inherited from the rifting and still existed in the Lower Jurassic (Fig. 17a), either due to the inability of the Hegre Group sedimentation to keep up with the created fault block subsidence, or perhaps due to some degree of fault movement taking up some of the thermal subsidence once the main rifting episode had finished. This is demonstrated by the thickness changes across faults observed in the Dunlin Group claystones (Fig. 17b), as seen on seismic. Well data shows the Dunlin Group to thicken from c. 250 ft TVT along the structural high formed by the Penguins Ridge, to at least 500 ft TVT in the hanging wall of this major fault, into the Penguins Basin. Although the base of the overlying Brent Group can be erosional in some areas, the Dunlin Group thickness changes are depositional and cannot be attributed to erosion, as the correlation of the internal Dunlin formations demonstrates. The deposition of the deep marine Dunlin Group claystones was followed by deposition of the shallower marine and deltaic Brent Group sandstones. Although conformable across much of the area, in some locations the base of the Brent can be seen as being erosional in nature, cutting down and eroding off the uppermost Dunlin Group claystones. Deposition of the lower part of the Brent, the Broom and Rannoch formations, took place mostly under quiescent tectonic conditions when the basin was undergoing thermal subsidence after the Triassic rifting (Fig. 17c). Fault activity was re-established immediately after deposition of the Rannoch Formation, as evidenced by seismic and well data from the Penguin D field. Many of these were newly formed faults in the middle Jurassic as their throw profiles demonstrate. Fault block rotation accompanied the fault movement, causing uplift of the footwalls that were then subjected to subaerial conditions, erosion of the uppermost Rannoch, and peneplanation. This faulting episode resulted in the top Rannoch unconformity. Immediately after, or perhaps coeval with the upper Rannoch erosion, was the deposition of the shallow marine to fluvial Etive Formation across the area (Fig. 17d). Its absence over areas of the Penguin C field is attributed to post-depositional fault movements and erosion. The Jurassic rifting, although initiated during the middle Jurassic, experienced its climax and most of the fault activity during the late Jurassic, as demonstrated by the syn-rift thickening of the late Jurassic-dated Humber Group claystones against major north–south-trending extensional faults (Fig. 17e). Many of these Jurassic faults are reactivated older features, probably Triassic in origin, as demonstrated by the difference in throw observed at different levels. The West Penguin Fault is a clear example, where the interpretation of the top of the
basement shows a much larger throw at top Triassic than at top Jurassic level. The Upper Jurassic Magnus Sandstone Member of the Kimmeridge Clay Formation, the oil reservoir in the Penguin A field, is syn-rift in origin. The provenance of this deep-marine turbiditic fan has been interpreted as being from either the north or the west. This sedimentary package thickens against north–southtrending faults over the Penguin A field (see Fig. 7), demonstrating its syn-rift origin, and it pinches out towards the West Penguin Fault, indicating these faults were moving as the turbiditic sediments tried to ‘climb’ towards the structural highs. The upthrown Penguins Ridge must have acted as a barrier to submarine sediment transport towards the west, where the Magnus Member is absent from all wells in the Penguin C, D and E fields. These submarine fans must have been deflected towards the south by the emerging topography during the late Jurassic, into the Penguin Basin, where it has been encountered as far south into the Penguins Basin as well 211/13-5. The late Jurassic fault activity caused oblique-slip reactivation of some Penguin C faults. Where bends or jogs were encountered along the strike of the fault, as mapped in some of the Penguin C field faults, accommodation problems resulted in transpressional or transtensional fault movements, depending on the relative orientation between the fault strike and the maximum stress vector. These movements created the positive flower structure drilled by the PENG-C02 well, and its strike-related negative flower structure in the central part of the field drilled by PENG-C01. Some authors (Booth et al. 1992; Thomas & Coward 1995) argue that the East Shetland Basin was affected by an episode of basin inversion during the latest Jurassic – early Cretaceous, originating a number of positive flower structures along NE–SW-trending faults. These faults suggest a component of strike-slip during the reactivation of pre-existing structures under a compressional tectonic regime. Booth et al. (1992) interpret the Penguins Horst as an inversion feature, a large-scale ‘pop-up’ developed due to transpressional or compressional movements as a result of strike-slip reactivation of faults during the late Jurassic to earliest Cretaceous (see fig. 12 in Booth et al. 1992). Thomas & Coward (1995), on the other hand, interpret the Penguins Ridge as being delineated with both normal and reverse faults, with positive flower structures, or ‘pop-ups’, developed at either side of the Penguins Horst (see fig. 14 in Thomas & Coward 1995). They argue that inversion may be a consequence of either compression of the basin as a whole, or alternatively, due to transpressional uplift along offset NW–SE systems. These past interpretations are based on 2D seismic
PENGUINS CLUSTER STRUCTURAL EVOLUTION
lines of limited quality. A re-interpretation of more recent, better quality 3D seismic data along the same lines as those used by Thomas & Coward (2005), and Booth et al. (1992) shows the Penguins Horst to be an extensional feature bounded by normal faults on either side which have accumulated throws of at least 2000 feet true-vertical depth at the BCU level. No reverse throws are indicated in the new seismic dataset at either side of the horst, and the previous interpretation of ‘pop-ups’ is here revisited as a series of extensional faults which, in some cases, can be traced down to the top of the basement. The detailed geometry of these faults at depth is uncertain due to deterioration in the seismic quality, and they could also be interpreted as synthetic splays from the main West Penguin Fault, rather than domino-faults. Detailed interpretation of a better quality 3D seismic volume over the Penguins Area shows that the main structural elements that define the Penguins Cluster are extensional in origin, although oblique-slip reactivation of existing normal faults during the late Jurassic has created at least one case of a transpressional flower structure, or ‘pop-up’, in the Penguin C field. Fault activity during the Permo-Triassic and Jurassic rifting episodes reactivated an underlying lineament with a NW –SE trend that runs between the south of the Penguin A field and the boundary between the Penguin C and D fields. This is probably a basement shear zone of Caledonian age that caused the Mesozoic faults overlying it to curve and become subparallel to this NW–SE trend. This basement lineament seems to have accommodated extension in different ways north and south of it. To the north, most of the stretching seems to have focused on one or two major normal faults, originating the Penguins Horst. South of the lineament, the stretching has been distributed among several north–south-trending normal faults. The Penguin Lineament is interpreted as the northwestern expression of the Northern Transfer Zone of the East Shetland Basin described by Lee & Hwang (1993). This is described as a NW–SE trending regional-scale Tornquist lineament that transects the northern area of the East Shetland Basin, and across which basin polarity changes occur. The Jurassic rifting was followed by an episode of thermal relaxation and subsidence of the basin between the Cretaceous and the Cenozoic, with the accummulation of a post-rift megasequence up to c. 3.5 km over Penguins that preserved the late Jurassic palaeotopography from significant erosion. I would like to thank Shell UK Ltd and its partner in the Penguins Cluster, ExxonMobil, for permission to publish this paper. I would also like to thank all of my fellow Penguins team members throughout the years, and in particular K. Fletcher, R. Shelton, D. Bateman and P. Watt.
47
References B ADLEY , M. E., E GEBERG , T. & N IPEN , O. 1984. Development of rift basins, illustrated by the structural evolution of the Oseberg feature, Block 30/6, offshore Norway. Journal of the Geological Society, London, 141, 639–649. B ADLEY , M. E., P RICE , J. D., R AMBECH D AHL , C. & A GDESTEIN , T. 1988. The structural evolution of the northern Viking Graben and its bearing upon extensional modes of basin formation. Journal of the Geological Society, London, 145, 455–472. B EACH , A. 1987. A regional model for linked tectonics in north-west Europe. In: B ROOKS , J. & G LENNIE , K. (eds) Petroleum Geology of North West Europe. Graham & Trotman, London, 43– 48. B OOTH , A., S TOCKLEY , F. J. & R OBBINS , J. A. 1992. Late Jurassic Structural Inversion in the North Viking Graben and East Shetland Basin, UK North Sea. Oryx Energy Company Internal Publication. C OWARD , M. P., D EWEY , J. F., H EMPTON , M. & H OLROYD , J. 2003. Tectonic Evolution. In: E VANS , D., G RAHAM , C., A RMOUR , A. & B ATHURST , P. (eds) The Millenium Atlas: Petroleum Geology of the Central and Northern North Sea. Geological Society, London, 17– 33. D E ’A TH , N. G. & S CHUYLEMAN , S. F. 1981. The Geology of the Magnus Oilfied. In: I LLING , L. V. & H OBSON , G. D. (eds) Petroleum Geology of the Continental Shelf of Northwest Europe. Institute of Petroleum, Heyden, London, 342– 351. D EEGAN , C. E. & S CULL , B. J. 1977. A standard lithostratigraphic nomenclature for the Central and Northern North Sea. Institute of Geological Sciences Report 77/25. F ÆRSETH , R. B. 1996. Interaction of Permo-Triassic, and Jurassic extensional fault-blocks during the development of the northern North Sea. Journal of the Geological Society of London, 153, 931– 944. G ABRIELSEN , R. H., O DINSEN , T. & G RUNNALEITTE , I. 1999. Structuring of the North Viking Graben and the Møre Basin; the influence of basement structural grain, and the particular role of the Møre-Trøndelag Fault Complex. Marine and Petroleum Geology, 16, 443– 465. J ONES , G., R ORISON , P., F ROST , R., K NIPE , R. & C OLLERAN , J. 1999. Tectono-stratigraphic development of the southern part of the UKCS Quadrant 15 (eastern Which Ground Graben): implications for the Mesozoic– Tertiary evolution of the Central North Sea Basin. In: F LEET , A. J. & B OLDY , S. A. R. (eds) Petroleum Geology of Northwest Europe: Proceedings of the 5th Conference. Geological Society, London, 133–151. L EE , M. J. & H WANG , Y. J. 1993. Tectonic Evolution and Structural Styles of the East Shetland Basin. In: P ARKER , J. R. (ed.) Petroleum Geology of Northwest Europe: Proceedings of the 4th Conference. Geological Society, London, 1137–1149. O DINSEN , T., R EEMST , P., VAN DER B EEK , P., F ALEIDE , J. I. & G ABRIELSEN , R. H. 2000. Permo-Triassic and Jurassic extension in the northern North Sea: results from tectonostratigraphic forward modeling. In: N ØTTVEDT , A. (ed.) Dynamics of the Norwegian
48
R. DOMI´NGUEZ
Margin. Geological Society, London, Special Publications, 167, 83– 103. P ARTINGTON , M. A., M ITCHENER , B. C., M ILTON , N. J. & F RASER , A. J. 1993. Genetic sequence stratigraphy for the North Sea Late Jurassic and early Cretaceous: distribution and prediction of Kimmeridgian-Late Ryazanian reservoirs in the North Sea and adjacent areas. In: P ARKER , J. R. (ed.) Petroleum Geology of Northwest Europe: Proceedings of the 4th Conference. Geological Society, London, 441– 470. R ATTEY , R. P. & H AYWARD , A. P. 1993. Sequence stratigraphy of a failed rift system: the Middle Jurassic to Early Cretaceous basin evolution of the Central and Northern North Sea. In: P ARKER , J. R. (ed.) Petroleum Geology of Northwest Europe: Proceedings of the 4th Conference. Geological Society, London, 215– 249. R AVNA˚ OS , R., N OTTVEDT , A., S TEEL , R. J. & W INDELSTAD , J. 2000. Syn-rift sedimentary architectures in the Northern North Sea. In: N OTTVEDT , A. (ed.) Dynamics of the Norwegian Margin. Geological Society, London, Special Publications, 167, 133–177. R OBERTS , A. M., Y IELDING , G. & B ADLEY , M. E. 1990. A kinematic model for the orthogonal opening of the late Jurassic North Sea rift system, Denmark-Mid Norway. In: B LUNDELL , D. J. & G IBBS , A. D. (eds) Tectonic Evolution of the North Sea Rifts. Oxford Science Publications, Oxford, 181–199. R OBERTS , A. M., Y IELDING , G., K USZNIR , N. J., W ALKER , I. & D ORN -L O´ PEZ , D. 1993. Mesozoic extension in the North Sea: constraints from flexural backstripping, forward modelling and fault populations. In: P ARKER , J. R. (ed.) Petroleum Geology of Northwest Europe: Proceedings of the 4th Conference. Geological Society, London, 1123– 1136. R OBERTS , D. G., T HOMPSON , M., M ITCHENER , B., H OSSACLE , J., C ARMICHAEL , S. & B JORNSETH , H.-M. 1999. Palaeozoic to Tertiary rift and basin dynamics: mid-Norway to the Bay of Biscay – a new concept for hydrocarbon prospectivity in the deep water frontier. In: F LEET , A. J. & B OLDY , S. A. R. (eds) Petroleum Geology of Northwest Europe: Proceedings of the 5th Conference. Geological Society, London, 7–40.
R ICHARDS , P. C., L OTT , G. K., J OHNSON , H., K NOX , R. W. O’B. & R IDING , J. B. 1993. Jurassic of the Central and Northern North Sea. In: K NOX , R. W. O’B. & C ORDEY , W. G. (eds) Lithostratigraphic Nomenclature of the UK North Sea. British Geological Society, Birmingham. S HEPHERD , M., K EARNEY , C. J. & M ILNE , J. H. 1991. Magnus Field. In: B EAUMONT , E. A. & F OSTER , N. H. (eds) Structural Traps II. Traps associated with tectonic faulting. American Association of Petroleum Geologists, Treatise of Petroleum Geology Atlas of Oil and Gas Fields, 95–125. T AYLOR , S. R., A LMOND , J., A RNOTT , S., K EMSHELL , D. & T AYLOR , D. 2003. The Brent Field, Block 211/29, UK North Sea. In: G LUYAS , J. G. & H ICHENS , H. M. (eds) United Kingdom Oil and Gas Fields, Commemorative Millennium Volume. Geological Society, London, Memoirs, 20, 233– 250. T HOMAS , D. W. & C OWARD , M. P. 1995. Late Jurassic-Early Cretaceous inversion of the northern East Shetland Basin, Northern North Sea. In: B UCHANAN , J. G. & B UCHANAN , P. G. (eds) Basin Inversion. Geological Society, London, Special Publications, 88, 275–306. Y IELDING , G. 1990. Footwall uplift associated with late Jurassic normal faulting in the northern North Sea. Journal of the Geological Society, London, 147, 219–222. Y IELDING , G., B ADLEY , M. E. & R OBERTS , A. M. 1992. The structural evolution of the Brent province. In: M ORTON , A. C., H ASZELDINE , R. S., G ILES , M. R. & B ROWN , S. (eds) Geology of the Brent Group. Geological Society, London, Special Publications, 28, 477–493. Z ANELLA & C OWARD , 2003. Structural Framework. In: E VANS , D., G RAHAM , C., A RMOUR , A. & B ATHURST , P. (eds) The Millenium Atlas: Petroleum Geology of the Central and Northern North Sea. Geological Society, London, 45–59. Z IEGLER , P. A. 1988. Evolution of the Artic-North Atlantic and the Western Tethys. American Association of Petroleum Geologists, Memoirs, 43.
Characterizing and producing from reservoirs in landslides: challenges and opportunities A. I. F. WELBON1,2, P. J. BROCKBANK1, D. BRUNSDEN3 & T. S. OLSEN1 1
StatoilHydro, Forusbeen 35, Forushagen, N-4035 Stavanger, Norway
2
BG norge, Løkkeveien 103, Stavanger, Norway (e-mail:
[email protected]) 3
Vine Cottage, Chideock, Dorset, England, UK
Abstract: Landslides can consist of rotational slips, translational glide blocks, topples, talus slopes, debris flows, mudslides and compressional toes which can combine in different proportions to form complex landslides. The mass movement can be subaerial or submarine, occur over wide ranges of scale and can vary in rate from creep to catastrophic failure. Complexity of the landslide reflects the controlling factors including the strength of the deforming material and triggering mechanisms such as earthquakes, imposed load, increasing topographic relief and removal of toe material. Processes of landslide deformation include slip on discrete surfaces, distributed shear within the landslide, vertical thinning and lateral spreading through shear, fluidization, porosity collapse and loss of material from the top or toe of the complex. These processes control the quality of the resultant reservoirs. This leads to a greater range of reservoir types than conventional faulted reservoirs, with a proportionate upside and downside potential and difficulty in quantifying uncertainty. This paper uses examples from the literature, outcrops and subsurface datasets (including the Statfjord Field and the Halten Terrace in Norway) to outline the complexity of reservoirs in landslides and the challenges and opportunities in finding and producing them. We present workflows for seismic and subseismic characterization for exploration and reservoir scale based on geomorphological principles. Seismic mapping is achieved by classifying the form of the reflectors (both slip surfaces and the bounding envelope of the landslide) from an atlas of geometric and structural styles and is applied to both the Halten Terrace example and the Statfjord Field. We present a new workflow for reservoir characterization in which integration of structural, biostratigraphic, sedimentological and dynamic data gives key information on process, timing and heterogeneity of the reservoir. For the Statfjord field, important maps of the landslide block stratigraphy derived from a subcrop map and communication maps based on a c. 130 well dataset can be correlated to outcrop analogues and used to develop a predictive tool for landslide reservoir extent and quality, both in this field and others.
Landslides have been documented and studied for centuries (Buckland 1840), generally following onshore events that had catastrophic consequences for local communities (e.g. Frank Slide, Canada 1903). Oil and gas companies have been producing from palaeolandslide complexes (commonly referred to in the industry as ‘slumps’ and ‘degradation complexes’) since shortly after the start of the hydrocarbon industry, but it has only been in the last two decades that seismic imaging technology, together with new integrated techniques, have led to detailed imaging and correct identification of many of these reservoirs. In many hydrocarbon provinces, the presence and magnitude of landslides has been underestimated until late in the development of the fields. In the Tampen Spur area of the North Sea, the number of identified landslide complexes on fields in production has risen from a handful in the late 1990s (Stewart 1997; Underhill et al. 1997; Berger &
Roberts 1999; Hesthammer & Fossen 1999; McLeod & Underhill 1999) to over 20 today. The downslope products of these landslides collect in the hanging wall of fault systems and are now key exploration plays around existing infrastructure. With the advent of high quality 3D seismic data, on many deep-water passive margins landslide products are commonly recognized as the main form of deformation, including the formation of major extensional and compressional landslide systems (Prior & Coleman 1982; Heinio & Davies 2006). Landslides are common on salt diapirs and associated folded sediments. Due to their topographic expression, carbonate reservoirs such as pinnacle reefs are subject to landslide processes on their flanks, and result in modifying the expected recoverable volumes and influencing production strategies to avoid early water breakthrough. Chalk mass movement deposits, synchronous with sedimentation, affect chalk fields in the North Sea.
From: JOLLEY , S. J., BARR , D., WALSH , J. J. & KNIPE , R. J. (eds) Structurally Complex Reservoirs. Geological Society, London, Special Publications, 292, 49– 74. DOI: 10.1144/SP292.3 0305-8719/07/$15.00 # The Geological Society of London 2007.
50
A. I. F. WELBON ET AL.
Development of these reservoirs was commonly postponed whilst other, more conventional, sources were exploited; these sources obeyed established concepts of sequence stratigraphy and basin development and were readily imaged in standard seismic data. With the emphasis of the industry now on the hard to access, non-conventional reservoirs as we enter the second half of the oil age (Campbell & Laherrere 1998; Hall et al. 2003), landslides are a key target both in exploration and production. In this paper we present a review of landslide types and methods for their characterization involving geomorphological analysis of seismic and well data, use of empirical datasets from subsurface landslides and outcrops as analogues, integrated studies which include sedimentology, biostratigraphy, structural geology and the use of production data. We then describe the challenges and opportunities these types of reservoirs present, using examples from various fields.
Landslide type, processes and challenges in identification Landslides have been described (Cruden 1991) as a movement of a mass of rock, earth or debris down a slope. Brunsden (1984) pointed out that these mass movement features do not necessarily have a transport medium (for example rock falls). Thus, landslides can be either subaerial or submarine; indeed on many coastlines today they originate onshore and extend offshore. There are several definitions and classification schemes for landslides (Hutchinson 1968; Varnes 1978; Dikau et al. 1996) which attempt to capture process; this paper follows the classification of Dikau et al. (1996). Landslide processes are varied and often complex. Figure 1 illustrates a series of landslide processes and the resulting morphology of the products. The nature of the process or combination of processes that cause a landslide will control the quality of the resultant hydrocarbon reservoirs. The degree of recovery can be dependent on preservation or enhancement of the original, pre-landslide reservoir character. Thus, landslides (or mass movements) can be classified into eight principle types (Fig. 1). These are slides (rotational and translational) (Fig. 1a, b), topples and falls (Fig. 1c), mudslides and flows (Fig. 1d), lateral spreads (Fig. 1e), and complex landslides (Fig. 1f), extrusions associated with cambering (Fig. 1g) and compressional toes (Fig. 1h) (EPOCH 1993). In addition, they can be further classified on the basis of the material involved, such as rock, debris and soil. Associated with all of these types of landslide can be processes such as creep, extrusion,
fluidization and expulsion of fluids or gas (Seed 1968; Varnes 1978; Varnes et al. 1989). Many of these landslide types can be mapped in seismic, but many fall below seismic resolution. The challenge for the hydrocarbon industry is to map features in seismic data, interpret the process and then predict the range of reservoir type and quality at, and below, seismic resolution. Mass movements are generally triggered through mechanisms such as: loading, removing support from a toe area, increasing pore fluid pressure in the material, or changing the chemistry of the substrate. In addition, in a rift, a thrust belt or on a passive margin, specific types of trigger are common: earthquakes, tsunami loading, loading by creation of topography above regional (footwall uplift, inversion) (Berger & Roberts 1999), glaciations (Bryn et al. 2005), movement on unstable siliceous oozes or salt (density contrast), gasification and increase in pore pressure due to maturation of hydrocarbons (Cobbold et al. 2005). The most common landslides types that can be mapped in seismic, particularly in fault footwalls in rifts and on passive margins, are rotational and translational slides. They are common near the onset of the gravitational collapse (often a back scar) and further downslope are replaced by either compressional toes, mudslides, and debris flows, mixed in with locally derived rock falls from exposed scarps and clifflines. Rotational slides are blocks where movement generally occurs on a spoon-shaped slip surface and the block develops a back tilt as a result (Fig. 1a). Translational slides (or glides, Fig. 1b) form where a block moves on a low angle shear surface with little or no resultant rotation. Rotational slides can be connected to translational slides downslope, in so-called compound failures (Dikau et al. 1996), but commonly, the movement of a translational slide block produces a chasm or ‘graben’ up-dip from the block (Pitts & Brunsden 1987). Rock falls and topples (Fig. 1c) are commonly found at cliff edges or at the foot of steep slopes, and are usually locally derived. Talus slopes develop as a result of the falls, or more generally erosion, of cliff-faces and these become rotated by any subsequent movement of the landslide underneath. This means the original critical taper of the talus slope may not be maintained and the top surface may have a lower dip than expected. As a result, seismic interpretations of falls are rare, often being misinterpreted as debris flows. Toppling failures are common in fractured/jointed rocks and occur at cliff faces. If buttressing material (e.g. rockfalls) prevents major rotation the toppled block may maintain a steep dip. Steep structures will be difficult to image in seismic. Where no buttress exists, the low angle slab of toppled material
CHARACTERIZING LANDSLIDE RESERVOIRS
51
Fig. 1. An atlas of landslide types, partly based on the classification of Dikau et al. 1996. (a) rotational slide; (b) translational slide; (c) topples and rock falls; (d) debris flows and mudslides; (e) lateral spread; (f) complex slides; (g) extrusion and cambering; and (h) compressional toe.
that often undergoes break-up from such a failure will also be hard to image, and again may be incorrectly interpreted as a debris flow or mudslide. Downslope from what are considered ‘intact’ blocks are a series of reworked and mixed rock types (Fig. 1c & d). Dependent on the failure type and process, within a few hundred metres of the onset of a metric to hectometric scale landslide, the original rock type is not recognizable. Breakup occurs as result of the rate of movement and a spreading/shearing process. Reworked sediments are mostly debris and mud flows and slides, lateral spreads and break-up products from toppling failures and rock falls. Mudslides are a mass movement where softened argillaceous to fine sand debris advances on a discrete surface. They are characterized by a source, track and a lobe with lateral shear surfaces and contractional features in the accumulation zone (Brunsden 1984). Debris flows consist
of coarse and fine material in a plastic or fluid medium. Another category of landslide is a lateral spread, (Fig. 1e) which originates from low angle slopes where thixotropic material liquefies and failure results. The original instability can be driven by loading, removal of toe material or chemical changes in, for example, quick clays. Overprinting of landslide processes is common, giving rise to complex slides (Fig. 1f) which can include any combination of landslide types. Examples include translational slides that become rotational as they move from a low angle slip surface to a steep, curved slip surface, and debris flows that become incorporated into rock falls at cliff edges. Additionally, a process that is often overlooked is extrusion and cambering at an emergent, often submarine, cliff face. Loading results in extrusion of ductile rock types causing cambering of the overlying, stiffer rock mass (Fig. 1g).
52
A. I. F. WELBON ET AL.
Down-dip within the landslide where there is a change in slope, or where buttresses occur such as earlier landslide or fault blocks, a compressional toe can form (Fig. 1h). A common feature in subsurface landslide complexes, given the rapid lateral and down-dip changes in rock type and geometry is the difficulty of imaging the complex. The top of a complex is often a high acoustic impedance layer that is underlain by small, steep, anisotropic blocks and discontinuous layers of material that may contain high porosity, fractured or vuggy rocks. As a result limited return of seismic energy from the complex and scattering of seismic waves is common and therefore processing and seismic imaging is challenging. A surface mapped in seismic data at the top or within the complex may represent a composite of two or more different types of landslide, or a complex slide resulting from overprinting processes. For the purpose of 3D geological model building this means that traditional methods need to be adapted, namely that surfaces from the seismic dataset may have to be split into the component landslide parts, where the thickness, net: gross and porosity reflect the rapid change in landslide type and character.
Outcrop analogues The use of outcrop analogues in the interpretation of landslide structures has been common practice for many years. During work on the Statfjord Field in the North Sea, outcrop analogues were used from the mid-1980s onwards. Outcrops allow examination of processes not seen in seismic, and historical events provide constraint on timing and evolution (Brunsden & Jones 1976; Farrell 1984; Pettinga 1987a, b; Brunsden 1996; Dikau et al. 1996). By using the observations from outcrops to aid in seismic interpretation and by linking to empirical databases of reservoir type and production behaviour, it is possible to narrow the uncertainty range of reservoir types in a landslide complex. Classic exposures of landslides occur along the south coast of Britain, with more than eight thousand recorded (Brunsden & Jones 1976; Brunsden 1996). These are developed within Jurassic and Cretaceous rocks similar to those found in the North Sea and Norwegian Sea. The exposures consist of all the landslide types illustrated in Figure 1 and illustrate the extreme variation of net: gross and thickness possible in a landslide complex. Two exposures of landslides in West Dorset are end members for common types of geomorphological features mappable in seismic data, the
Stonebarrow Hill area and the Black Ven landslide. (Figs 2 & 3). The Stonebarrow Hill landslides are contained within a cuspate back scar that is asymmetric and originates from the erosion of the sea cliff. The asymmetry stems from the landslides detaching at the Jurassic/Cretaceous stratigraphic contact and other bedding slip surfaces which dip 2–38SE, promoting oblique movement downslope (Brunsden 1996). Similar control on the form of slip surfaces is observed in seismic data as a result of regional tectonic tilt, or more commonly, fault related uplift. Pre-existing faults also control landslide geometry, with differential development taking place on either side of the inherited structure, which is also a common feature in seismic datasets. The cuspate form of the slip surfaces is replicated at various scales, down to metres (Fig. 3a), and is also found in seismic data. Where measurable, cuspate slip surfaces, even when near the onset of collapse, generally have significantly higher displacement gradients than those on faults related to crustal stretching (Kim & Sanderson 2004). In other areas, more linear landslip patterns occur where translational glide blocks develop and they can exhibit lower displacement gradients than rift-related faults (Fig. 3b). These aspects can be used as criteria for identification of landslides in the subsurface. Further into the Stonebarrow complex, rotational slides, toppled blocks and translational slide blocks are common, reflecting the increasing influence of a low angle basal slip surface (which is penetrated in boreholes), and resulting in uneven topography of the top of the complex. However, the most important observation from a reservoir prediction point of view is the thinning of the rock mass, driven by loss of material from the top of the blocks (toppling, rock falls), vertical thinning and lateral spreading (see cross-section, Fig. 4). Over a 200– 400 m distance the original stratigraphy thins to 10–20% of the original thickness, a feature that is also common in the subsurface. Further downslope the landslides break up into a series of very narrow (1–5 m wide) blocks separated by open fractures and shear surfaces (Fig. 4, photograph B). Beyond this zone of degradation, which is generally 20– 40 m wide, the blocks break up into debris flows, mudslides, and rock falls, and during heavy rainfall results in run off of these sediments occurs. In the subsurface, these rock types are often referred to as ‘reworked’ sediments. A boundary can be mapped between areas where intact blocks are present and only reworked material exists (Figs 3a & 4), which is a crucial delineator for estimates of reservoir uncertainty and risk in the subsurface. In the region of fault block break-up and beyond are amphitheatre-like areas of landslides with spurs
CHARACTERIZING LANDSLIDE RESERVOIRS
53
Fig. 2. South coast of Britain and the location of the Lyme Regis area. Black Ven is 1 km east of Lyme, and Stonebarrow Hill is 1 km east of Charmouth, both within the indicated area.
of degraded blocks between (Fig. 3a). This area is characterized by both movement towards the eroding sea cliff and oblique movement caused by local topography. The implication for reservoir provenance is clear: it will not always be possible to relate areas of lost material up dip to a downdip equivalent. Across the sea cliff at the southern edge of the landslide complex, the material from uphill moves over to the beach below (C on Fig. 4). This would be equivalent to crossing a fault scarp in a rift or passive margin setting. In addition, rock falls and topples from the cliff mix with the up-dip derived material. Mudslides and debris flows develop into compressional toes in response to under draining on the beach and/or changes in topography. In these exposures high frequency changes in the net: gross and particle size reflect the complex and mixed origin of the material (Fig. 4). In the subsurface this can lead to many surprises: thin, poor quality sands that produce large volumes of
hydrocarbons and local thick sands that deplete rapidly since they connect to small volumes. The Black Ven outcrops to the west of Stonebarrow, immediately East of Lyme Regis, are significantly different. They are characterized by a narrow strip of block slides adjacent to the back scar, faster degradation down-dip than on Stonebarrow and a greater proportion of mudslides (Brunsden 1984). Large landslide events and a higher event frequency reflect a greater rate of cliff retreat in this area of the coast (Brunsden 1996). The whole landslide area is characterized by a stepped profile reflecting the presence of aquicludes at stratigraphic boundaries. The crosssectional area is also smaller, below seismic resolution in typical 3D datasets. Seismic interpreters can often miss the presence of these thin complexes, identifying them as stepped unconformities though they may contain reservoir rocks and provide a migration pathway from source rocks on the flanks into a fault block. A key method for
54
A. I. F. WELBON ET AL.
Fig. 3. Map view patterns of (a) the Stonebarrow Hill complex, note the cuspate nature of the slip surfaces and the boundary between areas of intact landslide blocks and no landslide blocks; and (b) a linear slip surface pattern associated with a translational slide block developed east of the Stonebarrow complex.
CHARACTERIZING LANDSLIDE RESERVOIRS
55
Fig. 4. A sketch cross-section and photographs of the Stonebarrow Hill landslide showing the evolution of the landslide blocks. (a) The area immediately beneath the onset of landslides; (b) mid-section with progressively thinner fault blocks; (c) lower area showing a local cliff, debris flows and mudslides on the beach. Two landslide events here resulted in the outlined deposits that have significantly different net to gross values. The graph illustrates the net: gross distribution of an equivalent subsurface dataset from the Statfjord East Flank.
identification of these structures is the amphitheatrelike form of the landslide. Since pre-existing heterogeneities and rock types can have a dramatic effect on the form of
landslides it is useful to consider other outcrop examples. In New Zealand, the Waipoapoa landslide in Southern Hawke’s Bay has been described by Pettinga (1987b). In that case, the intersection
56
A. I. F. WELBON ET AL.
of joint patterns developed in limestone beds controlled the landslide development resulting in a serrated edge to the back scar, but again with an overall cuspate form. Lithology also had a clear effect, as in the Dorset outcrops, controlling the position of the slip surface at the base on the complex. This detachment dips back into the complex indicating the material at the toe moved uphill during translation, driven by the force of the collapsing rock mass at the top of the mountain. Rock type can also prevent the development of the classic cuspate slip surface patterns, particularly when the rock types are cemented or the rate of collapse is slow. The collapse structures in the Canyonlands of Utah (Trudgill & Cartwright 1994) are often used as analogues for extensional faults in rifts, having similar map patterns. These rocks are bounded by slip surfaces which detach in salt and are formed as the result of cutting-down of the Colorado River into the detachment zone and are therefore landslides. Thus they are more akin to translational glide blocks with relatively straight, parallel slip systems linked by breached relay ramps.
Comparison of fault systems and landslides Extensional fault systems related to stretching or shortening in the crust are well studied and understood. Their growth, scaling relationships and seismic geometry, including variation, are documented in detail (Wells & Coppersmith 1994; Cartwright et al. 1995; Walsh et al. 2002). Similar databases exist for thrusts, which may have longer trace lengths (Boyer & Elliott 1982). Slip occurs on a metre scale over large areas of the fault surface during seismic events, and larger slip events occur on larger faults. Growth takes place over hundreds of thousands to millions of years and for rift events can occur over a period of 20 Ma (Jarvis & McKenzie 1980). The geometry of the fault systems is determined by through-going major faults that extend down to the brittle–ductile transition, c. 15 km in the crust (Morley 1995), and fault systems are dominantly planar. Linking occurs via overlap zones that form during fault growth, forming relay ramps and with continued slip, breaching of the ramp occurs. The key controls on the dimensions of faults are the degree of extension, wall rock strength and the mechanical layer thicknesses. These types of fault systems have distinct displacement patterns that reflect growth and linkage. Landslide systems have a geometry that also reflects the process and rate of formation. They develop over minutes to years, but typically hours
to days, through unconstrained movement and have associated strain patterns. Periods of landslide activity can stretch beyond the time interval of the driving mechanism, such as a rift event, which may be 10–20 Ma in duration. Detachment dominated systems prevail where the landslides are coherent (slides, slips) whereas with break-up and formation of incoherent mass movement the translation can either be predominantly at the base or within the landslide. A comparison between landslide blocks and fault blocks is shown in Table 1. Key geometrical differences exist, with landslides being typically curvilinear (cuspate) in map view, occasionally more linear if related to translational glide blocks. Normal fault systems and thrusts exhibit a more linear trend, with a dominant trend perpendicular to the principle stress direction, and a subordinate pattern of linking structures which formed during breaching of relay ramps (Morley 1995). The landslide map form can be asymmetrical, reflecting an oblique dip of the underlying shear surface. Hinge slips develop at the edge of the back scar and generally have relatively high displacement gradients reflecting high movement rates and wall rock strains. Initial map view geometries are cuspate, and with the resulting lateral unloading the map pattern evolves into so called ‘butterfly slips’ as a result of a widening or back stepping of the collapse area (Brunsden 1996). Normal faults in the field and in seismic occur as planar or listric in cross-section, with displacement values increasing with depth reflecting the origin of the failure (crustal stretching) and there is a tendency to map landslide slip surfaces with a similar style and displacement gradient. However, slip surfaces can form initially by compaction across a discontinuity, which leads to decreasing displacement values with depth, and when the landslips move they often form by movement on a planar slip surface which detaches on a stratigraphic boundary. This angular, non-listric geometry is likely to be a common alternative to the classic seismic interpretation of a rotational slip. The lateral extent of landslides can exceed the range for faults (in excess of 300 km), and as a reflection of the strain rate, fault seal properties may also differ. The recovery factors for landslide reservoirs may be higher than the range of fault blocks, as a consequence of the difference in the formation process. Finally, reworked material is associated with landslides and is likely to have a higher proportion of mudslides, debris flows and rockfalls than syn-rift sediments, reflecting the degradation of the fault scarp and the local nature of the sediment input, in contrast to the rift related sediments that may have a higher proportion of footwall derived or far travelled sediments.
CHARACTERIZING LANDSLIDE RESERVOIRS
57
Table 1. A comparison of landslide blocks and faults driven by plate forces Slip surfaces Formation mechanism Rate of formation Range of fault styles Scale (map) Map pattern Associated strain Fault seal potential Recovery factor of blocks Associated features
Faults
Gravitational processes Minutes to 1000s years Listric and planar, detachment dominated 100s kilometres to metres Cuspate or linear Open fractures, small scale faults, or none Yes, but mostly juxtaposition seal From very low to high
Crustal stretching 10 000 to 100 000þ years Dominantly planar but also listric, can reach 15 km depth 100s kilometres to metres Linear, with relay ramps Damage zone around fault dominates
Reworked material
Syntectonic sediments
Characterizing landslides using integrated analogue and subsurface data: an example from the Statfjord field The Statfjord Field is located in the Tampen Spur area of the North Sea (Fig. 5). The structure is a
Yes Moderate to high
rotated extensional fault block with a landslide complex on its eastern flank (Fig. 6). The stratigraphic units on the main field include the Brent and Dunlin Groups and the Statfjord Formation (Kirk 1980; Roberts et al. 1987; Hesthammer et al. 1999). The fault block is approximately
Fig. 5. A location map of the Statfjord Field in the North Sea.
58
A. I. F. WELBON ET AL.
Fig. 6. A seismic cross-section through the Statfjord field outlining the structural elements, including the Main Field, Main Bounding fault, the East Flank landslide complex, and the hanging wall. (See Fig. 7. for location of profile.)
25 km long and 10 km wide, and the landslide complex, termed the East Flank, is up to 5 km wide (Fig. 7). Although little has been published on this structure, (Roberts et al. 1987; Hesthammer & Fossen 1999) a unique dataset has been collected through the drilling of c. 130 wells during 25 years of production, planning secondary and tertiary recovery mechanisms and through several equity
Fig. 7. A map of the base Cretaceous unconformity, Statfjord Field illustrating the cuspate nature of the landslide complex. (SFA, SFB and SFC refer to platform locations on the Statfjord Field.)
redeterminations. Consequently, this is one of the world’s best studied subsurface landslide datasets. The Statfjord landslide complex developed in response to middle –late Jurassic rifting which spanned the late Callovian –Oxfordian era. Apart from a limited area of erosion in the north and south of the Statfjord fault block, which may have been above wave base, the landslide complex is interpreted to have been submarine, based on the conformable relationship of lower Heather Formation (Bathonian) pre-landslide shales above many landslide blocks and the in-situ Brent Group reservoirs in the main field. Movement continued to a lesser extent in post-Oxfordian times. The triggering mechanisms were likely to include the development of the footwall topography of the Statfjord fault block, earthquakes and associated tsunami loading, changes in fluid pressure and type related to diagenesis/compaction, hydrocarbon maturation and sediment loading. In this analysis, the seismic interpretation and reservoir characterization of the Statfjord East Flank was completed using geomorphological principles where the atlas of geometric and structural styles in Figure 1 was used. The database for the work incorporated well data, biostratigraphic classification of landslide material into ‘intact’ or ‘reworked’ material (from cores, sidewall core and cuttings), and production data (particularly pressure and fluid composition). In the seismic data, the landslide complex consists of the following classic features: cuspate slip surfaces in map view, a rapid thinning downslope and irregular topography at the top of the complex (Figs 7 & 8). In addition the basal slip surface to the complex has a stepped form, backtilted during rotation of the Statfjord fault block. In detail, the geomorphological features identified in the East Flank are rotational slides, translational glide blocks, and reworked material of talus slopes, rock falls, turbidites and mass movement systems. By integrating well, seismic,
CHARACTERIZING LANDSLIDE RESERVOIRS 59
Fig. 8. A cross-section through the Statfjord East Flank based on (a) seismic and (b) a geological model constructed using well data. The East Flank is divided into A, B and C blocks, defined by progressively deeper levels of detachment on the Base of Slope Failure (BSF). Each block has characteristic structural geometries and stratigraphic content.
60
A. I. F. WELBON ET AL.
sedimentological, biostratigraphic and structural information it is possible to create a cross-section through the landslide complex (Fig. 8). This crosssection is derived from depth converted seismic interpretations of slip surfaces and horizons, supported by horizontal and vertical well data to identify missing/repeated stratigraphy, indicating slip surfaces and unconformities. The cross-section is a generalized representation of the types of landslides found in the East Flank. It includes a stepped underlying slip surface which tends to follow the shale dominated stratigraphic units, this being termed the ‘base of slope failure’. The East Flank is divided into the A, B and C blocks based on the detachment level within the stratigraphy. Rotational slips occur near the back scar, translational glide blocks in the centre of the section and progressively thinner and more deformed landslide blocks are seen downslope before the slipped material becomes broken up into mudslides, debris flows and rockfalls of the reworked Brent complex. The proportion of material lost from the top of the blocks increases downslope, as does the proportion of reworked Brent relative to intact Brent rocks. Local accommodation space created by block topography becomes in filled by rockfalls, mudslides and debris flows. With increasing distance down-dip, the base of slope failure cuts through deeper stratigraphic units which then become part of the landslide complex. Since the main driving mechanism of the landslides is the topography of the major rift fault scarp and this developed episodically over time, there is also an implication in terms of the material incorporated into the landslide complex (Fig. 9). Early in the rift history, where topography on the fault scarp is relatively low, the dominant detachment surfaces are developed within shales of the Ness or lower Brent rocks. With increasing slip on the fault and widening of the footwall uplift zone, the deeper stratigraphic units (Dunlin Group, Statfjord Formation) become part of the complex. The implication for hanging wall sedimentation is that it is likely to be Brent dominated at the base and progressively more influenced by Dunlin and Statfjord rocks as the master fault grows and unroofing progresses. A further implication is that Brent rocks will dominate the hanging wall volumetrically since they form the highest proportion of lost material in the footwall. Based on the extensive database of well data and maps, local detailed seismic interpretation of the East Flank is possible. Two examples are presented in Figure 10. Figure 10a illustrates how stratigraphic and seismic reflector patterns vary from west to east. In the innermost part of the East
Flank, the data indicates rotational slips are present, in the central part there is a thinned layer of low dipping, translational blocks. Further downslope the reflectors actually have a counter regional dip interpreted as either successively deeper eroded blocks, break-up of fault blocks into debris flows or cambering resulting from extrusion of shales in exposed cliff faces at the front of blocks. The interpretation is supported by core data containing thinned and reworked sediments with mixed biostratigraphic assemblages. Figure 10b illustrates a seismic line with an interpretation also supported by well picks in well 33/9-A-7-B. Here a thin block of upper Brent Group stratigraphy has been emplaced on top of older, Dunlin Group rocks and the stratigraphy between is missing. The well data supports the flat reflector interpretation of the Dunlin block, but also indicates the original overlying Brent stratigraphy has been removed, and replaced by a separate Etive block derived from landslides up-dip. As with landslide complexes seen in Dorset and other outcrop analogues, the Statfjord landslide thins down flank. However, in contrast to outcrops, the Statfjord dataset has over 100 wells from an area of c. 20 km 5 km which allow maps of landslide block stratigraphy, thickness and presence of ‘reworked material’ (debris flows, mudslides, talus slopes etc.) to be created. One of the most powerful predictive tools is a subcrop map of the Heather Fm, Draupne Fm, reworked Brent constructed from well logs and seismic information (Fig. 11). This map represents the uppermost stratigraphic unit (generally the first well penetration) of intact blocks of the landslide complex. Changing patterns of stratigraphic units can be used to predict reservoir type down-dip, where well data is sparse. Boundaries between the mapped stratigraphic units are based on the last easternmost occurrence of the relevant well pick and the fault pattern. The map shows that the tops of landslide blocks contain progressively older stratigraphic units down dip, so that Tarbert Fm. is lost first, then Ness and Etive then followed by Rannoch. This pattern is also seen in cross – section (Fig. 8b). Beyond a certain distance down-dip from the onset of landslides, usually 1.5–2 km, there are no more intact Brent landslide blocks left and all wells penetrate Dunlin Gp rocks, directly beneath the Upper Jurassic Heather and Draupne Formations. As the landslide blocks in the East Flank thin and eventually disappear, the thickness of the reworked rocks generally increases but varies rapidly dependent on local topography and accommodation space. After characterizing the landslide complex through geomorphological classification and mapping, it is possible using empirical data from
CHARACTERIZING LANDSLIDE RESERVOIRS
61
Fig. 9. A sketch of the development of the Statfjord East Flank. Footwall uplift of the main bounding fault results in the successive emergence of the Brent, Dunlin and Statfjord rocks. Three stages of detachment development occur as the major shale horizons are exposed. As the footwall uplifts a wider zone becomes unstable, Brent rocks are progressively incorporated into the landslide, as well as deeper rocks. Early emplaced blocks in the hanging wall are of the higher stratigraphic units, progressively older rocks are also incorporated as the complex develops and the proportion of mudslides, debris flows increases with time.
62
A. I. F. WELBON ET AL.
Fig. 10. Two examples of seismic interpretation in the East Flank. (a) Transition from rotational slides to translational and thinning down slope. (b) On this section in well 33/9-A-7-B Etive rocks are emplaced on the Dunlin Gp rocks, with lower Brent absent, indicating stacked landslide blocks, far-travelled landslides and multiple events (complex slides). (See text for details.)
wells to define trends in net: gross and porosity (Fig. 12). Although observed patterns are very dependent on the stratigraphic unit and lithology, the presented examples are illustrative. Using data from the Brent intact blocks a pattern of increasing then decreasing porosity occurs with distance into the landslide complex in the Etive Fm (Fig. 12a), and generally decreasing porosity is seen in the Rannoch Fm (Fig. 12b). Based on geomorphological observations of block break-up in outcrop and core, the Etive distribution could be interpreted as fracture porosity enhancement as the landslide blocks decrease in size in the middle of the complex, and then reduction occurs as increased mixing of clays and porosity collapse takes place downslope. An interpretation of the Rannoch Fm
results, which was finer grained and more cemented at the time of deformation and of closer proximity to the shale dominated Dunlin, would be that porosity mixing of clays and grain packing dominates the process. A plot of porosity v. distance into the landslide complex for the reworked Brent (Fig. 12c) has the clearest trend, the gradual degradation in rock properties reflecting an increasing mixing of clay and mud downslope as more and more of the Dunlin rocks become incorporated. Similar patterns are also seen in net: gross. Note that the porosity reduction from 0.3 to 0.22 is greater than that expected from compaction processes, which would predict a reduction from 0.27 to 0.23 for sandstones in this depth range (Schlater & Christie 1980).
CHARACTERIZING LANDSLIDE RESERVOIRS
63
Fig. 11. A subcrop map of the Draupne/Heather indicating the uppermost stratigraphic unit in the landside blocks. The green area indicates no intact blocks have been encountered in wells or would be predicted. This map is from 2000; new updates of this map confirm the prediction, with minor discrepancies.
64
A. I. F. WELBON ET AL.
(Well No. 3, Fig. 13c) has sands with good permeability measured in the well, but after the well was set on production, no significant amounts of hydrocarbons flowed. These responses can be interpreted as the result of penetrating pockets of poor or good reservoir that are in connection with better or worse reservoir whether they be intact blocks or reworked sands. In other words, contrasts between the penetrated stratigraphy and that supplying hydrocarbons to the well are the main control on flow response. Other dynamic and well data can be used to characterize the landslide reservoirs of the East Flank. For example, production and well data collected over more than twenty years has enabled communication patterns to be mapped within the landslide complex (Fig. 14). A study was performed on the rock type that separates the reworked Brent from underlying landslide blocks for all wells in the East Flank. In most wells, particularly those in the outermost part of the landslide complex, the two reservoir sands are separated by shale that is at least 2 m thick (salmon pink areas on the map). This data was then combined with observations of pressure differences between the intact blocks and reworked Brent to see if there was a pressure barrier between the two. The conclusion was that in most areas of the east flank the reworked Brent is not in communication with the underlying landslide blocks and there is no relation between that fact and the presence of compartments in the underlying landslide blocks defined as having restricted or no communication with the Main Field (marked in blue). Fig. 12. Porosity variations in the East Flank with distance into the landslide complex. (a) Etive Fm; (b) Rannoch Fm; (c) Reworked Brent.
Although landslide reservoir characteristics can to some extent be predicted based on outcrop, subsurface and geomorphological analysis, well data indicate that in nature these reservoirs are highly variable in quality and spatial extent, especially if the system is small, and thereby are difficult to predict in detail. This can be illustrated in the reservoir and production characteristics of three wells on Statfjord in the East Flank (Fig. 13), where reworked reservoirs (debris flows, mudslides, etc.) are thought to be in close proximity to intact landslide blocks. All three wells were drilled above the contemporary oil–water contact. Well No.1 (Fig. 13a) has a relatively poor permeability measured from the log suite (2– 200 mD) yet produced oil. Well No. 2 has a very thick sand package with high measured log permeability in the reworked Brent but had poor production characteristics (Fig. 13b). The final well
Regional scale landslides: the Halten Terrace example Where well data is absent or limited the principles of geomorphological analysis can still be used to make predictions about reservoir type and quality. For example, well developed examples of subsurface landslide complexes are present in the Triassic and Jurassic of the Halten Terrace on the midNorway margin (Fig. 15). The structures are developed in Lower –Middle Jurassic pre-rift and Upper Jurassic syn-rift sediments, and formed in response to late Jurassic and early Cretaceous basementinvolved rifting along the Norwegian margin. There are three types of landslide geometry and process found on Halten Terrace that mirror the styles found in outcrop and on the Statfjord field. 1) Simple, unconfined rotational slides and translational glides, produced by gravity gliding. These geometries are related to the translation of a rigid body down a sloping detachment. The deformation occurred during movement
CHARACTERIZING LANDSLIDE RESERVOIRS
65
Fig. 13. Plots of three wells in the Statfjord East Flank illustrating the reservoir quality encountered in the well and production characteristics. (a) Poor measured permeability in the sands, good production; (b) high permeability sands with poor production; (c) good permeability sands which failed to produce.
66
A. I. F. WELBON ET AL.
Fig. 14. A communication map, typical stratigraphy and pressure development in the Statfjord East Flank. A vertical pressure barrier between intact blocks and reworked sediments is recognized based on dynamic data and identification of shale at the base of the reworked rocks. On the communication map, green areas illustrate contact between landslide blocks and the reworked reservoir, whereas pink areas have no communication. Blue polygons indicate compartments (restricted or no communication) between intact landslide blocks and the main field.
CHARACTERIZING LANDSLIDE RESERVOIRS
67
Fig. 15. A regional depth structure map of the base Cretaceous/top Middle Jurassic on the Halten Terrace, mid-Norway, (note rotation of map). The terrace is defined by two major fault zones, the Klakk and Bremstein Fault systems, which consist of dominantly NNE-striking segments typical of the Jurassic rift trend on this part of the margin. In the eastern parts of the Terrace, however, a number of large fault blocks are bounded by NNW or north-striking faults, which reflect the regional dip of a deeply buried Triassic detachment surface within a major relay ramp in the Klakk Fault system. In this area the dominant structural styles are formed by gravity driven transport rather than continental rifting.
68
A. I. F. WELBON ET AL.
over low basement topography, usually towards the west and SW. These are the most common landslide structures in the study area. 2) Complex, unconfined deformed slide blocks. These are characterized by the vertical collapse and lateral spreading of a body on a basal shear surface. These structures form more intensely deformed landslide blocks, indicating unconfined movement and higher strain rates over higher basement relief. This landslide type is less common, found only in the western parts of the terrace. 3) Contractional toe-structures where compression and uplift can occur in downslope settings, effectively balancing upslope extension. These structures are recognized only locally. These types of landslide are seen on a smaller scale in the Statfjord Field and in many of the discussed outcrops. The common element to all these three structural styles is the presence of a mechanically weak Triassic evaporitic sequence, which acts as a basal shear or detachment surface and decouples post-salt ‘cover’ geometries from pre-salt ‘basement’ fault systems to differing degrees. From well data it is known that two evaporitic sequences are present in the Ladinian and Carnian (Hollander 1984; Jacobsen & van Veen 1984). Middle–Upper Triassic thickness variations can also be observed in seismic data, and the distribution of landslide structures is closely associated with the depositional distribution of evaporites.
Landslide types on the Halten Terrace Rotational slides on the Halten Terrace are simple rotated fault blocks and graben with little or no internal deformation, formed by gravitational gliding on Triassic evaporite detachments (Fig. 16a). They are usually bounded by planar faults that terminate abruptly within the upper evaporitic section. The base evaporite reflector has little or no relief so that block rotation is taken up by lateral salt movement and the formation of thick salt ‘keels’ beneath the bounding faults. Locally the Jurassic stratigraphy within the fault block has grounded on the base evaporite surface. Translational slides are relatively large horst blocks bounded by detached rotational slides. They form by gravitational gliding and lateral translation on Triassic evaporite detachments (Fig. 16a). The base of the evaporitic section has little or no relief, so that the cover section is the least deformed of all structural styles in the study area. The slide blocks follow the regional dip at Top Upper Salt, so that Top Middle Jurassic is more or less conformable with Top Upper Salt, both of which are often
gently folded. The bounding slip surfaces may be planar or listric, either type terminating in the uppermost evaporites or occasionally lower evaporate section (Fig. 16b). Although the lateral movement of slide blocks is not obvious, it is implied by the adjacent linear collapse graben which form in response to extension both up-dip and down-dip. Some of the largest hydrocarbon discoveries on the Halten Terrace are structures of this type (e.g. Kristin, Tyrihans, Trestakk and Onyx). Semi-regional seismic data suggest structural styles associated with gravity gliding occur in areas with a thick isopachous Upper Salt, low pre-rift basement topography and moderate regional dip at base Upper Salt. The orientation of slip surfaces bounding the slide blocks can also be used to imply regional dip patterns on the basal shear surface at the time of deformation. In western parts of the Halten Terrace and Sklinna Saddle, regional dip at base Upper Salt is to the SW, and dominant slip surface strike/dip is perpendicular/ synthetic to this dip (Fig. 15). Where landslide processes develop on relatively steeply dipping basal detachment slopes, and are unconfined in the downslope direction, the common characteristic is a series of strongly rotated blocks detached on an inclined Triassic evaporite substrate (Fig. 16b). In the upslope domain the blocks are not significantly deformed and are similar to rotational slides and translational glides already described. However, in the downslope direction the blocks become progressively more deformed and thinned through internal shear. It is likely that a full Jurassic section is preserved in the landslide blocks, but reflection character is lost due to small-scale faulting, fracturing and porosity collapse and internal stratigraphy is not recognizable. The most distal blocks exhibit chaotic seismic character and are possibly affected by deformation processes associated with the formation of debris flows, slides and rock falls. As described previously, contractional structures can also be significant elements of gravity driven deformational systems. One excellent example of this is seen in the hanging wall to the large basement-rooted Smørbukk Fault, west of the Smørbukk Field (Fig. 16c). A large anticline can be mapped with a fold axis parallel to the Smørbukk Fault and a trace length of over 25 km (Fig. 15). The fold is best developed within the Lower –Middle Jurassic section, but amplitudes are diminished at base Cretaceous and the top Triassic Salt is also more or less unfolded. The fold is strongly asymmetrical, with a short east-dipping steep limb, and a longer west-dipping limb with lower dips. The western limb passes into a syncline with rotational slides on its western flank. Most of
CHARACTERIZING LANDSLIDE RESERVOIRS
69
Fig. 16. Regional seismic lines from the eastern part of the Halten Terrace illustrating typical geometries in landslide complexes. (a) A profile through the internal, mostly confined, and less deformed part of a landslip complex. The dominant structural styles are rotational and translational slide blocks, gliding over a weakly dipping detachment. (b) A more mature and external part of the same system, characterized by a more steeply dipping detachment, less confinement and a transition from gravity gliding to gravity spreading towards the SW. (c) is a Cross-section through the Smørbukk Field and adjacent hanging wall block. The dominant fault (the Smørbukk Fault) is a basement involved normal fault which generated strong rotation of hanging wall stratigraphy during Jurassic rifting. The resultant steep dips within the Triassic evaporites generated gravitational instability and collapse of the overlying Jurassic stratigraphy, reflected in both up-dip extensional collapse and down-dip compressional uplift above the local regional.
70
A. I. F. WELBON ET AL.
the fold crest seems to be elevated above a regional defined by the erosional unconformity on the Northern Sklinna High and average elevation for the Smørbukk Field, although positioning the regional? in the Sklinna area is difficult as the degree of truncation is uncertain. However, the geometry is interpreted to be an extensional ‘roll-over’ anticline modified by later gravity-driven compression. As a result of the strong rotation of the hanging-wall block during basement rifting, the entire post-salt stratigraphy became gravitationally unstable and began to move downslope towards the east. This resulted in detached up-dip extensional sliding paired with downslope contractional folding, buttressed by the Smørbukk Fault.
Controls on structural style The landslips on the Halten Terrace were active in the late Jurassic and early Cretaceous (based on growth patterns across fault blocks). This implies initiation during major regional rift-related subsidence, which did not begin until Bathonian time. Accommodation space in graben and rotated hanging walls is generated by down dip gravitational collapse on a basin scale directed towards the Møre-Vøring Basin in the west. Driving mechanisms would be similar to those in the Viking Graben, (e.g. earthquakes and associated tsunami loading, build-up of fault block topography, maturation of hydrocarbons). Present day structural geometries can be linked to tectonic controls at the time of deformation, and these concepts can be used to aid seismic interpretation in areas of poor data, and to predict the type and intensity of deformation within local structures. For the landslide complexes observed on the Halten Terrace, the most important structural controls are the distribution of Triassic evaporites (which impact the degree of linkage between sub-salt ‘basement’ and post-salt ‘cover’ faulting) and the ‘triggering’ of gravity-driven collapse by deep-seated crustal extension determining the dip and orientation of evaporitic substrates beneath the Jurassic sediment pile. Thick-skinned crustal extension in the late Jurassic generated sufficient basement topography to create unstable slopes within the evaporite section, and caused gravity collapse of the cover stratigraphy. Lower strain rates during earlier Jurassic extension do not appear to have been sufficient to trigger any cover deformation or evaporite mobility. Also, since the evaporites were buried to several hundred metres at the time of deformation, they were likely to be immobile and functioned mostly as detachment surfaces. On the Halten Terrace, moderately low syntectonic sediment
accumulation rates (of the Melke and Spekk Formations) mean that sediment loading is unlikely to have driven downslope displacement. The geometry of the basal slope itself and the load of the existing rock mass are likely to have been the most significant factors.
Discussion Landslides in the subsurface are an important set of reservoirs that form an increasing part of the hydrocarbon industry’s exploration and development (E&P) portfolio in deep water passive margins, rifts and collisional belts. A wealth of information from geomorphological observations and research can be used to characterize landslide evolution. The geomorphological understanding combined with data and empirical observations from the subsurface, such as thinning patterns and net: gross statistics provides a tool to predict reservoir quality. Landslides have large ranges of scale from metres to kilometres and within a complex can exhibit highly variable net: gross and permeability. A classification of landslides can be made on the basis of form and material type, in the context of tectonic setting, timing, and depositional environment data derived from structural, biostratigraphic and sedimentological studies. There are several predictive tools that can aid in predicting landslide reservoir type and quality. Seismic mapping is the most important, based on interpreting structures in 3D using geomorphological principles and an atlas of landslide types (Fig. 1). Where well data exist, subcrop maps can be used to map the content of intact landslide blocks and maps can be made of the distribution of reworked material (e.g. debris flows, mudslides, rockfalls). Integrated studies are the key to reservoir characterization, which can include for example biostratigraphic, sedimentological and dynamic studies to classify intact versus reworked sediments, or seismic inversion of the landslide complex to predict net: gross. When the landslide complex has been mapped, an attempt can be made to model the range of thickness or net: gross for new well locations. In the example of the Statfjord East Flank, the geometry of the landslide complex is clear from the form of the Base Cretaceous Unconformity and the multiple detachment levels the Ness Fm, Dunlin Gp and Statford Fm shales. The thinning of the complex is apparent in both seismic and well data and the associated changes in net: gross are documented from a subcrop map (Fig. 11). Reworked sediments are delimited on a communication map (Fig. 14).
CHARACTERIZING LANDSLIDE RESERVOIRS
Using these datasets, together with a simplified stratigraphy, an idealized representation of the processes described in this paper can be generated (Fig. 17). The impact of different processes on reservoir distribution and quality can be estimated, and a trend for net: gross and likely reservoir thickness with distance into the landslide constructed. To quantify the processes at work, empirical data from the Statfjord East Flank can be used (see Figs 12 & 13). The example presented shows three processes documented in outcrop: (1) stratigraphic thinning, whilst maintaining net: gross; (2) erosion from the top of the blocks; and (3) break-up and formation of reworked material. Starting with an arbitrary undeformed ‘input’ stratigraphy with a net: gross of 0.5, the thinning model leads to progressively reduced stratigraphic thicknesses within the complex but no decay in net: gross, in line with outcrop and Statfjord observations. In the erosion model all reservoir units are lost by c. 1 km into the complex, and net: gross decreases proportionally. Conversely in the reworked sediment model, thicknesses are controlled by the amount of erosion and break-up of
71
the landslide blocks, and by local fault block topography. The result is a pattern of variable but general thickening downslope, with a large range in net: gross. The products of different types of landslides have significantly different implications for hydrocarbon potential. At one extreme the highly complex and variable mix of landslide types found in the Statfjord East Flank have a large impact, both in positive and negative ways. At the other extreme, a large, slowly emplaced translational glide block, such as those seen locally on the Halten Terrace, will have a lesser impact on net: gross and stratigraphic thickness, which may be very similar to the initial undeformed stratigraphy, improving prospectivity. The impact of landslides on fault seal processes and prediction techniques has rarely been discussed in publications. Occasionally fault seal has been considered for large gravitationally driven reverse faults on passive margins or for normal faults that may well be landslide structures (Welbon et al. 1997). The impact of landslides on fault seal is dominated by the likely geometrical and
Fig. 17. A method to predict reservoir ranges based on empirical data and geomorphological process understanding. Using an idealized net to gross input data (0.5) and empirical data on thinning factors, erosion and development of reworked sediments from the Statfjord East Flank, predictions of net to gross and thickness can be made based on end member processes. (See text for details.)
72
A. I. F. WELBON ET AL.
petrophysical characteristics of slip surfaces, namely the slip rate and displacement gradient, material properties in the slip zone, wall rock and matrix strain, the presence of sand injectites or thief sands within reworked sediments. Geometrical differences between faults and landslide slip surfaces are important, as juxtaposition across the slip surface is likely to be the primary control on across structure flow (Allen 1989). Where high displacement gradients exist on landslide slip surfaces, (higher than faults), then there is a higher likelihood of leakage, given appropriate scales of seal thicknesses (James et al. 2004). With higher strain rates, statistically higher numbers of relay ramps and lenses may exist leading to higher leakage tendencies. Conversely, slow moving translational slide blocks bounded and containing low displacement gradient structures may be more continuous and have fewer leakage points, but may be affected by many open fractures in response to the pull apart process at chasms bounding the slide block. As with faults, the stratigraphic architecture of the reservoir rocks (amalgamation ratio, net: gross) will be key (Bailey et al. 2002; Manzocchi et al. 2007) and landslides developed within reworked sediments will present particular challenges since high variability of net: gross, topography and scale means amalgamation ratio will be variable and difficult to predict. Since slip surfaces in landslides can form under very high strain rates, the materials in the slip zone are likely to have quite different characteristics to faults. One key difference is the fact that faults formed by plate tectonic forces occur by seismic slip, on patches of the fault surface which coalesce, and are followed by creep (Wells & Coppersmith 1994). The landslide slip surface may move entirely by non-seismic creep processes, or have individual movement events at a seismic rate but with much higher slip magnitudes during one event. For example, faults may move up to 20– 30 m in a single seismic event; landslides can often move hundreds of metres, imparting more strain in the slip zone. Another difference is the proportion of injected sands and shales in a landslide versus a fault zone. Both sand and shale injection have been documented for faults but is mostly associated with polygonal fault systems. Based on outcrops and an understanding of process, landslides are likely to have higher proportions of injected fluids along the slip surface since higher strain rates in the wall rocks and underlying material result in sediment remobilization, and detachment dominated slip systems can move laterally, opening up the fault to injection from below, or the infill from sediments at the sea floor above.
The consequences of this are that the traditional algorithms for fault seal (e.g. shale smear factor and shale gouge ratio; Lindsay et al. 1993; Yielding et al. 1997) may not be as generally applicable to landslides and must be used with caution. If possible, local calibration of the fault rocks and displacement gradients must be employed in estimating permeability of the fault zone.
Conclusions Landslides are an increasing component of the hydrocarbon industry’s E&P portfolio, on passive margins, in rifts and on occasion, in mountain belts. Landslide reservoirs are difficult to characterize, having larger ranges of scale and complexity than conventional fault-related reservoirs. In rifts they are often at the limits of seismic detection or poorly imaged because of geological complexity. Landslides have a range of distinct structural styles generally different from fault systems in rifts and this is reflected in reservoir quality. Using an atlas of structural styles, adapted surface mapping techniques, empirical databases and integrated studies it is possible to identify the landslide types and make a prediction of reservoir presence (for example from subcrop maps), thickness and net: gross. To characterize landslide reservoirs effectively, it is necessary to train the subsurface team and have robust data collection and use. Optimally acquired and processed seismic data is important (Ocean Bottom Cable being preferred) as is the acquisition of well data including image log and biostratigraphic information. Traditional techniques of sequence stratigraphy and reservoir characterization and uncertainty modelling do not apply. 3D geological models need to be based on modified workflows that can accommodate the lack of mappable surfaces in seismic, or the fact that single horizons can represent multiple landslide and reservoir types, and capture the rapid change in reservoir properties. Established algorithms of fault seal need to be adapted and the connectivity of the slide surfaces mapped. The focus of the industry in many rifts has been towards characterizing landslides reservoirs on the crest of existing fault block accumulations, which are often the last target of incremental oil recovery studies, but focus now has to switch to the other products of landslides (e.g. the deposits in hanging walls which constitute exploration plays). On passive margins, both extensional and compressional structures are seen, many on a large scale, often associated with salt or mud diapirism. Rates of landslide development in this case control sedimentation and reservoir quality. Again documenting and understanding evolution of
CHARACTERIZING LANDSLIDE RESERVOIRS
geometry based on geomorphological principles is important, which will indicate the deep water system depositional slope, as is measurement of the rate of slip surface development which helps in fault seal assessment. Landslide reservoirs are a challenge, reflecting the larger range of uncertainty and risk involved in characterizing and developing them. However, since world wide oil production has exceeded new discoveries since the early 1980s and this situation is likely to perpetuate a high oil price environment, the extra costs of characterizing and exploiting landslides are likely to be offset by the value of the reservoirs. Many people have contributed the understanding of subsurface landslides presented in this paper, in particular the Statfjord landslides. Amongst the pioneers were the original operators in Mobil and their partners and C. Jourdan in Statoil. Others who worked systematically on the East Flank include S. Becker, A. Cullum, H. Fossen, K. Gibbons, J. Hesthammer, S. E. Morterud, P. E. Nielsen and T. Vangsnes. S. Becker and T. Vangsnes are acknowledged for Figure 13 and S. E. Morterud for part of Figure 14. The Halten Terrace interpretations presented here build upon earlier work in Statoil by Ø. S. Kløvjan, M. Larsen, S. Hansen and co-workers, whose contributions to the geological understanding of this region we would also like to acknowledge. The significant contribution of Statoil’s partners is acknowledged they have been vital in building an understanding of subsurface landslides, including C. Kiven and J. Vold.
References A LLEN , U. S. 1989. Model for hydrocarbon migration and entrapment within faulted structures. Bulletin of the American Association of Petroleum Geologists, 73, 803–811. B AILEY , W. R., M ANZOCCHI , T., W ALSH , J. J. ET AL . 2002. The effect of faults on the 3D connectivity of reservoir bodies: a case study from the East Pennine Coalfield, UK. Petroleum Geoscience, 8, 263–277. B ERGER , M. & R OBERTS , A. M. 1999. The Zeta Structure: a footwall degradation complex formed by gravity sliding on the western margin of the Tampen Spur, Northern North Sea. In: F LEET , A. J. & B OLDY , S. A. R. (eds) Petroleum Geology of Northwest Europe. Proceedings of the 5th Conference. Geological Society, London, 107–116. B OYER , S. & E LLIOTT , D. 1982. Thrust systems. Bulletin of the American Association of Petroleum Geologists, 66, 1196– 1230. B RUNSDEN , D. 1984. Mudslides. In: B RUNSDEN , D. & P RIOR , D. B. (eds) Slope Stability. Wiley, Chichester, 363–418. B RUNSDEN , D. 1996. Landslides of the Dorset coast: some unresolved questions. Proceedings of the Ussher Society 9, 001–011.
73
B RUNSDEN , D. & J ONES , D. K. C. 1976. The evolution of landslide slopes in Dorset. Philosophical Transactions of the Royal Society Series A, 283, 605–631. B RYN , P., S OLHEIM , A., B ERG , K., S EJRUP , H. P. & M IENERT , L. (eds) 2005. Ormen Lange: an integrated study for safe field development in the Storegga submarine area. Marine and Petroleum Geology, 22, 1–318. B UCKLAND , W. 1840. On the landslipping near Axmouth. Proceedings of the Ashmolean Society 1, A2– 8. C AMPBELL , C. J. & L AHERRERE , J. H. 1998. The end of cheap oil. Scientific American, 278(March), 78– 83. C ARTWRIGHT , J. A., T RUDGILL , B. D. & M ANSFIELD , C. S. 1995. Fault growth by segment linkage: an explanation for scatter in maximum displacement and trace length data from the Canyonlands Grabens of SE Utah. Journal of Structural Geology, 17(9), 1319– 1326. C OBBOLD , P. R., M ORGUES , R. & B OYD , K. 2004. Mechanism of thin-skinned detachment in the Amazon fan, assessing the importance of fluid overpressure and hydrocarbon generation. Marine and Petroleum Geology, 21, 1013–1025. C RUDEN , D. M. 1991. A simple definition of a landslide. Bulletin International Association for Engineering Geology, 43, 27– 29. D IKAU , R., B RUNSDEN , D., S CHROTT , L. & I BSEN , M. L. 1996. Landslide Recognition: Identification, Movement and Causes. John Wiley. EPOCH. 1993. Temporal Occurrence and forecasting of landslides in the European Community 3 Volumes. European Community. F ARRELL , S. G. 1984. A dislocation model applied to slump structures, Ainsa basin South Central Pyrenees. Journal of Structural Geology, 6, 727–736. H ALL , C., T HARAKAN , P., H ALLOCK , J., C LEVELAND , C. & J EFFERSEN , M. 2003. Hydrocarbons and the evolution of human culture. Nature, 426, 318–322. H EINIO , P. & D AVIES , R. J. 2006. Degradation of compressional fold belts: deep-water Niger Delta. Bulletin of the American Association of Petroleum Geologists, 90, 753– 770. H ESTHAMMER , J. & F OSSEN , H. 1999. Evolution and geometries of gravitational collapse structures with examples from the Statfjord Field Northern North Sea. Marine and Petroleum Geology, 16, 259–281. H ESTHAMMER , J., J OURDAN , C. A., N IELSEN , P. E., E KERN , T. E. & G IBBONS , K. A. 1999. A tectono-stratigraphic framework for the Statfjord Field, northern North Sea. Petroleum Geoscience, 5, 241–256. H OLLANDER , N. B. 1984. Geohistory and hydrocarbon evaluation of the Haltenbank area. In: S PENCER , A. M. et al. (eds) Petroleum Geology of the Northern European Margin. Graham and Trotman, London, 383– 388. H UTCHINSON , J. N. 1968. Mass movement. In: F AIRBIDGE , R. W. (ed.) Encyclopedia of Geomorphology. Reinhold, New York, 688– 695. J ACOBSEN , V. W. & VAN V EEN , P. 1984. The Triassic offshore Norway north of 628N. In: S PENCER , A. M. et al. (eds) Petroleum Geology of the Northern European Margin. Graham and Trotman, London, 317– 327. J AMES , W. R., F AIRCHILD , L. H., N AKAYAMA , G. P., H IPPLER , S. J. & V ROLIJK , P. J. 2004. Fault-seal analysis using a stochastic multifault approach.
74
A. I. F. WELBON ET AL.
Bulletin of the American Association of Petroleum Geologists Bulletin, 88, 885–904. J ARVIS , G. & M C K ENZIE , D. 1980. Sedimentary basin formation with finite extension rates. Earth and Planetary Science Letters, 48, 42–52. K IM , Y.-S. & S ANDERSON , D. J. 2004. The relationship between displacement and length of faults: a review. Earth Science Reviews, 68, 317–334. K IRK , R. H. 1980. Statfjord field – a North Sea giant. In: H ALBOUTY , M. T. (ed.) Giant oil and gas fields of the decade 1969–78. American Association of Petroleum Geologists, Memoir 30, 95– 116. L INDSAY , N. G., M URPHY , S. C., W ALSH , J. J. & W ATTERSON , J. 1993. Outcrop studies of shale smears on fault surfaces. International Association Sedimentologists Special Publication, 15, 113 –123. M ANZOCCHI , T., W ALSH , J. J., S TRAND , J. A., T OMASSO , M., C HILDS , C. & H AUGHTON , P. 2007. Static and dynamic connectivity in bed scale models of faulted and unfaulted turbidites. In: J OLLEY , S. J., B ARR , D., W ALSH , J. J. & K NIPE , R. J. (eds) Structurally Complex Reservoirs. Geological Society, London, Special Publications, 292, 309–336. M C L EOD , A. E. & U NDERHILL , J. R. 1999. Footwall scarp instability and the processes and products of degradation Northern Brent Field Northern North Sea. In: F LEET , A. J. & B OULDY , S. A. R. (eds) Petroleum Geology of Northwest Europe. Proceedings of the 5th Conference. Geological Society, London, 91–106. M ORLEY , C. K. 1995. Developments in the structural geology of rifts over the last decade and their impact on hydrocarbon exploration. In: L AMBIASE , J. (ed.) Hydrocarbon Habitat in Rift Basins. No. 80. Geological Society, London, Special Publications, 1–32. P ETTINGA , J. R. 1987a. Ponui Landslide: a deep seated wedge failure in Tertiary weak-rock flysch, Southern Hawke’s Bay, New Zealand. New Zealand Journal of Geology and Geophysics, 30, 415–430. P ETTINGA , J. R. 1987b. Waipoapoa Landslide: a deep seated complex block slide in Tertiary weak rock flysch, southern Hawkes Bay, New Zealand. New Zealand Journal of Geology and Geophysics, 30, 401–414. P ITTS , J. & B RUNSDEN , D. 1987. A reconsideration of the Bindon Landslide of 1839. Proceedings of the Geologists Association, 98, 1 –18. P RIOR , D. B. & C OLEMAN , J. M. 1982. Active slides and flows in underconsolidated marine sediments on the slopes of the Mississippi delta. In: E MBLEY , R. W., S AXOV , S. & N IEUWENHUIS , J. K. (eds) Marine Slides and Other Mass Movements. Plenum Press, New York, 21–49. R OBERTS , J. D., M ATHIESON , A. S. & H AMPSON , J. M. 1987. Statfjord. In: S PENCER , A. M., H OLTER , E.,
C AMPBELL , C. J., H ANSLIEN , S. H., N ELSON , P. H. H., N YSÆTHER , E. & O RMASSEN , E. G. (eds) Geology of the Norwegian Oil and Gas Fields. Graham and Trotman, 319– 340. S CHLATER , J. G. & C HRISTIE , P. A. F. 1980. Continental stretching: An explanation of the post mid-Cretaceous subsidence of the Central North Sea. Journal of Geophysical Research, 85, 3711– 3739. S EED , H. B. 1968. Landslides caused by soil liquifaction. Journal of Soil Mechanics and Foundations Division, 94, 1055– 1122. S TEWART , S. A. 1997. Relationship between basementlinked and gravity-driven fault systems in the UKCS salt basins. Marine and Petroleum Geology, 14, 581–604. T RUDGILL , B. & C ARTWRIGHT , J. 1994. Relay-ramps forms and normal-fault linkages, Canyonlands National Park, Utah. Geological Society of America Bulletin, 106, 1143–1157. U NDERHILL , J. R., S AWYER , M. J., H ODGSON , P., S HALLCROSS , M. D. & G AWTHORPE , R. L. 1997. Implications of fault scarp degradation for Brent group prospectivity, Ninian field Northern North Sea. Bulletin of the American Association of Petroleum Geologists, 81, 999–1022. V ARNES , D. J. 1978. Slope movement: types and processes. In: S CHUSTER , R. L. & K RIZEK , R. J. (eds) Landslide Analysis and Control. Transportation research board special report 176 Washington, D.C. 11–33. V ARNES , D. J., R ADBRUCH -H ALL , D. H. & S AVAGE , W. Z. 1989. Topographic and structural conditions in areas of gravitational spreading in the Western United States. U.S. Geoogical Survey, Professional Paper, 1496, 1 –27. W ALSH , J. J., N ICOL , A. & C HILDS , C. 2002. An alternative model for the growth of faults. Journal of Structural Geology, 24, 1669–1675. W ELBON , A. I., B EACH , A., B ROCKBANK , P. J., F JELD , O., K NOTT , S. D., P EDERSEN , T. & T HOMAS , S. 1997. Fault seal analysis in hydrocarbon exploration and appraisal: examples from offshore mid-Norway. In: M ØLLER -P EDERSON , P. & K OESTLER , A. G. (eds) Hydrocarbon Seals: Importance for Exploration and Production. NPF Special Publication 7. Norwegian Petroleum Society, 125–138. W ELLS , D. L. & C OPPERSMITH , K. J. 1994. New empirical relationships among magnitude, rupture length, rupture width, rupture area and surface displacement, 84, 974–1002. Y IELDING , G., F REEMAN , B. & N EEDHAM , D. T. 1997. Quantitative fault seal prediction. Bulletin of the American Association of Petroleum Geologists, 81, 897–917.
The fused fault block approach to fault network modelling K. S. HOFFMAN1 & J. W. NEAVE2 1
Roxar, Inc., 14701 St Mary’s Lane, Houston, TX 77079 USA (e-mail:
[email protected])
2
Roxar, Inc., 2201 Walnut Ave, Suite 240, Fremont, CA 94538 USA
Abstract: Fault network modelling of complex faulted structures, those containing hundreds or even thousands of faults, can be an extremely difficult and time-consuming process. Although techniques for mapping and modelling faulted structures have been in existence for nearly forty years, asset teams still struggle to create correct portrayals of such complex faulted reservoirs due to the limitations of the commonly used techniques. We have developed a new approach to fault network modelling, using a new concept of ‘fused’ fault blocks. The identification of fault– fault intersections is based not on a manually drawn fault network or table of relationships, but rather is derived from the fault surfaces themselves. The calculated intersection lines are then used to truncate faults against each other. Because the truncation information can be stored with the fault model, this process yields a repeatable and easily updatable fault model. The name of our technique ‘fused fault blocks’ refers to the fact that when a section of a fault is removed, the two fault blocks that had been created by the fault are then fused together, forming a single fault block. The resultant fault model can then be used to create a 3D reservoir grid, one in which the fault geometry has not been compromised, and therefore better reflects the actual structure. The speed of the fault-building process ‘seconds or minutes, even for models with hundreds of faults’ also allows multiple interpretations, placing the emphasis of the fault network building on the evaluation of the interpretation and the effects of compartmentalization, and not on the manipulation of software.
One of the primary goals of computer mapping and modelling techniques is to produce a model that is structurally possible, internally consistent, and a viable representation of the interpretation. Twodimensional mapping techniques first became available in the late 1960s and early 1970s; commonly these mapping algorithms did not use standard contouring rules to create the 2D grids and therefore could produce impossible or unreasonable structures. With the advent of 3D computer graphics in the 1980s, mapping and modelling expanded into the 3D world as well. Sophisticated algorithms for distributing petrophysical properties or facies in 3D quickly became indispensable, as these techniques provide better information for reserve calculation and well planning than simple 2D maps. However, the modelling of complex structures continues to be a problem. Several methods have been developed for fault surface and fault network modelling, each of which has advantages and disadvantages. Most methods have practical, if not absolute, limitations to the number of faults that can be incorporated into a model simply due to the size of the resultant model or the complexity of building the network. Many also have limitations as to the types of fault intersections that can be modelled. Exploration and development continues to expand into increasingly complex and risky areas,
where the accuracy of the models becomes more and more important. If the fault network on its own was the final desired result, models with hundreds or thousands of faults could easily be created on a routine basis. Many problems arise in using the fault framework to create the full, layered structural model and in using that complete framework for reservoir gridding and subsequent petrophysical and facies modelling. The fused fault block method of fault network modelling has been developed to address the limitations of current methodologies; to eliminate the restrictions on numbers of faults and types of fault intersections, to increase the speed of the process, and to allow the accuracy of the structural framework to be carried into the reservoir grid. Creating a fault model requires two basic steps: calculating the fault surfaces and calculating truncations. Fault surface modelling is the process of creating a 2D grid surface from a set of input data points. The surfaces are calculated independently of one another, and may cross or intersect each other. Where faults intersect, one fault may be truncated against the other. Calculating these truncations is therefore the process of specifying which fault truncates against the other. Faults are not required to truncate at intersections; faults that do cross, such as X faults, are allowed.
From: JOLLEY , S. J., BARR , D., WALSH , J. J. & KNIPE , R. J. (eds) Structurally Complex Reservoirs. Geological Society, London, Special Publications, 292, 75– 87. DOI: 10.1144/SP292.4 0305-8719/07/$15.00 # The Geological Society of London 2007.
76
K. S. HOFFMAN & J. W. NEAVE
Fault surface modelling The creation of fault surfaces is a straightforward procedure, as it is primarily a question of turning a set of XYZ points into a surface; most of the complexity comes later during the construction of the fault network. Surfaces are generated using either a parametric or a triangulation approach. Parametric surfaces are generally created using B-splines (Akima 1970) in a regular, 2D grid. Discrete smooth interpolation (Mallet 1992) may be used for triangulation. Both methods have advantages and disadvantages.
Single-valued v. multi-valued Most faults are single-valued surfaces, having only one Z value at any given XY location. However, multi-valued faults may be encountered in compressional structures where faults are near vertical (such as flower structures) or where faults have been folded. Triangulation methods are well-suited to model multi-valued surfaces; true three-dimensional shapes such as salt domes or folded faults do not have to be treated as special cases. Many 2D gridding algorithms used to create a parametric surface are based on a grid that is regular and orthogonal in XY space. These grids are by definition single-valued and cannot be used to represent an overturned surface or a completely vertical surface. Methods that use this approach (Belcher 1994) therefore have difficulty with vertical or near-vertical faults and cannot model multivalued fault surfaces. To overcome this limitation, the fault data can be transformed into a space where they are single-valued (Zoraster & Bayer 1993). This method, which is used in the fused fault block technique, determines a plane which best fits the XYZ data, transforms the data so that the best-fit plane is a horizontal surface, and calculates the fault surface using standard parametric gridding techniques in the transformed space. Consider the data shown in Figure 1a. The fault is a single-valued surface with a curved shape. In Figure 1b, a best-fit plane has been calculated for these data (dashed line). A transform function is calculated which rotates this plane so that it becomes horizontal; the data are moved into this transform space (Fig. 1c) and a standard parametric gridding algorithm is used to calculate a surface. The extrapolation of the surface using standard gridding algorithms may be more controlled in this transformed space than in actual XYZ space. Vertical faults thus become ‘horizontal faults’ and are easily modelled; even multi-valued faults can often be transformed into a space where they
Fig. 1. (a) Cross-sectional view of input data points for a fault surface (circles). (b) A plane that approximates the overall trend of the data (dashed line). (c) Data transformed into a space where the best-fit plane is a horizontal surface.
are single-valued. A truly vertical fault is impossible to model using a gridding algorithm that works in XY space, as the fault would never
FUSED FAULT BLOCK NETWORK MODELLING
extend over even a single grid cell. Faults that are nearly vertical but change dip direction slightly also cannot be modelled in XY space using a gridding algorithm, as the 2D grid is a single-valued surface. The resulting surface would contain spikes (Fig. 2a); the ‘solution’ would be to edit the surface or delete data points and the resulting surface would no longer honour the actual structure. This type of fault can only be modelled using a triangulation method or by transforming the data into a coordinate system where it is a single-valued surface (Fig. 2b). Although a triangulation method would appear to have an advantage in modelling multi-valued fault surfaces, it is computationally expensive and does not provide all of the flexibility of a parametric surface, particularly in interactive modification
77
(Segonds et al. 1998). Parametric techniques are standard 2D gridding algorithms, may offer the interpreter a wider choice of algorithms, and may be faster.
Visualization Computer graphics libraries display surfaces as a series of triangles. Surfaces may appear smooth by using Gouraud shading (Gouraud 1971). A triangulated surface may be visualized directly without any further processing. Parametric surfaces are usually triangulated for speed of visualization and also for performing surface– surface intersection calculations. The choice of methodology sometimes becomes a trade-off between the initial steps and the later use. The triangulated surfaces are difficult to modify and lack the type of continuity needed for calculation of derivatives (Charles et al. 1995). They often need to go through a parameterization step in order to build a reservoir grid. The parametric approach needs to go through a triangulation step for visualization, but provides more flexibility in surface manipulation and a more direct workflow to get to a reservoir grid.
Methods used in the fused fault block approach The fused fault block method uses parametric surfaces for the faults. These are created in a transformed space, so that vertical, near-vertical and multi-valued faults can be modelled. Each fault in a model has its own transform function. For visualization purposes, the faults are back-transformed into the actual coordinate space and triangulated. This approach provides the speed and flexibility of the standard parametric technique and also the ability of the triangulation technique to handle multi-valued surfaces.
Fault network modelling
Fig. 2. (a) Cross-sectional view of data points (circles) for a near-vertical fault and the surface that would be generated using an XY-based gridding algorithm. (b) The same data as (a), but a surface generated in a transformed space.
Once the individual fault surfaces are made, they must be joined together into a fault network. The fault network forms the basis for stratigraphic modelling and it is the complete faulted stratigraphic model which is used for reservoir gridding and the subsequent attribute modelling. The most commonly used methods for fault network modelling can be divided into two categories: pillar or node-based methods and binary tree methods. Each of these has its advantages and disadvantages, but neither method can handle the full range of fault intersections or truncations.
78
K. S. HOFFMAN & J. W. NEAVE
Pillar method At its simplest, a pillar is a line that exists in 3D space. Pillars may be used in several places within the geological modelling process, including fault modelling and reservoir grid building. In a reservoir grid, a pillar is a straight line that ties cells together from the top to the bottom of the grid (Fig. 3). Although some flow simulators do allow curved segments in pillars, they are commonly straight lines. Because a modelling workflow very often includes attribute modelling and flow simulation, pillars are also used to control the shape of the fault surface and the intersections or truncations of faults (Fig. 4). Fault pillars are usually allowed to contain 2, 3, or 5 nodes. Although the use of pillars in fault surface generation ensures that the fault model and the reservoir grid are consistent, it has the disadvantage of sometimes simplifying the fault shape, which can adversely affect detailed geosteering applications, among others, as the faults as represented in the model may not capture important small structural details. Geosteering is a technique, primarily used in horizontal wells, where logging while drilling (LWD) or measurement while drilling (MWD) are used in real time to keep a directional wellbore within a particular formation or reservoir zone. Details such as proximity to faults thus become important, as it may be desired, for example, to keep a wellbore on one side of a fault. If the fault has been moved due to the restrictions of the modelling system, the wellbore could cross the fault and miss the target. Defining fault relationships in a pillar-based method may be done by drawing a sketch that
Fig. 3. Pillars are straight lines in a 3D reservoir grid. Pillars tie together cells from the top to the base of the grid and exist not only along a fault surface but also internal to the grid.
Fig. 4. Pillars along a fault surface. Each pillar has two nodes, one at the top and base of each pillar.
shows the truncation of one fault against another. In some respects, this sketch is similar to the 2D polygons or lines that are used to represent faults on a map, although the network is not required to be associated with a particular horizon. The sketch consists of a series of lines, one for each fault; faults that intersect or cross must share a common node (Fig. 5). Pillars extend from these nodes, and where the faults intersect or cross it is not only the node which must be shared, but also the pillar. The pillar method has several advantages. First, for a small number of faults, it is a very quick and easy method. It is graphical and interactive, and the relationships between the faults are easy to see and clearly defined. Second, the lateral extent of the faults can be changed quickly. It is easy to trim an unwanted portion of a fault (to make it truncate against another fault, for example) by deleting a node. It is also easy to extend a fault to intersect another by moving or adding a node. Some implementations of this method use nodes on 3D polygons to control the intersection and truncation of the faults. Third, because of the interactivity, the geoscientist or modeller can easily alter or
FUSED FAULT BLOCK NETWORK MODELLING
79
Fig. 5. A two-dimensional sketch of a fault network. Faults are represented by different line types. Individual nodes are squares; shared nodes are triangles.
change the fault relationships based on soft knowledge without having to create or add any actual data. However, the pillar method also has disadvantages. First, the relationships between the faults are mostly defined in a two-dimensional sense, much like a structural map on one horizon. Once a truncation has been made, it cannot change lower or higher in the section and therefore fault relationships cannot change either in depth or along the length of a fault. Second, the shared node and pillar at the fault intersections can limit the types of intersections that can be modelled. A very low angle intersection, for example, can distort the shape of the shared fault pillar and subsequently the shape of the faults themselves. The third problem is that some types of intersections are treated as special cases. For example a Y-intersection (commonly an antithetic fault truncating against a synthetic fault) requires special construction of the nodes and pillars. The simple example shown in Figure 6 illustrates this problem. This is a simple synthetic/antithetic fault pair, where the darker antithetic fault truncates against the larger synthetic fault. The truncation of the antithetic fault against the synthetic is defined by sharing nodes between the two faults (circles). This requires that the antithetic fault is no longer than the synthetic. The intersections become much more complicated when there are multiple, nested Y-intersections. Lastly, pillars are commonly required to extend from the top to the base of the reservoir and cannot terminate within the reservoir. Some pillar-based methods have attempted to solve this problem by introducing two types of pillars (de
Fig. 6. A pair of synthetic and antithetic faults. Network nodes for the antithetic fault are cubes; shared nodes are circles; and nodes for the synthetic fault only are cylinders.
Jager & Pols 2006) but this adds yet another layer of complexity to the modelling. The ‘solution’ to some of these problems, such as the low angle intersections and Y-intersections, is to move the faults so that the intersections become higher angle or so that the faults do not actually intersect one another. The resulting model is no longer a correct representation of the subsurface, which can impact hydrocarbon volume and recovery calculations, well placement and reservoir simulation. Manual edits of grid cell properties may be necessary, for example, to insert transmissibility multipliers into cells which should have been faulted but are not in the final simulation grid. Calculations that require knowledge of fault information, such as fault seal analysis, are also impacted. The pillar method is not well-suited for models containing hundreds of faults, as the shared pillars cause significant distortion of the surfaces and intersections, and the compromises made by moving the faults have too great an impact on the accuracy of the model.
Binary tree method The term ‘binary tree’ refers to a piecewise approach to identifying fault relationships. In this
80
K. S. HOFFMAN & J. W. NEAVE
method, a fault divides a volume into hanging-wall and footwall blocks. Faults are added one by one, and each fault must be placed into one of the preceding blocks (hence ‘binary’ as there are only two choices for the placement of the new fault). The new fault forms its own hanging-wall and footwall blocks, which together make up the previous block. This hierarchy looks somewhat like a diagram of a tree, with branches spreading out as more and more faults are added (Fig. 7). Each fault in a model has an explicit relationship to any other fault in the model and the placement in the hierarchy defines the allowable truncations. In the example in Figure 7, Fault 1 is the primary fault; it may truncate any other fault in the model but cannot itself be truncated by any fault. Fault 2 can only be truncated by Fault 1; Fault 4 may be truncated by Fault 2 or Fault 1. The various levels in the hierarchy can, in some ways, be related to the ages of the faults, with the faults at the top being younger, although this is not necessarily a strict age relationship. Manually defining this type of hierarchy can be a difficult task when there are hundreds of faults. Methods of determining the relationships automatically, usually based on the distribution of input data, have been developed, but these solutions must always be checked for erroneous truncations.
The primary advantage of this method is that the intersections are not controlled by pillars: therefore the shape of each fault is independent of the intersecting faults, and intersections such as Y-intersections are no longer a special case. The binary tree method is better suited to models containing hundreds of faults than the pillar method, although even here there are practical limitations to the number of faults that may be incorporated into a model. There are some disadvantages to an explicit binary tree as well. First, the fault relationships are not defined in an interactive, graphical way. Often the relationships are based on a table or a treelike diagram, where editing the relationships is quite difficult. Second, crossing faults are treated as a special case. Because a fault is defined as existing only in the hanging wall or footwall of a previous fault, the crossing fault must be specifically added to both blocks. Potentially, the crossing fault might have to be added in multiple places and it can be difficult to find all of the blocks within a complex tree where the crossing fault needs to be specified. Even methods that build trees automatically cannot always place a crossing fault in all of the correct blocks, especially when there are few data points. Third, non-intersecting faults may interfere with one another. Binary tree methods are generally fault
Fig. 7. Diagram of an explicit binary tree. Each fault subdivides a volume into hanging wall and footwall blocks. Subsequent faults are placed into one of these blocks. A fault may be truncated by any fault preceding it in the tree.
FUSED FAULT BLOCK NETWORK MODELLING
block based methods; that is the faults create specific fault block volumes bounded by the fault surfaces. Even where faults do not intersect one another, the extrapolated fault surface is used to define the boundaries of the fault block. Fault blocks are extremely useful entities for visualization, volume calculations and other purposes, and the extension of the fault surfaces is a useful and intelligent shortcut to creating the fault block boundaries. The drawback is that a strict binary approach always requires one fault at the beginning of the tree. If the ‘wrong’ fault is chosen as the primary fault in a tree, it could erroneously truncate faults further down in the tree, as shown in Figure 8. Here, if the dashed fault is chosen as the primary fault, the extension of this surface to the west of the solid fault could truncate a valid portion of the solid fault. In a simple model, this problem is easy to diagnose and solve, usually by choosing a different fault as the primary fault. However, in cases where there are hundreds of smaller, subparallel faults with no one fault being an obvious candidate as the primary fault, the problem can be persistent, difficult to find, and only solvable by introducing crossing relationships. The fourth problem is similar: self-truncating faults are impossible to model. The primary fault in a binary tree is that against which all other faults might truncate. The primary fault itself cannot be truncated. Therefore, in a case of self-truncating faults, there will always be part of the primary fault which is incorrectly included in the model. The incorrect overlap can be minimized, but the fact that it
Fig. 8. Non-intersecting faults may erroneously truncate one another when using an explicit binary tree. The dashed fault should actually stop short of the intersection with the solid fault; but if the dashed fault is selected as the primary fault, its extension (used to create fault blocks and shown as a dotted line) may truncate the solid fault.
81
exists at all means that additional blocks are created and that subsequent faults might be incorrectly truncated. This Escher-like problem is illustrated in Figure 9. Here, the dotted parts of each of the faults should be truncated. None of these faults can be chosen as the primary, non-truncated fault. Although this triangular problem is quite simplistic, this type of situation can often arise in areas where there are a number of subparallel or en echelon faults. Finally, relationships between faults cannot change along the length of the fault. It is difficult, if not impossible, to create a model where an antithetic fault truncates against a synthetic fault in one part of the model, but where the relationships are reversed in another part of the model.
Methods used in fused fault block approach The fused fault block method is a technique that preserves the advantages of the existing techniques while eliminating (as much as possible) the problems associated with them. This method is based on using parametric surfaces to represent the faults. It also uses an implicit binary tree to define fault relationships. An implicit binary tree differs from the explicit tree described above in that the relationships between the faults are not explicitly defined in a hierarchical diagram or table. The name of the method ‘fused fault block’ refers to the process of joining, or fusing, fault blocks together where appropriate. Two crossing faults, for example, would create four volumes, or fault blocks. If a section of one of the faults is
Fig. 9. Map view of a series of self-truncating faults. Each of these three faults should truncate on one of the other two and in turn be truncated by the remaining fault, a situation which is impossible to model in an explicit binary tree.
82
K. S. HOFFMAN & J. W. NEAVE
removed, there are only three volumes remaining; two of the original blocks have been merged into one (Fig. 10). Although the concept of ‘fusing’ fault blocks in and of itself is not new technology, the implementation of the method in fault and horizon modelling uses a unique application.
The implicit tree has two major advantages over the explicit tree method. First, faults that do not intersect each other do not need to have any relationship defined (Fig. 11). Non-intersecting faults therefore cannot interfere with one another and cause incorrect truncations. A series
Fig. 10. Derivation of the name of the fused fault block method. (a) Two crossing faults, which create the four fault blocks in (b). (c) One fault has now been truncated against the other; the number of fault blocks has now been reduced to three (d).
FUSED FAULT BLOCK NETWORK MODELLING
Fig. 11. Two non-intersecting faults. In the fused fault block method, no relationship needs to be defined between these two faults. In an explicit binary tree, one would need to be selected as the primary fault.
of relatively short en echelon faults thus becomes much easier to model as the interpreter is not faced with having to select one of the faults as the primary fault in an explicit tree. The fused fault block method also more fully defines sections of intersecting faults than an explicit tree does, providing more flexibility in defining fault truncations (Fig. 12). Fault truncation specifications for the pillar method are based on eliminating nodes of a fault network line; the binary tree method defines a fault as existing in only one block of a previous fault. Thus each of these methods removes all of a fault on one side or the other of the truncating fault. In the fused fault block method, intersection lines split a fault into areas that are fully defined
83
by the intersecting faults. That is, instead of an area of a fault having been defined as existing only in the footwall or hanging wall of the intersecting fault, an area may be defined by its compound relationship to several faults. In the example shown in Figure 12, the yellow-coloured fault in the middle of the model is crossed by three other faults. These intersections define four separate, independent areas of the light-coloured fault. Any one of these four areas may be selected for removal, thus allowing any type of truncation between the faults. The example shown in Figure 12, with the middle section of a fault removed, is no doubt a geologically implausible model. However, the ability to remove any particular section of a fault provides flexibility in defining fault truncations, and has applications to various complex truncations, such as the L-faults discussed below. This technique provides a solution to many of the intersection and truncation problems that exist with the pillar or explicit binary tree methods.
Examples The fused fault block method is not simply a technique for fault modelling. The method is an integral part of horizon modelling. The horizon modelling process uses a fault block based approach to create horizon surfaces from input data. Some modelling methodologies go directly to a 3D grid after creating the fault model (Chambers et al. 1999), but this approach requires making decisions about the grid geometry very early in the modelling process. Different aspects of modelling (facies modelling, rock attribute modelling, or flow simulation, for example) may require different grid
Fig. 12. (a) The yellow fault in the middle of the model is crossed by the three darker faults, creating four sections of the fault surface (A, B, C, D). (b) One of the middle sections has been removed from the fault model. Although removing this particular segment is not geologically reasonable, it illustrates the flexibility of the fault truncation process.
84
K. S. HOFFMAN & J. W. NEAVE
geometries. In a direct-to-grid approach, the entire horizon model must be rebuilt for each purpose. A second method is to generate parametric surfaces of the horizons. This method can have two variations. One variation creates a single 2D surface for each horizon. Where onlap or truncation due to an unconformity occurs, the surfaces become coincident. This can give rise to problems in the 3D reservoir grid, as grid cells may then become extremely thin or collapse near the truncations. In addition, these surfaces, being single 2D grids, cannot model repeated section. With a fault block based approach, such as that used in our method, the horizons are represented by a patchwork of smaller parametric surfaces. This allows easy modelling of repeated section. The information contained in the fault truncations is used to ensure that the individual parametric surfaces are completely continuous at fault terminations. Our method also does not require that surfaces be coincident in areas of onlap or erosion; the surfaces are instead truncated, which creates a 3D grid without collapsed cells. The fault block based, layered structural model thus provides the basis for creating a reservoir grid. The resultant grid represents the structure more accurately, as shortcuts and compromises in fault intersections and truncations do not have to be made as they often are in other fault modelling methods.
Y-faults The fused fault block method provides a technique to build models of multiple nested Y-faults quickly (Fig. 13a). In this example there are two antithetic faults which truncate against the main synthetic fault, and a smaller synthetic fault that truncates
against one of the antithetic faults. An explicit binary tree method would be able to model these faults, but the pillar method would not be able to include the small synthetic fault, as it creates a Y-on-Y fault situation. In this case, one of the faults would have to be removed from the model, or the surfaces changed so that the faults no longer intersect. Both of these would change the basic structure of the model, and therefore would also change the volumes of the fault blocks. These changes do not have to be made using the fused fault block method, as Y faults are not special cases and have no restrictions on the number of Y intersections. When creating a reservoir grid from a Y-fault model, the faults may be treated as linear pillar faults, or the faults may be regularized as stair-step faults. With Y-fault geometries, it is advisable to treat at least one of the faults as a stair-step fault in order to avoid the problems of collapsed or twisted cells at the fault intersection (Fig. 13b). In this example, the synthetic faults have been treated as stair-step faults and the antithetic faults as pillar faults. All faults could as easily have been treated as stair-step faults if desired.
l faults l faults are just Y-faults turned upside-down, but they are generally treated as even more of a special case than Y-faults when using the pillar approach to fault modelling. This type of fault geometry can also pose problems during horizon modelling, as data for the uppermost horizon in the triangular section of the l will not exist. This type of intersection is not considered to be a special case using the fused fault block method. l faults
Fig. 13. Y-fault model. (a) Fault surfaces showing truncation of two antithetic faults against a main synthetic fault, and a smaller synthetic fault truncating against an antithetic fault. (b) Reservoir grid created from the Y-fault model in (a). Colours represent fault blocks. The synthetic faults have been treated as stair-step faults and the antithetic faults as pillar faults.
FUSED FAULT BLOCK NETWORK MODELLING
85
Fig. 14. l fault model. (a) The younger, dark fault in the centre of the model has offset two older faults. The lighter coloured faults on the right side of the model are Y faults and the darker sections at the left are the l faults. (b) Reservoir grid created from fault model in (a). Colours represent fault blocks. The younger fault is a pillar fault and the remaining faults are stair-step.
may often be created where a younger fault has offset a series of older faults (Fig. 14a). These faults could be modelled correctly using an explicit binary tree method, but pillar methods would need to modify the fault surfaces in order to eliminate some of the intersections. The fused fault block method does not require any modifications of the fault surfaces or intersections, and has no limits as to the number of l faults that may be included in a model. As with Y-faults, l faults may be treated as either pillar faults or stair-step faults in the reservoir grid, although to avoid collapsed or twisted cells in the triangular areas, it is advisable to treat some of the faults as stair-step (Fig. 14b). In this example, the younger fault is a pillar fault, and the Y and l faults are stair-step. Because of the dip of the younger fault, the edges of the cells along this fault may pose a problem in flow simulation and the performance might be improved by treating this fault as stair-step as well. Our method allows all possible combinations of fault intersections in the 3D grid: stair-step on stair-step, stairstep on pillar, pillar on stair-step, and pillar on pillar, although the last may not be desirable in certain fault configurations.
Self-truncating faults One of the geometries that is possible in a pillar fault method, but impossible in an explicit binary tree method is a series of self-truncating faults. In an explicit binary tree, one fault must always be selected as the primary fault; this primary fault cannot be truncated by any other fault in the
model. A situation such as that shown in Figure 15, where each of the three faults truncates and is truncated by another fault, would therefore not be possible. The fault geometries would need to be simplified so that one of the intersections is removed, or a small extension beyond the intersection would have to be allowed. Both of these situations could negatively impact the reservoir model, either in terms of fault block volumes or in transmissibility across the faults. The fused fault block method does not require that any fault be selected as the primary. A fault is divided into areas based on the intersections with other faults and these areas may then be manipulated independently of one another. This approach makes it possible to model a series of self-truncating faults correctly. Although this example might appear to be synthetic and unrealistic, crossing conjugate faults have been observed and documented in many locations, and their process of formation has been described by Ferrill et al. (2000).
Changing relationships Faults that change truncation relationships along their length are perhaps some of the most difficult geometries to model correctly. Neither the pillar method nor the explicit binary tree can handle this situation without splitting each fault into at least two separate faults. In an L-type intersection, the truncation relationships between synthetic and antithetic faults can change along the length of the faults (Fig. 16). In one part of the model, the synthetic fault truncates against the antithetic, and in
86
K. S. HOFFMAN & J. W. NEAVE
Fig. 15. A series of self-truncating faults. (a) Faults prior to truncation. Each fault crosses the other two faults. (b) After truncation. The light-coloured fault at the north end of the model truncates against the dark fault at the east. The dark fault truncates against the medium-coloured fault, which in turn truncates against the light-coloured fault.
another part the opposite relationship is true. In addition, the faults do not exist over the same XY area; the truncated section of the fault is shorter than the truncating section of the other fault. Pillar methods cannot handle such a situation, because one and only one of the faults can be specified as the Y fault, and in this case each fault makes a Y against the other. In addition, a Y fault cannot extend beyond the truncating fault in a pillar-based model, as there must be a common set of pillars in the Y area. An explicit binary tree also has difficulty with this geometry, as a clean truncation between faults can only be accomplished by placing a fault in one of the two blocks created by its predecessor in the tree. Allowing the second fault to cross does not create a clean truncation, and putting a fault into the tree twice can lead to problems with coincident surfaces or sections of faults that do not exactly match. However, this complex situation can be modelled quite easily using the fused fault block technique (Fig. 16). This type of geometry can only be modelled correctly and easily when using a method that fully defines the compound areas of faults and allows them to be manipulated independently.
Conclusions The fused fault block technique for fault network modelling provides a solution to the problems that are inherent in existing methods of fault network modelling. The limitations as to the types of fault intersections that are possible to model have been removed. The fused fault block method can model Y and l faults, which are difficult with the pillar method; it can model self-truncating faults, which are difficult with the explicit binary tree method; and it can model L-shaped intersections, which are difficult or impossible in either the pillar or the explicit binary tree methods. It is now possible to create fault networks of these complex truncations and also to take these networks through to reservoir gridding without having to simplify or alter the fault relationships. Some limitations on reservoir gridding may still remain. With this method, there are no longer any artificial constraints on the simulation grid due to modelling restrictions, but there may be limits to the number of cells that are acceptable for flow simulation. Although advances are continually being made both in hardware and software, there
FUSED FAULT BLOCK NETWORK MODELLING
87
Fig. 16. L-shaped intersections. (a) Model viewed from the south, showing the truncation of the light fault against the dark fault and the extension of the dark fault beyond the light fault. (b) Model viewed from the north, showing the truncation of the dark fault against the light fault and the extension of the light fault beyond the dark fault.
are practical limits to the size of a simulation grid. This upscaling is a general problem that applies to almost all simulation grids and is not specific to any one gridding method. The simplicity of building and editing the fault relationships means that models can be updated quickly with new data or even new faults without losing the previously defined fault relationships. In addition, it is possible to test and store different scenarios of fault relationships. The speed of this process places the emphasis of the fault network building on the evaluation of the interpretation and the effects of compartmentalization, and not on manipulation of software. Preserving the true geological structure in the fault network produces more accurate reserve calculations, provides a less risky model for well planning and improves the results used in reservoir simulation. The authors wish to acknowledge the contribution of other members of the team, E. H. Nilsen and E. Sverdrup and would like to thank Roxar AS for permission to publish.
References A KIMA , H. 1970. A new method of interpolation and smooth curve fitting based on local procedures. Journal of the Association for Computing Machinery, 17, 589–602.
B ELCHER , R. C. 1994. Geospatial modelling techniques for understanding internal geometries of complexly faulted reservoirs. Society of Petroleum Engineers Paper, 27543. C HAMBERS , K. T., D E B AUN , D. R., D URLOFSKY , L. J. ET AL . 1999. Geologic modeling, upscaling, and simulation of faulted reservoirs using faulted stratigraphic grids. Society of Petroleum Engineers Paper, 51889. C HARLES , S., D ENY , L., B OMBARDE , S. & M ALLET , J. L. 1995. Toward an interface between triangulated surfaces and parametric representations. Society of Exploration Geophysicists Expanded Abstracts, 14, 1251– 1254. D E J AGER , G. & P OLS , R. J. W. 2006. A fresh look at integrated modelling software. First Break, 24, 75–81. F ERRILL , D. A., M ORRIS , A. P., S TAMATAKOS , J. A. & S IMS , D. W. 2000. Crossing conjugate normal faults. American Association of Petroleum Geologists Bulletin, 84, 1543–1559. G OURAUD , H. 1971. Continuous shading of curved surfaces. IEEE Transactions on Computers, 20, 623–628. M ALLET , J. L. 1992. Discrete smooth interpolation in geometric modelling. Computer Aided Design Journal, 24, 178– 191. S EGONDS , D., B ENNIS , C. & M ALLET , J. L. 1998. 3-D structural modelling: a new approach to interactively modify complex surfaces. Society of Exploration Geophysicists Expanded Abstracts, 17, 707 –709. Z ORASTER , S. & B AYER , S. 1993. Three-dimensional fault modelling from cross-sectional data. American Association of Petroleum Geologists Bulletin, 77, 207.
Editing faults within tetrahedral volume models in real time A. L. TERTOIS & J. L. MALLET Gocad Research Group, ASGA ENSG-INPL/CRPG-CNRS, Rue du Doyen Marcel Roubault - BP 40, 54501 Vandoeuvre-les-Nancy, France (e-mail:
[email protected]) Abstract: Accurately positioning faults in a geological model is a major concern because they are responsible for offsets of geological sequences. In the tetrahedral models studied in this paper, faults are discontinuities: faces of tetrahedra on either side of a fault are disconnected. Building tetrahedral models can require a large amount of time, especially when there are many faults. We present a tool for making small, real-time, modifications of faults in tetrahedral models arising from geometrical changes required either by new subsurface data or by new interpretations of existing subsurface data. Fault editing is achieved by moving control points on the fault in the tetrahedral grid and by computing a distortion property over an editable volume relative to the control point and spreading this distortion to neighbouring points using the Discrete Smooth Interpolation technique. The editable volume in which tetrahedron vertices are allowed to move is defined by a given distance to the fault. This approach provides a means of editing faults and fault-related features, such as branch-lines.
Grid building remains one of the main challenges of reservoir modelling. Geometric accuracy must sometimes be overlooked in favour of simple grids, which require fewer computing resources. Generally, reservoirs are modelled as curvilinear grids – Cartesian grids in which the layers of cells are deformed to match horizon and fault geometry. These grids offer a fair compromise between precise geometry and easy computation for property simulation, fluid flow and velocity modelling. However, in some instances, reservoirs are so complex that a curvilinear grid does not reflect the geometry of layers and faults accurately, and sometimes cannot even be constructed without making unacceptable approximations. One option for these complex geometries is to use unstructured and irregular meshes, such as polygonal or tetrahedral grids (Lepage 2002; Pre´vost et al. 2004). The resolution of the mesh can be adapted to the structures to be represented: the mesh is coarse in simple regions and denser where the structures or heterogeneity are more complex. Geological properties are also affected by faults. Some properties, usually related to the geometry of the model, for example distance from a well, may be continuous from one side of the fault to the other. In this case, values on both sides of the fault are identical. Other properties, generally related to rock type, such as porosity, may be continuous before displacement by the fault, but may not be continuous after fault displacement. A method involving vectorial links (Moyen 2005) restores the continuity of the property using fault throw (Fig. 1). Points that were at the same location before faults offset the geological sequences are linked so that property values are equal on both sides of faults. In the first
part of this paper, tetrahedral grid building and volume distortion methods are reviewed. Based on the concepts defined in these methods, a fault editing algorithm is presented in the second part. Finally, examples of consistent geological model editing are explained and illustrated.
Building and editing tetrahedral grids Geological applications based on tetrahedral grids Because their geometry is flexible, tetrahedral meshes have several applications in geology. Tetrahedra can be made to fit the geometry of geological structures and as they are the simplest possible simplexes in three dimensions, tetrahedra are interesting to perform computations on. Velocity modelling. Velocity models play a key role in processing seismic data, especially in preor post-stack time and depth migration, as well as in time-to-depth conversion (Yilmaz 1987). Quantitative interpretation of seismic data uses elastic properties derived from velocities. Comparing the results of velocity modelling with real seismic data for a study area can highlight defects in the velocity model or in the layout and geometry of geological structures. Macy & Smith (1998) use a tetrahedral model in which several regions with different geological attributes are separated by velocity discontinuities. The ray path is calculated in each intersected tetrahedron in turn, using the gradient of the velocity function. Velten (1998) defines tetrahedron columns inside the volume model and
From: JOLLEY , S. J., BARR , D., WALSH , J. J. & KNIPE , R. J. (eds) Structurally Complex Reservoirs. Geological Society, London, Special Publications, 292, 89– 101. DOI: 10.1144/SP292.5 0305-8719/07/$15.00 # The Geological Society of London 2007.
90
A. L. TERTOIS & J. L. MALLET
(a)
(b)
Fig. 1. Continuity of properties across a fault. Both sides of the fault are shown apart. Squares in shades of grey represent different values of a geological property. (a) Continuity of a property across a fault. Points that have the same geometrical coordinates also have a common property value. (b) Continuity of a property using fault throw. Points that have the same geometrical coordinates do not have the same property value: for instance point H would have a value between white and light grey. Continuity using fault throw is ensured by vectorial links (dashed grey lines).
computes ray paths using several options for the velocity law inside columns. Three-dimensional restoration. Three-dimensional structural restoration provides strain fields that help validate structural interpretations. Muron developed a restoration method for tetrahedral models based on volume conservation and strain minimization, taking into account mechanical rock properties (Muron 2005). Most other volume restoration methods are not fully three-dimensional: stacks of horizons linked by geological constraints are restored simultaneously (Griffiths et al. 2002). The strain fields generated by three-dimensional restoration can be used as input for fracture simulation in tetrahedral models (Mace´ et al. 2005, 2004). These fractures are modelled using a combination of stochastic methods and geological rules (Cacas et al. 1990; Josnin et al. 2002). Fluid flow modelling. Flow simulation is a key step in understanding the dynamic of an oil or water reservoir. Verma (1996) discusses flow simulation in reservoirs using several types of grids, including Voronoi grids. A Voronoi grid is the dual mesh of a tetrahedral grid. It is obtained by replacing tetrahedra with their barycentres, and linking these points together (Okabe et al. 2000). Voronoi grids are used to model fluid flow in reservoirs using the
finite elements or finite volume methods (Palagi et al. 1993). Two and three-dimensional unstructured cells can be adapted to reservoir heterogeneity (Pre´vost et al. 2004). Streamline-based simulations are an alternative to finite elements and finite volume methods (Matringe et al. 2006) and help assess the accuracy of upscaling in these grids and provide an insight into flow behaviour (Baker et al. 2001; Blunt et al. 1996). Geochron framework. Geometry of geological structures and property modelling can be separated using the Geochron framework (Mallet 2002b, 2004; Mallet et al. 2004). A three-dimensional parameterization with geological considerations such as fault– fault and fault–horizon contacts and sedimentary discontinuity surfaces is computed on a tetrahedral model of the study area. This parameterization associates each node in the tetrahedral grid to its position in the depositional space. If no fault was active during deposition the layers are perfectly flat in the depositional space, so geological properties can be modelled in this space using a fine Cartesian grid. Computing petrophysical properties such as porosity in the depositional space reduces errors introduced by gridding limitations. The parameterization enables mapping high resolution geological properties in the depositional space onto the coarser tetrahedral volume.
FAULT EDITING IN TETRAHEDRAL MODELS
Creating tetrahedral grids There are three main groups of tetrahedral grid creation methods (Owen 1998): octree-based, Delaunay-based and advancing front methods. The most popular meshing methods for geological applications are Delaunay-based. First, interpreters create fault and horizon surfaces from seismic and well data and from geological and regional knowledge (Fig. 2a). Faults represent the structural information in the study area, and partition the subsurface in fault blocks (Fig. 2b). If required, horizon surfaces can be integrated in the model as implicit surfaces. Then, a tetrahedral mesh of the model may be generated using a Delaunay criterion (Lepage 2003). As shown in Figure 2c, faults do not have to extend through the whole tetrahedral volume. Because they are unstructured, tetrahedral volumes can represent complex geometries incorporating, for example, discontinuous faults and fault branching. Rock properties can be stored and visualized directly on the tetrahedral volume or using the Geochron framework (Mallet 2002b, 2004; Mallet et al. 2004), a three-dimensional texture mapping enables
91
visualization of properties stored in a regular grid (Fig. 2d). Tetrahedral models in geology can be divided into different fault blocks. Inside a fault block, tetrahedra are all connected, and each tetrahedron shares vertices and faces with its neighbours. A fault is modelled as a topological discontinuity in the volume (Fig. 3a). Tetrahedra on both sides of a fault do not share faces and vertices are duplicated, although the geometrical position of the nodes is the same on both sides. To make property modelling easier on a tetrahedral volume, faces of tetrahedra that are on faults have the same geometry as the triangulated surfaces from which the model was created. In other words, on a fault, three entities have exactly the same shape and location: the triangle of the fault surface and the faces of tetrahedra on both sides of the fault (Fig. 3b).
Model updating New information obtained for a geological volume can make a model obsolete. For instance, a new interpretation of seismic data can show that the
Fig. 2. (a) Seismic and well data with surrounding box. (b) Structural model, with fault network. (c)Tetrahedral volume built from structural model. Faults highlighted in black. (d) 3D texture mapping of porosity simulation on tetrahedral volume. Data courtesy of Schlumberger.
92
A. L. TERTOIS & J. L. MALLET
(a)
(b)
Fig. 3. (a) Fault in a tetrahedral volume. Edges of tetrahedra (grey lines) are constrained by the fault (bold black line). Three geological layers are depicted in this model using a red, white and a green colour. Tetrahedra are not constrained by geological layers: tetrahedron edges can cross these interfaces. (b) Schematic representation of data structures on a fault. Triangle ABC is a triangle of a fault surface. Two tetrahedron faces are associated to this fault triangle: face DEF in tetrahedron DEFG and face HJK in tetrahedron HJKL. Point triplets (A, D, H), (B, E, J) and (C, F, K) have the same coordinates.
geometry of a surface is not accurate, or a new well can provide another data point for the location of a surface. Faults are discontinuities in the grid, hence fault editing requires modification of the geometry and possibly the topology of the tetrahedral mesh. The current way of dealing with this problem is to edit the structural model and build a new tetrahedral volume. However, obtaining a high-quality mesh that conforms to geological structures requires time and effort. Local editing techniques on geological models provide a way of making small adjustments to models without having to go through the whole meshing process. Volume distortion is a very active research field, especially in computer graphics and animation. Freeform deformation (FFD) incorporates a number of techniques for modifying the shapes of objects by the manipulation of control nodes or curves. Barr (1984) introduced a combination of simple objectediting transformations. FFD was made easier to use by Sederberg & Parry (1986), Coquillart (1990) and Chang & Rockwood (1994), who used different kinds of control lattices or curves in order to distort objects. But these methods did not take into account the nature of objects being edited. Hirota et al. (1999) added physical laws to FFD in order to preserve the total volume of objects during geometrical distortion. Borrel & Rappoport (1994) introduced simple constrained deformation (SCODEF). Constraint points possessing a radius of influence and desired distortion determine local B-spline basis functions. The entire space, along with the objects in it, is then distorted according to a linear combination of these
functions, creating bumps in space. Real-time geometrical distortion is possible with this method. Grosse (2002) adapted FFD and SCODEF to geological problems and described a number of solutions for interactive editing of geological models. Caumon et al. (2004) developed tools for editing a geological model in real-time whilst retaining other essential constraints, such as contacts between horizons and fault surfaces. One of the problems faced when modifying the geometry of a set of vertices in a tetrahedral volume is that tetrahedra may flip over. When that occurs, the mesh is no longer valid, because the sign of the algebraic volume of inverted tetrahedra has changed, and because some edges may intersect faces of these tetrahedra. Figure 4 shows an example of tetrahedron inversion where one point is moved through the opposing face of the tetrahedron. Shontz & Vavasis (2003) had to tackle this problem when they modelled the beating of a canine heart using a tetrahedral mesh. They applied geometrical distortion to the triangulated surface bounding the tetrahedral mesh. Then the points inside the volume were moved using linear weighted Laplacian smoothing. A constraint on the function used to deform the boundary of the volume ensured that the mesh did not flip over. They obtained satisfactory results for the canine heart, but the linear weighted Laplacian smoothing could not prevent inversion when the geometrical distortion was large. In conclusion, some of these volume editing methods can be adapted with varying degrees of success to geological models. In FFD methods, control curves are distorted and the object being
FAULT EDITING IN TETRAHEDRAL MODELS
(a)
(b)
93
(c)
Fig. 4. Inversion of a tetrahedron. Only point A is moved, points B, C and D are fixed. (a) Tetrahedron before inversion. Point A is moved toward face BCD. Edge AD is hidden. The volume of tetrahedron ABCD is V(ABCD) ¼ (1/ 6)(AC AD). AB, and is here positive. (b) Edges AB, AC and AD are hidden. Point A is moved through face BCD. (c) Edge CD is hidden. V(ABCD) is now negative: the tetrahedron has inverted.
edited is updated accordingly. SCODEF is an interesting method but it creates bumps centred on constrained points, whilst editing fault geometry requires smooth geometrical distortion in the control volume. Our method tries to solve these problems and to provide an intuitive, real-time distortion method adapted to geological problems.
Fault-editing method The fault-editing tool presented here permits small adjustments to fault geometry in real-time. An editable volume around the fault is defined and it is distorted by picking and dragging arbitrary control points. The geometrical distortion of the tetrahedral mesh is instantly visible and editing can be stopped when the new geometry is acceptable. Although the fault-editing tool was developed as a plug-in to the GOCAD software, similar approaches could, in principle, underpin methods implemented in other software systems.
Fault editing Before modifying fault geometries in a tetrahedral model, user-defined control points must be created on the fault which is to be edited and the volume in which the tetrahedron nodes will be allowed to move must be selected. In order to define this volume, a property is computed from the distance from the fault and stored at the nodes of the tetrahedral mesh. This distance property can be adapted to provide different shapes of control volumes. If only the distance from the fault is used, the control volume has boundaries that are parallel to the fault. Using the distance from a point at the centre of the fault will provide a spherical control volume. Any combination of these distance functions can be computed, thus varying the shape of
the control volume. A user-defined threshold value for the distance property defines the editable volume: tetrahedra with property values below this threshold are inside and tetrahedra with property values above the threshold are outside the editable volume and are therefore fixed. A low value enables small changes on a part of the fault, whereas a higher value enables general reshaping of the fault. If the threshold is too low, there will not be enough tetrahedra inside the editable volume to enable fault editing. In order to make the volume selection step easier, visualization tools can also be used to display the distance property on surfaces within a model (e.g. Frank 2006). As a control point is edited, by user-defined dragging inside the control volume, a distortion vector is computed between the starting position of the control point and its current position. This vector V provides distortion direction and magnitude, with the direction W ¼ (1/jVj). V stored for later use and the magnitude computed as a onedimensional property on the control volume. This distortion property is interpolated from the value set by the control point to zero on the borders of the editable volume using the Discrete Smooth Interpolation technique (see below; Mallet 1992, 2002a). This means the distortion varies smoothly from a maximum value at the control point to zero on the borders of the editable volume. Tetrahedra outside the editable volume are not distorted, whereas the vertices of the tetrahedral mesh that are inside the editable volume are moved along W, using the distortion property computed for each vertex. If required, W can be modified so that each vertex is moved in a geologically consistent way. This is of importance when a Geochron parameterization is available: horizons in the tetrahedral mesh are represented by a pseudo-time function. If this function is distorted, so is the geometry of the
94
A. L. TERTOIS & J. L. MALLET
horizons. In order to solve this problem, when a Geochron parameterization is provided, the component of vector W that is normal to the horizons is computed for each vertex and removed. This ensures that the vertices are moved along constant values of the pseudo-time function, thus preserving the geometry of the horizons. Vertices must also be moved in a geometrically consistent way, otherwise tetrahedra could invert. In order to prevent such inversions in the tetrahedral mesh, checks are performed at each step of the distortion. If there are no mesh inversions, editing can proceed and the geometry is updated. If the mesh does invert, the distortion is computed from the distortion vector stored in the previous step and the tool is terminated. Editing is carried out by moving control points around in the selected control volume. When a control point is moved, the editing tool enters a loop in which the distortion property is computed over the control volume and the geometry is updated: these editing steps are shown in Figure 5.
used as a linear solver to compute the distortion function. The tetrahedral mesh provides a finite set of interconnected nodes constituting a linear model. The DSI method minimizes the degree of violation of soft constraints set on the linear model. A constraint is a set of equations linking the nodes in the linear model and translating an idea on how the interpolated property should vary between the nodes that are involved. For instance, one of the basic DSI constraints is the roughness constraint, which states that the interpolated function should vary smoothly from one node of the model to the other. This roughness constraint is translated as a set of coefficients linking neighbouring nodes in the linear model. The DSI problem can be expressed as solving the set of linear equations
Background on Discrete Smooth Interpolation
with Aci a vector containing the coefficients of constraint ci at each node of the linear model, here the tetrahedral mesh. w is the unknown function at the nodes. k is the number of constraints on function w. The goal is to find, for each constraint ci, the
The Discrete Smooth Interpolation (DSI) method, described in detail in Mallet (1992, 2002a), is
Control Point dragged
Atc1 w bc1
(1)
.. . Atck w bck
New Direction Vector
Previous Direction Vector
Update of Constraints on Displacement
Update of Constraints on Displacement
Displacement Interpolation
Displacement Interpolation
No Inversion
Inversion
Geometry Update
Step Backwards
Geometry Update
Fig. 5. Editing loop. As a control point is dragged, the direction vector is computed. This vector is transcribed as constraints on the distortion function, which is then interpolated using DSI. If this distortion invalidates the mesh, the direction vector from the previous editing loop is retrieved, the distortion property is recomputed and the tool is terminated. If the distortion does not invalidate the mesh then the geometry is updated and editing can resume.
FAULT EDITING IN TETRAHEDRAL MODELS
value of w for which the linear combination Atck . w is closest to target value bci. The system of linear equations is solved using a least squares method which minimizes an error criterion computed from the values of the unknown function at the nodes of the tetrahedral mesh. This method has two main features that make it particularly useful for interpolating the distortion function. First, it has proved its robustness in many different problems because it is implemented in the Gocad software. Second, designing new constraints for the DSI method is only a matter of translating mathematical formulae into C þ þ code that can be managed by the software. The mathematical framework of DSI makes it easy to develop new geological constraints as well as geometrical constraints. In addition to the soft constraints described above, which are honoured in a least squares sense, the DSI method can also honour hard constraints such as: Atci w ¼ bci
(2)
Atcj w bcj
Interpolating distortion function The most important step in the editing loop illustrated in Figure 5 is interpolation of the distortion function. This section explains which DSI constraints are set on the distortion function for the interpolation to yield the expected result. Three constraints are set on distortion property w, defined for each node a of the tetrahedral volume V. For a constraint set on the distortion function, Equation 1 can be written as Atci w ¼
X ae V
Aci (a) w (a) bci :
(3)
A distance constraint transfers the distortion of the control points to the volume. Assuming that control point P is located inside tetrahedron T with nodes fa0, a1, a2, a3g, that fg0, g1, g2, g3g, are the barycentric coordinates of P inside T and that the distortion property value at control point P is wp, the DSI constraint associated to control point P can be written as:
A smoothness constraint ensures the gradient of w is constant over the volume, thus propagating the distortion induced by the control points to the editable volume. On the border of the editable volume, a hard constraint specifies that the distortion is zero. The distortion property is constrained to vary smoothly from the value given by the control point to zero on the border of the editable volume by the constant gradient constraint. This constraint was described in (Mallet 2003; Moyen 2005), and it is set on each pair of tetrahedra inside the editable volume having one common face. As a tetrahedral model is edited, tetrahedra must not be flattened or inverted. If that were to happen, tetrahedron edges would cross and algebraic tetrahedron volumes would change signs. A non-inversion constraint for the tetrahedra ensures that they do not invert when pushed against the border of the editable volume or against a tetrahedron pinned by another control point. The distorted tetrahedron T has vertices a0, a1, a2, a3 located at positions x(a0), x(a1), x(a2), x(a3). (vectors x are three-dimensional coordinate vectors, w is the intensity of distortion in direction W. T * is the tetrahedron after distortion). The goal of the non-inversion constraint is to prevent any point from getting on the other side of the opposite face. On a tetrahedron, this amounts to preventing the sign of the algebraic volume from inverting. After distortion, the algebraic volume V (T *) of tetrahedron T * is: 1 V(T ) ¼ ½(x(a1 ) þ w(a1 ) W 6 x(a0 ) w (a0 ) W) (x(a2 ) þ w (a2 ) W x(a0 ) w (a0 ) W) (x(a3 ) þ w (a3 ) W x(a0 ) w (a0 ) W):
(5)
If we define three vectors such that: d1 ¼ x(a1 ) x(a0 ) d2 ¼ x(a2 ) x(a0 ) d3 ¼ x(a3 ) x(a0 );
(6)
equation (5) simplifies to:
Atc w ffi bc : Ac (ai ) ¼ gi if i [ f0; 1; 2; 3g Ac (ai ) ¼ 0 if i f0; 1; 2; 3g bc ; ¼ wp
95
(4)
6V(T ) ¼ ½(d1 þ (w (a1 ) w (a0 ))W) (d2 þ (w (a2 ) w (a0 ))W) ½d3 þ (w (a3 ) w (a0 ))W :
(7)
96
A. L. TERTOIS & J. L. MALLET
Noting that d1 d2 d3 ¼ 6V(T ) and keeping only first order terms yields: 6V(T ) 6V(T ) þ (w (a2 ) w (a0 )) (d3 d1 ) W þ (w (a1 ) w (a0 )) (d2 d3 ) W þ (w (a3 ) w (a0 )) (d1 d2 ) W: (8) Our goal is to ensure that the sign of V (T *) is the same as the sign of V (T ). One way of achieving this is to constrain w so that: V(T ) V(T ) 0:
(9)
Using equation (8): 0 6V 2 (T ) þ (w (a2 ) w (a0 ))V(T ) (d3 d1 ) W þ (w (a1 ) w (a0 ))V(T ) (d2 d3 ) W þ (w (a3 ) w (a0 ))V(T ) (d1 d2 ) W: (10) In terms of a DSI constraint on w: Ac (a1 ) ¼ V(T ) (d2 d3 ) W Ac (a2 ) ¼ V(T ) (d3 d1 ) W Atc w bc : Ac (a3 ) ¼ V(T ) (d1 d2 ) W Ac (a0 ) ¼ Ac (a1 ) Ac (a2 ) Ac (a3 ) Ac (ai ) ¼ 0 if i f0; 1; 2; 3g (11) bc ¼ 6V 2 (T )(1 1) with 1 [ ½0; 1 : As second-order terms are neglected, bc is chosen as 6V2 (T ) (1 2 1) with 1 0:1. The value of 1 controls how close a tetrahedron can go towards inverting. A low value for 1, such as 0.05 for example, will result in nearly flat tetrahedra. A higher value, such as 0.2, will ensure all tetrahedra keep a certain volume, but will stop editing closer to the starting point. This constraint is installed on all tetrahedra inside the editable volume. The non-inversion constraint has dramatic effects on tetrahedral volumes. In two dimensions, the user can evaluate when triangles will invert and stop before that point. But in three dimensions, it is much more difficult to evaluate the instant when editing goes too far and tetrahedra begin to flip over. The non-inversion constraint stops editing when flipping begins and goes back one
step in the editing to a state where the model is valid.
Consistent editing of geological models In this part, three fault editing examples are shown. The interpreted fault in the first example is not consistent with seismic data for the study area. Its geometry is therefore edited using the fault-editing tool and the result is a better match to the seismic data. In the second example, a single fault is edited while a porosity simulation is mapped on the tetrahedral model. Steps must be taken so that the final model is still geologically valid, so that it honours both the seismic data and the porosity simulation data. In the last example, a fault branch line is edited and it is shown how editing of a fault branch line in a geologically meaningful way is more complicated than editing a single fault.
Adjusting a fault to seismic information After the tetrahedral model was created from interpreted surfaces, fault surfaces within the tetrahedral mesh can be displayed along with seismic data. This can show discrepancies between the seismics and the geometry of the tetrahedral model (Fig. 6b). If this happens, the horizon and fault surfaces used for the creation of the tetrahedral mesh can be edited, and the model can be tessellated again. An alternative time-saving process is to use the fault-editing tool to adjust the geometry of the fault to seismic data. As it is difficult to show a tetrahedral mesh being distorted, only a slice of the model is displayed. Figure 6a shows a global view of the sliced tetrahedral model and the point of view the other images were taken from. Displaying a section in the seismic cube provides a convenient basis for editing, as the user can move control points along a seismic section. Once a control point is set on the fault to be edited (Fig. 7a) it can be dragged until the geometry of the fault matches seismic information (Fig. 7b). The control point is constrained to move along the normal to the fault being edited. The vertices of the tetrahedral mesh are moved along this direction minus the normal to the horizons at each vertex, which is computed from the Geochron parameterization pseudo-time function. This ensures that the pseudo-time function is not distorted, thus preserving horizon geometry. Figure 8 shows superimposed slices of the initial tetrahedral model and of the tetrahedral model after the fault was distorted. The fault geometry is much more consistent with seismic information after editing.
FAULT EDITING IN TETRAHEDRAL MODELS
97
Fig. 6. A single fault with seismic amplitude, in the same dataset as in Figure 2. (a) Location of displayed fault and seismic slice in the tetrahedral model. The model is cropped along the red line close to the seismic slice in further images. A red arrow shows the point of view of further images. (b) Initial fault geometry (black line) and seismic amplitude. The geometry of the fault does not match seismic information (black arrows). Data courtesy of Schlumberger.
Fig. 7. Single fault editing. (a) Control point for fault editing. The green sphere is a control point positioned on the initial fault prior to editing (shown in yellow). The initial tetrahedral mesh is displayed in black. (b) Final and initial fault geometries (shown in green and black respectively). The control point was dragged towards the right and the geometry of the fault was updated accordingly. The resulting fault in the tetrahedral model matches the seismic data. Data courtesy of Schlumberger.
Fig. 8. Initial and distorted tetrahedral meshes, showing the initial mesh (grey) and fault (yellow), together with the final mesh (black) and fault (pink). Vertices of the tetrahedral mesh were moved along with the edited fault. Data courtesy of Schlumberger.
98
A. L. TERTOIS & J. L. MALLET
Three-dimensional property mapping The three-dimensional mapping in the Geochron model introduces another challenge. In the Geochron framework, geological data are mapped from an image of the depositional space, where horizons are flat, onto a faulted and folded geometry. If a property simulation is mapped to the tetrahedral volume, this simulation can have constraints such as values known from well data. These values must remain constant throughout editing. The texture properties which define the threedimensional mapping on the tetrahedral volume are attached to the vertices of the tetrahedral mesh. As their geometry is independent of the attached texture, the texture properties must be updated in order to avoid distorting the mapping along with the geometry. To achieve this, the texture properties are copied to a background volume before editing begins. As the vertices in the editable volume move, their texture properties are retrieved from the background volume. The vertices move, but the property mapping is not distorted because their texture properties are kept up-to-date. Figure 9a shows a slice in a tetrahedral model with a mapped porosity simulation in which a fault intersects a channel. When the fault is edited, the texture properties are updated so that the channel is not distorted. Not updating the properties distorts the porosity mapping and the channel is stretched on one side of the fault and compressed on the other (Fig. 9b). When the texture properties are updated, the channel keeps its original shape and is transferred in part from one side of the fault to the other (Fig. 9c). If other faults are detected inside the editable volume, the value of the distortion property on their nodes is fixed to zero so that their geometry is not modified.
Editing fault branch lines Editing faults in a geological model is more demanding than just ensuring that the tetrahedra do not flip over. If the object to be edited is not a single fault but a fault branch line, it must be edited in such a way that the main fault does not change shape, so that a secondary fault glides along the main fault without deforming it. When a control point is close to a fault branch line, or even on the branch line, additional constraints ensure that the shape of the main fault is preserved during editing. Branch lines are detected automatically in the initialization phase. If such a line is found inside the editable volume, the normal to the main fault is interpolated as a three-dimensional vector field on the editable volume. If other faults are inside the editable volume, their geometry is fixed. When a control point moves inside the
Fig. 9. Illustration of the editing of a mapped property. A vertical slice in a Geochron model showing colour coded simulated values of porosity and cross-cutting faults (thin black lines). (a) Property mapping before editing, in which a fault intersects a channel. (b) Property mapping after fault editing but without updating texture properties. Fault was translated toward the right from its original position (grey dashed line). Because the texture properties were not updated, the channel has yet to be transformed. (c) Property mapping after editing and including updating of texture properties. Data courtesy of Total.
editable volume, the component of the distortion vector W which is colinear with the normal to the main fault is removed for each vertex (Fig. 10). This ensures vertices only move along the main fault, preserving the general shape of the main fault while editing the secondary fault. Figure 11 shows an example of branch line editing, in which the branch line is straightened out whilst retaining the basic shape of the main fault.
FAULT EDITING IN TETRAHEDRAL MODELS
99
N
W WN (a)
(b)
Fig. 10. Editing points close to the branch line between a main fault (grey) and a secondary fault (white). (a) The initialization step, in which the normal to the main fault (black arrows) is interpolated on the editable volume. This vector is then available for each point of the editable volume. (b) Enlargement of box shown in (a). N is the normal to the main fault. W is the distortion vector retrieved from user editing before projection. WN ¼ W 2 (W . N)N is the distortion vector in which the component along the normal to the main fault has been removed.
Fig. 11. Illustration of branch line editing from two different perspectives. (a) Branch line before editing, showing four control points (red spheres) along its length. (b) Editing branch line in which a selected control point (blue) is allowed to move along a line (grey). (c) Branch line after the editing of four control points, with initial branch line also shown (thick black line). (d) Control points on branch line before editing. (e) Control points during branch line editing. in which selected control points are moved in turn and separately. (f) Control points on branch line after editing, with initial branch line also shown (thick white line). Data courtesy of Total.
100
A. L. TERTOIS & J. L. MALLET
Discussion and conclusions When modelling faulted study areas with tetrahedral meshes, a small geometrical correction to a fault may require a complete rebuild of the model and can be very time-consuming. The fault-editing tool presented in this paper enables direct geometrical adjustments in real-time. The user can control the aspect of the mesh precisely while editing, whereas existing FFD techniques usually involve control lattices or control curves. The magnitude of the editable volume and additional control points can be used to specify precisely which regions of the model can change, and which regions should not be modified. The fault-editing tool also features safety precautions which prevent the introduction of inconsistencies in the model. Tetrahedra are prevented from flipping over and making the mesh invalid. Vertices are moved along the pseudo-time function of the Geochron parameterization, thus preserving horizon geometry on both sides of edited faults. The texture properties for the Geochron parameterization are updated as the model is edited so that texture-mapped properties are not distorted. And finally, fault branch lines are edited in such a way that the secondary fault glides along the main fault, leaving the general shape of the main fault undisturbed. The tool provides graphic feedback fast enough to enable comfortable fault editing. Table 1 gives benchmarks of the tool. The time given for each set of nodes inside the editable volume is the average time spent in one tool loop, from the update of the coordinates of the control point to the graphic update (Fig. 5). The frame rate decreases when the editable volume expands, but a volume with about 700 nodes (about 4200 tetrahedra) can be edited in real-time on standard
Table 1. Tool performance on an Intel Dual Xeon 2.40 GHz - 2.39 GHz 2.00 GB RAM with NVIDIA Quadro4 900 XGL 128 MB running Windows XP Service Pack 2. The editing loop starts with the movement of the control point. The distortion property is smoothly interpolated on the nodes inside the editable volume. Then the geometry of the volume is updated. The loop ends with the graphic update that provides user feedback. The given time is the global time for all these operations. The frame rate is above five frames per second, which provides enough feedback for comfortable editing. Nodes in editable volume 403 705
Average editing loop time (ms)
Frame rate (fps)
124.4 169.8
8.036 5.890
hardware. The frame rate only depends on the number of nodes inside the editable volume: the actual tetrahedral mesh usually is much larger than the sub-volume used for editing (about 50000 tetrahedra for the model tested here). In some cases, larger modifications have to be made to tetrahedral models. A fault may have to be translated over such a distance that keeping existing tetrahedra would lead to tetrahedra of very poor quality. Faults may also have to be added to or removed from the mesh. These changes cannot be made while keeping the same tessellation: if a fault is added to the mesh, a topological discontinuity has to be introduced in the model, and if a fault is removed, this discontinuity has to be sealed. The next step for fast model editing is to develop tools that enable topological changes in tetrahedral models without having to recompute the whole tessellation. This work is part of a Ph.D. sponsored by the Association Scientifique pour la Ge´ologie et ses Applications through the G OCAD consortium. Consortium members are hereby acknowledged. Thanks to EarthDecision for providing the G OCAD development environment. Thanks to T. Frank for visualization tools, R. Moyen for the Geochron parameterization and B. Leflon for the porosity simulation.
References B AKER , R., K UPPE , S., C HUGH , S., B ORA , R., S TOJANOVIC , S. & B ATYCKY , R. 2001. Full-field modeling using streamline-based simulation: Four case studies. Paper SPE 66405, presented at the SPE Reservoir Simulation Symposium, Houston, Texas. B ARR , A. H. 1984. Global and local deformations of solid primitives. In: Proceedings of the 11th Annual Conference an Computer Graphics and Interactive Techniques SIGGRAPH 84. ACM Press, New York, 21–30. B LUNT , M., L IU , K. & T HIELE , M. 1996. A generalized streamline method to predict reservoir flow. Petroleum Geosciences, 2, 259–269. B ORREL , P. & R APPOPORT , A. 1994. Simple constrained deformations for geometric modeling and interactive design. ACM Transaction on Graphics, 13, 137–155. C ACAS , M.-C., L EDOUX , E., DE M ARSILY , G., ET AL . 1990. Modeling fracture flow with a stochastic discrete fracture network: Calibration and validation 2 1. The flow model. Water Resources Research, 26, 479– 489. C AUMON , G., L EPAGE , F., S WORD , C. & M ALLET , J.-L. 2004. Building and editing a sealed geological model. Mathematical Geology, 36, 405– 424. C HANG , Y. & R OCKWOOD , A. P. 1994. A generalized de Casteljau approach to 3D free-form deformation. In: Proceedings of the 21st Annual Conference an Computer Graphics and Interactive Techniques SIGGRAPH 94. ACM Press, New York, 257–260. C OQUILLART , S. 1990. Extended free-form deformation: a sculpturing tool for 3D geometric modeling. Technical
FAULT EDITING IN TETRAHEDRAL MODELS Report RR-1250, Inria, Institut National de Recherche en Informatique et en Automatique. F RANK , T. 2006. Advanced vizualization and modeling of tetrahedral meshes. Ph.D. thesis, Institut National Polytechnique de Lorraine, Nancy, France. G RIFFITHS , P., J ONES , S., S ALTER , N., S CHAEFER , F., O SFIELD , R. & R EISER , H. 2002. A new technique for 3-D flexural-slip restoration. Journal of Structural Geology, 24, 773–782. G ROSSE , O. 2002. Remise en cohe´rence d’un mode`le ge´ologique 3D. Ph.D. thesis, Institut National Polytechnique de Lorraine, Nancy, France. H IROTA , G., L IN , M. C. & M AHESHWARI , R. 1999. Fast volume-preserving free form deformation using multilevel optimization. In: Proceedings of the 5th ACM Symposium on Solid Modeling and Applications 0 SMA 99. ACM Press, New York, 234–245. J OSNIN , J.-Y., J OURDE , H., F´ ENARD , P. & B IDAUX , P. 2002. A three-dimensional model to simulate joint networks in layered rocks. Canadian Journal of Earth Sciences, 39, 1443– 1455. L EPAGE , F. 2002. Triangular and tetrahedral meshes for geological models. In: International Association of Mathematical Geology. IAMG 7th International Conference Proceedings, 15– 20 September. Springer, Berlin, Germany. L EPAGE , F. 2003. Ge´ne´ration de maillages tridimensionnels pour la simulation des phe´nome`nes physiques en ge´osciences. Ph.D. thesis, Institut National Polytechnique de Lorraine, Nancy, France. M ACE´ , L., S OUCHE , L. & M ALLET , J.-L. 2004. 3D fracture characterization based on geomechanics and geologic data uncertainties. In: 9th European Conference on the Mathematics of Oil Recovery (ECMOR). 30 August – 2 September, Cannes, France. EAGE, Houten, The Netherlands. M ACE´ , L., M URON , P. & M ALLET , J. -L. 2005. Integration of fracture data into 3D geomechanical modeling to enhance fractured reservoirs characterization. SPE Paper 95827 presented at the SPE Annual Technical Conference and Exhibition, Dallas, Texas, U.S.A., 9–12 October, 1 –9. M ACY , B. K. & S MITH , B. A. 1998. 3-D anisotropic ray tracing on a tetrahedral mesh. SEG Technical Program Expanded Abstracts, 1973– 1974. M ALLET , J.-L. 1992. Discrete smooth interpolation in geometric modeling. Computer-Aided Design 24, 4, 178–191. M ALLET , J. -L. 2002a. Geomodeling. Oxford University Press, USA. M ALLET , J.-L. 2002b. Space/time mathematical framework for sedimentary geology. In: Gocad Meeting Proceedings, 24– 25 June 2002, 10–11 June 2003. ASGA, Nancy, France. M ALLET , J.-L. 2003. Constraining a piecewise linear function defined on a 3D complex (Applications to 3D restoration). In: Gocad Meeting Proceedings, 24–25 June 2002, 10–11 June 2003. ASGA, Nancy, France.
101
M ALLET , J.-L. 2004. Space-time mathematical framework for sedimentary geology. Mathematical Geology, 36, 1 –32. M ALLET , J.-L., M OYEN , R., F RANK , T., L EFLON , B. & R OYER , J.-J. 2004. Getting rid of stratigraphic grids. In: 66th EAGE annual meeting, 7–10 June, Paris. EAGE, Houten, The Netherlands. M ATRINGE , S. F., J UANES , R. & T CHELEPI , H. A. 2006. Tracing streamlines on unstructured grids from finite volume discretizations. SPE Paper 103295 in Proceedings of the SPE Annual Technical Conference and Exhibition. M OYEN , R. 2005. Parame´trisation 3D de l’espace en ge´ologie se´dimentaire : le mode`le GeoChron. Ph.D. thesis, Institut National Polytechnique de Lorraine, Nancy, France. M URON , P. 2005. Me´thodes nume´riques 3D de restauration des structures ge´ologiques faille´es. Ph.D. thesis, Institut National Polytechnique de Lorraine, Nancy, France. O KABE , A., B OOTS , B., S UGIHARA , K. & C HIU , S. N. 2000. Spatial Tessellations – Concepts and Applications of Voronoi diagrams. John Wiley & Sons, England. O WEN , S. J. 1998. A survey of unstructured mesh generation technology. In: International Meshing Roundtable. 14– 17 September, Sandia National Laboratories, Santa Fe, NM, USA, 239–267. P ALAGI , C., B ALLIN , P. & A ZIZ , K. 1993. The modeling of flow in heterogeneous reservoirs with voronoi grids. In: 12th SPE Symposium on Reservoir Simulation, 23 February – 3 March, New Orleans, Louisiana, 291– 299. SPE, Richardson, TX, USA. P RE´ VOST , M., L EPAGE , F., D URLOFSKY , L. & M ALLET , J. -L. 2004. Unstructured 3D gridding and upscaling for coarse modeling of geometrically complex reservoirs. In: The European Conference on the Mathematics of Oil Recovery. Cannes, France. S EDERBERG , T. W. & P ARRY , S. R. 1986. Free-form deformation of solid geometric models. In: Proceedings of the 13th Annual Conference an Computer Graphics and Interactive Techniques SIGGRAPH 86. ACM Press, New York, 151– 160. S HONTZ , S. M. & V AVASIS , S. A. 2003. A mesh warping algorithm based on weighted laplacian smoothing. In: International Meshing Roundtable. 26–28 October, Sandia National Laboratories, Santa Fe, NM, USA, 147–158. Dearborn, Michigan, USA. V ELTEN , W. 1998. Effective seismic modelling in 3D Earth models. Ph.D. thesis, Institut National Polytechnique de Lorraine, Nancy, France. V ERMA , S. 1996. Flexible grids for reservoir simulation. Ph.D. thesis, Stanford University, California, USA. Y ILMAZ , O. 1987. Seismic Data Processing. SEG Investigations in Geophysics 2, Society of Exploration Geophysicists.
Mechanics of fault and expulsion rollover systems developed on passive margins detached on salt: insights from analogue modelling and optical strain monitoring C. KRE´ZSEK1,2, J. ADAM1 & D. GRUJIC1 1
Dalhousie University, Department of Earth Sciences, Life Sciences Centre, B3H 4J1 Halifax, Nova Scotia, Canada (e-mail:
[email protected]) 2 Present address: StatoilHydro, Oil and Energy, Global Exploration, 31 Kjørboveien, N-0246 Oslo, Norway Abstract: Scaled analogue experiments with layered brittle and ductile materials have been used to simulate the development of listric growth-fault and expulsion rollover systems during gravitational spreading of a passive margin sedimentary wedge detached on salt. The experiments were performed with varying sedimentation patterns and rates to simulate different depositional scenarios. Deformation monitoring with 3D optical image correlation techniques was used to quantify the 3D surface evolution and strain history of model structures. Our results indicate that rollover structure kinematics is strongly coupled to sedimentation patterns and rates. Whereas differential loading governs the margin-scale state of stress and extensional spreading in the experiments, more localized feedback between the dynamic depositional systems, fault-controlled subsidence, and salt mobilization control the strain history of local fault structures. This is reflected in the characteristic succession of extensional structures that evolve from symmetrical grabens through early, mature and late (collapsed) basinward listric growth-fault and rollover systems into landward listric growth-fault and rollover systems. A lack of sedimentation enhances reactive diapir rise and passive diapirism, whereas low sedimentation rates favour development of longlived basinward listric growth-fault or expulsion rollover systems. Conversely, high sedimentation rates lead to the development of landward listric growth-fault and rollover systems.
Thin-skinned extension of passive continental margins is controlled by gravity-driven down-slope spreading of the brittle overburden above a viscous (visco-plastic) substratum, such as salt (e.g. Jackson 1995). Down-slope gravitational spreading is induced by the pressure difference DP ¼ P1 2 P2 due to differential sediment loading (Fig. 1). This causes the flow in the viscous substratum and the induced shear causes a drag force at the base of the overburden and triggers extensional failure of the overburden (e.g. Gemmer et al. 2004). The up-dip extension may be accommodated by down-dip shortening (e.g. Burrolet 1975; Jackson & Cramez 1989; Worrall & Snelson 1989; Wu et al. 1990, see fig. 1). Some of the most common structures that develop in the extensional domain are listric growth-fault and rollover systems (e.g. Burrolet 1975; Bally et al. 1981; Shelton 1984; Jackson & Cramez 1989; West 1989; Jackson & Talbot 1991; Xiao & Suppe 1992; Seni 1992; Rowan 1993; Roberts & Yielding 1994; Schuster 1995). Rollover anticlines have been of great interest for oil and gas exploration in salt basins around the world (e.g. Demercian et al. 1993; Diegel et al. 1995; Peel et al. 1995; Cainelli & Mohriak 1999;
Cobbold et al. 2001; Tari et al. 2002; Shimeld 2004) because several play types can be associated with them (e.g. Bally & Tari 2004) (Figs 1 & 2). Thus, a key issue is to understand their structural evolution and mechanics (e.g. Rouby et al. 2000, 2002; Brown et al. 2004). In general, rollover systems may be grouped into two kinematic families: fault rollovers and expulsion rollovers (Fig. 3) (Hossack 1995; Rowan 1993; Ge et al. 1997). Fault rollovers develop because of geometric and space compatibility problems caused by extensional displacement along a listric growth-fault and subsequent down-bending of the hanging wall strata. Listric growth-fault and rollover systems can also be differentiated by the dip of the listric growth-fault relative to the down-slope oriented gravitational spreading direction, into basinward (i.e. down-slope; BLS) or landward (i.e. up slope; LLS) listric growth-fault and rollover systems (Mauduit & Brun 1998). The synsedimentary fault activity is indicated by thickening (i.e. growth) of synkinematic strata toward the listric fault (e.g. Bally et al. 1981). Thus, the kinematics of fault rollover anticlines is closely related to the amount and rate of extension in the overburden. Conversely, expulsion rollovers develop due to salt withdrawal.
From: JOLLEY , S. J., BARR , D., WALSH , J. J. & KNIPE , R. J. (eds) Structurally Complex Reservoirs. Geological Society, London, Special Publications, 292, 103–121. DOI: 10.1144/SP292.6 0305-8719/07/$15.00 # The Geological Society of London 2007.
C. KRE´ZSEK ET AL.
104
shelf
basin
P1 P2
Brittle overburden Viscous salt Extension
Contraction
L
KS
L
KS
A A S
S
Fig. 1. Schematic diagram illustrating the effect of differential pressure on the generation of different salt-related structures. Differential pressure (P1 2 P2) in a passive margin wedge (brittle overburden) above a viscous salt (ductile detachment) will trigger flow in the viscous layer and the resulting drag forces will cause extension in the overburden. The up-dip extension may be accommodated by down-dip contraction. Listric growth-fault and rollover systems (here shown as model data) are one of the most characteristic structures that develop in the extensional domain (the scale bar is 1 km) and consist of listric growth-fault (L), salt roller (S), rollover anticline (A) and associated key-stone graben (KS). The white marks represent possible HC play types.
Lateral salt removal (decrease of cross-sectional area, see Fig. 3) forces the progressive sinking of the overburden below the initial regional level (e.g. Schuster 1995; Ge et al. 1997). During lateral salt expulsion, the overburden is passively down-building. Expulsion rollovers do not accommodate any overburden sec
extension but are instead caused by bending due to salt withdrawal (Ge et al. 1997). Although both rollover mechanisms can act together in nature, under certain conditions one of them can dominate. It is not an easy task to discriminate between the different rollover structures and differentiation can only be sec
2km
1.6
1.6
2.0
2.0
2.4
2.4
2.8
pre-salt basement
3.2 pre-salt basement
Oligocene Santonian Turonian Cenomanian
2.8
3.2 Upper Albian Lower Albian Aptian evaporites
Fig. 2. Line drawing of a complex listric growth-fault and rollover from offshore Angola (redrawn based on Rouby et al. 2002, p. 785, fig. 2) that shows an early basinward listric growth phase followed by a late landward listric growth phase.
MECHANICS OF ROLLOVER SYSTEMS
Fault rollover extension “Pinned”
Expulsion rollover no extension “Free” b
a
105
“Pinned”
“Pinned” b
a
Amount of horizontal extension
Amount of salt withdrawal
a
b
a
a
b
a
b
b downbuilding
Overburden Passive diapir Space created by extension or salt withdrawal Fig. 3. Overview of rollover mechanisms, showing how fault rollovers involve extensional faulting of the overburden whilst expulsion rollovers develop due to salt withdrawal.
performed by careful structural restoration (e.g. Seni 1992; Xiao & Suppe 1992; Rowan 1993; Rouby et al. 2000). The kinematic concepts derived from analogue (see review in Jackson 1995) and numerical modelling (e.g. Gemmer et al. 2004) have been proven to be very helpful in understanding and validating the structural evolution of passive margin sedimentary wedges detached on salt. We apply scaled analogue experiments to simulate thin-skinned extension and gravity spreading of a passive margin sedimentary wedge detached on salt. Our aim is to develop an improved understanding of the kinematics and mechanics of fault and salt structures that develop by thin-skinned extensional deformation. This analogue modelling approach with high-resolution strain monitoring enables us quantitatively to assess: (i) the strain history of fault and salt structures; (ii) the role of sedimentation in deformation; and (iii) the changing mechanical conditions during rollover deformation. The experimental results demonstrate that the extensional structures and the margin-scale kinematic segmentation that develop during thin-skinned gravity spreading reflect the mechanical coupling between the ductile and brittle layers.
Modelling technique We use scaled analogue experiments consisting of silica sand and silicone elastomer, and 3D optical deformation analysis (PIV, Particle Imaging Velocimetry; Adam et al. 2005, 2006) to study: (1) the fault mechanisms; (2) the role of sedimentation; and (3) the mechanical coupling between the ductile and brittle layers.
Analogue materials Sifted silica sand (grain size distribution: 0.02– 0.45 mm; angle of internal friction: 348; density: 1.6 g cm23, strain softening c. 10–20%) was used to simulate the non-linear frictional-plastic deformation behaviour of brittle sedimentary rocks (Lohrmann et al. 2003). Coloured silica sand was used as marker horizons for the structural interpretation of the final deformation in model sections. A silicone elastomer (PDMS, polydimethylsiloxane; Wacker Elastomer NA USA; viscosity, 6 104 Pa s; density, 0.99 g cm23) with linear viscous behaviour under experimental strain rates simulates viscous flow of salt sediments under gravitational loading.
C. KRE´ZSEK ET AL.
106
Scaling The models are scaled to allow the quantitative comparison of the model geometry, kinematics and stresses to natural prototypes (e.g. Hubbert 1937; Ramberg 1981; Weijermars et al. 1993; Brun 1999; Costa & Vendeville 2002). The geometric scaling factor l * is 105 deduced from the density and cohesion of the sand material and gravitational acceleration. Therefore, 1 cm in the experiments equals 1 km in nature. Because gravitational acceleration is equal under experiment and natural conditions (i.e. the gravity ratio g* is 1), the stress ratio s * in the brittle overburden is computed using the geometric scaling factor l* and the density ratio r * (rmodel/rprototype ¼ 0.7):
s ¼ l r g ¼ ðlmodel =lprototype Þ ðrmodel =rprototype Þ ¼ 6:96 106 where: lmodel is 1 cm, lprototype is 1 km, rmodel is 1.6 g cm23, rprototype is 2.3 g cm23. The viscosity ratio h* between the model silicone (6 1024 Pa s) and natural salt (c. 1018 Pa s), and the stress ratio s* define the strain ratio 1*: 1 ¼ s =h ¼ 1:16 108 The scaling factor of time t* (i.e. tmodel/tprototype) is inversely proportional to the strain ratio 1*: t ¼ 1=1 1:4 109 Therefore, one hour in our models represents about 140 000 years in nature. The model density ratio of sand and silicone (rsand/rsilicone ¼ 1.61) is greater than the density ratio between sediments and salt in nature for rsalt ¼ 2.2 g cm23) (rsediment/rsalt ¼ 1.05 leading to higher buoyancy forces in the silicone in the model. However, salt mobilization due to buoyancy forces is considered to be minor compared with salt mobilization due to extension of the overburden. As is the case for natural passive margin thin-skinned extensional systems, the salt structures (e.g. reactive and passive diapirs) develop passively due to overburden deformation rather than actively pushing through the overburden due to buoyancy (Vendeville & Jackson 1992).
Experiment setup The 3D experiments were performed on a horizontal rig (120 cm long 90 cm wide). The base of each model consists of a pre-kinematic silicone layer of
1 cm thickness, which represents the initial salt basin with a typical area of 100 65 cm (Fig. 4a). The transparent silicone layer is covered with a thin preexperiment sand layer (c. 0.2 cm) in order to facilitate optical monitoring of the experiment surface. This experimental set up represents a passive margin sedimentary wedge that is prograding over a 80 –100 km wide basin filled with 1km thick evaporites. During the experiment, sand layers were sieved manually with precisely defined time laps and thickness onto the pre-experiment layer to simulate the progradation of a passive margin sedimentary wedge (Fig. 4). The sieving procedure provides homogeneous mechanical conditions (Lohrmann et al. 2003) in the sand layer and allows the structural evolution of the model surface to be linked to specific sedimentation patterns. In order to study the role of sedimentation on the development of the extensional structures in the passive margin sedimentary wedge, we performed experiments with different sedimentation patterns and rates (Table 1), while all other parameters (e.g. thickness and lateral extent of the ductile layer, basement configuration, etc.) were kept constant. Aggradation and progradation were simulated (Models 1 to 4, Fig. 4b, c). Models 1 and 2 (Fig. 4b, Table 1) investigate an aggradation setting where sedimentation was restricted to the initial shelf (i.e. the shelf/slope break was kept fixed). Aggradation was simulated by the incremental deposition of 0.5 cm (Model 1) and 0.25 cm (Model 2) thick sand layers in one-hour time intervals until the final sediment thickness of 4 cm on the shelf was reached. After the shelf build-up, no more sand was added in Model 1 to simulate a long sedimentation condensation interval. In contrast, in Model 2, sedimentation continued after the shelf build-up, but was restricted to active grabens where additional accommodation space was created by extension and/or silicone withdrawal. In Models 3 and 4 sediment aggradation and progradation was simulated. During the shelf build-up, aggradation on the shelf occurred with a rate of 0.12 cm/hour (Model 3) or 0.5 cm/hour (Model 4) coeval with progradation until the shelf succession reached a thickness of 4 cm. After the shelf build-up, sedimentation on the shelf was restricted to active grabens and on the prograding slope area basinward of the shelf –slope break. Progradation consisted of basinward movement of the shelf –slope break at a fixed rate (0.5 cm/hour in Model 3 and 1 cm/hour in Model 4; Table 1) while a constant slope profile was maintained (Fig. 4c). The sedimentation patterns were similar in Models 3 and 4, but Model 4 was performed with higher sedimentation rates. After completion, the experiments were cut perpendicular to the strike of the shelf to generate a set of parallel cross-sections at 5 cm intervals. For the
MECHANICS OF ROLLOVER SYSTEMS
107
Fig. 4. Experimental setup for the models presented. (a) Initial silicone basin with 1 cm thick silicone bordered by 3 cm wide sand border. The silicone surface was covered by a thin (c. 0.2 cm) thick pre-experiment sand layer. A height-adjustable guide and rail system was used to model incremental sedimentation with specific patterns and rates. Models 1 to 4 represent different sedimentary scenarios (Table 1). (b) In Models 1 and 2 no progradation is simulated and the shelf is built up by aggradation to 4 cm high with 0.5 cm/hour (Model 1) or 0.25 cm/hour (Model 2) sedimentation rates. After the shelf build up, in Model 1 sedimentation was stopped and in Model 2 sedimentation continued in the active grabens on the shelf and slope. (c) In Models 3 and 4 aggradation was simulated with a rate of 0.12 cm/hour (Model 3) or 0.5 cm/hour (Model 4). Progradation was continuous with a rate of 0.5 cm/hour (Model 3) and 1 cm/hour (Model 4). Aggradation on the shelf was stopped when the sedimentary thickness reached 4 cm. Afterwards the sedimentation on the shelf was restricted to active grabens.
sectioning procedure, the model surface was covered by a thick protective post-experiment sand layer, which removed burial differences and therefore any pressure gradient that would cause further migration of the silicone. Then the model was sprayed with water and finally sectioned and imaged with a digital camera. The section images were digitally edited to aid the structural interpretation. For this paper, we have selected the representative centre sections from the experiments (Fig. 5). Although the post-experiment layer in the central section of Model 4 was too thin to stop diapirism
at the toe of the slope, so that a diapir later pierced the post-experiment layer (Fig. 5d), this late stage diapirism does not impact the syn-experiment extensional structures and can therefore be neglected for the purposes of this study.
Optical strain monitoring 3D optical deformation and surface flow monitoring techniques along with displacement data analysis (PIV) were used to monitor the incremental and total deformation and surface evolution of
Table 1. Experimental setup and parameters Model No.
1 2 3 4
Silicone L W H (cm)
100 100 100 100
65 65 65 65
1 1 1 1
Shelf build up to 4 cm high
Post-shelf build-up
Aggradation (cm/h)
Progradation (cm/h)
Aggradation
Progradation (cm/h)
0.5 0.25 0.12 0.5
None None 0.5 1
None Active grabens Active grabens Active grabens
None None 0.5 1
Duration (hours)
72 56 56 98
Fig. 5. Overview of experiments and extensional evolution. For every experiment, one representative central section is shown together with its associated structural interpretation. The evolution of strain is illustrated as diagrams with incremental horizontal strain and finite horizontal strain v. time (see text for discussion). In all experiments, extensional structures are developed on the shelf and upper slope whereas the deep basin area is characterized by thickened silicone with small-scale buckle folds. Early extensional faults were always planar and evolved coeval with reactive diapir rise. Some of the planar faults, like in Models 2 and 4 were transformed to basinward listric growth-faults (BLS). However, most of them were abandoned early in the experiments (annotated with R). (a) Model 1 ran with high sedimentation rates (0.5 cm/hour) and aggradational sedimentary patterns until 8 hours, with no sedimentation afterwards (Fig. 4, Table 1). Grabens are labelled G1, G2 and G3. (b) Model 2 ran with moderate sedimentation rates (0.25 cm/hour) and aggradational sedimentary patterns until 16 hours. Post-16 hours, the sedimentation continued restricted to active grabens (Fig. 4, Table 1). Basinward listric growth-fault and rollover systems (BLS1 and 2), fault rollovers preserving some early normal faults (R) and keystone grabens (KS) are labelled. (c) Model 3 ran with low sedimentation rates (0.12 cm/hour) and progradational sedimentary patterns (Fig. 4, Table 1). Common extensional faults (R) and expulsion rollovers (ERS1-3) are labelled. (d) Model 4 ran with high sedimentation rates (0.5 cm/hour) and progradational sedimentary patterns. Several early grabens (G1-G3), a basinward listric growth-fault and rollover system (BLS) and a landward listric growth-fault and rollover system (LLS).
Fig. 6. Development of extensional structures in the aggradation (a, b) and progradation (c, d) experiments illustrated by time-series of incremental maximum principal horizontal strain and vertical displacement data. The white lines on the images mark the location of the structural reconstructions. The silicone is black on the structural reconstructions.
108
C. KRE´ZSEK ET AL.
the experiments (e.g. Adam et al. 2005, 2006). The monitoring equipment consisted of two computercontrolled high-resolution monochrome CCD cameras that were mounted in a stereo set up over the experiments. Stereo images of the experiment surface taken in intervals of 10 minutes were processed, analysed and cross-correlated using a dedicated image correlation and deformation analysis software adapted for analogue experiments (Strain Master, LaVision#). From successive images, the incremental 3D displacement vector field was calculated with submillimetre accuracy. This 3D incremental displacement field formed the basis for the further analysis of displacement components (e.g. subsidence, vz) and derived strain components (e.g. horizontal strain exx). For additional details of the PIV monitoring method, including technique, computation algorithms, resolution, and accuracy refer to Adam et al. (2005).
Structural and mechanical interpretation The structural interpretation of final model deformation was completed for each section. The integration of the structural model data with the strain data of active faults enabled reliable 3D fault correlation and mechanical analysis of the evolution of the complex fault and expulsion rollover systems. Selected fault and salt structures were restored using conventional balancing techniques (Rowan 1993). Their kinematic evolution was analysed in one-hour intervals using the incremental horizontal strain data (exx) and vertical displacement data (vz). These displacement and strain components yield insights into the accumulation and distribution of incremental strain (e.g. localization of faults) and quantify the incremental vertical movements (e.g. subsidence). The map of the incremental strain data shows extensional faults as linear zones of maximum negative horizontal strain (green), and the hinges of compressional folds appear as trends of maximum positive horizontal strain (blue) (Fig. 6a– d). The negative values of vertical displacement document active subsidence (grabens in blue and magenta colours), whereas the positive values show uplifting areas (passive diapirs and folds in green colours) (Fig. 6a–d). Quantitative analysis of the individual strain history of grabens, fault and expulsion rollover systems (Fig. 5) during the experiment facilitated the study of the characteristic differences and similarities in the kinematic development of these extensional structures. The kinematic analysis shows that distinct combinations of strain and subsidence patterns are characteristic of different extensional structures (e.g. symmetric grabens, listric growthfault and expulsion rollover systems; Fig. 7).
Kinematic development of graben systems In all experiments brittle failure of the sedimentary wedge was caused by flow in the silicone layer arising from the increase in differential load during the shelf build-up. As a result, extensional structures developed on the shelf and the upper slope. The deep basin experienced simultaneous contraction, represented by thickening of the autochthonous silicone and small-scale buckle folding of the thin pre-kinematic sand and silicone layers (Figs 5, 6). All faults terminate in the silicone layer, which therefore functions as the main detachment level. The early extensional evolution in all experiments was very similar. Extension was distributed on the shelf in zones of diffuse strain trending perpendicular to the downslope oriented maximum principal extension direction (Fig. 7a). The strain and subsidence patterns recorded in Model 2 at 5 hours (Figs 6b, 7a) exemplify this configuration. Because of diffuse extension, little or negligible subsidence was observed and the faults and grabens had yet to develop. Early faults developed as strain localized on the edges of the diffuse extensional zones. These faults triggered the formation of symmetrical grabens with a set of conjugate planar normal faults. These early grabens are characterized by symmetrical subsidence patterns with the highest subsidence located centrally (Fig. 7b). Important differences between the experiments occurred in the timing, distribution and frequency of the early extension structures. The finite and incremental strain histories of the early grabens indicate an increase of the incremental strain, but with different strain rates (Fig. 5). Initiation of overburden failure occurred earliest in Model 1, where the first grabens formed after 2 hours. Although Model 1 and 2 had similar aggradational sedimentary patterns, but with different sedimentation rates (Table 1), the early graben systems developed on the outer shelf and upper slope of both models. In Model 1 the first graben system (G1) initiated on the outer shelf. After 5 hours the graben developed into a composite graben system (Fig. 5a) when a second graben system (G2) initiated on the slope (Figs 5a, 6a). In a similar way, in Model 2 the symmetrical graben and basinward listric growth-fault and rollover systems (BLS1, BLS2 and G3 in Figs 5b, 6b) initiated in a basinward breaking sequence. The initial position of the graben systems in both models was similar. However, the one-hour incremental strain rates of the grabens in Model 2 were one order of magnitude lower than those observed in Model 1 (Fig. 5a, b). This is also indicated by the incremental horizontal strain data of Model 2 (Fig. 6b), which was
MECHANICS OF ROLLOVER SYSTEMS
109
Fig. 7. Structural reconstruction overlain by vertical displacement (middle part) and the down-slope oriented principal horizontal strain (upper part) data recorded over one-hour intervals. The combination of these patterns gives important insights into the structural development and constrains the reconstructions of section data. The pattern combinations are characteristic of the development of the following structures: (a) pre-faulting; (b) symmetric simple graben; (c) early basinward listric growth-fault and rollover system; (d) mature basinward listric growth-fault and rollover system; (e) collapse of the rollover anticline; (f) early landward listric growth-fault and rollover system; (g) complex graben; (h) inactive graben with passive diapir; and (i) expulsion rollovers with passive diapir. Note that frequently some compression is indicated near major extensional faults. Compression recorded by the monitoring system adjacent to major extensional faults arises from sediment fall off active fault scarps into the grabens, and therefore does not truly represent tectonic compression of the model.
characterized by narrow graben systems in comparison to Model 1 (Fig. 6a). The progradational sedimentation pattern of Models 3 and 4 differed significantly from the aggradational pattern of Models 1 and 2 (Fig. 4). The positions of the early grabens were similar in Models 3 and 4 (Fig. 6c, d), but different from Models 1 and 2 (Fig. 6a, b). In Models 3 and 4,
the early grabens were located not only on the outer shelf and on the slope as observed in Models 1 and 2, but also occurred on the inner shelf (e.g. ERS1 in Model 3; G1-G3 in Model 4). In general, in Models 3 and 4 more grabens developed on the shelf than in Models 1 and 2. The sedimentation rates were highest in Model 4 and lowest in Model 3 (Table 1), a feature which
110
C. KRE´ZSEK ET AL.
correlates with the larger number of graben systems (i.e. four; Fig. 6d by 5 hours) and the higher strain rates (Fig. 5d) recorded in Model 4. Conversely, Model 3 deformed with the lowest strain rates of all of the models (Fig. 5c). In Model 1, sedimentation ceased after 8 hours with the extension of the graben systems continuing afterwards but with decreasing rates (Figs 5a, 7g). By 11 hours, most of the diapirs pierced the graben floors and emerged to the surface (e.g. Figs 5a, 6a). This initiation of passive diapirism was coeval with the complete cessation of fault activity along the graben shoulders (e.g. Fig. 6a by 11–13 hours; 7 h). After 11 hours, the extension focused entirely on the passive diapirs (Fig. 7h) leading to the continuous widening of the passive diapirs (Fig. 5a). The development of Model 2 was entirely different from the post 8-hours evolution of stagnant Model 1. In Model 2, due to the continuous sedimentation, only the reactive diapir of the graben system G3 was able to actively pierce its graben floor. This most likely occurred because the graben formed on the lower slope, where the overburden was thin (Fig. 5b). In the other grabens (BLS1 and 2), the diapirs never pierced the graben floor. In contrast, after 13 hours, continuous sedimentation transformed the initial symmetrical grabens into basinward listric growth-fault and rollover systems (Fig. 5b). This change is observable as a characteristic subsidence and extensional strain pattern. The subsidence pattern is highly asymmetrical with the location of maximum subsidence shifted toward the listric growth-fault (Fig. 7c), and with new secondary normal faults in the developing hanging wall rollover (key-stone graben in Fig. 7c). These fault and subsidence patterns correlate well with the characteristics of basinward listric growth-fault and rollover systems and are consistent with the structures observed in the model sections. In Model 4, as in Model 2, the graben systems initiated on both the outer shelf and upper slope were rapidly transformed into basinward listric growth-fault and rollover systems (G3 and BLS on Figs 5d, 6d). However, in Model 4 the transformation occurred earlier than in Model 2 (8–9 hours rather than 11 hours). Coeval with the transformation, the extension in the landward part of the Model 4 decreased and the grabens on the inner shelf (G1, G2) became progressively inactive (Figs 5d, 6d). In Model 3, which is characterized by the lowest sedimentation rates of all of the models (Table 1), the transition of the early symmetrical graben systems into basinward listric growth-fault and rollover systems never occurred and the reactive diapirs pierced the central part of the graben floor after
11 hours (Fig. 6c). Similar to Model 1, all faults of the early graben systems became inactive as the diapirs emerged to the surface and further extension was localized along the axis of the passive diapirs. In this progradation scenario with low sedimentation rates, the surface depressions formed by the active grabens in the shelf area were not filled completely during the sedimentation events. Here, a thin layer of sediments (0.12 cm/hour) was supplied to the grabens that covered the top of the passive diapirs. During the subsequent nonsedimentation interval, ongoing extension along the diapir axis formed minor conjugate faults in the thin cover sediments above the diapir. The minor faults bound small grabens, which the incremental vertical displacement data highlight as narrow zones of high subsidence bordered by zones of maximum strain (i.e. faults) (Fig. 6c). However, the evolution of the minor grabens was very short-lived, because the diapirs quickly pierced the new sediment in the graben and emerged again to the surface (Fig. 7i). This cyclic process of diapir burial, minor faulting of the cover and re-emergence of the diapir was observable after every sedimentation event. The total extension accumulated by the conjugate faults above the diapir crest does not account for the finite extension of the grabens because most of the extension was accommodated by the diapirs themselves (Fig. 5c). In Model 1, after 15 hours, a new graben system (G3) started to develop on the inner shelf (Fig. 5a) landward of G1 and G2. By that time, the extension of the composite graben G1 was negligible and only minor in G2. In comparison, the peak strain rate of G3 was about two orders of magnitude smaller than the maximum rates observed for G1 and G2 (Fig. 5a). This suggests that sedimentation had a significant effect on the strain rates, because G3 developed during the non-sedimentation stage. In Models 2 and 3, by contrast, no major structural changes occurred during the late stage of the experiment run. Most of the structures reflect peak incremental strain rates between 15 –20 hours (Fig. 5b, c), after which the strain rates progressively decrease. In Model 4 most of the extension was accumulated by basinward listric growth-fault and rollover systems (BLS in Fig. 5d). During the early evolution (8–9 hours), maximum subsidence was located near the main basinward listric growth-fault and was almost negligible in the incipient keystone grabens (Figs 6d, 7c). This configuration changed between 9–11 hours, when the maximum subsidence progressively shifted from the listric main fault into the keystone graben of the hanging wall rollover (Fig. 7d). The shift of subsidence site was coeval with the decrease in strain rate of the main listric growth-fault. As a result, after 13 hours the
MECHANICS OF ROLLOVER SYSTEMS
subsidence associated with displacement along the main basinward listric growth-fault became almost negligible (Figs 6d, 7e). Thus, ongoing extension in the overburden was not accommodated by the listric main fault but within the keystone grabens. Some planar normal faults of the early reactive diapir stage that were inactive during the basinward listric rollover phase were reactivated in this stage (Fig. 7e). The extension in keystone grabens was controlled by the collapse of the rollover anticline as consequence of the basinward translation of the rafts in the slope area rather than the bending of rollover hanging wall (Figs 6d, 7e). Subsidence in the keystone grabens stopped when one of the landward dipping faults became dominant and developed into the main fault of the newly formed landward listric growth-fault and rollover (Figs 5d, 6d). This transition towards the landward listric growth-fault and rollover is observable in the characteristic asymmetrical subsidence pattern of the hanging wall with the maximum subsidence close to the basinward main fault (Figs 6d, 7f). The rollover collapse and the early phase of the landward fault rollover systems are characterized by the highest incremental extension rates, which decreased progressively during the final stages of the experiments (Fig. 5d).
Listric growth-fault and rollover kinematics Basinward and landward listric growth-fault and rollover systems are characterized by very different kinematic evolutions and contrasting fault mechanisms (Fig. 8). Displacement on the basinward listric growth-faults requires movement of the hanging wall block along the entire fault length including the basal detachment (BB0 ). In the passive margin setting, this is possible if the basinward translation of the hanging wall block is faster in comparison to the footwall block (v2 . v1 and v2 ¼ v3). This requires relatively efficient gliding of the hanging wall block along its basal detachment (Fig. 8a). In contrast, displacement on a landward listric growthfault is possible if the footwall block moves faster than the trailing hanging wall segment (v3 v2). This implies relatively slow (or no) gliding of the hanging wall block (Fig. 8b). Thus, the evolution of landward listric growth-fault and rollover systems, i.e. the growth of the hanging wall block, depends on the basinward translation of its footwall block rather than displacement of the hanging wall (e.g. Mauduit & Brun 1998). Any upslope oriented displacement along the rollover base is only apparent because the segment L0 L00 is merely an ‘old fault trace’ but has not accumulated any true
111
displacement in comparison to the active landward main fault segment (LL0 ) (Fig. 8b). Thin sheet stability analysis (Lehner 2000; Gemmer et al. 2004) shows that the hanging wall movement will be maintained when the horizontal component of the downslope oriented extensional force (FO) triggered by the differential pressure in the overburden is high enough to overcome the horizontal force balance in the ductile layer (FD). The horizontal force balance in the ductile layer is given by: FD ¼ Fp Fc
ð1Þ
where Fc and Fp are forces at the base of the overburden induced by Couette (shear) and Poiseuille (channel) flows of the silicone (i.e. viscous salt substratum) (Fig. 8a). The Couette and Poiseuille forces can be defined as: V ðx2 x1 Þ hc
ð2Þ
hc rgðh1 h2 Þ; 2
ð3Þ
Fc ¼ h
Fp ¼
where h is viscosity of the ductile substratum, r is density of the overburden, g is acceleration due to gravity, V is the horizontal gliding velocity of the overburden; (x2 2 x1) is the length over which Fc is acting; hc silicone thickness; and h1 2 h2 is differential thickness of the overburden. Consider an active basinward listric growthfault and rollover system as part of a margin-scale gravitational spreading system (Fig. 8a). During gravity spreading, most of the accommodation space is generated by the basinward listric growthfault (BB0 ). Therefore, continuous sedimentary loading of the hanging wall block will cause pronounced salt withdrawal beneath the hanging wall block. The thinning of the ductile layer (hc) increases the Couette force and decreases the Poiseuille force (see Eqns 2, 3) which results in increased basal coupling of the hanging wall (FD; Eqn 1). Therefore, in order to maintain displacement along the basinward listric growth-fault (BB0 ), i.e. to overcome the increase of basal coupling, the extensional force FO in the overburden has to increase. The extensional force FO can be achieved by increased differential load due to sedimentation or other external forces (e.g. change of basement tilt; Mauduit et al. 1997). With constant FO, the displacement on basinward listric growthfaults will subsequently decrease in time and will eventually stop due to weld formation beneath the hanging wall rollover.
C. KRE´ZSEK ET AL.
112
Down-slope
Up-slope footwall v1