Reservoir Compartmentalization

Reservoir Compartmentalization The Geological Society of London Books Editorial Committee Chief Editor Bob Pankhurst...

Author: S. J. Jolley | Q. J. Fisher | R. B. Ainsworth | P. J. Vrolijk | S. Delisle

187 downloads 1703 Views 18MB Size Report

This content was uploaded by our users and we assume good faith they have the permission to share this book. If you own the copyright to this book and it is wrongfully on our website, we offer a simple DMCA procedure to remove your content from our site. Start by pressing the button below!
Report copyright / DMCA form

DOWNLOAD PDF

Reservoir Compartmentalization

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) Rick Law (USA) Phil Leat (UK) Nick Robins (UK) Randell Stephenson (UK) Society Books Advisors

Mike Brown (USA) Eric Buffetaut (France) Jonathan Craig (Italy) Reto Giere´ (Germany) Tom McCann (Germany) Doug Stead (Canada) Gonzalo Veiga (Argentina) Maarten de Wit (South Africa)

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., Fisher, Q. J., Ainsworth, R. B., Vrolijk, P. J. & Delisle, S. (eds) 2010. Reservoir Compartmentalization. Geological Society, London, Special Publications, 347. McKie, T., Jolley, S. J. & Kristensen, M. B. 2010. Stratigraphic and structural compartmentalization of dryland fluvial reservoirs: Triassic Heron Cluster, Central North Sea. In: Jolley, S. J., Fisher, Q. J., Ainsworth, R. B., Vrolijk, P. J. & Delisle, S. (eds) Reservoir Compartmentalization. Geological Society, London, Special Publications, 347, 165 –198.

GEOLOGICAL SOCIETY SPECIAL PUBLICATION NO. 347

Reservoir Compartmentalization

EDITED BY

S. J. JOLLEY Shell Canada Energy, Calgary, Canada

Q. J. FISHER University of Leeds, UK

R. B. AINSWORTH University of Adelaide, Australia

P. J. VROLIJK Exxon Mobil Upstream Research Company, Houston, USA

and S. DELISLE Total CSTJF, Pau, France

2010 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 10 000. 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. þ 44 (0)20 7434 9944; Fax þ 44 (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. þ 44 (0)1225 445046, Fax þ 44 (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 2010. 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 Ltd, Saffron House, 6-10 Kirby Street, London EC1N 8TS, UK. Users registered with the Copyright Clearance Center, 222 Rosewood Drive, Danvers, MA 01923, USA: the item-fee code for this publication is 0305-8719/10/$15.00. British Library Cataloguing in Publication Data A catalogue record for this book is available from the British Library. ISBN 978-186329-316-5 Typeset by Techset Composition Ltd, Salisbury, UK Printed by MPG Books, 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. þ 1 800-972-9892 Fax þ 1 802-864-7626 E-mail: [email protected] For individual and corporate orders: AAPG Bookstore, PO Box 979, Tulsa, OK 74101-0979, USA Orders: Tel. þ 1 918-584-2555 Fax þ 1 918-560-2652 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. þ 91 11 2327-9113/2326-4180 Fax þ 91 11 2326-0538 E-mail: [email protected]

Contents JOLLEY, S. J., FISHER, Q. J. & AINSWORTH, R. B. Reservoir compartmentalization: an introduction

1

FOX, R. J. & BOWMAN, M. B. J. The challenges and impact of compartmentalization in reservoir appraisal and development

9

SMALLEY, P. C. & MUGGERIDGE, A. H. Reservoir compartmentalization: get it before it gets you

25

PAÉZ, R. H., LAWERENCE, J. J. & ZHANG, M. Compartmentalization or gravity segregation? Understanding and predicting characteristics of near-critical petroleum fluids

43

CHUPAROVA, E., KRATOCHVIL, T., KLEINGELD, J., BILINSKI, P., GUILLORY, C., BIKUN, J. & DJOJOSOEPARTO, R. Integration of time-lapse geochemistry with well logging and seismic to monitor dynamic reservoir fluid communication: Auger field case-study, deep water Gulf of Mexico

55

GILL, C. E., SHEPHERD, M. & MILLINGTON, J. J. Compartmentalization of the Nelson field, Central North Sea: evidence from produced water chemistry analysis

71

GAINSKI, M., MACGREGOR, A. G., FREEMAN, P. J. & NIEUWLAND, H. F. Turbidite reservoir compartmentalization and well targeting with 4D seismic and production data: Schiehallion Field, UK

89

TOZER, R. S. J. & BORTHWICK, A. M. Variation in fluid contacts in the Azeri field, Azerbaijan: sealing faults or hydrodynamic aquifer?

103

SCOTT, E. D., GELIN, F., JOLLEY, S. J., LEENAARTS, E., SADLER, S. P. & ELSINGER, R. J. Sedimentological control of fluid flow in deep marine turbidite reservoirs: Pierce Field, UK Central North Sea

113

WONHAM, J. P., CYROT, M., NGUYEN, T., LOUHOUAMOU, J. & RUAU, O. Integrated approach to geomodelling and dynamic simulation in a complex mixed siliciclastic –carbonate reservoir, N’Kossa field, Offshore Congo

133

MCKIE, T., JOLLEY, S. J. & KRISTENSEN, M. B. Stratigraphic and structural compartmentalization of dryland fluvial reservoirs: Triassic Heron Cluster, Central North Sea

165

AINSWORTH, R. B. Prediction of stratigraphic compartmentalization in marginal marine reservoirs

199

HOVADIK, J. M. & LARUE, D. K. Stratigraphic and structural connectivity

219

YIELDING, G., BRETAN, P. & FREEMAN, B. Fault seal calibration: a brief review

243

FREEMAN, S. R., HARRIS, S. D. & KNIPE, R. J. Cross-fault sealing, baffling and fluid flow in 3D geological models: tools for analysis, visualization and interpretation

257

IRVING, A. D., CHAVANNE, E., FAURE, V., BUFFET, P. & BARBER, E. An uncertainty modelling workflow for structurally compartmentalized reservoirs

283

VAN HULTEN, F. F. N. Geological factors effecting compartmentalization of Rotliegend gas fields in the Netherlands

301

vi

CONTENTS

CORONA, F. V., DAVIS, J. S., HIPPLER, S. J. & VROLIJK, P. J. Multi-fault analysis scorecard: testing the stochastic approach in fault seal prediction

317

RICHARDS, F. W., VROLIJK, P. J., GORDON, J. D. & MILLER, B. R. Reservoir connectivity analysis of a complex combination trap: Terra Nova Field, Jeanne d’Arc Basin, Newfoundland, Canada

333

Index

357

Reservoir compartmentalization: an introduction S. J. JOLLEY1*, Q. J. FISHER2 & R. B. AINSWORTH3 1

Shell Canada Energy, 400 4th Avenue S.W., Calgary, Alberta T2P 0J4, Canada

2

Centre for Integrated Petroleum Engineering and Geoscience, School of Earth and Environment, University of Leeds, Leeds, LS2 9JT, UK

3

Australian School of Petroleum, University of Adelaide, Adelaide 5005, Australia *Corresponding author (e-mail: [email protected])

Reservoir Compartmentalization – the segregation of a petroleum accumulation into a number of individual fluid/pressure compartments – occurs when flow is prevented across ‘sealed’ boundaries in the reservoir. These boundaries are caused by a variety of geological and fluid dynamic factors, but there are two basic types: ‘static seals’ that are completely sealed and capable of withholding (trapping) petroleum columns over geological time; and ‘dynamic seals’ that are low to very low permeability flow baffles that reduce petroleum crossflow to infinitesimally slow rates. The latter allow fluids and pressures to equilibrate across a boundary over geological time-scales, but act as seals over production time-scales, because they prevent crossflow at normal production rates – such that fluid contacts, saturations and pressures progressively segregate into ‘dynamic’ compartments. Thus, reservoir compartmentalization impacts the volume of moveable (produceable) oil or gas that might be connected to any given well drilled in a field, which restricts the volume of reserves that can be ‘booked’ for that field. Booking of reserves is tightly regulated by government authorities because it is a key measure used by stock analysts and investors to value an oil company. This places reservoir compartmentalization studies, and the predictive science and technology applied to them, at the heart of company valuation. Unexpected compartmentalization can also seriously impact the profitability of a field: with more data acquisition, more study, more wells, more time being required to produce less oil and gas than was originally anticipated. In extreme cases, this might even lead to early field abandonment. However, unexpected or misunderstood reservoir compartmentalization has been an industrywide experience for over 30 years, and it is clear that there is great value in learning from past, often expensive mistakes (e.g. Smith 2008). This learning process has often driven developments in geoscience, engineering and related

technology – enabling operating companies to identify and predict ‘new’ untapped volumes in old fields, make general improvements to field management (e.g. Gainski et al.; Wonham et al.), and apply this knowledge to other similar, but lessmature fields in their portfolio. Reservoir compartmentalization is therefore a major uncertainty that should be accurately assessed during the appraisal of petroleum reservoirs, in order to avoid unexpected compartmentalization at the production stage. In the past few decades, there have been major advances in data, detection and surveillance methods, and our geoscience and reservoir engineering approaches. However, reservoir compartmentalization can still be underestimated during field appraisal, and can still give surprises that force a re-think of the field development and production plan (e.g. Smith 2008). This is not necessarily a defect in the science or technology – but, rather, it may also be a result of ineffectual data appraisal or discipline integration within subsurface workflows. Thus, it is not unusual for subsurface teams to place too much emphasis on one aspect of the evidence, or to make early assumptions that bias data acquisition, analyses and interpretations later on. For example, Fisher & Jolley (2007) point out that non-existent faults or improbable fault seal capacities might be invoked to explain variable petroleum contacts and dynamic fluid/pressure behaviour – when an alternative or combination of factors, or other unrelated explanations might be more appropriate for the data (e.g. hydrodynamic tilting of petroleum contacts, Tozer & Borthwick). Research initiatives and collaborations between oil companies, service groups and academic institutions vary considerably in scope and content, and there is often a creative tension between the competitive advantages, motivations and knowledge gaps perceived by the sponsoring companies and their research partners. In addition to this, the scientific insights that can be gained from the

From: Jolley, S. J., Fisher, Q. J., Ainsworth, R. B., Vrolijk, P. J. & Delisle, S. (eds) Reservoir Compartmentalization. Geological Society, London, Special Publications, 347, 1– 8. DOI: 10.1144/SP347.1 0305-8719/10/$15.00 # The Geological Society of London 2010.

2

S. J. JOLLEY ET AL.

experience of operating compartmentalized fields, are often under-valued and/or infrequently published. However, it has become increasingly apparent that reservoir compartmentalization is a consequence of many factors, such that the successful approach is almost always the integrated one. The aim of this book (and the conference it was inspired by) has therefore been to bring papers together from different areas of the compartmentalization theme, in order to see the ‘bigger picture’; and to highlight both the necessity for effective specialist discipline integration, and the value of learning from operational experience of reservoir compartmentalization. This is reflected in the integrated nature of the data and science described in the papers within this book. This particular paper aims to provide the reader with a brief introduction into the topic of reservoir compartmentalization; a basic context for the papers; and a taste of their content (papers cited in bold here, are those presented in this book). For ease of discussion, we have sub-divided the papers in this introduction, into the dominant themes of: (a) detection and monitoring of compartmentalization; (b) stratigraphic and mixed-mode compartmentalization; and (iii) fault dominated compartmentalization. The paper by Fox & Bowman presents a highlevel summary of technical challenges and lessons learnt from practical experience of appraising and operating compartmentalized fields. The paper, which speaks on behalf of the oil and gas industry, gives BP’s perspective, written in a way that will resonate with many readers. The authors emphasize the need for compartmentalization to be evaluated in the early appraisal stages of the E&P workflow, in order to avoid surprises later in the development stage. They explain the importance of acquiring surveillance data early and continuously throughout the field life-cycle; and they describe the value of integrated workflows that capture the range of subsurface scenarios and uncertainties. The style of static and dynamic compartmentalization varies dramatically between different fields and geological settings – and the compartmentalization behaviour of a field evolves, driven by production-induced geomechanical and relative permeability changes for example. Inadequate data, inflexible development planning, and over-standardization of workflows and toolkits can therefore hamper efforts to evaluate, track and respond to the evolving, situation-specific nature of reservoir compartmentalization. Finally, the authors urge us to exercise our science and experience in assessing and planning for reservoir compartmentalization – pointing out that reservoir modelling tools are ‘. . . not a proxy for generating well thought-out geological scenarios’.

Detection and monitoring of compartmentalization Fluid properties (e.g. pressure, chemistry, density, viscosity, etc.) are commonly used in an attempt to assess the level of compartmentalization and identify the position of compartment boundaries during the appraisal and production of petroleum reservoirs. Indeed, it is often assumed that a reservoir is compartmentalized as soon as fluid-property differences are observed between wells or between reservoir units in a single well. However, this assumption can be misleading. Four papers published within this volume provide a more sophisticated insight as to how fluid-property data should be interpreted in terms of their implications for reservoir compartmentalization. Chuparova et al.; Gill et al.; Paez et al.; and Smalley & Muggeridge provide excellent insights into how fluid data can be used to improve understanding of reservoir compartmentalization from the appraisal to the production stage of development – and also show the importance of modelling fluid data to avoid jumping to erroneous conclusions about the level of reservoir compartmentalization. In particular, these papers show how differences in fluid composition may exist within reservoirs that are not significantly compartmentalized (also Scott et al.). It should also be emphasized that the contrary is also true – highly compartmentalized reservoirs may have very uniform fluid compositions. In other words, provided fluid data is interpreted with great care, it can be a cheap and very useful tool to identify compartment boundaries and assess the level of compartmentalization of a field. Smalley & Muggeridge describe the impact of reservoir compartmentalization on oil recovery. They argue that it is important to identify the extent of compartmentalization at an early stage in the appraisal and development of a reservoir. The authors use simple analytical equations to determine the time taken for a variety of fluid properties (e.g. pressure, density, composition, etc.) to equilibrate. Some properties, such as pressure differences within aquifers, are shown to equilibrate very rapidly (e.g. ,10 years). On the other hand, other fluid properties, such as the isotopic composition of pore fluids, would be expected to take tens of millions of years to equilibrate throughout a reservoir. The implication of these calculations is that one would expect to find differences in fluid properties that are slow to equilibrate even in reservoirs that are not compartmentalized. However, if there are differences in fluid properties that should equilibrate rapidly, they should be taken as a serious indication of reservoir compartmentalization. Paéz et al. present several examples from offshore West Africa that highlight the importance of thoroughly

INTRODUCTION

understanding Pressure –Volume–Temperature (PVT) behaviour, when using petroleum fluid phase compositional data in early assessment of compartmentalization in reservoirs containing critical or near-critical fluids. It is argued that processes such as gravity segregation can occur within near-critical fluids, which can result in a reservoir containing a spectrum of fluid compositions from the crest to the petroleum –water contact. In such cases, it might be easy to misinterpret the fluid data as indicating compartmentalization between wells, whereas careful application of PVT modelling would show that compositional differences can be expected even in a non-compartmentalized reservoir. An example is provided from a reservoir in which the fluids taken from three wells had very large differences in their gas –oil ratio (i.e. 2500 to 8500). A gravity-segregation model can explain these data in terms of a highly connected reservoir with a gas condensate gradually transitioning to a volatile oil. Chuparova et al. describe how the integration of time-lapse organic geochemistry with 4D seismic data can be used to improve understanding of fluid communication within petroleum reservoirs. The authors illustrate the method using data from the Auger Field in the deep water Gulf of Mexico. Fluids were regularly sampled and analysed from wells across the field. Statistical analysis of the evolution of fluid compositions with time provides evidence for the gradual mixing of oil and condensate from two reservoir units that were believed to be separate prior to the start of production. The geochemistry results were integrated with 4D seismic to up-date both the static and dynamic geocellular reservoir models, which provided the basis for improving the field development strategy. Gill et al. present an example of how time-lapse water geochemistry data can be used to improve understanding of reservoir compartmentalization. The paper focuses on the use of the inorganic composition of produced water, particularly the chloride ion concentration, from the turbidite reservoirs of the Nelson Field in the North Sea. The chemistry of the produced water is used to identify drainage cells within the field. The data is then used to create drainage charts that provide a framework for identifying regions within the reservoir that are likely to contain bypassed oil. The time-lapse (4D) seismic technique involves repeating 3D seismic surveys over a producing field – replicating exactly the acquisition and processing that was used for the original ‘base’ survey. Given this ‘repeatability’, differences (seismic amplitudes, two-way-time arrivals) between the base and monitor data cubes are related to production-induced fluid and pressure effects (e.g. Landrø 2001; Tura et al. 2005) – that are used to map fluid movement, identify flow barriers, and

3

‘condition’ integrated reservoir modelling workflows. Gainski et al. describe the 4D seismic method, and its integration with other surveillance data to identify, map and ‘condition’ modelling of compartmentalization in the turbidite reservoirs of the Schiehallion Field (west of Shetlands, UKCS). They discuss how this 4D integrated reservoir modelling, and a dedicated surveillance data acquisition programme has helped to decode the unexpected complexity of compartmentalization, caused by a combination of channel and sheet turbidite sandbody architectures and sealing faults. This work has extended field life and ultimate recovery by identifying unswept infill drilling targets. The paper provides a fascinating account of lessons learnt, and technologies and workflows developed from the reservoir compartmentalization experience. Variation of petroleum–water contact (PWC) heights around a field can result from hydrodynamic tilting of the contact, driven by water flowing out of a pressure dynamic acquifer (e.g. Dennis et al. 2000, 2005). Such contact variations could easily be misinterpreted as evidence for compartmentalization. Tozer & Borthwick describe pressure variations and the tilting of a PWC by this phenomenon, in the fluvio-deltaic reservoirs of the Azeri Field, Caspian Sea. Earlier workers on the field had proposed fault sealing to explain water pressure and oil –water contact (OWC) variations, however, the lack of seismic scale faulting and the high net-to-gross of the reservoir sequence preclude a fault sealing mechanism. Having projected the pressure data to a common seal-level datum, the authors were able to fit a planar grid surface to the data to determine the gradient and trend direction of the aquifer overpressure; and to convert the grid surface to a calculated OWC. There is good correspondence between the calculated OWC surface and a ‘best-fit plane’ that passes through the petroleum contacts measured in the field; and good correspondence between the line of intersection of that best fit plane with top reservoir – and the line of an abrupt change in seismic amplitude. The authors conclude that the contact was tilted by water flowing outwards from the compacting, overpressured basin centre. The paper provides a splendid example where a compartmentalized interpretation of fluid data variations might in fact be misleading.

Stratigraphic and mixed-mode compartmentalization In the second section of the book stratigraphic and mixed-mode (combined stratigraphic and structural) compartmentalization aspects (Bailey et al. 2002;

4

S. J. JOLLEY ET AL.

Ainsworth 2006) are addressed in a series of papers that range from detailed studies of compartmentalized fields to suggested methods for predicting compartmentalization potential, and theoretical modelling studies. Some fields that have hydrodynamically tilted PWCs may also be partially compartmentalized – these being the most difficult to interpret. For example, the explanation for variation in PWC heights in the salt-domed Pierce Field, central North Sea, has been an active debate for over a decade. This is an unusual and interesting case, because the debate can be tracked within public domain literature. The PWC variation has previously been explained by simple hydrodynamic tilting (Dennis et al. 2000, 2005); fault sealing (e.g. Hempton et al. 2005); or a combination of similar mechanisms (e.g. Fisher & Jolley 2007 and references therein). The paper by Scott et al. describes the most recent evolution of ideas on Pierce. They show how the depositional architecture of the deep marine Forties Sandstone reservoir, was influenced by contemporary salt diapir movement. In addition to PWC variations, the authors describe differences in fluid composition between wells in the field. The wells, drilled into the steep flanks of the field’s twin salt domes, sample different parts of the fluid column – with variations reflecting filling history and compositional segregation rather than indicating the presence of compartments. The authors attribute the PWC variations and stratified oil compositions to the interplay between the architecture of the channelized reservoir, the petroleum charge/filling history, and hydrodynamic flow of the aquifer. They conclude that the dynamic aquifer water and production flow are mainly controlled by the sand-body architectures (with limited influence from fault seal in one quadrant of the field) – to create a modified tilted contact pre-production and local flow compartmentalization during production. Wonham et al. provide an important dataset – in which they detail a study of compartmentalization in a mixed siliciclastic and carbonate marginal marine reservoir from offshore Congo, west Africa. Compartmentalization in the N’Kossa Field is attributed to combined stratigraphic and structural mechanisms. Multiple laterally extensive stratigraphic barriers to vertical fluid flow are recognized whilst large throw faults compartmentalize the reservoir laterally. However, prior to initial production, compartmentalization was not considered to be an issue in this field. After eight years of production, multiple problems were being encountered and it was very clear that the field was indeed compartmentalized. This study therefore clearly demonstrates the value of assessing potential compartmentalization issues prior to field

development. However, it also illustrates how relatively late recognition of compartmentalization during production can sometimes be addressed and development plans changed to mitigate these issues. In this example, a very detailed field review and integration of static and dynamic data permitted a revision of the reservoir management plan and modifications to the field development strategy. McKie et al. examine the stratigraphicand fault-related controls on compartmentalization of Triassic dryland fluvial reservoirs in the HPHT Heron cluster fields (Central North Sea). They describe an integrated study of the sedimentology and structural geology, 3D– 4D seismic, fluid pressure and geochemistry within the field cluster and the wider region. They conclude that reservoir compartmentalization is essentially stratigraphic in nature, being controlled by the deposition of a laterally extensive shale package during a major regional change in palaeo-environmental conditions. The fields in the cluster all show the same change in fluid chemistry and fluid pressure across this shale boundary – with the integrity of the stratigraphic compartments being preserved despite faulting in the fields. The authors point to extrinsic environmental factors as being a key control on the deposition of potential stratigraphic barriers at major gradational sequence boundaries in dryland fluvial systems. They also conclude that that smearing of the composite sand-shale reservoir sequence during early fault movement in the fields was the likely deformation mechanism, explaining why the integrity of the stratigraphic compartments was preserved despite the widespread intrareservoir faulting (i.e. the faults do not breach the shale boundaries or induce significant porosity/ permeability collapse in the sands). Since stratigraphic connectivity and compartmentalization is a complex three-dimensional problem, 3D reservoir models are the norm for analysing these issues. There are three basic steps that are required before a reliable 3D reservoir model can be generated: (a) determine key depositional processes and environments; (b) determine the range of sand-body and heterogeneity geometries; and (c) generate multi-scenario 3D–4D depositional concept models. Ainsworth presents a methodology for predicting and ranking the stratigraphic compartmentalization potential of siliciclastic marginal marine deposits. The author reviews and synthesizes a number of ideas that have until now been scattered within the literature – and integrates them with his own work into a clear, conceptual framework. The paper uses a hierarchical approach, displayed as a matrix, and relies on depositional process-based interpretations, combined with accommodation-to-sediment supply ratio information. This results in a tool that is easy

INTRODUCTION

to understand and use, which ranks stratigraphic compartmentalization potential. This ranking approach is particularly useful when a series of prospects are available or perforation strategies are being designed for multiple reservoir zones within a field. Hovadik & Larue analysed and modelled both stratigraphic and structural connectivity. They suggest that static connectivity is the best parameter to assess, and that dynamic connectivity should be avoided where possible since it is dependent upon fluid type, permeability heterogeneity, time and other factors, and hence confuses connectivity with tortuosity, sweep- and displacementefficiency. They also suggest that from percolation theory experiments, the most important factor addressing stratigraphic connectivity is a netto-gross threshold ranging from 30 to 45%. For a range of depositional geometries, their studies indicate that above this threshold connectivity is generally good, whilst below this threshold connectivity is generally poor. A flow diagram for analysing connectivity uncertainty is presented. This diagram can be utilized as a useful tool for defining key uncertainties regarding both potential stratigraphic and structural connectivity of a reservoir, and could be very useful in the appraisal and data acquisition phase of field development.

Fault dominated compartmentalization Fault compartmentalization is generally considered on two levels: static (geological time-scale) sealing in which the fault sealing capacity is sufficient to trap a petroleum column; and dynamic (production time-scale) sealing in which a fault might ‘leak’ on geological time-scales, but prevents cross-faultflow at commercial production rates. There are many published and proprietary approaches to fault seal analysis, and a huge literature has developed on the topic. Sealing and baffling of crossfault-flow is controlled by juxtaposition between reservoir and non-reservoir rocks across a fault. This, together with the 3D geometry of the reservoir architecture and its structural form (e.g. folding), defines the essential ‘plumbing’ of the field. Cross-fault-flow is also controlled by capillary sealing or baffling of flow due to the reduced porosity, permeability and relative permeability of fault rocks to different fluid phases. It is now well established that there is a relationship between sealing capacity of a fault and the clay content of lithic fragments and fault rocks it contains (e.g. see Fisher & Knipe 2001; Sperrevik et al. 2002 and references therein). Consequently, many fault seal algorithms calculate the clay content of seismic-scale faults by integrating fault throw with the clay content of host stratigraphy – e.g. Shale

5

Gouge Ratio (SGR, Yielding et al. 1997); Clay Smear Potential (CSP, Bouvier et al. 1989); Shale Smear Factor (SSF, Lindsay et al. 1993); Probabilistic Shale Smear Factor (PSSF, Childs et al. 2007); Effective Shale Gouge Ratio (ESGR, Freeman et al.). Yielding et al. provide a review of fault rock capillary seal mechanisms and industry standard clay-content prediction algorithms, and they compare and contrast deterministic and empirical approaches to predicting static fault sealing potential. The deterministic approach correlates laboratory-measured capillary threshold pressure and clay content of fault rock samples – with predicted clay content of seismic-scale faults (e.g. by using modelled SGR values as a proxy for clay content). The empirical approach compares SGR values modelled on known sealing faults in a field, with PWC and pressure differences measured in the field. Building on the work presented in Yielding (2002), the authors show that there is some similarity in the results from these two calibration methods, which might be coincidental, given the heterogeneous nature of fault zones. They conclude that whilst further work is needed to improve understanding of fault zone heterogeneity, the deterministic and empirical approaches are not mutually exclusive – but, rather, on a practical level they provide alternative scenarios within an acceptable range of subsurface uncertainty. In recent years, there have been major advances in our understanding and modelling of 3D depositional architectures and the heterogeneity of fault damage zones and fracture arrays. It is now common practice for appraisal and development teams to work on a range of complex geological concepts and for them to construct detailed model scenarios – in order to capture subsurface uncertainty. Freeman et al. address the characterization and visualization of fault zone heterogeneity in fault seal analysis and reservoir modelling. It has been possible to incorporate some of these complexities into existing modelling tools (e.g. Manzocchi et al. 2010). However, the authors contend that new analytical and intuitive visualization methods are needed to enhance or replace those that were developed over a decade ago when requirements were simpler. The paper presents enhanced visualization techniques – and provides a formal definition of the ESGR algorithm (evolved from SGR) – in the context of a detailed comparative review of commonly used clayprediction algorithms (SGR, ESGR, CSP, SSF, PSSF). The authors also introduce ‘intuitive’ parameters, such as fault hydraulic resistance, and Effective Cross Fault Permeability and Transmissibility (ECFP, ECFT) – that take into account the flow properties of the host reservoir, and the complex flow properties of the fault zone, and the

6

S. J. JOLLEY ET AL.

fault-damaged hanging wall and footwall rock volumes. When tested in a flow simulator, these parameters provide a more robust correlation with cross-fault fluid flux than host- or fault rock permeabilities alone. Irving et al. note that a key problem with modelling fluid flow in petroleum reservoirs is that subsurface data used to populate models is limited to spatially extensive low resolution seismic data or sparse high resolution well data. A workflow is presented to fill this sampling gap by the generation of multiple realizations in which fault geometry and properties can be varied in geological models so the impact of these uncertainties on production forecasts can be assessed. Static properties such as total connected gas volume are calculated for each model. A selection of these models was then incorporated into production simulation models to allow the assessment of structural uncertainty on predictions of the dynamic behaviour of the reservoir. The workflow is applied to a North Sea reservoir, in which it is shown that fault geometry and permeability are the most important properties although the relative significance of these properties varies throughout field life. Although a reasonably good ‘history match’ (i.e. a correspondence between numerically simulated and actual time-lapse production data) is achieved using the workflow, the paper indicates that further improvements may be achieved by incorporating stress and saturation dependent fault properties into the simulation model. The giant and prolific gas fields of the Permian Rotliegend Group were discovered in the Netherlands c. 50 years ago, and the Energie Beheer Netherland (EBN) organization was formed to represent the national interest in developing these fields and managing their production. Today, EBN participates in over 250 fields, and has been able to study the compartmentalization of these high net-to-gross fluvio-aeolian sandstones across a wider acreage than any individual operating company has available. Van Hulten provides a summary of EBN’s review of the mechanisms controlling reservoir compartmentalization in the Rotliegend fairway. Compartmentalization was not considered in early development plans or recognized as a common experience for over a decade after initial development drilling in the region. Since that time, however, there has been an active proprietary and literature debate on Rotliegend compartmentalization. The study that Van Hulten describes, was aimed at identifying the key controls on compartmentalization, and mapping them out across the Netherlands acreage. He describes a basic consensus amongst industry and academic workers, that there are three major controls: (a) the dominant mechanism being fault compartmentalization – with juxtaposition of reservoir to non-reservoir

rocks accounting for c. 75% of the compartment boundaries, and sealing at sand-sand fault contacts (caused by capillary rise of water into tight cataclastic fault rocks) accounting for c. 25% of the boundaries; (b) stratigraphic barriers are developed in a narrow east –west belt across the region, where the upper reservoir sands shale-out towards a sabkha lake in the basin centre; and (c) a similar belt of diagenetic impairment, with authigenic mineral growth being associated with specific facies belts developed close to the sabkha shoreline. There is a general literature consensus, that structural configuration, depositional architectures and fault juxtapositions define the basic ‘plumbing’ of a reservoir – and that this is a critical framework for understanding and modelling reservoir compartmentalization in any reliable sense (e.g. see worked ‘real field’ examples in Jolley et al. 2007a, b). Approaches such as the ‘stochastic multi-fault’ method (James et al. 2004; Corona et al.), and Reservoir Connectivity Analysis (RCA, Vrolijk et al. 2005; Richards et al.) emphasize the control of the basic fluid pressure ‘plumbing’ of the reservoir system on compartmentalization behaviour. Whilst many methodologies have been proposed for estimating the fluid flow properties of faults over geological time-scales (e.g. Allan 1989; Knipe et al. 1997; Yielding et al. 1997; James et al. 2004), few studies are published that discuss the success of these methodologies. The paper by Corona et al. is therefore refreshing in that as well as describing a methodology for assessing the sealing capacity of faults, it also provides statistics as to how successful the methodology has been when applied to a range of ‘real field’ examples. The fault seal methodology uses a stochastic juxtaposition-based approach (James et al. 2004), that identifies and tracks the myriad of spill –leak points within 3D geological models, to provide a quantitative prediction of petroleum contacts in structurally complex reservoirs with stacked sands and a variety of fault densities. The study is based on the analysis of 41 prospects from different parts of the world, with detailed description of examples from the Dutch Rotliegend fairway (cf. Van Hulten). Comparison of pre-drill predictions and post-drill results in these cases, showed that the methodology had a 76% success rate – leading the authors to conclude that the most important fault seal process is cross-fault juxtaposition, and that examples of capillary sealing by fault rocks are fewer and more difficult to predict. Richards et al. present a detailed ‘real field’ example of Reservoir Connectivity Analysis (RCA, Vrolijk et al. 2005), being used to define connections between reservoir compartments in the structurally and stratigraphically complex Terra Nova Field, offshore eastern Canada. The reservoir is composed of a stack of

INTRODUCTION

six fluvial sand packages, separated by laterally persistent shales. Major compartments are defined by faults and stratigraphic barriers, with diagenetic mechanisms also playing a role in the compartmentalization of the field. The distribution of reservoir fluids in this field is unusual when compared to others in the Jeanne d’Arc Basin. However, the connectivity diagram approach of RCA allows an understanding of the fluid distributions by systematically describing the multiple compartments and connections within the field and thereby driving key reservoir management decisions. This book and the energy to create it, were inspired by the presentations and stimulating debate that occurred at a 2-day international conference on ‘Reservoir Compartmentalization’ (March 2008, The Geological Society, Burlington House, London). SJJ thanks Shell and John Marshall, in particular, for encouraging and supporting this enterprise. We thank our authors for investing their time and science in this book – and for their perseverance in getting their manuscripts and revised manuscripts through their company external publication approval processes – and those of their partner companies . . . we appreciate this took time and effort. We also give sincere thanks to the people who gave their free time and expertise to reviewing manuscripts for us: Jennifer Adams, Andy Aplin, Wayne Bailey, Mark Bentley, Knut Bjorlykke, Bryan Bracken, Alton Brown, Chris Cornford, Tony Crook, Russell Davies, Steve Dee, Pete D’Onfro, Fred Dula, John Fisher, Quentin Fisher, Martha Gerdes, Carlos Gattoni, Gary Hampson, Peter Hennings, Steve Jolley, Bob Krantz, Paul Mankiewicz, John Marshall, Tom McKie, Trey Meckel, Don Medweffe, Steve Naruk, Tim Needham, Michael Poppelreiter, John Snedden, Rob Staples, Dick Swarbrick, Woluter van der Zee, Paul Ventris, Jennifer Wadsworth, Chris Wibberley, Peter Winefield, and several referees who asked to remain anonymous. Colour production of this book was funded from generous sponsorship provided to the original conference for this purpose by BP, Chevron, Total, Shell, ExxonMobil, StatoilHydro, Rock Deformation Research Ltd, and Badley Geoscience Ltd.

References Ainsworth, 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. Allan, U. S. 1989. Model for hydrocarbon migration and entrapment within faulted structures. American Association of Petroleum Geologists Bulletin, 73, 803–811. Bailey, W. R., Manzocchi, T. 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. Bouvier, J. D., Sijpesteijn, K., Kleusner, D. F., Onyejekwe, C. C. & Van der pal, R. C. 1989.

7

Three-dimensional seismic interpretation and fault sealing investigations, Nun River field, Nigeria. American Association of Petroleum Geologists Bulletin, 73, 1397–1414. Childs, C., Walsh, J. J. et al. 2007. Definition of a fault permeability predictor from outcrop studies of a faulted turbidite sequence, Taranaki, New Zealand. In: Jolley, S. J., Barr, D., Walsh, J. J. & Knipe, R. J. (eds) Structurally Complex Reservoirs. Geological Society, London, Special Publications, 292, 235– 258. Dennis, H., Baillie, J., Holt, T. & Wessel-berg, D. 2000. Hydrodynamic activity and tilted oil–water contacts in the North Sea. In: Ofstad, K., Kittilsen, J. E. & Alexander-Marrack, P. (eds) Improving the Exploration Process by Learning from the Past. Norwegian Petroleum Society, Special Publication, Elsevier, Singapore, 9, 171–185. Dennis, H., Bergmo, P. & Holt, T. 2005. Tilted oil– water contacts: modelling the effects of aquifer heterogeneity. In: Dore´, A. G. & Vining, B. A. (eds) Petroleum Geology: North-West Europe and Global Perspectives. Proceedings of the 6th Petroleum Geology Conference. Geological Society, London, 145– 158. Fisher, Q. J. & Jolley, S. J. 2007. Treatment of faults in production simulation models. In: Jolley, S. J., Barr, D., Walsh, J. J. & Knipe, R. J. (eds) Structurally Complex Reservoirs. Geological Society, London, Special Publications, 292, 219–233. Fisher, Q. J. & Knipe, 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. Hempton, M., Marshall, J., Sadler, S., Hogg, N., Charles, R. & Harvey, C. 2005. Turbidite reservoirs of the Sele Formation, Central North Sea: geological challenges for improving production. In: Dore´, A. G. & Vining, B. A. (eds) Petroleum Geology: North-West Europe and Global Perspectives. Proceedings of the 6th Petroleum Geology Conference. Geological Society, London, 449– 459. James, W. R., Fairchild, L. H., Nakayama, G. P., Hippler, S. J. & Vrolijk, P. J. 2004. Fault-seal analysis using a stochastic multifault approach. American Association of Petroleum Geologists Bulletin, 88, 885– 904. Jolley, S. J., Barr, D., Walsh, J. J. & Knipe, R. J. (eds) 2007a. Structurally Complex Reservoirs. Geological Society, London, Special Publications, 292. Jolley, S. J., Dijk, H., Lamens, J. H., Fisher, Q. J., Manzocchi, T., Eikmans, H. & Huang, Y. 2007b. Faulting and fault sealing in production simulation models: Brent province, northern North Sea. Petroleum Geoscience, 13, 321–340. Knipe, R. J., Fisher, Q. F. et al. 1997. Fault seal analysis: successful methodologies, application and future directions. In: Møller-Pedersen, P. & Koestler, A. G. (eds) Hydrocarbon Seals: Importance for Exploration and Production. Norwegian Petroleum Society, Special Publication, Elsevier, Singapore, 7, 15– 40. Landrø, M. 2001. Discrimination between pressure and fluid saturation changes from time-lapse seismic data. Geophysics, 66, 836– 844.

8

S. J. JOLLEY ET AL.

Lindsay, N. G., Murphy, F. C., Walsh, J. J. & Watterson, J. 1993. Outcrop studies of shale smears on fault surfaces. In: Flint, S. T. & Bryant, A. D. (eds) The Geological Modelling of Hydrocarbon Reservoirs and Outcrop. International Association of Sedimentologists, Special Publication, 15, 113– 123. Manzocchi, T., Childs, C. & Walsh, J. J. 2010. Faults and fault properties in hydrocarbon flow models. Geofluids, 10, 94– 113. Smith, P. 2008. Studies of United Kingdom Continental Shelf fields after a decade of production: how does production data affect the estimation of subsurface uncertainty? American Association of Petroleum Geologists Bulletin, 92, 1403– 1413. Sperrevik, S., Gillespie, P. A., Fisher, Q. J., Halvorsen, T. & Knipe, R. J. 2002. Empirical estimation of fault rock properties. In: Koestler, A. G. & Hunsdale, R. (eds) Hydrocarbon Seal

Quantification. Norwegian Petroleum Society, Special Publication, Elsevier, Singapore, 11, 109– 125. Tura, A., Barker, T. et al. 2005. Monitoring primary depletion reservoirs using amplitudes and time shifts from high-repeat seismic surveys. The Leading Edge, 24, 1214–1221. Vrolijk, P. J., James, B., Myers, R., Maynard, J., Sumpter, L. & Sweet, M. 2005. Reservoir connectivity analysis – defining reservoir connections and plumbing. Society of Petroleum Engineers, SPE Paper 93577. Yielding, G. 2002. Shale Gouge Ratio – calibration by geohistory. In: Koestler, A. G. & Hunsdale, R. (eds) Hydrocarbon Seal Quantification. Norwegian Petroleum Society, Special Publication, Elsevier, Singapore, 11, 1 –15. Yielding, G., Freeman, B. & Needham, D. T. 1997. Quantitative fault seal prediction. American Association of Petroleum Geologists Bulletin, 81, 897– 917.

The challenges and impact of compartmentalization in reservoir appraisal and development R. J. FOX1* & M. B. J. BOWMAN2 1

BP Exploration, 501 Westlake Park Boulevard, Houston, Texas, 77450, USA

2

BP Exploration, Chertsey Road, Sunbury on Thames, Middlesex, TW16 7LN, UK *Corresponding author (e-mail: [email protected]) Abstract: As an industry we have been poor at identifying and predicting the effect of reservoir compartmentalization on fluid flow throughout field life. In the context of harder-to-find reserves and rising development costs it is vital to have a well-rounded strategy in place to identify and mitigate uncertainties and risks associated with compartmentalization. The key challenge of today is therefore to improve predictive capability. Historically we have relied too heavily on ‘single complex’ linear modelling approaches to understand the impact of compartmentalization. Lately, we have begun to place a greater emphasis on using a forensic level of reservoir analysis coupled with the use of dynamic signals from production data and constant down hole monitoring of fluid-type, pressures and temperatures. Evidence is mounting that as fields deplete they evolve mechanically over production time-scales leading to changes in fault behaviour, stress configuration, compaction and hence compartmentalization; such factors are commonly not predicted at start-up. Our challenge has been to develop toolkits and workflows which integrate an appropriate range of geological models iteratively coupled with dynamic data. We need to develop analytical approaches that enable real-time updates from the evolving reservoir & fluid system to iteratively modify our models and improve their predictive power. This will allow us to make better-informed decisions at every stage of field life.

The upstream exploration and production business is commonly driven by the expectation that we can now get more from our older existing fields, continue with positive reserve replacement ratios (the ratio between: reserves booked from discoveries, field extensions and improved recovery schemes; and production over a given period), manage rising costs, and make better informed decisions when faced with increased geological complexity in new developments. Thus, production teams are pressured to deliver on promises derived from geological models and production rate profiles experienced and/or predicted earlier in a field’s Appraisal or Development phase. There are, however, many examples in the literature where compartmentalization has unexpectedly impacted field development (e.g. Knott 1993; Gibson 1994; Leveille et al. 1997; Knai & Knipe 1998; Hesthammer et al. 2002; Zoback & Zinke 2002; Porter et al. 2004; Barr 2007; Smith 2008; Gainski et al. 2010; McKie et al. 2010). It follows that successful production optimization is controlled by our ability to characterize and predict reservoir performance and fluid flow behaviour over the life-cycle of a hydrocarbon field. In many cases, this understanding comes through experience, relatively late in field life (e.g. Barkved et al. 2003; Clifford et al. 2005; Moulds et al. 2005; Barr et al. 2007;

Gainski et al. 2010). Understanding the impact of compartmentalization on fluid flow and the ability to predict it prior to drilling wells is a long-term intention of many companies, but it is consistently under-evaluated and underestimated. It would appear that we still have some way to go before we consistently predict the impact of faults, fractures, stress and other reservoir heterogeneities ahead of the drill bit with any certainty. As an industry, we commonly fail to realize at the appraisal or development stages, the full risks from sedimentological, structural, geochemical, geomechanical or time-dependent production-induced compartmentalization (Corrigan 1993; Smalley & England 1994; Larter & Aplin 1995; Heffer et al. 1995; Smalley & Hale 1996; Weber 1997; Reynolds et al. 1998; Milkov et al. 2007; Smalley & Muggeridge 2010). Operating companies are driven by business challenges to deliver rapid solutions that commonly do not capture the full range of subsurface uncertainties or comprehensively understand the implications of key geological heterogeneities on reserves, production rate and the selection of a development scheme. We commonly fail to collect sufficient observational data (i.e. core or pressure data) at an early stage in the exploration and appraisal programs or conduct adequate integration of data between sub-disciplines. Many of our toolkits and

From: Jolley, S. J., Fisher, Q. J., Ainsworth, R. B., Vrolijk, P. J. & Delisle, S. (eds) Reservoir Compartmentalization. Geological Society, London, Special Publications, 347, 9– 23. DOI: 10.1144/SP347.2 0305-8719/10/$15.00 # The Geological Society of London 2010.

10

R. J. FOX & M. B. J. BOWMAN

workflows are often incomplete, immature in their development or fail to link in a seamless manner that would provide a fully integrated solution. Many times we place an over reliance on modelling and simulation results based on limited or poor quality data. We are now beginning to place a greater emphasis on capturing appropriate ranges of geological scenarios and associated uncertainty ranges, with the incorporation of dynamic signals from production data (Bentley & Smith 2008; Smalley et al. 2008). Evidence is also mounting that as fields deplete under production they evolve mechanically: leading to changes in injectorproducer well communication; fault seal integrity; fault zone conductivity; fracture permeability; near and far-field stress configurations and orientations; and reservoir permeability heterogeneity (Heffer et al. 1995; Heffer 2002; Main et al. 2006; Zhang et al. 2007). These effects are commonly not predicted at the early stages of field development. In order to adequately address these issues we must have a structured approach that encompasses a rigorous early and continuous data collection program; accompanied by a ‘forensic level’ of geological data analysis with a range of geological models linked and calibrated by early time production tests (Drill Stem Tests & Pressure Build-Up tests). There must be full integration between disciplines (geologists, geophysicists, petrophysicists, reservoir engineers, drillers etc), producing a range of multiple coherent geological models that fully explain all the available data, and describe the range of possible scenarios, which can be coupled with dynamic data to enable real-time updates of the evolving reservoir and fluid system over production time-scales. An ability to visualize, rationalize, describe and defend a range of conceptual 3D geological models signifies a team’s readiness to progress to numerical model building for reservoir simulation studies (Bentley & Smith 2008). However, we must be wary of ‘black box’ model solutions, particularly those that only cross-correlate dominant data trends or apply linear extrapolations to inadequate data sets. Thus geological model building tools are not a proxy for generating well thought-out geological scenarios. We advocate the full use of all data sets, rigorous laboratory testing, the incorporation of field analogue data, coupled with the appropriate development of new tools and innovative workflows to support existing tool kits and ‘decision driven’ work flows; all framed by an understanding of the key risks and range of uncertainties associated with these risks. This will enable more informed decisions to be taken at every stage of field life, steer the development strategy, guide the appraisal program, feed well planning and optimize long-term production prediction.

In this keynote paper we discuss the application of a number of aspects to our ‘forensic’ geological analysis approach. We show how our integrated toolkits have evolved and link into the management of geological uncertainty during field appraisal and development and in understanding and predicting compartmentalization.

The integrated reservoir and fluids description toolkit We provide geoscientists and engineers with tools, best practices, lessons learned, guide-lines on fit for purpose approaches, access to the latest training and quality software to improve our models and their predictive powers. Integration of multiple data sets, modelling tools and techniques on seamless common software platforms are vital to enable these first steps to happen (Williams et al. 2004). The use of common processes and characterization of the full range of uncertainty in all our data sets, allows us to develop as complete as possible an understanding of the subsurface and its key risks (Smalley et al. 2008). Many companies follow this approach and hence have developed proprietary approaches and technology within this area to integrate multiple data sets (Fig. 1). However, a distinctiveness of the compartmentalization topic is that it needs to call upon a broad range of toolkits rather than specific workflows, which can be used in a variety of combinations, because of the breadth of issues we face differing from situation to situation. As we explore further into more complex geological environments, the necessity to collect early time dynamic flow data becomes more apparent and should always be considered at the exploration or early appraisal stage. Direct reservoir pressure measurements and formation fluid sampling, using wire-line tools such as a Modular Formation Dynamics Tester (MDT), can collect accurate pressure measurements and high quality fluid samples to infer the presence of baffles or barriers in a reservoir (Pelissier-Combescure et al. 1979). Together with reservoir fluid flow-back data from completion testing or early well-flow tests it is vital to calibrate static models created early in the appraisal stage of field life (Smalley & Hale 1996; Brown 2003). Several areas of significant recent advances within reservoir and fluids description of structurally complex reservoirs have occurred and are the subject of a number of special thematic publications (e.g. Møller-Pedersen & Koestler 1997; Coward et al. 1998; Jones et al. 1998; Nieuwland 2003; McClay 2004; Swennen et al. 2004; Boult & Kaldi 2005; Shaw 2005; Sorkhabi & Tsuji 2005; Fossen et al. 2007, Jolley et al. 2007a, 2010;

CHALLENGES AND IMPACT OF COMPARTMENTALIZATION

11

and strain analysis and modelling – incorporating the use of micro-seismicity for surveillance (Heffer 2002; Rutledge et al. 2004; Sanderson & Zhang 2004; Main et al. 2007; Zhang et al. 2007; Osorio et al. 2008).

Structural reservoir description

Fig. 1. A holistic approach to ‘The Integrated Reservoir and Fluids Description Toolkit’ at the Appraisal and Development stages of oil field development.

Lonergan et al. 2007; Wibberley et al. 2008a). In understanding the uncertainties and risks from compartmentalization these largely lie in the areas of: reservoir scale structural analysis (Fisher & Knipe 1998; Dee et al. 2005; Fossen & Bale 2007); fault and fracture growth, fault zone architecture and characterization (Knipe et al. 1997; Maerten et al. 2006; Wibberley et al. 2008b); fault seal prediction (Bretan et al. 2003; Dee et al. 2007); the incorporation of faults and fault zone properties in reservoir models and simulation (Manzocchi et al. 1999, 2002, 2008; Harris et al. 2005, 2007; Maerten & Maerten 2006; Childs et al. 2007; Fisher & Jolley 2007; Zijlstra et al. 2007); integrated fluid description (Smalley & England 1994; Larter & Aplin 1995; Smalley et al. 2004); the use of geomechanical data along with field and inter-well scale stress

Regional and basin scale structural geology studies and 2D and 3D restorations are a prerequisite for exploration. However, reservoir scale structural geology studies at the drill-core and individual fault plane scale often do not start until compartmentalization problems arise during production. In retrospect, we have learnt from experience in our mature producing fields and new developments, that there is value in applying reservoir scale structural geology tools and techniques early in an appraisal program. These studies should include detailed fault and fracture logging of core and image log data combined with micro-structural, micro-porosity and mineralogical analysis of fault rock products. These studies should integrate with fault architecture analysis, incorporating assessment of fault plane juxtaposition and fault rock prediction (e.g. Shale Gouge Ratio calibration). Mud loss, wellbore breakout and drilling data (weight on the bit, rate of penetration, etc) are also key sources of insights into the controls of compartmentalization. Discrete Fracture Network modelling (DFN) and object orientated statistical fault zone modelling, at both seismic and subseismic scales are also important, together with geomechanical rock testing and numerical stress and strain modelling and prediction. Collecting sufficient core and applying destructive geomechanical rock strength studies, (e.g. unconfined compressive strength tests, uniaxial and triaxial testing and pore volume compressibility), stress-strain analysis and fault rock property testing from an exploration well takes a degree of courage on the part of exploration or appraisal leadership teams, but the results can have a significant impact on understanding the risk and impact of compartmentalization (Ruddy et al. 1989; Knipe et al. 1997; Fisher & Knipe 1998; Gibson 1998; Jones et al, 1998; Dewhurst & Jones 2002; Couples 2005). Commonly, up to a third or more of a whole core can be preserved at rig-site for fluid analysis and the remainder is sampled back in the lab for porosity and permeability analysis and then slabbed; often half to a third is set in resin for preservation, leaving little whole core remaining for rock mechanics testing or fault rock microstructural analysis. Seismic and core data can indicate that structural compartmentalization is a key risk to depletion optimization and the timely implementation of

12


secondary and tertiary recovery mechanisms such as waterflooding and Enhanced Oil Recovery (e.g. chemical injection or steam flooding). The ability to predict the behaviour of faults, fractures and in-reservoir stress responses, together with the sub-seismic distribution of these heterogeneities, is critical to predict and mitigate the impact of compartmentalization. However, whilst it is unlikely that we will ever be able to predict actual individual fault and fracture distributions at the core scale across a reservoir, there have been advances in developing analytical proxies. The advent of high performance computing capability enables advances in numerical geomechanical modelling techniques together with rock property studies (Koutsabeloulis & Hope 1998; Nieuwland 2003; Zhang et al. 2007). Resulting numerical stress and strain simulation and forward prediction capabilities offer the opportunity to better model the spatial distribution of the processes controlling the formation of faults and fractures (McClay et al. 2002; Maerten & Maerten 2006; Maerten et al. 2006; Main et al. 2007; Wilkins 2007). Application of these approaches at a scale an order of magnitude or more below seismic resolution, coupled with forward modelling of strain from restored reconstructions of 3D structural restorations, enables an understanding of subseismic fault predictions to be determined with an understanding of the related uncertainty range. With advances in non-orthogonal grid manipulation techniques and the ability to incorporate local grid refinement in reservoir engineering simulation software, we have an opportunity to integrate these geomechanical predictions with multiple sedimentological, geochemical, fault and fracture property realizations (Harris et al. 2007; Jolley et al. 2007b; Main et al. 2007; Ma¨kel 2007; Zhang et al. 2007; Manzocchi et al. 2008). When the modelling system is linked to real-time (down hole) reservoir monitoring data, we will be able to make live updates of our models and forward prediction of primary, secondary and tertiary recovery behaviour and sweep efficiency (Reddick 2006). We have begun to place greater emphasis on understanding what production data is telling us about fluid communication, reservoir geology and the subsequent changes in reservoir properties that occur over production time-scales. Production and injection data that historically provided information on fluid flow rate are now recognized as containing dynamic behaviour signals (Heffer et al. 1995; Heffer 2002; Main et al. 2007). As an oil field matures it can often display increasing mechanically influenced dynamic responses. As soon as the first well is drilled, the mechanical equilibrium of the field is disturbed. Once water or gases are introduced into a reservoir, as part of a secondary recovery processes to boost

oil or gas recovery, their introduction can lead to changes in the permeability field and the physical properties of the reservoir (Jansen & Kelkar 1997; Pizarro & Lake 2001; Rutledge et al. 2004; Main et al. 2006). Faults that were invisible to fluid flow during primary depletion can become problematical once water is placed in the system. The cooling effects of injected water can cause the faults to become critically stressed or reactivated (Barton et al. 1995; Heffer et al. 1995; Koutsabeloulis & Hope 1998). Significant drilling fluid losses can occur in and around faults in late field life that originally had no impact on drilling in early field life. These can result from changes in the stress field acting on the fault plane over time (Barton et al. 1993). Fractures can close rapidly during depletion once reservoir pressure is reduced to a particular threshold below the fracture closure pressure. This can be an irreversible effect that can significantly reduce production (e.g. Lisburne Field Alaska). Studies on flood directionality controls (Heffer et al. 1995) show that commonly stress fields surrounding injector wells can be perturbed and influence fluid flow away from the injection well. At the pore scale changes in pore throat diameter or wettability can cause alterations in flow paths, isolating large volumes of hydrocarbons and short cutting flood fronts (Jerauld & Rathmell 1997; Casabianca et al. 2007).

Integrated fluids description Geochemical fingerprinting and time-lapse geochemical analysis of reservoir fluids, enables in-depth understanding of fluid filling history and mixing across reservoirs, and the identification of individual reservoir compartments (Smalley & England 1994; Larter & Aplin 1995; Smalley & Hale 1996; Smalley et al. 2004). Collection of early appraisal and pre-production fluid samples can be utilized to map out initial fluid typing linked to seismic mapping. A subsequent surveillance program in which fluid samples are taken from producing wells at regular intervals, fingerprinted and compared to pre-production baseline samples, has shown to be highly valuable in the identification of production compartments, thus helping effective reservoir management (Fig. 2a, b) (Milkov et al. 2007). Analysis of minor compounds such as alkylbenzene carbon chains plotted on relatively simple spider diagrams often show unique signatures in oil composition (Fig. 2a). In a number of fields these techniques have been of significant value in helping to identify compartmentalization (Peters & Fowler 2002; McKie et al. 2010). When these techniques are integrated with standard PVT analysis,


13

stresses over production time-scales (EberhartPhillips & Oppenheimer 1984; Osorio et al. 2008). Micro-seismic arrays allow monitoring of production or injection induced seismicity within a reservoir over the life of a field, revealing or confirming the influence of pre-existing structures (Fig. 3) at or below resolution of reflection seismic data. In areas of a reservoir that have critically stressed faults (Barton et al. 1993; Zoback & Zinke 2002) or stress-sensitive production & injection (Fig. 4; Heffer et al. 1995), this surveillance technique can be an important reservoir management tool, used in well planning and operations, and helping to condition reservoir and flow simulation modelling (Fig. 5). In the examples shown in Figures 3 and 4 a micro-seismic array was deployed 16 years ago during the development of an onshore BP oilfield. Surface seismometers were installed with the

Fig. 2. (a) Variations in alkyl-benzene fingerprinting from a BP Field (Fig. 2b) during early appraisal. Note the significant difference in high concentrations of M_10/(10 þ 11) compounds in Wells A & D. (b) Sand against sand compartmentalization (green against brown compartments) in a BP field determined during early appraisal from the results of alkyl-benzene finger printing of early production fluids based upon the results plotted in Figure 2a, combined with pressure data, PVT analysis and seismic mapping.

pressure measurements and standard industry appraisal practices, it can become apparent that barriers exist between wells even where there are clean sand-on-sand contacts with minimal shale fault gouge (Fig. 2b) (Milkov et al. 2007).

Micro-seismicity for surveillance The active use of permanent micro-seismic arrays both onshore and off shore has given tremendous insight into the behaviour of faults, fractures and

Fig. 3. Trends of induced micro-seismic events collected over the life-cycle of a BP field (Upper Plot). The NW– SE trends of the data clusters parallel the local principal horizontal stress directions (ShMax from well data in red) seen in the lower plot.

14


Fig. 5. Results from the passive micro-seismic monitoring project (see Figs 3 & 4) are used to condition the waterflood characteristics of a Reservoir Simulation model to achieve a history match and predict conformance, sweep and related compartmentalization. High permeability corridors allowing preferential waterflood directionality were added based upon the micro-seismic results.

used to condition field-wide reservoir modelling and flow simulation grids, with placement of high permeability corridors in the models (Osorio et al. 2008). This led to a better match between real and simulated production history in the models. It also helped identify and characterize poor sweep and by-passed oil zones (Fig. 5). Thus, incorporation of the micro-seismic data into reservoir modelling helped to inform reservoir management strategies and depletion planning.

Fig. 4. Impact of water injection on the generation of micro-seismic events (see Fig. 3). (a) Before water injection they showed relatively few events. (b) In contrast, during maximum water injection a large number of seismic events were recorded at the surface.

Managing compartmentalization uncertainty

objective to detect seismic events on faults for the purpose of wellbore stability. Similar techniques have been used by BP to monitor seismic events associated with platform subsidence in the North Sea (Kristiansen et al. 2000). Micro-seismic events, both natural and induced were plotted for a period during primary depletion, prior to injection start-up, providing a base line to compare the production-induced effects. One observation from the data collected during production was that faults near water injectors were reactivated during injection. In Figure 3 studies show that a strong correlation exists between the distribution of recorded seismic events and the regional direction of present day maximum horizontal stress. The results have been

One of the key challenges we face today involves understanding the impact of uncertainty in our subsurface descriptions – and using this to optimize more flexible field development plans (e.g. Smalley et al. 2008). Here, uncertainty in the prediction of compartment size and distribution together with fault sealing potential and fluid flow characteristics can have a significant impact on recovery and commerciality. In the past specialists have tended to work alone or in asset cross-disciplinary teams, but within the paradigm of linear working, model building and project planning to solve the most difficult problems of the day. Historically the industry has tended to focus on ‘single-complex’ linear all encompassing modelling approaches that usually misrepresent or inadequately represent the uncertainties inherent in our interpretations. This can


15

Fig. 6. Flexuring of a top reservoir horizon about a single well penetration when uncertainty in your seismic depth conversion occurs away from your control points and can have significant impact on Bulk Rock Volume (BRV) and STOOIP (Stock Tank Original Oil In Place) volumes.

lead to a failure to extract all the available or relevant information from our data. Increasingly, we are realizing the value of generating, testing and refining multiple geological realizations of the subsurface that honestly reflect the uncertainty in our constraining data. This approach demands real integration of individual disciplines from structural geology, geophysics, geomechanics, sedimentology, petrophysics, petroleum systems analysis, fluid description and geochemistry. This leads to linked, merged workflows and a range of tool kits that address all the pieces of the subsurface puzzle relevant to compartmentalization. The final piece of this approach is the need for continual cross-discipline feedback loops at each stage of a study to constrain model ranges and outcomes in real-time. Discrete iterative geological models are needed with reasonable ranges of uncertainty. Together these enable plausible static-dynamic scenarios that inform data acquisition programmes, and help to deliver the most likely base case business model, and minimize subsurface development risks. For example, consider the uncertainty we are often faced with immediately after a discovery well has been drilled. In deep basins we commonly have uncertainty in the depth conversion of our seismic data and until additional data is acquired we must carry a range of possible scenarios for the 3D shape of a discovered reservoir. Flexing a top reservoir horizon to show the degradation in our interpretation confidence away from well control, can result in considerable uncertainty in Bulk Rock Volume (BRV) and STOOIP (Stock Tank Original Oil In Place), that is, the total hydrocarbon

content of an oil reservoir (Fig. 6). This information can then be combined with other uncertainties (e.g. the density or connectivity of faults in a structural model) to determine the full possible range of petroleum volumes and compartmentalization of those volumes. For example, Figure 7 shows how multiple fault models are combined with uncertainty in a top reservoir map to give a range of BRV scenarios. The pessimistic case combines a minimum BRV with maximum fault density, whereas the optimistic case represents the reverse scenario. All models were equally viable given the range and quality of data available at that stage of the appraisal programme. They were tested during the subsequent appraisal and drilling programmes, with early production data suggesting the base case to be close to reality.

Field appraisal impacts The appraisal stage in the life-cycle of an oil or gas field should focus upon identifying the full range of possible reservoir outcomes, together with corresponding field development concepts to be applied in each case. Not fully appreciating the impact of compartmentalization during the appraisal stage can cause value erosion by driving up costs or introducing significant development risks and in extreme cases, this can lead to irreversible consequences or a project failure. It is therefore vitally important to describe the full range of reservoir outcomes during an Appraisal program. The reference geological model will be the most plausible description that honours all available data considering the

16


Fig. 7. Optimistic, base-case and pessimistic geological models in a deep water reservoir. These maps show how different fault scenarios and uncertainty in the position of top reservoir result in a large range of potential STOOIP (Stock Tank Original Oil In Place) volumes.

impact of stratigraphic compartmentalization, faulting, fracturing, mineralogical changes, stress changes and variations in fluid distribution. The amount of work effort at this stage together with the quality of data available and its integration are critical to understand the degree and range of uncertainty. Thus, we propose a level of study at this stage of field life that we term ‘Forensic Reservoir Analysis’. This involves creating a range of discrete plausible geological models in which the analysis should always consider compartmentalization and its impact on the full range of scenarios. Upside (optimistic) and downside (pessimistic) cases will focus appraisal activity on key uncertainties and ensure that the appropriate data collection programs, surveillance strategies and mitigation options are considered. Together these will inform field development concept selection and provide confidence in delivering the base production plan. However, we often underestimate the degree of complexity in a reservoir or field during appraisal. This is often because we focus on over simplistic single static models that disregard outlying uncertainties, or fix on an incorrect base case to begin with (Bentley & Smith 2008). This is particularly true of poorly imaged traps and reservoirs as in many of today’s deep water fields. Figure 8 shows the evolving understanding of complexity in a deepwater Gulf of Mexico field from pre-discovery to early production. It illustrates the evolution of our understanding of the field’s complexity through time with increasing well penetrations and data acquisition. The cartoons are taken from the reference case for the main reservoir in the field. It illustrates the importance of continuing to acquire, process and integrate data as well as appreciating the range of uncertainty possible in a field. Prior to discovery the field was believed to be a simple

4-way dip closure. Ten years later and with an addition of 17 wells, a clearer picture emerges of the degree of compartmentalization present and its impact on the development. Fortunately, compartmentalization was carried as a major uncertainty and used to formulate data collection activities and re-working of the subsurface description. Data collection was aggressively pursued from the exploration well through to start-up and beyond. Maximum value was extracted from each data set, while multiple ranges of models were carried through the appraisal and development stages – integrated multidisciplinary studies were a key part of the subsurface appraisal work. Rigorous, detailed investigation coupled with the application of state-of-the-art technology provided confidence to progress the project from discovery into production at a natural pace. Given the high level of irreducible subsurface uncertainty prior to start-up, a phased development approach was adopted. Initial development focused mainly on the southern flank with new seismic data volumes acquired and old data volumes re-processed to illuminate the remaining areas of this field (Fig. 8).

Learning from early production Reservoir performance prediction begins very early in field appraisal. However, static data alone rarely identifies all risks for compartmentalization in a reservoir or field. Whilst the range of compartmentalization should be captured in the uncertainty assessment, we must prepare to make fundamental changes to a development plan following start-up as early production data becomes available. Thus, baseline data must be collected prior to production start-up as reference for future performance


17

Fig. 8. The evolution of complexity over time in a deepwater development in the Gulf of Mexico demonstrates the impact of data acquisition on interpretation and reveals an evolving view of degree of compartmentalization. With increasing well numbers together with improvements in seismic imaging, the realization of the degree of compartmentalization has grown together with insight into its impact on fluid flow. The image north of the yellow lines in each figure is significantly obscured by an overlying complex salt body.

18


Fig. 9. A holistic approach to ‘Reservoir Management’ at the Production stages of oil field development. Critical early field life production and dynamic data will allow calibration of the geological and geocellular models and allow iterative feedback loops for real time updates between engineering, geological and geophysical data.

prediction and to test the robustness of the geological models and their inherent uncertainties. Surveillance plans are based upon the static description and any analogue field performance available. This is updated and revised with early time dynamic and production data along with evidence of sweep and

conformance between wells. Production time-scale effects must always be considered. Faults which support pressure or fluid compositional variations across the fault plane may be sealing over geological time-scales, but may not support large pressure differentials caused by production (or vice-versa). Tracer data, well testing, production logging, leak-off tests and down hole pressure monitoring are key data sources to indicate compartmentalization, cross field communication or anisotropic flow behaviour. During the early stages of production as a field moves towards mid-life production rate plateau, it is important that the surveillance plan includes acquisition of additional data to evaluate remaining compartmentalization, and to identify possible unswept opportunities and mitigation actions as appropriate. At this stage, an ‘Integrated Reservoir Management Production focused Toolkit’ and appropriately interconnected work flow must be applied to optimize recovery and field performance in the face of compartmentalization. In this, we use a similar holistic approach of inputs and outputs (described earlier for the appraisal stage, in Fig. 1), but this time with greater emphasis on dynamic data. Pressure measurements, interference

Fig. 10. Sedimentological and seismic interpretation of a stacked channel sequence in a deep water fan depositional setting. Modular Formation Dynamics Tester (MDT) pressure data plots are correlated to the fan architecture and demonstrate the sealing effectiveness of the intra-fan bounding shales. Multiple vertical barriers can be inferred from pre- and post-production pressure data.


testing, 4D seismic, benchmarking against analogue fields with longevity of production data and the integration with reservoir engineering simulation are key elements of the production toolkit (Fig. 9). Interpretation of compartments and their complexity often change once hydrocarbon production begins. These ideas then mature with time as field production plateaus, and later declines. Thus, in mature fields there is often less ambiguity as a function of static and dynamic data collected over field life. At start-up, with few producers and injectors, pressure trends and flow paths can rarely be defined without ambiguity and uncertainty. Early production schemes, extended well tests (EWT’s) and interference tests between wells help reduce early uncertainty; and early pressure drawdown and buildup data are used to determine heterogeneity within a reservoir compartment, as well as the impact and distance to boundaries. When early dynamic and production data indicate a variation from the expected base-case geological model, development teams must be quick to react. The deepwater Gulf of Mexico

19

reservoir illustrated in Figures 10 and 11 provide an excellent illustration of this. During the appraisal phase, reservoir complexity and compartmentalization were underestimated. Initial production suggested a complex heterogeneity at the limits of the predicted uncertainty range at field start-up. Static and dynamic pressure data (Fig. 10), pressure transient analysis build-up tests (Fig. 11) and well interference tests all suggested lateral baffling between geobodies and pressure isolated compartments between specific sedimentary shale layers impacting vertical connectivity (Fig. 10). Thus, the effect of complex sedimentological architecture was under evaluated. The main reservoir intervals were formed of stacked channels in individual, laterally offset fan lobes, bounded by condensed shales. Post depositional erosion of reservoir units, identified from seismic facies mapping further complicated well performance. In addition, core studies linked to fluid flow data indicate that a later structural fabric would reduce near-wellbore productivity. Integrating this information with dynamic data allowed a rapid reassessment and

Fig. 11. Seismic facies mapping is used to define architectural sedimentological elements. Pressure transient analysis indicates the close proximity of baffles and barriers representing channel margins, internal shale layers and faulting (Well B).

20


re-modelling of the reservoir heterogeneities. The importance of extensive thin shale layers (reducing vertical permeability) was recognized as being responsible for complex flow displacement and tortuosity in the reservoir. This required alternative well placement and a fundamental change to the development plan. Early calibration to the well data (Fig. 11) allowed a deeper understanding of well performance and a renewed confidence in the alternative depletion model and development plan.

Future challenges and key messages It is an industry wide experience that compartmentalization is one of the fundamental risks to the delivery of today’s production and the development of tomorrow’s fields. We find that we are managing increasing uncertainties in petroleum reserves, production rate and the profile of production over time due to the enhanced impact of sedimentological and structural features as existing fields mature and because new fields are discovered in increasingly complex geological settings. In order to manage the balance between risks and opportunities posed by reservoir compartmentalization, it is vitally important to capture the uncertainty that accompanies our data, knowledge base and ideas. An appropriate range of geological models is therefore required to minimize reservoir surprises and test our knowledge limits. It is important to ensure that we think widely enough and as realistically as possible to capture and integrate an appropriate range of factors and key uncertainties controlling the distribution and impact of compartmentalization. Developing the models to adequately capture the range of uncertainties requires increasingly seamless integration between our disciplines, our subsurface descriptions; and toolkits. To achieve this we suggest that the basic requirements are: † a ‘forensic’ level of data analysis; † improved techniques in structural reservoir description, sedimentological description, geomechanical modelling; † better use of fluid data; † greater dynamic and static data integration; † the right integrated tools; and † continual development of leading edge technology to combine these together in a common platform. The distinctiveness of the topic of compartmentalization is that it calls upon a wide range of toolkits and flexibility within any formulated workflow. Standardized workflows, which commonly follow a logical order and are often prescriptive, do not apply well to the challenges posed by compartmentalization – since the optimal techniques needed to solve these challenges often vary

dramatically from situation to situation. Thus, we do not advocate applying every reservoir description technique to every situation in the same order or with the same recurrence. It must be a fieldspecific judgement with the option of a toolbox containing a diverse range of specialist technical approaches as per (but not limited to), the examples described in this paper. Finally, we need a skilled work force with the right behaviours unconstrained by business process. Fresh ‘blue sky’, out-of-the-box lateral thinking must be encouraged and time allowed for it despite business time pressures, if we are to reduce the occurrence of regrettable reservoir surprises and development failures caused by underestimating the effects of compartmentalization. The authors wish to thank BP for permission to publish this paper. We also particularly wish to thank Tim Buddin and Sam Johnson for their help and many of our BP colleagues for support and discussions. The views expressed here are those of the authors and not necessarily of BP. The help and comments from Steve Jolley and reviewers Mark Bentley and John Marshall to improve this manuscript are greatly appreciated.

References Barkved, O., Heavey, P., Kjelstadli, R., Kleppan, T. & Kristiansen, T. G. 2003. Valhall field – still on plateau after 20 years of production. Society of Petroleum Engineers, SPE Paper 83957. Barr, D. 2007. Conductive faults and sealing fractures in the West Sole gas fields, southern North Sea. In: Jolley, S. J., Barr, D., Walsh, J. J. & Knipe, R. J. (eds) Structurally Complex Reservoirs. Geological Society, London, Special Publications, 292, 431– 451. Barr, D., Savory, K. E., Fowler, S. R., Arman, K. & McGarrity, J. P. 2007. Pre-development fracture modelling in the Clair Field West of Shetland. In: Lonergan, L., Jolly, R. J. H., Rawnsley, K. & Sanderson, D. J. (eds) Fractured Reservoirs. Geological Society, London, Special Publications, 270, 205– 225. Barton, C. A., Zoback, M. D. & Moos, D. 1993. Identification of hydraulically conductive fractures from the analysis of localized stress perturbations and thermal anomalies. EOS Transactions, 74, p43, 568. Barton, C. A., Zoback, M. D. & Moos, D. 1995. Fluid flow along potentially active faults in crystalline rock. Geology, 23, 683–686. Bentley, M. & Smith, S. 2008. Scenario-based reservoir modelling: the need for more determinism and less anchoring. In: Robinson, A., Griffiths, P., Price, S., Hefre, J. & Muggeridge, A. (eds) The Future of Geological Modelling in Hydrocarbon Development. Geological Society, London, Special Publications, 309, 145 –159. Boult, P. & Kaldi, J. 2005. (eds) Evaluating Fault and Cap Rock Seals. American Association of Petroleum Geologists Hedberg Series, 2. Bretan, P., Yielding, G. & Jones, H. 2003. Using calibrated shale gouge ratio to estimate hydrocarbon

CHALLENGES AND IMPACT OF COMPARTMENTALIZATION column heights. American Association of Petroleum Geologists Bulletin, 87, 397–413. Brown, A. 2003. Improved interpretation of wireline pressure data. American Association of Petroleum Geologists Bulletin, 87, 295–311. Casabianca, D., Jolly, R. J. H. & Pollard, R. 2007. The Machar oil field; waterflooding a fractured chalk reservoir. In: Lonergan, L., Jolly, R. J. H., Rawnsley, K. & Sanderson, D. J. (eds) Fractured Reservoirs. Geological Society, London, Special Publications, 270, 205– 225. Childs, C., Walsh, J. J., Manzocchi, T., Strand, J., Nicol, A., Tomasso, M. & Schopfer, M. P. J. 2007. Definition of a fault permeability predictor from outcrop studies of a faulted turbidite sequence, Taranaki, New Zealand. In: Jolley, S. J., Barr, D., Walsh, J. J. & Knipe, R. J. (eds) Structurally Complex Reservoirs. Geological Society, London, Special Publications, 292, 235– 258. Clifford, P. J., O’donovan, A. R., Savory, K. E., Smith, G. & Barr, D. 2005. Clair field – managing uncertainty in the development of a waterflooded fractured reservoir. Society of Petroleum Engineers, SPE Paper 96316. Corrigan, A. F. 1993. Estimation of recoverable reserves: the geologists’ job. In: Parker, J. R. (ed.) Petroleum Geology of Northwest Europe: Proceedings of the 4th Conference. Geological Society, London, 1473–1482. Couples, G. G. 2005. Seal: The role of geomechanics. In: Boult, P. & Kaldi, J. (eds) Evaluating Fault and Cap Rock Seals. American Association of Petroleum Geologists Hedberg Series, 2, 87–108. Coward, M. P., Daltaban, T. S. & Johnson, H. (eds) 1998. Structural Geology in Reservoir Characterization. Geological Society, London, Special Publications, 127. Dee, S. J., Freeman, B., Yielding, G., Roberts, A. & Bretan, P. 2005. Best practices in structural analysis. First Break, 23, 49–54. 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. Dewhurst, D. N. & Jones, R. M. 2002. Geomechanical, microstructural, and petrophysical evolution in experimentally reactivated cataclasites: applications to fault seal prediction. American Association of Petroleum Geologists Bulletin, 86, 1383– 1405. Eberhart-Phillips, D. & Oppenheimer, D. H. 1984. Induced seismicity in the Geysers Geothermal area, California. Journal of Geophysical Research, 89, 1191–1207. Fisher, Q. & Jolley, S. J. 2007. Treatment of faults in production simulation models. In: Jolley, S. J., Barr, D., Walsh, J. J. & Knipe, R. J. (eds) Structurally Complex Reservoirs. Geological Society, London, Special Publications, 292, 219–233. Fisher, Q. & Knipe, R. J. 1998. Fault sealing processes in siliciclastic sediments. In: Jones, G., Fisher, Q. J. & Knipe, R. J. (eds) Faulting, Fault Sealing and Fluid Flow in Hydrocarbon Reservoirs. Geological Society, London, Special Publications, 147, 117– 133.

21

Fossen, H. & Bale, A. 2007. Deformation bands and their influence on fluid flow. American Association of Petroleum Geologists Bulletin, 91, 1685–1700. Fossen, H., Schulz, R. A., Shipton, Z. K. & Mair, K. 2007. Deformation bands in sandstone – A review. Journal of the Geological Society, London, 164, 755– 769. Gainski, M., Macgregor, A. G., Freeman, P. J. & Nieuwland, H. F. 2010. Turbidite reservoir compartmentalisation and well targeting with 4D seismic and production data: Schiehallion Field, UK. In: Jolley, S. J., Fisher, Q. J., Ainsworth, R. B., Vrolijk, P. J. & Delisle, S. D. (eds) Reservoir Compartmentalization. Geological Society, London, Special Publications, 347, 89– 102. Gibson, R. G. 1994. Fault zone seals in siliclastic strata of the Columbus Basin, offshore Trinidad. American Association of Petroleum Geologists Bulletin, 78, 1372– 1385. Gibson, R. G. 1998. Physical character and fluid-flow properties of sandstone-derived fault zones. In: Coward, M. P., Daltaban, T. S. & Johnson, H. (eds) Structural Geology in Reservoir Characterization. Geological Society, London, Special Publications, 127, 83– 97. Harris, S. D., Fisher, Q. J., Karimi-Fard, M., Vaszi, A. Z. & Wu, K. 2005. Modelling the effects of faults and fractures on fluid flow in petroleum reservoirs. In: Ingham, D. B. & Pop, I. (eds) Transport Phenomena in Porous Media. Elsevier, Netherland, III, 441–476. Harris, S. D., Vaszi, A. Z. & Knipe, R. J. 2007. Threedimensional upscaling of fault damage zones for reservoir simulation. In: Jolley, S. J., Barr, D., Walsh, J. J. & Knipe, R. J. (eds) Structurally Complex Reservoirs. Geological Society, London, Special Publications, 292, 353–374. Heffer, K. J. 2002. Geomechanical influences in water injection projects: an overview. Oil & Gas Science and Technology Review. Institut Français du Petrole, 57, 412 –422. Heffer, K. J., Fox, R. J. & Mcgill, 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, SPE Paper 30711, 91– 98. Hesthammer, J., Bjørkum, P. A. & Watts, L. 2002. The effects 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. Jansen, F. E. & Kelkar, M. 1997. Non-stationary estimation of reservoir properties using production data. Society of Petroleum Engineers, SPE Paper 38729. Jerauld, G. R. & Rathmell, J. J. 1997. Wettability and relative permeability of Prudhoe Bay: a case study in mixed-wet reservoirs. Society of Petroleum Engineers, SPE Paper 28576. Jolley, S. J., Barr, D., Walsh, J. J. & Knipe, R. J. 2007a. (eds) Structurally Complex Reservoirs: An Introduction. Geological Society, London, Special Publications, 292. Jolley, S. J., Dijk, H., Lamens, J. H., Fisher, Q. J., Manzocchi, T., Eikmans, H. & Huang, Y. 2007b. Faulting and fault sealing in production simulation models:

22


Brent province, northern North Sea. Petroleum Geoscience, 13, 321–340. Jolley, S. J., Fisher, Q. J., Ainsworth, R. B., Vrolijk, P. J. & Delisle, S. D. (eds) 2010. Reservoir Compartmentalization. Geological Society, London, Special Publications, 347. Jones, G., Fisher, Q. J. & Knipe, R. J. (eds) 1998. Faulting, Fault Sealing and Fluid Flow in Hydrocarbon Reservoirs. Geological Society, London, Special Publications, 147. Knai, T. A. & Knipe, R. J. 1998. The impact of faults on Fluid Flow in the Heidrun Field. In: Jones, G., Fisher, Q. J. & Knipe, R. J. (eds) Faulting, Fault Sealing and Fluid Flow in Hydrocarbon Reservoirs. Geological Society, London, Special Publications, 147, 269–282. Knipe, R. J., Fisher, Q. J. et al. 1997. Fault seal analysis: successful methodologies, application and future directions. In: Møller-Pederson, P. & Koestler, A. G. (eds) Hydrocarbon Seals: Importance for Exploration and Production. Norwegian Petroleum Society, Special Publications, 7, 15–40. Knott, S. 1993. Fault seal analysis in the North Sea. American Association of Petroleum Geologists Bulletin, 77, 778–792. Koutsabeloulis, N. C. & Hope, S. A. 1998. Coupled stress/fluid-thermal multi-phase reservoir simulation studies incorporating rock mechanics. Society of Petroleum Engineers, SPE Paper 47393. Kristiansen, T. G., Barkved, O. & Patillo, P. D. 2000. Use of passive seismic monitoring in well and casing design in the compacting and subsiding Valhall Field, North Sea. Society of Petroleum Engineers, SPE Paper 65134. Larter, S. R. & Aplin, A. C. 1995. Reservoir geochemistry: methods, applications and opportunities. In: Cubitt, J. M. & England, W. A. (eds) The Geochemistry of Reservoirs. Geological Society, London, Special Publications, 86, 5 –32. Leveille, G. P., Knipe, R. J. et al. 1997. Compartmentalisation of Rotliegendes gas reservoirs by sealing faults, Jupiter Fields area, Southern North Sea. In: Ziegler, K., Turner, P. & Daines, S. R. (eds) Petroleum Geology of the Southern North Sea; Future Potential. Geological Society, London, Special Publications, 123, 87–104. Lonergan, L., Jolly, R. J. H., Rawnsley, K. & Sanderson, D. J. (eds) 2007. Fractured Reservoirs. Geological Society, London, Special Publications, 270. Maerten, L., Gillespie, P. & Daniel, J. M. 2006. Threedimensional geomechanical modelling for constraint of subseismic fault simulation. American Association of Petroleum Geologists Bulletin, 90, 1337– 1358. Maerten, L. & Maerten, F. 2006. Chronological modelling of faulted and fractured reservoirs using geomechanically based restoration: technique and industry application. American Association of Petroleum Geologists Bulletin, 90, 1201– 1226. Main, I. G., Li, L., Heffer, K. J., Papasouliotis, O. & Leonard, T. 2006. Long-range, critical-point dynamics in oil field flow rate data. Geophysical Research Letters, 33, L18308. Main, I. G., Li, L., Heffer, K. J., Papasouliotis, O., Leonard, T., Koutsabeloulis, N. C. & Zhang, X. 2007. The Statistical Reservoir Model: calibrating faults and fractures, and predicting reservoir response

to water flood. In: Jolley, S. J., Barr, D., Walsh, J. J. & Knipe, R. J. (eds) Structurally Complex Reservoirs. Geological Society, London, Special Publications, 292, 469–482. Manzocchi, T., Heath, A. E., Palananthakumar, B., Childs, C. & Walsh, J. J. 2008. Faults in conventional flow simulation models: a consideration of representational assumptions and geological uncertainties. Petroleum Geoscience, 14, 91– 110. Manzocchi, T., Heath, A. E., Walsh, J. J. & Childs, C. 2002. The representation of two phase fault-rock properties in flow simulation models. Petroleum Geoscience, 8, 119– 132. Manzocchi, T., Walsh, J. J., Nell, P. & Yielding, G. 1999. Fault transmissibility multipliers for flow simulation models. Petroleum Geoscience, 5, 53–63. Ma¨kel, G. H. 2007. The modelling of fractured reservoirs: constraints and potential for fracture network geometry and hydraulics analysis. In: Jolley, S. J., Barr, D., Walsh, J. J. & Knipe, R. J. (eds) Structurally Complex Reservoirs. Geological Society, London, Special Publications, 292, 375– 403. McClay, K. R. (ed.) 2004. Thrust Tectonics and Hydrocarbon Systems. American Association of Petroleum Geologists, Memoir, 82. McClay, K. R., Dooley, T., Whitehouse, P. & Mills, M. 2002. 4-D evolution of rift systems: insights from scaled physical models. American Association of Petroleum Geologists Bulletin, 86, 935– 960. McKie, T., Jolley, S. J. & Kristensen, M. B. 2010. Stratigraphic and structural compartmentalization of dryland fluvial reservoirs: Triassic Heron Cluster, Central North Sea. In: Jolley, S. J., Fisher, Q. J., Ainsworth, R. B., Vrolijk, P. J. & Delisle, S. D. (eds) Reservoir Compartmentalization. Geological Society, London, Special Publications, 347, 165– 198. Milkov, A. V., Goebel, E., Dzou, L., Fisher, D. A., Kutch, A., Mccaslin, N. & Bergman, D. F. 2007. Compartmentalization and time-lapse geochemical reservoir surveillance of the Horn Mountain oil field, deepwater Gulf of Mexico. American Association of Petroleum Geologists Bulletin, 91, 847– 876. Møller-Pedersen, P. & Koestler, A. G. (eds) 1997. Hydrocarbon Seals: Importance for Exploration and Production. Norwegian Petroleum Society, Special Publication, 7. Moulds, T. P., Trussell, P., Haseldonckx, S. A. & Carruthers, R. A. 2005. Magnus field: reservoir management in a mature field combining waterflood, EOR, and new are developments. Society of Petroleum Engineers, SPE Paper 96292. Nieuwland, D. A. (ed.) 2003. New insights into Structural Interpretation and Modelling. Geological Society, London, Special Publications, 212. Osorio, J., Pen˜uela, G. & Ota´lora, O. 2008. Correlation between micro-seismicity and reservoir dynamics in a tectonically active area of Colombia. Society of Petroleum Engineers, SPE Paper 115715-MS. Pelissier-Combescure, J., Pollock, D. & Wittmann, M. 1979. Application of repeat formation tester pressure measurements in the Middle East. Society of Petroleum Engineers, SPE Paper 7775-MS. Peters, K. E. & Fowler, M. G. 2002. Applications of petroleum geochemistry to exploration and reservoir management. Organic Geochemistry, 33, 5– 36.

CHALLENGES AND IMPACT OF COMPARTMENTALIZATION Pizarro, J. O. S. & Lake, L. W. 2001. Dominant Communications trends determined by nonparametric statistics. Journal of Canadian Petroleum Technology, 40, 61–67. Porter, J. R., McAllister, E. et al. 2004. Impact of fault-damage zones on reservoir performance in the Hibernia Oilfield (Jeanne d’Arc Basin, Newfoundland): an analysis of structural, petrophysical and dynamic well test data. In: Hiscott, R. & Pulham, A. (eds) Petroleum Resources and Reservoirs of the Grand Banks, Eastern Canadian Margin. Geological Association of Canada, 43, 129 –142. Reddick, C. 2006. Field of the future: making BP’s vision a reality. Society of Petroleum Engineers, SPE Paper 99777. Reynolds, A. D., Simmons, M. D. et al. 1998. Implications of outcrop geology for reservoirs in the Neogene productive series; Apsheron Peninsula, Azerbaijan. American Association of Petroleum Geologists Bulletin, 82, 25–49. Ruddy, I., Andersen, M. A., Pattillo, P. D., Bishlawi, M. & Foged, N. 1989. Rock compressibility, compaction, and subsidence in a high-porosity chalk reservoir: a case study of Valhall Field. Journal of Petroleum Technology, 41, 741 –746. Rutledge, J. T., Phillips, W. S. & Mayerhofer, 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. Sanderson, D. J. & Zhang, X. 2004. Stress-controlled localization of deformation and fluid flow in fractured rocks. In: Cosgrove, J. W. & Engelder, T. (eds) The Initiation, Propagation, and Arrest of Joints and Other Fractures. Geological Society, London, Special Publications, 231, 299– 314. Shaw, R. P. (ed.) 2005. Understanding the Micro to Macro behaviour of Rock-Fluid Systems. Geological Society, London, Special Publications, 249. Smalley, P. C., Begg, S. H., Naylor, M., Johnsen, S. & Godi, A. 2008. Handling risk and uncertainty in petroleum exploration and asset management: an overview. American Association of Petroleum Geologists Bulletin, 92, 1251–1261. Smalley, P. C. & England, W. A. 1994. Reservoir compartmentalization assessed with fluid compositional data. Society of Petroleum Engineers Reservoir Engineering, 9, 175. Smalley, P. C., England, W. A., Muggeridge, A., Abacioglu, Y. & Crawley, S. 2004. Rates of reservoir fluid mixing; implications for interpretation of fluid data. In: Cubitt, J. M., England, W. A. & Larter, S. R. (eds) Understanding Petroleum Reservoirs: Towards an Integrated Reservoir and Engineering Approach. Geological Society, London, Special Publications, 237, 99– 113. Smalley, P. C. & Hale, N. A. 1996. Early identification of reservoir compartmentalization by combining a range of conventional and novel data types. Society of Petroleum Engineers Formation Evaluation, 11, 163–170. Smalley, P. C. & Muggeridge, A. 2010. Reservoir compartmentalization: get it before it gets you. In: Jolley,

23

S. J., Fisher, Q. J., Ainsworth, R. B., Vrolijk, P. J. & Delisle, S. D. (eds) Reservoir Compartmentalization. Geological Society, London, Special Publications, 347, 25–41. Smith, P. J. 2008. Studies of United Kingdom Continental Shelf fields after a decade of production: how does production data affect the estimation of subsurface uncertainty? American Association of Petroleum Geologists Bulletin, 92, 1403– 1413. Sorkhabi, R. & Tsuji, Y. (eds) 2005. Faults, Fluid Flow and Petroleum Traps. American Association of Petroleum Geologists Memoir, 85. Swennen, R., Roure, F. & Granath, J. W. (eds) 2004. Deformation, Fluid Flow, and Reservoir Appraisal in Foreland Fold and Thrust Belts. American Association of Petroleum Geologists Hedberg Series, 1. Weber, K. J. 1997. A historical overview of the efforts to predict and quantify hydrocarbon trapping features in the exploration phase and in field development planning. In: Møller-Pederson, P. & Koestler, A. G. (eds) Hydrocarbon Seals: Importance for Exploration and Production. Norwegian Petroleum Society, Special Publications, 7, 1– 13. Wibberley, C. A. J., Kurz, W., Imber, J., Holdsworth, R. E. & Collettini, C. (eds) 2008a. The Internal Structure of Fault Zones: Implications for Mechanical and Fluid-Flow Properties. Geological Society, London, Special Publications, 299. Wibberley, C. A. J., Yielding, G. & Di toro, G. 2008b. Recent advances in the understanding of fault zone internal structural: a review. In: Wibberley, C. A. J., Kurz, W., Imber, J., Holdsworth, R. E. & Collettini, C. (eds) The Internal Structure of Fault Zones: Implications for Mechanical and Fluid-Flow Properties. Geological Society, London, Special Publications, 299, 5– 33. Wilkins, S. J. 2007. Fracture intensity from geomechanical models: application to the Blue Forest 3D survey, Green River Basin, Wyoming, USA. In: Jolley, S. J., Barr, D., Walsh, J. J. & Knipe, R. J. (eds) Structurally Complex Reservoirs. Geological Society, London, Special Publications, 292, 137– 157. Williams, G. J. J., Mansfield, M., MacDonald, D. G. & Bush, M. D. 2004. Top-down reservoir modelling. Society of Petroleum Engineers, SPE Paper 89974. Zhang, X., Koutsabeloulis, N. C., Heffer, K. J., Main, I. G. & Li, L. 2007. Coupled geomechanics-flow modelling at and below a critical stress state used to investigate common statistical properties of field production data. In: Jolley, S. J., Barr, D., Walsh, J. J. & Knipe, R. J. (eds) Structurally Complex Reservoirs. Geological Society, London, Special Publications, 292, 453 –468. Zijlstra, E. B., Reemst, P. H. M. & Fisher, Q. J. 2007. Incorporation of fault properties into production simulation models of Permian reservoirs from the southern North Sea. In: Jolley, S. J., Barr, D., Walsh, J. J. & Knipe, R. J. (eds) Structurally Complex Reservoirs. Geological Society, London, Special Publications, 292, 295–308. Zoback, M. D. & Zinke, J. C. 2002. Production-induced normal faulting in the Valhall and Ekofisk oil fields. Pure and Applied Geophysics, 159, 403–420.

Reservoir compartmentalization: get it before it gets you P. C. SMALLEY1* & A. H. MUGGERIDGE1,2 1

BP, Chertsey Road, Sunbury-on-Thames, Middlesex TW16 7LN UK

2

Department of Earth Science and Engineering, Imperial College, London SW7 2AZ UK *Corresponding author (e-mail: [email protected]) Abstract: This paper examines the impact of compartmentalization on oil recovery, the importance of identifying it during field appraisal, and methods to evaluate it using fluid data. The impact on recovery factor is highlighted using a global database of oil field recovery factors as a function of reservoir complexity and compartmentalization, and emphasized in two case studies. The effect of compartmentalization on oil recovery demonstrates the benefit in characterizing compartmentalization correctly during appraisal, so that the field can be developed in an optimal manner. Early characterization of field compartmentalization requires making maximum use of available fluid data during appraisal. When interpreting fluid data to identify compartmentalization, it is critical to take into account the different time-scales for various fluid signals (pressure, contacts, density, composition) to equilibrate, and to be able to extrapolate to field production timescales. This is essential to avoid false negatives (compartments assumed absent due to homogeneous fluid properties, when in fact fluids would have equilibrated even in the presence of compartments), false positives (where fluid differences are interpreted as evidence of compartments when in fact there has not been sufficient time for equilibration to occur), and to resolve apparently conflicting data (some fluid indicators are at equilibrium, others are not). Rigorous simulation of fluid equilibration is a complex multiphase multidimensional process, and is generally reserved for specialist in-depth studies. However, order-of-magnitude evaluations can be made using analytical solutions in minutes, allowing many ‘what-if’ scenarios to be considered and uncertainty to be assessed. Analytical solutions for estimating the time required for spatially-varying fluid properties to revert to steady state distributions are reviewed. All these mixing processes are shown to be diffusive in character. An effective diffusion coefficient for each process can be calculated from the reservoir rock and fluid properties. For an isothermal system, the different time-scales and distances for each fluid-property variation to attain equilibrium can be compared on a single graph. Where the time elapsed since fluid-perturbation is known, analytical solutions can be used to estimate the degree of compartmentalization (e.g. permeability of barriers). These solutions lend themselves to the development of simple practical compartment-assessment tools for industry practitioners.

A reservoir is said to be compartmentalized if the reservoir fluids cannot flow freely from one part of the reservoir to another over production time-scales (typically years to tens of years). Mechanisms of compartmentalization include faulting or depositional heterogeneity. Unidentified reservoir compartmentalization can have a profound, usually adverse, effect on oil or gas recovery. It is thus vital to characterize reservoir compartmentalization as early as possible in a field’s life, ideally during appraisal, before large investment decisions are made. This paper examines: (a) how reservoir compartmentalization affects hydrocarbon recovery and thus why it is so important to characterize early; and (b) what can be done to make maximum use of the available data so as to get the earliest possible indications of compartmentalization.

The challenge is to identify compartmentalization from the limited data that are available during appraisal. The really definitive data about compartmentalization – long-term surveillance during production – are, of course, not available prior to production start-up. The types of data available usually consist of 3D seismic, well logs and core, some indication of hydrocarbon and water composition and contacts, and fluid pressure data. Occasionally, but not always (notably in the Gulf of Mexico) a well test may have been performed. Seismic mapping may show some, though not necessarily all, of the faults present in a field but does not provide direct information about the transmissibility of those faults until calibrated by dynamic data. Logs will show if there are low permeability shales present in the field. These shales

From: Jolley, S. J., Fisher, Q. J., Ainsworth, R. B., Vrolijk, P. J. & Delisle, S. (eds) Reservoir Compartmentalization. Geological Society, London, Special Publications, 347, 25–41. DOI: 10.1144/SP347.3 0305-8719/10/$15.00 # The Geological Society of London 2010.

26

P. C. SMALLEY & A. H. MUGGERIDGE

may act as barriers to vertical flow if they are laterally extensive, but a shale layer may not be completely continuous even if it can be correlated across several appraisal wells. Although lateral facies differences may be inferred from seismic or core data, they can be very difficult to prove without dynamic data. Even where there are well test data, this may not be sufficient to identify compartments definitively. The distance that a well test can ‘see’ away from the well into the formation is related to the duration of the test and the amount of fluid produced or injected. The cost of testing, which increases with test duration, often results in shorter-than-optimal tests where the distance investigated into the formation may only be sufficient to identify flow barriers close to the well. Observations of non-equilibrium fluid-related properties in different parts of the reservoir, such as pressure, pressure gradients, water–oil or gas– oil contacts, fluid physical properties (e.g. density, gas–oil ratio (GOR)) and fluid composition (chemistry of gas, oil, water), provide possible supporting evidence that there are physical barriers to flow. Hydrocarbon reservoirs are subject to a variety of natural processes that can result in non-uniform fluid properties both vertically and laterally. Reservoir filling almost inevitably results in hydrocarbons of different compositions arriving at the reservoir over time as the source rock matures (Stainforth 2004). Indeed, some reservoirs are believed still to be filling (e.g. Matava et al. 2003). This will result in spatially varying oil composition (and thus variations in density, GOR and bubble point) and may also affect pressure and fluid contact depths. In-situ biodegradation may have a similar effect (e.g. Larter et al. 2003; Huang et al. 2004). Regional or local hydrodynamic conditions (caused by differential compaction or recharge at the surface) may affect reservoir pressure and contact depth as a function of the direction of fluid flow (e.g. Berg et al. 1994; Stenger et al. 2001; Zawisza 2004) as well as the water chemistry. Reservoir restructuring due to uplift or erosion can alter both pressure distributions and fluid contacts (e.g. Jiao & Zheng 1998; Sorenson 2005) whilst heterogeneity in mineral distribution and diagenetic processes can affect the water chemistry (e.g. Ziegler et al. 2001) and hydrocarbon chemistry (e.g. Bennett et al. 2007). A problem with interpreting such data is that, even if these fluid-perturbation processes have ceased, the resulting distributions of pressure, fluid contacts or composition may still take hundreds of thousands or millions of years to attain a steadystate condition, even in the absence of barriers or baffles to flow (England et al. 1987, 1995; Muggeridge et al. 2004; Smalley et al. 2004; Stainforth 2004). Clearly, it is important to establish the

time-scales necessary for the various fluid-property variations to revert back to equilibrium (once the processes causing these variations have stopped) through the relevant mixing processes, and to compare these mixing time-scales with reservoir age or the likely time since any fluid-perturbation processes ceased in order to determine whether any fluid variations reflect compartmentalization or simply incomplete mixing. Previously, this type of fluid data has been interpreted in three progressively more sophisticated ways: (1)

(2)

(3)

Some of the earliest studies used only one data type, or sometimes a small range of data types, and used the data as an on/off compartment indicator. Such ‘fingerprinting’ studies might identify statistical differences in fluid composition or pressure between different parts of a field, and use that as an indicator of compartmentalization (e.g. Kaufman et al. 1990a, b; Stølum & Smalley 1992; Stølum et al. 1993; Hwang & Baskin 1994; Hwang et al. 1994). Without knowledge of mixing rates, however, such an approach was susceptible to false positives (differences were attributed to compartmentalization rather than insufficient mixing time) or false negatives (lack of differences assumed to indicate lack of compartments rather than rapid mixing). Further, this approach fosters a black/white interpretation of ‘compartmentalized or not’ rather than a more useful interpretation of the probability and extent of compartmentalization. Later studies (e.g. Smalley & Hale 1996) integrated a wide range of different data types, including seismic and log data, oil and water chemistry, pressure, PVT properties and sedimentological and biostratigraphic data. Although not including a quantitative view of fluid mixing rates, the integration of many fluid properties that all had different mixing rates made the interpretations much less prone to false positives and negatives. Most recently, studies have included a quantitative analysis of mixing rates, using analytical models for fluid mixing, sometimes backed up by numerical simulations (England et al. 1995; Smalley et al. 1995, 2004; Muggeridge et al. 2004, 2005).

In this paper we attempt to: (a) demonstrate the importance of compartmentalization; and (b) extend the third-mentioned approach to early compartment identification by integrating simple analytical solutions for estimating the mixing rates for all of the main relevant data types.

GET IT BEFORE IT GETS YOU

27

How compartmentalization affects recovery To understand the impact of reservoir compartmentalization on recovery, it is useful to represent the degree of compartmentalization in a common way that can be compared between different reservoirs. One approach that has been used in the oil and gas industry is to codify the various elements of reservoir complexity that affect recovery using a numerical scoring system to generate a so-called complexity index. This approach grew out of the initial work of Dromgoole & Speers (1997). Several companies use this complexity approach, though little has been published in the open literature for confidentiality reasons. The framework used in this paper is to assign a numerical score for each of 20 complexity factors that typically influence oil and gas recovery. These factors include a number of geological and/or engineering controls such as structural complexity (e.g. faulting, fracturing); depositional complexity (e.g. depositional continuity); reservoir quality (e.g. permeability, heterogeneity); fluid quality (e.g. viscosity); and reservoir energy (e.g. pressure, aquifer strength). These factors are weighted for importance and added to create an overall complexity score that relates well with oil recovery factor (Fig. 1), where recovery factor refers to the proportion of oil initially in place that is expected to be produced with the currently sanctioned development. Other factors besides reservoir complexity affect recovery factor, particularly recovery process (e.g. depletion, waterflooding, enhanced oil recovery) and well type and spacing. Figure 1 thus illustrates only a subset of data from one situation (waterflooded sandstone oil reservoirs with ,10 cP oil viscosity, .10 mD permeability and 1 –4 km2 well spacing). The relationship of decreasing recovery factor

Fig. 1. Recovery factor v. complexity index (weighted sum of 20 individual factors) for mature waterflooded sandstone oil reservoirs with ‘normal’ properties, that is, excluding very viscous oils and low permeability reservoirs. In this and subsequent figures, recovery factor refers to the proportion of oil expected to be produced from the existing sanctioned development, from existing wells and facilities.

Fig. 2. Recovery factor v. the structural and depositional complexity factors extracted from the complexity index from the same dataset as Fig. 1.

with increasing complexity index holds for most types of oil and gas reservoirs. Two of the complexity factors incorporated into the complexity index specifically relate to compartmentalization: a structural compartmentalization factor and a depositional continuity factor. These two factors are extracted from the overall complexity index and plotted against recovery factor in Figure 2. There is a clear inverse relationship between the compartmentalization ‘score’ and recovery factor (Fig. 2). A statistical analysis of a much larger database (not reported here) indicates that c. 35% of the total variation in recovery factor can be accounted for by the degree of reservoir compartmentalization. Figure 3 shows the compartmentalization complexity scores for a large global database (n . 400) of producing oil and gas fields. This shows that c. 1/3 of the reservoirs fall in the lower two categories (minimal compartmentalization), about 1/3 fall in the middle category (moderate compartmentalization) and the remaining 1/3 fall into the

Fig. 3. Frequency histogram of structural plus depositional complexity scores, extracted from the complexity index for a large database of .400 producing oil and gas reservoirs, covering all depositional environments, structural settings and recovery processes. The data in Figures 1 & 2 are a small subset of the data shown here. The scores are grouped in intervals of 0– 2, 2 –4, 4 –6, 6 –8 and 8 –10.

28


highest two categories (high to extreme degrees of compartmentalization). This indicates that compartmentalization is a very widespread phenomenon, and thus an important control on oil and gas recovery globally. Figure 1 showed a relationship between reservoir complexity and recovery factor. In Figures 2 and 3 we focused on the x-axis of Figure 1 (Complexity Index), to show that a large part of the complexity effect on recovery factor is due to those elements of complexity that relate to compartmentalization. However, it is also instructive to focus on the y-axis of Figure 1 (Recovery Factor) to understand exactly how compartmentalization exerts control on recovery factor. One approach is to break recovery factor into four efficiency factors, where each efficiency factor is represented by a number between 0 and 1, and the product of the four efficiency factors when multiplied together equals the recovery factor (Fig. 4 – see Smalley et al. 2007 for a more detailed explanation). The efficiency factors are: † Eps: The Porescale Displacement efficiency factor that relates to the efficiency of the recovery process to displace oil at the pore scale. This is (Sinitial – Sresidual)/Sinitial, where Sinitial is the initial oil or gas saturation and Sresidual is the residual oil or gas saturation after the recovery process has been applied at the microscopic scale, that is, ignoring large-scale reservoir effects. Eps is often estimated from core-flood experiments. † Ed: The Drainage efficiency factor, representing the proportion of the field connected to a producing well. Ed is measured from reservoir pressure distributions during production. † Es: The Sweep efficiency, representing the degree to which oil is swept from the drained volume into a producing well. Es is measured from fluid saturations during production, often with the aid of reservoir simulation models. † Ec: The Cut-off efficiency, which measures the ability to continue producing the field to include the complete tail of the production profile. Ec is calculated by projecting the field production profile to theoretical completion, and evaluating the fraction lost by ceasing production at a practical economic (or other) cut-off threshold. In this efficiency factor framework, it is Ed and Es that are potentially affected by compartmentalization. Where there are distinct compartments in a field, drainage will be affected when there are more compartments than there are producing wells, because those compartments without a well will be undrained. Sweep is affected by compartmentalization in situations where there are barriers or baffles between producers and injectors.

Fig. 4. Framework for breaking recovery down into 4 efficiency factors: Porescale displacement (Eps), Drainage (Ed), Sweep (Es) and Cut-offs (Ec). Recovery Factor is the product Eps*Ed*Es*Ec.

Using the Ed and Es efficiency factors it is possible to analyze the influence of compartmentalization on recovery factor (Fig. 2). Figures 5 and 6 respectively show Ed and Es plotted against field compartment complexity scores. Figure 5 reveals only a vague correlation between drainage efficiency and compartment score. The fields with the lowest degree of compartmentalization have

Fig. 5. Drainage efficiency factor (Ed) v. compartment complexity score for mature waterflooded fields (cf. Figs 1 & 2).


29

compartmentalization. The full cost impact of managing the compartmentalization can then be incorporated into the field economics and wellfounded decisions made about whether or how to develop, or perhaps divest, the field. In the next two sections we examine two case studies that illustrate the impact of not properly or completely evaluating compartmentalization during field appraisal. Fig. 6. Sweep efficiency factor (Es) v. compartment complexity score for mature waterflooded fields (cf. Figs 1 & 2)

perfect drainage (Ed ¼ 1). However, at the highest degrees of compartmentalization, the drainage is very variable, with Ed varying between 0.4 and 1.0. Indeed, throughout the complete range of compartmentalization scores, there is a healthy proportion of fields with high Ed, approaching 1.0. The reason for this is that the data plotted in Figure 5 are for mature oilfields. In mature fields it is common practice to drill infill wells to produce undrained compartments that have been identified by reservoir surveillance after reservoir production has started, thus raising Ed. Figure 5 therefore shows that, to some degree, the effect of compartmentalization on drainage can be mitigated by drilling more wells. Figure 6 shows a strong relation between Es and degree of compartmentalization, with the least compartmentalized fields having high sweep efficiencies (0.6–1.0) and the most compartmentalized lower Es values (0.3– 0.7). Unlike Ed (Fig. 5), high Es values are restricted to fields with little or no compartmentalization. This indicates that it is much more difficult to mitigate the effects of barriers and baffles on sweep than on drainage. The combined view from Figures 5 and 6 is that the relation between compartmentalization and recovery factor (Fig. 2) is largely governed by the way compartmentalization affects drainage and sweep, and that poor drainage is more easily mitigated by subsequent field activity (e.g. drilling more wells) than is poor sweep. All of the oilfields analysed in Figures 5 and 6 are mature fields that have been producing successfully for several years or more. Clearly then, even fields that are highly compartmentalized can still be successfully developed, even though drilling more wells to mitigate the compartmentalization may affect the economics by requiring more capital expenditure. If the degree of compartmentalization can be assessed quantitatively during appraisal, then this can be taken into account in the design of the field development, including adjusting the amount, position and type of wells to mitigate the

Case-study 1: onshore turbidite oilfield This case-study concerns an onshore oil discovery in turbidite sandstones deposited as numerous channels in a muddy submarine fan system covering c. 80 km2 at a depth of c. 3000 m. The reservoir is overpressured at 6400 psi. The discovery well flowed 4200 BOPD of 1.5 cp crude. Individual turbidite beds in core ranged in thickness from 1 –30 cm, amalgamating in places into massive sand. Core and log data revealed a highly layered comparatively poor-quality reservoir: porosity ranged from 11–24% and permeabilities from 1 –400 mD. Three appraisal wells encountered similar sandstones, and calibration of wells to seismic supported the presence of reservoir away from the wells (Fig. 7a). Oil viscosity varied from 1.5 to 10 cp (GOR of 600– 700 scf/bbl) and short term production rates varied correspondingly from 300 to 1700 BOPD. The presence of different oil types (i.e. wide range in viscosity) suggested some degree of compartmentalization or gravity segregation. Major uncertainties recognized before development were oil quality distribution, net sand distribution, compartmentalization and recovery factor. In-place volumes associated with individual channel sands were quite small, so the development depended upon good connectivity between the sand bodies so that several bodies could be drained by individual producers, and could be swept from injectors to producers. The high net-to-gross predicted from seismic calibrated to the appraisal wells (Fig. 7a) indicated that connectivity should be adequate. It was estimated at sanction that 35 fracturestimulated wells were necessary to drain the reservoir effectively. The various reservoir uncertainties led to a wide range of uncertainty in predicted reservoir performance. The stock-tank-oil-initiallyin-place (STOIIP) uncertainty varied +30% around the base case and ultimate recovery per well +40%. The recovery factor base case was 25%, with an uncertainty range of 17 –33%. The variation in oil composition probably should have been a warning of compartmentalization, but when the first phase of six development wells was implemented the results were worse than even the downside estimates (Fig. 7b). The sands were

30


continuity in vertical and lateral directions, net-togross) scored maximum values. It is difficult to estimate the actual values of the drainage and sweep efficiency factors, due to the early shut-down of the field, but the drainage efficiency was estimated at ,0.1. If this could have been correctly identified before development, this field may have been developed differently, or perhaps not at all.

Case-study 2: offshore turbidite oilfield

Fig. 7. Case-study 1 (a) pre- (upper) and (b) post(lower) development views, showing seismicallyderived net sand thickness. White circles are appraisal wells, black circles are initial development well locations. The development was centred on three thick amalgamated sand areas, but success demanded the whole area to be drained. The post-development view shows that the predicted thick amalgamated sands are in fact absent.

thinner and of poorer quality, with lower net-togross, leading to extensive compartmentalization due to depositional discontinuity. The STOIIP was only 34% of the original base case, the recovery factor only 2% and the recovery per well only 5% of the pre-drill base case. Results were so disappointing that further development was halted and the field shut in. Subsequent post-mortem analysis suggested the degree of compartmentalization could have been estimated more realistically with better calibration of the seismic data (from which net-to-gross was calculated), longer-term well testing, and a quantitative interpretation of the oil compositional variations. The complexity analysis of the reservoir postdevelopment showed that, although the fault-related compartmentalization risk was low, the depositional compartmentalization factors (depositional

This field lies in water depths of c. 500 m, in clastic reservoirs that consist of multiple turbiditic sands that vary between highly channelized to laterallyamalgamated sheet-like sands. The oil is trapped partly by sealing faults that completely offset the reservoirs, partly by stratigraphic pinch-out of the reservoir sands and partly by dip closure. The field is cross-cut by east–west faults, which define several structural segments (Fig. 8). Reservoir quality varies, the more massive sands being of better quality. The sands have 25 –30% porosity and 800–1600 mD permeability. The oil gravity is in the range 22– 288 API. The discovery well was followed by five appraisal wells, then a sixth appraisal well that was put on extended well test (EWT). The early appraisal wells were aimed at understanding the volume of oil in place, while subsequent appraisal and development drilling focused on understanding and reducing the remaining key uncertainties, including the degree of connectivity between the channel complexes and between channel and inter-channel areas. If the connections were found to be poor, then each channel might need a dedicated producer–injector pair, whereas if the connectivity was good, then the field could be developed with fewer wells. The EWT indicated that some of the larger faults were barriers to flow, and this was taken into account in the planning of development well locations (Fig. 8a). Even this very strong set of dynamic appraisal data, absent in the first case-study, did not yield unambiguous results, and interpretations of the degree of connectivity varied. However the base case assumed good connectivity and that most of the field could be drained with 12 horizontal/high angle production wells, supported by 10 water injectors. The pre-development estimate of recovery factor was 36%. Shortly after first oil, it was clear that compartmentalization was greater than expected (Fig. 8b), as demonstrated by rapidly declining well bottomhole flowing pressures, declining production rates and rising GOR in some production wells, and by rising tubing head pressures and decreasing injectivity in some water injectors. Many of the injectors were failing to support their designated producers. As a result, the expected recovery from the initial


31

Fig. 9. Drainage efficiencies in Case-study 2 increase through time as more wells are drilled. The overall increase from 0.60 to 0.85 relates to a doubling of well count.

Fig. 8. Case-study 2, (a) pre- (upper) and (b) post(lower) development views. Blue thick lines are horizontal producing well locations; blue circles are injector locations. Thinner lines are fault locations. Areas are coloured based on the extent of connectivity to producing wells: green, well-connected; yellow/orange, possibly connected; red/purple, poorly connected. In the lower map, faults are coloured blue-red-black in order of increasing likelihood of sealing behaviour. Ticks and crosses refer to presence or absence of communication between injector-producer pairs based on early production data.

stock of wells was reduced from 36 to 22%, and the recovery per well was reduced to 79% of the predevelopment estimates. On the positive side, the STOIIP was estimated to be 46% higher than originally thought. Clearly, compartmentalization had been underestimated during appraisal, despite the extensive appraisal programme. Nevertheless, this field is now highly successful, demonstrating that it is possible to mitigate compartmentalization by drilling additional thoughtfully-placed wells. The initial drainage efficiency was estimated at only 0.60. Doubling the well count over the next few years raised Ed to about 0.85 (Fig. 9). What would have been done differently if the degree of compartmentalization was better understood during appraisal? In this case the field would still undoubtedly have been developed, but the development well number and placement may have been different. The timing of the wells may

also have been optimized, with a greater number of initial development wells achieving good production profiles. These two case studies illustrate the ways in which compartmentalization can affect reservoir performance, and how development decisions could be improved greatly if compartmentalization was better understood during appraisal. The following sections deal with how an understanding of fluid mixing rates can help improve the early diagnosis of compartmentalization.

Reservoir fluid mixing processes As described earlier, it has commonly been assumed that spatially varying, fluid related properties are indications of either barriers to flow (e.g. Kaufman et al. 1990a; Elshahawi et al. 2005) or recent external processes affecting the reservoir (e.g. Montel et al. 1993; Berg et al. 1994; Zawisza 2004), if they diverge from a steady state distribution. Examples of these properties include pressure, contact depth, formation water composition, hydrocarbon density, GOR or composition. In the past, effort has been focused primarily on methods for identifying differences (e.g. Kaufman et al. 1990b; Hwang & Baskin 1994; Hwang et al. 1994) or whether observed property variations are statistically significant (e.g. Stølum & Smalley 1992; Stølum et al. 1993). In reality it may take a significant period of time for a given distribution of these properties to relax back to the steady state once the process that caused the perturbation has stopped. As well as identifying that there is a non-steady state distribution of a given fluid-property, it is crucial to be able to: (1) (2)

Estimate the time-scale over which the property-distribution of interest will equilibrate to the steady state. Integrate this with mixing times estimated for other fluid-property distributions in the reservoir.

32


(3)

Compare the mixing time-scales with the age of the events that led to the disturbance in fluid properties. If this is not known then the age of the reservoir can be used as a maximum time limit. If the mixing time-scale is greater than the age of the reservoir (or the fluid-perturbation event) then the disequilibrium distribution may simply be a relict of reservoir filling or other dynamic processes that affected the reservoir in the past. If the mixing timescale is less than the age of the perturbation event then either there is a barrier or baffle preventing reservoir mixing or there is a more recent (or ongoing) geological process that has acted on the reservoir. For example, suppose there is an observation of different oil –water contact depths in two wells in a newly discovered reservoir. If the time taken for the observed oil–water contact depths to equilibrate in the absence of barriers is much less than the age of the reservoir then it is possible that there is a barrier or baffle to flow between the wells or that the reservoir may be affected by regional hydrodynamics (e.g. Tozer & Borthwick 2010). Comparison of mixing times for different properties and their observed distributions can clarify the likelihood of compartmentalization in that reservoir in the event of conflicting data, for example, when the pressure distribution appears to be steady state but the water composition shows spatial variability. In this section we review a series of analytical equations that have been used previously for providing order of magnitude estimates of time-scales for observed pressure distributions, tilted contacts or water and hydrocarbon compositional variations to return to their steady state distributions. We shall assume that: (1)

(2)

The reservoir is isothermal so thermal diffusion and convection effects are neglected. These influences have been discussed by Jacqmin (1990), Padua (1999) and Montel et al. (1992, 2007), amongst others. Montel et al. (2007) use estimates of characteristic times to show that thermal convection will be most important for fluids that are initially compositionally homogeneous, but these characteristic times are not appropriate for estimating reservoir mixing time-scales. Flow is single phase. The presence of more than one fluid in the pore-space will tend to reduce the permeability of the porous medium to the fluid of interest by a factor given by the relative permeability of that fluid’s saturation if the mixing is driven by Darcy flow. If the flow is driven by molecular diffusion then the effect of another fluid in the pore-space is to increase the tortuosity of the porous media. In

(3)

(4)

the first instance it is assumed single phase fluid properties will provide a reasonable order of magnitude estimate of mixing times. There are no ongoing external processes affecting reservoir flow, for example regional hydrodynamics. The existence of such processes would require solution of the flow equations to include the presence of a source or sink term. The purpose of using the analytical solutions described in the following sections is to assess mixing times in the absence of barriers/baffles or external processes. If the mixing times are much less (by at least an order of magnitude) than the probable time since the reservoir was affected by external geological processes then the presence of nonsteady state fluid distributions indicates that the engineer/geoscientist needs to perform further investigations to determine the likelihood of barriers/baffles or external processes. The mixing of one fluid-property by a particular process is unaffected by the mixing of any other property. For example, time-scales for mixing of hydrocarbon compositional gradients by molecular diffusion are independent of any mixing that might occur by gravitational overturning resulting from associated density gradients.

The analytical expressions presented provide reasonable, order of magnitude, preliminary estimates of mixing time-scales that are suitable for comparing with the estimates of reservoir age (or time since a fluid-perturbation event) during field appraisal. Inevitably they rely on an estimate of various reservoir properties that may not be well known and indeed may vary across the reservoir. Moreover they have been derived for idealized reservoir geometries and distributions. They are intended to enable engineers and geologists to examine very rapidly, using a spreadsheet or calculator, † the probability of reservoir compartmentalization based on observed fluid-property distributions, † the sensitivity of mixing time-scale to uncertainties in the range of possible reservoir properties, † baffle/barrier properties needed to maintain observed fluid-property distributions. Further, more involved investigations of the reservoir may be required to characterize fully the inferred compartmentalization or clarify whether the reservoir is affected by other external processes such as regional hydrodynamics. The time-scales for mixing are compared here using the simple model reservoir shown in Figure 10 and the data shown in Table 1. The fluid


Fig. 10. Model reservoir used to assess time-scales for reservoir mixing for pressure, gravitational overturning and molecular diffusion. The reservoir is homogeneous with a uniform permeability of 100 mD and porosity of 0.2.

properties in Table 1 were chosen to span a range of fluid types: water, gas, black oil, and heavy/viscous oil. Using this range of fluids highlights the impact of differing viscosity, density and compressibility on mixing time.

Molecular diffusion Some reservoirs exhibit lateral and/or vertical variations in hydrocarbon and formation water composition. The hydrocarbon compositional variations may be subtle or may affect more obvious properties such as density, bubble-point pressure or GOR.

33

Such variations can be the result of gradual maturation of the source rock(s) during the reservoir filling history and the geometry of the reservoir with respect to the source rock, or even the result of adsorption of different components onto different minerals in the rock matrix during migration. Hydrocarbon composition is determined in the laboratory from fluid samples obtained via Repeat Formation Testers (RFT) and Modular Formation Dynamics Testers (MDT) or during production tests. The spatial resolution of such data is limited by the availability of fluid samples. Composition can also be determined from oil or water extracted from core samples. Data obtained by this means will have greater spatial frequency but may be less reliable due to contamination or fractionation effects. Changes in composition can also be inferred from coarse estimates of density obtained from vertical pressure gradients. Water composition variations can result from chemical reactions between water and the rock matrix (e.g. dissolution of minerals) and regional flows (e.g. Berg et al. 1994; Bachu 1995). Considerable research has been dedicated to determining which fluid components are both easy to measure and most likely to be sensitive to

Table 1. Mathematical symbols, property description, property values and units used in the mixing calculations described in the text Symbol L L H HB f K kB G cw cho co cg mw mho mo mg rw rho ro rg rr D KD t

Property

Value/units

Length of reservoir Width of reservoir Thickness of reservoir Thickness of baffle Porosity Reservoir permeability Baffle permeability Acceleration due to gravity Effective compressibility of water-filled reservoir Effective compressibility of heavy oil filled reservoir Effective compressibility of black oil filled reservoir Effective compressibility of gas filled reservoir Water viscosity Heavy oil viscosity Black oil viscosity Gas viscosity Water density Heavy oil density Black oil density Gas density Rock density Molecular diffusion coefficient Distribution coefficient Tortuosity

1000 m 1000 m 100 m 1m 0.2 100 mD 1026 D 9.81 m s22 7 10210 Pa21 8 1029 Pa21 2.5 1028 Pa21 2.5 1027 Pa21 1 cP 20 cP 0.9 cP 0.05 cP 1000 kg m23 850 kg m23 700 kg m23 200 kg m23 2000 – 2800 kg m23 1029 m2 s21* 2 – 6 1023 m3 kg21 2 – 8†

*Typical order of magnitude value for hydrocarbon diffusion in liquid phase, see Poling et al. (2001). † Wang et al. (2005).

SI equivalent (value/units)

10213 m2 10218 m2

1023 Pa s 2 1022 Pa s 9 1024 Pa s 5 1025 Pa s

34


reservoir compartmentalization (Kaufman et al. 1990b; Smalley & Hale 1996; Elshahawi et al. 2005). Other workers have focused on methods for determining the final, vertical steady state distribution of hydrocarbon components due to the interaction between gravitational segregation and molecular diffusion (Schulte 1980; Holt et al. 1983; Høier & Whitson 2001; Ratulowski et al. 2003) and sometimes thermal effects (Jamet et al. 1992; Pederson & Lindeloff 2003). Lateral changes in the vertical hydrocarbon compositional profile are then used to infer compartmentalization. This is despite the fact that England et al. (1987) showed that lateral diffusional mixing is very slow on reservoir time-scales. A commonly used tracer in formation waters for detecting formation water variations and thus investigating compartmentalization is the 87Sr/86Sr isotope ratio (e.g. Smalley et al. 1995; Mearns & McBride 2001). This ratio is particularly useful because it is not affected by dilution or precipitation reactions, enabling formation water compositions to be measured from core samples, even when longterm storage has meant the water has evaporated leaving behind only salt residues. Molecular diffusion will tend to smooth out compositional variations within a fluid. In a binary mixture in 1D the process is described by the diffusion equation: @ci @2 c ¼D 2 @t @x

(1)

where D is the diffusion coefficient, c is the concentration of the component of interest, t is time and x is distance. The time-scale for the complete mixing of two, initially segregated, miscible fluids in a porous medium via molecular diffusion is given approximately by t

L2 t D

(2)

where L is the distance over which the two fluids are mixing and t is the tortuosity of the pore space. However, the time-scale may be considerably different depending on the initial concentration distribution, the geometry and properties of the

container and other boundary conditions (see Crank 1979 or Carslaw & Jaeger 1959). In formation waters the mixing of ions, for example identified by differences in strontium isotopes from residual salts in core, is slowed further because of adsorption of strontium ions onto the pore walls. If the pore space is completely filled with formation water then the mixing time is given approximately by (Smalley et al. 1995)

t

L2 t (1 f) 1þ rr KD D f

where f is the porosity, rr is the density of the rock and KD is the distribution coefficient between the adsorbed and free ions. This assumes that the timescale for the partition of the strontium ions between the connate water and the pore walls is much quicker than the timescale for diffusion of ions between pores. Previously published comparisons of experimental work with analytical models for the nuclear waste industry suggest this is a reasonable model (Ho¨ltta¨ et al. 2001; Mell et al. 2006). As discussed earlier, the presence of hydrocarbon fluids in the pore spare will tend to increase the time-scale for mixing by increasing the tortuosity. The other parameters in equation (3) will not be affected by the presence of hydrocarbon as the partition of strontium ions into the oil is negligible. Table 2 gives estimates of single phase mixing times for oil and formation water via molecular diffusion, using the data given in Table 1. As shown, compositional variations in the 87Sr/86Sr isotope ratio may mix vertically (in other words, at the intrawell scale) via molecular diffusion in times that are short compared to reservoir ages, with the possible exception of low porosity rocks with a high distribution coefficient (i.e. highly clay-rich lithologies). Thus, vertical compositional changes within a well that are different from the expected steady state distribution will indicate barriers to vertical flow. However, lateral mixing over distances typical of an offshore well spacing takes a much longer time. Thus observations of compositional differences between appraisal wells may simply reflect the filling history of the reservoir rather than the existence of barriers to lateral flow.

Table 2. Time-scales for mixing of oil or formation water composition by molecular diffusion Oil compositional gradient L ¼ 10 m L ¼ 1000 m

6000 years 60 000 000 years

(3)

Formation water (low)

Formation water (high)

100 000 years 1.1 109 years

4 000 000 years 4 1010 years


35

Pressure Fluid pressure is one of the most commonly measured properties in oil and gas reservoirs. RFT and MDT can measure pressure to sub-psi precision at multiple depths in the reservoir. Accuracy is usually less good but this is less of an issue as in most cases it is the change in pressure with depth and from well to well for a given depth that is of most interest. Pressure disequilibrium can be caused by a number of processes (Osborne & Swarbrick 1997). These include changing rates of burial or uplift across a field, ongoing migration of hydrocarbons, local production or breakdown of hydrocarbons due to thermal cracking or microbial action, hydrodynamic effects and diagenetic reactions as well as depletion due to production from a neighbouring connected field. The dissipation of a given pressure distribution p(x) in an isothermal reservoir is described by the pressure diffusion equation: ce f

@p k @2 p ¼ @t m @x2

(4)

where f is the porosity, k is the permeability, m is the viscosity of the fluid in the reservoir, x is distance, ce ¼ (cb =f þ cf ) is the effective compressibility of the reservoir –fluid system and is a function of the bulk rock compressibility cb, the porosity and the fluid compressibility cf. In the absence of any barriers or baffles pressure will return to its steady state distribution over distance L in time t

L2 mfce k

(5)

where time is calculated in seconds if the other properties are entered in SI units. Thus the time taken for abnormal pressures to dissipate increases with the square of the reservoir length (or thickness), the fluid viscosity, the rock porosity and the effective compressibility. The time decreases as the permeability increases. If an abnormally pressured compartment of the reservoir is separated (as in Fig. 11) by a thin, low transmissibility fault or shale (imagine the figure rotated through 908) from the rest of a normally pressured reservoir then Luo & Vasseur (1997) and Muggeridge et al. (2005) showed that the time for abnormal pressures to dissipate can be approximated by

t

mfce LHB dP ln 2kB DP

(6)

Fig. 11. Schematic of an abnormally pressured compartment separated from the main, normally pressured reservoir by a thin, low permeability barrier or baffle.

where HB is the baffle thickness, kB is the baffle permeability, dP is the precision of the pressure measurement and DP is the observed pressure difference. Muggeridge et al. (2005) derived this expression from a solution given in Carslaw & Jaeger (1959) to the diffusion equation for boundary and initial conditions relevant to this problem. This expression is only valid if the baffle thickness and permeability are much smaller than the reservoir length and permeability. Table 3 compares the time taken for pressure to dissipate in the absence of a baffle (eq. 5) with that taken when there is a low permeability baffle (eq. 6) for the rock and fluid properties given in Table 1. In both cases excess pressures dissipate back to a steady state distribution very quickly on geological time-scales. Pressure dissipates most slowly in a heavy oil reservoir due mainly to the high viscosity of the oil. It dissipates most quickly if the formation is filled with water due to the very low compressibility of the water. Abnormal pressure distributions will dissipate in roughly the same time in reservoirs filled with black oil or gas because although the gas has a much lower viscosity it also has a much higher compressibility. All fluids show pressure equilibration times that are relatively short on a geological time-scale, but are significant on a production time-scale. Thus an observation of a steady state pressure distribution in a reservoir does not guarantee that the reservoir is well connected. There may still be low permeability baffles compartmentalizing the reservoir and affecting recovery. Alternately, observations of any abnormal pressure differences between wells or layers are a strong indication of either a barrier to flow (compartmentalization) or perhaps

Table 3. Time-scales for abnormal pressures to dissipate over 1 km Fluid Water Black oil Heavy oil Gas

Time, no baffle

Time, with baffle

8 days 260 days 1800 days 160 days

5 years 165 years 2150 years 100 years

36


regional aquifer flow. The latter possibility may be confirmed if there are also observations of differences in contact depth consistent with the observed abnormal pressure gradients.

Gravitational overturning Gravitational overturning affects reservoirs that have lateral fluid density variations. This applies to hydrocarbons that have inherited a spatially varying density due to the filling history and geometry of the reservoir with respect to the source rock. It also applies to gas–oil, gas–water and oil –water contacts that have been tilted during reservoir filling due to filling from different directions (England et al. 1987) or subsequent restructuring. Pressure communication with other, nearby producing fields (see Coutts 1999 for example) or regional aquifer flux may also result in tilted contacts (e.g. Berg et al. 1994; Stenger et al. 2001; Zawisza 2004). As discussed in the section on molecular diffusion, hydrocarbon and water composition can be determined from fluid samples or core extracts. Fluid contacts are usually detected via resistivity logging although they can also be inferred from pressure gradient data and core saturation measurements. Contacts can also be identified by flat spots in seismic data but the vertical resolution may not be sufficient to identify subtle variations in contact depth as an indicator of reservoir compartmentalization. An analytical expression for density driven overturning was tested and used by England et al. (1995) to infer a significant transmissibility barrier between the Forties and South East Forties Fields. The expression used was

t

25 L2 mf 4 kgHDr

(7)

where, as in previous expressions, time is calculated in seconds if other parameters are input in SI units. L is the reservoir length, m is the fluid viscosity, f is the reservoir porosity, g is the acceleration due to gravity, H is the reservoir thickness and Dr is the density difference; k is the horizontal permeability. This describes the time taken for an initially vertical interface to rotate until it is inclined at an angle H/L to the horizontal, as shown in Figure 12. It is only valid when the vertical transient flow is much smaller than the horizontal transient flow, that is, the vertical permeability is large. This expression was derived from earlier work by Gardner et al. (1962) who showed that the equation of motion of the interface, at depth h from the reservoir top, between two miscible

Fig. 12. Schematic of gravitational overturning as described by equation (7). The reservoir is initially filled with two fluids of different density separated by a vertical interface. Gravitational forces then cause the interface to rotate so that the denser fluid lies under the lighter fluid.

fluids at later times was described by a diffusion equation of the form @j @ @j ¼ D @t @x @x

(8)

where D¼

kDrgH (1 j) fm

(9)

and j ¼ H/h. Note that in this case the diffusion coefficient is also a function of interface depth. The diffusive nature of buoyancy driven frontal dynamics has been re-discovered more recently by Seón et al. (2007a, b). Using the reservoir and rock properties given in Table 1 we can estimate that the time taken for a vertical interface between two miscible oils of viscosity 0.9 cP and with a density difference of 10 kg m23 to relax back to an angle of H/L to the horizontal is about 36 000 years. Equation (7) can also be used to estimate the time taken for an initial vertical contact between two immiscible fluids to overturn due to gravity (England et al., 1995). The overturning times are given in Table 4. Table 4. Estimates of time for an initially vertical interface between two fluids to relax back to an angle of H/L to the horizontal Fluids

Time (years)

Oil – oil (density difference 10 kg m23) Oil – water Heavy oil– water Gas – water

36 000 12 000 54 000 2500


These times are clearly several orders of magnitude greater than those estimated for abnormal pressures to revert to their steady state distribution but still significantly less than reservoir ages. The heavy oil–water contact overturning time is greater than that for regular oil–water contacts due to the greater viscosity of the heavy oil and the lower density contrast with water. Overall the mixing times indicate that observations of horizontal density gradients will only be observed in reservoirs if there is either a barrier or baffle preventing or reducing horizontal flow, or some process, such as a dynamic aquifer, maintaining that density contrast, or possibly a reduced vertical permeability.

Discussion Comparison of the mixing times calculated for the different mixing processes shows that mixing of unsteady state pressure distributions is quickest, and mixing via molecular diffusion is the slowest process. Molecular diffusion is so slow that it is unlikely that horizontal compositional gradients will mix by this process alone over typical timescales since reservoirs filled. Even vertical compositional gradients are unlikely to have mixed purely by molecular diffusion if the vertical length scale is much greater than 10 m. The mixing times for gravitational overturning are intermediate between those for pressure and molecular diffusion but are typically two or three orders of magnitude less than potential reservoir ages. Based on these estimates it would appear that: † Non-equilibrium distributions of pressure, hydrocarbon density or fluid contacts will often reflect compartmentalization or the influence of very recent (or ongoing) processes such as regional aquifer flow, reservoir filling, restructuring, and so on. † Non-equilibrium variations of water composition (e.g. 87Sr/86Sr isotope ratios) over length scales greater than c. 10 m do not necessarily indicate compartmentalization if molecular diffusion is the only operating mixing process, as it is very slow over these length scales. In many cases, of course, diffusion will occur in addition to other mixing processes. † Non-equilibrium horizontal variations in hydrocarbon composition may indicate compartmentalization (or a recent/ongoing perturbation) if they are associated with a significant density gradient. In this case mixing will be primarily via gravitational overturning which is a relatively quick process. If the associated density gradients are small then the main mixing mechanism will be molecular diffusion, which is very slow. In these cases the non-equilibrium gradients may

37

simply be a relict of reservoir filling or other processes. † Observations of normal or constant hydrostatic pressure gradients across a reservoir do not in themselves preclude the existence of low permeability baffles that will reduce recovery. In these circumstances it is worth analysing compositional (density) and contact depth distributions to determine if these are steady state or not. † Observations of uniform contact depths across a reservoir also do not preclude the existence of baffles or barriers, as these also return to steady state quickly compared with geological timescales although more slowly than do pressures. Again, other static (log, seismic) and dynamic (pressure, hydrocarbon and formation water composition) data should be integrated into the interpretation. These mixing estimates neglect the influence of thermal convection on the mixing of hydrocarbon composition. Thermal convection mixes by changing the density of the hydrocarbon as a function of temperature and as a result the time-scales are rather short compared with a typical age of a reservoir (Montel et al. 1992; Nasrabadi 2008). If present, it will establish a horizontal steady state gradient in hydrocarbon composition. However it is likely to be ineffective if the hydrocarbons are vertically segregated with the proportion of heavier components increasing with depth, and where there are no horizontal temperature gradients (Jacqmin 1990; Montel et al. 2007).

A unified diffusive model for reservoir fluid mixing All the equations for estimating mixing time-scales described above have been obtained by solving a partial differential equation that has the same form as that for molecular diffusion (eq. 1). Thus each mixing process can be modelled as a diffusive process by modifying the effective diffusion coefficient appropriately, as shown in Table 5. Examination of the effective diffusion coefficients for the different mixing processes shows that, in agreement with the foregoing discussion, the effective diffusion coefficient for pressure mixing is largest and that for mixing of water composition (the 87 Sr/86Sr isotope ratio) is smallest. The ranges of equilibration-times and distances of these diffusionbased processes are shown graphically in Figure 13. This plot also provides an easy means for assessing whether or not reservoir fluids should be in steady state. The effective diffusion coefficient(s) can be calculated using the expressions given in Table 5 by substituting the appropriate rock and fluid properties. The intersection of the effective

38


Table 5. Expressions for mixing time, effective diffusion coefficient and range of values (order of magnitude)

Pressure (no baffle) Gravitational overturning Molecular diffusion (hydrocarbon) Molecular diffusion (87Sr/86Sr ratio)

Time to mix (seconds)

Effective diffusion coefficient, De (m2 s21)

Range of De for k ¼ 100 mD (m2 s21)

1 L2 mfce 2 k 25 L2 mf 4 kgHDr L2 t D

2k mfce 4 kgHDr 25 mf D t

10 – 1022

L2 t (1 f) 1þ rr KD D f

D t(1 þ (1 f=f)rr KD )

1025 – 1028 1029 – 10210 10211 – 10213

These were calculated for a reservoir permeability of 100 mD, using the values given in Table 1.

diffusion coefficient(s) with the estimated age of the reservoir (or time since perturbation, if known) provides an indication of the distance over which different fluid related properties will be expected to have mixed. For example, ordinary black oil with viscosity and density as in Table 1 in a 100 mD reservoir would have the following expected effective diffusion coefficients: † DP ¼ 4 1022 m2 s21, pressure equilibration † Dg ¼ 3 1026 m2 s21, gravitational overturning of an oil–water contact † Dg ¼ 3 1028 m2 s21, gravitational overturning due to an oil density difference of 9 kg m23 † Dm ¼ 6 10210 m2 s21, molecular diffusion Figure 14 illustrates that, if the time available for equilibration is about 2 million years, compositional variations in the oil will equilibrate by molecular

diffusion over length scales less than 1 km. Density driven overturning would equilibrate these variations over length scales less than 10 km. Overturning of the water–oil contact would take less than 10 000 years and pressures would equilibrate in less than 10 years over a 10 km length scale. Thus any observed unsteady state distributions in pressure or the oil –water contact are indications of compartmentalization or a more recent perturbation. Variation in oil composition over distances of ,1 km may also be suggestive of compartmentalization, but changes over greater distances may simply be due to the reservoir not having been in existence long enough for these variations to have mixed. Using this approach, if the age of the perturbation event is known (e.g. where detailed basin modelling constrains the timing of reservoir filling), then the time-scale is known and the relations described above can be used to calculate the permeability of the barrier that would be needed to maintain fluid disequilibrium over that timescale. This is an important step in the quantification of reservoir compartmentalization.

Implications for field appraisal

Fig. 13. Graph showing mixing time as a function of length scale for different values of effective diffusion coefficients ranging from 1023 m2 s21 to 10213 m2 s21. The mixing processes typical of particular ranges of effective diffusion coefficient are shown, highlighting the fact that molecular diffusion is the slowest mixing process and pressure-driven flow is the quickest. Values were calculated using rock and fluid properties given in Table 1.

This relatively simple overall view of fluid mixing lends itself to the production of simple spreadsheetbased tools to relate fluid disequilibrium to the quantitative interpretation of reservoir compartmentalization. Such tools would be of great value to industry practitioners when evaluating appraisal data such as those described earlier in the two case studies. In Case-study 1, the different oil densities between the appraisal wells about 1 km apart (Fig. 7), if interpreted using the methods described herein, would have indicated an equilibration


39

barriers (Fig. 13), very rapidly compared to the oil emplacement time of .10 million years, indicating compartmentalization. In this case such knowledge would not have jeopardized the development, but development wells could have been placed more optimally so that drainage and sweep efficiencies were maximized.

Conclusions

Fig. 14. Illustration of how the previous figure may be used to assess reservoir compartmentalization. If the age of fluid-perturbation (maximum value taken as age of reservoir) is 2 million years, then any hydrocarbon compositional variation will have had time to mix over length scales less than c. 1 km by molecular diffusion (Dm) and length scales less than 5 km by gravitational overturning (Dgoo), for the calculated values of the effective diffusion coefficients. Pressure variations (DP) will have equilibrated over length scales of 10 km in about 10 years and overturning of the water–oil contact (Dgwo) would take 10 000 years for a 10 km length scale. Both these time-scales are much shorter than the time since the original fluid was perturbed as shown by the fact that the contours for these diffusion coefficients (shown as dashed lines) do not intersect the line for the estimated age of the reservoir.

time for density-driven mixing of 103 –105 years (Fig. 13) assuming reservoir permeabilities of the order of 100 mD. Compared to the basin modelling information that suggests at least 4 106 years since oil filling, this could have been seen as a strong indicator of compartmentalization (permeabilities would need to be less than 2 mD for gravitational overturning to still be mixing the fluids), leading to a modified development case, or a decision not to develop. In Case-study 2, there was a reliance on an extended well test, generally a very high quality source of information. Pressure depletion in nearby (,1 km) wells as a result of production from the EWT was interpreted as indicative of good connectivity. However, application of the methodology presented here would have shown that pressure transmission is so rapid on such length scales that such pressure depletions would be expected even where connectivity is sufficiently poor on a production timescale to result in poor sweep efficiency. On the other hand, slight differences in oil– water contacts in different parts of the field could have taken on new significance if it had been understood that these would have equilibrated in thousands of years in the absence of

This paper has examined how reservoir compartmentalization affects oil recovery and how quantification of reservoir fluid mixing can be used to assess the likelihood and degree of compartmentalization. A review of compartmentalized fields has shown that, in general, compartmentalization reduces recovery by reducing both drainage and sweep efficiencies. The effect on drainage efficiency can be effectively mitigated by drilling additional wells, though at extra cost. Mitigating reduced sweep efficiency appears to be more difficult. Two specific case studies have highlighted the potential impact on field performance of compartmentalization not being fully appreciated during appraisal. Methods for estimating the mixing times of initially non steady state distributions of pressure, fluid density, fluid contacts and hydrocarbon and formation water composition have been reviewed. A simplified model reservoir has been used to show that: † The mixing times for pressure variations are much quicker than for density driven overturning which in turn is much quicker than molecular diffusion. † Interpretations of steady state or unsteady state fluid-property distributions may be misleading if not seen in the context of fluid mixing times. For example a steady state distribution of pressure does not guarantee the reservoir is not compartmentalized. It may simply mean that there has been sufficient time for pressure to equilibrate through existing baffles over geological time-scales. † The analysis of all available types of fluidproperty data should be integrated in order to provide a better assessment of the likelihood of compartmentalization. † All key mixing processes can be modelled as diffusive in nature. This enables a single simple graphical representation of reservoir mixing over different time and length scales by calculation of the appropriate effective diffusion coefficients for the different mixing processes and reservoir properties. We thank BP for supporting the development of these mixing models and for permission to publish this work.

40


References Bachu, S. 1995. Synthesis and model of formation-water flow, Alberta Basin, Canada. American Association of Petroleum Geologists Bulletin, 78, 501–518. Bennett, B., Lager, A., Potter, D. K., Buckman, J. O. & Larter, S. R. 2007. Petroleum geochemistry proxies for reservoir engineering parameters. Journal of Petroleum Science and Engineering, 58, 355–366. Berg, R. R., Demis, W. D. & Mitsdarffer, A. R. 1994. Hydrodynamic effects on mission canyon (Mississippian) oil accumulations, billings nose area, North Dakota. American Association of Petroleum Geologists Bulletin, 79, 1159–1178. Carslaw, H. & Jaeger, J. C. 1959. Conduction of Heat in Solids. 2nd edn. Oxford University Press, Oxford. Coutts, S. D. 1999. Aquifer behaviour during Brent depressurization and impact on neighbouring fields. Society of Petroleum Engineers Reservoir Engineering and Evaluation, 2, 53–61 Crank, J. 1979. The Mathematics of Diffusion. 2nd edn. Oxford University Press, Oxford. Dromgoole, P. & Speers, R. G. 1997. Geoscore: a method for quantifying uncertainty in field reserve estimates. Petroleum Geoscience, 3, 1 –12. Elshahawi, H., Hasem, M., Mullins, O. C. & Fujisawa, G. 2005. The missing link – identification of reservoir compartmentalization through downhole fluid analysis. Society of Petroleum Engineers, SPE Paper 94709, presented at the Society of Petroleum Engineers Technical Conference & Exhibition held in Dallas, Texas, USA, 9–12 October. England, W. A., Mackenzie, A. S., Mann, D. M. & Quigley, T. M. 1987. The movement and entrapment of petroleum fluids in the subsurface. Journal of the Geological Society, London, 144, 327– 347. England, W. A., Muggeridge, A. H., Clifford, P. J. & Tang, Z. 1995. Modelling geological mixing rates in petroleum reservoirs to detect flow barriers: theoretical considerations and case history from the Forties Field (UKCS). In: Cubitt, J. M. & England, A. A. (eds) The Geochemistry of Reservoirs. Geological Society, London, Special Publications, 86, 185– 201. Gardner, G. H. F, Downie, J. & Kendall, H. A. 1962. Gravity segregation of miscible fluids in linear models. Society of Petroleum Engineers Journal, 95, 95–104. Høier, L. & Whitson, C. H. 2001. Compositional grading – theory and practice. Society of Petroleum Engineers Reservoir Engineering & Evaluation, 3, 525–535. Holt, T., Lindeberg, E. & Ratkje, S. K. 1983. The effect of gravity and temperature gradients on methane distribution in oil reservoirs. Society of Petroleum Engineers, SPE Paper 11761. Ho¨ltta¨, P., Siitari-Kauppi, M., Hakanen, M. & Tukiainen, V. 2001. Attempt to model laboratoryscale diffusion and retardation data. Journal of Contaminant Hydrology, 47, 139–148. Huang, H. P., Larter, S. R., Bowler, B. F. J. & Oldenburg, T. B. P. 2004. A dynamic biodegradation model suggested by petroleum compositional gradients within reservoir columns from the Liaohe basin, N.E. China. Organic Geochemistry, 35, 299–316.

Hwang, R. J., Ahmed, A. & Moldowan, J. M. 1994. Oil composition variation and reservoir continuity: Unity Field, Sudan. Organic Geochemistry, 21, 171–188. Hwang, R. J. & Baskin, D. K. 1994. Reservoir connectivity and oil homogeneity in a large-scale reservoir. Geoscience Geo94, 2, 529–541. Jacqmin, D. 1990. Interaction of natural convection and gravity segregation in oil/gas reservoirs. Society of Petroleum Engineers Reservoir Engineering, May issue, 5, 233– 238. Jamet, P., Fargue, D., Costesque, P. & Cernes, A. 1992. The thermo-gravitational effect in porous media: a modelling approach. Transport in Porous Media, 9, 223–240. Jiao, J. J. & Zheng, C. 1998. Abnormal fluid pressures caused by deposition and erosion of sedimentary basins. Journal of Hydrology, 204, 124–137. Kaufman, R. L., Kabir, C. S., Abdul-Rahman, B., Quttainah, R., Dashtl, H., Pederson, J. M. & Moon, M. S. 1990a. Characterizing the Greater Burgan field with geochemical and other field data. Society of Petroleum Engineers Reservoir Engineering & Evaluation, 3, 118–126. Kaufman, R. L., Ahmed, A. S. & Elsinger, R. L. 1990b. Gas chromatography as a development and production tool for fingerprinting oils from individual reservoirs: applications in the Gulf of Mexico. In: 9th Research Conference Proceedings. Gulf Coast Section of the Society of Economic Paleontologists and Mineralogists, Earth Enterprises, Austin, Texas, USA, 263–282. Larter, S., Wilhelins, A. et al. 2003. The controls on the composition of biodegraded oils in the deep subsurface – part 1: biodegradation rates in petroleum reservoirs. Organic Geochemistry, 34, 601–613. Luo, X. & Vasseur, G. 1997. Sealing efficiency of shales. Terra Nova, 9, 71–74. Matava, T., Rooney, M. A., Chung, H. M., Nwanko, B. C. & Unomah, G. I. 2003. Migration effects on the composition of hydrocarbon accumulations in the OML 67–70 areas of the Niger Delta. American Association of Petroleum Geologists Bulletin, 87, 1193– 1206. Mearns, E. W. & McBride, J. J. 2001. Strontium isotope analysis can help define compartmentalization. Oil & Gas Journal, 99, 70–75. Mell, P., Megyeri, J., Riess, L., Ma´the´, Z., Ha´mos, G. & La´za´r, K. 2006. Diffusion of Sr, Cs, Co and I in argillaceous rock as studied by radiotracers. Journal of Radioanalytical and Nuclear Chemistry, 268, 411–417. Montel, F., Bickert, J., Lagisquet, A. & Galie´ro, G. 2007. Initial state of petroleum reservoirs: a comprehensive approach. Journal of Petroleum Science and Engineering, 58, 391– 402. Montel, F., Caillet, G., Pucheu, A. & Caltagirone, J. P. 1993. Diffusion model for predicting reservoir gas losses. Marine and Petroleum Geology, 10, 51–57. Montel, F., Calatagirone, J. P. & Pebayle, L. 1992. A convective segregation model for predicting reservoir fluid compositional distribution. In: King, P. R. (ed.) The Mathematics of Oil Recovery. Clarendon Press, Oxford.

GET IT BEFORE IT GETS YOU Muggeridge, A. H., Abacioglu, Y., England, W. A. & Smalley, C. 2004. The dissipation of abnormal pressures in the sub-surface. Journal of the Geophysical Research (Solid Earth), 109, B11104, doi: 10.1029/ 2003JB002922. Muggeridge, A. H., Abacioglu, Y., England, W. A. & Smalley, C. 2005. The rate of pressure dissipation from abnormally pressured compartments. American Association of Petroleum Geologists Bulletin, 89, 61– 80. Nasrabadi, H., Firoozabadi, A., Oliveira, R. & Vieira, A. J. M. 2008. Interpretation of an unusual bubblepoint pressure variation in an offshore field. Society of Petroleum Engineers, SPE Paper 113574. Osborne, M. J. & Swarbrick, R. E. 1997. Mechanisms for generating overpressure in sedimentary basins: a reevaluation. American Association of Petroleum Geologists Bulletin, 81, 1023– 1041. Padua, K. G. O. 1999. Nonisothermal gravitational equilibrium model. Society of Petroleum Engineers Reservoir Engineering, 2, 211– 217. Pederson, K. S. & Lindeloff, N. 2003. Simulation of compositional gradients in hydrocarbon reservoirs under the influence of a temperature gradient. Society of Petroleum Engineers, SPE Paper 84364. Poling, B. E., Prausnitz, J. M. & O’Connell, J. 2001. The Properties of Gases and Liquids. McGraw-Hill, New York. Ratulowski, J., Fuex, A. N., Westrich, J. T. & Sieler, J. J. 2003. Theoretical and experimental investigation of isothermal compositional grading. Society of Petroleum Engineers Reservoir Engineering & Evaluation, June, 6, 168–173. Schulte, A. M. 1980. Compositional variation within a hydrocarbon column due to gravity. Society of Petroleum Engineers, SPE Paper 9235. Seón, T., Znaien, J., Salin, D., Hulin, J. P., Hinch, E. J. & Perrin, B. 2007a. Transient buoyancy-driven front dynamics in nearly horizontal tubes. Physics of Fluids, 19, 123603. Seón, T., Znaien, J., Perrin, B., Hinch, E. J., Salin, D. & Hulin, J. P. 2007b. Front dynamics and macroscopic diffusion in buoyant mixing in a tilted tube. Physics of Fluids, 19, 125105. Smalley, P. C., Dodd, T. A., Stockden, I. L., Raheim, A. & Mearns, E. W. 1995. Compositional heterogeneities in oilfield formation waters: identifying them, using them. In: Cubitt, J. M. & England, W. A. (eds) The Geochemistry of Reservoirs. Geological Society, London, Special Publications, 86, 59–69. Smalley, P. C., England, W. A., Muggeridge, A. H., Abacioglu, Y. & Cawley, S. 2004. Rates of reservoir fluid mixing: implications for interpretation of fluid data. In: Cubitt, J. M., England, W. A. & Larter, S. (eds) Understanding Petroleum Reservoirs: Towards an Integrated Reservoir Engineering and Geochemical

41

Approach. Geological Society, London, Special Publications, 237, 99–113. Smalley, P. C. & Hale, N. A. 1996. Early identification of reservoir compartmentalization by combining a range of conventional and novel data types. Society of Petroleum Engineers, Formation Evaluation, September 1996, 163 –169. Smalley, P. C., Ross, B., Brown, C. E., Moulds, T. P. & Smith, M. J. 2007. Reservoir Technical Limits: A Framework for Maximizing Recovery from Oil Fields. Society of Petroleum Engineers, SPE Paper 109555. Sorenson, R. P. 2005. A dynamic model for the Permian Panhandle and Hugoton fields, western Anadarko basin. American Association of Petroleum Geologists Bulletin, 89, 921 –938. Stainforth, J. G. 2004. New insights into reservoir filling and mixing. In: Cubitt, J. M., England, W. A. & Larter, S. (eds) Understanding Petroleum Reservoirs: Towards an Integrated Reservoir Engineering and Geochemical Approach. Geological Society, London, Special Publications, 237, 115– 132. Stenger, B. A., Pham, T. R., Al-Sahhaf, A. A. & Al-muhaish, A. S. 2001. Assessing the oil-water contact in Haradh Arab-D. Society of Petroleum Engineers Paper 71339, presented at the SPE Annual Technical Conference and Exhibition held in New Orleans, Louisiana, USA, 30 September– 3 October. Stølum, H. H. & Smalley, P. C. 1992. A deterministic method for assessing reservoir communication based on strontium fingerprinting. Society of Petroleum Engineers, SPE Paper 25007. Stølum, H. H., Smalley, P. C. & Hanken, N. M. 1993. Prediction of large-scale communication in the Smorbukk fields from strontium fingerprinting. In: Parker, J. R. (ed.) Petroleum Geology of Northwest Europe. Proceedings of the 4th Conference. Geological Society, London, 1421–1432. Tozer, R. S. J. & Borthwick, A. M. 2010. Variation in fluid contacts in the Azeri field, Azerbaijan: sealing faults or hydrodynamic aquifer? In: Jolley, S., Fisher, Q. J., Ainsworth, B., Vrolijk, P. J. & Delise, S. (eds) Reservoir Compartmentalization. Geological Society, London, Special Publications, 347, 103–112. Wang, R., Ravlin, T., Rosen, M. S., Mair, R. W., Cory, D. G. & Walsworth, R. L. 2005. Xenon NMR measurements of permeability and tortuosity in reservoir rocks. Magnetic Resonance Imaging, 23, 329–331. Zawisza, L. 2004. Hydrodynamic conditions of hydrocarbon accumulation exemplified by the Pomorsko and Czerwiensk oil fields in the Polish lowlands. Society of Petroleum Engineers, SPE Paper 90586. Ziegler, K., Coleman, M. L. & Howarth, R. J. 2001. Palaeohydrodynamics of fluids in the Brent Group (Oseberg Field, Norwegian North Sea) from chemical and isotopic compositions of formation waters. Applied Geochemistry, 16, 609–632.

Compartmentalization or gravity segregation? Understanding and predicting characteristics of near-critical petroleum fluids ´ EZ1*, JOHN J. LAWERENCE2 & MEI ZHANG3 RACHEL HANNAH PA 1

ExxonMobil Development Company, 12450 Greenspoint Drive, Houston, Texas 77060, USA 2

ExxonMobil Upstream Research Company, PO Box 2189, Houston, Texas 77252, USA 3

ExxonMobil Exploration Company, 23 Benmar, Houston, Texas 77060, USA *Corresponding author (e-mail: [email protected])

Abstract: Compartmentalized reservoirs can be identified based on a variety of criteria most commonly using structural geometries and stratigraphic barriers. This paper reviews several case studies from West African fields in which an additional variable, fluid chemistry, is used to identify compartmentalization at an early stage in the production history. Fluid characterization is important to constrain productivity, connectivity, facilities planning, and commercial value. Near-critical, single-phase fluids offer a case in which underestimation fluid PVT behaviour and variability can have a significant impact on all of these elements – resulting in incorrect interpretations of resource type and distribution.

Defining and characterizing compartments in petroleum-bearing reservoirs is a critical step in development planning, with significant business impact on well count and facilities design. In the literature, reservoir compartments are cast in either structural or stratigraphic terms; based on fault geometry or depositional facies. This paper addresses a third, equally important method to identify and characterize compartmentalized reservoirs – that of the fluid phase composition. The economic drivers to the prediction of the vapour phase v. the liquid phase of a petroleum reservoir (‘gas v. oil’) are large; development costs and facility design depend heavily on being able to predict the physical properties of the produced fluid, such the ratio of gas to oil (GOR) and the liquid yield of the petroleum. The definition of a fluid-compartment within a reservoir requires wellbore data, such as fluid samples, pressure data, and well bore measurements (resistivity/density/velocity logs). Potential fluidcompartments are identified by obvious changes in GOR that impact log-readings or pressure data; the simplest case being a gas cap on an oil leg in a continuous blocky sand. Increasing uncertainty in fluidcompartment identification can occur when fluids within reservoir rock are separated in the wellbore by a shale interval or when the fluid chemistry of the vapour and liquid phases become very similar. In these cases, it becomes less evident whether the two reservoir fluids are separated into two compartments and do not communicate on a production time-scale, or whether they have a connection point. To develop an appropriate development plan for fluid production, it is critical to predict

fluid-compartments away from the wellbore using secondary data. In appraisal and development drilling, one of the most relied upon tools for fluid-compartment identification is the analysis of 3D seismic amplitude v. offset (AVO) data. In conjunction with information about depth of burial and rock properties, AVO analysis can be used to identify compartments with different fluid types – vapour-phase (reservoir gas), liquid-phase (reservoir oil), water. AVO-based fluid predictions work best when there are two petroleum phases with significantly different densities – a hydrocarbon gas (C1 to C6) and heavy oil (high proportion of C7þ). However, the density differences between critical and near-critical fluids can be more subtle, and quantitative seismic analysis such as AVO is not as successful. The goal of this paper is to discuss the suite of technical tools available (and their uncertainties) to use in the identification, behaviour, and impact of compartmentalization of fluids with subtle density differences. The methodology presented for identification of fluid-compartments is applicable in many offshore West African and South American basins. There are two key assumptions in the casestudies presented in this paper. First, temperature fluctuations are minor and therefore processes such as thermal convection are not in play. Many reservoirs with large petroleum column heights (.100 ft) can have a thermal gradient of 1 –48F/ 1008F (Firoozabadi 1999). In these cases thermal diffusion and thermal convection can have a large influence on compositional changes within the


44

´ EZ ET AL. R. H. PA

petroleum. However, in the case-studies we present, there is a relatively low closure height of the petroleum traps and dramatic swings in composition occur within 3 m; these large compositional gradients over short distances are not typical of thermally-driven processes. Second, petroleum charge is on-going, and the source-rock remains in the gas and oil window. All case studies presented have a similar hydrocarbon source-type. Additional changes to fluid properties in basins in which petroleum charge has ceased are not incorporated in this discussion.

Case-study During the first phase of deepwater West African oil and gas exploration, most discovered petroleum resources were dual-phase, relatively shallow, and associated with aquifers at hydrostatic pressure conditions, less than 4000 psia. In deepwater Nigerian exploration there were only a handful of fields discovered that fell into a single-phase regime. Reservoir pressures in these fields greater than 10 000 psia, and these fluids were initially interpreted to be hydrocarbon gas of limited economic interest. In the last several years, a combination of deeper drilling and expanded fluid-property databases has resulted in a re-evaluation of perceptions regarding deepwater petroleum. In particular, the increased focus on phase-behaviour and the pressure-regime of target reservoirs has allowed for an improved prediction of the range of expected fluid properties – saturated v. undersaturated, gas condensate v. volatile oil. Data presented in this paper outlines examples from several West African fields where it was originally difficult to classify the fluid based on previously outlined definitions. Figure 1 is a phasediagram of West African discoveries relative to a generative source fluid. Over ten reservoirs contained fluids in which laboratory analysis confirmed a gas condensate clearly associated with a volatile oil, such as the example shown in Figure 2. Based on the early ‘saturated oil’ bias, this data presented a conundrum – how can two single-phase fluids co-exist in the same reservoir, and at pressure– temperature conditions that are very similar? Are these reservoirs compartmentalized?

Fig. 1. Phase diagram of West African, offshore deepwater generative fluid. Pressure– temperature points for representative discoveries plot in the dual phase, volatile oil, and gas condensate regimes. The focus of this study is outlined in the red-circle. Reservoir X, Y, and Z are examples discussed in Figures 2, 9 and 10.

Definitions and terminology The definition of fluid phase is based on the chemical composition (volume), pressure, and temperature at reservoir conditions (Cosse 1993). Figure 1 is a Pressure –Volume–Temperature (PVT) plot for a generic, generative West African fluid, with key elements labelled; the bubble point curve, dew point curve, and critical point. Fluid definitions are

Fig. 2. Reservoir pressure data, plotted in depth TVD SS plotted against pressure (psia) for Well 1, Reservoir X. Gamma-Ray (GR) log measurements plotted against depth to distinguish sand/shale interval. Within the same blocky sand interval, data support a clear relationship between two single-phase fluids; a gas-condensate and a volatile oil.

FLUID COMPARTMENTALIZATION

45

Table 1. Composition and characteristics of typical reservoir fluids (modified after Cosse 1983)

Component mole % Methane, CH4 Ethane, C2H6 Propane, C3H8 Butanes, C4H10 Pentanes, C5H12 Hexanes, C6H14 C7 þ C7 þ mol. wt. GOR, SCF/STB LGR, STB/MMSCF Stock tank oil gravity, API

Black oil

Volatile oil

Gas condensate

Gas

49.1 2.8 1.9 1.5 1.1 1.5 42.1 225 625 1600 34.3

64.2 7.5 4.7 4.1 3.0 1.4 15 180 3000 –5000 50 50.1

87.1 4.4 2.3 1.7 0.8 0.6 3.1 140 18 200 55 60.8

95.8 2.7 0.3 0.5 0.1 0.1 0.5 110 105 000 9.5 68.0

based on reservoir pressure –temperature (P–T) conditions relative to saturation pressure –temperature conditions and are unique for a given composition of the fluid. Table 1 summarizes typical properties for four generic petroleum fluid types: black oil, volatile oil, gas condensate, and gas. Single-phase, undersaturated fluids have reservoir P –T conditions that are higher than the bubble point or dew point P–T curves, and these are the focus of this paper. Dual-phase, saturated fluids have reservoir P, T conditions that are below the bubble-point and dew-point P– T curves (Fig. 3a). The intersection of these two curves is called the critical point. Single-phase fluids will migrate upward through the geological strata of a basin as an undersaturated fluid until they enter the two-phase region (Fig. 3b). At this point there will

be two saturated phases: vapour-phase and liquidphase petroleum. Additional phase transitions (and associated compositional changes) may occur while the fluids migrate, but fluids should remain saturated. There should never be a significantly undersaturated fluid below the saturation pressure of the generative (source) fluid, unless burial conditions of the host reservoir change (either through tectonic uplift or continued burial) causing changes in P –T conditions. Classification of single-phase reservoir –fluid type is based on isothermal decompression (Fig. 3c; Standing 1977) and where a fluid crosses the bubble point or the dew point curve. Upon isothermal decompression, reservoir liquids that cross the bubble point curve are called either volatile oils or black oils. Due to the uniqueness of each

Fig. 3. (a) Pressure– Volume– Temperature (PVT) plot of a generic, West African generative fluid. (b) PVT plot of a migrating, single-phase fluid into the dual-phase regime – Point 1. The single-phase begins exolving gas, and the resultant oil is heavier than the generative fluid – Point 2. Continued migration of the saturated oil to lower pressure conditions will result in more exsolved gas and heavier saturated oils. Note that the bubble-point and dew-point changes as the chemical composition of the oil changes with removal of the light components. (c) Classification of single-phase, undersaturated fluids is dependent on isothermal decompression. Circles represent starting pressure– temperature conditions, lines represent path of isothermal decompression and where fluid types would cross phase-boundaries.

46


PVT-plot based on the chemical composition of the fluid, single-phase black oils are simply defined as fluids that have significantly lower temperature than the critical point temperature. The boundary between black oil and a volatile oil cannot be as arbitrarily defined as suggested in Figure 3c. Reservoir gases that cross the dew point curve upon isothermal decompression are called gas condensates (sometimes referred to as retrograde gases). Reservoir gases that never intersect the dew-point curve are referred to as wet or dry gasses. The term ‘wet’ applies to a gas that does not release condensate in the reservoir, but would at surface conditions; ‘dry’ refers to gases that do not release condensate at surface conditions (McCain 1988). Figure 4 is phase diagram with schematic PVT curves for single-phase fluids. Single-phase reservoir fluids that cross either the dew point curve or the bubble point curve close to the critical point are referred to as near-critical or critical fluids. Near-critical fluids are a complex association of undersaturated gas-condensates and volatile oils with saturated fluids. Near-critical dual-phase petroleum fluids and single-phase fluids have relatively minor differences as compared to each other; in contrast dual-phase petroleum fluids that are well below the saturation line can have large differences in density. Typical saturated gas caps in West Africa have gradients ranging from 0.07–0.11 psia/ft and oil gradients ranging from 0.27 to 0.34 psia/ft. Undersaturated gas condensate densities range from 0.14 –0.16 psia/ft and volatile oil data points ranged from 0.18– 0.22 psia/ft. This similarity in fluid density affects AVO and rock-property modelling, PVT modelling, field-data analysis, and the ability to acquire representative samples of reservoir fluids. Due to subtle density differences in near-critical fluids, secondary sorting such as gravity segregation can occur within a reservoir to separate the fluid

Fig. 4. Schematic end-member phase curves for different fluid types. Red curves represent the dew-point, green the bubble-point curve, and CP stands for the critical point. Petroleum fluids investigated in this study range in behaviour, as outlined by the dashed lines.

compositionally based on molecular weight differences. During this process, the heavier molecular components settle to the ‘bottom’ (near hydrocarbon–water contact) of the reservoir, and the lighter molecular components rise to the ‘top.’ For near-critical fluids, dramatic gravity segregation can occur at essentially the same pressure and temperature conditions. In the case of gravity segregation, there is a spectrum of fluid compositions present in the same reservoir and the pure end-member fluids presented in Figure 4 capture only the lightest fluid near the crest of the structure and the heavier fluid near the hydrocarbon–water contact. As will be outlined in the case-studies, establishing where within this spectrum wellbore fluid samples are taken is critical to understanding the potential compartmentalization (or lack of) within a field.

Using seismic analysis to define fluid-compartments Defining fluid-compartments based on seismic amplitudes can be highly successful in dual-phase fluids where density differences result in strong impedance contrasts between oil, gas and water sands over a large porosity range. However, densitydriven impedance contrasts are not apparent in single-phase fluids, therefore quantitative seismic analysis can lead to incorrect interpretations of fluid compartmentalization. The pre-drill prediction and post-drill results for Well 1, Reservoir X exemplifies this kind of interpretation error. The petroleum fluid in Reservoir X plots roughly 200 psia above the critical point in Figure 1 (see orange circle). Figure 5 is an interval amplitude extraction of the far offset gathers from the 3D seismic data-set. As there are no strong changes in relative amplitude strength across the reservoir, the pre-drill prediction was for a single compartment of dual-phase oil. The well discovered both a gas-condensate and a volatile oil co-existing in the same 20 m sand. To develop these fluids, it was important to define the resource size of both fluid types and how they might be compartmentalized. The post-drill analysis of this reservoir was hampered by three apparently contradictory pieces of data. First, as shown in Figure 5, there is no strong variation in seismic amplitudes of these fluids. Second, two trends are clearly indicated by the pressure data (Fig. 2), suggesting that there is an oil-leg and a gas cap. Third, laboratory analysis of these fluids confirmed that the ‘gas cap’ 0.144 psia/ft fluid is actually a gas-condensate and the ‘oil leg’ is a volatile-oil with 0.22 psia/ft gradient. Using measured fluid-property data and appropriate rock properties, it was possible to complete AVO modelling to see ‘what went wrong’ with the


Fig. 5. Interval extraction of 3D Far offset data, maximum amplitude for Reservoir X (GR log-curve shown in Fig. 2), penetrated by Well 1 and Well 1 ST. Green dashed line represents the apparent ‘gas– oil’ contact suggested by pressure gradient analysis.

pre-drill prediction. As shown in Figure 6, the nature of these single-phase fluids do not lend themselves to direct detection using AVO modelling. It was possible to detect an AVO signature of hydrocarbon v. water over a range of porosities (10– 26%), but it was not possible to distinguish between a gas condensate or a volatile oil.

Using phase-behaviour modelling to define fluid-compartments Given the results of Well 1, it is important to identify other methodologies for defining fluidcompartments in near-critical fluids. First, however,

47

it is critical to define where in the stratigraphic column near-critical fluids might occur; this requires fluid-phase modelling (PVT modelling). Choosing the appropriate fluid model is one of the key uncertainties in modelling the behaviour of critical fluids. Simple black oil equations of state (EOS) do not account for the compositional variability present in a near-critical fluid, and material balance equations based on black-oil assumptions for near-critical fluids are unrealistic (Moses 1986). A multi-stage flash calculation, as described using a Peng-Robinson EOS (Tsai & Chen 1998), is more appropriate. Fevang et al. (2000) propose a systematic approach to choosing the correct fluid model for PVT modelling of critical fluids. We modelled the phase-behaviour of typical Nigerian fluids, testing for the vertical variability of single-phase fluids at different P –T conditions, using a modification of the Peng-Robinson EOS (Fig. 7). The goal of this experiment was to identify fluids that might lend themselves to incorrect interpretations of fluid-compartments, such as that outlined for Reservoir X. The composition of our model fluid is based on laboratory analysis from a field in the Western Niger Delta. The model investigates the vertical variation of fluid properties – such as gas –oil ratio (GOR) – within the petroleum column at different P–T conditions. Reservoirs 1 and 2 have P–T conditions similar in three penetrations (Fig. 7a) and have pressures significantly higher than the saturation line (.3000 and .8000 psia respectively). The P –T conditions of Reservoir 3 are the similar to those of 10 reservoirs penetrated in the Western and Central Niger delta, all with pressures slightly above saturated (Fig. 7a).

Fig. 6. AVO modelling for Reservoir X modelling the near-far seismic offset response for a range of reservoir porosities (10–25%). There is no significant difference between a gas-condensate or a volatile oil AVO signature; however, it is still possible to distinguish between these two single-phase fluids and water.

48


Fig. 7. (a) Phase diagram of West Africa data and phase envelope for the common source generative fluid. The blue square is the modelled petroleum fluid with chemistry, pressure and temperature similar to discovered West African near-critical Petroleum Fluids. The pink and red squares are the modelled petroleum fluids with the same chemistry and temperature as the type (blue square), but with higher pressures. (b) PVT model of representative West African near-critical fluid, at different pressure conditions, ranging from over 10 000 psi undersaturated to only 100 psi undersaturated. Gas –Oil-Ratios (GOR) were then modelled over a range of pressure conditions; fixing temperature and chemistry. To compare the potential GOR variation within the hydrocarbon column at three different pressure conditions, GOR is plotted relative to column mid-point (for a 2000 ft hydrocarbon column, regardless of depth, column midpoint will be at ‘0’, 21000 will represent the crest, and þ1000 will represent the hydrocarbon–water contact). The models with pressure conditions significantly above the saturation line have minimal GOR variation. The model with pressure just above the saturation line can have dramatic swings in GOR over a very small depth range. Labelling of ‘gas injector data,’ ‘discovery well data’ and ‘appraisal well data’ ties to modelled interpretations of this data in Figure 8a, b.

Plotted relative to a mid-point in a petroleum column, fluids in Reservoir 1 and 2 have only minor variability in GOR (a representative fluidproperty which can be related to density; Fig. 7b). In contrast, the fluid in Reservoir 3 has significant GOR variability – from 2000 scf/stb to well over 8000 scf/stb within a 200 ft interval. This is an

extreme case of gravity segregation, which commonly occurs in tall, saturated oil columns. As mentioned earlier, it is important to note that modelling of phase behaviour such as gravity segregation is often subject to significant uncertainty due to the data quality constraints and the limitations on our ability to adequately model different aspects of the

Fig. 8. (a, b) Schematic cross sections presenting two alternate interpretations of the PVT model presented in Figure 7b using the same three reference labels.


problem. However, this model does provide a theoretical backdrop to interpret the results of Well 1 (Fig. 2), and illustrates how models of phasebehaviour can impact interpretation of fluid data in terms of reservoir compartmentalization. On Figure 7b, three data points are labelled – gas-injector, exploration well and appraisal well. In the absence of a gravity-segregation model, the interpretation of this data – based on single, isolated data points – would suggest a highly compartmentalized reservoir with much heavier oil at the base and perhaps a gas cap, such as that shown in Figure 8a. In contrast, using a gravity-segregation model, the resultant connectivity story would be very different – a highly connected reservoir with a gas condensate gradually transitioning to a volatile oil (Fig. 8b). These two models would result in very different depletion plans, as is discussed below.

Using pressure analysis to define fluid-compartments Rapid-turn around of the complete laboratory fluid analysis required for phase-behaviour modelling is often not possible. This can result in a data-vacuum in which the interpretations based on fluid data in terms of reservoir compartmentalization are made. Initial evaluations of a resource based on expectations of relatively constant fluid properties (such as Reservoirs 1 and 2 in Fig. 7b) can be proved incorrect too late in the process to impact business decisions such as facilities design. We present two methodologies in which evaluation of reservoir data available immediately after drilling a well, in light of phase-behaviour modelling, can be used to show the critical differences in fluid properties

49

necessary for compartment identification: excesspressure analysis (Fig. 9) and well-site PVT analysis (Fig. 10). Figure 9a is a pressure– depth plot for data for Well 2, Reservoir Z. Reservoir Z is a confined channel complex in which the sands were interpreted to connect both laterally and vertically, with an average net-to-gross of around 45%. In this view, the pressure data from Reservoir Z is consistent with that model, with a vertically connected column of volatile oil with an average pressure gradient of 0.21 psia/ft, and has a pressure 50 psia above the saturation line (Fig. 1, green square). In light of the phase-behaviour modelling, a strong gravity segregation signature is expected, but is not immediately obvious based on the pressure data. An alternate way of interpreting pressure data is using an excess pressure plot (Fig. 9b), which removes the effect of buoyancy on the data and allows for detailed interpretation of vertical compartmentalization. Reservoirs that are in complete vertical communication with homogenous fluid properties will plot on a vertical line when plotted to the average pressure gradient. Plotted in excess pressure-space relative to that 0.21 psia/ft gradient, the data from Well 2 systematically deviates from vertical, verging to the right, or towards a ‘lighter’ fluid, and a few samples at the base of the oil column have a trend breaking from left to right with increasing depth, which would indicate a slightly heavier fluid. This is evidence for strong degree of gravity segregation and compositional variability in what was previously viewed as a homogenous, ‘single phase’ fluid. With just the log suite and pressure data from the well, however, it is difficult to prove gravity

Fig. 9. Two methods for plotting down-hole pressure data can provide insight into fluid compartmentalization. (a) Depth v. absolute pressure; reservoirs look to be relatively homogeneous with good vertical communication. (b) Depth v. excess pressures; reservoir fluids appear to have a strong gravity segregation trend.

50


Fig. 10. (a) Plot of absolute pressure v. depth for Well 3, Reservoir Z. Strict pressure analysis would suggest a homogenous, vertically connected fluid with a gradient of 0.146 psi/ft. Plotted saturation pressures (both dew point and bubble point) suggest a compositionally graded, single-phase fluid. (b) Plot of depth v. condensate yield for the gas condensate of Reservoir Z, Well 3. Well-site PVT data confirmed interpretation of a gravity segregated hydrocarbon column with variable condensate yield.

segregation in a connected column, as some changes in fluid properties occur at shale breaks at the well. For example, the transition from the Lower, 0.23 psia/ft fluid, to the Middle, 0.189 psia/ft, fluid occurs at a shale. However, when combined with detailed interpretation in 3D seismic datasets, the erosional behaviour of this channel complex allow for multiple communication points for the sands away from the Well 2 penetration point.

While impossible to ‘prove’, the integration of these data leads to a most-likely interpretation of a single, non-compartmentalized, column that is strongly gravity segregated. Well-site PVT analysis can also be used to assist in rapid identification of fluid-compartments. Wellsite PVT analysis is a quick-look approach, and makes several assumptions of fluid composition to make fluid-property predictions. While there is a


high degree of uncertainty in the raw data, the relative differences between the data can be used for interpretation of fluid-compartments. Figure 10 has the pressure –depth data for Well 3, which penetrates Reservoir Y, a confined channel complex interpreted to have a net-to-gross sand of greater than 50%. The P–T conditions for Reservoir Z are 100 –120 psia above the inferred critical point for a generative West African fluid (Fig. 1, yellow triangles). Reservoir sands A, B, and C contain gas condensates with what appears to be a connected column of gas condensate with a pressure gradient of 0.146 psia/ft. Reservoir sand D contains a volatile oil with a 0.255 psia/ft pressure gradient, separated by a significant shale break. Given the presence of two single-phase fluids with slightly different P– T conditions, these sands could be viewed as vertically compartmentalized, rather than connected. Four samples were collected for quick-look well-site PVT analysis, and are also plotted in Figure 10a. In a dual-phase, saturated oil with a connected gas cap, the saturation pressure equals the reservoir pressure at the gas–oil contact. However, in a gravity-segregated near-critical fluid, the transition from heavier ‘volatile’ oils and lighter ‘gas condensates’ while abrupt, the saturation pressure is always lower than the reservoir pressure. The well-site PVT saturation pressure data from Reservoir Z plots lower than the reservoir pressure in a classic pattern evidenced by other near-critical, gravity segregated fluids. For gas condensates in particular, these two contrasting interpretations of vertically compartmentalized v. vertically connected fluids has a significant impact on estimations of condensate yield and therefore resource size (Fig. 10b). Without wellsite PVT data, resource estimations of a vertically compartmentalized gas condensate would be based on empirical relationships between a pressure gradient inverted to density, in this case 20 bbls/mcf. However, well-site PVT data suggests a much richer liquid yield, which increases towards the base of the gas condensate column to a maximum of

51

78 bbls/mcf. This systematic increase supports the alternate interpretation of a connected fluid rather than a compartmentalized reservoir with distinct fluids.

Data uncertainty in near-critical fluids As shown in Figures 5, 7, 9 and 10, single samples taken from any point in the petroleum column of a near-critical fluid will not be necessarily representative of the fluid properties throughout the entire column; it is necessary to put these data within the context of PVT modelling. It is important to note that well-site sampling procedures (i.e. the length of the test), combined with the relative mobility of these fluids, can result in an underestimation of the liquid yield of gas condensates. Generally, down-hole pressure measurements use opening pressures significantly lower than reservoir pressures to facilitate fluid migration into the sample chamber. For single-phase fluids, low opening pressures results in the reservoir fluid dropping below the saturation line (which is often only a few hundred psia) and preferentially filling the sample chamber with a higher GOR fluid. Table 2 contains data from offshore deepwater West Africa, in which condensate yield estimates for the same sample are measured four ways: pressure gradient, inversion of GOR from a well-site (100 cc) single-stage flash PVT analysis, measurement from a full laboratory PVT multi-stage separator analysis, and a drill stem test. One of the key differences in these methodologies is the length of the test – either actual (20 minutes v. 3 days for MDT sample v. drill-stem test) or modelled (instantaneous for single-stage flash v. stepped for separator analysis). For all samples, the condensate yield (CGR) of the sample increased with the length of the test, in one case four-fold, increasing the potential economic viability of these fluids. These data are consistent with results reported by Siemek & Nagy (2004), who quote up to a 50% range in CGR measurements for the same reservoir based on estimation or

Table 2. Condensate yield ratios listed based method of measurement/estimation for six near-critical fluids from the Niger Delta. Sample 3 is from Reservoir Z, Well 3 CGR estimated from pressure gradient/ density (bbls/mcf)

CGR estimated from single-stage flash GOR (bbls/mcf)

CGR estimated from multi-stage seperator GOR (bbls/mcf)

CGR based on drill-stem test (bbls/mcf)

25 30 20 110 100 5

23 28 15 100 92 12

45 63 82 132 n/a 54

n/a n/a n/a 230 195 n/a

Sample 1 Sample 2 Sample 3 Sample 4 Sample 5 Sample 6 n/a, not applicable.

52


Fig. 11. Map view schematic of two fluid compartmentalization models and production implications.

sampling techniques. While the gas phase property variation reported is small due into improper well conditioning and sampling (4%), the variation in condensate yield ranged from 220% to þ70%.

Discussion Gradational changes in fluid properties such as those described in these cases can result in significant underestimation of volumetric reserves. Several authors have argued that only 25% of the reserve uncertainty is associated with petrophysical (fluid) properties (Bu & Damsleth 1996). These statistics were built with dominantly dual-phase fluid data,

and are not broken out by fluid type. In single-phase fluids, the affects of gravity segregation or the use of a non-representative fluid sample to constrain variables such as GOR or shrinkage (FVF) can result in miscalculation of the liquid yield. For example, in gas condensate/volatile oil systems, Siemek & Nagy (2004) argue that the uncertainty of FVF might rise as high as 50%. In addition, the interpretation of these nearcritical systems as connected or compartmentalized can have a significant impact on development plan (Singh et al. 2005). Figure 11 is the map-based corollary of the schematic cross-sections presented in Figure 8a, b, with pre- and post-production


interpretations. Figure 8a is a pre-production cross section of the dual phase model and Figure 8b is the pre-production cross section of a single phase model. If the reservoir is interpreted to have a gravity-segregated, near-critical fluid, it is necessary to maintain the reservoir pressure above the saturation pressure to prevent gas coming out of solution, in order to optimize liquid recovery. In a modelled development plan, early (immediate) water injector support would be required to drive condensate liquids to the producer (Fig. 11, 1 Phase Model). However, if the reservoir was interpreted to be a dual-phase system with a gas cap, it is likely that early injection support would not be in the development plan. Pressures would drop below the saturation line near the producer, gas would exsolve, potentially gassing out the production early. Up-dip, or away from the pressure sink, however, fluids would still appear as oil. This could lead to a false interpretation of structural or stratigraphic compartmentalization (Fig. 11, 2 Phase Model). Simulating the production life of these systems also presents a challenge. Fevang et al. (2000) suggest that a black-oil fluid model can be used to simulate the behaviour of critical fluids in most depletion-drive cases, but not in gas-injection cases due to incorrect and extreme differences in gas densities. Even in a depletion-drive case, their work stresses the importance of generating the black-oil model PVT-data correctly, with the knowledge that the field is truly a gravity segregated, critical fluid.

Conclusion Compartmentalized reservoirs can be variably defined, most commonly using structural geometries and stratigraphic barriers. We would recommend that an additional variable to investigate early in the life of the field is that of fluid compartmentalization. Fluid characterization is important to constraining productivity, connectivity, facilities planning and commercial value. Near-critical, single-phase fluids offer a case in which underestimation fluid PVT behaviour and variability can have a significant impact on all of these

53

elements – resulting in incorrect interpretations of resource type and distribution. This paper is the result of several years of analysis and discussion. West African generative source models and building of ‘type’ curves for source fluids builds from work completed by Paul Hicks and Scott Barboza at ExxonMobil’s Upstream Research Company. Field data and prospect analysis worked by Robert German, Mike Mann, Steve Sutton, Keith Conrad, Felipe Pontigo, Eric Wildermuth and Richard Barke. Herb Hyatt and Mike Davis provided invaluable assistance in developing appropriate resource assessments. Thank you to ExxonMobil Exploration management team for providing support for this work; Sherry Becker, Reggie Beasley, and Pamela Darwin.

References Bu, T. & Damsleth, E. 1996. Errors and uncertainties in reservoir performance predictions. SPE Formation Evaluation, 11, 194–200. Cosse, R. 1993. Basics of Reservoir Engineering: Oil and Gas Field Development Techniques. Institut Français du Petrole Publications, Editions Technip, Paris, 77–114. Fevang, Ø., Singh, K. & Whitson, C. H. 2000. Guidelines for choosing compositional and black-oil models for volatile oil and gas condensate reservoirs. Society of Petroleum Engineers, SPE Paper 63087. Firoozabadi, A. 1999. Thermodynamics of Hydrocarbon Reservoirs. McGraw-Hill Co., USA. McCain, W. D., Jr. 1988. The Properties of Petroleum Fluids. 2nd edn. Penn Well Publishing Col, Tulsa. Moses, P. L. 1986. Engineering applications of phase behaviour of crude oil and condensate systems. Journal of Petroleum Technology, 38, 715– 723. Siemek, J. & Nagy, St. 2004. Estimation of uncertainties in gas-condensate systems reserves by Monte Carlo simulation. Acta Montanistica Slovaca, 9, 289– 293. Singh, K., Fevang, Ø. & Whitson, C. H., 2005. Depletion oil recovery for systems with widely varying initial composition, Journal of Petroleum Science and Engineering, 46, 283–297. Standing, M. B. 1977. Volumetric and Phase Behaviour of Oil Field Hydrocarbon Systems, SPE. Richardson, TX 124. Tsai, J. C. & Chen, P. 1998. Application of a volumetranslated Peng-Robinson equation of state on vapour-liquid equilibrium calculations. Fluid Phase Equilibria, 145, 193–215.

Integration of time-lapse geochemistry with well logging and seismic to monitor dynamic reservoir fluid communication: Auger field case-study, deep water Gulf of Mexico E. CHUPAROVA1*, T. KRATOCHVIL2, J. KLEINGELD1, P. BILINSKI2, C. GUILLORY2, J. BIKUN1 & R. DJOJOSOEPARTO1 1

Shell International Exploration and Production, Subsurface Expertise & Deployment, 150 North Dairy Ashford, Houston, Texas, USA 2

Shell Exploration and Production Company, New Orleans, Louisiana, USA *Corresponding author (e-mail: [email protected])

Abstract: The present study illustrates the multi-disciplinary integration of time-lapse geochemistry with 4D seismic, production logging and pressure history analysis in the Auger Blue reservoirs. The integrated approach enables identification of dynamic fluid reservoir communication occurring after six years of primary production between an oil (Lower Blue O2) and a gas reservoir (Blue O Massive) indicated by pressure data as separate at static conditions, and points to cross-fault leakage with possible stratigraphic communication as the likely mechanism(s) by which dynamic fluid reservoir communication has occurred. The time-lapse geochemistry study was performed utilizing fluid samples collected during a period of over eight years of primary production. The results present evidence for gradual mixing of Lower Blue O2 oil with the condensates of the Blue O Massive reservoir. Time-lapse geochemistry results also pinpointed the timing when dynamic reservoir communication started to occur, with an initial mixing detected as early as six years after the start of production. Statistical evaluation of the geochemistry results showed the contribution of the oil to the gas condensate reservoir to significantly increase from 14– 23% to 60– 63% within a year from when mixing was initially detected. The results of the study assisted in implementing an improved field development strategy to increase production.

Detection and prediction of static reservoir compartmentalization at the appraisal stage, which reflects the presence of barriers to fluid flow established over a geological time-scale, is the first key task for optimizing reservoir development. During the last several decades, advances in reservoir geochemistry have shown that static reservoir compartmentalization can be detected based on compositional variability of the fluids (Slentz 1981; Kaufman et al. 1990; Hwang & Baskin 1994; Hwang et al. 1994; Larter & Aplin 1995). Consideration and understanding of reservoir filling history (England et al. 1987, 1995; Wilhelms & Larter 2004), compositional grading and fluid properties (Westrich et al. 1999; Wavrek et al. 2001; Stainford 2004; Smalley et al. 2004; Weissenburger & Borbas 2004; Wavrek & Mosca 2004), as well as multidisciplinary integration (Smalley & Hale 1996; England 2007) are essential to achieving higher confidence in reservoir geochemistry interpretations and predictions of static compartmentalization. Another key aspect in optimizing reservoir development is monitoring of dynamic reservoir performance over a production time-scale, after

the production has been initiated. The surveillance geochemistry application is referred to as ‘timelapse geochemistry’ or ‘4D geochemistry’, and can be applied to both oil and gas fluids (Milkov et al. 2007) as well as during primary, secondary (e.g. water flood) or tertiary (e.g. gas injection) recovery (Chouparova & Philp 1998).

What is time-lapse geochemistry? Time-lapse geochemistry is an approach utilizing geochemical techniques and methods to identify and monitor subtle compositional changes in produced fluid samples collected over a period of time that could signal and/or explain significant changes in dynamic reservoir performance and production. Production allocation monitoring, dynamic reservoir communication, depletion of compositionally graded fluid columns, and predictions of water breakthrough and water flood sweep efficiency are examples of some specific applications. The results of time-lapse geochemistry studies provide verification and/or improvement of static and


56

E. CHUPAROVA ET AL.

dynamic reservoir models as well as predictive capabilities. The predictive capability of time-lapse geochemistry comes from the fact that dynamic macroscopic phenomena (e.g. GOR changes, water breakthrough) in reservoir performance could be detected much earlier at the molecular and microscopic level of the fluids’ geochemistry. This predictive capability allows more time for potential re-adjustment of business development strategy, more effective and efficient reservoir management and ultimately higher recovered hydrocarbon volumes. The approach of time-lapse geochemistry requires availability of representative fluid samples (oil, gas, and water) collected from the same reservoir(s) over a period of time during appraisal, development and production. The appraisal or pre- (early) production samples (down hole or surface) and data are very important because they define the baseline or initial reference closest to the static condition of the fluids in the reservoir and serve for further comparisons. The optimum frequency of fluid sample collection for a particular reservoir depends on a number of factors, including, for example, type of reservoir fluid and drive, size of reservoir, and stage of production. The successful application of time-lapse geochemistry requires multi-disciplinary integration with other time-lapse techniques, and most commonly is integrated with pressure, PVT and production data, time-lapse production logging and 4D seismic, as well as static and dynamic

reservoir models. The uniqueness of time-lapse geochemistry in the whole integrated 4D toolkit is that it is the only source of information about the subtle compositional variations experienced by fluids (oil, gas, water) during production. In addition, acquiring produced fluid samples (at the surface) is much cheaper compared to the cost of 4D seismic surveys, production logging, and PVT analyses, and results in availability of sampling sets with higher frequency (e.g. several months, bi-annual, annual). A typical sequence of questions to be answered with a time-lapse geochemistry study includes: Is there areal and/or temporal variability in the produced fluid’s properties and composition by well and on a reservoir scale? What are the controlling factors and processes causing the variability? How can they be used in a predictive mode for reservoir management and surveillance purposes? Applied in an integrated way, this approach can significantly improve and assist the efficiency and effectiveness of reservoir development, surveillance and management, ultimately resulting in increased recovery of hydrocarbon volumes. In this paper, we present the Auger Blue casestudy of integration of time-lapse geochemistry with 4D seismic, pressure history and well logging illustrating dynamic reservoir communication occurring after six years of primary production between two reservoirs that are indicated by pressure data as separate at static conditions.

Fig. 1. Location map of the Auger field, deepwater Gulf of Mexico, USA.

INTEGRATION OF TIME-LAPSE GEOCHEMISTRY

Geological setting and production history of Auger Blue sands The Auger field is located in the deep water Gulf of Mexico (GOM) at 2862 ft water depth, 220 miles SW of New Orleans (Fig. 1) and has been in production since 1994. The reserves of the field are distributed in five main Miocene –Pliocene turbidite pay horizons – Pink (S), Golden Brown (R), Green (Q), Blue (O) and Yellow (N). The fluids range from volatile/black oils to condensate-rich gases trapped along a deep palaeo-ridge related to a shallow piercement salt dome and major fault systems. The reservoirs are formed in a turbiditic depositional environment and range from laterally continuous sheet sands at deeper levels (S through lower Q series) to amalgamated channel sands with overlying levee and overbank deposits at shallower levels (O and N series), Figure 2, (McGee et al. 1993). Condensed depositional zones that are most likely eustatically controlled form the caps to these sequences. The average porosity and permeability are 24% and 150 mD, and 29% and 800 –900 mD at deeper and shallower levels, respectively. The amount of faulting increases from deeper to shallower levels. The Blue O pay in the Auger field consists of five reservoirs – Lower Blue O2, Blue O Stray,

Fig. 2. Type log illustrating all pay levels in Auger field.

57

Blue O Massive, which is subdivided vertically by a shale barrier into Lower and Upper, and Blue O Laminate (Fig. 3). The laminated sand is approximately 30% net-to-gross while the massive sand is approximately 85–95% net-to-gross and represents the main producing reservoir. Average whole core measured properties for the massive sand are 29% and 781 mD stressed porosity and permeability, respectively. The hydrocarbon accumulations are situated at original pressures of over 10 000 psi and temperatures of 174–1768F (79 –808C) (Table 1). Under static conditions, based on appraisal RFT pressure and fluid gradient data, Lower Blue O2 was determined as originally oil-bearing reservoir and O Laminate, O Massive and O Stray – as gas-bearing reservoirs. During the exploration–appraisal stage, only one well had penetrated and measured formation pressures for both Blue O Massive and Lower Blue O2 sands, that is, well Garden Banks GB 426-1 (Fig. 4). The difference in original formation pressures between Blue O Massive and Lower Blue O2 reservoirs in GB 426-1 well was 48 psi. The fluid gradient in the O Massive was about 0.2 psi/ft while the O2 was about 0.27 psi/ft. These differences indicated that the reservoirs were not in pressure communication and represented separate gas and oil reservoirs prior to production. In addition, the seismic loop for the O Massive and O2 reservoirs are unique and amalgamate at no point except at the faults. The production of Blue O Massive sand started in 1994 with A14 well (Fig. 5). Wells A16ST and A19ST were added to production in 1999 and 2001, respectively. Down hole pressure gauges in the producing wells and formation tester pressures in subsequently drilled wells provided dynamic pressure information for the Blue sands postproduction. Pulsed Neutron Logs (PNC) were run post-production in several wells to evaluate sweep efficiency. PNC’s indicated a complex drainage pattern within the Blue reservoir. This observation resulted in breaking the Blue Massive into the O Massive Upper and O Massive Lower. The O Massive Upper did not appear to be draining and was subsequently perforated in the A19BP well to sweep this by-passed oil. Production from Lower Blue O2 started in 1997 at A9 well and lasted until 1999 when the reservoir watered out. The A19ST well, discussed in more detail later, is located up-dip from the single producing A9 well, and is the most up-dip penetration of Lower Blue O2 sand but has never produced this sand. During drilling of the A12ST well for deeper objectives, the MDT’s experimental (at the time) down hole Optical Fluid Analyzer (OFA) gargling technique was tested at several depths in the Blue sand. This technique is a method by which the

58

E. CHUPAROVA ET AL.

Fig. 3. Blue O pay in Auger field. (a) Seismic amplitude structure map with well locations; (b) Under static conditions, Blue O Massive is a gas bearing and Lower Blue O2 reservoir is an oil-bearing reservoir. Blue O Massive and Lower Blue O2 are separate reservoirs based on original formation pressures (see Fig. 4) and discussion in text. Under dynamic conditions, a comparison of early (1994) and late (2001) production, based on pulsed neutron well logging results, illustrates that the lower part of O Massive reservoir had been drained and watered out while the upper part of the reservoir had not (see Fig. 10).

reservoir fluid is allowed to pass through the MDT’s OFA and the type of fluid determined based on the optical properties observed (Smits et al. 1995; Felling & Morris 1998; Hashem et al. 1999; Dong et al. 2007). The results showed that the A12ST Blue Massive fluid had similar optical properties as the fluid in the A12ST Blue Stray. Those fluids were different from the heavier fluids measured at two depths in the A12ST Lower Blue O2. The observations were consistent with gas presence in O Massive and O Stray, and oil in O2. A subsequent PNC obtained in this well showed that the Blue Stray sand (which sits on water) was not drained while both the Lower Blue O2 and Blue O Massive were drained. A depth-formation pressure (measured in 1998) plot illustrates the pressure depletion in O2 and O Massive compared to O Stray sand (Fig. 6). The A13 well drilled in 1997 penetrated the Blue O Massive sand and what has been termed the O Stray sands. Down hole MDT pressures were taken throughout the Blue sands, including the O Stray (Fig. 7). The pressures in the O Massive Lower were depleted more than the O Massive Upper, and the pressures obtained in the O Stray were also depleted. In addition, a fluid sample was obtained in the O Stray, which yielded a condensate geochemically different from the O Massive producers A16ST, A14, and A19ST, which will be discussed in more detail later. The A13 well has never produced from the Blue sands.

The Blue sand is a strong water drive reservoir. Residual hydrocarbons were found down-dip of the main hydrocarbon accumulation. This residual zone has been penetrated and logged, and is seen and verified seismically. The residual zone is considered to be a result of palaeo-hydrocarbon–water contact, which changed as the trap rotated over geological time.

Samples and methods Samples Regular fluid sampling of producing wells in the Auger field has been carried out for over eight years on an annual or bi-annual basis starting in

Fig. 4. Static formation pressures in 426-1 well. Refer to Figure 5 for well location.

Table 1. PVT and basic oil properties of the reservoir and dead fluids from Blue O Massive, Lower Blue O2 and Blue O Stray sands

Date sampling Jul-94 Jun-97 Mar-97

Well A14 A13 A09

Sand O Mass Blue O Stray Blue O2 Lower Blue

Measured depth (ft) 17253 16362 16356 – 16410

Sample source

GOR (scf/bbl)

surface RFT Surface

3410 4012 1032

Tres (F)

Pres (psi)

Psat (psi)

Undersat.

176 175 174

10540 9951 10471

8000 8121 5555

2540 1830 4916

b. Dead fluids – basic oil properties Date sampling

Well #

Jul-01 Oct-01 Oct-01 Oct-01 Jun-97 May-97

A14 A14 A16ST A19ST1 A13 A-9

Sand

O mass Blue O mass Blue O mass Blue O mass Blue O Stray Blue O2 L. Blue

MD (ft)

Gravity (API)

Sulfur (wt%)

Ni (ppm)

V (ppm)

17 253 – 17 312 17 253 – 17 312 17 750 – 17 993 16 116 – 16 190 16 362 16 356 – 16 410

40.7 42.4 41.4 39.4 37.9 38.6

0.68 0.63 0.68 0.75 0.80 0.96

,0.9 n.a. n.a. n.a. 1.4 4.9

1.5 n.a. n.a. n.a. 3.1 13.1


a. Reservoir fluids – PVT properties

n.a., not analysed.

59

60

E. CHUPAROVA ET AL.

Fig. 7. Formation pressure– depth plot in A13 well. Pressure measurements were taken in 1997. The pressures in the O Massive Lower were depleted more than the O Massive Upper, and the pressures in O Stray were also depleted. Fig. 5. Seismic amplitude structure map of Blue O Massive sand with well locations. The main Blue O Massive producers are A14, A16 and A19 wells. The Lower Blue O2 sand has a single well producer, A9 well.

1994. Fluid samples were collected from producing wells of different Blue O reservoirs as follows: † Blue O Massive fluids were sampled from three wells (A14, A16ST, A19ST) for eight consecutive years from 1994 to 2001. † Lower Blue O2 fluids were collected at the A9 well as surface samples in 1997 (early production) and 1998. † Blue O Stray fluid was collected as a down hole sample at A13 well in 1997.

Methods 4D seismic. Following the acquisition of orthogonal baseline seismic surveys and the initiation of production in 1994, three 3D seismic surveys acquired

Fig. 6. Formation pressure–depth plot in A12 well illustrating the pressure depletion of O2 and O Massive compared to O Stray sand. The pressure measurements of Blue sands in A12 well were taken post-production (1998) while the well was drilling for deeper objectives. Pulsed neutron logs (PNC) obtained in this well showed that the Blue Stray sand was not drained while both O2 and O Massive sands were drained.

in 1997, 1999 and 2002 were integrated to monitor aquifer progression through the Auger reservoirs. The 2002 acquisition combined a high level of repeatability relative to previous surveys to illuminate 4D effects and technically advanced seismic acquisition and processing to image subtle stratigraphic and structural features. The aquifer influx is well defined by seismic amplitude changes in all gas and oil pay levels. A comparison between preproduction and 2002 data is discussed in this paper. The amplitude measurements used in the study are secant amplitude measurements on integrated (runsummed) data, which corresponds to the natural logarithm of the acoustical impedance, band-limited. The data was co-processed, that is, after the designature of the base and monitor survey to the same zero phase wavelet, the identical seismic processing parameters and migration velocities were applied to each (Fig. 8). A 2DSRME (2D Surface Related Multiple Elimination) was applied to reduced multiples. The acquisition parameters were nearly identical, but today a similar survey would have been shot in a more repeatable manner. Seismic acquisition repeatability is key to minimize variable overburden effects. The seismic wavelength and sand thicknesses are generally comparable for appropriate use of the integrated (runsum) seismic displays. The bandwidth of the data is roughly 8 –30 hz, 3 db/octave down, integrated polarity. The gross stratigraphic thickness of most of the objectives was 60 to 120 ft. The sands are softer than the bounding shales, whether the rock fluids are pure brines, at initial commercial hydrocarbon saturations, or at production or geological residual saturations. Commercially charged hydrocarbon sands are noticeably brighter (softer) than sands at residual saturation, which in turn are noticeably brighter than wet sands. In terms of the acoustical impedance responses for these unconsolidated rocks, the effects due to


61

complete gas chromatographic separation of twelve individual aromatic compounds, specifically the alkyl-benzenes in the C8 to C10 low boiling point oil fraction. This method allows comparison between samples even when heavily contaminated by oil-based drilling mud. The technique has excellent precision with better than 0.3% relative standard deviation and a long-term repeatability of around 1%. The long-term repeatability of the method is a major advantage for time-lapse geochemistry studies, where samples are collected at different times, and any newly collected samples can be compared with older ones without re-running the whole sets of old and new samples. The MDGC data are presented using various ratios of the individual alkyl-benzene compounds (numbered 1 through 12) and plotted on the axes of a star diagram (polar) plot. Fig. 8. The figure illustrates equating the pre-production and post-production 3D seismic surveys when the baseline was not designed for 4D purposes. (a) Baseline and monitor surveys, processing is without acquisition normalization seafloor seismic amplitude. Note the gap in coverage from Auger TLP (tension leg platform), which prevented the monitor survey from reproducing the baseline in acquisition. Today, a Baseline Streamer Survey is designed to address such obstructions. For 4D purposes, seismic surveys are equated to lowest common denominator for bandwidth, composite, and so on. Both are processed together. (b) baseline and monitor survey, processing is with acquisition normalization seafloor seismic amplitude. Only the repeatable parts of the acquisition were kept in the baseline. Cable station and wavelength shape manipulation were also done for normalization.

change in saturation are much higher than compaction and pressure changes. The exception is when the pressure change causes a change in the saturation, that is, when free gas develops in the geological residual aquifer area due to pressure drops. Several of the aquifers with known residual in the lower Pleistocene and upper Pliocene in deep water Gulf of Mexico have large magnitude increases in seismic amplitude after production starts, as the pressure is drawn down and gas is freed. It is our observation from other fields, especially when the gas is drier, that compaction effects are similar to saturation changes, but this was not the case with Auger. Geochemical fingerprinting. The geochemical fingerprinting data presented in this paper are obtained by Multi-Dimensional Gas Chromatography (MDGC) using an in-house proprietary analytical set-up, conditions, and methodology for data processing and interpretation, calibrated with large datasets. The MDGC method provides

Results and discussion 4D seismic and pulsed neutron logging The first observation that brought the question of potential inter-pay dynamic fluid communication between Lower Blue O2 oil-bearing and Blue O Massive gas condensate reservoirs came from the 4D seismic data. A comparison between the preproduction and 2002 seismic surveys shows a clear drainage of Lower Blue O2 sand (Fig. 9). It also illustrates that the volumes up-dip from the single producer well (A9) in Lower Blue O2 reservoir are drained without another up-dip drainage point in this reservoir. The single producing well (A9) for Lower Blue O2 is located down-dip from A19ST well (Fig. 9d) where pulsed neutron logging verified the drainage in Lower Blue O2 confirming the 4D seismic interpretation (Fig. 10). The drainage area up-dip of the producing well (A9) is large, but the change in structural elevation is small. It could be argued, given the small structural elevation change that some kind of cusping brought up-dip fluids into the A-9 wellbore. However, the drainage appears too complete areally to accept that just cusping has taken place. Further on, pulsed neutron logging in A19 well carried out in 1999 and 2001 revealed unexpected behaviour of the Blue O Massive reservoir. The lower O Massive part of the reservoir clearly showed drainage occurring between 1999 and 2001, while the upper O Massive part of the reservoir was not drained (Fig. 10) despite both members being perforated by up-dip producing wells (A-14 and A16ST). Detailed geological evaluation of the reservoir architecture suggested that favourable conditions exist for potential fluid communication between the Lower Blue O2 and the lower part of the Blue O Massive reservoir post-production, while leaving

62

E. CHUPAROVA ET AL.

Fig. 9. Comparison of pre-production and 2002 3D seismic results for Blue O Massive (a, b) and Lower Blue O2 (c, d) events. Note the reduction in seismic amplitudes due to production from the Blue O Massive sand (b). There is no drainage point up-dip from A9 well in O2 sand but seismic amplitudes are reduced indicating oil has been withdrawn (d). PNCs in A19 well detected drainage in Lower Blue O2 but the only O2 producer is A9 well located down-dip of A19 well (refer to Fig. 10).

the upper part of Blue O Massive reservoir isolated, that could explain the above discussed seismic and petrophysical observations. Two geological cases were identified that could promote dynamic inter-reservoir fluid communication, that is, via (a) cross-fault drainage, and/or (b) stratigraphic communication. Geological case 1: Cross-fault drainage. A seismic section through A9 well (the single well producer of Lower Blue O2 reservoir) illustrates an interpretation of several small-scale faults (Fig. 11). These small-scale faults are different from the North Auger fault indicated on Figure 5. A juxtaposition of the upper part of the Blue O Massive reservoir against shale and a juxtaposition of the Lower Blue O2 reservoir against the lower part of Blue O Massive along one of the faults are evident on the section. Fluid flow across this fault could explain the PNC observations in the up-dip A19 well where the Lower Blue O2 was drained without an up-dip producing well (Fig. 10). It can also explain the lack of drainage observed in the upper part of Blue O Massive in A19 well, considering

its juxtaposition against shale across the fault in A9 well. Analysis of the static and dynamic pressure history of the sands demonstrates that favourable conditions for fluid flow from O2 oil to O Massive gas reservoir have been created during production. The pressure differential across the fault has increased during production, as indicated by the formation pressure measurements during appraisal drilling before the start of production and static bottom hole pressure (SBHP) measurements during production (Fig. 12). The gaseous Blue O Massive sand initially was at c. 500 psi higher pressure than the oily Lower Blue O2 Sand in A9 well. As the Blue O Massive was pulled hard in production to stay ahead of water influx, a 21400 psi pressure difference between Blue O Massive and Lower Blue O2 sands across the fault was created, which introduced favourable dynamics for oil flow from Lower Blue O2 to O Massive sand across the fault. This geological situation for potential cross-fault drainage provides a consistent explanation of the 4D seismic and pulsed neutron logging observations discussed earlier: (1) drainage


63

utilizing previously collected fluid samples from these reservoirs.

Time-lapse geochemistry and fluid properties

Fig. 10. Comparison of pulsed neutron logs (PNC) in 1999 and 2001 at A19 well, the most up-dip penetration of the Lower Blue O2 oil sand (but not O2 producer). PNCs illustrate that the Upper Blue O Massive sand has not drained while the Lower Blue O Massive sand has (light blue area). The Lower Blue O2 sand has been drained (1999– 2001) without an up-dip take point (see also Fig. 5).

of the Lower Blue O2, without an up-dip from the A9 well production point, as detected by the 4D seismic and PNCs in A19 well, and (2) drainage only of the lower part of Blue O Massive sand as identified by the 4D well logging (PNC). Geological case 2: Stratigraphic communication. Another seismic section (Fig. 13) illustrates the interpreted position and extent of the Blue O Stray sand based on well control and seismic. This interpretation of the position of the Blue O Stray sand could provide an alternative conduit for stratigraphic communication of fluids between the O Stray and Blue O Massive sands. Based on the seismic and well log interpretation and integration, the main question to be answered was whether inter-pay fluid communication between the Lower Blue O2 oil (or O Stray) and the Blue O Massive condensates had actually occurred. To attempt to answer this question, the time-lapse geochemistry study was carried out

Under reservoir conditions, basic PVT properties (Table 1) illustrate that Blue O Massive fluids are undersaturated gas condensates and the Lower Blue O2 fluid is an undersaturated black to volatile oil. The undersaturation (the difference between reservoir pressure, Pres, and saturation pressure, Psat) of Blue O Massive fluids is 2540 psi. The gas –oil ratio (GOR) is 3410 scf/bbl, heptane plus (C7þ) fraction is 7 mol% of reservoir fluid and condensate gravity is higher than 408 API, all consistent with properties of rich condensate gases. During the period of sample collection discussed in this paper (1994– 2001), the reservoir pressure had not dropped below the fluid saturation pressure (8000 psi Fig. 12, Table 1) and for that reason condensate dropout in the reservoir is not expected. The undersaturation of Lower Blue O2 fluid is 4916 psi. The GOR (1032 scf/bbl), heptane plus (C7þ) fraction (18 mole %) and oil gravity (398 API) suggest black to volatile oil type of reservoir fluid. Detailed geochemical characterization of the fluids (not discussed here) supports a filling history of the reservoirs by multiple charges from a similar source rock with increasing maturity. No indications of major alteration processes have been detected in the fluids.

Lower Blue O2 and Blue O Massive fingerprinting Fluid geochemistry fingerprinting using MDGC of gasoline range aromatic compounds was performed for the available fluid samples collected from the Lower Blue O2 and Blue O Massive reservoirs between 1994 and 2001 to test the possibility of dynamic inter-reservoir fluid communication. The fingerprinting MDGC results, presented as star diagrams in Figure 14, demonstrate the following: † Blue O Massive condensates produced from A14 well did not change their composition during the first six years of production from 1994 to 1999 (the fluids are indicated by green colour lines on the star diagram). † Oils from the Lower Blue O2 (1997–1998) reservoir (light blue stars) are significantly different from the Blue O Massive (1994–1999) condensates (green stars), suggesting separate reservoirs, which is consistent with the original formation pressure data. † A subtle compositional change is detected in the Blue O Massive condensates produced in 2000 from A14 and A16 wells (black stars with red

64

E. CHUPAROVA ET AL.

Fig. 11. Seismic cross-section illustrating the geological case of potential cross-fault drainage between Blue O Massive (gas) and Lower Blue O2 (oil) reservoirs. A9 well, the only Lower Blue O2 producer, is shown on the cross section. Note the juxtaposition of the upper part of Blue O Massive against shale across the fault. This sand has not been drained between 1999– 2001 as indicated by pulsed neutron logs in A19 well (Fig. 10). Lower Blue O2 sand has been drained (see Figs 9 & 10) without an up-dip take point (see also Fig. 5). Note the juxtaposition of Lower Blue O2 sand against lower part of Blue O Massive sand across the fault that can facilitate inter-pay cross-fault fluid flow.

Fig. 12. Static and dynamic pressures in Blue O Massive and Lower Blue O2 sands measured in multiple wells through time. Static and dynamic pressures are measured via wireline tools and BHP (bottom hole pressure) gauges, respectively. The blue arrows indicate the pressure drawdown from production in the sands. The red arrows indicate a static pressure differential between Blue O Massive and Lower Blue O2 of c. þ500 psi, and a dynamic pressure differential created after seven years of production (in 2001) of c. –1400 psi, which introduces favourable dynamics for fluid flow from Lower Blue O2 to Blue O Massive sands.


65

Fig. 13. Seismic cross-section illustrating the geological case for possible stratigraphic communication between Blue O Massive and Lower Blue O2 reservoirs via O Stray sand.

Fig. 14. Time-lapse geochemistry fingerprinting results by MDGC (Multi-Dimensional Gas Chromatography) for Blue O Massive and Lower Blue O2 fluids. The axes of the star plot represent different ratios of twelve alkyl-benzene compounds (numbered 1 through 12). Ratios that contain compounds 2, 3 and 4 include xylenes. Each star represents a fingerprint formed by the values of the individual ratios for a specific oil or condensate. The plot illustrates that produced O Massive condensates were compositionally very similar during the first six years (1994–1999) of production (light green stars). The first slight compositional change in O Massive condensates is observed in 2000 (black stars with red symbols). An year and a half later (2001) the composition of produced O Massive condensates changes significantly (red stars) and behaves as a mixture between the initially produced O Massive condensates (light green stars) and Lower Blue O2 oil (blue stars).

symbols) compared to the 1994–1999 produced fluids from the A14 well (green stars). † A significant compositional change in the Blue O Massive condensates produced in 2001 (red stars), compared to fluids produced in the previous years, is detected in the fluids from A14, A16, and A19 wells. † The 2001 produced Blue O Massive condensates behave as a mixture between the Lower Blue O2 oils and the Blue O Massive initially produced condensates (1994–1999). This is indicated by the intermediate compositional parameter values (red stars, Fig. 14) of 2001 fluids relative to the Lower Blue O2 oils (light blue stars) and Blue O Massive condensates (1994–1999; green stars). † The 2001 produced Blue O Massive condensates from the three wells are similar to each other supporting good reservoir continuity. The time-lapse geochemistry results present strong evidence for gradual mixing of fluid similar to the Lower Blue O2 oil with the gas condensates of Blue O Massive reservoir and support the hypothesis of dynamic inter-reservoir fluid communication occurring during production between the oil and gas condensate reservoirs that were separate at static conditions. Moreover, the results pinpoint the timing when the dynamic reservoir communication would have started to occur, with an initial mixing detected as early as February 2000 and becoming more pronounced eighteen months later (10/2001).

66

E. CHUPAROVA ET AL.

Since time-lapse geochemistry uses molecular parameters and could detect very subtle compositional changes, an interesting question to clarify is how significant the contribution of Lower Blue O2 oil to Blue O Massive gas production in 2000 and 2001 would have been. To address this question, Lower Blue O2 oil collected in 1997 was used as one end member, and the condensates produced by A14 well during 1994– 1999, when no significant compositional changes in MDGC fingerprinting parameters were observed (Fig. 14), was used as a second end member fluid. For the latter, an average of the MDGC parameters for A14 produced fluids (1994–1999), was utilized. The calculations were performed using Shell proprietary statistical technology. The results showed the contribution of the Lower Blue O2 oil to the Blue O Massive condensate reservoir could have been in the 14–23% range (A16-A14 wells) in 2/2000 and increased three times to 60–63% (A14, A16ST and A19ST wells) in 10/2001. Examination of oil, gas, and water production data and producing gas–oil ratios (GORs), based on monthly well test data for the period 2000– 2001, does not show any obvious anomalies for A16 and A19 Blue O Massive producers. An exception is A14 well where the oil and gas have increased with c. 4000 bbls and 7.7 MMcf, respectively, from 2/2000 to 10/2001. However, several acid jobs were performed on the well in this period of time. Therefore, it becomes uncertain if the increase in oil and gas volumes could be attributed solely to the discussed inter-pay communication or it is an effect of the acid jobs performed on the well. A comparison of the timing of mixing indicated by time-lapse geochemistry with the pressure history of the sands in Figure 12 illustrates that at the time the first mixing (2/2000) between the Blue O Massive condensate and Lower Blue O2 oil was detected by time-lapse geochemistry (estimated O2 contribution of 14– 23%), the pressure difference between O Massive and O2 was in order of þ500 to 700 psi. A year and a half later (10/2001), when the mixing was detected to have become more pronounced by time-lapse geochemistry results (estimated O2 contribution of 60 –63%), the pressure difference between O Massive and Lower Blue O2 was exceeding c. 21400 psi (Fig. 12). During this period, well A19ST was brought on production (7/2001), and probably contributed to the increase in the pressure drawdown in Blue O Massive reservoir and increase in pressure difference between O Massive and Lower Blue O2 sands. An increase in pressure difference between O Massive and O2 sands can create favourable conditions for cross-fault fluid flow from Lower Blue O2 to Blue O Massive

at their juxtaposition sites, as discussed earlier. The pressure difference increase and its timing is consistent with the timing of oil– gas mixing between Lower Blue O2 and Blue O Massive detected by time-lapse geochemistry. The consistency of these two independent lines of evidence provides strong support for cross-fault fluid flow as the mechanism by which the dynamic interreservoir fluid communication between Lower Blue O2 oil and Blue O Massive gas condensates had occurred.

O Stray and Blue O Massive fingerprinting As discussed earlier, there is geological reasoning to support a possible stratigraphic communication between O Stray and Blue O Massive sands. A down hole sample from the O Stray sand (A13 well) was collected in 1997. Under reservoir conditions, the fluid is an undersaturated gas condensate with GOR of 4012 scf/bbl. The gravity of the dead liquid (condensate) from this fluid is 388 API, relatively low for a typical condensate. Formation pressures showed the reservoir was characterized by depleted pressure (9951 psi) when first penetrated by A13 well (Fig. 7, Table 1a, b), suggesting this sand has been in pressure communication with producing sands from the field. To further investigate the possibility of O Stray fluids contributing to Blue O Massive production and/or facilitating stratigraphic fluid communication between Lower Blue O2 and Blue O Massive reservoirs, a fluid from the O Stray sand collected at the A13 well in 1997 was included in the geochemistry study. The fingerprinting MDGC results compare O Stray fluid with Lower Blue O2 oil and Blue O Massive (2000) condensates, and are presented in Figure 15. The results illustrate that O Stray fluid is compositionally different from both the Lower Blue O2 oil and the initially produced Blue O Massive condensates (1994–2000). However, it shows closest compositional similarity to the mixed Blue O Massive fluids produced in 2001. The differences between O Stray and 2001 Blue O Massive fluids relate mainly to MDGC ratios that contain xylenes (ratios 1/1 þ 3, 2/2 þ 3, 4/4 þ 5), and could be influenced by any artificial xylene-containing addition in the well. A13 O Stray stock tank fluid has 2.7 wt% contamination from the oil-based drilling additive Syn-Teq, which does not contain xylene. Artificial introduction of other additives containing xylene in the well cannot be ruled out completely even though no production records were found stating addition of xylene. According to the MDGC ratios that do not contain xylene, A13 O Stray and 2001 produced Blue O Massive fluids have identical parameter values (see the red and dark blue stars in Fig. 15).


Fig. 15. Time-lapse geochemistry fingerprinting results by MDGC (Multi-Dimensional Gas Chromatography) for Blue O Stray, Blue O Massive and Lower Blue O2 fluids. The axes of the star plot represent different ratios of twelve alkyl-benzene compounds (numbered 1 through 12). Ratios that contain compounds 2, 3 and 4 include xylenes. Each star represents a fingerprint formed by the values of the individual ratios for a specific oil or condensate.

Despite the lack of evidence for xylene addition in the well, if the possibility of xylene contamination in the well were considered, then the ratios containing xylene (ratios 1/1 þ 3, 2/2 þ 3, 4/4 þ 5) would be an artificial overprint on the star plot of Figure 15. The remaining parameters would then suggest a strong fingerprinting compositional similarity between the A13 O Stray and 2001-produced Blue O Massive fluids. In this case, A13 O Stray fluid behaves as a mixture, with intermediate values of the fingerprinting parameters

67

on the axis of the star plot, between Lower Blue O2 oil and initially produced Blue O Massive (1994– 2000) condensates. This compositional behaviour could be a result of reservoir filling history, and would represent a geological fluid mix established in the reservoir prior to production, or indicate mixing of O2 and O Massive fluids through O Stray stratigraphic communication. In this case, a contribution of O Stray fluid to Blue O Massive 2001 production (A14, A16, A19ST wells) would be possible, and could explain the fingerprinting results as well as the depleted reservoir pressure in the A13 O Stray sand. The pressure depletion in A13 O Stray was detected in 1997, which would suggest that the mixing between O Stray, O2 and O Massive fluids had likely already started at that time. Based on time-lapse geochemistry fingerprinting results (Fig. 14), however, no compositional changes of Blue O Massive fluids are detected prior to 2000 (1994–1999). Figure 16 delineates an area of possible stratigraphic communication between the O Stray and Blue O Massive reservoirs based on well and seismic interpretations. The delineated area is approximately 1000 2000 ft (c. 46 acres), and does not extend areally to the location of Blue O Massive well producers A14, A19ST and A16. The estimated O Stray total aerial extent is in the range of 150– 300 acres. Considering these aerial estimates, the thickness (c. 5–15 ft), porosity (26 –28%) and water saturation (40–50%) of O Stray sand, the volumes that O Stray could have contributed to Blue O Massive production may have been sufficient to be detected in the Blue O Massive production at A14, A16

Fig. 16. Delineation of a possible area of Blue O Stray and Blue O Massive stratigraphic communication.

68

E. CHUPAROVA ET AL.

and A19 wells. However, it is difficult to explain with this scenario the preferential drainage of the lower part of the Blue O Massive reservoir, compared to the upper part of the Blue O Massive sand, as well as the Lower Blue O2 drainage in A19 well without an up-dip drainage point based on the PNC evidence (Fig. 10), as discussed earlier. If there were no xylene contamination in the A13 well sample, then the fingerprinting MDGC results would suggest that the O Stray fluid is different from all Blue O Massive and Lower Blue O2 fluids, implying a separate reservoir with no indication for stratigraphic fluid communication (Fig. 15). Even if stratigraphic fluid communication between O Stray and Blue O Massive is assumed, the original O Stray fluid composition may not be the one of the fluid sampled in A13 well, because of the known pressure depletion in the reservoir, which further increases the uncertainty with this scenario. Based on the above discussion, we conclude that stratigraphic fluid communication between the Blue O Massive and O Stray sands is in general possible but unlikely because it cannot satisfy all the observations and results. In contrast, the scenario of inter-reservoir fluid communication via cross-fault leakage provides a satisfactory explanation of all the observations based on seismic, production logging, static and dynamic pressure reservoir analysis and geochemistry, and for that reason is accepted as the preferred scenario. A combined mechanism of cross-fault leakage and stratigraphic fluid communication cannot be ruled out. The possibilities of condensate dropout or oil rim and/or compositional grading in the Blue O Massive gas condensate reservoir were also considered, which could be alternative explanations for the geochemistry fingerprinting results and dynamic reservoir communication interpretations. The reasons for not favouring any of these alternative explanations are discussed as follows. First, depletion of a compositionally graded column would result in a much smoother compositional change than the observed results (Fig. 14) but it could be consistent with a condensate dropout or oil rim presence. Second, a condensate dropout in the reservoir would require the reservoir pressure to decrease below the dew point (or saturation pressure), which for Blue O Massive fluids (A14) is at 8000 psi (Table 1). As illustrated in Figure 12 by the static bottom hole pressures (SBHP), the reservoir pressure during the production period discussed in this paper (until 10/2001) has not dropped below 8000 psi, and therefore condensate dropout in the Blue O Massive reservoir is not expected. Third, residual gas down-dip has been detected by both seismic and petrophysical evaluations during the pre-production appraisal drilling. The decrease in

pressure during production resulted in the release of the bound residual gas in the aquifer area. The free, expanded gas in turn caused the 4D seismic amplitude to bloom. This 4D observation is common for gas residuals in Pleistocene and Upper Pliocene aquifers through the greater Auger area. An oil rim cannot cause this phenomenon to occur, at least not locally. Finally, a compositionally graded column or oil rim in Blue O Massive cannot explain by itself the observed preferential drainage of the lower part of Blue O Massive reservoir, compared to the upper part of the Blue O Massive sand as well as the drainage of Lower Blue O2 in A19 well without an up-dip take point.

Conclusions The study presented in this paper illustrates application of multi-disciplinary integration of timelapse geochemistry with 4D seismic, production logging and dynamic pressure reservoir analysis. The integrated approach enables detection of dynamic reservoir communication occurring after six years of primary production between an oil and a gas reservoir that were separate at static conditions, and points to cross-fault leakage with possible stratigraphic communication as the likely mechanism by which dynamic reservoir communication had occurred. Time-lapse geochemistry results present strong evidence for mixing of Lower Blue O2 oil with the condensates of the Blue O Massive reservoir, thus supporting the hypothesis of dynamic inter-reservoir fluid communication. The results also pinpoint the timing when the dynamic reservoir communication would have started to occur, with an initial mixing detected as early as 2000 and becoming more pronounced a year later. Dynamic pressure data analysis suggests a pressure differential increase between the gas and oil sands with timing consistent with the geochemistry conclusions. Statistical evaluation of the geochemistry results suggests the contribution of Lower Blue O2 oil to Blue O Massive gas condensate reservoir to significantly increase from 14 –23% in 2000 (pressure difference between O Massive and O2 at that time was c. þ500 psi) to 60– 63% in 2001 (pressure difference between O Massive and O2 at that time was exceeding 21400 psi). The results of the study were integrated in building updated static and dynamic reservoir models and contributed to improvement of the field development strategy and hydrocarbon recovery efficiency. The present study illustrates the importance of early implementation of a fluid sampling programme and sustaining it long-term as part of field surveillance. The long-term commitment to


sustainable regular fluid sampling programmes in producing fields is sometimes difficult to achieve. The case presented in this paper provides an example to support this long-term commitment to fluid sampling and illustrates that significant fluid compositional changes may not occur for over half a decade of production. However, the low cost investment in long-term regular fluid sampling can bring a high rate of return when production and surveillance problems start to occur in mature fields by reducing the uncertainties in reservoir performance, achieving more effective and efficient reservoir management and ultimately increased recovered hydrocarbon volumes. We would like to thank several generations of geoscientists who have worked on Auger Basin, including Jim Booth, Dave McGee, Tim Beattie, Vu Cung, King Sim Lee, and especially the 2001–2002 Auger team who recognized and acknowledged the value of time-lapse geochemistry to surveillance. Special recognitions go to Alan Kornacki for initiating the fluid sampling programme at Auger field and Mohamed Hashem for collecting high quality appraisal down hole fluid samples even from the stray sand discussed in this study. Robert Elsinger’s constructive technical suggestions and useful discussions have brought additional insights and are highly appreciated. Al Killi, Ray Lesoon and Linda Peacock are acknowledged for the high quality production sampling and analytical support. We thank Shell EP America and Shell International EP for permission to publish this paper as well as for the long-term organizational commitment in fluid sample collection and multi-disciplinary integration, with special thanks to Dean Malouta, Val Brock, Iain Percival, Brad Kerr, and Min-Teong Lim who have been supporting the establishment and development of time-lapse geochemistry studies in the Americas and on a global scale. We thank an anonymous reviewer, Chris Clayton and Quentin Fisher for the useful comments and suggestions to improve the manuscript.

References Chouparova, E. & Philp, P. 1998. Geochemical monitoring of waxes and asphaltenes in oils produced during the transition from primary to secondary water flood recovery. Organic Geochemistry, 29, 449–461. Dong, C., O’Keefe, M. et al. 2008. New down hole fluid analyzer tool for improved reservoir characterization. Society of Petroleum Engineers, Reservoir Evaluation & Engineering, 11, 1107–1116. England, W. A. 2007. Reservoir geochemistry – a reservoir engineering perspective. Journal of Petroleum Science and Engineering, 58, 344– 354. England, W. A., MacKenzie, A. S., Mann, D. M. & Quigley, T. M. 1987. The movement and entrapment of petroleum fluids in the subsurface. Journal of the Geological Society, London, 144, 327– 347. England, W. A., Muggeridge, A. H., Clifford, P. J. & Tang, Z. 1995. Modeling density-driven mixing rates in petroleum reservoirs on geological time scales,

69

with application to the detection of barriers in the Forties Field (UKCS). In: Cubitt, J. M. & England, W. A. (eds) The Geochemistry of Reservoirs. Geological Society, London, Special Publications, 86, 185–201. Felling, M. M. & Morris, C. W. 1998. Characterization of in-situ fluid responses by use of optical fluid analysis. Society of Petroleum Engineers, Reservoir Evaluation & Engineering, 1, 297–302. Hashem, M. N., Thomas, E. C., McNeil, R. I. & Mullins, O. C. 1999. Determination of producible hydrocarbon type and oil quality in wells drilled with synthetic oil-based muds. Society of Petroleum Engineers, Reservoir Evaluation & Engineering, 2, 125– 133. Hwang, R. J., Ahmed, A. S. & Moldowan, J. M. 1994. Oil compositional variation and reservoir continuity: unity field, Sudan. Organic Geochemistry, 21, 171– 188. Hwang, R. J. & Baskin, D. K. 1994. Reservoir connectivity and oil homogeneity in a large-scale reservoir. Middle East Petroleum Geoscience Geo94, 2, 529– 541. Kaufman, R. L., Ahmed, A. S. & Elsinger, R. J. 1990. Gas chromatography a development and production tool for fingerprinting oils from individual reservoirs: applications in the Gulf of Mexico. In: Schumacher, D. & Perkins, B. F. (eds) Proceedings of the 9th Annual Research Conference. Gulf Coast Section of the Society of Economic Paleontologists and Mineralogists Foundation, 263– 282. Larter, S. R. & Aplin, A. C. 1995. Reservoir geochemistry: Methods, Applications, and Opportunities. In: Cubitt, J. M. & England, W. A. (eds) The Geochemistry of Reservoirs. Geological Society, London, Special Publications, 86, 5 –32. McGee, D. T., Bilinski, P. W., Gary, P. S., Pfeifer, D. S. & Scheiman, J. L. 1993. Geological models and reservoir geometries of Auger field, deepwater Gulf of Mexico. In: Weimer, P., Bouma, A. H. & Perkins, B. F. (eds) Submarine fan and turbidite systems – sequence stratigraphy, reservoir architecture, and production characteristics. Gulf Coast Association of Geological Societies 15th Annual Research Conference, 233– 244. Milkov, A. V., Goebel, E., Dzou, L., Fisher, D. A., Kutch, A., McCaslin, N. & Bergman, D. F. 2007. Compartmentalization and time-lapse geochemical reservoir surveillance of the horn mountain oil field, deep-water Gulf of Mexico. American Association Petroleum Geologists Bulletin, 91, 847– 876. Slentz, L. W. 1981. Geochemistry of reservoir fluids as a unique approach to optimum reservoir management. Society of Petroleum Engineers, SPE Paper 9582. Smalley, C., England, W. A., Muggeridge, A., Abacioglu, Y. & Cawley, S. 2004. Rates of reservoir fluid mixing: implications for interpretation of fluid data. In: Cubitt, J. M., England, W. A. & Larter, S. (eds) Understanding Petroleum Reservoirs: Towards an Integrated Reservoir Engineering and Geochemical Approach. Geological Society, London, Special Publications, 237, 99– 114. Smalley, C. & Hale, N. A. 1996. Early identification of reservoir compartmentalization by combining a range of conventional and novel data types. Society

70

E. CHUPAROVA ET AL.

of Petroleum Engineers, Formation Evaluation, 11, 163– 169. Smits, A. R., Fincher, D. V., Nishida, K., Mullins, O. C., Schroeder, R. J. & Yamate, T. 1995. In-situ optical fluid analysis as an aid to wireline formation sampling. Society of Petroleum Engineers, Formation Evaluation, 10, 91–95. Stainford, J. G. 2004. New insights into reservoir filling and mixing processes. In: Cubitt, J. M., England, W. A. & Larter, S. (eds) Understanding Petroleum Reservoirs: Towards an Integrated Reservoir Engineering and Geochemical Approach. Geological Society, London, Special Publications, 237, 115– 132. Wavrek, D. A. & Mosca, F. 2004. Compositional grading in the oil column: advances from a mass balance and quantitative molecular analysis. In: Cubitt, J. M., England, W. A. & Larter, S. (eds) Understanding Petroleum Reservoirs: Towards an Integrated Reservoir Engineering and Geochemical Approach. Geological Society, London, Special Publications, 237, 207– 220. Wavrek, D. A., Mosca, F., Chaouche, A. & Jarvie, D. M. 2001. Geochemical insights to engineering

problems: When does oil heterogeneity really indicate a compartment? Abstracts 20th International Meeting on Organic Geochemistry, Nancy, France. Weissenburger, K. S. & Borbas, T. 2004. Fluid properties, phase and compartmentalization: Magnolia Field case-study, Deepwater Gulf of Mexico, USA. In: Cubitt, J. M., England, W. A. & Larter, S. (eds) Understanding Petroleum Reservoirs: Towards an Integrated Reservoir Engineering and Geochemical Approach. Geological Society, London, Special Publications, 237, 231– 256. Westrich, J. T., Fuex, A. M., O’Neal, P. M. & Halpern, H. I. 1999. Evaluating reservoir architecture in the Northern Gulf of Mexico with oil and gas chemistry. Society of Petroleum Engineers, Reservoir Evaluation & Engineering, 199, 514– 519. Wilhelms, A. & Larter, S. 2004. Shaken but not always stirred. Impact of petroleum charge mixing on reservoir geochemistry. In: Cubitt, J. M., England, W. A. & Larter, S. (eds) Understanding Petroleum Reservoirs: Towards an Integrated Reservoir Engineering and Geochemical Approach. Geological Society, London, Special Publications, 237, 27–36.

Compartmentalization of the Nelson field, Central North Sea: evidence from produced water chemistry analysis C. E. GILL1*, M. SHEPHERD1 & J. J. MILLINGTON1,2 1

Shell UK Limited, 1 Altens Farm Road, Nigg, Aberdeen, AB12 3FY, UK

2

Nexen Petroleum U.K. Ltd., Charter Place, Vine Street, Uxbridge, Middlesex, UB8 1JG, UK *Corresponding author (e-mail: [email protected]) Abstract: Drainage cells are localized reservoir volumes that are bounded both laterally and vertically by permeability barriers. The subdivision of a reservoir volume into drainage cells provides a framework that allows a mature producing field to be screened for remaining oil volumes. Nine drainage cells have been defined in the Nelson field. The lateral edges of these drainage cells are stratigraphic in nature and correspond to the boundaries between individual macroforms, for instance, between channel complexes and interchannel sediments. A very large dataset of produced water chemical analyses has been used to help define the extent of the drainage cells. Provinciality, shown by areal variations in produced water compositions, is consistent with the inferred location of the cells. The Nelson field shows variation in the chloride ion concentration of produced water both vertically and laterally. Vertical variation can be detected by changes in produced water chemistry after water shut-off events at shale horizons which are thought to be laterally extensive within the reservoir. Lateral variation corresponds to patchwork areas that are consistent with individual macroforms such as channel complexes. An additional technique has been used to confirm the location and extent of drainage cells within the field. This involves the compilation of drainage charts, a quantitative volumetric method that involves comparing theoretical and actual oil –water contact changes within particular field areas.

The Nelson field is located in blocks 22/11, 22/6a and 22/12a in the UK Central North Sea (Kunka et al. 2003). Nelson is a simple dip closed structure and is one of a series of Paleocene Forties Sandstone Member oil accumulations on the Forties Montrose High (Fig. 1). Oil is produced from two main reservoir intervals. The upper interval is the T75 genetic sequence separated from the underlying T70 genetic sequence by the T75 maximum flooding surface (Fig. 2). The uppermost interval of the T75 reservoir unit is a field wide mappable unit termed the Last Chance Sand; so called because in the more mature parts of the field, this is the last sandstone interval left producing significant volumes of oil. A laterally extensive mudstone occurs at the base of the Last Chance Sand and has been called the Last Chance Shale. The principal areas of oil production are two NW–SE trending sandstone filled channels termed hereafter the Western and Eastern Channels (Fig. 3a). The T70 interval is dominated by a major central channel complex (Fig. 3b). Seawater was injected into the major channel axis in the water leg through four injection wells between 1998 and 2005. The field is also believed to have a strong aquifer through connectivity with the regionally extensive Forties Fan system.

The Nelson field is at the early mature stage of the producing life-cycle. A total of 37 production wells have been drilled since field start-up in 1994 and more wells are planned. It is not thought that the existing well stock will be sufficient to optimally recover the remaining oil volumes. The moderate to high degree of stratigraphic complexity in the channellized turbidite reservoir may be responsible for trapping a number of stranded oil volumes which are potentially large enough to justify drilling as infill wells. In 2006 a project was started with the objective of locating the remaining oil in the Nelson field. The intention was to use data integration techniques to find areas within the reservoir that may contain stranded oil volumes. Data integration is a method for co-analysing geology and production data (Bryant & Livera 1991). By combining production data with the geological interpretation, common patterns may be observed that show how the reservoir architecture is influencing the production pathways within the reservoir. The method mainly involves graphical overlays of two data types, production data and the geological interpretation. Common techniques include for instance, the overlay of cumulative production bubble plots onto fault maps or lithofacies maps (e.g. Hamilton et al. 1998).


72

C. E. GILL ET AL.

Fig. 1. Paleocene oil accumulations on the Forties-Montrose High, from Kunka et al. (2003).

Data integration methods have been used by geologists to screen oil fields for remaining oil potential since at least the 1960s. The techniques became better known after a series of papers were written in the 1990s by both staff from Shell in the UK and the Texas based Bureau of Economic Geology in the USA (e.g. Holtz & Hamilton 1998). In Shell, data integration techniques were used mainly to locate the remaining oil volumes within the Brent Group fields in the Northern North Sea. Here, the reservoirs show a layercake geometry, comprising stacked hydraulic units. The term hydraulic unit is used in the sense that these are specific reservoir intervals that show internal communication but are bounded at the top and base by extensive permeability barriers to vertical flow, typically shales (Haldorsen & Lake 1984). Hydraulic units combine with an orthogonal

intersecting fault system to create relatively simple box-shaped reservoir cells in the Brent Group fields. These have been called ‘shoeboxes’ by Shell staff and given the more generic name of drainage cells by Shepherd (2009). Drainage cells act as self-contained tanks, restricted laterally by permeability barriers to horizontal flow, principally sealing faults in the Brent province, and with permeability barriers to vertical flow, typically shale barriers, bounding the top and base of the cells. Whereas reservoir compartments are commonly defined as areally bounded, the term compartment has been used variously in the technical literature as a volume that is either restricted laterally; or restricted vertically; or both. The definition of a drainage cell in this paper is similar to the description of compartments by Larue & Hovadik (2006) as ‘non-connected parts of the reservoir’. The term

COMPARTMENTALIZATION OF THE NELSON FIELD

Fig. 2. Type log for the Nelson field.

‘drainage cell’ has been used in this paper by preference as this avoids the ambiguities involved in the use of the word ‘compartment’. There is also a pragmatic/operational element to the definition, in that a

73

drainage cell can in practice be slightly leaky, but if the leak is small by comparison to well drainage rates, then this will not be of great importance to a reservoir management team seeking to optimize recovery from a field. The concept of the drainage cell is useful in framing a reservoir for the purposes of locating stranded oil. The simple geometry of a Brent type drainage cell allows the reservoir volumes to be worked out with reasonable accuracy. For each drainage cell, the oil-in-place is calculated and the total amount of oil produced by each well accessing the drainage cell is summed up. From this, the remaining oil volumes for each drainage cell are derived by difference and tabulated. By this means, individual drainage cells can be screened for remaining oil, and those with significant remaining mobile oil volumes can be considered as worth investigating for potential infill well targets to recover this oil. It is not possible to apply this approach to the Nelson field in the same way it has been used for the fields in the Brent province. Although there are a small number of faults cutting the oil leg in Nelson, they do not have a significant influence on the reservoir performance. Nevertheless, difference in oil production rates, water cuts and the level of producing oil –water contacts indicate a patchwork variation suggestive of compartmentalization. This

Fig. 3. (a) Map showing macroforms and drainage cells for the T75 reservoir interval of the Nelson field. Yellow areas represent channel axes, green areas represent channel margins, grey areas represent the interchannel areas. (b) Map showing macroforms and drainage cells for the T70 reservoir interval of the Nelson field. Yellow areas represent channel axes, green areas represent channel margins, grey areas represent the interchannel areas.

74

C. E. GILL ET AL.

variation was considered likely to be controlled to a significant degree by the heterogeneous nature of the channellized turbidite depositional environment. A different approach to framing the reservoir for remaining oil volumes was required compared to previous practice in the Brent province. Two major assumptions were made, with the objective of creating the basic framework required to carry out data integration and to start a detailed analysis of the field performance. The first assumption was that the maximum flooding shales separating the genetic sequences serve to act as flow barriers. In particular, the T75 maximum flooding surface was assumed to act as a barrier between the upper T75 and lower T70 sequences. The second assumption was that the boundaries between individual macroforms within individual sequences are flow barriers. Macroforms are sediment bodies that are found at the macroscopic level of heterogeneity, that is, at a scale of hundreds of metres (Alpay 1972; Jackson 1975). Similar terms used in geological publications include genetic units and geobodies. In the Nelson Field reservoir, the major macroforms are turbidite channel complexes, channel margin sediments and interchannel sediments. Once data integration analysis started, the early indications were that these initial assumptions appeared to be valid. The T75 maximum flooding surface acts as a vertical permeability barrier. The Last Chance Shale in the upper part of the T75 sequence is a very extensive permeability baffle but not quite a total barrier to flow on the basis of production and pressure data. Flow behaviour character is also comprehensible at the level of individual macroforms. Nevertheless, a problem remained. There was a wish to frame drainage cells within the Nelson field with a high enough level of confidence such that they could be screened for remaining oil volumes. Whereas drainage cells in faulted Brent Group fields are easy entities to define as self contained volumes, the definition of drainage cells within a stratigraphically complex reservoir appeared to be a much more difficult task. Various techniques were therefore sought to validate the extent and volume of the individual drainage cells within the Nelson field. The main method used is the subject of this paper and this involved a detailed analysis of field-wide variation in produced water chemistry. An additional technique was also found to be useful in helping to define drainage cells, the drainage chart method; and this is also described.

Assumptions The objective of analysing produced water chemistry variation within the Nelson Field was to

determine whether there was any provinciality to water compositions corresponding to the inferred location of drainage cells within the reservoir. It is much more common to find oil geochemistry variation being used as a technique for inferring compartmentalization in oil fields. One of the earliest papers to describe the use of fluid geochemistry in reservoir management gave a case history on the use of produced water geochemistry in a Gulf Coast oil field in the US (Slentz 1981). The technique was used to fingerprint the reservoir interval sourcing water in a producing oil field with multiple pay zones. A number of assumptions have been made as to how produced water has been sourced in the Nelson field, based on a survey of existing technical papers, some of which are over 50 years old. The basic assumption is that the produced formation water that enters a Nelson production well is a sample of the connate water within the oil leg in the near wellbore area. Connate water is water that was originally present in the pores during the formation of a rock, although the water composition may subsequently have changed over geological time. Connate water is present both in the water leg and oil leg of the reservoir interval, the latter present as the water phase adhering to grain surfaces as a result of capillary forces. These forces have been sufficient to resist displacement of the connate water by oil during the filling phase of the Nelson field. This irreducible water phase constitutes about 20–25% of the total fluid volume above the oil –water contact. Although the irreducible water phase is immobile under static conditions, core studies have indicated that the connate water in the oil leg can be mobilized by injected water during a waterflood (Brown 1957; Kelley & Caudle 1966; Korsbech et al. 2006). The connate water is thought to form a mobile zone that separates injection water from the continuous oil volume. Although reservoir pressures in the Nelson field are supported by four wells injecting seawater, much of the water produced in the wells is connate water. Only three wells show evidence for seawater breakthrough. Produced seawater can be recognized, as it is distinctive in its chemistry compared to connate water; it has a lower salinity, is rich in sulphate and is deficient in barium.

Produced water Newly drilled Nelson field production wells typically show a year or more of dry oil production before water starts to break through. Water rates are initially small compared to the total well flow rate but are observed to gradually increase with time. Once the water cut (the percentage of the


water production rate to the total flow rate) reaches a high level in a well, production logs will be run to determine the source of the water influx. A rise in the oil–water contact as a result of production will typically leave the lowermost production perforations producing all or much of the water. These zones are then shut-off by setting a bridge plug opposite a suitable shale above the swept interval. The combination of a bridge plug inside the wellbore and a laterally extensive shale outside the casing provides a major baffle which acts to hold back the water in the underlying water-swept interval of reservoir. This type of operation has often proved successful in eliminating or reducing water production to a production well for a year or more. Nevertheless, continuing oil production will eventually draw in water up and around the edges of the shale baffle and water production will restart once the water enters the perforations above the shale. Once this water production becomes excessive, another bridge plug will be set providing there is a suitable shale interval in the reservoir to provide containment. The net effect of these operations is for the perforated interval in a production well to decrease in length with time as more and more water producing intervals are isolated lower in the well. It has proved necessary to closely monitor the chemistry of the produced water from production wells. A significant problem in North Sea wells is the precipitation of barium sulphate scale in the production tubing and surface manifolds as a result of the mixing of sulphate rich injected seawater with barium rich and sulphate deficient formation water. The scale can form in such large quantities, that the tubing and surface manifolds can be severely restricted or even blocked by the scale. Small concentrations of radioactive material within the barium sulphate scale can also cause a

75

safety hazard in offshore facilities. If the produced water analysis starts to indicate the early onset of seawater breakthrough to a production well, scale inhibitor treatment will be squeezed to the well to inhibit scale formation. The object of sampling produced water in the Nelson field has been for the specific operational requirements as described above. Nevertheless, this practice also provides a high-density dataset that can be used by the production geologist to help understand how the reservoir geology controls fluid movement within the reservoir.

Data Data sources The primary database used in this study comprises biweekly water samples taken from each production well on the Nelson Field throughout the well life. The samples are taken at the wellhead and analysed onshore for major element chemistry concentrations. This study uses principally chloride ion concentration data from these analyses as an indication of water salinity.

Data limitations Overall, it should be noted that the water samples are prone to contamination (Warren & Smalley 1993). The early (pre 2001) water chemistry data exhibits poor resolution and in general a characteristic water chemistry signature is difficult to identify at low water cuts. This is the case early in the field life, early in infill well history and after successful water shut-offs. Figure 4 shows the scatter associated with a typical Nelson well when the chloride ion concentration is analysed early in a wells lifetime.

Fig. 4. Chloride, barium and sulphate ion concentration over the lifetime of N12. Note scatter in early well life and smear of data at scale squeeze events (08/10/02 & 30/04/03). Note also the plateau in chloride ion concentration that was reached 5/6 years after the onset of production.

76

C. E. GILL ET AL.

In addition, the produced water chemistry is highly influenced by scale squeeze events, at which time multiple, closely spaced produced water samples are taken to monitor the concentration of chemicals introduced to the well. The chemicals contaminate the water and results in a large scatter of spurious data at multiple concentrations. These events must be removed prior to background signature analysis (Fig. 4). The data also show gaps when production in either an individual well or from the entire platform is shut in; for example when an individual well is waiting for workover. The gaps are a hindrance for data analysis but can also cause problems if an individual well shows cross-flow between two differently pressured reservoir units during a prolonged period of shut-in. This may result in a given well not demonstrating a true water chemistry fingerprint for several weeks after production has resumed. A significant limitation to this study is that it is not always possible to determine the exact depths of the reservoir interval that sourced the water within a perforated interval. The length of the perforated section in a well can vary between a short interval targeting a specific zone to an interval where every oil bearing zone is open to flow over a length of a hundred feet or more. The water may be sourced from anywhere within the perforated interval, maybe even from more than one reservoir zone, yet without production log information, it is not possible to know precisely where the water has been produced from in a well. Given this problem, the depth interval assigned to a set of samples can in practice correspond to a large depth interval of open perforations. The produced water sample may potentially be a mixture of water from all the reservoir sections open to the well. As more reservoir sections are shut-off, the length of open perforated interval decreases. In this instance, it is easier to assign a produced water composition to a more narrow range within the stratigraphic section. These sections tend to be towards the top of the reservoir section.

some sense as to the vertical variation in water compositions. The overall range of chloride ion concentrations of produced connate water from the Nelson field is circa 54 000–60 000 milligrams per litre (mg/l). The intent of this study has been to define provincial variation in compositions within this range. This has involved screening the data to identify specific water chemistry signatures corresponding to particular areas of the field and specific producing intervals. Most Nelson wells show a characteristic early scatter of chloride ion concentrations which over time becomes more consistent in composition. Assuming that no well interventions are carried out to modify the producing reservoir interval, the produced water signature will remain at a stable composition unless seawater breaks through to the well from the water injection wells. The early scatter is as a result of low initial water cuts and as water cut increases chloride ion concentration tends to increase until a stable plateau is reached. This occurs anything up to 5/6 years after the onset of production of an individual well; an example is shown in Figure 4. Periods of stability in produced water chloride ion concentration are punctuated by the setting of bridge plugs, which isolate water-producing perforations in the oil producers. Where these mechanical plugs are successful, the water cut will be reduced or eliminated. Subsequent water production will show a decrease in chloride ion concentrations. For example, well 22/11-N4 shows a stable chloride ion concentration during mid 2002 when the entire T75 section was open to flow. A successful water shut-off was carried out in the well in December 2002, leaving only the uppermost reservoir section (the Last Chance Sand) open to production. Following this water shut-off, the chloride ion concentration decreased and was stable at a lower value from 2003 until end 2005 (Fig. 5). Periods of stability such as these have been used to provide a chloride ion concentration value for each combination of open perforations in individual Nelson wells.

Results

Variable water composition laterally and vertically in the field

The high density sampling of produced water composition from each individual well throughout the lifetime of the field means that it is possible to plot the variation in produced water compositions between wells across the field area and in individual wells with time. The ability to monitor variations in water major element chemistry with time can be considered as a time-lapse geochemistry method akin to similar methods that use oil geochemistry data (Milkov et al. 2007). In addition, the changing intervals open to production in the wells gives

The areal and vertical variation in chloride concentrations within the reservoir is summarized by the mean values in Table 1 and the map in Figure 6. A second data set is also available that gives information on the water composition of the aquifer and in the oil–water transition zone. Samples were taken from well tests in two appraisal wells drilled before production start-up to provide a water resistivity (Rw) estimate for well log interpretation (Table 2). 22/6a-9 is an appraisal well


77

Fig. 5. Chloride ion concentration over the lifetime of N4. The N4 production well is located in the centre of the Eastern Channel area of the field. A bridge plug was set opposite the Last Chance Shale to isolate water producing perforations from the lower part of the T75 sequence. This left only the Last Chance Sand open to production. The chloride levels dropped from about 59 000 to 55 000 mg/l as a result of this operation (red arrow). An earlier (less successful) water shut-off event only managed to transiently reduce the chloride ion concentration (blue circle).

located in the Eastern Channel. A water sample was taken in the aquifer from a FMT log. A drill stem test (DST) also sampled water from the transition zone just above the oil–water contact. The mean chloride ion concentration is 56 975 mg/l. An appraisal well, 22/11-8Y, located in the Western Channel has a mean chloride ion concentration of 60 819 mg/l.

Analysis of results There is a systematic decrease in the produced water chloride ion concentration upwards through the reservoir. The uppermost unit in the reservoir, the Last Chance Sand, produces water with the lowest chloride ion concentration. These low-salinity water samples from the Last Chance Sand are also typically sampled from producing intervals at a significant height above the oil–water contact. The deepest samples (taken from the water leg in appraisal wells and from the lowermost perforated intervals in the oil leg) have the highest chloride

ion concentration. For instance, the Western Channel shows an aquifer composition of about 60 800 mg/l chlorides, whereas produced connate water from the Last Chance Sand in the Western Channel area shows a mean chloride concentration of 57 949 mg/l. Because of the depth resolution problem, there is not enough information to establish if there is a gradient of decreasing chloride ion concentration up through the reservoir or if field wide shales separate zones of formation water with distinctive chloride ion concentrations. Lateral compartmentalization is shown by a correspondence between a specific produced water composition from wells and the macroform in which they are located (Fig. 7). Laterally the Eastern Channel of the Nelson Field produces water with a lower chloride concentration than the Western channel at all reservoir levels. The Western Interchannel region is intermediate to both flanking channel bodies. A statistical test applying analysis

Table 1. Mean chloride concentrations in produced connate water by field area and by producing reservoir interval Field region Eastern Channel Eastern Channel N26 area South/T70S T70S Southern tip Western Channel Western Channel Western Channel Margin/T70 NW Western Channel Margin T70 NE Western Interchannel/T70 NW Western Interchannel Abbreviation: LCS, Last Chance Sand

Reservoir interval

Chloride (mg/l)

LCS T75 T75 T70 and T75 T70 T75 LCS T75 T70 and T75 T75 T70 T70 and T75 T75

54 444 57 039 61 046 59 431 61 292 57 798 57 949 59 662 59 648 60 227 62 065 58 896 56 001

78

C. E. GILL ET AL.

Fig. 6. Map showing chloride ion concentration for designated drainage cells of the Nelson field. Yellow areas represent channel axes, green areas represent channel margins, grey areas represent the interchannel areas. LCS: Last Chance Sand.

of variance (ANOVA) to the observed trends supports the interpretation that there are discrete populations of formation water chemistry present within the Nelson Field. ANOVA is closely related to the t test and is sometimes called an F test. It tests the difference between the mean of two or more

groups and produces the F statistic which is the ratio of between group variation to the within group variation. The table below shows the results of ANOVA on the data presented above (Table 3). The results are interpreted by evaluating the F ratio. If the F ratio is larger than the F critical

Table 2. Chlorine ion concentrations from appraisal well test samples taken in the aquifer and transition zone Well

Source Date Depth Chlorine ion concentration

22/6a-9 (Eastern Channel)



22/11-8Y (Western Channel)

22/11-8Y (Western Channel)

FMT 4/5/88 7806 ft MD BRT 57 030 mg/l

DST 19/5/88 7526–7596 ft MD BRT 57 060 mg/l

DST 19/5/88 7526 – 7596 ft MD BRT 56 835 mg/l

DST 29/6/88 9065 – 9097 ft MD BRT 60 790 mg/l

DST 29/6/88 9065 – 9097 ft MD BRT 60 848 mg/l


79

Fig. 7. West– east schematic cross section through the northern part of the Nelson field showing the variation in chloride concentration (in mg/l chloride) according to areal and stratigraphic location.

value, F crit, there is a statistically significant difference. If it is smaller than the F crit value, the score differences are best explained by chance. It can therefore be seen that the observed trends are in general supported by statistical analysis and that there are a number of different populations of formation water chemistry present within the Nelson field. There are two sets of produced water chloride ion concentrations which do not show

statistically different populations. The first exception is the Western Interchannel where a mixture of T70 and T75 water cannot be said to be different to T75 alone. The second is the T70 and T75 mixes in the different flow units across the field. These results are as expected as a mixture of two populations (T70 and T75 formation water) is unlikely to give a discrete signature that can be recognized as different to a single end member composition

Table 3. Analysis of variance results showing which populations of chloride ion concentrations are statistically different both vertically and laterally in the Nelson Field Region

Unit 1

Vertical variation within a hydraulic unit WCM T70 & T75 WCM WIC T70 & T75 WIC EC LCS EC WC LCS WC LCS LCS EC Reservoir Unit

Hydraulic units analysed

Unit 2

T75 WCM T75 WIC T75 EC T75 WC LCS WC

F

P-value

F crit

F . F crit?

4.6015 0.0012 98.6963 54.6615 7.5826

0.0335 0.9726 0.0000 0.0000 0.0075

3.9038 3.8624 3.8765 3.8772 3.9798

yes no yes yes yes

F

P-value

F crit

F . F crit?

0.3572

3.0118

no

0.0000

2.1091

yes

0.0075

3.9798

yes

Lateral variation within a reservoir interval but between hydraulic units T70 & T75 T70 & T75 SC 1.0315 T70 & T75 WCM T70 & T75 WIC T75 T75 EC 76.8534 T75 N26 T75 WC T75 WCM T75 WCS T75 WIC LCS LCS EC 7.5826 LCS WC

Abbreviations: WCM, Western Channel Margin; WIC, Western Interchannel; LCS, Last Chance Sand; EC, Eastern Channel; WC, Western Channel; SC, South Central.

80

C. E. GILL ET AL.

in the absence of the second end member concentration (T70).

Magnitude of compositional changes in produced water at water shut-off events A key reservoir management strategy in the Nelson field is to isolate water producing perforations by setting bridge plugs in the wellbore opposite shales in the reservoir. The anticipation is that the more extensive the shale, then the longer it should take for the water from the isolated zone to flow around of the edges of the shale and enter the perforations above the bridge plug. This idea has been used as the basis for making a qualitative estimate of the lateral extent of shales in producing reservoirs. Water shut-off performance was used by Hamilton et al. (1998) to evaluate the status of permeability barriers and baffles within the Jackson field in Australia. The result of each operation was recorded as a water shut-off table containing both the reduction in water cut and the length of time the operation was successful in achieving the reduction in water cut. This table gives a qualitative idea of the extent of individual flow barriers and baffles within the reservoir. By extension of this argument, it can be conjectured that the more extensive the shale, then the greater the chances are that the shale will separate reservoir intervals with different connate water compositions. Thus, analysis of the changes in chloride ion concentration observed before and after water shut-off events allows a comparison to be made between how much water is shut-off (water shut-off performance) and the magnitude of change in chloride ion concentration. In general, chloride ion concentration decreases at water shut-off events as the more saline lower waters are eliminated from the production stream. A greater reduction in water cut at a water shut-off event results in a greater decrease in the chloride ion concentration of the produced water (Fig. 8).

Fig. 8. Relationship between chloride ion concentration change and water cut change at water shut-off events.

Some wells show a distinctive pattern whereby water production following a water shut-off event results in lower salinity water composition (i.e. lower chloride concentrations), but with a trend towards higher salinity values with increasing water-cut and time. The salinity will increase towards values that were seen in the well before the bridge plug was set. An example of this is shown in Figure 9 where a water shut-off event is shown, and in which the recovery of the chloride ion concentration follows the gradual increase in water cut. In the same way, water shut-off events which are less successful will see both chloride ion concentration and water cut rise back to pre-water shut-off levels much more rapidly (Fig. 10). This pattern may indicate a mixing trend between connate water sourced from above the shale and with time an increasing component of more saline connate water from beneath the shale. Salinity v. water-cut plots are shown in Figure 11.

Discussion Reservoir compartmentalization The following observations are made: (1) Analysis of produced water chemistry samples shows that there is a large and systematic variation in the formation water composition within the Nelson field (the range is circa 54 000–60 800 mg/l chloride concentrations). (2) Water samples from the water leg in the appraisal wells show salinities which are at the higher end of the salinities by comparison to produced water samples from the oil wells (average: 60 819 mg/l chlorides in the Western Channel and an average of 56 975 mg/l chlorides in the Eastern Channel). (3) The lowest salinity water compositions are produced in wells accessing the stratigraphically (and generally structurally) highest interval of the reservoir, the Last Chance Sand (average: 57 949 mg/l chlorides in the Western Channel and an average of 54 444 mg/l chloride in the Eastern Channel). (4) The data indicates vertical variation in water composition with the highest water salinities seen in the water leg and the lowest salinities in the upper reservoir units. However, the end-member compositions shown by vertical salinity variation are seen to change across the field (e.g. from 60 819 to 57 949 mg/l chlorides in the Western Channel and from 56 975 to 54 444 mg/l chlorides in the Eastern Channel).


81

Fig. 9. Top: Well N7 Chloride concentration throughout the lifetime of the field. Bottom: N7 Oil rate, Water rate, and water cut throughout the lifetime of the field. Black bar represents water shut-off on 29/08/05. Note drop in chloride ion concentration and water cut at the water shut-off event in 2005 and the subsequent rise in both values over the following year.

It is considered an important observation that most of the produced water compositions from perforated intervals in the oil leg are significantly less saline than those observed in samples from the aquifer. This appears to be consistent with the old experiments from core studies that indicate that the irreducible connate water fraction in the oil leg can be mobilized during a waterflood (Brown 1957; Kelley & Caudle 1966). It is proposed that for each individual region of the Nelson field that there are two end member water compositions. The lowest salinity compositions are produced from the stratigraphically highest (and generally structurally highest) parts of the reservoir. The most saline compositions are sourced from well tests in the water leg and produced water samples from oil leg connate water in the stratigraphically lower intervals. There is not enough data to state that a linear compositional gradient is present with increasing height between the high and low salinity water volumes, although some of the time-lapse plots can be interpreted to show that this locally exists. There may be step changes in water salinity across major shales

within the reservoir and this is one explanation for the decrease in water salinities when bridge plugs are successful in isolating water production for a period of time. However, it is also feasible that there is a simple compositional gradient in water salinity with height that has developed over the tens of millions of years since the start of oil filling, yet when the connate water is mobilized on a production time-scale, the shales act as baffles to flow. Thus a high water cut well after several years of water production will be drawing in water from the more down-dip regions where the water is more saline, whereas once this volume has been isolated, the first produced water volumes are from the volume of connate water in the immediate vicinity of the remaining and structurally higher perforated intervals. Both explanations are consistent with laterally extensive shales tending to compartmentalize regions of differing water salinities. Detailed chloride ion analysis indicates a level of compartmentalization that delineates three reservoir units vertically and which laterally partitions the field into several distinct units, which correspond

82

C. E. GILL ET AL.

Fig. 10. Top: Well N20 chloride ion concentration throughout the lifetime of the field. Bottom: N20 Oil rate, water rate and water cut throughout the lifetime of the field. Black bars represent water shut-offs on 29/08/05 and 25/12/04. Green circle shows rapid recovery of chloride ion concentration and water cut after the water shut-off event in 200. The pink circle shows slow (ongoing) recovery.

to depositionally controlled macroforms. The variations in chloride ion concentration suggest that there is a level of compartmentalization both laterally and vertically within the Nelson field. These discrete formation water chemistry signatures as characterized by chloride ion concentration variations appear to correspond to drainage cells. The compartments indicated by the variation in chloride ion chemistry are the major sedimentological macroforms identified as controlling fluid flow in the Nelson Field. These sedimentological bodies are defined by the deepwater gravity flows that deposited them. The major macroforms in the Nelson Field reservoir are turbidite channel complexes, interchannel sediments and channel margin sediments. It is the discontinuity between the channel sediments (both channel axis and channel margin deposits) and the interchannel sediments that compartmentalizes the formation water chemistry of the field. Channel axis sands and their marginal equivalents are in hydraulic continuity with each other. The axial sands are genetically linked to the thinner marginal sands and are believed to have received a simultaneous oil charge. By

contrast, the interchannel deposits are not in hydraulic continuity with the channel sands and as such act as discrete compartments where fluid chemistries (both oil and water) have not equilibrated on a geological timescale. This lack of continuity is due to the primary depositional architecture. The interchannel regions sit on palaeo-structural highs and are generally separated from the axial regions by low permeability sediment. Deposition in the interchannel areas is largely unrelated to deposition in the channel axis and as such a direct hydraulic link is not present and the field is sedimentologically compartmentalized as a result. The suggested explanation for the water salinity variations observed in the Nelson are similar to those observed in an unnamed field in the North Sea where a lower salinity connate water within the oil leg overlies a more saline aquifer. It was suggested that oil migration into the reservoir immobilized the connate water fraction as the reservoir filled. The more dilute pore-waters in the oil zone may represent ‘fossilized’ palaeo-pore water with similar compositions to the connate water present in the sediment at the time of oil filling (Coleman 1992).


83

Fig. 11. From top down: (a) Western Channel chloride ion concentration. Note that the T75 has a higher chloride ion concentration than the Last Chance Sand at all water cuts. (b) Western Interchannel chloride ion concentration. The T70 and T75 intervals have higher chloride ion concentrations than the T75 at all water cuts. (c) Eastern Channel chloride ion concentration. Note that the T75 has a higher chloride concentration than the Last Chance Sand at all water cuts.

In the Nelson field, the lateral variation in chloride ion concentration may reflect a temporal difference in the oil charging history of the field. For example, the western and eastern channels are separated by the Forties-Montrose high and may have

been charged at different times and from different source areas. The oil geochemical fingerprint for samples from the Western and Eastern Channels shows them to be distinct from each other. Although the geochemical fingerprints suggest maturation

84

C. E. GILL ET AL.

from the same evolving source, the timing of charge may have varied between different parts of the Nelson reservoir thus producing a less or more evolved oil type.

Drainage charts Produced water chemistry has proved successful in helping to validate the location of drainage cells within the Nelson field. In addition, a second method was used, which backed up the results from produced water analysis.

A drainage cell shows hydraulic communication throughout. Therefore if a known volume of oil is produced from a drainage cell, then the resulting rise in the oil –water contact should be predictable if the oil-in-place is known for the drainage cell. The idea is analogous to a fuel gauge measuring the remaining petrol in a petrol tank. This principle forms the basis for the drainage chart method (Shepherd 2009). A drainage chart is constructed in a number of steps. The first step is to make an oil volume v. height graph for the drainage cell. The height is

Fig. 12. Drainage chart for the Western Channel, T75 interval. The chart compares the estimated oil– water contact rise due to production with actual well data from production log data.


Fig. 13. Location of wells in the Western Channel, T75 interval.

85

defined as the height above the initial oil– water contact (assuming it to be a flat surface). The 3D geological model for the Nelson field was interrogated to get this information. The oil volume used is the mobile oil volume rather than the oil-inplace. The mobile oil volume is the oil-in-place minus the residual oil volume. The volume –height graph will therefore show where the producing oil –water contact should ideally be located after a specific volume of oil has been produced from the drainage cell. The trend line defines the theoretical drainage path for the cell as the field depletes with time. The next step is to work out the cumulative oil production from the wells in the drainage cell at the end of each year of production. Note that if any of the wells penetrate more than one drainage cell it will be necessary to estimate the flow allocation for the drainage cell in question. The end of year cumulative volumes should be spotted on the drainage path at the appropriate point. The drainage path will therefore be calibrated according to year-by-year production.

Fig. 14. Fuel tank display for the T75 interval of the Nelson Field. A full fuel tank is equivalent to the total moveable oil in the drainage cell. The unrecovered mobile oil is the volume of oil estimated to be left behind at the end of field life after subtracting both the cumulative production to-date and remaining reserves.

86

C. E. GILL ET AL.

The final step is to mark on the drainage plot, any actual producing oil–water contact available for the wells in the drainage cell. This may come directly from production logs or from open hole logs in infill wells. The values are located on the graph at the appropriate height of the oil –water contact rise and in a vertical line with the appropriate date on the drainage path. Data where the oil– water contact has reached the upper bounding surface of the cell in a particular well should be considered as a minimum value for the oil –water contact rise. The end result will be a drainage chart showing both the theoretical and actual drainage paths. If the drainage cell has been defined more or less correctly, then the theoretical and actual drainage paths should lie close to each other. The actual oil –water contact data should also ideally fall on or close to a common drainage path (Fig. 12). It is unlikely that the theoretical and actual drainage paths will coincide exactly. Indeed, it will be fortuitous if they do so. For the two paths to coincide a number of ideal conditions would need to be satisfied. These are: (1) The volume of the drainage cell should be correct. (2) The drainage cell should be sealed with no leakage. (3) The volume of residual oil saturation subtracted from the oil-in-place needs to be the correct value. (4) The macroscopic sweep efficiency within the cell should be 100%. (5) The oil–water contact rise should be perfectly even and unimpeded by any internal baffles within the cell. An example of a drainage cell in the Nelson field is shown in Figure 13. This has been made for the Western Channel cell. The proximity of the theoretical and actual drainage paths suggests that the volumetric definition of the Western Channel is not too far out.

to the total mobile oil volume of the drainage cell. The level of the tank is reduced as a result of production with the gauge indicating the amount of mobile oil estimated to be remaining at the end of field life (Unrecovered Mobile Oil) unless extra well activity is planned to access this volume (Shepherd 2009).

Conclusions The Nelson Field shows variation in chloride ion concentration of connate water compositions both vertically and laterally. The overall range of chloride ion concentration is circa 54 000 to 61 000 mg/l. Vertical variation can be detected by changes in produced water chemistry after water shut-off events. Lateral variation can be detected by variation on a fieldwide scale. Produced water chemistry is independent of water cut. Provincial variation in formation water composition can be used as indicators of reservoir compartmentalization and to distinguish laterally restricted and laterally extensive shales, both in the oil and water leg (Warren & Smalley 1993). Such step changes can be observed in the Nelson Field. The drainage chart method has proved beneficial in validating the location and size of drainage cells as indicated by produced water chemistry. Having framed the reservoir geology in terms of drainage cells, subsequent screening has allowed significant volumes of remaining oil to be localized within five of the nine drainage cells defined within the Nelson field. The authors thank the Nelson Field partners Shell Exploration & Production, Esso Exploration and Production UK Ltd, Total E & P UK plc, Petro Summit Investment UK Ltd and Svenska Petroleum Exploration U.K. Ltd for publishing permission. We also acknowledge the input of all who have worked on Nelson during the course of this study; and Andy Aplin, Alton Brown and Quentin Fisher for their helpful reviews of the manuscript.

Screening drainage cells in the Nelson field for remaining oil

References

Detailed data integration on the Nelson field has shown that it is possible to define drainage cells even in fields with complex reservoir geometries. Thus volumetric assessments can be carried out to determine which drainage cells show the largest volumes of remaining oil. Of the nine drainage cells defined in the Nelson field, five show significant remaining oil volumes that are worth investigating further for infill well locations. This is illustrated by the fuel tank display shown in Figure 14. A full fuel tank on this map corresponds

Alpay, O. A. 1972. A practical approach to defining reservoir heterogeneity. Journal of Petroleum Technology, 24, 841– 848. Brown, W. O. 1957. The mobility of connate water during a water flood. Petroleum Transactions, 210, 190– 195. Bryant, I. D. & Livera, S. E. 1991. Identification of unswept oil volumes in a mature field by using integrated data analysis: Ness Formation, Brent Field, UK North Sea. In: Spencer, A. M. (ed.) Generation, Accumulation and Production of Europe’s Hydrocarbons. Special Publication of the European Association of Petroleum Geoscientists, SpringerVerlag, Berlin, 1, 75–88.

COMPARTMENTALIZATION OF THE NELSON FIELD Coleman, M. 1992. Water composition variation within one formation. In: Kharaka, Y. & Maest, A. S. (eds) Water–Rock Interaction – Proceedings of the 7th International Symposium on Water– Rock Interaction – WRI-7/Park City Utah/USA/I3– 18 July 1992. A. A. Balkema, Rotterdam, 1109–1112. Haldorsen, H. H. & Lake, L. W. 1984. A new approach to shale management in field-scale models. Society of Petroleum Engineers, SPE Paper 10976. Hamilton, D. S., Holtz, M. H., Ryles, P., Lonergan, T. & Hillyer, M. 1998. Approaches to identifying reservoir heterogeneity and reserve growth opportunities in a continental-scale bed-load fluvial system: Hutton Sandstone, Jackson Field, Australia. American Association of Petroleum Geologists Bulletin, 82, 2192–2219. Holtz, M. H. & Hamilton, D. S. 1998. Reservoir characterization methodology to identify reserve growth potential. Society of Petroleum Engineers, SPE Paper 39867. Jackson, R. G. 1975. Hierarchical attributes and a unifying model of bed forms composed of cohesionless material and produced by shearing flow. Geological Society of America Bulletin, 86, 1523– 1533. Kelley, D. L. & Caudle, B. H. 1966. The effect of connate water on the efficiency of high-viscosity waterfloods. Journal of Petroleum Technology, 18, 1481–1486. Korsbech, U., Aage, H. K., Andersen, B. L., Hedegaard, K. & Springer, N. 2006. Measuring and

87

modelling the displacement of connate water in chalk core plugs during water injection. SPE Reservoir Evaluation and Engineering, 9, 259– 265. Kunka, J. M., Williams, G. et al. 2003. The Nelson Field, Blocks 22/11, 22/6a, 22/7, 22/12a, UK North Sea. In: Gluyas, J. G. & Hichens, M. H. (eds) United Kingdom Oil and Gas Fields, Commemorative Millennium Volume. Geological Society, London, Memoir, 20, 617– 646. Larue, D. K. & Hovadik, J. 2006. Connectivity of channelized reservoirs: a modelling approach. Petroleum Geoscience, 12, 291–308. Milkov, A. V., Goebel, E., Dzou, L., Fisher, D. A., Kutch, A., McCaslin, N. & Bergman, D. F. 2007. Compartmentalization and time-lapse geochemical reservoir surveillance of the horn mountain oil field, deep-water Gulf of Mexico. American Association of Petroleum Geologists Bulletin, 91, 847– 876. Shepherd, M. 2009. Oil Field Production Geology. AAPG Memoir 91. American Association of Petroleum Geologists, Tulsa. Slentz, L. W. 1981. Geochemistry of reservoir fluids as a unique approach to optimum reservoir management. Society of Petroleum Engineers, SPE Paper 9582. Warren, E. A. & Smalley, P. C. 1993. The chemical composition of North Sea formation waters: a review of their heterogeneity and potential applications. In: Parker, J. R. (ed.) Petroleum Geology of Northwest Europe. Proceedings of the 4th Conference. Geological Society, London, 1347–1353.

Turbidite reservoir compartmentalization and well targeting with 4D seismic and production data: Schiehallion Field, UK M. GAINSKI1*, A. G. MACGREGOR1, P. J. FREEMAN1,2 & H. F. NIEUWLAND1,3 1

BP Exploration Operating Company Ltd, 1 Wellheads Avenue, Dyce, Aberdeen, AB21 7PB, UK 2

Nexen Petroleum UK Ltd, Charter Place, Vine Street, Uxbridge, UB8 1JG, UK

3

BP Azerbaijan, Villa Petrolea, 2 Neftchilar Prospekti, AZ1003 Baku, Azerbaijan *Corresponding author (e-mail: [email protected]) Abstract: This paper discusses integration of production surveillance techniques, focusing on the use of 4D seismic data to identify reservoir compartmentalization. We present two examples of recently drilled compartments that were successfully identified following integration of surveillance data with detailed reservoir modelling work. Our examples are from the faulted, Paleocene channelized turbidite reservoirs of the Schiehallion oil field, offshore West of Shetlands, U.K. The first example provides a good case history of a 4D Integrated Reservoir Modelling (4D IRM) approach – which involved integration of dynamic well data with 4D seismic, and iterative revision of geological and reservoir simulation models. Two newly identified targets were drilled and completed successfully using these techniques. The second example illustrates a situation where 4D seismic interpretation was key in identifying a new infill target. Production in the Schiehallion Field started in 1998, and the current development scheme totals 46 wells (22 producers and 24 water injectors). During the early years of production it became apparent that geological connectivity, fluid flow and pressure communication between wells was not as inter-connected as expected. As a result the number of wells required to maximize the recovery has more than doubled to that specified in the original development plan, and the number is expected to increase further as the field matures. Continuous collection of bottom hole pressure data from permanently installed gauges, well testing and production logging (PLT) supported by a regular time-lapse (4D) seismic programme are used to update conceptual thinking and thus constrain geological and flow simulation modelling. This data integration results in improved understanding of the static and dynamic reservoir compartmentalization and well connectivity.

The Schiehallion Field is located in the West of Shetlands area of the North Atlantic (Fig. 1). The field infrastructure consists of four sub sea drill centres with 22 oil producers and 24 water injectors connected through a system of subsea wellheads, manifolds and pipelines to the Floating Production Storage and Offloading vessel (FPSO). Crude is exported via shuttle tankers taking crude to Sullom Voe terminal on the Shetland Islands. The field schematic layout (Fig. 2) and pertinent data (Table 1) are shown below. Production of oil and associated gas comes from relatively thin, Paleocene turbidite channel and sheet sands (slope/proximal basin floor setting) ranging in thickness from 5 to 30 m (Chapin et al. 2000). During Palaeogene times the Faroe– Shetland Basin was a major depocentre, and the Flett sub-basin in which the Schiehallion oilfield is located. Accommodation space was controlled by thermal sag and faulting that remained active throughout the Paleocene. A thermal plume under the Scottish Highlands and West Shetland Platform at that time was generating volcanic activity and

uplift that, in turn, created a coarse clastic supply to the deepwater basins west of Britain. The bulk of the clastic input occurred during the Lower Paleocene T30 sequence, which is approximately equivalent to the Andrew Member of the Lista Group in the North Sea (Freeman et al. 2008). The T30 sequence comprises siliciclastic turbidites sands transported from the shelf in seismically resolvable channels to the basin floor. Within the Schiehallion field the main oil bearing sands occur in the T25, T28, T31, T34 and T35 sequences. The focus of this paper is on the T31 reservoirs. The Schiehallion trap comprises a stratigraphic pinch-out in the east and dip closure along the northern and western margins to an oil –water contact (OWC) of 2064 m TVDSS. Reservoir sands are sealed on the up-dip southern edge by east –west normal faults that completely offset the reservoirs against basinal mudstone lithologies. The combination of south to north trending channel sand-body geometry and west –east trending normal faults significantly reduce connectivity between these reservoirs (Fig. 3). Barriers and baffles are also created


90

M. GAINSKI ET AL.

Fig. 1. (a) Schiehallion field location; West of Shetland Islands in North Atlantic Ocean, United Kingdom continental shelf. (b) Stratigraphic log showing Tertiary Lower Paleocene age T31U and T31L reservoirs.

by sealing or partially sealing faults or local mud barriers (slumps and drapes) at the edges and bases of the main sand channels. As a consequence of this complexity, the number of wells has more than doubled to that specified in the original

development plan as a result of poorer than predicted connectivity between mapped sand bodies. Reservoir pressure is close to bubble point (185 bar) so effective pressure maintenance is required to avoid excessive gas breakout from the oil as the

Fig. 2. Schematic of field infrastructure showing the major components: Floating Production Storage and Offloading (FPSO) vessel, semi-submersible drilling rig, subsea drill centres with several subsea wellheads connecting field producers and water injectors through a system of flow lines to the FPSO. A shuttle tanker transports the oil from the FPSO to Sullom Voe.

RESERVOIR COMPARTMENTALIZATION

Table 1. Schiehallion field facts and statistics First oil – 1998 UKCS Blocks: 204/20a, 25a & 205/16a, 21b Water depth – 400 m 4 Drill centres (DC): 22 producers and 24 water injectors c. 2 billion bbls in place, .270 mm bbl produced to-date c. 140 m bbls/day, Low gas –oil ratio oil (340 scf/stb) Baseline 3D survey – 1996 Six monitor surveys – 1999, 2000, 2002, 2004, 2006 and 2008

reservoir is depleted. This relies on the presence of good hydraulic connectivity between water injector and oil producer wells. Poor connectivity could result in reduced well performance and high solution gas levels (Parr et al. 2000). The Schiehallion field has experienced poorer than predicted connectivity since start-up in 1998. This has focused the Schiehallion subsurface team on identifying and quantifying key factors affecting the field connectivity. The field’s data-rich environment and the presented approach to detecting, modelling and dealing with compartmentalization provide a good example of how our understanding of connectivity can be improved. Continuing acquisition and integration of field surveillance data should result in further improved understanding of

91

the reservoir connectivity – leading to a more effective field management and successful infill drilling. This paper discusses the surveillance techniques used to identify field compartmentalization and describes two examples of recently drilled compartments which were successfully identified following detailed subsurface work. The analysis involved integration of static and dynamic well data with 4D seismic to revise ‘static’ geological models and re-run ‘dynamic’ reservoir flow simulations. This method of conditioning static and dynamic modelling with 4D seismic data, is similar to 4D integrated reservoir modelling (4D IRM) workflows described elsewhere (e.g. Staples et al. 2002, 2005; Lygren et al. 2003; Waggoner et al. 2003).

Field surveillance An integrated surveillance plan is the basis for continuous improvement of our understanding of the field – an outline of the data sources follows: In the Schiehallion field, a permanent down-hole gauge is installed as an integral part of the production tubing in each production well. Permanent down-hole pressure gauges have been installed in all but one production well (the gauge was not installed in this well due to operational issues). Water injection wells have wellhead pressure and temperature gauges. Field water injection frequently continues while oil producers are shut in during

Fig. 3. Depositional environment and Schiehallion main T31 sand reservoir. Higher net-to-gross (N/G) is highlighted in red. Seismic section A– A0 shows typical channel geometry of the NW trending depositional system. The channels are cut by east–west trending faults (grey lines). Dashed rectangles indicate the two example areas discussed in this paper.

92

M. GAINSKI ET AL.

periods of FPSO maintenance work. This allows for pressure interference testing between wells across the field, which can provide insights to interwell connectivity. The production wells in the Schiehallion field do not have continuous flow meters installed as a part of their design. However, well tests are performed on regular basis to allocate production rates to individual wells. During a well test, a single production well is flowed under stable conditions to a well test separator for several hours (normally 12 hours) where the total liquid rate, its composition and gas production rate are measured. The well test separator is sometimes used as an additional production separator during periods when the throughput of the normal production separators is reduced due to operational problems. The water cut of the well is determined by taking a liquid well head sample during the well test. This is required since the test separator measures only liquid rate and gas rate. The oil rate is calculated from the test liquid rate and water cut from the well head sample. Isotopic tracer elements are incorporated into injection water, and detected in production wells after water breakthrough. This helps to identify and calibrate seismically defined compartments, and is especially useful in areas where sand bodies extend across major faults. Most of these faults are considered to be sealing. However, in a few cases communication across these faults has been confirmed as a result of tracers appearing in a producer located in the adjacent fault block. Communication across the fault could be a natural phenomenon (the fault never was a barrier) or was induced later by increased differential pressure across the barrier due to production and injection. Two possible theories explain the latter case: overcoming relative permeability capillary pressure of fault rocks on draw-down (Fisher et al. 2001), or mechanical reactivation of ‘critically stressed’ faults (Zoback 2002). Production logging tools (PLTs) are used to measure flow contribution across the completed interval – a part of the well that is open to produce from or inject into the reservoir. Analysis of the PLT data provides detailed information on which reservoir zone is producing (or accepting) fluid and what type of fluid is produced (oil, water or gas). Most production and injection wells are multi-zone completions in the Schiehallion field so this type of surveillance is critical to our understanding of how a reservoir is being swept and helps identifying bypassed and un-swept areas. PLT logs are also required to understand the zonal split and corresponding material balance. The zonal material balance is simply a volume balance on a single reservoir zone which equates the total production to the difference between the initial volume of

hydrocarbons in the reservoir and the current volume (Dake 1978). Log derived permeability thickness (kh) is used in absence of the measured zonal split. In Schiehallion, PLT logging is relatively expensive due to the sub-sea location of the wellheads that until recently could only be accessed with a semi-submersible rig. As can be seen in Figure 2, the wellheads are located on the seabed and there is no direct access into the well. All the well surveillance jobs compete for time, resource and investment with a nearly continuous drilling programme of production and injection wells. As a consequence, only two PLTs have been acquired to-date (since start-up in 1998). Our understanding of the performance of the wells in which the PLTs were run has increased significantly. In addition, the PLT data is used to calibrate and clarify interpretations of ambiguous 4D seismic responses. A purpose built light well intervention vessel (Nordbø Jøssang et al. 2008) for utilization in the West of Shetland area is due for delivery in 2009. This type of vessel can position itself directly above the wellhead and should be able to conduct standard logging jobs as well as some well interventions. This removes the necessity for the semisubmersible rig, and should result in a higher number of PLTs in Schiehallion in future. Formation pressure tests are taken in all the new wells. A small chamber is pushed against the newly drilled formation and a small volume of reservoir fluid is produced into it. During this short flow period typically lasting a few minutes, pressure of the fluid in the chamber is measured. The pressure from the build-up after the flow period is the formation pressure at this depth in the wellbore. Several pressure points are taken in a well and used to calculate a formation pressure gradient. The calculated pressure gradient often shows some breaks, indicating presence of vertical barriers that separate the reservoir into thinner flow units leading to the definition of previously undetected sub flow units. These new sub flow units are detected as a result of differential depletion of reservoirs with vertical barriers and baffles. This information is extremely important in later field life as these sub flow units become future infill targets that can be newly drilled or re-perforated in existing wells to maximize oil recovery from the reservoir. Improved reservoir description of all flow units allows better placement of new wells by targeting specific zones or performing well interventions where individual zones are shut-in or opened up. In Schiehallion, flow barriers can be geologically subtle, and difficult to image and interpret within existing 3D seismic data. Connectivity therefore cannot often be confirmed from a static 3D seismic image alone. Flow connectivity in the production simulation model is continuously assessed


by evaluation of the identified barriers and baffles. We use all the 4D seismic surveys acquired to-date to test and enhance the dynamic reservoir model predictions (cf. Staples et al. 2002, 2005). 4D seismic (or time-lapse 3D seismic) involves repeat acquisition of 3D seismic after a period of oil, gas or water production and water and/or gas injection. Seismic acoustic impedance is a function of rock bulk density and propagation velocity. When the rock matrix density remains constant, the bulk density is controlled by changes in the fluid type present in pore spaces and its saturation. When water replaces oil or gas phase, the rock bulk density increases and the corresponding acoustic response becomes ‘harder’. Increasing reservoir pressure reduces propagation velocity and has an opposite effect on the rock acoustic response by making it ‘softer’. The field production and injection activities alter the fluid saturation in the reservoir and its pressure. These changes in the reservoir acoustic impedance are recorded as seismic amplitudes and represent a snapshot of the reservoir at this particular time. Thus, 3D seismic surveys, repeated at regular time intervals, are compared with each other to monitor changes in the reservoir caused by fluid movements and pressure variations. Seismic 4D responses are calculated by comparing amplitudes between the base survey (first 3D seismic survey acquired preferably before the field start up) and monitoring surveys (repeat surveys) at required time intervals. In Schiehallion six monitor surveys were acquired since the baseline 1996 3D survey: 1999, 2000, 2002, 2004, 2006 and 2008. The seismic quality and repeatability of the monitor surveys are very good (Campbell et al. 2005). The 4D seismic signal is strong for most of the reservoirs allowing even moderate changes in pressure and fluid movement to be detected between consecutive surveys. A qualitative approach to interpret 4D responses was used to capture changes introduced to the reservoir by continuous production and injection. This method primarily involves analysis of changes in magnitude of the seismic responses with time for a given reservoir. These responses are described as hardening, softening or showing no change (e.g. increased water saturation causes hardening whereas increase in pressure or gas saturation results in softening). All available surveillance and well data are integrated in order to correctly interpret these 4D seismic responses (e.g. differentiation between gas saturation increase or pressure increase of the softening response). Floricich et al. (2008) describes a seismic 4D based method that can be used to analyse flow barriers between reservoir compartments. Coherence analysis (measure of similarity between the adjacent seismic traces) is used to assess the spatial variability of the time-lapse

93

seismic. Production induced discontinuities along the strong flow barriers will remain constant with time as opposed to the weak barriers that show different seismic responses on each monitor survey.

Field compartments Key in identifying compartmentalization is integration of all the surveillance data sources. The latest dynamic full field model (FFM) was created in 2003 (Freeman et al. 2008). This model is regularly updated to include new surveillance and dynamic data gathered to-date. Occasionally, we identified some new channel bodies that were previously not visible on the seismic data. These new bodies are included in revisions of the simulation model. Nonetheless, we reached the limits of flexibility for updating the existing reservoir simulation model. The existing grid has limited flexibility for adding new channel bodies and flow barriers. The FFM 2003 grid is based on proportional gridding techniques with control lines for the majority of identified channels. As a result, there are several pinch-out locations in the grid that cause modelling and subsequent flow simulation difficulties. These difficulties are often a result of large volume contrasts between the small pinch-out cells and much bigger neighbouring grid cells which cause numerical instability and convergence errors during flow simulation. At the time of writing, a new static FFM is being built for delivery in 2009. The new FFM will be based again on proportional gridding technique, but with fewer control lines to reduce the volume of pinch out cells in the simulation grid. The channels and sand bodies will be identified by discrete channel parameters. The result is essentially a structurally conformable sugar-cube-like grid that can easily be updated for new sand bodies and flow boundaries because they are not irrevocably fixed to grid constructional surfaces. Thus, our understanding of the field connectivity has evolved since the geological static model was created in 2003/4, and simulated production flow was matched to real historical production data in the full field simulation model in 2004/5. During the flow simulation ‘history match’ process, items like barrier strength and reservoir permeability are adjusted in the model in order to reproduce the observed surveillance data in the flow simulation (e.g. well initial pressures, well water breakthrough time, 4D seismic responses). In 2005, the history match was achieved with a relatively small number of faults and barriers (Fig. 4) when compared to the total number of potential features identified (Freeman et al. 2008). With limited dynamic and surveillance data, barriers were carefully selected to reflect likely geological connectivity

94

M. GAINSKI ET AL.

Fig. 4. T31 sand barriers (black) and baffles (red) used in 2005 history match of the 2003 Full Field Model. Also shown are field producers (green lines) and water injectors (blue dots).

and in some cases supported by production, 4D and tracer information. However, as discussed by Fisher & Jolley (2007), history matches do not ‘prove’ a model’s validity – they can be achieved in models that bear little resemblance to the ‘true’ geology, and do not necessarily predict true flow and depletion behaviour. The convergence between an FFM and this ‘true’ subsurface comes from: continuous critical assessment of the geological concepts being modelled and evidence based inclusion of features such as faults; rapid incorporation of new geological, surveillance and production data; and iterative updates between 3D–4D seismic interpretation, static and dynamic models. This is the basis for 4D Integrated Reservoir Modelling (4D IRM) approaches (e.g. Staples et al. 2005). In the case of the Schiehallion IRM, additional compartmentalization information was provided by the new 4D surveys, infill wells, PLTs and dynamic production data post 2005. Consequently the simulation model was updated with new barriers and baffles and history matched again in 2007. Some flow barriers have only been identified on the later monitor surveys (2006 and 2008) after extended periods of production and water injection affected acoustic properties around them. As a result, existing barriers have been modified in the FFM (location and transmissibility) and new barriers added. The update of the reservoir model with the new set of barriers permitted a better match between the reservoir simulation and the observed 4D seismic effects (Floricich et al. 2008). The overall result is that connectivity has

reduced in the FFM as more compartments have been revealed.

Time-lapse seismic (4D) The use of 4D seismic as a tool for production surveillance, guiding reservoir management and infill drilling to access un-swept compartments, has been highly effective in Schiehallion. Figure 5 shows a distribution of the main T31U channel sands and three amplitude difference maps generated for the 2002, 2004 and 2006 monitor surveys. Seismic amplitudes extracted from the main T31U Sand reservoir are indicative of the reservoir net-to-gross ratio (Fig. 5a). Better quality hydrocarbon filled sands are shown in red whereas amplitudes highlighted in blue are indicative of intra-channel, poorer quality sands, silts and shales. An amplitude switch-off at the OWC can be clearly seen in the NW corner of this map. Also shown are the field producers (black bars) and water injectors (blue dots) drilled to the end of 2002. Figure 5b –d show amplitude differences between the 1996 base survey and each monitor survey. Black polygons (A, B and C) are used to help reader focus on areas where 4D responses have changed most with time. The 1996– 2002 4D difference map (Fig. 5b) reveals large areas of the reservoir that are coloured red that were interpreted as gas coming out of solution due to de-pressuring around the producing wells. This interpretation suggests that the producers were insufficiently

RESERVOIR COMPARTMENTALIZATION Fig. 5. (a) Seismic amplitude map for Schiehallion T31U sand. Major channel sands (A, B and C areas) are highlighted by amplitudes shown in red (red corresponds to higher netto-gross reservoir sand, blue is low net-to-gross). An amplitude switch-off at the OWC can be clearly seen in the north–west corner of this map. (b), (c) & (d) 4D responses (amplitude differences between the 1996 base survey and each monitor survey) in 2002, 2004 and 2006. Areas A, B and C show good examples of softening (red) and hardening (blue) of seismic responses due to production and injection. 95

96

M. GAINSKI ET AL.

pressure-supported by the existing water injectors. In other parts of the field there are also red areas around some water injectors where the 4D responses are interpreted to be caused by pressure increase (softening). A good example of such a pressure compartment can be seen in the north-western corner of polygon C (Fig. 5c). Hence, the seismic 4D responses should not be analysed in isolation, their interpretation must be integrated with all other surveillance data types to give the most probable explanation. Interpretation of the 2002 4D responses (Fig. 5b) led to the drilling of five new water injectors in order to provide a better pressure and sweep support to the existing producers in these compartments. Two channel sands are highlighted in Figure 5 by black polygons A and C and will be used to demonstrate a varying degree of communication between a producer and injector pair. Figure 5b shows a red 4D response in the channel complex outlined by Polygon A. This is as a result of lack of pressure support from the existing water injector located to the NW, just outside of the main channel. The seismic 4D evidence was also supported by the pressure build-ups in a producer located in the southern end of Polygon A. This producer showed a significant reduction in reservoir pressure and increased gas–oil ratio (GOR) during the same time period. A new water injector was drilled in the northern part and in the middle of the mapped sand channel. Figure 5c, d show a greatly reduced gas signature in this channel. The channel sand has dimmed as a result of increased pressure support and gas going back into solution. The producer also showed a reservoir pressure increase and a reduced GOR after a new injector came online. Figure 5d (2006 4D) shows a clear water sweep between the water injector/producer pair where the channel sand has turned blue indicating an increase in water saturation. This seismic response also agreed with the increasing water cut in the producer and detection of the unique tracer injected into the new water injector at start up. Figure 5b shows a large red area outlined by Polygon C. This 4D response corresponds to a channel complex with an oil producer located approximately in its central part. This producer also suffers from insufficient pressure support. Pressure build-ups in this well show reduction in the reservoir pressure and increased GOR. Based on the interpretation of the 2002 4D, a new water injection well was drilled in the north-western end of this channel to improve pressure support to the producer. Figure 5c shows the location of the new water injector and the 2004 4D response after well start up. The area around it turned bright red as a result of water injection and corresponding pressure increase. The strong pressure overprint of this

injector, drilled in a ‘sealed box’, is significantly reduced after it was shut in prior to the 2006 4D survey acquisition. Figure 5d shows a blue patch (increased water saturation) to the east of the injector indicating some level of connectivity with the producer. The communication with the producer was also confirmed by the detection of the unique isotopic tracer which was pumped into this injection well at start up. The 2002 4D seismic interpretation played a key role in our review of the total water injection distribution among the available injectors. The available field water injection capacity was targeted at the high value water injection wells (wells delivering most pressure support and barrels of oil per barrel of water injected). The 2006 4D snapshot (Fig. 5d) clearly shows a significantly improved pressure support and sweep pattern for most parts of the field. The blue colour is indicative of increasing water saturation which covers an increasing area of the T31 reservoir. The 4D interpretation not only helped us in more optimal placing of new water injectors but also assisted us in a development of the field wide very successful water injection strategy.

Infill target example 1 (see Fig. 3 for location) The requirement for two infill injectors was identified through the integration of static data with available surveillance information. Figure 6 shows an example where two new injectors (WW12 and WW16) were eventually needed in a compartment in order to provide better pressure support and more effective sweep to two up-dip oil producers (WP02 and WP14). This is an illustration of an area where the initially mapped and modelled reservoir connectivity did not agree with subsurface ‘truth’, and the geological and flow simulation models had to be significantly modified to include new baffles and barriers in order to match all the surveillance data in this area such as formation pressures, PLTs and seismic 4D surveys used to define new reservoir compartments. Originally the water injector WW06 was drilled to support the dual zone producer WP02. Both wells started injection and production respectively in 1998. The 2002 4D response in the T31 upper sand around the heel of the producer was originally interpreted as increasing water saturation (Fig. 5b). A new ‘attic’ producer WP14 (Fig. 5d) was drilled up-dip from WP02 in 2005 and encountered significantly depleted T31U sand with no water responses identified on well logs. A PLT was run around the same time in the dual zone WP02 producer which had a 20% water cut. The PLT confirmed that only

RESERVOIR COMPARTMENTALIZATION 97

Fig. 6. Example of infill targets – two new water injectors WW12 and WW16. (a) T31U sand seismic amplitude map showing location of the target area. (b) 1996–2002 4D difference map. The blue area on the 4D map was interpreted as increased water saturation. (c) 1996–2004 4D difference map. The blue area around WP02 is not present. A reservoir barrier just to the south of the WW06 injector is clearly imaged. Water produced in WP02 must come from the T31L sand. (d) 1996–2006 4D difference map showing two new water injectors (WW12 & WW16) drilled to the South of the identified 4D barrier to provide pressure support and sweep oil to WP02 and newly drilled WP14 producer. Grey polygon indicates a ‘data hole’, an area located below the FPS where the 2006 4D survey was not acquired.

98

M. GAINSKI ET AL.

the lower sand was connected to the WW06 injector (indicated by large cross flow of 4000 barrels of liquid per day during shut in from the lower zone, which was pressure supported into the unsupported upper zone; and by water entering the well in the lower sand only). An appraisal well (W11) which was targeting a deeper horizon was opportunistically used to penetrate the T31 reservoir about 250 m from the heel of WP02. The formation pressure data in this well showed a 500 psi pressure difference between the T31U and T31L sands (see a stratigraphic column – Fig. 1b) but no water response was logged. This was a clear indication that the shale between the T31U and T31L sands was a strong vertical barrier in this compartment. Prior to the collection of surveillance data, this scenario was considered unlikely. Detailed mapping of the intra-T31 upper to lower shale had identified large areas where the barrier shale was considered to be very thin or absent due to erosion by channels in the upper sand. Small faults with throws of ,5 m were also expected to create hydraulic communication

between the upper and lower sands (cf. Manzocchi et al. 2007). This scenario was reflected in the geological model, which at that time provided hydraulic connectivity between the T31 upper and T31 lower. The W11 pressure data combined with WP02 PLT and the new WP14 well results proved that an alternative geological model was required to match the pressure data. The initial processed 2004 seismic data did not show a clear barrier between the WW06 injector and WP02 producer. The 2004 monitor survey was re-processed in 2005/6, using improved processing techniques. This enabled the team to pick up a new reservoir barrier based on the 4D interpretation (Fig. 6c). This barrier was supported by evidence from the WP02 PLT, W11 formation pressures and WP14 formation pressures. The data integration resulted in drilling two new water injectors, WW12 and WW16, dedicated to the T31U sand only, to provide a better oil sweep and pressure support towards the WP02 and WP14 producers. Figure 7 shows a production history plot for WP02 well and how 4D responses are calibrated

Fig. 7. Production history and calibration of 4D responses. The horizontal time axis covers a period from the field start up to end of 2007. The upper graph shows daily oil production in green (left axis in thousand of bbls), water cut in blue (right axis in %) and produced GOR in red (right axis in standard cubic feet/bbl). The black arrows indicate when monitor surveys were acquired so the calculated 4D responses around WP02 can be correctly interpreted. The lower graph shows the bottom hole pressure (BHP, blue curve and red dots; values on the left axis in psi) which remained constant until the new water injectors went on line. The BHP trend is gradually going up as a result of this additional support. The well head pressure (WHP) is in magenta. This shows that the well is normally operated at full open flow.

RESERVOIR COMPARTMENTALIZATION 99

Fig. 8. Example of infill targets – a new oil producer. (a) 1996– 2002 4D difference map for the T31U sand. Also shown are seismic responses around the CW11 injector (red pressure response) and the CP05 producer (red gas response due to the lack of support from CW11). The grey dashed line indicates extend of the erosive channel cutting into the T31L sand. (b) 1996– 2004 4D difference map for the T31U sand. CP23 target in Sand 1 is highlighted by red seismic response due to increase pressure as a result of water injection in CW16. Area around CP05 has significantly reduced gas response indicating pressure support coming from CW16 and pushing it back to the oil phase. CW11 is shut in but water response (blue) is clearly seen in a small compartment around it. (c) 1996– 2002 4D difference map for the T31L sand. Water injector CW11 does not provide adequate support to CP05 where gas comes out of solution. A new water injector CW16 is drilled to provide pressure support to both sands. (d) 1996–2004 4D difference map for the T31L sand. Seismic response around CP05 is now indicating reduced gas saturation and increased pressure support by CW16. The 2004 4D images also show 2 areas of pressure increase denoted by CP23. The blue arrows indicate communication lines with CW16. This is the new infill target CP23.

100

M. GAINSKI ET AL.

with the existing dynamic well data. The black arrows indicate when seismic monitor surveys were acquired in order to correctly interpret the calculated 4D responses around WP02. Such calibration is more difficult if a given well is producing from or injecting into multiple zones.

Infill target example 2 (see Fig. 3 for location) The second example is of a new oil producer CP23 as shown in Figure 8. The identification of new reservoir compartments provided the team with new infill opportunities. In this example a dual zone target was identified through detailed remapping using a new deposition model and integrating 4D data from the re-mapped intervals. This analysis identified two vertically stacked sands that were over-pressured and untapped by the existing producer CP05. The T31U target (Sand 1) is clearly shown on Figure 8b (2004 4D) where it is marked by a red seismic response centred on the CP23 location. This is interpreted as increased pressure, after the CW16 injector went on line. There is no seismic response for the same area on the 2002 4D (Fig. 8a) indicating that there was no pressure communication with the CW11 injector and suggesting a possible barrier between these two areas. The red area around CW11 is interpreted as pressure increase caused by injection into a small compartment. This is also confirmed by seismic responses around the CP05 producer. Softening of the seismic response here is caused by gas coming out of solution due to the lack of pressure support coming from CW11. There is also some evidence on the 2004 4D (Fig. 8b) that CW16 water injector was providing enough pressure support to the CP05 area where exsolved gas was successfully pressurized back into solution. The T31L target (Sand 4) shows similar seismic responses to those for the overlying Sand 1. Figure 8d (2004 4D) shows a red seismic response in the CP23 area. This sand-body is separated into two areas by the erosive channel. The 4D response is interpreted here as increased pressure, after the CW16 injector went on line. There is no seismic response for the same area on the 2002 4D (Fig. 8c) indicating that there was no pressure communication with the CW11 injector during this time. The seismic cross-section in Figure 9 shows the CP23 well trajectory and two sands that were interpreted to be over-pressured on the 4D responses. The drilling results confirmed this interpretation but pressures encountered in both sands were even higher than predicted. Also shown here is a gamma ray log and vertical barriers found within

Fig. 9. 3D seismic cross-section along the CP23 producer showing infill targets in T31U and T31L sands. The solid green lines are the completion intervals in CP23. The blue curve shown along the well trajectory is the gamma ray log with vertical barriers indicated. The T31L sand is partially eroded by a younger channel.

the T31U sand. Figure 10 shows the formation pressure data points with the CPI (Computer Processed Image) log to further emphasize the vertical barriers found. The sand is about 30 m thick and due to the limited seismic resolution cannot be subdivided further. However, the log and formation pressure data points clearly show that Sand 1 is subdivided in 3 different flow units that are not

Fig. 10. Vertical compartments, formation pressure data points v. the CPI (Computer Processed Image) log interpretation showing sand, shale, water and oil saturation. The T31U sand is imaged as a single sand-body on the 3D & 4D seismic. However, the well logs and pressure data clearly show two shale breaks in the T31U. These shales appear to be vertical barriers to flow as the formation pressures have a step change of c. 150 psi between them.


currently included in the geological model. The pressure difference between the units is around 150 psi. This type of vertical barrier is difficult to map and predict. In future, more surprises like these can be expected. The down hole flow control in this well is located between the T31U and T31L Sands. This enables the sweep and production from these two sands to be optimized. However, there is nothing in place in the well to optimize the sweep of the three different flow units found in the T31U Sand, so these are treated as one flow unit. An additional issue for these sands was the pressure prediction before drilling. Detailed seismic mapping indicated limited areas for flow connectivity to the existing CW16 injector and this was incorporated into the static and dynamic modelling. However, this resulted in model predictions that were less connected than actual pressure data showed. This was good news from the perspective of injection support but for drilling and completions it required an increase mud-weight with barite to manage the pressure, resulting in unplanned potential for formation and/or completion damage. After completion, the well was not immediately flowed back since the sub sea well head was not yet connected to the flow lines that run from the drill centre to the FPSO (Fig. 2). The well was brought on production 3 months after completion, however, only one zone flowed and the production rate was very poor. This was caused by blocked completion tubing and production screens as a result of incompatible drilling mud which was left in the well for the period of three months. The well was successfully sidetracked and completed in 2009 and is currently being produced.

Conclusions Geological barriers to flow are still a major uncertainty after 10 years of production in the Schiehallion Field. Integration of a diverse and continuously growing surveillance data set is essential for identifying reservoir compartments. Lateral barriers such as faults and mud-draped or slumped channel margins can be detected by integration of surveillance data like pressures from permanently installed down-hole gauges, isotopic tracers and 4D seismic. Vertical barriers, however, are more difficult to detect and require PLTs and formation pressure tests. Identification of these vertical barriers will become increasingly important in later field life as the field matures and production becomes more reliant upon access to residual unswept compartments. The key to identifying these more challenging infill targets will be further 4D acquisition and production logging in more wells. Seismic

101

reprocessing advances will also play an important role as will challenging commonly accepted depositional models and interpretations. However, many features will remain below detection limits without a step change in seismic resolution as demonstrated by the CP23 pressure plot in Figure 10. Quantitative 4D seismic predictions compared to simulation model predictions will be required to calibrate pressure forecasts for infill wells and reduce the risk of formation damage during well operations. The Schiehallion Field development experience has clearly shown that compartmentalization can have a large impact on static and flow-connected volume estimates, and the associated development planning. In hindsight, the initial field development plan underestimated reservoir complexity and a level of compartmentalization present across the field. Our understanding of the Schiehallion field has significantly improved as a result of the last ten years of its production history and surveillance data gathered during this time. The field development plan has been updated accordingly to provide a better day-to-day field production management and to maximize its ultimate hydrocarbon recovery. We would like to thank the Schiehallion Field partners (BP, Shell, Hess, StatoilHydro, Murphy Oil and OMV) for providing the data and permission to publish this work.

References Campbell, S., Ricketts, T. A. et al. 2005. Improved 4D seismic repeatability – a West of Shetlands towed streamer acquisition case history, UKCS. SEG Annual Meeting, Expanded Abstracts, 24, 2394– 2397. Chapin, M., Terwoght, D. & Ketting, J. 2000. The from seismic to simulation using new Voxel body and geologic modelling techniques. The Leading Edge, 19, 408 –412. Dake, L. P. 1978. Fundamentals of reservoir engineering, Developments in Petroleum Science, Elsevier, London, 8. Fisher, Q. J., Harris, S. D., McAllister, E., Knipe, R. J. & Bolton, A. J. 2001. Hydrocarbon flow across faults by capillary leakage revisited. Marine and Petroleum Geology, 18, 251–257. Fisher, Q. J. & Jolley, S. J. 2007. Treatment of faults in production simulation models. In: Jolley, S. J., Barr, D., Walsh, J. J. & Knipe, R. J. (eds) Structurally Complex Reservoirs. Geological Society, London, Special Publications, 292, 1–24. Floricich, M., Evans, A., McCormick, D., Jenkins, G. & Stammeijer, J. 2008. Adding the Temporal Coherence Dimension to 4D Seismic Data: Assessing Connectivity in the Schiehallion Field. 70th EAGE Conference & Exhibition, Rome, Italy, 9– 12 June 2008. Freeman, P. J., Kelly, S., MacDonald, C., Millington, J. & Tothill, M. 2008. The Schiehallion field: lessons

102

M. GAINSKI ET AL.

learnt modeling a complex deepwater turbidite. In: Robinson, A., Griffiths, P., Price, S., Hegre, J. & Muggeridge, A. (eds) The Future of Geological Modelling in Hydrocarbons Development, Geological Society, London, Special Publications, 309, 205– 219. Lygren, M., Fagervik, K. et al. 2003. A method for performing history matching of reservoir flow models using 4D seismic data. Petroleum Geoscience, 9, 85–90. Manzocchi, T., Walsh, J. J., Tomasso, M., Strand, J., Childs, C. & Haughton, P. D. W. 2007. Static and dynamic connectivity in bed-scale models of faulted and unfaulted turbidites. In: Jolley, S. J., Barr, D., Walsh, J. J. & Knipe, R. J. (eds) Structurally Complex Reservoirs. Geological Society, London, Special Publications, 292, 309–336. Nordbø Jøssang, S., Friedberg, R., Buset, P. & Gramstad, B. 2008. Present and future well intervention on subsea wells. Society of Petroleum Engineers, SPE Paper 112661.

Parr, R., Marsh, M. & Griffin, T. 2000. Interpretation and Integration of 4D Results Into Reservoir Management, Schiehallion Field, UKCS. 70th SEG Annual Meeting, Expanded Abstracts, 1464– 1467. Staples, R., Hague, P., Cooke, G., Ashton, P., Stammeijer, J., Jolley, S. J. & Marshall, J. D. 2002. Integrating 4D seismic to optimize production. Society of Petroleum Engineers, SPE Paper 78346. Staples, R., Stevens, T., Leoutre, E., Jolley, S. J. & Marshall, J. D. 2005. 4D seismic history matching: the reality. Paper D009, 67th EAGE Conference, Madrid. Waggoner, J. R., Cominelli, A., Seymour, R. H. & Stradiotti, A. 2003. Improved reservoir modeling with time-lapse seismic data in a Gulf of Mexico gas condensate reservoir. Petroleum Geoscience, 9, 61–72. Zoback, M. D. & Zinke, J. C. 2002. Production-induced normal faulting in the Valhall and Ekofisk oil fields. Pure and Applied Geophysics, 159, 403– 420.

Variation in fluid contacts in the Azeri field, Azerbaijan: sealing faults or hydrodynamic aquifer? R. S. J. TOZER1,2* & A. M. BORTHWICK1 1

BP Exploration & Production, Chertsey Road, Sunbury on Thames, TW16 7LN, UK

2

BP Exploration & Production, 240-4th Avenue SW, Calgary, Alberta, T2P 2H8, Canada *Corresponding author (e-mail: [email protected]) Abstract: The Azeri field in the South Caspian Sea, offshore Azerbaijan, is a periclinal anticline 20 km in length containing multiple stacked reservoirs of Pliocene age. Appraisal wells that were drilled at the eastern end of the structure identified multiple oil–water contacts and fluid pressure gradients in both of the principal reservoirs, the Pereriv B and D. At the time, these data were interpreted to indicate the presence of compartments at the eastern end of the field as a result of sealing faults within the aquifer. This local compartmentalization seemed to be in marked difference to the majority of the field where pressure connectivity had been observed. A new analysis of the pressure data for the Pereriv B shows that aquifer pressures at sea-level datum define a simple water potential gradient. As a result of this, the oil – water contact in this reservoir is gently inclined towards the NNE. The precise inclination and orientation of the oil–water contact has been determined geometrically using the depths and coordinates of free-water levels and oil–water contacts from around the field. The best-fit inclined oil– water contact for the Pereriv B also provides a good fit to the contact observed from seismic amplitudes. The new analysis provides a more optimistic view of reservoir connectivity, and the conceptual geological model for the eastern end of the field is now consistent with observations made in the rest of the Azeri field.

Variations in pressure, fluid contacts and fluid compositions within hydrocarbon reservoirs are commonly attributed to reservoir compartmentalization as a result of sealing faults or stratigraphic complexity (see Knipe et al. 1998; Jolley et al. 2007 and papers in the same volumes). However, in his seminal paper on hydrodynamics, Hubbert (1953) suggested that many off-structure hydrocarbon accumulations which have been classified as fault or stratigraphic traps, may in fact be hydrodynamic traps. All of the natural examples originally described by Hubbert (1953) occur in basins where the flow of water is towards the basin centre, driven by the topographic gradient from surrounding highlands; other examples of this process have since been documented from many other basins around the world (e.g. Alem et al. 1998 and Crossley & McDougall 1998; see also Dennis et al. 2000 for a review). The opposite situation, where the flow of formation water is driven away from the basin centre by sediment compaction, has also been described in the South Caspian Basin (Bredehoeft et al. 1988) and North Sea (Dennis et al. 2000). A combination of topographic and compaction-driven flow has been modeled in the San Joaquin Basin (Wilson et al. 1999). Other mechanisms that are believed to be responsible for flow of water in the subsurface include tectonism (Estrada & Mantilla 2000; Seggie et al. 2003), and

variations in aquifer salinity in response to changes in regional geothermal gradient (Stenger 1999). In this paper we will look at a case-study from the Azeri field, offshore Azerbaijan, where variations in fluid contacts are observed around the field.

ACG background The Azeri, Chirag and Gunashli (ACG) fields lie in the South Caspian Sea (Fig. 1). These three fields are all located at the crest of a major periclinal anticline 50 km in length. First production was from the Shallow Water Gunashli (SWG) field in 1982, and was followed by first production from Chirag in 1997, Azeri in 2005 and Deep Water Gunashli in 2008. Initial development of the Azeri field has focused on the Pereriv B reservoir; this unit is 40 –60 m thick, has high net-to-gross, and is composed of Pliocene fluvio-deltaic quartz arenites. The reservoir has excellent porosity (20– 28%) and permeability (0.1–1D) and good lateral continuity. Early in the appraisal history of the field, the oil– water contact was found to be deeper on the northflank of the structure (at 3406 mTVDSCS (metres true vertical depth sub Caspian Sea)) than on the south flank (at 3076 mTVDSCS). In order to understand this change in oil– water contact, in 2000 an


104

R. S. J. TOZER & A. M. BORTHWICK

Fig. 1. (Top) Digital elevation model showing the regional location of Azerbaijan and the Caspian Sea. The white rectangle shows the location of the map below. (Bottom) Location of ACG (Azeri, Chirag, Gunashli) fields in the South Caspian Sea. Abbreviations: SWG, Shallow Water Gunashli; DWG, Deep Water Gunashli. Also shown is the regional aquifer overpressure (from well data) in the Pereriv B reservoir, contour values in bar. Note the trend of decreasing overpressure towards the Aspheron Peninsula and AGC fields.

AZERI AQUIFER

105

Fig. 2. (a) Fault juxtaposition diagram (‘triangle plot’) for the Pereriv B reservoirs generated using the Vshale log from the GCA6 well. Due to the low Vshale content of the sands only faults which completely offset the main reservoir intervals (Pereriv B and D) would have sealing potential (shale gouge ratio .0.2; Freeman et al. 1998). (b) Representative seismic line across the eastern part of the Azeri field (see Fig. 5 for location). The top of the Pereriv B reservoir is marked in green and the base would correspond to the blue reflector immediately below this. Note that the reservoir is not offset by any faults which would be large enough to have sealing potential. The seismic data are towed streamer, full-offset reflectivity, pre-stack depth migrated, and have been tied to well depths.

106


appraisal well (GCA6) and two planned sidetracks (GCA6Z & GCA6Y) were drilled at the eastern end of the structure where the anticline plunges gently towards the SE. These three wells identified two apparently anomalous oil –water contacts at elevations intermediate between the established north and south flank contacts. At the time, the different oil –water contacts and aquifer pressures in this reservoir and the underlying Pereriv D reservoir were believed to indicate the presence of lateral sealing faults within the aquifer. In turn, these faults were thought to indicate smaller compartments than assumed prior to appraisal. The interpretation of sealing faults to explain the apparently anomalous results from appraisal drilling is in contrast to the absence of faults with sealing potential in the eastern part of the Azeri field. The data which demonstrate this are shown in Figure 2. The fault juxtaposition diagram (‘triangle plot’; Fig. 2a) shows that, due to the low Vshale content of the sands, only faults which entirely offset the main reservoirs (the Pereriv B and D) would be associated with a shale gouge ratio of greater than the 0.2 sealing threshold (Freeman et al. 1998). However, a representative seismic line across the eastern part of the Azeri field

(Fig. 2b) shows that no faults with this amount of offset are present in this area of the field. In addition, good reservoir connectivity has been noted in the majority of the field. This is illustrated by a plot of pre-production pressure decline for the period 1987–2005 (Fig. 3), caused by production from the adjacent SWG and Chirag fields to the north-west. Although there is some variation in the fit of the data to the best fit trend of depletion (due to a combination of spatial variation of the data and periods of increased oil production and/ or decreased water injection), in general the plot shows a steady decline in oil pressure of 1.4 bar/a throughout the Azeri field. This supports the view that the reservoir shows a good degree of connectivity throughout the field. Furthermore, the regional aquifer overpressure shows a clear trend of decreasing overpressure away from the central part of the South Caspian Basin (to the SW of the ACG fields; Fig. 1) where up to 10 km of sediment was deposited during the Pliocene and Quaternary (Brunet et al. 2003). The extremely rapid rate of sediment deposition resulted in overpressure due to disequilibrium compaction (Bredehoeft et al. 1988; Davies & Stewart 2005), and present-day subsurface flow is driven

Fig. 3. Pre-production pressure (at 2900 mTVDSCS datum) against time for the Pereriv B and D reservoirs in the Azeri field. There is a steady decrease in pressure throughout the field due to production from the SWG and Chirag fields to the NW. This suggests that both reservoirs are well connected on a regional scale.

AZERI AQUIFER

towards surface and submarine outcrops around the Aspheron Peninsula. This pattern is consistent with the more elevated oil–water contact on the south flank of the Azeri field. For these reasons, an alternative hypothesis (that a hydrodynamic aquifer is present beneath the field) was proposed to explain the apparently anomalous results from the east Azeri appraisal wells. The following sections describe the methods which were used to explore this hypothesis, together with its implications for field development.

Pressure data: analysis and results A plot of Pereriv B reservoir pressure against depth for four key appraisal wells (Fig. 4) shows a single trend of oil pressure in all three East Azeri appraisal wells (GCA6, GCA6Z, GCA6Y), suggesting that these wells are in common communication in the

107

oil leg. However, three different water pressure trends and associated oil –water contacts are present (north flank aquifer pressures have also been projected onto this figure); these are the data which were originally interpreted to indicate compartmentalization in the aquifer. In order to assess the spatial organization of aquifer pressure, the data points from each well were first projected to a common sea-level datum using the formation water density (0.098 bar/m) determined from laboratory analysis of well samples (PVT). Having reduced the data for each of the six wells to a single value, a best-fit plane of aquifer overpressure was fitted to the points using EarthVision software (Fig. 5a). The best-fit plane has a gradient of 3.5 bar/km and dips towards 0198N; the isobars are parallel to structure contours on the south flank, and show a pattern of decreasing overpressure towards the north flank. Overpressure also decreases from east to west along the north flank of the structure. The

Fig. 4. Plot of Pereriv B formation pressure against depth for the GCA appraisal wells. Measurements of oil pressure lie on a single line, but there are four aquifer pressure trends. Abbreviations: FWL, free-water level; OWC, oil– water contact.

108


Fig. 5. (a) Map of Pereriv B aquifer pressure at sea-level datum. A simple south to north pressure gradient provides a good fit to all six data points. (b) Map of Pereriv B oil –water contact. A planar contact dipping gently towards the NE provides a good fit to the data, with the exception of the GCA6Z oil– water contact which is shallower than predicted. See Figure 1 for regional location.

AZERI AQUIFER

fit of the planar surface to the data can be assessed by the position of the real data relative to the best-fit plane. The aquifer pressure gradient determined in this way can be converted to a predicted inclination of the oil –water contact, u: tan u ¼ rhowater =(rhowater rhooil ) DH=DL: Where rhowater is the density of the formation water, rhooil is the density of the oil, and DH and DL are the change in elevation and change in distance respectively. The term rhowater/(rhowater 2 rhooil) is known as the tilt multipler and DH/DL is known as the hydrostatic gradient (Hubbert 1953). Using densities determined from laboratory analysis for oil (0.69 g/cm3 ¼ 0.068 bar/m) and water (1.00 g/cm3 ¼ 0.098 bar/m), and the hydrostatic gradient of 0.036 (3.5 bar/km), the predicted tilt of the oil–water contact using this method is 6.68 towards 0198N (Fig. 5a).

Oil – water contact and free-water level data: analysis and results An independent analysis of the spatial organization of the oil–water-contact elevations from around the field has also been carried out. Although the oil – water contact in the Pereriv B is directly penetrated in only two wells (GCA6 and GCA6Z), the freewater level can be determined from paired oil/ aquifer wells at four other locations around the field using the gradient intercept technique (this is explained in Jahn et al. 1998). These locations are in the south-central, NW, north-central and NE sectors of the field. The free-water level depths derived from pressure data in this way provide a valid proxy for the depth of the oil– water contact. This is demonstrated by the very close correspondence of the oil–water contact depths defined by electric logs in the GCA6 and GCA6Z wells (3310 and 3162 mTVDSCS respectively) with the free-water level depths derived from pressure data in the same wells (3307 and 3161 mTVDSCS respectively). The close correspondence is consistent with the excellent reservoir permeability already described. The spatial organization of the two oil –water contacts and three free-water levels was determined by applying basic trigonometry to the depths and coordinates of the known contacts. Three data points from the south-central, north-central and NE sectors of the field were first used to define a single plane. The result was then tested for fit against the two oil –water contacts penetrated by the GCA6 and GCA6Z wells, and also against the GCA6Y well in which only oil was encountered in

109

the Pereriv B reservoir. The best-fit plane dips at 4.48 towards 0268N, and closely honours both the GCA6 oil–water contact and the oil-filled reservoir in GCA6Y. However, this plane is 68 m deeper than the known oil– water contact at the GCA6Z well location. A single plane containing all four free water levels and the oil– water contact in GCA6 was also constructed using EarthVision software in order to provide independent check of the trigonometric analysis; this is shown in Figure 5b. The plane provides a good fit to all five data points, and also honours the oil-filled reservoir in GCA6Y.

Interpretation The single oil pressure trend for the Pereriv B gives confidence that this reservoir is in pressure communication between wells GCA6, GCA6Z and GCA6Y (Fig. 4). The presence of multiple offset water pressure trends and multiple fluid contacts was initially interpreted to indicate compartmentalization in the aquifer. However, the simple south to north water potential gradient (3.5 bar/km, strike 1098) that is defined by the datumed water pressures (Fig. 5a) strongly suggests that a hydrodynamic aquifer is present beneath the oil leg. The orientation of the oil–water contact that results from this is defined by the plane containing the coordinates of the free-water levels and oil– water contacts; this dips gently towards the NE (04.48/0268N) and has a strike sub-parallel to the north flank structure contours (hence the consistent free-water level along this flank). The angle of tilt is comparable to several other global examples compiled by Dennis et al. 2000 (see their Fig. 1). Note that the strike of the oil –water contact plane is rotated 78 clockwise of the water potential grid. This may reflect greater pressure depletion in the western part of the field due to production from the Chirag and SWG fields to the NE; it is well known that pressure differences equilibrate more rapidly than fluid contacts (Smalley et al. 2004). The only anomaly is the oil–water contact penetrated in GCA6Z which lies 68 m above its predicted depth. The underlying Pereriv D reservoir also has an unusually shallow oil–water contact at this well location; both anomalies could be explained by a local area of increased water pressure close to the point of maximum aquifer flux. This would cause a local increase in elevation of the oil –water contact, which therefore does not fit the simple planar model presented in Figure 5b. We admit that the anomalous GCA6Z data point is a weakness in our argument, but it cannot be explained as a fault-controlled compartment because no faults with sufficient displacement can be

110


observed in seismic reflection data (Fig. 2b), and stratigraphic complexity has not been observed in seismic or well data.

Independent test: seismic amplitude An independent test of the best-fit inclined oil– water contact can be provided by seismic attributes. These are known to provide a reliable indicator of the oil– water contact in the Chirag (Robinson et al. 2005) and Gusashli (Manley et al. 2005) fields to the NW. Figure 6 shows the sum of negative amplitude from a 48 msec window centred on the Pereriv B reservoir, derived from a fluid impedance seismic volume. The line of intersection between the best-fit oil –water contact and the Pereriv B top structure has been overlaid, together with the well data points which have been used to define this plane. The line corresponds to an abrupt change in seismic amplitude even in the

eastern part of the Azeri field where the angle between structure and the contact is at its greatest. A range of different seismic attributes, window sizes and seismic volumes show a similar correspondence to the best-fit oil–water contact, providing confidence that the hydrodynamic model is valid. Bright amplitudes are also present below the present-day oil–water contact (Fig. 6); these are clearest on the north flank and may represent a palaeo oil–water contact. These bright amplitudes are oblique to the present-day structure and deepen towards the eastern end of the field; this can be interpreted to indicate that the structure was tilted (downwards in the east) after it was charged with hydrocarbons. As a consequence of tilting, hydrocarbons may have been spilled into the fields to the NW and lost from the Azeri field. This could explain why the palaeo oil–water contact lies deeper than the present-day oil – water contact.

Fig. 6. Map showing the sum of negative amplitude (SNA) from a 48 msec window centred on the Pereriv B reservoir. The line of intersection between the present-day (hydrodynamic) oil–water contact and the Pereriv B top structure has been overlaid on this (heavy black line), together with the well data points. The line corresponds to an abrupt change in seismic amplitude even in the eastern part of the Azeri field. Below the present-day oil–water contact a possible palaeo oil–water contact is also shown (narrow black line; dashed where uncertain). See Figure 1 for regional location.

AZERI AQUIFER

Implications for field development The degree of compartmentalization is of critical importance for any field development (Jolley et al. 2007). Unexpected compartmentalization can reduce production rates and recoverable volumes; conversely there may be a positive impact if compartmentalization is less extensive than expected. Although the risk of compartmentalization cannot be eliminated, as a result of this study the base-case structural model for the eastern part of the Azeri field can contain fewer sealing faults. In terms of field development, it is anticipated that fewer water injection wells will be needed to supply pressure support for Pereriv B production wells. Later in field-life it is possible that there will be a reduced need for infill drilling (additional production wells and/or sidetracks) since it is more likely that the initial production wells will access a greater volume of oil if reservoir connectivity is as extensive as this study suggests. This scenario would result in comparable ultimate recovery from fewer wells, with corresponding savings in development costs. The approach described in this paper also contributes to the description of initial fluid contacts. In some regions of the Azeri field there is limited well control, and therefore there is greater uncertainty in predicting the depth of the oil –water contact. If compartmentalization is suspected but difficult to define, then this uncertainty is even greater. Conversely, if we have confidence in the hydrodynamic hypothesis then the position of the initial oil–water contact can be predicted away from areas with well control. This will help development wells to be correctly positioned.

Conclusions In the Azeri field, aquifer pressures at sea-level datum in the Pereriv B reservoir define a clear south to north water potential gradient of 3.5 bar/ km inclined towards 0198N. The changes in pressure strongly suggest that a dynamic aquifer is present; this is believed to be driven by northward flow of water from the more highly overpressured sediment depocentre in the South Caspian Basin to the south of the field location. As a result of this the oil–water contact is gently tilted towards the NE. The precise inclination and orientation of the contact (4.48 towards 0268N) has been defined by a broad span of oil –water contacts and free-water levels from around the field. Only one well in the south-eastern part of the field is inconsistent with this model. In this well the penetrated oil–water contact is anomalously shallow; a possible explanation of this is a local increase in elevation of the

111

oil –water contact around the point of maximum aquifer flux. The hydrodynamic model can also be tested independently using an attribute map of seismic amplitude. This map shows an abrupt change in seismic amplitude at the predicted oil– water contact, and provides confidence that the model is valid. The new model will be taken into consideration during development of this part of the field in order to optimize well placement. The results from this study illustrate how new analysis of existing data supports a more optimistic model of reservoir connectivity without the need to invoke fault-sealed compartments. We are grateful to all the partners of the Azerbaijan International Operating Company for granting permission to present and publish this case-study. We also thank many of our colleagues within BP for useful discussions and feedback. The map of regional aquifer overpressure in Figure 1 was compiled by N. Abdullayev, S. Duppenbecker, T. Harrold & G. Riley, and prepared by M. Joly. The top reservoir structure grid in Figure 5 was created by M. Norris. Constructive reviews by R. Swarbrick, T. Needham, and the editors are gratefully acknowledged.

References Alem, N., Assassi, S., Benhebouche, S. & Kadi, B. 1998. Controls on hydrocarbon occurence and productivity in the F6 reservoir, Tin Fouye´-Tabankort area, NW Illizi Basin. In: MacGregor, D. S., Moody, R. T. J. & Clark-Lowes, D. D. (eds) Petroleum Geology of North Africa. Geological Society, London, Special Publications, 132, 175– 186. Bredehoeft, J. D., Djevanshir, R. D. & Belitz, K. R. 1988. Lateral fluid flow in a compacting SandShale Sequence: South Caspian Basin. American Association of Petroleum Geologists Bulletin, 72, 416– 424. Brunet, M.-F., Korotaev, M. V., Ershov, A. V. & Nikishin, A. M. 2003. The South Caspian Basin: a review of its evolution from subsidence modelling. Sedimentary Geology, 156, 119– 148. Crossley, R. & McDougall, N. 1998. Lower Palaeozoic reservoirs of North Africa. In: MacGregor, D. S., Moody, R. T. J. & Clark-Lowes, D. D. (eds) Petroleum Geology of North Africa. Geological Society, London, Special Publication, 132, 157– 166. Davies, R. J. & Stewart, S. A. 2005. Emplacement of giant mud volcanoes in the South Caspian Basin: 3D seismic reflection imaging of their root zones. Journal of the Geological Society, London, 162, 1 –4. Dennis, H., Baillie, J., Holt, T. & Wessel-Berg, D. 2000. Hydrodynamic activity and tilted oil–water contacts in the North Sea. In: Ofstad, K., Kittilsen, J. E. & Alexander-Marrack, P. (eds) Improving the Exploration Process by Learning from the Past. NPF Special Publication, Elsevier, Amsterdam, 9, 171– 185.

112


Estrada, C. & Mantilla, C. 2000. Tilted oil water contact in the Cretaceous Caballos Formation, Puerto Colo Field, Putumayo Basin, Colombia. Society of Petroleum Engineers, SPE Paper 59429, 1– 12. Freeman, B., Yielding, G., Needham, D. T. & Badley, M. E. 1998. Fault seal prediction: the gouge ratio method. In: Coward, M. P., Daltaban, T. S. & Johnson, H. (eds) Structural Geology in Reservoir Characterization. Geological Society, London, Special Publications, 127, 19–25. Hubbert, M. K. 1953. Entrapment of petroleum under hydrodynamic conditions. American Association of Petroleum Geologists Bulletin, 37, 1954–2026. Jahn, F., Cook, M. & Graham, M. 1998. Hydrocarbon Exploration and Production. Developments in Petroleum Geoscience, Elsevier, Amsterdam, 46. Jolley, S. J., Barr, D., Walsh, J. J. & Knipe, R. J. 2007. Structurally complex reservoirs: an introduction. In: Jolley, S. J., Barr, D., Walsh, J. J. & Knipe, R. J. (eds) Structurally Complex Reservoirs. Geological Society, London, Special Publications, 292, 1– 24. Knipe, R. J., Jones, G. & Fisher, Q. J. 1998. Faulting, fault sealing and fluid flow in hydrocarbon reservoirs: an introduction. In: Jones, G., Fisher, Q. J. & Knipe, R. J. (eds) Faulting, Fault Sealing and Fluid Flow in Hydrocarbon Reservoirs. Geological Society, London, Special Publications, 147, vii– xxi.

Manley, D. M., Mohammed, S. F., Robinson, N. D. & Thomas, R. W. 2005. Structural interpretation of the deepwater Gunashli Field, facilitated by 4-C OBS seismic data. The Leading Edge, 24, 922– 926. Robinson, N., Ford, A., Howie, J., Manley, D., Riviere, M., Stewart, S. & Thomas, R. 2005. 4D time-lapse monitoring of Chirag Field. The Leading Edge, 24, 928–932. Seggie, R. J., Ainsworth, R. B. et al. 2003. The SunriseTroubadour gas-condensate fields, Timor Sea, Australasia. In: Halbouty, M. T. (ed.) Giant Oil and Gas Fields of the Decade 1990– 1999. American Association of Petroleum Geologists Memoir, 78, 189–209. Smalley, C., England, W. A., Muggeridge, A., Abacioglu, Y. & Cawley, S. 2004. Rates of reservoir fluid mixing: implications for interpretation of fluid data. In: Cubitt, J. M., England, W. A. & Larter, S. (eds) Understanding Petroleum Reservoirs: Towards an Integrated Reservoir Engineering and Geochemical Approach. Geological Society, London, Special Publications, 237, 99–113. Stenger, B. A. 1999. Regional temperature gradient: a key to tilted OOWC. Society of Petroleum Engineers, SPE Paper 53197, 1– 15. Wilson, A. M., Garven, G. & Boles, J. R. 1999. Paleohydrogeology of the San Joaquin basin, California. Geological Society of America Bulletin, 111, 432– 449.

Sedimentological control of fluid flow in deep marine turbidite reservoirs: Pierce Field, UK Central North Sea E. D. SCOTT1,2*, F. GELIN1,3, S. J. JOLLEY1,4, E. LEENAARTS1,5, S. P. SADLER1,6 & R. J. ELSINGER1,5 1

Shell UK Limited, 1 Altens Farm Road, Nigg, Aberdeen, AB12 3FY, UK 2

Marathon Oil, 5555 San Felipe, Houston, Texas 77056, USA 3

Total, Avenue Larribau, 64000 Pau, France

4

Shell Canada, 400 – 4th Avenue S.W., Calgary, Alberta, T2P 0J4, Canada

5

Shell International E&P, 1 Kessler Park, Rijswijk, AB 2280, Netherlands 6

Shell E&P, 701 Poydras St, New Orleans, LA 70139, USA

*Corresponding author (e-mail: [email protected]) Abstract: The Pierce Field in the Central UK North Sea is a twin diapir structure that produces from the Paleocene Forties Sandstone Member (Forties Sandstone). Different hydrocarbon– water contacts encountered in the wells around both diapirs have been variously ascribed to a hydrodynamically tilted oil – water contact or else some form of stepped (compartmentalized) contact. Recent reinterpretation of the structure, sedimentology and fluid geochemistry has indicated that the stratigraphic architecture of the reservoir is the prime control on fluid flow over both geological and production time-scales. These depositional architectures deflect the hydrodynamic flow of aquifer water around the field, resulting in a modified-tilted-contact. The same depositional architectures control the flow of fluids under production. The Forties Sandstone was emplaced by turbidity flows influenced by pre-existing seafloor topography that funneled the flows into discrete sediment corridors and into the Pierce area. The rising twin diapirs further influenced the flows by forming: (a) a small salt withdrawal basin between the diapirs that captured sediment; and (b) enough seafloor topography to prevent the bulk of the flows from depositing significant amounts of sand over the crest of the diapirs. Because the bulk of the high permeability sands were deposited in a rim around the diapirs, the aquifer and injected water does not always flow to structurally higher elevations, but follows the geometry of the channelized sands. While faults are present on both South and North Pierce, they are not extensive and do not appear to play a major role in the compartmentalization of the field. From production data, pressure communication can be inferred around North Pierce and around the majority of South Pierce, the main exception being a block bound by large throw faults in the SE of the southern diapir. Geochemical fingerprinting of the hydrocarbons in Pierce shows families of oils that suggest that the northern and southern parts of the reservoir are separate oil compartments, which is a result of the interaction of the filling history and the stratigraphic and structural architecture of the reservoir.

Fluid flow in hydrocarbon reservoirs is often determined, at least in part, by a combination of structural and stratigraphic components. Over geological time, hydrocarbons migrate to the highest possible structural position in a trap. This buoyancy effect also generally holds true during production of hydrocarbons from a reservoir. However, fluid movement towards the highest point of a structural trap is not always the easiest and preferred path for hydrocarbons. The causes and controls of reservoir compartmentalization (structural and/or stratigraphic architectures) in any given hydrocarbon field are often a

matter of considerable study and debate. Examples can be found in the Forties Sandstone reservoirs in general, and the Pierce Field in particular. Previous studies on the Pierce Field (Fig. 1) have invoked fault sealing to explain variations in pressures and fluid contacts within this twin-diapir field (e.g. Birch & Haynes 2003) – although more recent studies imply faults have little impact, and suggest that stratigraphic architecture is the controlling factor (Fisher & Jolley 2007). The current study of the Forties Sandstone in the Pierce Field further defines the influence of the stratigraphic architecture on the flow of hydrocarbons and water.


114

E. D. SCOTT ET AL.

Fig. 1. (a) Location of the Pierce Field in the UK Central North Sea. (b) The Top Sele Formation to Top Ekofisk Formation isochore generated from seismic showing the distribution of Paleocene depositional systems (Reds/Yellows, thicks; Blues, thins).

Regional depositional setting The Forties Sandstone has long been recognized as the deposits of a series of turbidity current events in a deep marine depositional system (e.g. Ahmadi et al. 2003; Hempton et al. 2005). These events punctuated a generally quiescent regime in which the ‘background’ sediment accumulation involved the passive settling of fines from suspension. The Forties Sandstone system is thought to have developed in response to sediment supplied by extensive deltaic feeders located to the north and west of the main depocentre (Bowman 1998). Sediment influx appears to have been initiated by significant (up to 1 km) uplift of Scotland and the East Shetland Platform in response to thermal doming associated with rifting of the Greenland–European plate. The rejuvenation of Mesozoic hinterlands and basin margins culminated in the dispersal of substantial volumes of clastic material both eastwards into the Northern and Central North Sea and NW into the Faroe– Shetland Basin (Bowman 1998). Although depositional substrates were consistently located

below storm wave-base, there is no evidence to suggest the development of a deep water or ‘abyssal’ setting at this time. The laminated black shales that represent the distinctive signature of the Forties Sandstone interval are more likely to reflect limited circulation and bottom water anoxia. Regionally, the Forties Sandstone is dominated by an elongate NW –SE oriented depositional system, reflecting the south-easterly axial re-working of the sediment influx into the Central Graben (Fig. 1). The absence of the classical ‘fan’ morphology is largely a function of the scale, orientation and topographic expression of the graben complex to which the system was largely constrained. Sand distribution patterns are complicated by the interaction of the axial system with east – west oriented ‘side fans’, the most prominent of which is the Gannet –Guillemot system (Fig. 1). Data from the Forties Sandstone and Nelson Fields suggest deposition within large-scale, NW – SE trending ‘fairways’ characterized by an anastomosing complex of submarine channels (Wills & Peattie 1990; Whyatt et al. 1991).

SEDIMENTOLOGICAL CONTROL OF FLUID FLOW

Channel complexes are thought to have persisted up to the margins of the fan system. The abundance of submarine channel-fills, regarded as the defining characteristic of the Forties Sandstone succession, has been related to the effects of shelf bypass and a dominance of ‘fluidal’ rather than cohesive flow dynamics (Den Hartog Jager et al. 1993; Jennette et al. 2000). In contrast, studies have demonstrated that the distal reaches of the ‘side fans’ were characterized by the accumulation and compensational offset stacking of sand sheets from unconfined turbidity currents which constructed both sand-prone and heterogeneous lobes. Within this framework, the Forties Sandstone at the Pierce Field appears to occupy a medial to distal location within the NW–SE oriented axial system. In this location, it is dominated by channel and channel margin depositional environments (Ahmadi et al. 2003; Birch & Haynes 2003; Hempton et al. 2005).

115

Fig. 2. Structure map on the Top Sele surface showing twin salt diapirs about 3.5 km apart, dips next to diapirs in excess of 608. The Pierce field is developed with horizontal producers and injectors from two drill centres.

Pierce Field The Pierce Field is located at the eastern margin of the UK Sector of the Central North Sea and is split between blocks 23/22a and 23/27 (East Central Graben). It lies approximately 265 km east of Aberdeen, on the western flank of the Jaeren High against which the Forties Sandstone system thins and eventually shales out (Fig. 1). The field is characterized by the accumulation of hydrocarbons on the flanks of two discrete salt diapirs that are separated by a structural saddle approximately 1.5 km wide (Figs 2 & 3). These

accumulations were discovered independently (in 1975 and 1990) and were originally considered as two separate fields (with the southern diapir being Ranger’s Pierce Field and the northern structure forming BP’s Medan Field). It can now, however, be demonstrated that these accumulations are linked through the saddle zone and that Pierce, acquired as a single entity by Enterprise Oil prior to their acquisition by the Shell Group, should be considered as a somewhat exceptional ‘twin diapir’ field.

Fig. 3. Seismic line through North and South Pierce showing the steep dips on the flanks of the rising diapirs. Note the thinning of the Forties Sandstone interval towards the diapirs. See Figure 2 for location of cross-section line.

116

E. D. SCOTT ET AL.

A total of 13 vertical and deviated appraisal wells were drilled by BP and Ranger, although the original development strategy involved the drilling of 6 long-reach horizontal producer wells and 2 crestal gas injectors. Sub-sea producer wells are tied back to a floating production and storage offloading vessel from which the oil is exported by shuttle tanker. Produced gas is re-injected back into the gas cap. In 2003, Shell drilled the first horizontal water injection well in the ‘saddle’ zone and pursued this strategy in 2003 and 2004 with 2 longreach horizontal water injectors designed to support the main producers on the western and southern flanks of the South Pierce diapir (Fig. 2).

Stratigraphy After deposition of the Ekofisk Formation, siliciclastic sedimentation commenced in the Pierce field area with predominately silts and muds, most likely in the distal portions of submarine systems deposited in other parts of the Central Graben, of the Maureen and Lista Formations of the Montrose Group (Fig. 4). The overlying Upper Paleocene Moray Group holds the Sele and Balder Formations. Sedimentation during the Sele Formation is dominated by the deposition of a submarine fan system (the Forties Sandstone) that covers a large portion of the Central Graben. At the Pierce Field the

Fig. 4. Stratigraphic column of the Paleocene in the North Sea (from Ahmadi et al. 2003). The section in the red box is included in this study.


Forties Sandstone is separated into the Lower Forties Sandstone and the Upper Forties Sandstone and can be defined biostratigraphically by the UKOOA biozones S1a and S1b. The overlying Balder Formation contains numerous volcanic tuff deposits.

Dataset The Pierce Field is covered by several vintages of 3D seismic volumes comprised of near, full and far offset data sets. The pre-stack time migration seismic volume was the primary volume utilized to define seismic geometries in the current study. Although seismic data quality is adversely affected throughout the field by the presence of the salt diapirs, analysis of a wider area has provided valuable information on the larger-scale distribution and orientation of sand accumulations which can be extrapolated into the diapir area. The Forties Sandstone Member has been extensively cored in nine of the appraisal wells located on both North and South Pierce diapirs. A total of 661 m of Forties Sandstone core are currently available.

Seismic stratigraphy The fluid flow in any turbidite reservoir will be controlled in part by the stratigraphic architecture of the sandstones and intervening muddier layers. To

117

understand the distribution of sediments at the Pierce Field, investigation of the seismic stratigraphy shows that the section holding the Forties Sandstone interval contains varying seismic facies with complex relationships of the reflectors (Figs 5 & 6). Numerous seismic reflectors terminate onto onlap surfaces at the base of, and internal to, the Forties Sandstone. The onlap surface at the base of the Forties Sandstone indicates complex sea floor topography at the onset of sedimentation. An internal onlap surface shows the continued influence of the existing topography on sandstone deposition. The sea floor topography that influenced the deposition of the Forties Sandstone is not present in the underlying Ekofisk Formation (Figs 5 & 6). The Top Ekofisk horizon is consistent, gently undulating, with minor fault offsets. Deposited on the Ekofisk Formation, the top of the overlying Maureen Formation shows some low angle onlap surfaces indicating minor sea floor topography at the onset of deposition of the Lista Formation. The resultant deposits of the Lista Formation exhibit significant topography which influenced deposition of the subsequent Sele Formation and the Forties Sandstone. The deposits of the Forties Sandstone filled-in the pre-existing topography. Consequently, the top surface of the Sele Formation is a consistent, gently undulating reflector with some minor fault offsets.

Fig. 5. Seismic line (east–west) showing onlaps between the Top Ekofisk and Top Sele in the Pierce area that defines three surfaces. The middle surface is interpreted to be the Base Lower Forties Sandstone surface while the upper surface is the Base Upper Forties Sandstone surface.

118

E. D. SCOTT ET AL.

Fig. 6. Seismic line (north– south) showing onlaps between the Top Ekofisk and Top Sele in the Pierce area that defines three surfaces. The middle surface is interpreted to be the Base Lower Forties Sandstone surface while the upper surface is the Base Upper Forties Sandstone surface.

Seismic surfaces

Lower Forties Sandstone

Past evaluation of the Pierce seismic volumes by previous Shell staff have identified the main sediment fairways and general relationships between the Forties Sandstone deposits and the sub-regional basin setting primarily based on the Top Ekofisk to Top Sele isopach map along with supporting evidence from core and seismic amplitude extractions. Flattening the seismic volumes on the Top Sele surface reveals three onlap surfaces in the Pierce area (Figs 5 & 6). The lowest surface is in the Maureen/Lista interval and is not part of this study. The middle surface (Base Forties Sandstone) most likely represents the base of the Lower Forties Sandstone unit. The upper onlap surface (Mid Forties Sandstone) is the base of the Upper Forties Sandstone unit. Mapping the two upper onlap surfaces around the seismic cube provides the opportunity to further resolve the depositional settings and the sedimentation history of the Forties Sandstone in the Pierce area. While the Base Forties Sandstone and Mid Forties Sandstone surfaces (Figs 5 & 6) have been influenced by subsequent tectonic and salt movement, they still give an indication of the nature and amount of palaeo-topography at the onset of the Lower Forties Sandstone and Upper Forties Sandstone sedimentation respectively. The main features of the salt diapirs and sediment fairways can be identified. Isopach maps and amplitude extractions from these surfaces form the basis for many of the conclusions of this study.

In the Pierce area, sand distribution in the Lower Forties Sandstone is influenced by the existing sea floor topography at the top of the Lista Formation. Sediment pathways during deposition of the Lower Forties Sandstone, as shown by the thicker isochron trends between thinner isochron trends, come from the north of Pierce along the pathways indicated by the isopach map on Figure 1 and predominately run down the west side of the Pierce diapirs with some secondary pathways that went on the eastern side (Fig. 7). A Root Mean Square (RMS) amplitude extraction from the seismic data over the Lower Forties Sandstone shows that the seismic facies over the palaeo-highs vary from what is seen in the sediment pathways (Fig. 8). The palaeo-highs are defined by areas of chaotic higher amplitudes that can be contrasted with the lower amplitude, more continuous reflectors of the sediment fairways (Fig. 8). The palaeo-highs range in size from 200 m by 200 m to 2 km by 8 km and up to 100 m in height. While they appear random in their distribution, they generally show a NW to SE orientation, similar to the palaeo-current direction of the Forties Sandstone (Milstead 2007). The differing seismic facies, along with the onlap of seismic reflectors in the Lower Forties Sandstone interval against the top of the Lista Formation and the isopach trends show that most of the accommodation space during Lower Forties Sandstone deposition was defined by the existing sea floor topography


119

Fig. 7. Base Forties Sandstone surface (time) with Lower Forties Sandstone isochron drape showing areas of palaeo-topographic highs and sediment fairways through the Pierce area.

already present by the end of Lista Formation deposition. The Pierce salt diapirs appear to also have had an influence on the deposition of the Lower Forties Sandstone. Thinning of the seismic package containing the Forties Sandstone can be seen moving

from the sediment fairways towards the diapirs, sometimes with internal onlapping surfaces (Fig. 3). Although it is evident that the diapirs had sufficient seafloor expression to influence deposition of the Forties Sandstone, due to seismic illumination issues, salt overhangs and steep structural

Fig. 8. Lower Forties RMS amplitude map from the Base Forties to the Mid Forties (a) uninterrupted and (b) with interpreted palaeo-highs that appears to have influenced sand distribution.

120

E. D. SCOTT ET AL.

dips, it is difficult to determine the amount of thinning of the Forties Sandstone interval close to the diapirs. It is also difficult to gauge the full extent of the influence the diapirs had on the sea floor topography at that time. However, a well cross-section starting in the sediment fairway to the west of the diapirs and going up the flank of the South Pierce diapir shows significant thinning of the Lower and Upper Forties Sandstone (Fig. 9). While the overall thickness of the Forties Sandstone diminishes along the well cross-section, there is still reservoir quality sand deposited even in one of the most crestal well locations around the diapir suggesting a significant influence by the diapirs regarding accommodation space and sediment pathways but not to the extent of precluding sand deposition possibly even at the crest of any sea floor topography created by the rising diapirs. Influence of salt movement during the Forties Sandstone deposition can also be inferred from a seismic package that can be defined close to the

salt diapirs. A number of continuous seismic reflectors onlap onto a single surface that is in the area between and to the east of the two diapirs (Fig. 10). The seismic package’s lateral distribution and facies indicate accommodation space creation in this area and imply that underlying salt was moved from this locale to feed the rising diapirs and formed topographic lows that were subsequently filled by sediment gravity flows of the Forties Sandstone.

Depositional environments The sea floor topography during deposition of the Lower Forties Sandstone influenced the type of depositional environment in the Pierce area. The Lower Forties Sandstone thickness map shows the dominant sediment fairways come from the north and go past the west side of the salt diapirs (Fig. 11). Secondary pathways are present on the east side of the diapirs and are defined by slightly thinner zones in the interval (Fig. 11). The main

Fig. 9. Well cross-section showing the thinning of the Lower and Upper Forties Sandstone from the main sediment fairway on the west side of the Pierce Field going towards the South Pierce diapir.


121

Fig. 10. Seismic line illustrating the seismic facies found in between and to the east of the salt diapirs interpreted to be ponded sediments in accommodation space created by salt withdrawal. Map shows the distribution of the interpreted ponded seismic facies around the Pierce salt diapirs in the Lower Forties Sandstone.

sediment fairways to the west of the diapirs predominately received deposition from high density turbidity currents, with numerous interpreted amalgamated contacts, as seen in the 27/23-10 well and core (Fig. 12). These features are attributed to flows maintaining internal turbulence and high energy due to the corridors set up by the Lista Formation highs. As suggested by previous studies (Ahmadi et al. 2003; Birch & Haynes 2003; Hempton et al. 2005) and supported by the well information and the restricted nature of the sediment fairways, it is apparent that these corridors comprise a channelized depositional environment. With the restricted nature of the sediment fairways, successive channels had limited space to shift laterally and eroded into previous channel deposits resulting in amalgamated sand contacts, which increased connectivity vertically along the length of the fairway. While it is difficult to image clear individual channels in the Lower Forties Sandstone in the seismic data sets, there are potential candidates that can be interpreted as channel features. To the south of the South Pierce diapir the fairways become laterally more extensive. This would have resulted in sediment gravity flows expanding laterally, slowing down and depositing sediment in a shingled distributary complex. Between the two diapirs and to the east, the topography was different to the sediment fairways to the west due to the accommodation space created by the salt movement into the rising diapirs. The created accommodation space between the diapirs captured

sediment gravity flows coming down the main fairways and directed them into the salt withdrawal basin on the east side of the diapirs. The depositional environments in this setting, based on seismic facies and thickness maps, would develop deposits similar to a mini-basin fill and spill scenario (e.g. Prather et al. 1998; Badalini et al. 2000), first ponding sediments and then experiencing more bypass channelized sedimentation as the accommodation space was filled. Recent studies, based on core interpretation from the North Sea Central Graben in general and Pierce in particular, have recognized that a portion of the stratigraphy consists of hybrid flows (or linked debrites) (Haughton et al. 2003; Davis et al. 2009). In general, these deposits occur in sediment gravity flows where there is lower energy in the flow (the fringes or distal portions outside of the higher energy core of the flow) where the mud content becomes great enough to suppress turbulence and transform it into a debris flow. Hybrid flows are abundant in the Lower Forties Sandstone in wells on the east side of the Pierce diapirs while they are essentially absent in wells in the main sediment fairway on the west side of the diapirs (Davis et al. 2009). This supports the concept that sediment gravity flows were captured by the salt withdrawal minibasin, decreased in flow velocity and deposited sediment in a confined space. The muddier portions of these hybrid flows will introduce vertical baffles and/or barriers into the reservoir; however the

122

E. D. SCOTT ET AL.

Fig. 11. Map of the Lower Forties Sandstone depositional environments showing their distribution based on seismic and well data.

distribution of these low permeability portions of the deposits away from the wellbore and how they will affect fluid flow is not as well understood at this time.

facies from the sediment pathways (Fig. 8). The palaeo-highs are defined by areas of chaotic higher amplitudes that can be contrasted with the lower amplitude, more uniform reflectors of the sediment fairways. The Upper Forties Sandstone distribution shows similar patterns to the Lower Forties Sandstone and consists of the main and secondary sediment fairways between the topographic highs; however they are wider and more laterally extensive. It appears that the Lower Forties Sandstone filled in the lower portion of the available accommodation space allowing the Upper Forties Sandstone to be appreciably more widespread. There is little indication of ponding of sediments between or to the east of the diapirs, indicating that no accommodation space associated with salt withdrawal was created before deposition of the Upper Forties Sandstone. The well cross-section starting in the sediment fairway to the west of the diapirs and going up the flank of the North Pierce diapir again shows significant thinning of the Lower and Upper Forties Sandstone indicating influence of the rising diapirs on sedimentation (Fig. 14). Although the overall thickness of the Forties Sandstone diminishes along the well cross-section on the flank of the North Pierce diapir similar to the well cross section on the flank of the South Pierce diapir (Fig. 9), the proportional amount of thinning in the Upper Forties Sandstone is significantly less than what is observed in the Lower Forties Sandstone. This suggests that reservoir quality sand was deposited even at the crest of any sea floor topography created by the rising North Pierce diapir during deposition of the Upper Forties Sandstone. However, the cross section on the flanks of the South Pierce diapir shows that the Upper Forties Sandstone is absent implying that there was enough topographic relief to preclude sedimentation at the crest of the structure (Fig. 9).

Depositional environments Upper Forties Sandstone Similar to the Lower Forties Sandstone, sand distribution in the Upper Forties Sandstone was highly influenced by pre-existing sea floor topography and co-eval rising salt diapirs. While the sea floor topography at the onset of the Upper Forties Sandstone does not appear to be as extensive as during deposition of the Lower Forties Sandstone, the sediment fairways of the Upper Forties Sandstone generally follow along the same trends as those seen in the Lower Forties Sandstone. An RMS amplitude extraction from the seismic data over the Upper Forties Sandstone (Fig. 13) shows a similar seismic facies pattern to the Lower Forties Sandstone. Palaeo-highs show different seismic

The depositional environment in the Pierce area during deposition of the Upper Forties Sandstone continued to be influenced by the sea floor topography. Probable channels in the Upper Forties Sandstone are easier to image and identify in the seismic volumes than in the Lower Forties Sandstone (Fig. 15). Amalgamated blocky sandstones representing channels remain dominant in the Upper Forties Sandstone in the wider but still laterally restricted sediment fairway thicks that originates from the north and goes past the west side of the salt diapirs (Fig. 16). Channels in secondary pathways are present on the east side of the diapirs. They are defined by slightly thinner zones in the interval (Fig. 16) and became more established during Upper Forties deposition.


123

Fig. 12. Sands deposited from high density turbidity currents as seen in the 27/23-10 well located in the main sediment fairway on the west side of the South Pierce diapir.

The most significant difference between the depositional environments in the Upper Forties Sandstone to the Lower Forties Sandstone is the clearly imaged channelization on the east side of the diapirs. This channelization occurred during deposition of the Upper Forties Sandstone when there was ponding of sediments in a restricted salt withdrawal mini basin.

Geochemistry Multi-Dimensional Gas Chromatography (MDGC), a Shell proprietary geochemical fingerprinting technique, was used to evaluate reservoir connectivity based on similarity of dead crude oil from the

various exploration and production wells in the Pierce Field. The aromatics between C8 and C10 in an oil sample are separated and are displayed in ratios on a starplot to correct for any minor evaporation effects. Very small differences can be used to infer potential reservoir separation. In Pierce, six major MDGC groupings have been determined. The coloured bands along the well paths shown in Figure 17 highlight the perforated producing intervals with their corresponding colourcoded oil fingerprints. The colour-coding is based on statistical cluster analysis of the twelve ratios that are displayed in the starplot on Figure 18. The ratios consist of peak numbers that represent the aromatic compounds analysed by the MDGC (e.g. peak 1 is ethyl-benzene). The cluster diagram on

124

E. D. SCOTT ET AL.

Fig. 13. Upper Forties RMS amplitude map from the Mid Forties to the Upper Forties (a) uninterrupted and with (b) interpreted palaeo-highs that appear to have influenced sand distribution.

Figure 19 provides a statistical visualization of the similarity of oils to provide guidance as to how similar two oils are to each other. A cut-off value of 0.03 in the linkage distance has been established using calibrations of non-biodegraded oils in other North Sea oil fields. A linkage distance above 0.03 suggests reservoir separation while linkage distances below 0.03 suggest reservoir connectivity. Within each of the coloured groupings the similarities would suggest reservoir connectivity, but does not rule out reservoir compartmentalization. The geochemical fingerprint differences between the individual groupings from South and North Pierce are large enough to suggest that they represent separate oil compartments, which could be the result of structural or stratigraphic baffles or barriers. In addition, the fingerprint from well 23/ 22A-A7Z, which is located in the saddle between the two salt domes, is even more different than all other fingerprints in South and North Pierce. This strongly suggests that there is no connection between the 23/22A-A7Z and 23/22A-A5 well, which seems to be supported by the very limited production from this well and the lack of pressure support to South Pierce after the well was turned into a water injector. Because the API gravity is lower (348 v. 378) and represents the ‘end-member’ on the starplot (Fig. 18), the A7Z oil could represent

the earliest charge of lower maturity oil into the Pierce field. In South Pierce, another important observation can be made. The geochemical signature of the oil in the toe of the 23/27-B2 well does not match the production sample higher up in the wellbore suggesting that the toe section of the well is in a separate compartment (Fig. 20). Samples for geochemical analysis have been collected from Pierce through time to enable time-lapse geochemistry studies. Samples from South Pierce did not undergo any significant changes through time. Samples from North Pierce, however, alter slightly through time, potentially indicating fluid movement along some preferential stratigraphic pathways. The starplot in Figure 21 shows how the geochemical fingerprint of oils from the 23/ 22A-A3X and 23/22A-A6 wells have changed from early production (gold lines) to more recent production (red lines). The most recent well at North Pierce, 23/22a-A9, exhibits a geochemical fingerprint close to the early production from the 23/22A-A3X and 23/22A-A6 wells suggesting that this is the geochemical signature of the original oil in the vicinity of the 23/22A-A3X, 23/22A-A6 and 23/22a-A9 wells in North Pierce. In other words, because the shape of the 23/22a-A9 star is closest to the most extreme 23/22A-A7Z star


125

Fig. 14. Well cross-section showing the thinning of the Upper Forties Sandstone from the main sediment fairway on the west side of the Pierce Field going towards the North Pierce diapir is proportionately less than occurs in the Lower Forties Sandstone (see Fig. 9).

(Fig. 18), which is believed to represent the early charge into the reservoir, this 23/22a-A9 fingerprint can be regarded as being representative for the original fingerprint of oil in North Pierce prior to being affected by longer-term production as is observed for 23/22A-A3X and 23/22A-A6. This would suggest that over production time, the contribution of oil from the original drainage area in 23/ 22A-A3X and 23/22A-A6 is decreasing and that recent production is more similar to oil from the more up dip 23/22A-A2Z exploration well. This becomes apparent when zooming in on the shape of the more recent 23/22A-A3X and 23/22A-A6 stars that are getting closer to the 23/22A-A2Z star (purple lines on Fig. 18).

Fault seal potential The Pierce Field lies in the immediate footwall to a major regional NNW–SSE-trending basement

fault, which defines the eastern boundary of the Central Graben of the North Sea. The fault is segmented at a regional scale, such that the Pierce diapirs are situated in the footwall ‘corner’ to an orthogonal jog in the fault trace (which trends NNW –SSE west of the diapirs and ENE –WSW to the south of them). This is a common structural setting for salt diapirs within obliquely rifted basins (Dooley et al. 2005, and references therein). Significant thickening of the Paleocene sequence across this basement structure indicates that it was active during the accumulation of the Forties Sandstone, with depositional accommodation space in the subsiding hanging wall to the south and west. Intra-reservoir faulting is dominated by similar orthogonal NNW- and ENE- ‘basement’ fault trends (Fig. 22); with smaller (near seismic throw resolution) more discontinuous, radial fault arrays (Figs 20 & 21). The radial faults are spatially related to polygonal faulting, developed within shale rich packages in the Tertiary stratigraphy.

126

E. D. SCOTT ET AL.

Fig. 15. Upper Forties RMS amplitude map 42 mss below the Top Sele surface (a) uninterrupted and (b) with interpreted channels running NE to SW through the Pierce area.

The orthogonal ‘basement’ trending faults are interpreted in seismic data to pass downwards into the Mesozoic section, where expansion of reflectors into their hanging walls indicates syn-rift growth of the sediments. The north and south Pierce diapirs are centred on the deformed footwalls to two of these ENE –WSW faults, close to their intersection with NNW–SSE faults of similar age. Thus, in common with other salt structures in the Central North Sea, the Pierce salt diapirs are rooted to a Permian Zechstein Group source, through a late Jurassic fault system, and are fringed by smaller radial and polygonal fault arrays within the upper structural levels, related to subsequent passive down-building of the Tertiary sediments around the rising salt plugs (Stewart 2007). Movement of salt, principally from the hanging wall into the footwall of the ENE –WSW Mesozoic faults, would have caused reactivation and upward propagation of the fault tips into the Tertiary section; but also up-doming and de-activation of these fault planes, and progressive development of a ‘horse-tail’ array of faults in the higher structural levels (cf. illustration and analysis of Pierce fault geometries by

Davison et al. 2000). In summary, the reservoir sands are impacted by small discontinuous radial faults that have limited connectivity and do not fully isolate compartments; and elements of a more orthogonal, connected pattern of larger faults inherited from reworked Mesozoic fault systems beneath the Forties. The nature of compartmentalization in the Pierce Field has been a subject of debate since the early days of field appraisal and development. As summarized by Dennis et al. (2005) and Fisher & Jolley (2007), ideas have evolved over the years, with different groups of workers explaining the variation of pressures and fluid contacts by sealing faults, stratigraphic compartments, or hydrodynamic tilting of the oil – water contact. So do the faults in the Pierce Field seal and isolate compartments? Hempton et al. (2005) indicated that sealing of faults in neighbouring analogue fields was difficult to justify – whereas sealing of a (more complex) radial fault pattern on Pierce might explain compartmentalization of the field. At the time, they were reporting a preliminary fault seal study, which used the Shale Gouge Ratio algorithm (SGR,


127

Fig. 18. Six major groupings of oils in the Pierce Field based on ratios of hydrocarbon elements.

Yielding et al. 1997) to estimate the flow retarding clay content of the faults. These early results suggested that the faults might seal. However, the analysis was conducted on a structural framework model in which the radial fault throws were much larger than the small throws observed in the seismic data; and it used a very shale-rich Vshale scenario. Taken together, these erroneous inputs to the SGR analysis over-predict fault clay content and imply that the faults are more sealing than they are likely to be. Results from subsequent infill drilling, revision of sedimentological concepts Fig. 16. Map of the Upper Forties Sandstone depositional environments showing their distribution based on seismic and well data.

Fig. 17. Six major groupings determined by MDGC geochemical fingerprinting (see Fig. 19) are highlighted by the coloured bands along the well paths.

Fig. 19. Cluster analysis of 58 MDGC oil fingerprints at Pierce to determine the similarity of oils. A linkage distance above 0.03 suggests reservoir separation while linkage distances below 0.03 suggest reservoir connectivity. Colour groupings are displayed along well paths in Figure 17.

128

E. D. SCOTT ET AL.

Fig. 20. The oil sample from the toe-only section of the 23/27B-2 well does not match the production sample coming from the along the length of the horizontal well suggesting a separate compartment in the toe section of the well.

and updated geological modelling were less supportive of fault compartmentalization, and more indicative of either stratigraphic controls or a hydrodynamically tilted petroleum contact. Indeed, incorporation of more ‘realistic’ weakly baffling permeabilities into the faults in fluid flow

simulation models at the time provided far better matches between simulated production history and actual production data than models containing explicitly sealed faults (Al-Busafi 2005). Pressure data collected from the wells at Pierce indicates only a minor control on fluid flow from faulting.

Fig. 21. Samples have been collected through time for time-lapse geochemistry. South Pierce oils did not change through time whereas North Pierce fingerprints show variations that indicate fluid movement along stratigraphic pathways from north to south.


129

Fig. 22. Interpreted fault pattern around North and South Pierce. The salt utilized the weak zones inherited from the regional structural trends. Only two sealing faults have been identified at the Pierce field defining a structural compartment in the SE quadrant of South Pierce.

Good pressure communication is seen during production around all of the North Pierce wells and also communication between the South Pierce wells to the north and west. Only the SE quadrant of South Pierce has shown any evidence of fault compartmentalization. Data from the 29/22A-A1 and 23/27-B2 wells indicate probable sealing faults toward the toes of these two penetrations, corresponding to the mapped position of two large intersecting NNW –SSE and ENE –WSW faults that pass downwards into underlying Mesozoic basement structures (Fig. 22).

Sedimentological control on fluid flow Reservoir connectivity between the 23/27-B3 and the up-dip 23/27-B2 producer To understand the flow of fluids in the Pierce field, prior work involved construction of a static model which was based on facies interpretation of the

logs calibrated to core analysis and structural/ stratigraphic interpretations from the seismic data. Based on the static model, dynamic modelling was undertaken to history match the produced hydrocarbons from the Pierce Field. In the dynamic model, on the southern side of the Pierce Field the 23/27-B2 well produced more water and experienced water breakthrough earlier than the actual production history. The apparent cause for this early water breakthrough and greater produced water can be explained by the distribution of the sedimentary facies between the producing 23/ 27-B2 well and the associated 23/27-B3 water injection well (Fig. 23). Both wells are interpreted to contain amalgamated channel sands that are present through the Upper and Lower Forties Sandstone. The static model honours these logs but also distributes the amalgamated sand facies away from the wellbore such that high permeability sands in the 23/27-B2 producer are in direct communication with sands in the water leg and the 23/27-B3 injector. In addition to not attaining an

130

E. D. SCOTT ET AL.

Fig. 23. View from a Petrel model showing the Forties Sandstone section between the 23/27B-2 and 23/27B-3 wells. Distribution of the interpreted amalgamated channel sand facies in the wells not relying on a geologically accurate model resulted in the two wells being connected not reflecting the production history of the wells.

accurate history match for the 23/27-B2 well, tracers added to the injected water in the 23/ 27-B3 injector were not seen in the up-dip 23/ 27-B2 producer further indicating that the static model did not accurately reflect the subsurface architecture. Interpreted channels from seismic data (Fig. 16) are present in this area. The edge of the seismically defined channel lies between the 23/27-B3 injection well and 23/27-B2 production well (Fig. 24) suggesting the presence of a geological boundary. The 23/27-B3 injector well is situated in this

Fig. 24. The 23/27B-3 injector is in a main channel fairway that wraps around South Pierce to the west and has limited sand-body connectivity to the 23/27B-2 producer due to the change in sedimentary facies from channel axis to channel margin. Another option is that the 23/27B-2 well is in a separate channel axis that is not connected to the 23/27B-3 channel axis.

channel that comes along the west side and wraps around the SE quadrant of the South Pierce diapir before turning south. The 23/27-B2 production well lies outside of the seismically define channel fairway. Definition of the exact lateral extent of the channel from the available well and seismic data is difficult. This potential boundary, along with the dynamic data indicating a barrier to flow between the two wells, may represent a change in sedimentary facies from channel axis to channel margin that has limited sand-body connectivity and inhibits communication between the two wells. It may also be the case that the 23/27-B2 well is in a channel system separate from the 23/27-B3 well that is isolated by a shale drape on the base of the channel or some other impermeable layer.

Aquifer support Over production time the Pierce Field, in general, has seen stronger aquifer support on the west side of the diapirs than on the east side. With the main sediment fairway running north to south down the western side of the field there is good connectivity of the amalgamated sands allowing water to readily flow into the Pierce structure to provide aquifer support (Fig. 25). In particular, the southwest portion of the field should see excellent connectivity down-dip following the sediment fairway on the west side into the sedimentary distributary complex to the south of the diapirs. The wells on the east side of the diapirs are linked to thinner secondary sediment pathways and the laterally constrained sands deposited in the salt withdrawal mini-basin. While the connectivity of the sands on the eastern side may be similar to the west side, the limited extent of these sands will result in weaker pressure support from a limited aquifer. Connecting the sandstones in the wells down dip to a strong aquifer will result in early water breakthrough in a dynamic reservoir model, particularly on the east side of the field. Incorporating the channelized nature of the deposits with stratigraphically defined flow paths will result in delayed water breakthrough and better match well performance. The 509 m of difference from the shallowest oil –water contact in the south to the deepest oil – water contact in the north has been attributed to a hydrodynamically tilted oil–water contact (Dennis et al. 2005) or else some form of stepped, structurally compartmentalized contact. As shown by pressure and geochemical data, there is only one area (the SE quadrant of South Pierce) that can be interpreted to be structurally compartmentalized from the rest of the field. While it is still possible to have a stepped oil –water contact due to faulting of some type at Pierce, it is difficult to determine


Fig. 25. Sediment ponding of the Lower Forties Sandstone in the salt withdrawal mini-basin limits the extent and strength of the aquifer that is connected to the wells on the east side of Pierce. Along the sediment fairways from the south there is strong aquifer support.

the nature or the cause due to the poor seismic data close to the diapirs. Alternatively, the aquifer drive coming from the sedimentary distributary complex to the south of the diapirs appears to be significantly greater than the aquifer drive coming from the east or the north of Pierce, which could drive the oil– water contact higher in the south. At this time it appears that the main cause for the difference seen in the oil–water contacts across the field is driven by the dominant southern aquifer and focused by the Forties depositional architecture.

Summary Fluid flow in deep water turbidite reservoirs will, in part, reflect the sedimentary architecture of the

131

reservoir. At the Pierce Field, it appears that structural control on fluid flow only has a minimal impact on the flow of fluids. Only in the SE quadrant of South Pierce do we see pressure and geochemical differences that indicate reservoir compartmentalization associated with faulting. Instead, the depositional architecture of the Forties Sandstone appears to be the primary control on fluid flow. The channelized depositional environments at Pierce, particularly on the west side of the diapirs, dictated the location of the high permeable trends in the field along which any fluid (both hydrocarbons and water) will preferentially flow. These preferential flow paths dominate over structural control as can be seen from the example of the injected water from the 23/27-B3 well which does not travel up-dip to the 23/27-B2 producer, but instead most likely travels along the channelized sediment pathway and wraps around the South Pierce diapir towards the NW. The variable oil – water contact at Pierce appears to be the result of a strong aquifer drive from the south that exploits the same flow path. Understanding the depositional history and three-dimensional architecture of the sandstones and mudstones is integral in deciphering the production history of the field. While the wells encountered numerous amalgamated sandstones, determining the lateral boundaries of the sand bodies and the candidate locations for fluid flow barriers is necessary to understand the history of the production of the field. The geochemical studies of the oils at Pierce were valuable in not only confirming the structural control of the reservoir compartmentalization of the SE quadrant of South Pierce but also indicating the movement of hydrocarbons that are following the sand bodies around North Pierce. The authors would like to thank the past and present members of the Pierce Asset Team and the Geological and Geophysical Service Teams, particularly John Marshall, at Shell Aberdeen for all of their efforts that this study was based on. We also thank the reviewers, Trey Meckel, Paul Ventris and Bruce Ainsworth for their comments and suggestions that greatly improved the paper.

References Ahmadi, Z. M., Sawyers, M., Kenyon-Roberts, S., Stanworth, C. W., Kugler, K. A., Kristensen, J. & Fugelli, E. M. G. 2003. Paleocene. In: Evans, D., Graham, C., Armour, A. & Bathurst, P. (eds) The Millennium Atlas: Petroleum Geology of the Central and Northern North Sea. The Geological Society, London, 235– 259. Al-Busafi, B. 2005. Incorporation of fault rock properties into production simulation models. PhD thesis, University of Leeds, UK.

132

E. D. SCOTT ET AL.

Badalini, G., Kneller, B. & Winker, C. D. 2000. Architecture and processes in the late Pleistocene Brazos-Trinity turbidite system, Gulf of Mexico continental slope. In: Weimer, P., Slatt, R. M., Coleman, J., Rosen, N. C., Nelson, H., Bouma, A. H., Styzen, M. J. & Lawrence, D. T. (eds) Deep-water Reservoirs of the World. GCSSEPM 20th Research Conference, Houston, 16–34. Berggren, W. A., Kent, D. V., Swisher, C. C. & Aubry, M. P. 1995. A revised Cenozoic geochronology and chronostratigraphy. In: Berggren, A. T., Kent, D. V., Aubry, M. P. & Hardenbol, J. (eds) Geochronology, Time Scales and Stratigraphic Correlation. SEPM Special Publication, 54, 129– 212. Birch, P. & Haynes, J. 2003. The Pierce Field, Blocks 23/ 22a, 23/27, UK North Sea. In: Gluyas, J. G. & Hichens, H. M. (eds) United Kingdom Oil and Gas Fields, Commemorative Millennium Volume. Geological Society, London, Memoir, 20, 647–659. Bowman, M. B. J. 1998. Cenozoic. In: Glennie, K. W. (ed.) Petroleum Geology of the North Sea; Basic Concepts and Recent Advances. 4th edn. Blackwell Science, Oxford, 350– 375. Davis, C., Haughton, P., McCaffrey, W., Scott, E., Hogg, N. & Kitching, D. 2009. Character and distribution of hybrid sediment gravity flow deposits from the outer Forties Fan, Paleocene Central North Sea UKCS. Marine and Petroleum Geology, 26, 1919–1939, doi: 10.1016/j.marpetgeo.2009.02.015. Davison, I., Alsop, I. et al. 2000. Geometry and latestage structural evolution of Central Graben salt diapirs, North Sea. Marine and Petroleum Geology, 17, 499–522. Den Hartog Jager, D., Giles, M. R. & Griffiths, G. R. 1993. Evolution of Palaeogene submarine fans of the North Sea in space and time. In: Parker, J. R. (ed.) Petroleum Geology of Northwest Europe. Proceedings of the 4th Conference. Geological Society, London, 59–71. Dennis, H., Bergmo, P. & Holt, T. 2005. Tilted oil – water contacts: modeling the effects of aquifer heterogeneity. In: Dore´, A. G. & Vining, B. A. (eds) Petroleum Geology: North– West Europe and Global Perspectives—Proceedings of the 6th Petroleum Geology Conference. Geological Society, London, 145– 158. Dooley, T., McClay, K. R., Hempton, M. & Smit, D. 2005. Salt tectonics above complex basement extensional fault systems: results from analogue modeling. In: Dore´, A. G. & Vining, B. A. (eds) Petroleum Geology: North– West Europe and Global Perspectives—Proceedings of the 6th Petroleum Geology Conference. Geological Society, London, 1631–1648. Fisher, Q. J. & Jolley, S. J. 2007. Treatment of faults in production simulation models. In: Jolley, S. J., Barr,

D., Walsh, J. J. & Knipe, R. J. (eds) Structurally Complex Reservoirs. Geological Society, London, Special Publications, 292, 219– 234. Haughton, P. D. W., Barker, S. P. & McCaffrey, W. D. 2003. ‘Linked’ debrites in sand-rich turbidite systems – origin and significance. Sedimentology, 50, 459– 482. Hempton, M., Marshall, J., Sadler, S., Hogg, N., Charles, R. & Harvey, C. 2005. Turbidite reservoirs of the Sele Formation, Central North Sea: geological challenges for improving production. In: Dore´, A. G. & Vining, B. A. (eds) Petroleum Geology: North– West Europe and Global Perspectives, Proceedings of the 6th Petroleum Geology Conference. Geological Society, London, 449–459. Jennette, D. C., Garfield, T. R., Mohrig, D. C. & Cayley, G. T. 2000. The interaction of shelf accommodation, sediment supply and sea level in controlling the facies, architecture and sequence stacking patterns of the Tay and Forties Sandstone/Sele basinfloor fan, Central North Sea. In: Weimer, P., Nelson, H., Slatt, R. M., Bouma, A. H., Coleman, J., Styzen, M. J., Rosen, N. C. & Lawrence, D. T. (eds) Deep Water Reservoirs of the World. GCSSEPM Foundation Annual Research Conference. GCSSEPM, Houston, Texas, 402–421. Milstead, D. P. 2007. The geology of the Maureen and Lista Formations and their influence on Forties Fan deposition in the Pierce Field. MSc thesis, Imperial College London. Prather, B. E., Booth, J. R., Steffens, G. S. & Craig, P. A. 1998. Classification, lithologic calibration, and stratigraphic succession of seismic facies of intraslope basins, deep-water Gulf of Mexico, American Association of Petroleum Geologists Bulletin, 82, 701– 728. Stewart, S. A. 2007. Salt tectonics in the North Sea Basin: a structural style template for seismic interpreters. In: Ries, A. C., Butler, R. W. H. & Graham, R. H. (eds) Deformation of the Continental Crust: The Legacy of Mike Coward. Geological Society, London, Special Publications, 272, 361– 396. Whyatt, M., Bowen, J. M. & Rhodes, D. N. 1991. Nelson – successful application of a development geoseismic model in North Sea exploration. First Break, 9, 265–280. Wills, J. M. & Peattie, D. K. 1990. The Forties Sandstone Field and the evolution of a reservoir management strategy. In: Buller, A. T., Berg, E., Hjelmeland, O., Kleppe, J., Torsaeter, O. & Aasen, J. O. (eds) North Sea Oil & Gas Reservoirs – II. Norwegian Institute of Technology. Graham and Trotman, London, 1– 23. Yielding, G., Freeman, B. & Needham, D. T. 1997. Quantitative fault seal prediction. American Association of Petroleum Geologists Bulletin, 81, 897– 917.

Integrated approach to geomodelling and dynamic simulation in a complex mixed siliciclastic – carbonate reservoir, N’Kossa field, Offshore Congo J. P. WONHAM1*, M. CYROT1, T. NGUYEN1, J. LOUHOUAMOU1 & O. RUAU2 1

Total E&P, 2 place de la Coupole, La De´fense 6, 92078 Paris 2

Total E&P Congo, Pointe Noire, Congo

*Corresponding author (e-mail: [email protected]) Abstract: The Lower Sendji Carbonate (Albian age) of the N’Kossa field is a mixed siliciclastic– carbonate reservoir exhibiting very heterogeneous reservoir properties and the development of extensive vertical flow permeability barriers. The reservoir is a succession of interstratified dolomite, limestone, sandstone and shale lithologies of sabkha, tidal flat and lagoonal origin. Early, synsedimentary dolomite cements are extensively developed, particularly over local palaeohighs. Challenges to hydrocarbon production include: (1) the vertical variability of reservoir properties; (2) the presence of many, laterally extensive vertical flow barriers; and (3) variable connectivity between fault-bounded compartments. This complexity was underestimated at the appraisal stage, the initial development plan calling for a relatively simple scheme of pressure support to the critical reservoir fluid via injection into the gas cap and the water leg. This decision was supported by the identification of a single hydrocarbon column (.400 m thick) over the entire structure which suggested a lack of vertical compartmentalization. Recent studies, which include a geological synthesis of more than forty wells and a dynamic data synthesis to determine the production mechanism and identify key heterogeneities, show a lack of pressure support to the oil leg from both water and gas injector wells. Today, integration of static and dynamic data together with an improved geological understanding of the stratigraphic control on vertical flow barriers and mapping of high permeability sandstone layers is the key to identifying unswept zones of the field and future infill drilling targets.

The N’Kossa field is located in the Republic of Congo’s Haute Mer permit, 60 km off the West African Atlantic margin (Fig. 1). The field was one of the largest oil discoveries in West Africa of the 1980s (Cameron et al. 1999; Cameron & White 1999). Situated on the modern shelf slope between 170 and 350 m water depth, it was one of the first deep offshore fields to be developed in a region lately characterized by very high hydrocarbon exploration activity. USGS estimates (Brownfield & Charpentier 2006) indicate that around 30 billion barrels of oil in place were discovered in the deep water offshore zone of central West Africa in the last ten years, approximately doubling cumulative oil volume in the region. The N’Kossa field represents a particularly prolific type of hydrocarbon play found in the region: the listric faulted and rafted blocks with Albian age carbonate reservoir. These rafted blocks developed during the Late Cretaceous to Tertiary thin-skinned extension of the margin. The field was discovered in 1984. After appraisal, volumes were estimated at 900 million barrels of oil and condensate in place. The principal

reservoir is the Sendji Carbonate. This is an Albian-age, mixed siliciclastic –carbonate formation consisting of interbedded dolomite, limestone, sandstone and shale. The reservoir is several hundred metres thick. Reservoir layer thickness varies laterally due to synsedimentary tectonism. Following deposition, the reservoir rocks underwent a phase of major extension during the Cenomanian that was driven by gravity tectonics and sediment loading. The N’Kossa rafted fault-block structure (Fig. 2) moved down the margin slope on an underlying layer of Aptian-age Loeme Salt for a distance of 25 km. Drilling appraisal of the N’Kossa field indicated the presence of a single hydrocarbon column with a water –oil contact at 23425 m TVDSS and a gas– oil contact at 23283 m TVDSS over the entire structure (Fig. 3). The hydrocarbon column of the field is over 400 m thick and consists of a 38–418 API oil leg (140 m thick column) with an overlying condensate transition zone (40 m thick column) and a thick, critical fluid gas cap. Pressure data collected from all wells conforms to this fluid gradient with minor variation interpreted to relate to pressure


134

J. P. WONHAM ET AL.

Fig. 1. Location of N’Kossa Field, Offshore Republic of Congo.

Fig. 2. Seismic inline (time section) crossing the N’Kossa raft structure at NKSM-3 well location. The Albian age Sendji Limestone reservoir (outlined in red) is completely surrounded by Albo-Cenomanian marls and major listric structures are seen both to the east and west of the field. Extensive structural deformation above the field reduces seismic image quality at the reservoir level. Section location is shown in Figure 1.

RESERVOIR DYNAMICS, N’KOSSA FIELD

135

Fig. 3. Pressure data from wells in different sectors of the N’Kossa field conforms to a common hydrocarbon column. GOC, gas– oil contact; OWC, oil –water contact.

gauge and depth inaccuracies. Natural aquifer support is limited due to the situation of the field on a rafted fault block with a surface area of 20 km2. The reservoirs identified by appraisal drilling were thick dolomitic units and thinner sandstone layers. Both were considered to be laterally extensive and tabular. Many thin beds of micritic limestone and shale were identified from the core and identified as potential barriers to vertical fluid flow. However, these were considered to be discontinuous and likely to be offset by faults. The base case development concept for the field was that it would produce as a single, reasonably well connected reservoir. It was planned that reservoir pressure would be maintained by gas injection in the gas cap and water injection in the aquifer. The field came on stream in 1996. After a short plateau, production went quickly into decline. Eighteen months after start-up of the field, formation pressure data from infill drilling wells showed that several sandstone reservoir levels, each a few metres thick, had become strongly depleted and that surrounding dolomite layers were only poorly depleted. The initial development plan based on pressure support by gas and water injection was not suited to the multi-layer behaviour of the field. In particular, the layers into which water

was being injected proved to be very poorly connected to the overlying producing layers with a consequent lack of pressure support. Vertical flow barriers between injector and producer layers were much more laterally continuous and hence effective than had been prognosed. In 2004, after 8 years of production, challenges to the management of the field were multiple and included in particular: (1) the difficulties of reservoir lithology characterization and hence porosity and permeability distribution; (2) the difficulties of well integration and correlation, notably where inclined wells passed through multiple faults leading to missing and repeat stratigraphy; (3) the need to hierarchize vertical flow barriers of different kinds before defining a flow unit framework; and (4) the need to understand fluid and pressure distribution in the context of unexplained phenomena related to complex lateral connection between faultbounded compartments. The production mechanism was not well understood and there remained major uncertainties about the degree to which water had swept the reservoir, the level of pressure maintenance and the location of remaining hydrocarbons for targeting infill drilling. An integrated geoscience and dynamic synthesis was therefore initiated to improve the reservoir

136

J. P. WONHAM ET AL.

characterization and suggest a way forward for the future development of the field. The results of this synthesis form the basis of this paper. The synthesis aimed to better understand recovery mechanisms and to close the loop between the static and dynamic models. Following its completion in 2008, the reservoir management plan has been adapted to target remaining untapped potential of the field. Today, new well drilling is based on an enhanced understanding of the field and a more predictive dynamic model. This paper shows the importance of not underestimating the extent and effectiveness of vertical flow barriers in the Albian mixed carbonate– siliciclastic reservoirs of offshore West Africa. These reservoirs form a very extensive play type in both the Congolese and Angolan offshore zones. Both the tectono-sedimentary controls on dolomite distribution and the processes controlling the formation of barriers to vertical flow described by this paper should be widely applicable in the region.

Data available This study was carried out by a multi-disciplinary team of geophysicists, geologists’, petrophysicists, production engineers and reservoir engineers. The team aimed to integrate all available data on the field ranging from 3D seismic to well and core data and all available pressure, flow and production data. Data was incorporated in an iterative fashion: initially to build a geomodel and flow unit framework, and later to focus on particular problems of dynamic history matching. The 3D seismic data utilized was acquired in 1993 and reprocessed using post-stack depth migration in 1996. The dominant frequency of the seismic is typically around 25 Hz. Sonic velocity in the Albian carbonates is c. 4040 m s21. Consequently, vertical resolution at reservoir levels is 40–80 m leaving limited scope for detailed seismic stratigraphic analysis. However, interpretation of this dataset provides both a regional understanding of structural evolution and a framework for reservoir model building. Regionally picked horizons were: Base Oligocene, Santonian Discordance, Intra-Cenomanian, Top Albian, Top Loeme Salt and Base Loeme Salt horizons. The top of the Lower Sendji Carbonate is marked by the Top S4D2 horizon, the base by the Top Loeme Salt. There are three intra-reservoir seismic horizons: Top S2F3, Top S3C2 and Top S2B5 (see later for stratigraphic definition). The seismic data shows clearly the rafted character of the Albian reservoir (Fig. 2) and provides a structural framework for reservoir model building.

Normal faults with throws of 20 m have been recognized from well data but not on seismic. Faults with throws greater than 30 m can be mapped based on seismic (Figs 3 & 4). Thickness mapping of major seismically-defined reservoir layers also provides useful insights into structural controls on lateral variability of reservoir quality. Direct use of seismic attributes for reservoir characterization is limited due to: (1) the reservoir depth of 3000– 3500 m; and (2) the presence of a large number of structural features in the interval overlying the field (Fig. 2). Forty-three wells (locations shown in Fig. 4) were available for this study with logs of GR, neutron porosity, formation density, photo-electric factor (PEF) and resistivity. PEF and formation density logs are particularly important for lithological quantification. Identified lithologies were calibrated to core description and core mineralogy based on X-ray diffraction data. Five wells provide a total of 1650 m of core used for sedimentological studies and sedimentary sequence identification. Cores were plugged at 30 cm intervals to provide a large phi–K database for lithofacies characterization. A number of wells were strongly deviated, sub-horizontal wells. A few sub-horizontal wells passed both downwards and upwards through the stratigraphy requiring a close cooperation between geophysicist and geologist to identify correlatable stratigraphic patterns and traversed fault locations. An extensive dynamic database has been analysed. Formation pressure logging was carried out in twelve wells prior to start-up and 15 wells post-start-up. Production logging tools (PLTs) were run in nine wells pre-start-up and in twentyfour wells post-start-up. Eight wells had down hole pressure gauges which in some cases were able to test for well start-up interference. All the production wells had monthly production history and production test data. Interpreted results of twentyfour drill stems tests (DSTs) carried out in seventeen wells on a range of different flow units were available and, in some cases, have demonstrated the presence of no-flow boundaries in the reservoir. Applications of these various data will be discussed in more detail below.

Regional context of the Sendji Carbonate During the Albian to Cenomanian, the N’Kossa field was situated in a zone affected by extension due to the opening of the Atlantic Ocean. Thick salt deposits accumulated during the Aptian. The original thickness of the salt is uncertain. However, correlation with the onshore zone suggests that it was originally c. 500–700 m thick. The Lower to Middle Albian is dominated by shelf and littoral


137

Fig. 4. Location map of N’Kossa wells. Numbers beside the wells indicate well impact depth of surface Top S2B5 in metres TVDSS. Location of north– south and west– east aligned sections shown in Figures 9 and 10 are indicated in red. Figure 22 correlation shown in purple.

deposition, following the creation of a narrow seaway between the African and South American continental plates (Coward et al. 1999). The principal reservoir of the N’Kossa field and its satellite N’Kossa South is the Sendji Carbonate. It was deposited in shallow marine and littoral environments during the Lower to Middle Albian. These deposits are 600 –800 m thick. The Sendji Carbonate is age equivalent to the Pinda Group of the neighbouring Angolan offshore zone (Coward et al. 1999; Eichenseer et al. 1999; Valle et al.

2001). It is an important reservoir in several fields of the Congo offshore zone (Baudouy & Legorjus 1991; Van Horn 2001). Limestone and dolomite are the dominant lithologies. Sandstone and shale are also present and in some stratigraphic intervals, notably at the top of the Sendji Carbonate, they are the dominant lithologies. Source rocks are plentiful in the underlying Neocomian –Barremian syn-rift succession (Fig. 6). The lacustrine Pointe Noire Marl is probably the most prolific source rock but the Sialivakou Shale

138

J. P. WONHAM ET AL.

Fig. 5. Map showing extension of Albian Limestone (blue) creating raft structures across the Congo offshore zone.

RESERVOIR DYNAMICS, N’KOSSA FIELD Fig. 6. Regional setting of N’Kossa Field showing extensional and compressional domains created by thin-skinned extension, sediment loading and gravitational deformation of the Congolese margin. Location of this section shown by red line in Figure 5. N’Kossa field indicated by red arrow.

139

140

J. P. WONHAM ET AL.

and the Djeno Sandstone may also contribute (Baudouy & Legorjus 1991).

Structural setting The Loeme Salt played a fundamental role in the development of structures produced by gravitational tectonics. The N’Kossa field is located in a zone that experienced rapid thin-skinned extension during the mid-Albian to early Miocene. Up to 60–70 km of extension (Raillard et al. 1998) led to the development of rafts of Albian limestone surrounded by deposits of Cenomanian-age, typically siltstones and calcilutites (Figs 5 & 6). The rafting concept was first introduced by Burollet (1975) and has been mainly studied with reference to the neighbouring Angolan offshore zone (Fort et al. 2004). The N’Kossa field is a rafted structure that was displaced by some 25 km seaward with an associated element of clockwise rotation (Rouby et al. 2002, 2003). The structural development of the N’Kossa raft is considered to have progressed in three main steps: (1) passive subsidence influenced by the formation of salt diapirs; (2) local raft grounding and onset of extension on listric faults during the Aptian and before the end of Sendji Carbonate deposition; and (3) accelerated extension

during the Cenomanian leading to raft structure development (Fig. 7). The first and second phases had an influence on the sedimentology and early diagenesis of the Sendji Carbonate reservoir. This will be discussed below.

Overview of Sendji Carbonate stratigraphy The Sendji Carbonate can be subdivided into six major composite Transgressive– Regressive (T–R) sequences (S0 –S5), each characterized by lateral thickness changes and ranging from 10s of metres up to 200 m thick (Fig. 8). S0 –S5 are composed of two orders of smaller scale sequence. The smaller of these are high-order T –R sequences ranging from a few metres up to 15 m thick. More details of high resolution sequence development are discussed later after description of the facies of which they are composed. Regressive or transgressive trends in high order sequences identify midorder T –R sequences (around 40 m thick). All three orders of T–R sequence have been defined in core using facies stacking patterns and key stratigraphic surface identification. High order sequences can be correlated across the field with a good degree of certainty using lithological variation based on core-calibrated petrophysical log interpretation (Figs 9 & 10).

Fig. 7. Schematic structural cross-section (no vertical exaggeration) showing development of N’Kossa Field.


141

Fig. 8. Lower Sendji Carbonate (Albian) showing sequence hierarchy and large-scale composite sequences with principal bounding surfaces. MP/TS, Maximum Progradation/Transgressive Surface; MFS, Maximum Flooding Surface.

The use of the term ‘sequence’ is not intended to imply a sequence stratigraphic approach as described by Van Wagoner et al. (1990) where sequences made up specifically of lowstand, highstand and transgressive systems tracts are defined. In the current context, the use of the term ‘sequence’ is more analogous to the simpler sequence stratigraphic approach of Embry & Johannessen (1993) where transgressive and regressive systems tracts are identified. This reflects the need for a relatively straightforward stratigraphic approach to the description of such a thick and variable stratigraphic succession. The complete Sendji Carbonate succession shows long-term depositional changes probably related to the opening of the early Atlantic Ocean and a progressively less restricted depositional setting. The composite sequences S0 –S3 are mainly situated in the water and oil legs of the N’Kossa field and are the main focus of interest for this study. They are carbonate dominated and often dolomitic (up to 60% dolomite). The uppermost part of S3 is mainly non-reservoir micritic

limestone and forms an extensive, correlatable barrier at the scale of the field. S4 –S5 are dominated by sandstone and clastic carbonate deposits and are beyond the scope of this paper. They represent a period of renewed shoreline progradation. The interval S4 –S5 shows strong thickness variations in the study area and thickens considerably from the N’Kossa field towards N’Kossa South where it is the most important reservoir interval. The thickness variation of S4 –S5 indicates more rapid, fault controlled accommodation space creation compared to S0 –S3, where subsidence was largely controlled by the passive movement of salt and where composite sequences have a more constant thickness (as shown in Fig. 10).

Sedimentology of the interval S0– S3 Core data has been used extensively for sedimentological and diagenetic studies which help to constrain the geological 3D model. The studies of sedimentary facies and depositional environments of sequences S0-S3 are based on cores from wells

142 J. P. WONHAM ET AL.

Fig. 9. Variable dolomitization and extensive ‘sandstone drains’. Section location shown on Figures 4 and 15. Note strong dolomitization to the west created by structural palaeohigh. Section is flattened on S4D2 datum.

RESERVOIR DYNAMICS, N’KOSSA FIELD 143

Fig. 10. Variable dolomitization and extensive transgressive sandstones that serve as preferential reservoir drains. Section location shown on Figures 4 and 15. Section is flattened on S4D2 datum.

144

J. P. WONHAM ET AL.

NKSM-2, NKSM-3, NKSM-4, NKF1-01 and NKF2-02 (at least 300 m of core per well) and NKSM-4 (27 m). Well locations are shown in Fig. 4. A simplified description and interpretation of facies described in the core is given below. The facies (shown by photographs in Fig. 11) are described in order of interpreted increasingly distal position: Facies A: Laminated anhydrites alternating with dolomitic or silty claystones. Preserved thickness is typically 50 cm or less and not more than 1 m. Claystones may show the presence of dessication cracks. This facies is indicative of a supratidal sabkha environment deposited up to 2 m above mean sea-level. Deposition was sporadic by occasional marine inundation followed by evaporite precipitation during drying of the sabkha. Facies B: Dolomite, silty dolomite, calcaraeous dolomite and dolomitic limestone (mudstone to wackestone). Beds are usually metre-scale, forming intervals up to several metres thick. Laminations are normally horizontal or slightly inclined. Algal mats and small nodules of anhydrite are occasionally developed. This facies is indicative of an intertidal flat environment. Dolomite is early in origin and is discussed in more detail below. Studies of the Trucial Coast of Abu Dhabi, the type area for the sabkha dolomitization model (Butler et al. 1982;

Machel 2004), show that modern dolomitization on intertidal –supratidal flats normally occurs in the upper 2 m of the sediment column. Facies C: Bioturbated dolomitic siltstone or silty sandstone showing remnant low angle lamination and clay laminae. Dolomite cementation is generally weak compared to Facies B. This facies is interpreted to be indicative of a lower intertidal flat setting. Facies D: Very fine to medium-grained sandstone (units up to 10 m thick) with occasional clay laminae. Cross-bedding, current ripples or planar lamination may be present, but the sandstones often appear massive. Units composed of this facies can be correlated at the scale of the field. These sandstones are interpreted to be the deposits of decimetre to metre-scale dunes developed in a subtidal to intertidal channel and flat environment. Clay laminae provide possible indications of tidal influence. However, no strong tidal indicators such as spring-neap clay drape bundles have been identified. The relative stratigraphic position of this facies above laminated anhydrites (see above) suggests that may have originated by transgression in an estuarine environment. Facies E: Massive limestone (mudstone to wackestone), sometimes dolomitic, showing few sedimentary structures apart from variable levels

Fig. 11. Core photographs showing reservoir facies for Sequences S0–S3 with associated porosity and permeability characteristics.


of bioturbation. Beds are usually metre-scale forming intervals up to several metres thick. This facies is indicative of a subtidal lagoon environment with water depths of between 5 and 30 m. Where this facies is more dolomitized and bioturbated, shallower water deposition is interpreted. Facies F: Bioclastic, oolitic and oncolithic limestones (grainstone to packstone) containing red algae. Inclined laminations are commonly developed. Beds are usually metre-scale forming intervals up to several metres thick. This facies is interpreted to be the deposit of subtidal dunes developed in the external part of an estuary as oolitic shoals. Water depth may be around 5–10 m, but there are few bathymetric indicators. Facies G: Strongly bioturbated calcitic siltstone or mudstone. This facies is interpreted as a deposit of the external platform. It is only encountered as a cap rock above the reservoir. The sedimentological model is one of littoral sedimentation on a mixed siliciclastic –carbonate ramp. During periods of relatively high accommodation space creation and marine inundation, the ramp was dominated by carbonate deposition. At times when accommodation space creation slowed, the ramp was transformed to a mixed siliciclastic –carbonate setting with siliciclastic material derived from the land able to accumulate as intertidal deposits on the margins of barred lagoons. It is likely that structural movements played a role in the orientation and positioning of coastal features such as embayments and subaerially exposed areas. This is discussed in more detail below.

145

High resolution sequences of the S0– S3 interval The sedimentological model for a small-scale T– R sequence (5–15 m thick) of the S0 –S3 interval (Fig. 12) is based on the recognition of facies stacking patterns (Caline et al. 1994, 1995) and serves as the basis for correlation across the field as shown in Figures 9 and 10. Sandstones of the transgressive succession (Facies D) commonly overlie a transgressive coquina lag with small pebbles. In some cases, these transgressive sandstones can be subdivided into a basal very fine sandstone, often showing bioturbation, which is overlain by medium-grained sandstones showing cross-bedding. These sub-units can be traced across the field and are considered to reflect phases of transgressive deepening in a tidal estuarine setting. Passing upwards, siliciclastic deposition can become mixed with carbonate deposition as open marine influences invade the tidal flat area. A bed of bioclastic and/or oolitic limestone (Facies F) is sometimes developed at the top of the transgressive succession. Following sea-level rise, a lagoonal limestone may be developed (Facies E). The subsequent highstand period sees the progradation of intertidal-supratidal flats (Facies B and C) into lagoonal and bay areas. Shallowing eventually results in the development of features such as dessication cracks related to subaerial exposure, and eventual supratidal deposition as recorded by the presence of a dolomitic claystone

Fig. 12. Idealized small-scale transgressive –regressive sequence.

146

J. P. WONHAM ET AL.

interval (typically 50 cm thick) with anhydrite development (Facies A) interpreted to be formed by sabkha processes. Dolomite cementation commonly occurs in the interval below the claystone and will be described and interpreted in more detail in the following sections. No specific evidence for a lowstand phase (e.g. fluvial incision) or lowstand deposits (e.g. fluvial deposits) has been recognized. Generally, the highstand phase is directly overlain by transgressive deposits. The sequence described above is an idealized one not commonly present in the complete form described. In reality, the facies which make up high order sequences are developed in accordance with their relative position within low order sequences. This means that either the transgressive or regressive phase of the high order sequence may be accentuated in terms of thickness. Equally, more proximal siliciclastic or more distal carbonate deposition may dominate. It is by such changes in the characteristics of high order sequences that midorder and low-order sequences are recognized.

Description of diagenetic textures in the interval S0 – S3 The following section describes diagenetic textures and changes in diagenetic texture across the field. Interpretations are provided in the following section. From a reservoir point of view, the most significant diagenetic transformation is the dolomitization of limestones. This dolomitization is widespread and responsible for local improvements of reservoir quality across the field through the formation of secondary microporosity and microvuggy porosity. Dolomite also occurs widely as a cement within siltstones and sandstones and is developed within thin (dm-thick) corrensite-rich mudstones. Along side the dolomitization, other notable phenomena include: (1) the precipitation of graincoating calcite cements as isopachous fringes, sometimes later dolomitized; (2) intergranular dissolution of ooids creating oomouldic porosity, which is partially developed in the oolite shoal facies and often developed throughout where the oolites occur within sandstones; (3) sparry calcite cements, post-dolomitization and after grain dissolution; and (4) poikilotopic anhydrite cements partially replacing ooid or bioclast fragments. The dolomite (sometimes ferruginous) is typically rhombohedric, subhedral to euhedral, with the size of the crystals varying from 5– 60 mm (average 15 mm). The degree of dolomitization is variable, even at the microscopic scale: all stages of dolomitization exist between early crystal growth, cloudy replacement texture or total

dolomitization. The dolomitization is also selective, guided by macrotextural features (preferential dolomitization of beds showing evidence of subaerial exposure for example) or microtextural features, for example: selective dolomitization of micritic matrix or resistance to dolomitization by ooids and bioclasts. The consequence of this discontinuous character of dolomitization is the generation of reservoir heterogeneity at different scales. Isotopic and chemical data used to determine the origin and nature of the dolomitizing fluids show a strong vertical variability at the scale of the reservoir. Dolomitization is variable both areally and vertically. For example, some stratigraphic levels show very strong dolomitization associated with formation of a micro-vuggy texture (vugs c. 200 mm) within dolomite. Such strong dolomitization is notably developed at various stratigraphic levels within S2, but less within S1 and S3. It is also preferentially developed on the western part of the field in a zone of radius 2 km around the NKSM-2 well. In addition, well correlation panels traversing N’Kossa field (Figs 9 & 10) shows that between surfaces S2D2 and S3B4, dolomite development became increasingly areally restricted until it was eventually only strongly developed in certain areas of the field, notably a small area in the western part of the field around NKSM-2 well. Comparison between maps of high resolution sequence thickness and maps which show the distribution of lithofacies (Fig. 13) suggest that in areas where sequences are relatively thin (hence low subsidence rates) dolomite is preferentially developed.

Interpretation of diagenetic effects in the interval S0– S3 Four factors indicate the early, synsedimentary nature of the dolomitization in a hypersaline, sabkha setting: (1) association of dolomite with laminated anhydrite development; (2) the stratabound nature of the dolomite with dolomitization mainly confined to the regressive portions of small-scale T–R sequences as shown in Figure 14; (3) a strong relationship between the d18O composition of the dolomites and its covariance with trace elements (such as Na) typically enriched in evaporitic brines (Walgenwitz et al. 1992); and (4) association with corrensite (Magnesium-rich claystone) which typically forms authigenically within a sabkha under near-surface conditions where magnesium is abundant and where temperature and pH are high enough to overcome the kinetic barriers to its formation (Andreason 1992). In general, dolomites indicative of the most hypersaline conditions are found at the base of the reservoir (S1 and the base of S2). This may


147

Fig. 13. One high order T– R sequence showing how dolomite is preferentially developed in zones of thinner reservoir, above palaeostructural highs where subsidence rates are reduced.

Fig. 14. Linked sedimentology and diagenesis of a high order T –R sequence (10 m thick). Facies colour scheme as for Figure 12.

148

J. P. WONHAM ET AL.

explain the strong and widespread dolomitization at these stratigraphic levels with the formation of vuggy dolomites in S2. From the top of S2 up to S4, the proportion of samples indicative of evaporation decreases progressively suggesting that the dolomitization occurred through the circulation of brines that were moderately hypersaline or close to sea water. Only S4 has a strontium isotope composition which conforms to the Albian oceanic norm (Walgenwitz et al. 1992). Synthesis of geochemical, stratigraphic and facies mapping data brings some precise constraints to the dolomitization model for N’Kossa. The diagenetic model combines two processes: (1) increase in the Mg/Ca ratio of cementing solutions due to precipitation of sulphates such as anhydrite in a sabkhatype supratidal evaporitic domain; and (2) gravitational reflux of high-density brines on the depositional slope. The dolomite potential decreases down slope due to: extraction of magnesium along the reaction pathway; mixing with marine waters; and the reinforcement of kinetic barriers which result from this (Fig. 15). The crystal textures remain overall microsparitic and favour a homogeneous distribution of intercrystalline porosity. The dolomitization front, which constrains the limit beyond which the kinetic barriers do not allow any more dolomitization, is mobile through time and is regularly displaced according to fluctuations of relative sea-level. As described above, the areal variability of dolomite cementation is well documented from the forty-one wells distributed over the field using layer-based mapping and well correlation panels. This variability is interpreted to have been influenced by the position of palaeo-structural highs where subsidence rate was slower (Wonham et al.

2005). Preferential dolomite development on the western part of the field in a zone of radius 2 km around the NKSM-2 well is interpreted to be a response to long-term transgression which gradually restricted the area of dolomitization to a smaller and smaller region situated on the palaeohigh: the focus point of the arcuate (plan view) listric fault as shown in Figure 15. The significance of understanding the depositional controls on preferential dolomite development is that the enhanced porosity of this reservoir type means that a large proportion of the field’s reserves are present in such facies. Given the widespread development of structures analogous to N’Kossa on the West African margin (both in the Congolese and Angolan offshore zones), this model of structural influence on dolomite reservoir creation has good potential for predicting reservoir quality distribution using only thickness maps generated from seismic data and sparse well control.

Extrapolation of core facies to all wells as electrofacies Lithofacies of the Sendji Carbonate are very variable as a function of: (1) original lithology, that is, dolomite, sandstone, limestone; (2) shaliness, that is, silty or clayey; and (3) diagenesis, that is, dolomitic or calcitic. The combination of these factors produces a great number of possible lithofacies. Core studies show that phi –K relationships are different for different lithologies, however, there are two major trends, one for granular lithofacies and one for the non-granular carbonate and dolomite lithologies (Fig. 16). For modelling purposes, three important exercises were carried out: (1) simplification of

Fig. 15. Early dolomitization in the structural context of listric fault development. Dolomite strongly developed in a palaeohigh area of relatively low accommodation space creation. This area is situated at the focus point of an arc created by the major listric fault in plan view.


149

Fig. 16. Poro-perm characteristics of lithofacies. Note that limestone can belong to either a high permeability trend (oolitic–bioclastic grainstone facies) or a low permeability trend (micritic wackestone/mudstone facies).

depositional facies described in core to a number of key facies types with distinctive lithology as described above; (2) verification that this simplification can support the definition of robust petrophysical groups; and (3) verification that this simplification will allow the calculation of electrofacies equivalent to the simplified lithofacies in noncored wells. It is necessary to simplify the lithofacies scheme due to the limitations of complexity that can be handled by the geological model. The facies previously described are simplified to five lithofacies groups: (1) shale (Facies A); (2) sandstone/siltstone (facies C and D); (3) dolomite (Facies B); (4) limestone (Facies F); and (5) compact limestone with total porosity ,5% (Facies E). These lithofacies are identified in core and used to calibrate a multimineral interpretation of the petrophysical logs. Cut-offs are applied on the log-derived mineral proportions in order to arrive at a representative lithofacies for any given interval. The resulting electrofacies are robust and can be used for sequence recognition, geological correlation (e.g. Figs 9 & 10) and lateral trend determination.

Reservoir characteristics of the S0– S3 interval Transgressive sandstones developed within S2 form a series of stacked, homogenous, highly continuous, very fine to medium-grained sandstone units (up to

10 m thick) which can be correlated at the scale of the entire field. These sandstones originate from estuarine transgression of the underlying anhydritic sabkha mudflat and extend across the whole field. Sandstone permeability is good. Core plug permeability is typically in the range of 10–1000 mD (Figs 11 & 16). Lower permeabilities are associated with occasional silty sandstones or dolomite cemented sandstones. Good permeabilites for sandstones have also been proved by numerous DSTs and are in the range of 60–350 mDm (Table 1). The sandstones possess notably better reservoir quality, in terms of permeability, than the dolomitic and clastic limestone deposits that surround them (Table 1). From a reservoir point of view, these transgressive sandstones act as levels of preferential fluid mobility. Dolomite reservoir has an intercrystalline porosity distributed in a manner which is generally homogenous and well connected. The pore size is of the order of 15 mm. The measured porosity of dolomite in plugs can be as much as 35% in zones where micro-vuggy porosity is developed, but the global average layer-based porosity varies from 18–20% which is similar to sandstone porosity and considerably higher than bioclastic/oolitic limestone porosity where global average layer-based porosity is around 13%. Dolomite permeability is, however, considerably lower than sandstone permeability, with core plug permeability typically in the range of 0.05– 50 mD (Figs 11 & 16) and interpreted DST permeability generally ,15 mDm (Table 1).

150

J. P. WONHAM ET AL.

Table 1. Summary of Sendji Carbonate reservoir properties for lithological types identified from petrophysical log evaluation

AvPHIE NET (%) S4 S3 S2 S0–S1 Kh DST (mD) S4 S3 S2 S0–SI

Sandstone

Limestone

Dolomite

Compact limestone

14 18 18 18

10 10 10 13

13 14 20 18

5 8 – 8

125 – 60 to 350 –

15 to 40 –

– 5 to 10 – 5 to 15

– – – –

Reservoir pressures can be recorded in the subsurface and well testing shows low to moderate permeability.

Extent of reservoir barriers and short-comings of earlier models Numerous vertical barriers to flow, ranging from 50 cm to tens of metres in thickness are recognized. A certain number, notably those that are shaly, have a thickness less than the resolution of the log data

–

which is around 70–100 cm. Nonetheless, the majority of these very thin barriers can be recognized from the historical formation pressure data and from descriptions of core data and plugs. The barriers are of two types: (1) maximum progradation/transgressive surfaces (MP/TS) produced by two superimposed stratigraphic events (Fig. 17): an initial fall in sea-level which creates a thin bed (50 cm thick) of supratidal anhydritic shale developed in a sabkha setting with a very large lateral extent (field wide). These shales are seen on gamma ray logs but sometimes appear

Fig. 17. Depositional origin and reservoir character of Maximum Progradation/Transgressive Surface (MP/TS) type vertical flow barrier.


porous on neutron logs due either to their thinness or to the presence of dolomite-related porosity within them. The initial fall in sea- level is followed by a transgression which introduces an extensive bed of bioclastic/oolitic grainstone. This grainstone is subsequently cemented to form a barrier. The two nonreservoir lithologies are superimposed and act together to create a significant reservoir barrier that can be correlated throughout the field despite its thin development (,1 m thick); (2) maximum flooding surfaces (MFS) which create a bed (1– 5 m thick) of tight micritic limestone of lagoonal origin (Fig. 18). These barriers have a relatively variable thickness since they have been created slowly (condensed beds) and hence their thickness is more prone to change in response to local tectonic movements. Barrier formation is linked to cycles of accommodation space creation with lagoonal limestone formed in periods of maximum accommodation space creation and anhydritic shales/cemented bioclastic/oolitic grainstones formed in periods of minimum accommodation space creation. Both barrier types show a lateral variability in thickness in relation to the structural control being, in general, thicker towards the East in proximity to the large listric fault which bounds the field. Where barriers are occasionally absent, it is commonly found to be due to post-depositional faulting.

151

Well-to-well correlation shows that both types of barrier are generally continuous across the field and this is in keeping with their stratigraphic origin. The process origin of barriers is a key element of reservoir characterization: barriers related to stratigraphic cyclicity (typical periodicity of 100s–1000s of years) are commonly more continuous than barriers or baffles produced by random events, for example shaly fluvial abandonment facies produced by river channel avulsion (typical periodicity of 10s of years). The continuity of barriers became apparent at a late stage in N’Kossa’s development through the amassing of well data and its integration with production data (as detailed later). It is worth reviewing here the historical aspects of barrier recognition on the field since this reveals how and why the continuity of barriers was initially underestimated. While the presence of vertical flow barriers was recognized by early studies of the field, prior to production start-up, the lateral continuity and possible effect of these barriers was difficult to quantify without any production data. Mapping of lateral continuity was difficult due to the small well database, precise correlation between wells situated kilometres apart being impossible. In addition, the pressure database suggested the presence of a single uncompartmentalized reservoir zone. Consequently, barriers were incorrectly represented in the

Fig. 18. Depositional origin and reservoir character of Maximum Flooding Surface (MFS) type vertical flow barrier.

152 J. P. WONHAM ET AL. Fig. 19. Core data from S2 (NKSM-2 well) showing facies and low permeability MP/TS type barriers. Photograph indicates core plug locations above and below probable low permeability zones.


early reservoir models as discontinuous ‘stochastic’ baffles. Detailed mapping of the barriers was carried out as more wells were drilled and used in the reservoir modelling. However, the identification of barriers was based only on static criteria: for example, porosity below 8% and thickness .1 metre. These cut-off values failed to recognize the field-wide continuity of low permeability beds that were as little as 0.5 m thick. Because the beds were thin, they were poorly characterized by logging tools. In addition, some shale barriers had elevated porosity values due to dolomitization while retaining extremely low permeability values. The early models which resulted from this approach continued to underestimate vertical compartmentalization and overestimate the water flood sweeping efficiency. Another reason for the underestimation of barrier presence is found in the treatment of core data. These data verify the existence of low permeability barriers (Fig. 19), but it is now noted that phi–K plugs were often not cut in intervals that were clay/silt-dominated (Fig. 19, inset photographs). This created a bias in the sampling and meant that very low permeability zones did not show up in petrophysical datasets that were later used to establish cut-offs.

Dynamic synthesis The dynamic synthesis utilized production data of different types collected from across the field (Table 2). The first task of this synthesis was to integrate updated geological correlations with production data in order to define the major flow units of the field. Since the geological synthesis had stressed the presence of very extensive and laterally continuous vertical barriers associated with MP/TS and MFS surfaces, evidence was sought for the influence of these flow barriers in the formation pressure data from infill drilling wells. This pressure data showed that the extensive transgressive sandstones

153

were becoming very depleted and that surrounding dolomite layers were poorly depleted (Fig. 20). The field’s dynamic behaviour was captured by a simplified model of ten flow units (Fig. 21) subdivided by extensive flow barriers. The flow units are named FU10– FU1 (top to base of the reservoir) and are defined by regrouping sedimentary layers that have shown the same dynamic behaviour, assessed from PLT, well tests and formation pressure data acquired after first oil. Within each flow unit, the lateral connectivity is assumed to be good, although there are likely to be baffles due to faulting seen on seismic and identified in core. Barriers to vertical flow are detected from formation pressure data which sometimes shows a deltaP (pressure difference) value of as much as 110 bars across a barrier bed of less than 1 m thick. The dynamic synthesis assessed the connectivity between flow units using the pressure v. time trend in order to characterize the relative importance of different geological types of barrier (Fig. 22). Rather than using porosity cut-offs, the dynamic synthesis of 2004 developed a new approach to adequately model the vertical barriers in the reservoir model. This approach involved transforming the calculated horizontal permeability log to a vertical permeability log using a Kv/Kh ratio that varied by barrier type (MP/TS or MFS). A harmonic average of vertical permeability was then calculated from the centre of one reservoir layer to the next with no cut-offs applied in order to capture, in the dynamic model, the low vertical permeability of any barrier layer present. A total of twenty-four DST tests carried out in seventeen wells on a range of different flow units. This data provides evidence of flow unit effective permeability and the presence or absence of no-flow boundaries related to faulting. In seven of the DSTs, indications were seen on seismic that a fault would be present in the radius of investigation. Six of these seven tests proved no-flow boundaries. This suggests that seismic-visible faults are likely to

Table 2. Dynamic data available for the dynamic synthesis Issue

Parameters

Vertical connectivity

Kv/Kh ratio Tight layers Thin, extensive barriers

Lateral connectivity

Faulting

Lateral anisotropy

Variation of facies Variation of permeability in a given facies

Dynamic data available Plug ratio, MDT multi-probe, DST MDT pressure profile, plug data, mini-permeameter, GWD, well correlation Pressure v. time plot derived from MDT, DST, fault plane analysis Sedimentary correlation Phi– K laws from core data, DST, mobility from MDT, PLT

154

J. P. WONHAM ET AL.

act as barriers to flow. Nine tests were carried out in areas considered to have no seismic-visible faults in their radius of investigation. Only one of these nine DSTs detected a no-flow boundary suggesting that faults below seismic resolution are either not present or have little influence on reservoir flow characteristics. Since only one seismic-visible fault can be proved as open to flow by the dynamic data, compartment-bounding faults of N’Kossa field are considered to be sealing unless proven otherwise. Examples have been observed, however, where injected gas has caused overpressuring in faultbounded compartments adjacent to the injector well compartment, supporting the idea that some faults should be modelled as open to flow, at least to gas. A geological model was created for N’Kossa (Fig. 21) based on updated correlations and facies mapping of the field using data from all the wells. The model captures stratigraphic trends by the

incorporation of a total of 77 layers. For each of these layers, electrofacies based on simplified lithology classes (as previously described) were mapped using well data. These electrofacies are not sufficient to describe the permeability variation of the field and permeability modifiers (based on core petrophysical measurements and DST data) are therefore applied on a layer by layer basis within the geological model. The dynamic model created from the geological model was calibrated to pressure data collected after the start-up of production on the field. Vertical transmissivity in the model was modelled between layers using an approach based on the genetic origin of the barriers (as described above). History matching was targeted towards matching both produced hydrocarbon volumes and observed pressure data. In order to arrive at a well calibrated model it was necessary to adjust permeability multipliers related to vertical transmissivity barriers between layers. In a few cases, faults recognized on seismic

Fig. 20. MDT pressure curve showing typical reservoir depletion (NKF1– 17).


Fig. 21. Capturing dynamic behaviour with 10 dynamic flow units. Figure at top right is an oblique 3D structural view of top reservoir surface S2B5 showing wells and section location. 155

156 J. P. WONHAM ET AL. Fig. 22. Mapping barriers in dynamic context. Position of correlation line shown in Figure 4. Distance between wells is roughly 1 km. Values in bar indicate differential pressure across barriers.


157

Review of the initial development plan

Fig. 23. Calibration of dynamic model to MDT pressure test points.

were opened to flow. The model history proved the necessity for introducing extensive flow barriers and finally resulted in a good model to data match (Fig. 23).

The first objective of the dynamic synthesis was to explain the dynamic evolution of the field since start-up. Using a revised flow unit description it was possible to return to the initial development concept and work out step-by-step the history of the field in a robust geological framework. The initial development concept is illustrated in Figure 24. Because the appraisal of the field had shown a single hydrocarbon column with a WOC at 23425 m over the entire structure (Fig. 3), the base case hypothesis was that the field would produce as a single, reasonably well connected reservoir. The initial development plan led to the drilling of 43 wells of which 25 were producers, 11 water injectors and 7 gas injectors (Fig. 4). The producers were completed with commingled production in flow units 1 to 5 with a stand off to the water and gas contacts. The sandstone layers were systematically perforated in order to ensure good productivity. Three types of producers were identified according to their location: † Type 1: located on the West flank and completed in FU 4-5-6 (water stand off) † Type 2: located in the middle of the reservoir and completed in FU 3-4-5. † Type 3: located towards the East flank and completed in FU 1-2-3 (gas stand off)

Fig. 24. Schematic showing initial development concept: gas injection at the top, bottom water drive, production commingling. Section location is similar to that of Figure 9.

158

J. P. WONHAM ET AL.

It was planned that reservoir pressure would be maintained with gas injectors in the gas cap and water injectors in the aquifer. The water injectors were located on the periphery (both West and East flank) and completed in FU1 –2. Initially, only 2 horizontal gas injectors were planned to be completed in FU10, however, as it became apparent that these injectors were isolated from the main producing intervals by tight limestone layers, it was decided to drill more injectors and complete them in FU5 –6. The initial development plan based on pressure support by gas and water injection was not suited to the ‘multilayer’ behaviour of the field, in particular the water injectors (FU1 –2) proved to be very poorly connected to the main producing flow units (FU3–5) with a consequent lack in pressure support during the first years of production. The field production history is shown in Figure 25. From 1996 to mid-1998, water injection was confined to the lower part of the reservoir in the West flank (FU1) with the aim of supporting reservoir pressures in the oil leg. However, the presence of extensive barriers to vertical flow meant that the main producing areas (FU3 to 5) were quickly depleted from 390 to 200 bars. The field potential declined abruptly and the FU1 was over-pressured (from 405 bars to more than 450 bars). Water injection on the East flank was quickly abandoned due to the poor characteristics of the reservoir and poor connectivity between injectors and producers, again due to the presence of barriers to vertical flow. By mid 1998, the over-pressurization in the FU1 was such that fractures with high vertical extension were suddenly generated near the water injectors creating a direct path between the injection point and the depleted flow units (Fig. 26). After the water injectors had fractured the overlying barriers, water swept the high permeability sandstone layers (FU2–4) and quickly reached the producers. Due to the low reservoir pressure in the invaded area and the absence of activation, most of the wells died shortly after the water breakthrough. Water advance though the high permeability zone has been mapped (Fig. 27), and is now providing the basis for identifying the remaining potential of the field. It is now evident that initial producer perforation policy was not selective enough. Wells were producing with 3–5 flow units commingled, with high productivity difference between sandstone and dolomite/limestone layers. No activation was planned in spite of the secondary recovery scheme by water injection. In addition to the problems of water breakthrough, two of the four gas injectors located at the top of the reservoir (FU10) were found to be completely isolated from producing areas by thick and tight limestone layers and

contributed only to the creation of overpressure in upper zones of the reservoir. These effects of the strong vertical heterogeneity had a significant impact on reducing the plateau of the field and the long-term production profile. In spite of a very significant effort of monitoring during the first years of production, acquisition issues and insufficient data integration prevented a clear understanding of the dynamics of the field. Apart from the problems of vertical flow barrier recognition which have already been described, key issues included the lack of a proper interference test at start-up to understand the communication between producers and between producers and injectors. Tracers could also have been used in the injected gas and water to monitor the injection front.

Actions and results arising from the dynamic synthesis Integration of geological correlation and mapping with dynamic data such as DST and formation pressure data provide a conceptual framework for mapping water or gas injection in each flow unit and serve as the basis for identifying the remaining potential of the field (Fig. 28). Actions launched following the dynamic synthesis follow three main strategies: † The drilling of new producers in FU6– 10 on the West Flank of N’Kossa (NKF1–25) to target undeveloped dolomitic flow units. Production from this zone will also be supported by the conversion of water injector wells previously perforated in FU1– 2 (NKF1 –16 and NKF1 –20). † The drilling of new producers (NKF1 –23 and NKF1 –26) in FU1 –2 which represent reservoirs of oolitic grainstone and dolomite that remains unproduced, notably on the eastern flank of the field. † Ensuring pressure maintenance through gas injection in FU6– 10 and FU1 –2 into which new producers are being drilled. Water injection across the field has been stopped. Initial results from these strategies are promising. Converted water injector wells NKF-1 –16 and NKF1 –20 are producing in FU6– 10 and show initial oil rates of 2900– 4250 bopd, which is considered a good performance for the dolomitic reservoirs at this level. The new producer NKF1– 25 has performed slightly less well with initial rates of 2100 bopd descending to 700 bopd a few months later. A high production of water (50% water cut) suggests this well may be receiving water injected into FU1 – 2 via fractures from nearby NKF1-05 injector which is located close to a large NW–SE striking fault (for location see Fig. 27).


Fig. 25. Field production history, June 1996 to May 2002. 159

160

J. P. WONHAM ET AL.

Fig. 26. Injector wells of the West flank showing 50 bars of over pressurization in well NKF1–16 due to water injection. Subsequent fracturing of the barriers at top FU1 around NKF1-05 led to uncontrolled water flooding in FU2, FU3 and FU4.

Fig. 27. Water front advance by flow unit showing first arrival dates of water advance. Note widespread advance of water in transgressive sandstones of FU3 following fracturing of underlying limestone barriers (situated above FU1).


161

Fig. 28. Actions since geological/dynamic synthesis. 2002 water flood status for FU3 is shown for reference (see Fig. 26).

Wells NKF1 –23 and NKF1 –26 targetting FU1 –2 have also produced at good rates. A production logging test (PLT) measured in NKF1 –26 showed 85% of production coming from the oolitic limestone and dolomitic limestone facies of FU1 and the rest from the silty dolomite and sandstone facies (thin sandstone beds) of FU2. A pilot gas injector has been drilled in FU1 –2. It is not yet known if this will be successful at maintaining pressure. If it is, it is intended that more gas injector wells will be drilled in FU1 –2 and FU6 –7.

Discussion The identification of a single hydrocarbon column from pressure data at the field appreciation stage, led to a hypothesis of a single, reasonably well connected reservoir which could be produced with pressure support from water injectors at the base of the reservoir and gas injection at the top. The presence of a single hydrocarbon column contrasted with other fields developed in the Sendji Carbonate reservoir such as the Sendji field (Baudouy & Legorjus 1991) where a greater number of wells were required to develop multiple reservoir zones with variable fluid content. It is possible that the N’Kossa field, being a rafted structure, was a closed pressure cell in which the hydrocarbon column had equilibrated over geological time,

thereby disguising the heterogeneity in an inherently complex reservoir type. Vertical permeability problems were expected in some parts of the reservoir, however interference testing planned at an early stage of the development in order to confront this uncertainty was not put into practice. In contrast to the Sendji Field, which according to Baudouy & Legorjus (1991) posed ‘no major production problems’, the N’Kossa field declined from plateau production more rapidly than expected. The principal remedial action of massive water injection into the aquifer (FU1 and FU2) did nothing to improve the poor pressure support in the overlying good reservoir quality transgressive sandstones (FU3 and FU4) where unsupported production had led to high GORs. Eventually, water injection increased reservoir pressure by around 50 bars and led to overpressure fracturing of the barriers that separated the aquifer from the reservoir zones and rapid water flooding of the transgressive sandstones ensued. This shows that, in fields where complex heterogeneity exists, a detailed monitoring of reservoir dynamic evolution needs to be put in place with regular update of the reservoir model. Importantly, if the development strategy is found not to be conforming to expectation, actions need to be put in place to update the development plan. In the case of N’Kossa, such an adaption would have been to close off water injection to

162

J. P. WONHAM ET AL.

prevent fracturing of the base FU3 barrier and the subsequent rapid, uncontrolled sweep of a few high permeability zones. Despite its difficulties, the N’Kossa field continues to offer opportunities for infill drilling notably in the S4 reservoir in the Central and Southern parts of the field (Mazel & Poitrenaud 2007). The S4 reservoir is more homogenous in character and dominated by sandstone and clastic limestone facies. In the S0 –S3 interval, drilling prospectivity also remains, notably in zones where water-flooding has been ineffective and has left untapped zones which are currently targets for production using horizontal wells.

Conclusions † Facies analysis and high-resolution sequence analysis of Units S0 –S3 of the Sendji Carbonate in the N’Kossa field show that these deposits originated in a lagoon-sabkha setting. Fluctuations in relative sea-level resulted in the development of many stacked small-scale T– R sequences. Early, syndepositional dolomite developed in a hypersaline sabkha setting and substantially improved reservoir characteristics of the original low porosity limestones which underlay the sabkha. † The syndepositional evolution of dolomite means that it is preferentially developed over palaeohighs where accommodation space creation is lower and subaerial exposure is more frequent and for longer periods. For N’Kossa field, the palaeohigh location was linked to the position of major listric faults that are typically arcuate in plan view. The palaeohigh is located at the focal point of the arc, and it is here that dolomitization occurs most strongly. † An abundance of well data together with field production data indicates that two types of laterally extensive ‘stratigraphic’ barrier exist which are usually continuous at the scale of the whole field. These barriers result from processes of relative sea-level change. † Barrier formation is linked to cycles of accommodation space creation. MFS barriers of lagoonal limestone are developed in periods of maximum accommodation space creation and MP/TS barriers of anhydritic shale overlain by trasngressive lags of cemented bioclastic/ oolitic grainstones are formed in periods of minimum accommodation space creation proceeded by marine transgression. † Dynamic modelling of the N’Kossa field illustrates that two main factors control hydrocarbon production: (1) high flow rates from good permeability transgressive sandstones; and (2) presence of many barriers to vertical flow which

impede injected water and gas intended to provide pressure support. † The dynamic synthesis of N’Kossa shows the importance of recognizing barriers to vertical flow at an early stage of field development. Attention needs to be placed on the full evaluation of these barriers right from the stage of taking core plugs, through to sedimentological evaluation of their continuity and possible implementation of interference testing to test their effectiveness as barriers. † Early decline of N’Kossa field shows that where vertical barriers are recognized and are seen to significantly compartmentalize the reservoir, a selective completion and injection strategy will normally need to be developed. Grateful acknowledgements are due to Total, Total E&P Congo and partners on the N’Kossa field: SNPC and Chevron, for permission to publish this study. Many Total geoscientists and engineers assisted with the geological/geophysical studies and dynamic synthesis work, notably: F. Walgenwitz, H. Eichenseer, A. Meyer, P. Biondi, B. Caline, S. Raillard, D. Poilleux, S. Vanneste, L. Schuwer, M.-C. Devilliers, C. Prinet, U. Ashari and M. Chikh. We thank all concerned for their contributions. We would also like to thank B. Ainsworth, M. Po¨ppelreiter and J. Wadsworth for reviewing and suggesting improvements to this paper.

References Andreason, M. W. 1992. Coastal Siliciclastic Sabkhas and related evaporative environments of the Permian Yates formation, North Ward-Estes Field, Ward County. Texas. The American Association of Petroleum Geologists Bulletin, Tulsa, 76, 1735–1759. Baudouy, S. & Legorjus, C. 1991. Sendji field – People’s Republic of Congo, Congo Basin. In: Foster, N. H. & Beaumont, E. A. (eds) Treatise of Petroleum Geology, Atlas of Oil and Gas Fields – Structural Traps V. American Association of Petroleum Geologists, Tulsa, 121 –149. Brownfield, M. E. & Charpentier, R. R. 2006. Geology and Total Petroleum Systems of the West– Central Coastal Province (7203), West Africa. U.S. Geological Survey Bulletin, 2207-B. Burollet, P. F. 1975. Tectonique en radeaux en Angola. Bulletin de la Socie´te´ Geólogique de France, 17, 503–504. Butler, G. P., Harris, P. M. & Kendall, C. G. St. C. 1982. Recent evaporates from the Abu Dhabi coastal flats. In: Handford, C. R., Loucks, R. G. & Davies, G. R. (eds) Deposition and Diagenetic Spectra of Evaporites. Society of Economic Palaeontologists and Mineralogists Core Workshop 3, Tulsa, 33–64. Caline, B., Eichenseer, H. & Calatayud, P. 1994. Application of sequence stratigraphy to a mixed carbonate– siliciclastic Albian reservoir (Offshore Congo). Proceedings of the Conference on High Resolution Sequence Stratigraphy: Innovations and Applications, University of Liverpool, UK.

RESERVOIR DYNAMICS, N’KOSSA FIELD Caline, B., Eichenseer, H., Calatayud, P. & Walgenwitz, F. 1995. Improved reservoir prediction of a mixed siliciclastic– carbonate platform by using highresolution sequence stratigraphy (N’Kossa field, Offshore Congo). American Association of Petroleum Geologists Bulletin, 79, 1201. Cameron, N., Bate, R., Clure, V. & Benton, J. 1999. Oil and gas habitats of the South Atlantic: introduction. In: Cameron, N. R., Bate, R. H. & Clure, V. S. (eds) The Oil and Gas Habitats of the South Atlantic. Geological Society, London, Special Publications, 153, 101–131. Cameron, N. & White, K. 1999. Exploration opportunities in offshore Deepwater Africa. Proceedings of the IBC Conference ‘Oil and gas developments in West Africa’, London, 25–26 October 1999. Coward, M. P., Purdy, E. G., Ries, A. C. & Smith, D. G. 1999. The distribution of petroleum reserves in basins of the South Atlantic. In: Cameron, N. R., Bate, R. H. & Clure, V. S. (eds) The Oil and Gas Habitats of the South Atlantic. Geological Society, London, Special Publications, 153, 101– 131. Eichenseer, H. T., Walgenwitz, F. R. & Biondi, P. J. 1999. Stratigraphic control of facies and diagenesis of dolomitized oolitic siliciclastic ramp sequences (Pinda group, albian, Offshore Angola). American Association of Petroleum Geologists Bulletin, 83, 1729–1758. Embry, A. F. & Johannessen, E. P. 1993. T –R sequence stratigraphy, facies analysis and reservoir distribution in the uppermost Triassic –Lower Jurassic succession, western Sverdrup Basin, Arctic Canada. In: Vorren, T. O., Bergsager, E., Dahl-Stamnes, O. A., Holter, E., Johansen, B., Lie, E. & Lund, T. B. (eds) Arctic Geology and Petroleum Potential, Norwegian Petroleum Society Special Publication, Elsevier, Amsterdam, 2, 121–146. Fort, X., Brun, J.-P. & Chauvel, F. 2004. Salt tectonics on the Angolan margin, synsedimentary deformation processes. American Association of Petroleum Geologists Bulletin, 88, 1523–1544. Machel, H. G. 2004. Concepts and models of dolomitization: a critical appraisal. In: Braithwaite, C. J. R., Rizzi, G. & Darke, G. (eds) The Geometry and Petrogenesis of Dolomite Hydrocarbon Reservoirs. Geological Society, London, Special Publications, 235, 7 –63. Mazel, J. M. & Poitrenaud, H. 2007. Acid Stimulation of Extended Reach Wells: Lessons Learnt from

163

N’Kossa Field. Society of Petroleum Engineers, SPE Paper 107760. Raillard, S., Biteau, J. J., Allix, P. & Chevalier, C. 1998. Lower congo Tertiary basin – Offshore West Africa structural zonation and evolution (Abstract). American Association of Petroleum Geologists Bulletin, 82, 1954–1955. Rouby, D., Guillocheau, F., Robin, C., Bouroullec, R., Raillard, S., Castelltort, C. & Nalpas, T. 2003. Rates of deformation of an extensional growth fault/raft system (Offshore Congo, West African Margin) from combined accommodation measurements and 3-D restoration. Basin Research, 15, 183– 200. Rouby, D., Raillard, S., Guillocheau, F., Bouroullec, R. & Nalpas, T. 2002. Kinematics of a growth fault/raft system of the West African margin using 3D restoration. Journal of Structural Geology, 24, 783– 796. Valle, P. J., Gjelberg, J. G. & Helland-Hansen, W. 2001. Tectonostratigraphic development in the eastern lower Congo basin, offshore Angola, West Africa. Marine and Petroleum Geology, 18, 909– 927. Van Horn, J. 2001. Sendji Formation reservoir delineation based on 2-D and 3-D inversion, Yombo Field, offshore Congo. The Leading Edge, 20, 435–441. Van Wagoner, J. C., Mitchum, R. D., Campion, K. M. & Rahmanian, V. D. 1990. Siliciclastic sequence stratigraphy in well logs, cores, and outcrops. AAPG Methods in Exploration Series, 7, 1–55. Vernet, R., Assoua-Wande, C. A., Massamba, L. & Sorriaux, P. 1996. Paleógeógraphie du cre´tace´ (Albien-Maastrichtien) du bassin coˆtier congolais. Elf Aquitaine Me´moire, 16, 39– 56. Walgenwitz, F., Le Dluz, A. & Eichenseer, H. 1992. Isotope and trace element record of relative sea level in Albian carbonates from the Congo Atlantic margin. In: Kharaka, Y. K. & Maest, A. S. (eds) Proceedings of the International Symposium on Water–Rock Interaction. Edmonton, 487– 491. Wonham, J. P., Ashari, U., Schuwer, L., Devilliers, M.-C. & Nguyen, T. 2005. Tectono-sedimentary controls on a mixed clastic– carbonate shallow marine reservoir: Lower Sendji Carbonate (Albian), N’Kossa field, Offshore Congo (Long Abstract). Proceedings of the 2005 AAPG International Conference, Paris, 11–14 September.

Stratigraphic and structural compartmentalization of dryland fluvial reservoirs: Triassic Heron Cluster, Central North Sea T. MCKIE1*, S. J. JOLLEY1,2 & M. B. KRISTENSEN1 1

Shell UK Limited, 1 Altens Farm Road, Nigg, Aberdeen, AB12 3FY, UK

2

Shell Canada Energy, 400 – 4th Avenue SW, Calgary, Alberta, T2P 2H5, Canada *Corresponding author (e-mail: [email protected]) Abstract: This paper describes the nature and relative significance of stratigraphic and structural compartmentalization in dryland fluvial reservoirs using data drawn from the Heron Cluster (Heron, Egret and Skua) oil fields in the UK Central North Sea. The Triassic Skagerrak Formation reservoir in these fields was deposited in a variety of dryland terminal fluvial settings, ranging from relatively arid terminal splay and playa to more vegetated, channel-confined systems with associated floodplain and palustrine facies. Laterally extensive floodbasin shales punctuate this terminal fluvial architecture. Static and dynamic data indicate that these fields are compartmentalized: geochemical data indicate significant fluid variations both between wells and vertically within individual wells; material balance calculations suggest production from restricted connected volumes, locally from a subset of the range of oils present; and re-perforation across significant shale boundaries access undepleted reservoir with different fluid compositions. Lateral variations could be ascribed to prominent structuration within these fields, but in general these high net:gross reservoirs do not have a viable fault seal mechanism. Early (syn-halokinetic) grounding of Triassic ‘pods’ between salt swells during salt withdrawal has resulted in zones of intense faulting along the zone of contact of the pod and the underlying basement, and also on the flanks of pods as the margins collapsed under further salt withdrawal. This deformation occurred under relatively shallow burial depths and is largely expressed by disaggregation zones and phyllosilicate fault rocks. Fault property averaging algorithms (e.g. shale gouge ratio), indicate that the sands should communicate across the juxtapositions, implying that the fluids and pressures should equilibrate between reservoir sands. However, the stratigraphic differences across major shales in both fluid geochemistry and pressure caused by draw-down are preserved despite the presence of these faults. The preservation of stratigraphic compartments indicates that for these faults the deformation mechanism was probably dominated by clay smear, in which the shale-prone sequence was smeared down the fault planes without losing its coherence. This style of stratigraphic compartmentalization occurs across several shale-prone intervals that are correlatable across the region. In some cases these mark the boundaries to major changes in fluvial depositional character, provenance and floodplain drainage, suggesting an extrinsic control that led to shale packages defining consistent barriers in all the fields. Other shale barriers do not show major changes in depositional character and, although correlatable, appear to be the product of semiregional advance and retreat of the fluvial systems, possibly combined with nodal avulsion. In contrast to reservoirs deposited by large exorheic rivers, the terminal nature of these dryland fluvial systems appears to have resulted in the repeated interfingering of fluvial and floodbasin facies over a scale of many tens of kilometres. As a result such terminal fluvial reservoirs are prone to stratigraphic compartmentalization. However, thinner shales are prone to breaching by fluvial erosion and as a result not all correlatable shale events form barriers and only a subset will compartmentalize. Mitigation against this compartmentalization requires a development strategy where well trajectory and perforation maximizes stratigraphic exposure.

Flow rates, pressures, fluid saturations and contacts often begin to provide evidence for independent reservoir compartments during the latter stages of production from an oil and/or gas field. Sealing faults are commonly invoked to explain such compartmentalization, partly because they are obvious candidates to explain the areal subdivision of a horizontal (or near horizontal) hydrocarbon column, and because it is relatively straightforward

to perform a range of uncertainty experiments on such features within numerical flow simulation models. However, over-emphasis on faulting as the causative mechanism can lead to inclusion of geologically unrealistic fault arrays and sealing capacities within these simulation models in order to force the modelled production history to mimic the observed compartmentalization. Depositional and stratigraphic heterogeneities that might impact


166

T. MCKIE ET AL.

flow in the subsurface can be over-simplified in these models and their role in reservoir compartmentalization consequently under-estimated (Fisher & Jolley 2007). Fault seal studies have attempted to understand the complexity of fault zones and the distribution of flow properties they contain from a variety of structural settings, basins and reservoir types and to distil this understanding into components that can be directly implemented within geological and production simulation models. However, it could be argued that there are fewer studies that specifically examine the stratigraphy and sandbody architecture of specific depositional systems in terms of fluid flow and compartmentalization. Recent studies have begun to approach depositional systems in this way (e.g. Ainsworth 2005; Howell et al. 2008), but in general, attention has focused on stratigraphic compartmentalization by flooding surfaces in shallow marine and paralic systems (Martinsen 1994; Ainsworth 2005), lobe connectivity in deepwater systems (e.g. Edman & Burk 1998; Milliken et al. 2008) and connectivity of channelized sandbodies in fluvial and deepwater settings (Larue & Hovadik 2006). Recent connectivity and modelling studies in these settings have also been used to test the interaction between faults and stratigraphy on fluid flow and compartmentalization (e.g. Ainsworth 2006; Manzocchi et al. 2007, 2008; Hovadik & Larue 2010). These studies have built on previous work (e.g. Bailey et al. 2002), and help guide more focused collection and integration of subsurface and outcrop data to improve geo-cellular modelling strategies and flow simulations. However, greater understanding of the stratigraphic components of compartmentalization is required, in addition to knowledge gained in deepwater and shallow marine reservoirs, in order to more fully characterize the variation in the style and severity of compartmentalization associated with different reservoir types and structural histories. In addition, fault zone studies have provided data on the flow properties of different fault rock species (e.g. Fisher & Knipe 1998, 2001; Sperrevik et al. 2002), and basic predictive algorithms that distil the complexity of fault zone architectures into more basic components that describe the distribution of fault rocks within a fault (e.g. Yielding et al. 1997). However, the more commonly applied fault seal algorithms calculate the distribution of flow-retarding clay content within fault planes using the modelled fault throw and stratigraphic content as critical input. Consequently, in addition to requiring a geometrically valid structural model, fault seal analysis is also reliant upon an accurate rendering of the geometry and property distribution of stratigraphy and depositional bodies

as input data. If these conditions are met it is therefore also possible to constrain and predict the severity and distribution of fault sealing properties and to more realistically model the compartmentalizing effect of faults on static trapping and production drawdown (e.g. Manzocchi et al. 1999, 2002; cf. Jolley et al. 2007; Zijlstra et al. 2007). As a contribution towards addressing these issues, in this paper we examine the relative sealing effects of stratigraphic and structural features, and their interplay in compartmentalizing dryland terminal fluvial reservoirs. To do this we use a well-constrained multi-disciplinary dataset drawn from the Heron Cluster fields (Heron, Egret and Skua) in the Central North Sea (Figs 1 & 2). These form part of an integrated development of seven fields tied back to a single processing facility over the Marnock Field (Pooler & Amory 1999). They are classified as high pressure and high temperature (HPHT) reservoirs, and occur at depths of up to 15 000 ft with initial reservoir pressures and temperatures in excess of 12 900 psi (890 bar) and 350 8F (180 8C) respectively. The main reservoir for all three fields is the sand-rich (50–80%) Triassic Skagerrak Formation, which in this area of the Central North Sea is dominated by dryland alluvial facies punctuated by widespread marsh and lacustrine intervals (Mckie & Audretsch 2005). In the early stages of appraisal and development the sand-rich, broadly fluvial nature of the Skagerrak succession led to the expectation that these reservoirs would be well connected both laterally and vertically. However, early production from the Heron Field indicated that the reservoir was in fact strongly compartmentalized. Initially this was considered to be caused by small, seismic-scale sealing faults. However, the required faults proved difficult to detect in the available seismic reflection data, and acquisition of geochemical and production data resolved a vertical component to the compartmentalization caused by stratigraphic barriers (McKie & Audretsch 2005). The purpose of this paper is to describe the nature of compartmentalization across the three fields in the Heron Cluster, and to critically assess the evidence for interaction between the stratigraphic and structural components of reservoir compartmentalization. We begin by describing the nature and depositional origins of the compartmentalizing intervals within the Skagerrak. We suggest that the shale-prone intervals have a variety of origins, some of which can be linked to extrinsic events affecting the stratigraphic evolution of the Skagerrak alluvial succession, and consequently represent barriers with widespread and predictable distribution. We then describe the style, timing and nature of the fault systems in the fields; and the nature, significance and likely distribution of flow retarding sub-structures within their fault

STRATIGRAPHIC COMPARTMENTALIZATION IN THE TRIASSIC HERON CLUSTER

167

Fig. 1. Location of the Heron Cluster field within the Central North Sea. (a) Regional distribution of Triassic basins within the Central North Sea. These earlier basins are cross-cut by a narrower Jurassic rift system with more focused extension. The Heron Cluster is located on the Jurassic Forties-Montrose High structure. (b) Detail of area outlined in yellow on the regional map showing the Triassic reservoirs on the Forties-Montrose High, including the producing cluster fields of Heron, Egret and Skua.

zones. We suggest that the intra-reservoir faults are unlikely to be capable of generating the observed flow/pressure barrier effects over either geological or production time-scales. Finally, we discuss the wider applicability of our results to other dryland reservoirs, and the general benefits of an integrated approach in the context of compartmentalization studies.

Gross depositional setting of the Triassic in the central North Sea Triassic depositional systems in NW Europe occur within a collage of linked rift basins, extending from eastern North America and West Africa, through the Iberian realm and west of Ireland to mid-Norway and eastern Greenland. Basins up to 4 km deep formed, with a variety of filling styles. To the north of the central North Sea area, between Greenland and Norway, a narrow, focused rift was initiated during early Triassic extension (Steel & Ryseth 1990; Tomasso et al. 2008). South of the central North Sea, in the southern North Sea area, the interaction of early Triassic extension with residual thermal subsidence of an older Permian basin and opening of Neotethys (Van Wees et al. 2000) led to a more diffuse rift

system with several phases of low magnitude, areally widespread extension (cf. Geluk 2005). Overall these Triassic basin-fills are dominated by alluvial facies, which vary from the deposits of large perennial rivers to more ephemeral systems, and are associated with variable proportions of aeolian facies, both perennial and ephemeral lacustrine facies, and evaporates. In addition, the southern Permian Basin also shows episodic marine penetration from Tethys to the south (via breaches in the remnant Variscan mountain chain), and in the middle Triassic the establishment of widespread marine carbonate facies of the Muschelkalk. The Triassic of the central North Sea area is dominated by thick alluvial successions (cf. Fisher & Mudge 1998) with no apparent marine penetration from the southern North Sea (Fig. 3). Although there is stratigraphic continuity between the two areas via the Danish sector (Michelsen & Clausen 2002), the Mid-North Sea High appears to have been an area of sufficiently lower subsidence rates that marine facies did not extend beyond this region (Fig. 3). Marine penetration may have also been suppressed by long-term sediment supply maintaining an alluvial gradient to the SE. The central North Sea area is largely lacking evidence for aeolian activity or evaporite deposition, and

168

T. MCKIE ET AL.

lake or clastic marine settings. Distal facies, where preserved, indicate fluvial termination in thinly bedded splay sandstones and floodbasin shales pervasively modified by soils, or in desiccated playa heterolithics. Perennial lacustrine facies are virtually absent. The overall depositional setting would therefore appear to comprise ephemeral to intermittent streams in a relatively dry climatic setting (albeit with climatic fluctuations cf. Ruffell & Shelton 1999), which formed an endorheic drainage into basinal playa and marshes. There is insufficient stratigraphic resolution to enable identification of the geometry of individual drainage systems and it is unclear whether they formed via the amalgamation of large fluvial fans (Moscariello 2005), more apron-like terminal alluvial plains (Aigner et al. 1995) or piedmont aprons (Smith 2000). However, the net result would appear to be the deposition of large aprons of transversely and axially supplied sediment into these basins.

Skagerrak stratigraphy

Fig. 2. Summary structure maps for the Heron, Skua (with slump structure highlighted in blue) and Egret fields. Scale bars 1 km.

palynological evidence from shales within the Skagerrak is indicative of episodic wetland conditions (Goldsmith et al. 2003). However, the widespread development of red-bed facies, calcic and vertic palaeosols, aeolian deposits and evaporites in adjacent basins also suggests prevailing semiarid conditions. Palaeogeographical reconstructions indicate that the fluvial systems of the Skagerrak are likely to have been terminal in character (Fig. 3b). They pass laterally into floodbasin facies or shallow marine carbonates and do not form major deltaic deposits associated with perennial

The variable behaviour of the Skagerrak reservoirs suggests that stratigraphic baffles and barriers are relatively common within the succession. These are likely to have multiple origins, and a synopsis of the stratigraphy and sedimentological architecture of the section is required in order to provide a framework for understanding the nature of these barriers. The Triassic stratigraphy of the Central North Sea area has been defined by Goldsmith et al. (1995, 2003) based on biostratigraphic calibration of wells from the J-Ridge area (Fig. 3a). The stratigraphy comprises an early Triassic succession of playa mudstone and fluvial sandstone (Smith Bank and Bunter Formations), overlain by the middle to late Triassic Skagerrak Formation. The Skagerrak Formation is composed of alternating sand-prone and mud-prone members, with the sandprone sections dominated by alluvial facies and the mud-prone members comprising a variety of nonmarine, basin-wide floodbasin facies (e.g. floodplain, marsh, playa and very rare lacustrine facies). Goldsmith et al. did not extend their analyses northwards into the Heron Cluster area, but additional biostratigraphic data between the J-Ridge and the Heron Cluster (29/5b-3Fz and 22/29-1S1) allow the dating of the Heron Shale at the top of the Skagerrak section in the Heron Cluster Fields (McKie & Audretsch 2005) as late Anisian/early Ladinian, and allow its correlation with the Julius Mudstone member in the J-Ridge area. This is supported by heavy mineral correlation, seismic (PSDM and regional megamerge datasets) and well log correlation of the main shale-prone packages across the central North Sea, and allows

STRATIGRAPHIC COMPARTMENTALIZATION IN THE TRIASSIC HERON CLUSTER 169

Fig. 3. Regional stratigraphic and palaeogeographical framework of the Triassic in the Central North Sea. (a) Regional stratigraphic framework extending from the Central North Sea southwards into the Danish and Dutch sectors of the North Sea. 1, Buntsandstein equivalent; 2, Judy Member; 3, Julius Mudstone Member; 4, Joanne Member; and 5, Heron Shale. The overall transition from Central North Sea ‘Skagerrak’ facies into the ‘Germanic’ succession of Bunter, Muschelkalk and Keuper occurs across the Mid North Sea High (MNSH) region, which appears to have acted as an area of reduced subsidence which prevented marine ingress into the CNS. Differential subsidence due to extensional faulting and halokinesis was most pronounced during deposition of the Smith Bank playa mudstones, with reduced differential subsidence during deposition of the largely post-rift Skagerrak Formation. Individual members of the Skagerrak record successive progradational pulses of terminal fluvial systems across the basin. (b) Middle Triassic palaeogeographical setting illustrating the dominantly alluvial facies in the Central North Sea and relationship to the marine carbonates of the Muschelkalk to the south in the Southern North Sea (line of stratigraphic section highlighted).

170

T. MCKIE ET AL.

the Skagerrak in the Heron Cluster to be placed in a regional stratigraphic and palaeogeographical framework. This provides a vital context to the compartmentalizing intervals seen in the reservoirs. Correlation of the shale-prone members of the Skagerrak into the Danish and Dutch sectors of the North Sea indicate that some are equivalent to marine flooding events in the southern North Sea (Michelsen & Clausen 2002), indicating that they have an origin linked to these events. The equivalent to the Rot Formation is a widespread playa shale separating the Bunter and Judy Members (Marnock Shale in the cluster area), whilst the marine middle to upper Muschelkalk equivalent in the central north sea is represented by anhydritebearing marsh facies of the Julius Mudstone Member/Heron Shale (Fig. 3a). Although the Rot section is largely composed of brackish marine to non-marine claystones (cf. Geluk 2005), Kovalevych et al. (2002) have shown that the Rot evaporites also have marine isotopic signatures, indicating widespread marine penetration many tens of kilometres into inland playa facies. Their ‘ponded’ distribution is suggestive of silled basins, possibly filled during minor eustatic sea-level rises (Aigner & Bachmann 1992). The intervening sandprone members of the Skagerrak Formation, which form reservoirs in a number of fields in the Central North Sea, comprise a range of fluvial facies. As mentioned above, aeolian and evaporitic deposits appear to be absent in these intervals, and palaeosols and biogenic traces (Porter 2006) are widespread (albeit within discrete stratigraphic intervals). This contrasts with the episodically recurring presence of evaporate facies in the time equivalent middle Triassic Rot and Muschelkalk, and suggests that there was sufficient freshwater influence, either through surface run-off or groundwater flow, in the Skagerrak basin to suppress aeolian and evaporate deposition and promote biogenic activity. In addition, whilst the early Triassic was marked by widespread aridity over NW Europe, marine penetration of the Southern North Sea in the middle and late Triassic, together with a general climatic amelioration during northward drift of Pangaea, appears to have been associated with increased vegetation cover in the Triassic across the Central North Sea, possibly in response to elevated groundwater and humidity.

Post Triassic tectonics and Skagerrak preservation The resultant Triassic stratigraphy is incompletely preserved as a result of a subsequent history of deformation and erosion. This history is fundamentally controlled by early Triassic extension and

halokinesis, regional thermal doming and erosion during the middle Jurassic, and upper Jurassic polyphase extension (Erratt et al. 1999; Davies et al. 2001). Early Triassic rifting in the Central North Sea resulted in widespread syn-depositional salt movement of the underlying Zechstein halite. This created the accommodation space to accumulate successions up to several thousand feet thick of Smith Bank mudstone in localized depocentres or ‘pods’ up to c. 2 –10 km across (Fig. 4; cf. Hodgson et al. 1992; Smith et al. 1993). The overlying Skagerrak Formation was largely deposited during post-rift thermal subsidence, and although the halokinesis triggered by rifting had largely ceased, these alluvial deposits show gentle thickening and thinning patterns which reflect ongoing movement due to sediment loading (Fig. 3a). In general this differential movement was insufficient to impede the Skagerrak fluvial systems and these sands are widespread across the basin. In areas of formerly thin halite the pods grounded on the underlying Rotliegend as a result of almost complete evacuation of salt from under the subsiding pods (Fig. 4). The flanks of these pods are heavily faulted as a result of subsequent salt withdrawal from the margins after the initial grounding of the pod axis (Hodgson et al. 1992; Smith et al. 1993). The widespread presence of disaggregation seams and phyllosilicate framework fault rocks in these grounded zones, with minor or no cataclasis (see below), suggests that the grounding occurred relatively early (or that the Triassic succession never achieved significant burial depths during the early burial history of the pods). Following grounding, basement strain was transmitted directly through the Triassic section without the dampening effects of an underlying ductile layer (Helgeson 1999). The Triassic depocentres have been regionally eroded by the middle Jurassic, ‘intra-Aalenian Unconformity’, which represents an episode of thermal doming and erosion of the central North Sea area (cf. Husmo et al. 2002). This doming was in response to mantle plume-induced thermal uplift (Underhill & Partington 1993), with the subsequent onlap of middle to upper Jurassic paralic deposits a result of deflation of the thermal dome combined with regional base level rise. Associated igneous activity is represented by a major magmatic centre in the Forties area (Latin et al. 1990; Smith & Ritchie 1993), and minor volcanic centres in the Acorn and Beechnut areas. The Heron Cluster fields lie c. 20 km to the south of the Forties centre and were close to the area of maximum uplift of the middle Jurassic dome. As a result the Triassic stratigraphy here is more deeply truncated by middle Jurassic erosion than the J-Ridge area c. 30 km further south (Fig. 3a) and the section is regionally truncated below the level of the Julius


171

Fig. 4. Illustrative seismic sections showing the preserved geometry of the Triassic in the Heron Cluster area. (a) Section across the Marnock, Skua and Seagull fields showing the broadly intact nature of Marnock, the faulted nature of Skua, which grounded on the margin of a basement terrace, and the deeply truncated Seagull structure which grounded on a shallow basement terrace. (b) Section between Seagull and Heron contrasting the grounded Seagull structure with Heron, which remains embedded in Zechstein halite. These structures record Triassic pod-style deposition during rift related halokinesis, middle Jurassic thermal doming and regional planation, and upper Jurassic extensional faulting.

Mudstone, which is only locally preserved in Heron and Egret. Upper Jurassic, polyphase extension resulted in the creation of tilted fault blocks at a range of scales. The Forties-Montrose High (Fig. 1) represents the footwall to a major tilted fault block, and subarial erosion and fluvial incision of this uplifted footwall occurred throughout the Jurassic. The crest of the structure is eroded to Smith Bank level, but off the crest of the structure the Skagerrak is variably preserved as truncated pods (Fig. 4). The Heron Cluster fields are located c. 10 km down-dip

to the east of the crest, and are the erosional remnants of the Triassic which were not fully onlapped by sediments until the Cretaceous. Smaller-scale faulting, detached into the underlying Zechstein halite, and exploiting the geometry of the Triassic pods resulted in the creation of the tilted fault blocks which define the fields in this area (Fig. 4). These show more localized footwall erosion of the Triassic, and also upper Jurassic coastal plain facies (Pentland Formation) which are variably distributed off the crest of these structures. The Triassic in the Heron Cluster area can therefore be regarded

172

T. MCKIE ET AL.

as erosional remnants of formerly basin-wide depositional systems, which in part owes its preservation to localized pod subsidence below the regional erosional base level during successive Jurassic tectonic events. The present day configuration has little bearing on the original stratigraphic architecture of the succession and needs to be filtered out in order to understand the nature of the stratigraphic compartmentalization.

Sedimentological framework The Skagerrak in the Central Graben has long been recognized as a dryland alluvial succession (Hodgson et al. 1992; Fisher & Mudge 1998), although the interpretation of the depositional systems has evolved with time from a simple braidplain model (Hodgson et al. 1992) to a more diverse suite of dryland depositional styles, and in some cases more wetland environments. In the Heron Cluster area the Skagerrak comprises a variably truncated Julius Mudstone (Heron Shale), Judy Member and Bunter equivalent sandstone, overlying a Smith Bank Mudstone succession. The Judy Member can be broadly subdivided into an upper interval, characterized by common channelfills and a variety of floodplain facies indicative of fluctuating water table conditions, and a lower interval showing deposition largely by unconfined flows, with a greater dominance of shales deposited in ephemeral lakes and ponds (Fig. 5a). The upper/ lower boundary is marked by a shale-prone interval, across which a change in heavy mineral assemblage occurs marking a transition from a restricted assemblage of Scottish Highland provenance to one indicative of a greatly enlarged catchment including Fennoscandia (Fig. 6, cf. Mange-Rajetzky 1995). This change is not abrupt, and does not have any precision as a stratigraphic marker, but instead appears to mark a relatively rapid shift in dominance between two co-existing fluvial drainage systems. It does, however, correspond to an overall change in depositional style in the Heron Cluster area from more dryland fluvial facies to a setting characterized by vegetated and episodically wetland environments. Although the Skagerrak shows systematic trends in depositional style both vertically and laterally, it can be rationalized into seven end-member facies associations. In the upper, fluvial section (Fig. 5) above the major provenance change the Skagerrak comprises: † Channel belt deposits. These are the dominant facies in the upper section and form vertically amalgamated intervals up to 100 ft thick composed of erosively based, cross-stratified and low angle plane-bedded sandstones with abundant lags of mudclasts and reworked calcic

palaeosol material scattered along basal contacts and foresets (Fig. 5d –f ). They are organized into fining-upward (or trendless) cycles 2– 10 ft thick capped by mud-prone, very fine-grained, palaeosol mottled sandstones with associated burrows, calcrete nodules and rhizoliths. Palaeocurrent data based on borehole images and oriented core are widely scattered, but shows a dominance of flow to the south and SW (Fig. 6). These deposits record channel-confined flow, with the presence of lag deposits, plane bedded and cross-stratified sands suggestive of fluvial systems comprising shallow bars and channels. There is no evidence of large-scale crossstratification or thick fining-upward cycles indicative of large rivers, and the overall impression throughout is of relatively shallow streams. The overall context of these deposits, in demonstrably terminal systems in a region prone to evaporate precipitation, suggests that they are largely the product of streams which had a highly erratic discharge which may have been ephemeral or intermittent in nature. However, the presence at some stratigraphic intervals of perennial lacustrine and palustrine shales may indicate more intermittent to continuous discharge, or sustained groundwater flow, in these sections. The palaeocurrent data, in combination with provenance information, indicate transverse flow from Fennoscandia towards the south west, and axial flow to the south east, presumably reflecting a combination of Scottish Highland and Fennoscandian derived streams. † Weakly-confined, biogenically disrupted sheet deposits occur in association with the channel belt deposits. These commonly have a structureless to disrupted appearance with abundant mottling by adhesive meniscate burrows and Skolithos, and common root traces (Fig. 5g). They comprise poorly sorted, very fine to fine grained sandstone containing a high proportion of intraformational clay and carbonate material. Sharp grain size breaks and palaeosol profiles, rather than clearly defined erosional surfaces, mark the bases of bedding on a scale of ,2 ft. Commonly, these form fining and coarsening-upward bedsets ,30 ft thick. Indistinct trough cross lamination and plane bedding is locally discernible. These sands are interpreted to represent weakly confined to unconfined fluvial or sheetflood deposition subject to pervasive palaeosol modification. Sharp bases and localized lags suggest some scouring and preservation of small-scale bedforms, but in general these deposits were probably deposited as poorly defined sheets. The association with channel facies, either within the same vertical section


173

Fig. 5. Stratigraphic section from Egret well 22/24d-10 showing the overall facies architecture of the Skagerrak (a) and representative core sections from across the Heron Cluster area showing the main facies. Core sticks (b) and (c) show Heron Shale palustrine facies, sticks (d), (e) and (f ) show upper Skagerrak channel belt sandstones, (g) illustrates bioturbated and rooted sheet deposits cemented by groundwater carbonate, (h) illustrates a well developed calcic palaeosol profile in floodplain mudrocks, core sticks (i) and ( j) show mudclast mantled scour and channel-fill deposits from the lower Skagerrak section, (k) shows micaceous, plane bedded and dewatered lower Skagerrak splay sandstones, and (l) shows heterolithic playa deposits. All core sticks 1 m long.

or as lateral equivalents (over distances of several kilometres to tens of kilometres) suggesting an origin as lateral and terminal splays to the channels (cf. Tooth 2005; Fisher et al. 2007). The evidence for shallow scouring could either be attributable to a network of shallow channels, or alternatively could be the product of erosion around floodplain vegetation for example, as shadow bars and scours (Rygel et al. 2004). † Floodplain deposits are dominated by siltstones and shales, forming intervals up to a few tens of feet thick. They are typically pervasively disrupted by root traces, burrowing and calcrete (Fig. 5h), although the burrows are typically restricted to monospecific assemblages of abundant and pervasive adhesive meniscate burrows (Porter 2006). The colouration of these palaeosols varies from red to green and grey, although in many cases this is not primary but instead related to reduction during hydrocarbon charge. Ped fabrics are common in the mud-rich intervals, and soil slickensides are locally present, but are generally uncommon. These mudrocks record vegetated floodplain conditions. The presence of calcrete, ped

fabrics and slickensides suggest generally well-drained conditions, but with common water table fluctuations. The overprinting by pervasive adhesive meniscate burrows suggests episodically higher soil moisture conditions favourable for insect larvae (cf. Smith & Hasiotis 2008; Smith et al. 2008) and indicates the former presence of abundant organic detritus derived from plant material. This bioturbation has been commonly interpreted as lacustrine in origin, leading to an over-emphasis of lacustrine processes in the Skagerrak (cf. Goldsmith et al. 2003). Although lacustrine facies are present, they are not common except within the major shales defining the Skagerrak stratigraphic subdivisions, and the bioturbation present can be entirely attributable to opportunistic organisms exploiting episodically high soil moisture. † Perennial lake and palustrine deposits form intervals up to c. 1 –200 ft thick of black coloured mudrocks with localized limestone (micrite) horizons ,1 m thick. This is the dominant facies within the Heron Shale and also occurs in thin intervals punctuating the upper 50 ft of the underlying Skagerrak fluvial facies. The basal 10– 20 ft of the Heron Shale, and the

174

T. MCKIE ET AL.

Fig. 6. Skagerrak facies architecture and palaeodrainage. (a) Stratigraphic section from 22/24b-5S1 showing vertical variations in percentage of tourmaline species based on shape, colour, internal structure and optical properties (modified after Mange-Rajetzky 1995). Fennoscandian detritus is highlighted in red, Scottish Highland tourmaline highlighted in grey. (b) Palaeocurrent data from the upper Skagerrak section in the Heron Cluster fields. The data are widely scattered overall, but show a dominance of flow towards the SW (from Fennoscandia) and to the south–SE, reflecting axial flow of streams draining the Scottish Highlands. (c) Middle Triassic palaeogeography with palaeo-drainage patterns synthesized from provenance and palaeocurrent data.

examples within the upper Skagerrak succession are typically more heterolithic in lithology (Fig. 5b, c) and show the local development of wave ripple lamination, bioturbation, algal laminites, flat pebble conglomerates and root trace disruption. Soft sediment deformation, in the form of clastic dykes and fissures are locally abundant. Palynological analysis of this facies indicates the presence of Lunatisporites, Calamospora and Protohaploxypinus. The black mudrocks record persistent subaqueous conditions in widespread perennial lakes. These water bodies appear to have been many tens of kilometres across (from the extent of the Heron Shale in well and seismic data), deep enough to preserve sediment below wave base (albeit with restricted wave fetch), and with reducing bottom sediments conducive to the preservation of organic matter. In the heterolithic sections more marginal conditions predominated, with the plant rooting, algal mats and early carbonate cementation indicative of

palustrine conditions. The palynological data suggest a mix of drought resistant shrubland and wetland environments, possibly with a maritime influence (cf. Kelber & van Konijnenburgvan Cittert 1998; Hubbard & Boulter 2000). In contrast to the upper section the lower Skagerrak interval (Fig. 5a) is almost entirely devoid of biogenic indicators, has limited palaeosol development, and appears to have a provenance restricted to the Scottish Highlands without any significant Fennoscandian input (Mange-Rajetzky 1995). An overall drier setting is envisaged with less extensive fluvial systems which supplied limited or infrequent discharge into the basin. The section broadly comprises: † Channel-fill deposits comprise sharply/ erosively based, fine-grained sandstones with very common mud flake lags, locally infilling gutter casts (Fig. 5i, j). Unlike the upper section reworked calcrete is absent. The sandbodies are generally ,10 ft thick and have an


internal fabric varying from structureless to indistinctly dewatered or dominated by horizontal to gently inclined planar lamination (with abundant mica). Cross bedding is uncommon. These deposits record channel-confined deposition under a variety of conditions. The abundance of mud flake lags may either be indicative of channel reworking of overbank muds or reworking of desiccated channel base mud drapes. The structureless to dewatered sandstones suggest catastrophic deposition, either through bank collapse or rapid deposition from waning, sediment-laden flows. † Terminal splay complexes comprise micaceous, very fine to fine grained, sandstones (Fig. 5k) organized into coarsening-upward cycles 5– 20 ft thick (in turn grouped into composite cycles up to 100 ft thick). At the base of these cycles are thin shales (,10 ft thick), which are well laminated or massive, commonly desiccated and rarely burrowed. Wave ripple lamination is locally developed and rare, delicately interlaminated shales and siltstones are locally present. The contact with the overlying sandstones is either sharp (with mud clasts or desiccation cracks), or gradational. The sandstones forming these cycles are dominated by plane bedding and climbing current ripple lamination, with dewatering fabrics locally developed. The tops of the cycles are either truncated by channel-fill deposits or sharply overlain by mudstone of a succeeding cycle. The basal shales to these cycles record subaqueous deposition, although the paucity of biogenic activity and common desiccation suggest ephemeral water bodies (although rare intervals of delicately interlaminated shales and siltstones indicate the persistence of water bodies through multiple flood events). The overlying plane bedded and climbing ripple laminated sandstones represent the product of deposition from shallow, unconfined flows, and their arrangement into coarsening upward cycles is suggestive of progradation. The variably sharp (with evidence of desiccation) or gradational contacts suggest that these progradational sand systems alternated between entering standing water bodies or subareally exposed surfaces. The overall setting appears to have been of a flood dominated system characterized by rapid deposition into ephemeral water bodies or dry lake floors. † Ephemeral lacustrine (playa) deposits form intervals up to c. 50 ft thick (with the Marnock Shale up to 200 ft), but commonly these are less than 3 ft thick. They are most common in the lowermost parts of the Skagerrak section. These deposits have a heterolithic lithology

175

(Fig. 5l), comprising mm to cm-scale alternations of very fine to fine grained sandstone and mudstone organized into metre-scale cleaning and coarsening-upward cycles. Wave and current ripple lamination, and desiccation fabrics are locally common, with dewatering fabrics present in thicker sandstone intervals. Biogenic structures and palaeosol fabrics are virtually absent. These mud-prone deposits represent low energy deposition in a distal setting characterized by episodic sand deposition and erratic water table conditions that fluctuated between subaqueous, damp and desiccated in response to ephemeral flood events. Overall the shales within the Skagerrak section provide a record of variable water table conditions ranging from drier playa conditions in the lower section, through well drained palaeosols to poorly drained marsh and episodic lacustrine conditions in the upper section. The fluvial systems were apparently terminal in character throughout both sections, but with more episodic discharge in a true dryland setting in the lower interval. The general depositional setting of the lower section may have been analogous to the terminal fluvial systems of the modern Lake Eyre basin (cf. Croke et al. 1996, 1998; Lang et al. 2004; Fisher et al. 2008). Heavy mineral analyses of the lower section (MangeRajetzky 1995) indicate a derivation from the west with extensive polycyclic reworking. In contrast, the upper section was characterized by more extensive streams and has a heavy mineral composition derived from an enlarged catchment including the Fennoscandian Shield (Fig. 6). Palaeocurrent data are indicative of south-westward directed, transverse flow from Fennoscandia and southward directed axial flow. Floodplain conditions were generally more frequently moist and capable of supporting plant and animal life (albeit limited in expression to pervasive adhesive meniscate burrows). The setting of the upper section may have been in part analogous to the semi-arid wetlands seen in the Okavango Delta in Botswana (Stanistreet & McCarthy 1993) and the Mesopotamian marshes of Iraq and Iran (Baltzer & Purser 1990; Baeteman et al. 2004) in having high water table environments within a generally semi-arid climatic setting characterized by seasonal fluvial discharge. However, the Skagerrak appears to have been lacking the major river systems seen in these locations. This change in character between the upper and lower Judy sections may be indicative of a change in precipitation from localized desert storms to more intermittent, or seasonal/monsoonal, hinterland runoff (van der Zwan & Spaak 1992) and the shale-prone package

176

T. MCKIE ET AL.

separating the two sections therefore records a significant basinal event in terms of climate and depositional response. The Heron Shale also represents a major change in basinal facies processes, with the widespread development of marsh facies. The lacustrine facies is restricted to this region and elsewhere marsh facies are present (with localized anhydrite recorded by Goldsmith et al. 2003, possibly indicative of drier, better drained conditions away from sites of groundwater flow). Despite the wide range of facies present within the Skagerrak succession, and gross alluvial setting, the overall sand:shale ratio of the succession (typically 50 –80%) and high proportion of channel bodies in the upper section would indicate that sandbody connectivity would in general be very high, with limited possibility of isolated bodies (King 1990; Larue & Hovadik 2006). However, within the shale-prone interval in the basal part of the upper Skagerrak sands typically comprise less than 30% of the c. 50– 200 ft thick succession, and within this interval poor lateral and vertical connectivity is likely. Lack of connectivity in this interval is likely to be compounded by the generally unconfined nature of the enclosed fluvial sands, and the lack of more erosive channel bodies which could facilitate vertical connectivity.

Structural framework The structural configuration of the Heron Cluster fields was developed by a combination of Triassic salt movement ‘grounding’ sediment pods, and reworking of the pod margins by upper Jurassic rift-related faulting. Each field differs in its overall form, and in the detail of the faulting it contains given the presence or absence of significant halite beneath the Triassic and the style and timing of salt movement. Structural interpretation and mapping of the fault arrays within the Skagerrak requires some caution. Seismic reflection data at Skagerrak level often contains peg-leg multiple energy that in places obscures the primary reflections and the imaging within ‘semblance’ cubes, and this can hamper fault interpretation (e.g. Fig. 7). Mapping of faults in the prominent base Cretaceous reflection only reveals the latest Jurassic structuration, and provides less information on the earlier intra-pod deformation that has been erosionally truncated. Many ‘edges’ seen in the base Cretaceous reflections, and semblance cube extractions near to them, are palaeo-topographic erosional features rather than fault scarps (e.g. scarp- and lee-slopes on sub-cropping Skagerrak beds, Figs 7b & 8), but recognition of these features provides useful information on the distribution of potential stratigraphic compartments (see below). Structure mapping at the base Cretaceous cannot therefore be

simply projected downwards. To identify the true details of the intra-Skagerrak fault system it is also necessary to interpret the intra-Triassic reflectors. The main aspects of the fault systems mapped in the Heron Cluster fields (shown in Fig. 2), are summarized below: † The Heron structure comprises a northward facing, upper Jurassic tilted fault block on the eastern flank of the Forties-Montrose High (Fig. 2). The structure has not grounded on basement and salt is present below the Triassic section across the field (Fig. 4b). A large salt ridge is located on the eastern part of the field in the footwall to the eastern bounding fault and this is likely to have exploited a former salt wall associated with the margins of a Triassic pod. Deformation associated with this ridge is localized and the Triassic appears to have broadly rotated as a coherent slab with relatively little internal faulting. The field contains two broadly conjugate arrays of en echelon faults with some breached relays, but no seismic-scale fault spans the full width of the field or isolates any of the wells. Following blanket perforation of the stratigraphy the crestal producers appear to have access to the entire in-place volume. This is borne out by 4D seismic shot seven years after field start-up, which shows a fieldwide, two-way-time (TWT) difference signal. † The Egret structure is a smaller Jurassic fault block, but in this case the structure has formed across a similarly scaled salt ridge. As a result Egret shows the deformation effects associated with this ridge across the field, expressed as an anticlinal flexure of the stratigraphy, cut at a high angle by a major westward-dipping fault system which defines the western boundary of the field. Internal structuration is by the intersection of penecontemporaneous NNE–SSW and east –west striking fault arrays, which span the field and subdivide it into a patchwork of discrete fault blocks (Fig. 2). In common with Heron the Egret structure has not grounded on salt, but is pervasively faulted as a result of the upper Jurassic structure forming across a Triassic faulted pod margin, with possible upper Jurassic reactivated salt movement. The faults offset reservoir section and isolate the wells into separate fault blocks. † The Skua structure is also defined by an upper Jurassic tilted fault block located against the eastern flank of the Forties-Montrose High. However, in this case the Triassic has grounded on the Rotliegend as a result of more pervasive withdrawal of the underlying Zechstein halite from a basement terrace (Fig. 4a). The internal structure of Skua (Fig. 2) is dominated by


177

Fig. 7. Illustrative seismic character of the Skagerrak reservoir in the Skua Field. (a) Peg-leg multiples of the Ekofisk Formation, overprinting reflectors at Skagerrak level. (b) Representative section summarizing the configuration of the Triassic and Jurassic section in Skua- relief on the top of the Skagerrak reflects an erosional scarp topography and not fault offsetting. This is overlain down-dip by fluvial Jurassic Pentland formation. Intra-Triassic faults tip-out before the Jurassic section, or are overstepped by Pentland channel/valley features.

east –west faulting related to the grounding on Rotliegend basement and is likely to have formed early in the burial history of the SkuaMarnock pod. Seismic scale faults in Skua are of sufficient size to offset the reservoir and/or juxtapose different sand layers. Some faults are sufficiently linked to create discrete fault blocks, although most of the blocks are not completely fault-bounded and contain ‘open’ margins where fluid and pressure communication between fault blocks would be possible around intervening blind fault tips. Significantly, fluids and pressures do not communicate between the upper and lower reservoir sands, despite sand –sand juxtapositions across several of the faults. This is examined in more detail below. † The Seagull structure is an undeveloped Triassic discovery to the south of Skua. It has a pod geometry, which in this case is entirely grounded on

the Rotliegend basement high which partly underlies Skua. As a result of limited preservation space on this basement high, the Skagerrak section is more deeply truncated by Jurassic erosion (Fig. 4a, b). The structure comprises a four-way dip closure with significant fault boundaries on all sides as a result of grounding and collapse of the pod flanks as salt was evacuated. Seagull shows the most pervasive inter-linked seismic-scale faulting, dominated by NNW– SSE striking faults, roughly coinciding with the trend of the basement faulting in the Rotliegend.

Compartmentalization The most compelling evidence for compartmentalization is the dynamic behaviour of these fields. In all cases initial production has shown connection to

178

T. MCKIE ET AL.

Fig. 8. Seismic attribute map of Skua field, showing seismically imaged subcrop pattern at base Cretaceous. (a) Seismic section showing the snapped horizon picks at top Marnock Shale (yellow); and the pre-cursor loop, lower zero-crossing and post-cursor loop of the base Cretaceous reflector (green, cyan, blue respectively). Where reflections from sand- and shale-prone Skagerrak layers subcrop against the base Cretaceous, this causes ‘tuning’ of reflection energy (constructive and destructive interference). Thus, the subcrop of a seismically resolvable Skagerrak layer is marked by a sharp oscillation in the two-way-time isochore between the zero-crossing and the post-cursor of the base Cretaceous. (b) Map of two-way-time isochore between the zero-crossing and post-cursor of the base Cretaceous reflection. Subcrops shown in dotted lines are mapped from this attribute, tied to wells and cross-checked against seismic section displays. Fault line pattern is an overlay of fault interpretations made from seismic sections and a range of other seismic attributes not illustrated in this paper.

limited volumes, although it was not immediately clear with the first development of Heron whether this was due to stratigraphic, structural, or combined effects. Fluid geochemistry shows significant vertical and lateral variations both between wells, and also along the stratigraphy penetrated by individual wells. In addition, the 4D seismic behaviour of each field is consistent with the degree of internal structural and stratigraphic heterogeneity, and the manner in which the field was developed.

Fault zone permeability structure Faults are generally mapped from seismic data and carried into geo-cellular models as discrete dislocation planes. In reality faults are not simple smooth surfaces – they are 3D zones of strain that develop during the evolution of a fault through its propagation, linkage, and slip consolidation phases

(cf. Wibberley et al. 2008, and references therein). These ‘fault zones’ are composed of smeared, disarranged and re-aggregated fragments of the host litho-stratigraphy (e.g. Foxford et al. 1998; Aydin & Eyal 2002; van der Zee et al. 2003; Davatzes & Aydin 2005; Kristensen et al. 2008) and a related assemblage of meso- to small-scale faults and fractures which host a variety of fault-rock types and mineral fills (e.g. Caine et al. 1996; Fisher & Knipe 1998; Beach et al. 1999; Odling et al. 2004; Berg & Skar 2005; Shipton et al. 2005). The total flow retarding/sealing effect of a fault is caused by the collective distribution of the re-arranged juxtapositions caused by lithological smearing and fragmentation, the increase of capillary entry pressures caused by permanent porosity and permeability collapse within fault rocks, and the extreme tortuosities imposed on potential fluid flow pathways by the complexity of the fault zone architecture. The exact


suite of fault rocks that is formed within a fault zone is related to the clay content and grain size of the protolith as well as the syn-kinematic and post deformation stress –temperature history (Fisher & Knipe 1998, 2001; Sperrevik et al. 2002; Fisher et al. 2003). For example, faults that were active in siliciclastic reservoirs at 3 km burial (90 8C) or more are prone to severe grain size reduction and porosity collapse through cataclasis, and further porosity occlusion through temperature-enhanced quartz cementation. Such faults form effective fault seals which compartmentalize reservoirs over production time-scales (e.g. Hesthammer et al. 2002; Zijlstra et al. 2007). Movement of these faults at shallower (lower temperature) conditions would cause brittle failure that may breach the older fault seals and/or form extensive open transmissive fracture networks (e.g. Barr 2007). Seismic data show that the majority of faults in the Skagerrak reservoirs described here experienced a single late Triassic movement phase and/or movement during Jurassic rifting. Microstructural and diagenetic observations confirm that these faults were active at relatively shallow (0–500 m) depths, and subsequently underwent passive burial within the reservoir to present day depths (see below). When faulting occurs at such near-surface, low stress conditions, the sands are generally underconsolidated and unable to support open fractures. Three basic deformation mechanisms dominate the formation of fault rocks under these conditions: (a) in high net-to-gross sands (0–15% clay), fault slip is accommodated by particulate flow in which individual sand grains slide over each other with negligible grain fracturing, and revert back to a packing order similar to the pre-deformation state once the fault has stopped moving to form fault rocks called ‘disaggregation’ zones. Permeability remains similar to the host sandstone; (b) with increasing amounts of clay particles within the sand (15–40% clay) the basic deformation mechanism is similar, except that framework grains passing over each other during particulate flow cause instantaneous dilation and influx of clay particles into the fault plane to form phyllosilicate-framework fault rocks (PFFR). This focused admixture of clay particles and framework grains in the fault causes a rapid decrease in porosity and permeability (Fisher & Knipe 1998, 2001); and (c) where host rock clay content is higher, concentrated into clay drapes or formed into discrete beds, the beds are plastically smeared into a fault plane, in some cases accompanied by injection of clay from the source layer to form clay or shale-smears. Clay smears are observed at all scales, and in detail probably form a large component of the fault rocks that are classified as PFFR in many existing fault rock databases. Permeabilities are reduced to less than

179

0.1 mD, below that of the source shale beds (Fisher & Knipe 2001; Eichhubl et al. 2005). Porosity and permeability may be further reduced in the fault zone during post-kinematic burial, since the concentrated presence of clay minerals enhances grain-contact quartz dissolution (Sverdrup & Bjørlykke 1992; Fisher & Knipe 1998, 2001; Sperrevik et al. 2002). Small-scale faults in Skagerrak drill cores form locally intense clusters of discrete to diffuse subplanar slip planes dominated by disaggregation and phyllosilicate framework fault rocks (PFFR) (Figs 9 & 10). In detail, the PFFR features are composed of various zones of re-packed framework grains and concentrated admixture of clay particles, and small- to micro-scale clay smears (Fig. 10a). Some larger-scale clay smears are also present in the cores, where individual shale beds have amalgamated into a composite, clay-rich strain zone composed of plastic deformation and secondary shear-bands which disaggregate and incorporate thin lenses of sand within the margins of the smear material (Fig. 9d). These small faults are occasionally seen to deflect around lithic fragments, and occasionally grade into granular injectite structures or show distorted geometries typical of early formation and subsequent passive deformation and/ or some degree of post-kinematic compaction (Figs 10b & 11). Patchy dolomite cement occurs locally within a few fault planes; cataclasite was not observed during core logging. Scanning- and back-scatter-electron microscopy (SEM, BSEM) analysis of fault rock specimens from Seagull and Skua cores show phyllosilicate framework fault rocks with minor framework grain breakage (mostly cracking of K-feldspar) in some faults, but no true cataclasites (Q. J. Fisher, pers comm. 2008). The fault rocks in these Skagerrak cores therefore formed as a result of particulate flow with only minor amounts of grain fracturing. In addition to mechanical compaction during burial (Fisher et al. 1999), the burial diagenesis experienced by the Skagerrak sands included the precipitation of K-feldspar, dolomite, chlorite, quartz, albite and illite, and late pressure solution of feldspar and quartz. K-feldspar overgrowth on detrital K-feldspar was one of the first authigenic minerals to precipitate, in micro-structural sites which are consistent with widespread observation of low temperature K-feldspar precipitation under nearsurface stress-temperature conditions (Kastner & Siever 1979; McKeever et al. 1992; Bjørlykke et al. 1995). Authigenic K-feldspar in the samples is seen to be subsequently overgrown by dolomite and quartz; fault rocks formed after K-feldspar precipitation, possibly during dolomite precipitation but before quartz and illite precipitation (Porter et al. 1998; Q. J. Fisher pers. comm. 2008). This

180

T. MCKIE ET AL.

Fig. 9. Skagerrak core photographs showing typical examples of small- to meso-scale faulting. (a– c) Pale disaggregation faults developed in sand-rich units, as thick (1– 10 cm) bands of fault rock and thinner discrete (1 – 5 mm) slip planes. These faults often form complex, mutually cross-cutting ladders of individual faults. (d) Meso-scale ‘composite’ clay smear, composed of plastically deformed clay containing deformed sand lenses, and concentrated clusters of secondary shear bands. Scale bars 0.05 m wide.

Fig. 10. Skagerrak core photographs showing typical examples of faults developed in the sands. (a) Classic phyllosilicate framework fault rock (PFFR) dominated faults. In detail each fault plane is a composite of zones of disaggregation (D), zones of admixture of clay particles into disaggregation fault rock (PFFR), and micro clay smears (CSm). Scale bar 0.1 m. (b) Early, shallow PFFR formation: liberation of lithic fragments from host matrix, into clay matrix (PFFR) supported granular flow zones. Scale bar 0.02 m.


181

Fig. 11. Early faulting fabrics in the Skagerrak. (a) and (b) Sand injection into dilatant disaggregation/PFFR faults. (c) Siderite nodules growing in/on a shear fracture in footwall to a meso-scale fault, cross-cut bedding and overprint more steeply dipping disaggregation faults. (d) Disaggregation/PFFR faults smear some clay clasts, but also deflected around and liberate small lithic pebbles from the host matrix. (e) Early disaggregation faults are occasionally seen to pass laterally into syn-kinematic fluid escape features. Scale bars 0.05 m.

indicates that the faults formed in poorly lithified sediments, under low effective stress conditions, at burial depths much less than 1 km. Despite some variability due to the effect of laboratory specimen expansion cracks, Hg-air pressure data for the fault rocks cluster around 400 psi and several thousand psi, up to c. 8500 psi (Porter et al. 1998). This probably reflects entry pressure measurement of the ‘admixture’ and micro-smear domains of the PFFR faults. Thus the PFFR fault rocks are likely to form baffles to production in the field but are unlikely to seal over geological time-scales. However, shale beds and shale smears derived from them clearly have the potential to form robust seals.

Fault seal potential A basic estimate of the clay content of a fault can be provided by calculation of ‘Shale Gouge Ratio’ (SGR; Fristad et al. 1997; Yielding et al. 1997; Fig. 12a). This algorithm estimates the amount of clay that is redistributed into the fault zone from the host sediments by mechanical admixture, assuming that the volume-percent of clay entrained into any point on a model-scale fault plane is equivalent to the proportion of shale within the stratigraphy that has slipped past that point. This seems to be a fair approximation, since comparison of SGR

predictions with geological mapping/logging of surface exposures, shows a reasonable overall match to the average clay content of natural faults (Doughty 2003; van der Zee & Urai 2005). Published calibration data for SGR-equivalent hydrocarbon retention (e.g. Yielding 2002; Yielding et al. 2010), suggest that faults formed at or buried to modest depths, are incapable of sustaining more than a few tens of psi across-fault pressure difference below 15% SGR, begin to impede fluid flow at around 25% SGR and only achieve stronger sealing capacities above 40% (perhaps retaining 1000 psi or greater at 50 –60% SGR-equivalent to the onset of clay smear formation). Triangle plots (also called juxtaposition diagrams, Knipe 1997), use well logs to calculate and display fault juxtaposition of sands and shales, and the distribution of clays and fault rocks as a function of throw. Figure 13a, b shows the juxtaposition of sands and shales, and SGR values for faults in the Skagerrak, indicating typical SGR values of 10 –15% (lower Skagerrak) and 25– 35% (upper Skagerrak). The plots show that as throw increases to 50 m or more, SGR trends towards a value equivalent to the Vshale of the Skagerrak sequence. SGR values on the faults therefore rapidly diminish below the theoretical ability of the faults to act as an effective membrane seal where sands are juxtaposed against sands, such that they would be flow baffles

182

T. MCKIE ET AL.

Fig. 12. Common algorithms, used to calculate clay content of fault zones include: (a) Shale Gouge Ratio (SGR), which assumes mechanical admixture of clay with sand in any given point within a fault zone, proportional to the stratigraphically held volume of shale that has slipped past that point of observation; and (b) Clay Smear Potential (CSP), which calculates the distance over which a plastically smeared shale remains intact within a fault zone, as a function of source bed thickness and fault throw. (from Jolley et al. 2007, modified after Yielding et al. 1997).

incapable of supporting the pressure drawdown experienced in the Heron Cluster fields (see below). However, the clay/shale smearing process shears, pumps and stretches clay from relatively pliable shale ‘source’ layers into the fault zone to generate more persistently very high clay content (e.g. Lehner & Pilaar 1997; Aydin & Eyal 2002; van der Zee et al. 2003; van der Zee & Urai 2005). In meso-scale to seismically imageable faults there is generally an area of complexity in the fault zone around a shale smear where multiple fault strands relay displacement across the shale layer, which deforms by a combination of flexural shearing and plastic shape change (Fig. 14). In section view shale layers essentially tongue into the fault and pinch-out some distance away from the source bed. Sand beds deformed under such shallow burial conditions are also smeared alongside the shale. Studies of small-scale, ‘soft’ sediment faulting show significant complexity in the 3D continuity of clay smear features and their lengths with respect to source-bed thickness (e.g. van der Zee et al. 2003; Childs et al. 2007; Kristensen et al. 2008). Notwithstanding difficulties in upscaling of material properties and deformation mechanisms from the scale of these observations to seismically-imageable fault dimensions, these

studies show that several source beds may feed into a single amalgamated smear, and that ‘holes’ may emerge in a smear where displacement becomes large enough to disengage smears from their source layers and they become entrained wholly or partially as ‘slugs’ within the fault plane. Where the shales occur as thin beds within a dominantly sand stratigraphy, or sand– shale layering is relatively fine (alternating cm –m thick beds), it can be argued that the smear and slug lengths will be up to orders of magnitude less than seismically imageable fault throw values and are thus accounted for within the SGR algorithm once stratigraphy is upscaled to the dimensions of a typical geo-cellular model. This implies that SGR can be used under the right circumstances as a catchall algorithm for estimating fault permeability (e.g. Childs et al. 2007). Given that clay smears have comparable or even lower permeability than their source shale beds, an understanding of the continuity of a smear on a fault surface should indicate the sealed portion of that fault. Thus, several published fault seal algorithms are designed to predict the continuity of a clay smear (seal) in faults formed in under-lithified sediments as a function of fault throw and the thickness of the source shale bed


Fig. 13. Diagrams showing juxtaposition of sands and shales (defined at 0.55 and 0.60 Vshale thresholds), and comparison of fault zone clay content predicted by SGR and CSP algorithms. Equivalent sealing capacities: Red-yellow, fully open to flow; Yellow-green, flow baffling; Blue, totally sealed. 1st column: Main sand and shale-prone packages (yellow/green, upper Skagerrak sandstones and shales respectively, pink/brown, lower Skagerrak sandstones and shales respectively); 2nd column: Vshale log with cut-off (red line); 3rd column: green bands show discrete shale beds derived from Vshale cut-off. Log depth in feet, throw in metres.

184

T. MCKIE ET AL.

There are three major shale-prone intervals in the upper Skagerrak that are capable of forming continuous sealing clay smears at this scale. These clay smears will have most impact on intra-reservoir flow units, diverting pressure/fluid communication around available open fault tips. However, the shale-prone interval that separates the upper and lower Skagerrak would be capable of sourcing a coherent clay smear in a fault of c. 3–500 ft of throw (Fig. 13c, d). Given that the disaggregation and PFFR fault rocks in the sand-prone intervals are relatively weak baffles, the sands would be expected to retain flow connectivity, whilst these major clay smears would be expected to preserve the integrity of the stratigraphic seal and stratabound compartmentalization of fluids and pressures between the upper and lower reservoirs in the fields.

Dynamic behaviour linked to structure Fig. 14. Schematic representation of typical fault zone architectures formed at shallow to modest burial depths. Ductile clay smears emanating from stratigraphic ‘source’ shales (green), are mantled by paired fault slip surfaces that relay fault displacement across the smear. The intervening sands (yellow-brown), are also smeared by inter-granular flow and repacking within the fault zone, and host a complex array of discrete slip planes and shear fractures that contain disaggregation and PFFR fault rocks. This style and arrangement of the various fault zone components is seen at all scales of observation (see text for details).

(e.g. ‘clay-smear-potential’ CSP, Bouvier et al. 1989; Fulljames et al. 1997; Fig. 12b). In their study of small- to meso-scale faults exposed in turbidite coastal cliffs, Childs et al. (2007) were unable to establish a simple relationship between fault throw and smear length or the disengagement and transport of smear material within the faults. However, in the case of decametre (.10 m scale) thickness shale units deformed by meso-scale to relatively small seismic-scale fault throws, general experience is that clay smears only appear to begin to lose their 3D coherence when the fault throw exceeds a value equivalent to 4 or 5 times the source bed thickness, such that clay smears of this scale are fully capable of separating different fluid types and pressures (e.g. Færseth 2006; Færseth et al. 2007). Under these circumstances the fault plane is completely sealed across the smear and stratigraphic continuity is essentially maintained (note that sands also smear and only contain weakly baffling fault rocks), such that the flow integrity of sands and sealing capacity of significant shales (i.e. stratigraphic compartmentalization) is preserved in the reservoir, despite faulting (e.g. Jolley et al. 2007).

Heron was the first field in the cluster to be developed, with three crestal producers. There are few seismic scale faults within this structure and none are of sufficient extent to link into compartments or discrete blocks. Material balance following blanket perforation of the entire stratigraphy indicates that the wells access the entire field. There is a field-wide TWT difference signal across the field and it would appear that the entire field is being depleted. In contrast, the heavily faulted Egret field was developed with a single vertical producer (Fig. 15a) perforated only in the upper Skagerrak section to avoid early water coning. The initial production from this well was below prognosis based on the assumption of bottom drive aquifer support. There was apparently no sign of pressure support, with c. 6000 psi of depletion after only c. 1.5 MMbbl was produced. Material balance calculations indicated that the well was connected to 8–12 MMbbl, which would be equivalent to the seismically resolvable fault compartment that the producer was drilled within. However, flattening of production decline occurred nine months after start-up and 6500 psi depletion, and following repeated shut-ins the well returned to production at higher rates and increased tubing head pressures (Fig. 15b). To date Egret has maintained this production behaviour and is clearly recharging from a larger connected volume than the initial material balance would suggest. The Heron and Egret fields represent two endmembers of structural style and fault densities within essentially similar reservoir rocks. The dynamic behaviour of Heron demonstrates that in an unfaulted Skagerrak reservoir the production wells do not see the full in-place volume without being perforated across the entire reservoir


185

Fig. 15. Egret production characteristics. (a) Illustrative seismic section showing the faulted nature of the Egret structure. (b) Oil rates from Egret producer 22/24d-P1. Egret production is repeatedly characterized by high, but rapidly falling oil rates following well shut-ins as a result of depletion of the fault compartments the P1 well is located in, followed by slow cross-fault re-charge during shut-in periods.

stratigraphy. The behaviour of Egret indicates that faults in this sand-rich succession do not seal under large differential pressures and are transmissible on a production time-scale.

Geochemical indications of fault block communication? There are geochemical variations across the Seagull and Skua fields which could be interpreted to be a reflection of fault compartmentalization. In both these fields two appraisal wells were drilled which encountered differences in Sr isotopic ratios and oil geochemistry. However, in both fields the wells also penetrate different parts of the Skagerrak stratigraphy over a lateral separation of several kilometres, and it is therefore uncertain whether these differences are due to genuine fault compartmentalization, stratigraphic compartmentalization or incomplete fluid mixing (given the possibility that over these separations diffusive mixing could conceivably be incomplete (England et al. 1987; Smalley et al. 2004). Compositional grading can

be discounted because the oils have similar densities, and the differences occur at similar depths. However, if there is unlikely to be an effective fault seal mechanism in the sand-rich Skagerrak the first scenario could be discounted, although significant baffling over several kilometres would retard diffusive mixing to result in the observed differences. Whilst these differences could argue for significant fault baffling, they are also consistent with the stratigraphy of the Skagerrak reservoir, and must remain as circumstantial evidence which supports, but does not prove, any interpretation of stratigraphic compartmentalization. The critical evidence remains the dynamic behaviour of the fields, which in the cases of Heron and Skua points towards effective stratigraphic boundaries.

Stratigraphic compartmentalization by shales Within the Skagerrak section there are shale-prone intervals which are regionally correlatable and are candidates for being field-wide compartment boundaries. The most significant of these are the

186

T. MCKIE ET AL.

basin-wide shale packages which define the members of the Skagerrak Formation. These appear to have an extrinsic origin linked to marine flooding events in the southern North Sea. Within the Judy Member the shale package separating the splay dominated lower section from the channel dominated upper section is well developed across the area and appears to have an extrinsic origin involving a major drainage re-organization of the Skagerrak fluvial systems and a change in discharge regime. There are also indications that shale packages within the upper and lower sections are capable of forming internal seals based on difference in fluid geochemistry in Skua and Seagull and the dynamic behaviour of the Heron and Skua fields (Fig. 16a –c). The Seagull field is undeveloped and lacking dynamic data to validate the indication of shale barriers. In Heron the initial production was from crestal producers which had only been perforated in the upper 200 ft of the Skagerrak section (Fig. 16). This was done to mitigate against aquifer ingress in a reservoir assumed to be fully connected both vertically and horizontally. Initial rates of pressure decline indicated that all three producers were connected to a limited volume (Fig. 16a, d). It was clear that the field was not cross-cut by seismic-scale faults which would be candidates for defining compartments (Fig. 2) and that the most likely explanation was that there were significant stratigraphic barriers within the field. A series of well interventions were undertaken which successively perforated deeper parts of the Skagerrak stratigraphy (McKie & Audretsch 2005). The first intervention added perforations in 22/30a-H3, c. 50 ft below the existing perforations (Fig. 16a, d), and straddling the shale interval separating the lower and upper Skagerrak sections. As a result of this intervention, tubing head pressures increased by c. 4000 psi. This was slightly lower than virgin pressure as a result of cross-flow of the newly perforated intervals into the overlying depleted zone. Decline curves and material balance indicated that the wells were accessing a greater connected volume, although still not the entire field volume. After three years production, with sufficient data to suggest that the possibility of aquifer ingress was lower than anticipated, a second intervention was undertaken to perforate deeper into the reservoir stratigraphy. These additional perforations were shot across a lower Skagerrak playa shale interval in 22/30a-H4, and accessed the lower part of the section c. 200 ft below the existing perforations in this well and deeper stratigraphically than the preceding H3 perforations (Fig. 16a, d). In this case tubing head pressures did not increase but, instead, the well which was approaching bubble point, saw a sharp reduction in GOR as a result of cross-flow into the overlying, more heavily depleted

interval. This response was seen across the crestal producers, indicating that lateral connectivity in the sand-rich fluvial and splay intervals is high. In addition, the geochemical fingerprint of the oil from Heron abruptly changed following this intervention (Fig. 16e). Drill stem tests from the Skua field tested intervals above and within the shale-prone section separating the upper and lower Skagerrak sections (Fig. 16a). DST 1 in 22/24b-7 sampled within this shale interval. Samples from DST 2a and DST 2b (commingled with 2a) were recovered from a channel-prone section c. 10 and 100 ft above (Figs 16a & 17). were analyzed from DST 2a and DST 3 (commingled with DST 2a). The DST 2a from 22/24b-9 sampled mainly the Lower Skagerrak splay facies whilst DST 3 included additional perforations in the basal upper Skagerrak (Fig. 17). The oil fingerprinting technique used was an analysis of the aromatic compounds in the C9-C10 alkane range using Multi-DimensionalGas-Chromatography (MDGC). Data are presented as relative ratios of components of relatively similar retention time. The geochemical analyses indicate that there are three significantly different groups of oils in the 22/24b-7 and 22/24b-9 wells (with differing PVT properties). The variations within 22/24b-7 occur between test samples over a 200 ft interval and are unlikely to be the result of compositional grading (there is no density difference between the top of the oil column and deeper levels) or incomplete mixing of different fluids in a reservoir without significant internal barriers. Over such a short vertical distance diffusive mixing within ,1 Ma would probably occur (England et al. 1987; Smalley et al. 2004) and since Skua was charged around c. 37–70 Ma (Winefield et al. 2005) any discrete oils should therefore have been completely mixed in the absence of stratigraphic barriers. Vertical compartmentalization is therefore strongly suggested across the shale-prone interval tested in 22/24b-7. There are also minor fingerprint differences between oils within individual groups of oils, which may also suggest that vertical mixing between sand bodies in this interval is poor. In contrast, the different oil fingerprints seen in 22/24b-9 compared to 22/24b-7 could be explained by stratigraphic or fault compartmentalization, or incomplete mixing over a lateral distance of 1.7 km. The data are inconclusive without clear sampling within the same stratigraphic interval. In Seagull the fingerprinting of oils recovered from five DST intervals in wells 22/29-2S1 and 22/29-3 again showed significant variations. In particular, in well 22/29-2S1 (Fig. 16a) the oil from DST1 was recovered across the same shale package separating the upper and lower sections,


Fig. 16. Correlation of wells (gamma, neutron, density) across the Heron Cluster area via Marnock, Skua, Seagull, 22/29-1S1, Heron and Egret (a) showing the overall change in depositional style through the Skagerrak section and datumed on the facies boundary between the lower splay and upper channel dominated sections. (b and c) Spider plots of test oil compositions for Skua and Seagull wells 22/24b-7 and 22/29-2S1, with compositions colour coded by test interval. (d) Pressure plot (tubing head pressure in black, bottom hole pressure in purple) and GOR plot (blue) for Heron Field, with the position of successive perforation intervals highlighted. (e) Spider plot of produced oil from Heron before and after the 1999 and 2002 perforations (with oil compositions colour coded according to perforation colours in (a) and (d).

188

T. MCKIE ET AL.

Fig. 17. Skua well section (gamma, density, neutron) with spider plots of test oils and 87Sr/86Sr plots from core derived residual salt analysis. Geochemical data and well tests have matching colours, and the Sr data conform to the core depths in the well sections. Horizontal production well 22/24b-S1 intersects the Skagerrak stratigraphy above the shale cluster defining the base of the upper Skagerrak section and has only produced oil with the geochemical fingerprint of the upper tests (2a and 2b) seen in 22/24b-7.

whilst DST2b was recovered from the upper interval, c. 150 ft above DST1. Again the test samples show significant differences between the upper and lower test intervals which suggest a compartmentalized vertical section in this well in a similar stratigraphic succession to that seen in Skua well 22/24b-7. In addition to the variations in oil composition within these fields the Sr isotopes from residual salt provide additional insights into possible barriers or baffles within these fields. In Skua the Sr isotopes from core from the two wells individually show invariant vertical profiles, but with a significant difference between each well (Fig. 17). This difference could be interpreted to reflect lateral fault compartmentalization, or severe baffling, which inhibited diffusion. However, when viewed stratigraphically, in combination with the oil geochemical data, it is likely that these data are a reflection of limited connectivity across the shale-prone interval separating the upper and lower sections, potentially in combination with reduced transmissibility across faults. The difference between the two Skua wells is more marked when viewed within the context of data from the adjacent Marnock field (Fig. 4), where 87Sr/86Sr values are identical to 22/24b-7 over a similar stratigraphic interval, while both differ from the section in 22/24b-9

below the shale package. This would appear to indicate that the upper Skagerrak section originally had a common formation water history through the upper Skagerrak section across the area, which differed from that below the medial shale package.

Shales as baffles In Heron, Sr data from core in the uppermost part of the section in 22/29-5RE show a change in gradient across a shale package which forms a correlatable precursor to the Heron Shale (cf. Webb & Kuhn 2003, Fig. 7). There is no step change in Sr isotopic values across this shale, suggesting that it either originally acted as a barrier during infill and the Sr isotopic profile was later homogenized by diffusion processes, or there was a change in the rate of hydrocarbon fill. However, the coincidence of a correlatable shale and the change in Sr isotopic gradient is an indication that this shale interval influenced the fill history, but that in this case functioned as a significant baffle, but not as a barrier. This example demonstrates that whilst some correlatable shale packages define fluid compartments, it does not necessarily follow that all correlatable shales are barriers. In this case it is likely that a formerly continuous mudstone was eroded by the overlying fluvial package and that there are minor breaches


in the continuity of this interval. In core the base of the overlying fluvial package has a significant accumulation of reworked calcrete detritus derived from the underlying interval, suggesting substantial downcutting and erosion. These reworked calcrete clast conglomerates are tightly cemented by post depositional groundwater calcite cementation, which could function as a barrier or baffle. This hypothesis was tested in Egret during a re-perforation exercise, when additional perforations were shot across such a thick, cemented section. In this particular case no difference was detected in well performance, suggesting that some cemented lags are not of sufficient extent to form significant barriers.

Shale-fault interaction in Skua Heron and Egret represent end-member reservoirs, with the former demonstrating stratigraphic compartmentalization in a relatively unfaulted field, and the latter showing through its dynamic behaviour that the internal faulting within the Skagerrak does not seal, and particularly with the large pressure differentials achievable in these high pressure fields. Skua represents a field in an intermediate situation-geochemical data indicate similar stratigraphic compartmentalization to Heron, but with seismic-scale faults present within the field (albeit not as pervasive as in Egret). The field was developed with a single horizontal producer adjacent to 22/24b-7, located in the upper Skagerrak section and blanket perforated along 1000 m (Figs 2, 17 & 18). The 4D signal in Skua (Fig. 18) shows a clear TWT difference around the S1 producer with a significant response clearly crossing the faults within the reservoir, indicating that as might be expected the faults in Skua do not form seals on a production time-scale. Whilst the 4D signal crosses faults these do tend to define the overall trend of the depletion signal, suggesting that they do form baffles. However, the boundaries to this depletion cannot be entirely attributable to faults, and locally there are sharp boundaries where no faults occur. These boundaries appear to follow the grain of the stratigraphy as it sub-crops beneath the Kimmeridge Clay top-seal at the eroded crest of the Skua structure (Figs 8 & 18). The 22/24b-S1 producer does not access the range of oils present within the field, and significantly, does not access the oil recorded by DST1 in 22/ 24b-7, which is only a few hundred feet from the heel of 22/24b-S1. The 4D signal induced by depletion through S1 is therefore the result of production from the upper Skagerrak section, and the shale-prone upper/lower boundary functions as an effective internal seal. This occurs despite the presence of faults within the structure, indicating that

189

faulting does not disrupt the integrity of the shale section as a fluid/pressure barrier.

Discussion Early development of the Heron Cluster was based on the assumption that such a high nett:gross fluvial reservoir would be well connected both vertically and laterally. Although a large part of the main reservoir section in all the fields in the cluster is indeed composed of fluvial deposits, their overall context is not one of exorheic fluvial systems flowing through the Central North Sea, but instead of endorheic systems which terminated in playa and marsh settings. This distinction is not necessarily discernable using data from a single well, or field (which would only indicate the fluvial nature of the succession), and requires the gross depositional environment to be established (North & Warwick 2007). The terminal setting of the Skagerrak appears to have resulted in the interfingering of sheet-like reservoir and non-reservoir facies over sufficiently large distances that the shale-prone horizons associated with fluvial retreat could be compartmentalizing (or form significant baffles). These compartmentalizing intervals can be broadly categorized into three main types: † Shales associated with major extrinsic events which affected the depositional system over a wide area. The Heron and Marnock shales can both be tentatively tied by limited biostratigraphic data to major marine transgressions in the southern Permian basin. They do not record marine penetration into the central North Sea, but do record more persistent (or frequent) water table rise, as playa in the more arid/evaporitic early Anisian, and as marshes and lakes in the wetter/more sustained marine flooding in the late Anisian/Ladinian. In the Heron Cluster the Heron Shale separates markedly different oils between the Skagerrak and the overlying Jurassic Pentland Formation (Winefield et al. 2005), indicating that the Heron Shale must have acted as a major stratigraphic barrier during oil generation and migration. In the Jade field, c. 30 km to the south, the equivalent Julius Mudstone also acts as an intra-Skagerrak barrier, supporting different reservoir pressures and contacts between the overlying Joanne and underlying Judy Members (Jones et al. 2005). In Heron this shale contains a minor reservoir sand interval (c. 50 ft thick) which also acts as a separate compartment. † At a finer scale, the shale-prone interval marking the transition from the lower dryland splay system to the overlying, ‘wetter’ and more channel-prone section within the Skagerrak

190

T. MCKIE ET AL.

Fig. 18. Skua 4D seismic TWT difference signal caused by stress arching in response to pressure depletion. This depletion cell is broadly constrained by flow baffling faults and by the subcrop of the shale-bounded Skagerrak stratigraphy against the top seal (as illustrated in Fig. 8).

records a substantial rise in water table (or frequency of episodes of high water table) coupled with a reduction in the basinward extent of the terminal fluvial systems in this area. This event was of a smaller magnitude than that recorded by the Heron and Marnock Shales (basin wide fluvial deposition did not cease), but still formed a widespread floodplain and marsh system into which a radically adjusted fluvial system drained under a different discharge regime (characterized by larger discharges and more perennial groundwater flow). This could be attributed to a change in climate in the headwaters of the fluvial system. The resultant shale package appears to form a significant barrier in all the fields where it occurs. † At the finest scale, particularly in the lower Skagerrak section, shales associated with splay cycles record the episodic expansion of playa/ marsh facies (Fig. 19). These intervals may have been a response to random, autocyclic terminal lobe switching over lateral scales of tens of kilometres or larger scale nodal avulsions, and possibly accumulated in areas of minor salt withdrawal. Alternatively, such shales could have been an extrinsic response to higher frequency water table or sediment supply fluctuations, although the lack of evidence for a sustained adjustment of the fluvial system to a new depositional regime suggests that this is less likely. The second well

intervention in Heron well 22/30a-H4 perforated across an example of this type of shale and encountered a different oil composition under virgin pressure. Jones et al. (2005) describe similar intra-Skagerrak shale intervals from the Jade field which also show geochemical differences across them, although in this field blanket perforation on the Skagerrak gives no indication as to whether these shales would be compartmentalizing during production drawdown. Such shales are not necessarily correlatable across the cluster area, are typically ,20 ft thick and prone to local fluvial erosion. However, it is also likely that within a stratigraphic section containing many of these packages that a subset would retain their local continuity across individual fields. Elsewhere in NW Europe similar Triassic alluvial reservoirs have been described. Whilst the sand:shale ratio of these reservoirs can be variable they typically comprise sheet-like terminal fluvial systems interbedded with floodbasin facies comprising playa, marsh, floodplain and lacustrine deposits. These can be field-wide in extent for example, in Alwyn (Harker et al. 2003), Wytch Farm (McKie et al. 1997, 1998; Hogg et al. 1999), Nevis (Farquharson & Gibson 2005), Beryl (O’Donnell 1993), Jade (Jones et al. 2005) and Chaunoy (Eschard et al. 1998), and can separate different fluids or


191

Fig. 19. Semi-regional shale packages which locally compartmentalize occur at the base of terminal splay progradational cycles in the lower Skagerrak section (a), and represent episodes of increased flood frequency and/or magnitude which resulted in playa expansion, and occur as floodplain facies in the upper section (b). The key to the log symbols is shown in (c). Abundant adhesive meniscate burrows and Skolithos signify that these intervals experienced episodically higher soil moisture levels compared to the background, well-drained calcic palaeosols. The example in (a) is from the lower Skagerrak section in Skua well 22/24b-9, and the example in (b) is from the upper section in Marnock well 22/24b-5S1.

support multiple contacts. Their presence suggests that field-wide, compartmentalizing shales, particularly in medial and distal reaches, may be a common element of endorheic, terminal fluvial reservoirs, in contrast to exorheic fluvial systems where channel belt connectivity and stacking in response to marine or perennial lake level fluctuations exerts a more fundamental control on reservoir behaviour. Although field-wide shales are not unknown in the latter, they do not show the regularity, frequency and continuity encountered in terminal systems. In addition, the most significant shales punctuating these terminal systems commonly also record major fluvial adjustments in response to hinterland climate change (which affected discharge), or episodes of sustained base level rise (or frequency and magnitude of base level rise episodes), and as a result the permeability structure and connectivity of the sand-prone layers commonly differs as a

result of changing fluvial style. This change in fluvial style can be reflected in the degree of channel confinement, channel belt dimensions, provenance or drainage pattern (axial v. transverse). Occasionally the response is subtler and in the case of the Triassic reservoir in the Corrib field, offshore Ireland, a field-wide shale occurs at a change in the isotopic composition of groundwater calcrete (Schmid et al. 2006). This change resulted from increased biogenic fractionation by vegetation, presumably reflecting a climatic change towards increased precipitation in the fluvial catchment. The basinal response was one of playa expansion, followed by a shift towards more channel confined deposition in response to increased fluvial discharge. Observations from outcrop studies reveal in greater detail the sheet-like nature of such terminal systems. Hornung & Aigner (1999, 2002a, b) have

192

T. MCKIE ET AL.

described late Triassic examples of such facies from the Stubensandstein Formation in Germany. These terminal fluvial deposits comprise sharplybased sand pulses c. 40 m thick extending several hundred kilometers into the basin, which back-step in response to base level rise with corresponding expansion of playa and floodplain facies during climatic wettening. The cycles are laterally extensive and in the subsurface the bounding shale tongues would clearly be capable of stratigraphic compartmentalization. Similarly, Geluk & Roehling (1997) have noted in the early Triassic Bunter Formation in the Netherlands and Germany that there is a high degree of correlatability of fluvial pulses in the Bunter section, which they ascribe to the terminal fluvial system response to climatic fluctuations. The sheet-like nature of the Bunter example may in part be due to widespread, low rates of thermal subsidence, but in more tectonically active basins mud-prone, floodbasin facies are still observed to episodically blanket fluvial systems over distances of several tens of kilometres, either through playa/ lake expansion (Nichols & Fisher 2007; Saez et al. 2007), or development of widespread floodplains and marsh. Whilst climate clearly exerted a major control on these systems, it is also likely that random nodal avulsions would also create geographical areas of fluvial abandonment and playa/marsh incursion. The shales resulting from such local abandonment could also have the potential to be field-wide, but could equally have limited areal extent. Such a process was inevitably part of the Skagerrak fluvial dispersal system and some examples clearly extend over several fields and mark major episodes of fluvial abandonment. However, in contrast to the shales associated with major extrinsic events the products of avulsion are not characterized by fundamental shifts in the depositional character of the system, and there are no diagnostic sedimentological criteria with which to distinguish extensive from more restricted examples. In addition, such shales tend to be relatively thin and as a result even extensive examples are prone to local fluvial erosion which would breach their capability as seals.

Conclusions The Skagerrak reservoirs in the Heron Cluster show aspects of their dynamic behaviour that may be typical of dryland fluvial reservoirs in general. Such reservoirs show interfingering of fluvial and playa/marsh facies over wide areas, which results in field-wide, compartmentalizing shale-prone intervals (Fig. 20). It is likely that this is an inherent property of this style of alluvial reservoir since outcrop analogues confirm that such interfingering

commonly occurs on a scale of 10 –100s of kilometres. Three scales of compartmentalizing shale are identified: † basin-wide shales representing the response to major changes in climate and base level. These are mappable over wide regions, are thickly developed and mark major changes in depositional environment within the basin. † semi-regional shales which signal widespread adjustments of the fluvial system to changing discharge regime. Their signature is more subtle, but may be marked by changes in fluvial style, provenance and floodplain drainage. † field-wide to semi-regional shales recording nodal avulsions, minor base level fluctuations and random development of floodplain lakes and ponds. Typically only a subset of these will be compartmentalizing; thin examples will be prone to fluvial erosion and, because they are discontinuous only a small proportion will have field-scale dimensions. The inability to detect the compartmentalizing subset of these shales drives a development strategy of blanket perforation of the stratigraphy, but at the cost of reducing the knowledge gained on reservoir behaviour by progressive well interventions. These compartmentalizing shales are widely correlatable, but it is likely that some examples may be prone to fluvial erosion and fail to function as compartment boundaries-correlatability does not equate with continuity. However, relatively thin shale intervals (which would not be anticipated to be laterally extensive based on their limited thickness) may be overlain by splay and sheetflood facies and lack the deep fluvial erosion which would breach their continuity, and as a result would compartmentalize. The fault zones in the Heron Cluster fields contain fault rocks that are typical of faults formed in under-consolidated siliciclastic sediments at shallow to near-surface conditions. These include disaggregation and phyllosilicate framework fault rocks which have a modest flow baffling effect and are incapable of withholding significant pressure differences, and ductile clay smears which have a similar or greater fluid/pressure sealing capacity to stratigraphic shales. Clay smears generated from the thicker, major sequence-bounding shales in the Skagerrak are likely to be continuous across the fault planes at throws up to 3–500 ft. Under these circumstances, a seismic-scale fault plane is completely sealed across the smear and stratigraphic continuity is essentially maintained (note that sands also smear) – such that the flow continuity of sands and sealing capacity of significant


Fig. 20. Conceptual model for the evolution of the Skagerrak reservoirs. (a) Deposition of the Skagerrak as sheet-like terminal fluvial systems draining off the Scottish Highlands and Fennoscandia. Changes in discharge regime and base level, possibly influenced by marine incursions into the southern north sea affecting local climate and lake level, resulted in widespread advance and retreat of these systems and deposition of extensive floodbasin muds which subsequently became compartmentalizing (cf. Fig. 3 colour key). (b) Regional thermal doming resulted in planation of the Skagerrak in the middle Jurassic. Thicker reservoir sections were preserved in areas of salt withdrawal (pods) where subsidence of the Skagerrak below the regional level of erosion preserved greater reservoir section. Where pods grounded the Skagerrak was more deeply truncated and experienced greater faulting and fracturing. At such shallow depths this is expressed by phyllosilicate fault rocks and disaggregation seams. (c) Upper Jurassic extensional faulting fragmented and rotated the pods into fault blocks, with associated resumption of halokinesis. (d) Schematic representation of fault zones in the Skagerrak. Ductile clay smears emanating from major sequence-bounding shales are expected to be continuous at fault throws up to 3 –500 ft. The clay smears have a similar or greater sealing capacity to the source shales, whilst disaggregation and phyllosilicate framework fault rocks that form in the intervening sands have a modest flow baffling effect. Thus stratigraphic continuity is essentially maintained (note that sands also smear) such that the flow continuity of sands and sealing capacity of significant shales (i.e. stratigraphic compartmentalization) is preserved in the reservoir across the faults.

194

T. MCKIE ET AL.

shales (i.e. stratigraphic compartmentalization) is preserved in the reservoir, despite faulting (Fig. 20). The authors acknowledge the contributions of Bob Elsinger and Oliver Kuhn to geochemical understanding of these fields, Paul Hague for 4D seismic analysis, Gerhard Makel for regional structural work and RDR (Quentin Fisher) for fault rock analysis. The reviews of Bryan Bracken and John Fisher, and editing by Bruce Ainsworth, are gratefully acknowledged.

References Aigner, T. & Bachmann, G. H. 1992. Sequence stratigraphic framework of the German Triassic. Sedimentary Geology, 80, 115–135. Aigner, T., Schauer, M., Junghans, W.-D. & Reinhardt, L. 1995. Outcrop gamma-ray logging and its applications: examples from the German Triassic. Sedimentary Geology, 100, 47– 61. Ainsworth, R. B. 2005. Sequence stratigraphic-based analysis of reservoir connectivity: influence of depositional architecture – a case study from a marginal marine depositional setting. Petroleum Geoscience, 11, 257–276. Ainsworth, 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. Aydin, A. & Eyal, Y. 2002. Anatomy of a normal fault with shale smear: implications for fault seal. American Association of Petroleum Geologists Bulletin, 86, 1367–138. Baeteman, C., Dupin, L. & Heyvaert, V. 2004. The Persian Gulf shorelines and the Karkeh, Kurun and Jarrahi Rivers: a geoarchaeological approach. Akkadica, 125, 155–215. Bailey, W. R., Manzocchi, T. 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. Baltzer, F. & Purser, B. H. 1990. Modern alluvial fan and deltaic sedimentation in a foreland tectonic setting: the lower mesopotamian plain and the Arabian Gulf. Sedimentary Geology, 67, 175–197. Barr, D. 2007. Conductive faults and sealing fractures in the West Sole gas fields, southern North Sea. In: Jolley, S. J., Barr, D., Walsh, J. J. & Knipe, R. J. (eds) Structurally Complex Reservoirs. Geological Society, London, Special Publications, 292, 309–336. Beach, A., Welbon, A. I., Brockbank, P. J. & McCallum, J. E. 1999. Reservoir damage around faults: outcrop examples from the Suez rift. Petroleum Geoscience, 5, 109–116. Berg, S. S. & Skar, 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. Bjørlykke, K., Aagaard, P., Egeberg, P. K. & Simmons, S. P. 1995. Geochemical constraints from formation water analyses from the North Sea and the Gulf Coast Basins on quartz, feldspar and illite

precipitation in reservoir rocks. In: Cubitt, J. M. & England, W. A. (eds) The Geochemistry of Reservoirs. Geological Society, London, Special Publications, 86, 33–50. Bouvier, J. D., Kaars-Sijpesteijn, C. H., Kleusner, D. F., Onyejekwe, C. C. & Van der Pal, R. C. 1989. Three-dimensional seismic interpretation, and fault sealing investigations, Nun River field, Nigeria. American Association of Petroleum Geologists Bulletin, 73, 1397–1414. Caine, J. S., Evans, J. P. & Forster, C. B. 1996. Fault zone architecture and permeability structure. Geology, 24, 1025–1028. Childs, C., Walsh, J. J. et al. 2007. Definition of a fault permeability predictor from outcrop studies of a faulted turbidite sequence, Taranaki, New Zealand. In: Jolley, S. J., Barr, D., Walsh, J. J. & Knipe, R. J. (eds) Structurally Complex Reservoirs. Geological Society, London, Special Publications, 292, 235–258. Croke, J. C., Magee, J. M. & Price, D. M. 1996. Major episodes of quaternary activity in the lower Neales River, northwest Lake Eyre, central Australia. Palaeogeography, Palaeoclimatology, Palaeoecology, 124, 1– 15. Croke, J. C., Magee, J. M. & Price, D. M. 1998. Stratigraphy and sedimentology of the lower Neales River, West Lake Eyre, Central Australia: from Paleocene to Holocene. Palaeogeography, Palaeoclimatology, Palaeoecology, 144, 331–350. Davatzes, N. C. & Aydin, A. 2005. Distribution and nature of fault architecture in a layered sandstone and shale sequence: an example from the Moab fault, Utah. In: Sorkhabi, R. & Tsuji, Y. (eds) Faults, Fluid Flow and Petroleum Traps. American Association of Petroleum Geologists Memoir, Tulsa, Oklahoma, 85, 153– 180. Davies, R. J., Tumer, J. D. & Underhill, J. R. 2001. Sequential dip-slip movement during rifting: a new model for the evolution of the Jurassic trilete North Sea rift system. Petroleum Geoscience, 7, 371–388. Doughty, 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. Edman, J. D. & Burk, M. K. 1998. An integrated study of reservoir compartmentalization at Ewing Bank 873, offshore Gulf of Mexico. Society of Petroleum Engineers, SPE Paper 49246, 653–668. Eichhubl, P., D’Onfro, P. S., Aydin, A., Waters, J. & McCarty, 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. England, W. A., Mackenzie, A. S., Mann, D. M. & Quigley, T. M. 1987. The movement and entrapment of petroleum fluids in the subsurface. Journal of the Geological Society, London, 144, 327–347. Eschard, R., Lemouzy, P., Desaubliaux, G., Bacchiana, C. & Smart, B. 1998. Combining sequence stratigraphy, geostatistical simulations, and production data for modeling a fluvial reservoir in the Chaunoy

STRATIGRAPHIC COMPARTMENTALIZATION IN THE TRIASSIC HERON CLUSTER Field (Triassic, France). American Association of Petroleum Geologists Bulletin, 82, 545–568. Erratt, D., Thomas, G. M. & Wall, G. R. T. 1999. The evolution of the central North Sea Rift. In: Fleet, A. J. & Boldy, S. A. R. (eds) Petroleum Geology of Northwest Europe; Proceedings of the 5th Conference. Geological Society, London, 63– 82. Færseth, R. B. 2006. Shale smear along large faults: continuity of smear and the fault seal capacity. Journal of the Geological Society, London, 163, 741–751. Færseth, R. B., Johnsen, E. & Sperrevik, S. 2007. Methodology for risking fault seal capacity: implications of fault zone architecture. American Association of Petroleum Geologists Bulletin, 91, 1231–1246. Farquharson, G. W. & Gibson, J. P. C. 2005. A significant satellite: the Nevis South Field in Block 9/13. In: Dore´, A. G. & Vining, B. A. (eds) Petroleum Geology: North-West Europe and Global Perspectives – Proceedings of the 6th Petroleum Geology Conference. Geological Society, London, 389–403. Fisher, J. A., Nichols, G. J. & Waltham, D. A. 2007. Unconfined flow deposits in distal sectors of fluvial distributary systems: examples from the Miocene Luna and Huesca systems, northern Spain. Sedimentary Geology, 195, 55–73. Fisher, J. A., Krapf, C. B. E., Lang, S. C., Nichols, G. J. & Payenberg, T. D. 2008. Sedimentology and architecture of the Douglas Creek terminal splay, Lake Eyre, central Australia. Sedimentology, 55, 1915–1930. Fisher, M. J. & Mudge, D. C. 1998. Triassic. In: Glennie, K. W. (ed.) Petroleum Geology of the North Sea. Blackwell, Oxford, 212–244. Fisher, Q. J. & Jolley, S. J. 2007. Treatment of faults in production simulation models. In: Jolley, S. J., Barr, D., Walsh, J. J. & Knipe, R. J. (eds) Structurally Complex Reservoirs. Geological Society, London, Special Publications, 292, 219– 234. Fisher, Q. J. & Knipe, R. J. 1998. Fault sealing processes in siliciclastic sediments. In: Jones, G., Fisher, Q. J. & Knipe, R. J. (eds) Faulting, Fault Sealing and Fluid Flow in Hydrocarbon Reservoirs. Geological Society, London, Special Publications, 147, 117– 134. Fisher, Q. J. & Knipe, 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. Fisher, Q. J., Casey, M., Knipe, R. J. & Clennell, M. B. 1999. Mechanical compaction of deeply buried sandstones of the North Sea. Marine and Petroleum Geology, 16, 605– 618. Fisher, Q. J., Casey, M., Harris, S. D. & Knipe, R. J. 2003. The fluid flow properties of faults in sandstone: the importance of temperature history. Geology, 31, 965–968. Foxford, K. A., Walsh, J. J., Watterson, J., Garden, I. R., Guscott, S. C. & Burley, S. D. 1998. Structure and content of the Moab fault Zone, Utah, USA, and its implications for fault seal prediction. In: Jones, G., Fisher, Q. J. & Knipe, R. J. (eds) Faulting, Fault Sealing and Fluid Flow in Hydrocarbon Reservoirs. Geological Society, London, Special Publications, 147, 87–103.

195

Fristad, T., Groth, A., Yielding, G. & Freeman, B. 1997. Quantitative fault seal prediction – a case study from Oseberg Syd area. In: Møller-Pedersen, P. & Koestler, A. G. (eds) Hydrocarbon Seals: Importance for Exploration and Production. Norwegian Petroleum Society (NPF) Special Publication, Elsevier, Amsterdam, 7, 107– 124. Fulljames, J. R., Zijerveld, L. J. J. & Franssen, R. C. M. W. 1997. Fault seal process: systematic analysis of fault seals over geological and production time-scales. In: Møller-Pedersen, P. & Koestler, A. G. (eds) Hydrocarbon Seals: Importance for Exploration and Production. Norwegian Petroleum Society, Special Publication, Elsevier, Amsterdam, 7, 51–60. Geluk, M. C. 2005. Stratigraphy and tectonics of PermoTriassic basins in the Netherlands and surrounding areas. PhD thesis, University of Utrecht. Geluk, M. C. & Roehling, H. G. 1997. High-resolution sequence stratigraphy of the lower Triassic ‘Buntsandstein’ in the Netherlands and northwestern Germany. Geologie en Mijnbouw, 76, 227– 246. Goldsmith, P. J., Hudson, G. & Van Veen, P. 2003. Triassic. In: Evans, D., Graham, C., Armour, A. & Bathurst, P. (eds) The Millenium Atlas: Petroleum Geology of the Central and Northern North Sea. Geological Society, London, 105– 127. Goldsmith, P. J., Rich, B. & Standring, J. 1995. Triassic correlation and stratigraphy in the South Central Graben, UK North Sea. In: Boldy, S. A. R. (ed.) Permian and Triassic Rifting in Northwest Europe. Geological Society, London, Special Publications, 91, 123– 143. Harker, S. D., Richardson, G., Sides, L. E. & Cooper, R. 2003. Alwyn North Triassic main gas-condensate: drilling deeper promotes production. Petroleum Geoscience, 9, 133– 143. Helgeson, D. E. 1999. Structural development and trap formation in the Central North Sea HPHT play. In: Fleet, A. J. & Boldy, S. A. R. (eds) Petroleum Geology of Northwest Europe; Proceedings of the 5th Conference. Geological Society, London, 1029– 1034. Hesthammer, J., Bjørkum, P. A. & Watts, 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. Hodgson, N. A., Farnsworth, J. & Fraser, A. J. 1992. Salt-related tectonics, sedimentation and hydrocarbon plays in the Central Graben, North Sea, UKCS. In: Hardman, R. F. P. (ed.) Exploration Britain; Geological Insights for the Next Decade. Geological Society, London, Special Publications, 67, 31–63. Hogg, A. J. C., Evans, I. J., Harrison, P. F., Meling, T., Smith, G. S., Thompson, S. D. & Watts, G. F. T. 1999. Reservoir management of the Wytch Farm oil field, Dorset, UK: providing options for growth into later field life. In: Fleet, A. J. & Boldy, S. A. R. (eds) Petroleum Geology of Northwest Europe, Proceedings of the 5th Conference. Geological Society, London, 1157– 1172. Hornung, J. & Aigner, T. 1999. Reservoir and aquifer characterization of fluvial architectural elements:

196

T. MCKIE ET AL.

Stubensandstein, Upper Triassic, southwest Germany. Sedimentary Geology, 129, 215–280. Hornung, J. & Aigner, T. 2002a. Reservoir architecture in a terminal alluvial plain; an outcrop analogue study (Upper Triassic, southern Germany); Part 1, Sedimentology and petrophysics. Journal of Petroleum Geology, 25, 3– 29. Hornung, J. & Aigner, T. 2002b. Reservoir architecture in a terminal alluvial plain; an outcrop analogue study (Upper Triassic, southern Germany); Part II, Cyclicity, controls and models. Journal of Petroleum Geology, 25, 151–178. Hovadik, J. M. & Larue, D. K. 2010. Stratigraphic and structural connectivity. In: Jolley, S. J., Fisher, Q. J., Ainsworth, R. B., Vrolijk, P. J. & Delisle, S. (eds) Reservoir Compartmentalization. Geological Society, London, Special Publications, 347, 219–242. Howell, J. A., Skorstad, A., MacDonald, A., Fordham, A., Flint, S., Fjellvoll, B. & Manzocchi, T. 2008. Sedimentological parameterization of shallow-marine reservoirs. Petroleum Geoscience, 14, 17–34. Hubbard, N. L. B. & Boulter, M. C. 2000. Phytogeography and paleoecology in Western Europe and Eastern Greenland near the Triassic-Jurassic boundary. Palaios, 15, 120– 131. Husmo, T., Hamar, G. P., Hoiland, O., Romuld, A., Spencer, A. M. & Titterton, R. 2002. Lower and Middle Jurassic. In: Evans, D., Graham, C., Armour, A. & Bathurst, P. (eds) The Millenium Atlas: Petroleum Geology of the Central and Northern North Sea. Geological Society, London, 129–155. Jolley, S. J., Dijk, H., Lamens, J. H., Fisher, Q. J., Manzocchi, T., Eikmans, H. & Huang, Y. 2007. Faulting and fault sealing in production simulation models: Brent province, northern North Sea. Petroleum Geoscience, 13, 321–340. Jones, A. D., Auld, H. A., Carpenter, T. J., Fetkovich, E., Palmer, I. A., Rigatos, E. N. & Thompson, M. W. 2005. Jade Field: an innovative approach to highpressure, high-temperature field development. In: Dore´, A. G. & Vining, B. A. (eds) Petroleum Geology: North-West Europe and Global Perspectives – Proceedings of the 6th Petroleum Geology Conference. Geological Society, London, 269–284. Kastner, M. & Siever, R. 1979. Low temperature feldspars in sedimentary rocks. American Journal of Science, 279, 435–479. Kelber, K.-P. & Van Konijnenburg-van Cittert, J. H. A. 1998. Equisetites arenaceus from the Upper Triassic of Germany with evidence for reproductive strategies. Review of Palaeobotany and Palynology, 100, 1– 26. King, P. R. 1990. Connectivity and conductivity of overlapping sand bodies. In: Buller, A. T., Berg, E., Hjelmeland, O., Kleppe, J., Torsaeter, O. & Aasen, J. O. (eds) North Sea Oil and Gas ReservoirsII. The Norwegian Institute of Technology, Graham & Trotman, 353 –362. Knipe, R. J. 1997. Juxtaposition and seal diagrams to help analyze fault seals in hydrocarbon reservoirs. American Association of Petroleum Geology Bulletin, 81, 187– 195. Kovalevych, V., Peryt, T. M., Beer, W., Geluk, M. & Halas, S. 2002. Geochemistry of early Triassic

seawater as indicated by study of the Rot halite in the Netherlands, Germany, and Poland. Chemical Geology, 182, 549 –563. Kristensen, M. B., Childs, C. J. & Korstga˚rd, J. A. 2008. The 3D geometry of small-scale relay zones between normal faults in soft sediments. Journal of Structural Geology, 30, 257 –272. Lang, S. C., Payenberg, T. H. D., Reilly, M. R. W., Kassan, J., Hicks, T. & Benson, J. 2004. Modern analogues for dryland sandy fluvial- lacustrine deltas and terminal splay reservoirs. Australian Petroleum Production and Exploration Association Journal, 44, 329–356. Larue, D. K. & Hovadik, J. M. 2006. Connectivity of channelized reservoirs: a modelling approach. Petroleum Geoscience, 12, 291 –308. Latin, D. D., Dixon, J. E., Fitton, G. & White, N. 1990. Mesozoic magmatic activity in the North Sea basin, implications for stretching history. In: Harman, R. F. P. & Brooks, J. (eds) Tectonic Events Responsible for Britain’s Oil and Gas Reserves. Geological Society, London, Special Publications, 55, 207–227. Lehner, F. K. & Pilaar, W. F. 1997. The emplacement of clay smears in synsedimentary normal faults: inferences from field observations near Frechen, Germany. In: Møller-Pedersen, P. & Koestler, A. G. (eds) Hydrocarbon Seals: Importance for Exploration and Production. Norwegian Petroleum Society (NPF) Special Publication, Elsevier, Amsterdam, 7, 39– 50. Mange-Rajetzky, M. 1995. Subdivision and correlation of monotonous sandstone sequences using highresolution heavy mineral analysis, a case study: the Triassic of the Central Graben. In: Dunay, R. E. & Hailwood, E. A. (eds) Non-biostratigraphical Methods of Dating and Correlation. Geological Society, London, Special Publications, 89, 23–30. Manzocchi, T., Walsh, J. J., Nell, P. & Yielding, G. 1999. Fault transmissibility multipliers for flow simulation models. Petroleum Geoscience, 5, 53–63. Manzocchi, T., Heath, A. E., Walsh, J. J. & Childs, C. 2002. The representation of two phase fault-rock properties in flow simulation models. Petroleum Geoscience, 8, 119– 132. Manzocchi, T., Walsh, J. J., Tomasso, M., Strand, J., Childs, C. & Haughton, P. 2007. Static and dynamic connectivity in bed-scale models of faulted and unfaulted turbidites. In: Jolley, S. J., Barr, D., Walsh, J. J. & Knipe, R. J. (eds) Structurally Complex Reservoirs. Geological Society, London, Special Publications, 292, 309– 336. Manzocchi, T., Carter, J. N., Skorstad, A. et al. 2008. Sensitivity of the impact of geological uncertainty on production from faulted and unfaulted shallow-marine oil reservoirs: objectives and methods. Petroleum Geoscience, 14, 3 –15. Martinsen, R. S. 1994. Stratigraphic compartmentalisation of reservoir sandstones: examples from the Muddy Sandstone, Powder River Basin, Wyoming. In: Ortoleva, P. J. (ed.) Basin Compartments and Seals. American Association of Petroleum Geologists Memoir, Tulsa, Oklahoma, 61, 273–296. McKeever, P. J., Carey, P. & Quinn, J. 1992. Authigenic K-Feldspar in the Permo-Triassic of northwest Britain: a pilot oxygen isotope study. In: Parnell, J. (ed.)

STRATIGRAPHIC COMPARTMENTALIZATION IN THE TRIASSIC HERON CLUSTER Basins on the Atlantic Seaboard: Petroleum Geology, Sedimentology and Basin Evolution. Geological Society, London, Special Publications, 62, 93– 96. McKie, T. & Audretsch, P. 2005. Depositional and structural controls on Triassic reservoir performance in the Heron Cluster, ETAP, Central North Sea. In: Dore´, A. G. & Vining, B. A. (eds) Petroleum Geology: North-West Europe and Global Perspectives – Proceedings of the 6th Petroleum Geology Conference. Geological Society, London, 285–298. McKie, T., Aggett, J. & Hogg, A. J. C. 1997. Sequence architecture of a Triassic semi-arid, fluvio-lacustrine reservoir, Wytch Farm Field, southern England. In: Shanley, K. W. & Perkins, B. F. (eds) Shallow Marine and Nonmarine Reservoirs: Sequence Stratigraphy, Reservoir Architecture and Production Characteristics. 18th Annual SEPM Gulf Coast Section Research Conference Proceedings, 197–207. McKie, T., Aggett, J. & Hogg, A. J. C. 1998. Reservoir architecture of the upper Sherwood Sandstone, Wytch Farm Field, southern England. In: Underhill, J. R. (ed.) The Development and Evolution of the Wessex Basin and Adjacent Areas. Geological Society, London, Special Publications, 133, 399–406. Michelsen, O. & Clausen, R. 2002. Detailed stratigraphic subdivision and regional correlation of the southern Danish Triassic succession. Marine and Petroleum Geology, 19, 563– 587. Milliken, W. J., Levy, M. & Strebelle, S. 2008. The effect of geologic parameters and uncertainties on subsurface flow: deepwater depositional systems. Society of Petroleum Engineers, SPE Paper 114099, 1–16. Moscariello, A. 2005. Exploration potential of the mature Southern North Sea basin margins: some unconventional plays based on alluvial and fluvial fan sedimentation models. In: Dore´, A. G. & Vining, B. A. (eds) Petroleum Geology: North-West Europe and Global Perspectives – Proceedings of the 6th Petroleum Geology Conference. Geological Society, London, 595 –605. Nichols, G. J. & Fisher, J. A. 2007. Processes, facies and architecture of fluvial distributary system deposits. Sedimentary Geology, 195, 75–90. North, C. P. & Warwick, G. L. 2007. Fluvial fans: myths, misconceptions, and the end of the terminal-fan model. Journal of Sedimentary Research, 77, 693 –701. O’Donnell, D. 1993. Enhancing the oil potential of secondary Triassic reservoirs in the Beryl A Field, UK North Sea. In: Spencer, A. M. (ed.) Generation, Accumulation and Production of Europe’s Hydrocarbons. Special Publication of the European Association of Petroleum Geoscientists, Oxford University Press, Oxford, 3, 37–44. Odling, N. E., Harris, S. D. & Knipe, R. J. 2004. Permeability scaling properties of fault damage zones in siliciclastic rocks. Journal of Structural Geology, 26, 1727– 1747. Pooler, J. & Amory, M. 1999. A subsurface perspective on ETAP- an integrated development of seven North Sea fields. In: Fleet, A. J. & Boldy, S. A. R. (eds) Petroleum Geology of Northwest Europe; Proceedings of the 5th Conference. Geological Society, London, 993–1006.

197

Porter, J., Fisher, Q. J., McAllister, E. & Knipe, R. J. 1998. Structural analysis of the Skagerrak Formation within Seagull wells. Rock Deformation Research Limited, Report 9848, 1–120. Porter, R. 2006. Ichnology and sedimentology of Triassic continental sequences: onshore and offshore UK. PhD thesis, University of Bristol, UK. Ruffell, A. & Shelton, R. 1999. The control of sedimentary facies by climate during phases of crustal extension: examples from the Triassic of onshore and offshore England and Northern Ireland. Journal of the Geological Society, London, 156, 779– 789. Rygel, M. C., Gibling, M. R. & Calder, J. H. 2004. Vegetation-induced sedimentary structures from fossil forests in the Pennsylvanian Joggins formation, Nova Scotia. Sedimentology, 51, 531– 552. Saez, A., Anadon, P., Herrero, M. J. & Moscariello, A. 2007. Variable style of transition between Palaeogene fluvial fan and lacustrine systems, southern Pyrenean foreland, NE Spain. Sedimentology, 54, 367– 390. Schmid, S., Worden, R. H. & Fisher, Q. J. 2006. Carbon isotope stratigraphy using carbonate cements in the Triassic Sherwood Sandstone Group: Corrib field, west of Ireland. Chemical Geology, 225, 137– 155. Shipton, Z. K., Evans, J. P. & Thompson, L. B. 2005. The geometry and thickness of deformation-band fault core and its influence on sealing characteristics of deformation-band fault zones. In: Sorkhabi, R. & Tsuji, Y. (eds) Faults, Fluid Flow and Petroleum Traps. American Association of Petroleum Geologists, Memoir, 85, 181– 195. Smalley, C., England, W. A., Muggeridge, A., Abacioglu, Y. & Cawley, S. 2004. Rates of reservoir fluid mixing: implications for interpretation of fluid data. In: Cubitt, J. M., England, W. A. & Larter, S. (eds) Understanding Petroleum Reservoirs: Towards an Integrated Reservoir Engineering and Geochemical Approach. Geological Society, London, Special Publications, 237, 99– 113. Smith, G. A. 2000. Recognition and significance of streamflow-dominated piedmont facies in extensional basins. Basin Research, 12, 399– 411. Smith, J. J. & Hasiotis, S. T. 2008. Traces and burrowing behaviors of the cicada nymph Cicadetta calliope: Neoichnology and paleoecological significance of extant soil-dwelling insects. Palaios, 23, 503–513. Smith, J. J., Hasiotis, S. T., Kraus, M. J. & Woddy, D. T. 2008. Naktodemasis bowni: new ichnogenus and ichnospecies for adhesive meniscate burrows (amb), and paleoenvironmental implications, Paleogene Willwood formation, Bighorn Basin, Wyoming. Journal of Paleontology, 82, 267–278. Smith, K. & Ritchie, J. D. 1993. Jurassic volcanic centres in the Central North Sea. In: Parker, J. R. (ed.) Petroleum Geology of Northwest Europe, Proceedings of the 4th Conference. Geological Society, London, 519– 531. Smith, R. I., Hodgson, N. & Fulton, M. 1993. Salt control on Triassic reservoir distribution, UKCS central North Sea. In: Parker, J. R. (ed.) Petroleum Geology of Northwest Europe; Proceedings of the 4th Conference. Geological Society, London, 547– 557.

198

T. MCKIE ET AL.

Sperrevik, S., Gillespie, P. A., Fisher, Q. J., Halvorsen, T. & Knipe, R. J. 2002. Empirical estimation of fault rock properties. In: Koestler, A. G. & Hunsdale, R. (eds) Hydrocarbon Seal Quantification. Norwegian Petroleum Society, Special Publication, Elsevier, Amsterdam, 11, 109–125. Stanistreet, I. G. & McCarthy, T. S. 1993. The Okavango Fan and the classification of subaerial fan systems. Sedimentary Geology, 85, 115 –133. Steel, R. & Ryseth, A. 1990. The Triassic– early Jurassic succession in the northern North Sea : megasequence stratigraphy and intra-Triassic tectonics. In: Hardman, R. F. P. & Brookes, J. (eds) Tectonic Events Responsible for Britain’s Oil and Gas Reserves. Geological Society, London, Special Publications, 55, 139– 168. Sverdrup, E. & Bjørlykke, K. 1992. Small faults in sandstones from Spitsbergen and Haltenbanken. A study of diagenetic and deformational structures and their relation to fluid flow. In: Larsen, R. M., Brekke, H., Larsen, B. T. & Talleraas, E. (eds) Structural and Tectonic Modelling and its Application to Petroleum Geology. Norwegian Petroleum Society Special Publication, Elsevier, Amsterdam, 1, 507– 517. Tomasso, M., Underhill, J. R., Hodgkinson, R. A. & Young, M. J. 2008. Structural styles and depositional architecture in the Triassic of the Ninian and Alwyn North fields: implications for basin development and prospectivity in the Northern North Sea. Marine and Petroleum Geology, 25, 588– 605. Tooth, S. 2005. Splay formation along the lower reaches of ephemeral rivers on the Northern Plains of arid central Australia. Journal of Sedimentary Research, 75, 636–649. Underhill, J. R. & Partington, M. A. 1993. Jurassic thermal doming and deflation in the North Sea: implications of the sequence stratigraphical evidence. In: Parker, J. R. (ed.) Petroleum Geology of Northwest Europe, Proceedings of the 4th Conference. Geological Society, London, 337– 346. Van der Zee, W. & Urai, 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. Van der Zee, W., Urai, J. L. & Richard, P. D. 2003. Lateral clay injection into normal faults. GeoArabia, 8, 501– 522.

Van der Zwan, C. J. & Spaak, P. 1992. Lower to Middle Triassic sequence stratigraphy and climatology of the Netherlands, a model. Palaeogeography, Palaeoclimatology, Palaeoecology, 91, 277– 290. Van Wees, J. D., Stephenson, R. A. et al. 2000. On the origin of the Southern Permian Basin, Central Europe. Marine and Petroleum Geology, 17, 43–59. Webb, P. J. & Kuhn, O. 2003. Enhanced scale management through the application of inorganic geochemistry and statistics. Society of Petroleum Engineers, SPE Paper 87458, 1– 15. Wibberley, C. A. J., Yielding, G. & Di Toro, G. 2008. Recent advances in the understanding of fault zone internal structure: a review. In: Wibberley, C. A. J., Kurz, W., Imber, J., Holdsworth, R. E. & Collettini, C. (eds) The Internal Structure of Fault Zones: Implications for Mechanical and Fluid-Flow Properties. Geological Society, London, Special Publications, 299, 5 –33. Winefield, P., Gilham, R. & Elsinger, R. 2005. Plumbing the depths of the Central Graben: towards an integrated pressure, fluid and charge model for the Central North Sea HPHT play. In: Dore´, A. G. & Vining, B. A. (eds) Petroleum Geology: North-West Europe and Global Perspectives—Proceedings of the 6th Petroleum Geology Conference. Geological Society, London, 1301–1315. Yielding, G. 2002. Shale Gouge Ratio – calibration by geohistory. In: Koestler, A. G. & Hunsdale, R. (eds) Hydrocarbon Seal Quantification. Special Publication of the Norwegian Petroleum Society, Elsevier, Amsterdam, 11, 1 –15. Yielding, G., Freeman, B. & Needham, D. T. 1997. Quantitative fault seal prediction. American Association of Petroleum Geologists Bulletin, 81, 897– 917. Yielding, G., Bretan, P. & Freeman, B. 2010. Fault seal calibration: a brief review. In: Jolley, S. J., Fisher, Q. J., Ainsworth, R. B., Vrolijk, P. J. & Delisle, S. (eds) Reservoir Compartmentalization. Geological Society, London, Special Publications, 347, 243 –255. Zijlstra, E. B., Reemst, P. H. M. & Fisher, Q. J. 2007. Incorporation of fault properties into production simulation models of Permian reservoirs from the southern North Sea. In: Jolley, S. J., Barr, D., Walsh, J. J. & Knipe, R. J. (eds) Structurally Complex Reservoirs. Geological Society, London, Special Publications, 292, 295 –308.

Prediction of stratigraphic compartmentalization in marginal marine reservoirs R. B. AINSWORTH Australian School of Petroleum, Centre for Tectonics, Resources and Exploration (TRaX), University of Adelaide, Adelaide 5005, Australia (e-mail: [email protected]) Abstract: Marginal marine depositional systems exhibit stratigraphic reservoir compartmentalization potential at three hierarchical scales. At each of these scales, stratigraphic compartmentalization potential can be related to the dominant depositional processes and accommodation:coarse sediment supply ratio (A/S) that are acting at the time of deposition. All three orders of compartmentalization potential must be considered in order to define optimal field development plans and completion strategies. The lowest order of compartmentalization is usually at the inter-parasequence scale. The parasequence is represented by a conformable succession of strata separated by marine flooding surfaces and as such it generally defines the basic flow unit in marginal marine systems. In systems tracts associated with relatively high A/S ratios, for example late Lowstand, Transgressive and early Highstand (steeply rising shoreline trajectories), vertical compartmentalization potential is relatively high because of the enhanced preservation potential of flooding surface shales under these conditions. In systems tracts associated with relatively low A/S ratios, for example late Highstand, Falling-stage and early Lowstand (flat, slightly rising and falling shoreline trajectories), vertical compartmentalization potential of parasequences is reduced because the potential for erosion of flooding surface shales by overlying deposits is high and hence potential for vertical sand–sand contact between parasequences is enhanced. The second level of compartmentalization hierarchy is the inter sand-body scale. Individual sand bodies are defined within parasequences. The lateral connectivity of these sand bodies is a product of the dominant depositional processes active at the time of their deposition (wave, tidal, fluvial). Wave-dominated systems tend to produce more laterally continuous sand bodies, fluvial-dominated systems more laterally restricted sand bodies and tide-dominated systems both laterally continuous and laterally restricted sand bodies. Vertical compartmentalization potential of these reservoir sand bodies is related to A/S regime. In high A/S regimes, sand bodies are more likely to be disconnected or compartmentalized. In low A/S regimes, erosional amalgamation of sand bodies is more likely thereby leading to lower compartmentalization potential. The third order of potential stratigraphic compartmentalization is the intra sand-body scale. This scale is represented by intra sand-body heterogeneities such as dipping or horizontal shales, carbonaceous-rich beds or laminae, shale abandonment plugs of channels and carbonate concretions. In high A/S regimes the preservation potential of these heterogeneities is relatively high leading to an enhanced potential for intra sand-body compartmentalization. Lower A/S regimes result in a greater likelihood of lateral and vertical erosion of these heterogeneities leading to a higher potential for reservoir connectivity.

Reservoir compartmentalization is a complex subsurface uncertainty that is the product of the combination of stratigraphic architecture, structural architecture, fault permeability and diagenesis. All of the above components must be considered to analyse reservoir compartmentalization fully (Knipe et al. 1998; Walsh et al. 1998; Ainsworth 2005, 2006; Jolley et al. 2007; Manzocchi et al. 2008; Hovadik & Larue 2010; McKie et al. 2010). In any compartmentalization study, the stratigraphic architecture of the reservoir forms the fundamental building block upon which the other compartmentalization uncertainties (structural, fault seal and diagenesis) must be applied. In fields with high fault densities, sealing faults and complex and heterogeneous diagenetic histories, stratigraphic

architecture can be of minor importance in determining the overall connectivity of the field. However, at the other end of the spectrum, in fields with low fault densities and simple burialrelated diagenetic histories, stratigraphic architecture can be the fundamental control on reservoir compartmentalization. In these scenarios, the primary depositional fabric and stratigraphic architecture of a depositional system defines the sandbody connectivity network and hence determines the stratigraphic compartmentalization potential of the reservoir. It is important to be able to predict stratigraphic compartmentalization potential in all types of depositional systems from different depositional environments. Channelized fluvial and deep marine


200

R. B. AINSWORTH

depositional systems have been the focus of many compartmentalization and reservoir connectivity studies (e.g. Budding et al. 1992; Cook et al. 1994; Jones et al. 1995; Mijnssen 1997; Keogh 2002; Bailey et al. 2002; Larue & Friedmann 2005; Larue & Hovadik 2006; Hovadik & Larue 2007; Manzocchi et al. 2007). However, stratigraphic compartmentalization potential in marginal marine depositional systems has received less attention. Notable exceptions include; Martinsen (1994), Ainsworth et al. (1999), Tye et al. (1999), Ainsworth (2003, 2005) and Howell et al. (2008). Given that these marginal marine deposits also form the reservoirs for a large proportion of global hydrocarbon reserves (Ahlbrandt et al. 2005), it is important to understand the key controls on stratigraphic compartmentalization potential in these coastal systems. The objective of this paper is to describe the underlying controls on stratigraphic connectivity in marginal marine systems and to provide a process for analysing and predicting the potential for stratigraphic compartmentalization in these types of reservoirs.

Importance of stratigraphic subdivision and correlation In order to sensibly analyse reservoir compartmentalization potential, a reservoir succession must be subdivided into key stratigraphic units. Sequence stratigraphy (Fig. 1; van Wagoner et al. 1990; Posamentier & Allen 1999) represents an excellent tool that when applied at reservoir scale, can be utilized to define correlatable units in marginal marine systems (Bryant & Flint 1993). Correlation of strata is the first step in defining potential connectivity or predicting compartmentalization. Within marginal marine systems, laterally extensive, correlatable shale intervals generally form the most efficient barriers to vertical fluid flow. Key shale layers occur at maximum flooding surfaces and parasequence flooding surfaces (Fig. 1). Carbonate cemented intervals may also be associated with the tops of sands bounded by these flooding shales and hence may enhance their ability to restrict vertical fluid flow (Taylor et al. 2000; Alsop & Ainsworth 2006; Howell et al. 2008). Parasequences, therefore, generally represent the key flow unit in the marginal marine realm (Bryant 1996; Ainsworth 2003, 2005; Larue & Legarre 2004). The second key issue with regards to connectivity is the correlation or otherwise of the sand bodies that occur within the parasequence framework. Connectivity of the sand bodies within parasequences is termed here ‘inter sand-body connectivity’ (Fig. 2).

Prediction of stratigraphic compartmentalization In terms of stratigraphic compartmentalization prediction, marginal marine deposystems produce a natural three-tiered hierarchy which can be related to primary depositional processes (Fig. 2). The lowest order of compartmentalization is the interparasequence level; the second order is the inter sand-body scale and the third order is the intra sandbody scale. This paper concentrates on marginal marine systems, hence, the levels of hierarchy classification do not correspond exactly with those of previous workers who have focused on reservoir heterogeneities in predominantly fluvial systems (Fig. 3). The key difference with fluvial classification schemes is that in fluvial systems, the parasequence scale of hierarchy (clearly defined in marginal marine systems) is often not easy to identify and hence does not usually produce a readily recognizable lithological unit. This study does not include microscopic heterogeneities (Fig. 3) since they are not considered to be a function of primary depositional processes acting in the marginal marine zone. These microscopic heterogeneities would have to be considered in a separate diagenetic study. Lasseter et al. (1986) described reservoir heterogeneities which impact fluid flow in reservoirs and also divided them into three scale subdivisions (Fig. 3). The parasequence scale detailed herein fits within the large scale of Lasseter et al. (1986). The inter sand-body scale is equivalent to both the large and medium scales of Lasseter et al. (1986; Fig. 3). Lasseter et al. (1986) include diagenetic effects and silts and clays of centimetre dimensions in their small scale category. The intra sand-body scale described in this paper does not account for pore-scale diagenetic heterogeneities as it is purely focused on heterogeneities generated by primary depositional processes. It therefore crosses both the medium and small scale zonation described by Lasseter et al. (1986; Fig. 3). Weber (1982) also discussed key heterogeneities that could influence fluid flow on the inter sand-body to intra sand-body scales described in this paper. He further developed this work and adapted the hierarchical sequence of heterogeneities based on scale, and developed by Pettijohn et al. (1973; Fig. 3), into a classification of seven heterogeneities based on size, genetic origin and influence on fluid flow (Weber 1986; Fig. 3). The parasequence scale utilized in this paper fits within the genetic unit boundary scale defined by Weber (1986; Fig. 3). The inter sand-body scale used in this paper combines the genetic boundary scale and the genetic unit scale of Weber (1986; Fig. 3). The intra sand-body scale encompasses both the baffles within genetic units defined by Weber

STRATIGRAPHIC COMPARTMENTALIZATION PREDICTION

201

Fig. 1. (a) Schematic marginal marine lithological cross-section and well log with key sequence stratigraphic surfaces. (b) Sequence stratigraphic interpretation of (a) showing systems tracts and key sequence stratigraphic surfaces. Note that in the marginal marine zone laterally extensive parasequence flooding surfaces commonly vertically separate reservoir sands and delineate reservoir flow units. This sequence stratigraphic breakdown is essential for predicting compartmentalization potential in coastal systems. No scale intended.

202

R. B. AINSWORTH

Fig. 2. Marginal marine compartmentalization hierarchy. The lowest order of compartmentalization is the inter-parasequence scale (see Fig. 1). The second order is the inter sand-body scale. At this scale it is critical to identify the key depositional processes acting at the time of deposition as these define the lateral continuity of sand bodies. The smallest scale is the intra sand-body scale where heterogeneities such as horizontal and dipping shales, concretions and shale abandonment plugs in channels.

Pettijohn et al. (1973) NA Width = 1 to 10 km Thickness = 100’s m Width = 100’s m Thickness = 10’s m

Lasseter et al. (1986)

Weber (1986) 1) Faults 2) Genetic unit boundaries 3) Permeability zonation within genetic units

NA Large scale heterogeneities

Inter sand-body Medium scale heterogeneities

4) Baffles within genetic units

Thickness = 1cm–10’s m

5) Lamination, cross-bedding

Thickness = 10–100’s µm

6) Microscopic heterogeneity

NA

7) Fracturing

This paper. Marginal Marine Systems NA Inter-parasequence

Intra sand-body Small scale heterogeneities NA NA

NA

Fig. 3. Comparison of reservoir heterogeneity schemes impacting fluid flow within a reservoir. Note that Pettijohn et al. (1973) utilized a geometry based classification scheme, which was modified by Weber (1986) to include faults and fractures. Scales used by Pettijohn et al. (1973) are approximate. Lasseter et al. (1986) further simplified these schemes. Also note that the hierarchy utilized in this paper is applied specifically to marginal marine systems. It therefore, crosses the scale definitions of the other authors who had mainly focused on fluvial channelized systems. Microscopic heterogeneities are not considered in this paper since they are not primary sedimentary structures related to marginal marine deposition. NA, not applicable.


(1986) and the lamination and cross-bedding heterogeneities (Fig. 3).

Role of accommodation:coarse sediment supply ratio (A/S) Compartmentalization potential at all three scales described in this paper (Fig. 2) can be related to the relative sea-level curve and to accommodation: coarse sediment supply ratios acting in the coastal zone at the time of deposition. Accommodation: coarse sediment supply ratio (A/S; Swift & Thorne 1991; Cross et al. 1993; Schlager 1993; Ainsworth 2003; Ainsworth et al. 2008) has been well documented as being a controlling factor in defining reservoir connectivity in the non-marine realm (Geehan 1993; Legaretta et al. 1993; Shanley & McCabe 1994; Posamentier & Allen 1999). In the marginal marine zone, the A/S curve is intimately linked to the relative sea-level curve (Fig. 4). Figure 4a shows the relative sea-level curve at a coastal location and the associated rate of accommodation change curve (Fig. 4b). The sediment supply curve at the coastline is shown in Figure 4c. Changes in sediment supply illustrated in this figure are solely related to changes in accommodation that influence sediment storage and supply on the shelf and coastal plain. The peak in sediment supply at the coastline occurs during the falling-stage systems tract and is associated with the inflection point on the falling limb of the relative sea-level curve (Fig. 4a). This is the point at which incision, erosion and bypass occur in the non-marine zone. The eroded sediment is transported to the coast and results in the sediment supply peak at the coastline (Fig. 4c). If the rate of accommodation development curve and the sediment supply curve are convolved, an accommodation:sediment supply ratio curve is produced (Fig. 4d). Relatively high A/S ratios are experienced in the early Highstand (eHST) and reduce in the late Highstand (lHST), being lowest during the Falling-stage Systems Tract (FST). A/S ratios are still low in the early Lowstand (eLST) and increase during the late Lowstand (lLST) and Transgressive Systems Tracts (TST; Fig. 4d). Low A/S ratios can be equated with flat, slightly rising and falling shoreline trajectories (Helland-Hansen & Martinsen 1996). High A/S ratios correspond to steeply rising shoreline trajectories. The concept of changing architecture in response to accommodation and sediment supply variations is independent of the causal mechanism of the accommodation or sediment supply changes. The causal mechanisms that result in the changes in sediment supply or accommodation are not important in defining the sedimentary architecture. Sediment supply changes induced by climatic variation or

203

hinterland uplift will produce the same changes in coastline architecture as will accommodation changes induced by eustatic sea-level variation or changes in rates of tectonic subsidence or uplift. That is, in the marginal marine zone, sedimentary architecture and hence the connectivity of the sand bodies is controlled by the ratio of accommodation: coarse sediment supply and the dominant depositional processes acting at the shoreline and is independent of the causal mechanisms of these changes.

Prediction of inter-parasequence compartmentalization The lowest order of compartmentalization in the marginal marine zone is the inter-parasequence scale (Figs 1 & 2). Vertical connectivity between parasequences is controlled by A/S ratios. Figure 5 shows a typical depositional sequence and its associated A/S curve. During periods of relatively high A/S (late Lowstand, Transgressive and early Highstand Systems Tracts) vertical separation of parasequences by relatively thick flooding shales is the norm (Fig. 5). Hence compartmentalization potential is high. During periods of lower A/S ratios (late Highstand, Falling-stage and early Lowstand Systems Tracts), lower accommodation and falling sea-levels result in a higher potential for downcutting and erosion from overlying parasequences into underlying parasequences by fluvial processes, resulting in incised valleys, and wave erosion processes (Plint 1988; Ainsworth 1994). Previously deposited flooding shales may therefore be eroded or, in very low accommodation systems that experience constant wave agitation, may never have been deposited. Hence compartmentalization potential at the inter-parasequence scale in the low A/S systems tracts is low. Since many parasequences can be traced over large along-strike distances (up to 100s of kms; Reynolds 1999) and down depositional-dip extents of kilometres to 10s of kilometres (Reynolds 1999), lateral connectivity of parasequences is generally not a concern over the areas of typically sized oil and gas fields (usually less than 10 km2; Reynolds 1999) and even giant oil and gas fields (usually less than 100 km2 in area; Reynolds 1999).

Prediction of inter sand-body compartmentalization The second order of compartmentalization is the inter sand-body scale (Figs 2 & 6). This is a function of the inter-connectivity of reservoir elements or sand bodies within a parasequence. In the marginal marine depositional setting there are two main controls on inter sand-body connectivity. The main

204

R. B. AINSWORTH

Fig. 4. Accommodation:sediment supply ratio (A/S) curves in the marginal marine zone. (a) Relative sea-level curve and systems tracts. (b) Relative rate of accommodation change curve derived from (a). (c) Relative rate of sediment supply curve derived from (a) and (b). N.B. the curve assumes that sediment supply variations are controlled by sediment storage and release on the exposed shelf or coastal plain due to changes in accommodation. That is, it assumes no independent variations in sediment supply (e.g. due to climate change or hinterland tectonics). (d) Accommodation:sediment supply ratio curve derived from convolving curves in (b) and (c). Note that maximum values of A/S occur at the inflection points on the rising limb of the relative sea-level curve (maximum flooding surface) and the falling limb of the curve within the Falling-stage Systems Tract (FST). HST, Highstand Systems Tract; LST, Lowstand Systems Tract; TST, Transgressive Systems Tract; StR, Steeply Rising; SR, Slightly Rising. Modified from Ainsworth et al. (2008).


205

Fig. 5. Relationship between inter-parasequence compartmentalization and A/S. The differentiation between high and low compartmentalization potential corresponds to the change of high to low A/S regime. Note that parasequence flooding shales are well developed during relatively high A/S periods (late Lowstand, Transgressive and early Highstand Systems Tracts) creating a high potential for inter-parasequence compartmentalization. Conversely during periods of low A/S (late Highstand, Falling-stage and early Lowstand Systems Tracts), the potential for eroding flooding shales and creating vertical communication pathways between parasequences is high. Hence for these systems tracts, inter-parasequence compartmentalization potential is relatively low.

control on lateral or horizontal connectivity is the dominant depositional process active at the time of deposition (fluvial, tidal or waves). These processes define the geometry of the reservoir sand bodies within the parasequences (Reynolds 1999). Dominant fluvial processes generally produce more laterally restricted geometries such as mouthbars, crevasse-splays and distributary channels. These processes result in jigsaw puzzle or labyrinth type reservoirs (Weber & van Geuns 1990). Dominant wave processes generally produce more laterally extensive sand bodies such as strandplains and beach-ridge complexes. These processes result in layer-cake type reservoirs (Weber & van Geuns 1990). Dominant tidal processes can result in laterally extensive bodies (e.g. tide-dominated shorefaces) or laterally restricted geometries (e.g. sand bars in estuarine complexes) and generally form any of the reservoir types from layer-cake to jigsaw puzzle to labyrinth type reservoirs (Weber & van Geuns 1990). The main control on inter sand-body scale vertical connectivity in the marginal marine depositional environment is accommodation:sediment supply ratio. Low A/S ratios typical of late Highstand, Falling-stage and early Lowstand systems tracts (flat, slightly rising and falling shoreline trajectories; Helland-Hansen & Martinsen 1996) are more likely to produce enhanced vertical

amalgamation of sand bodies within a parasequence (Fig. 6). Relatively high A/S ratios typical of late Lowstand, Transgressive or early Highstand systems tracts (steeply rising shoreline trajectories) on the other hand are more likely to result in greater vertical separation of sand bodies (Fig. 6). In high A/S settings, wave-dominated shoreface systems have a variable vertical compartmentalization potential in along-depositional strike v. a down-depositional dip sense (Fig. 7). In a depositional strike orientation, vertical compartmentalization potential is high (Fig. 7a). However, in an up-dip sense, compartmentalization potential is reduced due the amalgamation of bedsets in an onshore direction (Fig. 7b). This relationship has been well documented by numerous authors (e.g. Pattison 1995; Hampson 2000; Hampson & Storms 2003). Note that these bedset surfaces are effectively large scale dipping shale draped surfaces that amalgamate over up-dip distances of 100 to 6000 m (Hampson 2000). However, they should not be confused with those dipping shales observed at the intra sand-body scale in fluviallydominated mouthbar deposits (see intra sandbody compartmentalization prediction below). For wave-dominated shorefaces in low A/S settings, both up-dip and along strike compartmentalization potential is low (Fig. 7c, d).

206

R. B. AINSWORTH

Fig. 6. Relationship between inter sand-body compartmentalization and A/S. Relative sea-level curve depicts systems tracts related to high and low A/S ratio regimes. Lateral compartmentalization potential at this scale is a product of the dominant depositional processes acting the time of deposition. Low lateral continuity (top of figure) is a product of fluvial- and some tide-dominated systems. High lateral continuity (bottom of figure) is a product of wave- and some tide-dominated systems. Vertical compartmentalization potential at this scale is a result of A/S regime. High A/S regimes produce relatively high compartmentalization potential whereas low A/S regimes result in relatively low to moderate compartmentalization potential. Red, orange and blue vertical and horizontal arrows represent high, moderate and low vertical and horizontal compartmentalization potential respectively. Note that vertical compartmentalization potential in the High A/S, high lateral continuity scenario is classed as being moderate. This is due to the potential for up-depositional dip connectivity of sandstones that are vertically separated in a depositional strike sense. See Figure 7a, b. StR, Steeply Rising; SR, Slightly Rising. Abbreviations as in Figure 4.

Prediction of intra sand-body compartmentalization The third order of compartmentalization prediction is the intra sand-body scale (Figs 2 & 8). This scale is represented by heterogeneities that occur

within the individual reservoir sands. Features such as inclined or dipping shales on the foresets of mouthbars (Ainsworth et al. 1999; Tye et al. 1999) or individual bedforms (Visser 1980; Terwindt 1981), horizontal shales (Geehan 1993), carbonate concretions (Taylor et al. 2000; Alsop


207

Fig. 7. Inter sand-body connectivity complications in wave-dominated systems. In high A/S regimes along strike vertical compartmentalization potential is high (a). However, up depositional-dip connectivity potential is possible due to the up-dip amalgamation of bedsets (b). However, in low A/S systems, both strike and dip compartmentalization potential is low (c & d). Yellow is reservoir and grey is non-reservoir. The vertical scale is approximately 10–20 m and the horizontal scale 100 to 6000 m (Hampson 2000).

& Ainsworth 2006; Howell et al. 2008), and channel shale plugs (Eberth 1996; Ainsworth et al. 1999; Larue & Friedmann 2005) may form barriers or baffles to fluid flow at this scale. In more tide-dominated shoreface systems, laterally extensive carbonaceous inter-laminations also form potential barriers and baffles (Ainsworth 2003; Ainsworth et al. 2008). The impacts of A/S changes at this scale are not well documented and they are the subject of ongoing research. However, it can be speculated that in low A/S regimes, increased lateral and vertical erosion and scour would result in preferential erosion of any previously deposited shales or early cements

thereby decreasing the potential for compartmentalization (Fig. 8a, c). However, in high A/S regimes the opposite effect may occur. That is, increased rates of accommodation development may result in increased preservation potential and increased lateral and vertical distribution of heterogeneities thereby leading to decreased reservoir connectivity or more tortuous connectivity pathways (Fig. 8b, d). It can also be speculated that during periods of varying A/S, the angle of the dipping clinoform fronts in mouthbar or shoreface deposits may change (Fig. 8a, d). Enge (2008) demonstrated that the clinoform front gradients in the Panther Tongue, a low A/S (Falling-stage Systems Tract)

Fig. 8. Relationships between intra sand-body compartmentalization and A/S. (a) & (b) show the impact of dipping shales, (c) & (d) show the impact of horizontal shales or carbonaceous layers. (a) Cross-section through a mouthbar deposit in a low A/S regime. (b) Cross-section through a mouthbar deposit in a high A/S regime. Note that in (a) dips of foresets are lower than in (b) and that dipping shales do not extend to the top of the mouthbar. Red lines represent channels which are sand-prone in the low A/S setting and shale-prone in the high A/S setting. Intra sand-body compartmentalization potential is higher under the high A/S regime. (c) Laterally discontinuous shales more typical of low A/S settings. (d) Laterally continuous shales more typical of high A/S settings.

208

R. B. AINSWORTH

deposit, have much longer beds and are more shallowly dipping (mean dips ¼ 1.258) than the clinoforms from the Ferron Sandstone, a relatively high A/S (Highstand Systems Tract) deposit (mean dips ¼ 2.68). Both of the intervals studied in the Panther Tongue and the Ferron Sandstone are interpreted to be fluvial-dominated systems (Enge 2008) hence appealing solely to dominant depositional processes to account for the difference in clinoform dips may not be reasonable. An alternative possibility may be that under high A/S conditions the angle of the dipping clinoform mouthbar fronts may increase due to the more rapid rates of sediment accumulation and reduced periods of time allowed for depositional processes to equilibrate across the mouthbar front (Fig. 8b). At lower rates of A/S, accommodation space is developing at lower rates and more time may be available for processes operating on the mouthbar front to equilibrate and produce relatively lower angle equilibrium mouthbar dips (Fig. 8a). It may also be possible that the prevalent A/S regime may impact the balance of dominant depositional processes acting at the mouthbar front thereby resulting in different clinoform dips in different A/S regimes. Further study is required to determine whether or not there is a consistent relationship between clinoform dip angle and dominant A/S regime.

Compartmentalization potential matrix Figure 9 represents a matrix for predicting stratigraphic compartmentalization potential in marginal marine reservoirs at the three levels of hierarchy discussed above and shown in Figure 2. The horizontal axis represents the dominant depositional process acting at the shoreline at the time of deposition (fluvial, tidal or wave). This process domination determines the most likely lateral extent of the sand bodies produced, low for fluvial systems, low to high for tidal systems and high for wavedominated systems. Each process dominance column is subdivided into two columns, high and low A/S. The vertical axis represents the three scales of the compartmentalization hierarchy, interparasequence, inter sand-body and intra sand-body (Fig. 2). Compartmentalization potential for each level is listed at the bottom of each hierarchy row. Relative compartmentalization potential for the six classes of marginal marine system is detailed at the bottom of the table. It can be seen from the figure that fluvial-dominated systems in a high A/S setting have the highest stratigraphic compartmentalization potential at all three levels of the hierarchy. Wave-dominated systems in low A/S regimes fall at the opposite end of the spectrum and have relatively low compartmentalization potential at all levels of the hierarchy.

Fluvial-dominated systems in low A/S settings have a moderate compartmentalization potential since any connections likely to be made with other sand bodies will have a high probability of being tortuous jigsaw puzzle type reservoirs (Weber & van Geuns 1990). Strandplain systems in tidal- and wave-dominated, high A/S settings have a moderate vertical compartmentalization potential since amalgamation of bedsets in an up-dip direction is common (Fig. 7a, b). Inter-parasequence compartmentalization potential is high for all high A/S systems. Inter-parasequence and inter sand-body compartmentalization potential for tide-dominated systems in low A/S settings is ranked as moderate to low. This is due to the potential large variation in the lateral continuity of tide-dominated reservoir sand bodies. That is, tide-dominated shoreface systems (Ainsworth 2003; Ainsworth et al. 2008) can be laterally extensive but sand bodies in estuarine settings are relatively laterally restricted by comparison.

Discussion Implications for hydrocarbon extraction The above described scheme and predictive matrix for compartmentalization potential have practical applications for field development planning. Identification and prediction of the three key scales of reservoir compartmentalization potential is important to enable optimal well placement, completion strategies and in mature fields, identification of unswept volumes. All three scales of compartmentalization potential need to be considered in order to produce optimal field development plans and completion strategies. Larue & Legarre (2004) demonstrated the significance of the parasequence as a key flow unit in the Meren reservoir, Nigeria and were able to identify unswept volumes by modelling parasequence surfaces as compartmentalizing boundaries. Bryant & Livera (1991) also documented laterally extensive mudstones associated with parasequence boundaries that compartmentalize the Ness Formation of the Brent Field, North Sea. The recognition of parasequence surfaces as being potential flow barriers will influence the completion strategy for a field composed of a number of stacked parasequences. If static and dynamic pressure data indicate that parasequence boundaries are pressure seals that vertically compartmentalize a reservoir, then the parasequences may represent flow units that can potentially be treated as individual reservoirs. This information is critical to planning completion strategies for these types of fields and allows potential flexibility for zoned, sequential or combined production from the different parasequences.

STRATIGRAPHIC COMPARTMENTALIZATION PREDICTION 209

Fig. 9. Summary matrix of dominant depositional processes, A/S regime and compartmentalization potential. Note that high A/S settings have higher compartmentalization potentials than reservoirs deposited in low A/S settings. Also note that wave-dominated systems in low A/S settings have the lowest compartmentalization potential ranking whilst fluvial-dominated systems in high A/S settings have the highest compartmentalization potential ranking. Tide-dominated systems have an intermediate ranking. See text for further explanation and discussion.

210

R. B. AINSWORTH

Inter sand-body compartmentalization prediction is especially important when determining initial well spacings, locations and completion strategies. More wells (higher well densities) will be required to drain low continuity, jigsaw puzzle and labyrinth type reservoirs relative to the high continuity, layer-cake reservoirs (Fig. 9; Weber & van Geuns 1990). Ainsworth (2003, 2005) demonstrated the importance of inter sand-body connectivity from the Sunrise Field (Australia). He showed how in this field, inter sand-body connectivity for both low continuity and high continuity sand bodies (Fig. 6) can be related to dominant depositional processes acting at the time of deposition, position in the sequence stratigraphic hierarchy and A/S ratios. Wehr & Brasher (1996) illustrated the importance of recognizing dipping bedset correlations (Fig. 7) within the Cormorant Field (North Sea) and the potential impact these features can have on recovery factors if not included in 3D reservoir modelling strategies. Tye et al. (1999) demonstrated the impact of intra sand-body scale dipping mouthbar clinoforms (Fig. 8) on production from the Prudhoe Bay Field, Alaska. Ainsworth et al. (1999) similarly recognized the production impacts of dipping mouthbar clinoforms in the Sirikit Field, Thailand. They also suggested that perforation strategies that incorporated perforation of thin sands in the toesets of the clinoforms may improve recovery factors from these types of reservoirs.

Predictive subsurface tools This paper has focused on the prediction of compartmentalization from A/S parameters. However, accommodation and sediment supply are not parameters that can be directly measured from the rock record. This section details two simple tools that can be applied to enable the prediction of ranges of compartmentalization potential in ancient marginal marine systems. Sequence stratigraphy. Figures 4 –6 illustrate how sequence stratigraphy has a strong link to relative sea-level and accommodation:coarse sediment supply ratios. It therefore follows that sequence stratigraphic interpretations can be utilized to develop compartmentalization potential predictions. Figure 10 schematically summarizes compartmentalization potential at the three hierarchical scales described in Figure 2. Low reservoir compartmentalization potential is predicted within systems tracts deposited under relatively low A/S conditions (lHST, FST, eLST, flat, slightly rising and falling shoreline trajectories). Conversely, relatively high reservoir compartmentalization potential is predicted within systems tracts deposited under

relatively high A/S conditions (lLST, TST, eHST, steeply rising shoreline trajectories). The intermediate or moderate compartmentalization potential predictions can be made once depositional environment is also factored in to the scenario being analysed (Figs 6, 7 & 9). The key sequence stratigraphic surfaces that are usually identified in subsurface or outcrop studies (transgressive surface of erosion, maximum flooding surface and sequence boundary) all fall within the zones of low and high compartmentalization potential (Fig. 10). That is, they are not key surfaces for identifying switches between high and low compartmentalization potential. The key features for identifying these changes are the switches in stacking patterns which change from the early Highstand into the late Highstand (high to low compartmentalization potential) and from the early Lowstand into the late Lowstand (low to high compartmentalization potential). Well log data. Once a sequence stratigraphic framework has been generated, another tool can be used to analyse inter sand-body compartmentalization potential. Figure 11 utilizes the parasequence stratigraphy from the Sunrise Field, Australia (Ainsworth 2003, 2005, 2006). Plotting average parasequence thickness divided by average parasequence sand:shale ratios against the stratigraphy generated very similar connectivity patterns to the inter sand-body connectivity calculated from 3D reservoir models (Fig. 11b; Ainsworth (2003, 2005)). These parameters generated a better fit to the 3D model than just comparing sand:shale ratios or net sand to gross rock ratios as suggested by King (1990). The reason these parameters appear to generate similar patterns to the 3D models is that they are most likely acting as proxies for the two key controls on depositional connectivity, accommodation and coarse sediment supply (Figs 4–6). The average thickness of a parasequence represents an approximation of the accommodation space available at the time of deposition and if the wells are located in a palaeo-coastal location, the sand:shale ratios from those wells represent an approximation of the proportion of coarse sediment being supplied to the coastline. Averaging of the thickness and sand: shale ratios for all wells in a field can help to account for any areal variations in these parameters. This technique works best when most wells are located in palaeo-coastal positions. Figure 12 demonstrates why an increase in inter sand-body connectivity is generated moving vertically upwards through HST 1. In the early HST, rates of A/S are relatively high and inter sand-body compartmentalization is correspondingly relatively high (Fig. 6). However, towards the top of the HST in the late HST, A/S ratios are decreasing

STRATIGRAPHIC COMPARTMENTALIZATION PREDICTION 211

Fig. 10. Sequence stratigraphy as a predictive tool for stratigraphic compartmentalization potential in marginal marine systems. Note that high compartmentalization potential equates to high A/S ratios and low compartmentalization potential to low A/S ratios at all hierarchical levels. Also note that high Thickness divided by sand/shale ratios equate to high compartmentalization potential and low Thickness divided by sand/shale ratios equate to low compartmentalization potential. CP, compartmentalization potential.

212

R. B. AINSWORTH

Fig. 11. (a) Results from 3D reservoir modelling of the Sunrise Field (Australia; Ainsworth 2003, 2005, 2006) showing relative connectivity within parasequences (inter sand-body scale compartmentalization). This is a datumed model representing depositional or stratigraphic connectivity. High connectivity is represented by a relatively small number of clusters (groups of connected reservoir quality grid blocks) in the 3D model and low connectivity by a relatively large number of clusters. (b) Parasequences v. thickness divided by sand/shale ratio plot. Note that the connectivity trends in systems tracts (coloured dotted arrows) are the same as seen in the 3D model in (a). This is probably because the average thickness parameter is acting as a proxy for accommodation and the average S/Sh parameter is acting as a proxy for coarse sediment supply. See Figures 4, 5 and 10. Modified from Ainsworth (2003).

enabling lateral and vertical amalgamation of sand bodies thus resulting in fewer clusters (groups of connected reservoir quality grid blocks) and higher connectivity (Fig. 6). At the top of the succession shown in Fig. 12, there is a switch in depositional style to wave-dominated coastal systems. These systems occur in relatively low A/S settings and hence exhibit both low lateral and vertical inter sand-body compartmentalization potential (Figs 6 & 12). Since the thickness divided by sand:shale ratio parameter can be equated to A/S, it is also possible to apply the thickness divided by sand:shale ratio parameter to the predictive matrix shown in Figure 9 and hence also to develop a relative predictive compartmentalization matrix from parameters derived from well data alone. Note that these techniques will only work if a sand:shale ratio is utilized. If net sand to gross rock ratios are used where sands are discounted by reservoir cut-offs, these techniques will not work. This is because the deposited reservoir architectures are a function of the sands and shales, and bear no

relation to net sand values calculated once postdepositional diagenetic effects or dynamic flow restrictions have been applied to the reservoirs. Reliable sand:shale values can be derived from Vshale logs that have been calibrated to core data. Other logs or combinations of logs that define sandstone from shale and have been calibrated to core can also be utilized.

Coarse sediment supply, depositional processes and reservoir architecture An interesting point to note in terms of potential reservoir geometries is that changing the volume of coarse sediment supply to a coastline does not necessarily directly alter the depositional processes acting at that coastline. For example if a shoreline is wave-dominated and coarse sediment supply is then reduced, the coastline does not become fluvialdominated, it just becomes a relatively high A/S wave-dominated coastline (Fig. 13). Conversely, if coarse sediment supply is increased to the coastline, as long as wave energy is still greater than fluvial


213

Fig. 12. Results from 3D reservoir modelling of the Sunrise Field (Australia; Ainsworth 2003, 2005, 2006) showing relative inter sand-body connectivity by parasequence (Fig. 11). Note that the relative sea-level and A/S curve is superimposed on the plot. Schematic cross-sections on the bottom right illustrate the reasons for the increase in connectivity upwards through the fluvial-dominated HST-1 (blue dotted arrow on graph). Schematic cross-section at top right illustrates the reasons for good connectivity in the wave-dominated upper part of the succession. See Figure 6. Modified from Ainsworth (2003).

energy, the coastline will remain wave-dominated, with only the A/S ratio changing to a lower value (Fig. 13). The same principles apply for changing the accommodation parameter. If only accommodation is changed and sediment supply is kept constant then the end result is only a change in A/S and not a change in depositional style (Fig. 13). For changes in depositional style to be induced by coarse sediment supply changes, the supply changes that occur need to be significant such as rapid fluvial sediment increases or decreases at a point source. The resultant relative change in energy between fluvial and wave and tide processes can then precipitate changes in the depositional style of the coastline. Subtle changes in sediment supply to a coastline generally do not directly influence the architecture of the coastline. Subtle changes in sediment supply to a coastline can only alter the depositional architecture of that coastline in an indirect manner. For example, a change in supply may in turn alter the palaeogeography of a coastline which may then result in changes in dominant depositional processes which may alter the reservoir architectures and hence the potential for reservoir compartmentalization (Ainsworth 2003; Ainsworth et al. 2008).

Issues with mixed-process systems This paper has discussed compartmentalization potential in end-member wave-, fluvial- and tidedominated marginal marine systems. These endmember depositional models represent the extremes of potential compartmentalization scenarios but are not necessarily the most common models that will be encountered in nature. Indeed, recent work suggests that many naturally occurring marginal marine depositional systems are the product of mixed-process systems, that is, systems that exhibit strong degrees of secondary process influence (Ainsworth 2003; Bhattacharya & Giosan 2003; Yoshida et al. 2007; Ainsworth et al. 2008). Figure 14a, b shows two wave-dominated, fluviallyinfluenced deltas and the relatively complex reservoir architecture likely to be generated by these systems (Fig. 14c). The fluvial influence can be seen in the form of mouthbar deposits sitting behind wave-dominated spits or barriers. This level of complexity is not captured within the models presented in this paper. The connectivity of many marginal marine systems is therefore likely to lie within a spectrum between the endmembers depicted in this paper. Further work is

214

R. B. AINSWORTH

Fig. 13. Summary figure illustrating how simple changes of accommodation or coarse sediment supply can produce similar A/S ratios and hence similar compartmentalization potential. The figure also illustrates that changing coarse sediment supply rates does not necessarily result in a change of process dominance. For example, increasing coarse sediment supply in a wave-dominated system (bottom two right schematic cross-sections – Low S to High S) does not result in a fluvial-dominated system if wave energy still overrides fluvial energy, it simply results in a lower A/S ratio wave-dominated system. A, accommodation; S, coarse sediment supply.

required on this aspect of compartmentalization prediction in marginal marine systems.

(3)

Conclusions The key conclusions of this study are: (1) Marginal marine depositional systems exhibit stratigraphic reservoir compartmentalization potential at three hierarchical scales; interparasequence, inter sand-body and intra sandbody. All three scales are important to consider in terms of field development planning and reservoir completion strategies. (2) At each scale, stratigraphic compartmentalization potential can be related to the dominant depositional processes acting at the time of deposition and accommodation:coarse sediment supply ratios (A/S). Stratigraphic compartmentalization potential prediction is independent of the causal mechanisms driving the changes in accommodation and coarse sediment supply parameters.

(4)

At the lowest order of compartmentalization (inter-parasequence scale), in systems tracts associated with relatively high A/S ratios (steeply rising shoreline trajectories), for example late Lowstand, Transgressive and early Highstand, vertical compartmentalization potential is relatively high. In systems tracts associated with relatively low A/S ratios (slightly rising, flat and falling shoreline trajectories), for example late Highstand, Falling-stage and early Lowstand, vertical compartmentalization potential of parasequences is reduced. At the second level of compartmentalization hierarchy (inter sand-body scale), the lateral connectivity of reservoir sands is a product of the dominant depositional processes active at the time of their deposition (wave, tidal, fluvial). Wave-dominated systems tend to produce more laterally continuous sands, fluvialdominated systems more laterally restricted sands, and tide-dominated systems both laterally continuous and laterally restricted sands.


215

Fig. 14. Issues with mixed-process marginal marine systems. (a) and (b) show Google Earth images of wave-dominated deltas. In these examples of mixed-process systems, waves dominate regional shoreline deposition, but fluvial influence is locally very significant. This will result in relatively complex sedimentary architectures not predicted by this paper. (c) is modified from Bhattacharya & Giosan (2003) and shows the potential internal stratigraphic complexities of such a marginal marine system.

(5)

(6)

Vertical compartmentalization at an inter sand-body scale is related to A/S regime. In high A/S regimes, sand bodies are more likely to be disconnected or compartmentalized. In low A/S regimes, erosional amalgamation of sand bodies is more likely thereby leading to lower compartmentalization potential. At the third order of potential compartmentalization (intra sand-body scale), heterogeneities such as dipping or horizontal shales, carbonaceous-rich beds or laminae, shale plugs of channels and carbonate concretions form the key features capable of compartmentalizing the reservoir. In High A/S regimes the preservation potential of these heterogeneities is relatively high. In lower A/S regimes,

(7)

preservation potential is lower resulting in a greater likelihood of lateral and vertical reservoir connectivity. Further study is required to analyse the compartmentalization potential of mixed-process marginal marine depositional systems.

The author would like to acknowledge former colleagues at Shell International Exploration and Production Company for many discussions over the past seventeen years that have shaped some of the ideas detailed in this paper. However, all ideas presented herein are the sole responsibility of the author. I am grateful to reviewers Gary Hampson, Tom McKie and editor Steve Jolley for many useful comments and suggestions on an earlier version of the manuscript. Thanks are due to the Cooperative Research Centre for Greenhouse Gas Technologies

216

R. B. AINSWORTH

(CO2CRC) and to the members of the WAVE Consortium (Bapetco, BHPBP, Chevron, ConocoPhilips, Nexen, OMV, Shell, StatoilHydro, Todd Energy, Woodside Energy) for partial funding of this work. This paper represents TRaX publication #100.

References Ahlbrandt, T. S., Charpentier, R. R., Klett, T. R., Schmoker, J. W., Schenk, C. J. & Ummishek, G. F. 2005. Global Resources Estimates from Total Petroleum Systems. American Association of Petroleum Geologists Memoir, 86. Ainsworth, R. B. 1994. Marginal marine sedimentology and high resolution sequence analysis; Bearpaw – Horseshoe Canyon transition, Drumheller, Alberta. Bulletin of Canadian Petroleum Geology, 42, 26–54. Ainsworth, R. B. 2003. Sequence stratigraphic-based analysis of depositional connectivity using 3-D reservoir modelling techniques. PhD thesis, University of Liverpool, UK. Ainsworth, R. B. 2005. Sequence stratigraphic-based analysis of reservoir connectivity: influence of depositional architecture – a case study from a marginal marine depositional setting. Petroleum Geoscience, 11, 257–276. Ainsworth, 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. Ainsworth, R. B., Sanlung, M. & Duivenvoorden, S. T. C. 1999. Correlation techniques, perforation strategies and recovery factors. An integrated 3-D reservoir modeling study, Sirikit field, Thailand. American Association of Petroleum Geologists Bulletin, 83, 1535–1551. Ainsworth, R. B., Flint, S. S. & Howell, J. A. 2008. Predicting coastal depositional style: influence of basin morphology and accommodation to sediment supply ratio within a sequence stratigraphic framework. In: Hampson, G. J., Steel, R., Burgess, P. & Dalrymple, R. (eds) Recent Advances in Models of Siliciclastic Shallow-Marine Stratigraphy. Society of Economic Paleontologists and Mineralogists, Special Publication, Tulsa, 90, 237–263. Alsop, D. B. & Ainsworth, R. B. 2006. Predicting Calcite Cement Distribution in the Sunrise Gas Field Using Analogues. American Association of Petroleum Geologists International Convention, Perth, Abstract. Bailey, W. R., Manzocchi, T., Walsh, 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. Bhattacharya, J. P. & Giosan, L. 2003. Waveinfluenced deltas: geomorphological implications for facies reconstruction. Sedimentology, 50, 187– 210. Budding, M. C., Paardekam, A. H. M. & Van Rossem, S. J. 1992. 3D connectivity and sandstone architecture. SPE Paper 22342 presented at SPE International Meeting on Petroleum Engineering, Beijing, China. Bryant, I. D. 1996. The application of physical measurements to constrain reservoir-scale sequence stratigraphic models. In: Howell, J. A. & Aitken, J. F.

(eds) High Resolution Sequence Stratigraphy: Innovations and Applications. The Geological Society, London, Special Publications, 104, 51–63. Bryant, I. D. & Flint, S. S. 1993. Quantitative clastic reservoir geological modelling: problems and perspectives. In: Flint, S. S. & Bryant, I. D. (eds) The Geological Modelling of Hydrocarbon Reservoirs and Outcrop Analogues. International Association of Sedimentologists, Special Publication, Oxford, 15, 3– 20. Bryant, I. D. & Livera, S. E. 1991. Identification of unswept oil volumes in a mature field by using integrated data analysis: Ness formation, Brent Field, UK North Sea. In: Spencer, A. M. (ed.) Generation, Accumulation and Production of Europe’s Hydrocarbons. European Association of Geoscientists, Special Publication, Amsterdam, 1, 75– 88. Cook, T. W., Bouma, A. H., Chapin, M. A. & Zhu, H. 1994. Facies architecture and reservoir characterization of a submarine fan channel complex, Jackfork Formation, Arkansas. In: Weimer, P., Bouma, A. H. & Perkins, B. (eds) Submarine Fans and Turbidite Systems: Sequence Stratigraphy, Reservoir Architecture and Production Characteristics, Gulf of Mexico and International. Proceedings of Gulf Coast Section SEPM 15th Annual Research Conference. SEPM, Tulsa, 69–81. Cross, T. A., Baker, M. R. et al. 1993. Applications of high resolution sequence stratigraphy to reservoir analysis. In: Eschard, R. & Doligez, B. (eds) Subsurface Reservoir Characterization from Outcrop Observations. Proceedings of the 7th Exploration and Production Research Conference, Institute Francais du Petrole, Editions Technip, Paris, 51, 11–33. Eberth, D. A. 1996. Origin and significance of mud-filled incised valleys (Upper Cretaceous) in southern Alberta, Canada. Sedimentology, 43, 459–477. Enge, H. D. 2008. Deltaic clinothems– digital data capture, geometries, and reservoir implications. PhD thesis, University of Bergen, Norway. Geehan, G. 1993. The use of outcrop data and heterogeneity modelling in development planning. In: Eschard, R. & Doligez, B. (eds) Subsurface Reservoir Characterization from Outcrop Observations. Proceedings of the 7th Exploration and Production Research Conference, Institute Francais du Petrole, Editions Technip, Paris, 51, 53–64. Hampson, G. J. 2000. Discontinuity surfaces, clinoforms, and facies architecture in a wave-dominated shorefaceshelf parasequence. Journal of Sedimentary Research, 70, 325– 340. Hampson, G. J. & Storms, J. E. A. 2003. Geomorphological and sequence stratigraphic variability in wave-dominated, shoreface-shelf parasequences. Sedimentology, 50, 667–701. Helland-Hansen, W. & Martinsen, O. J. 1996. Shoreline trajectories and sequences: description of variable depositional-dip scenarios. Journal of Sedimentary Research, 66, 670– 688. Hovadik, J. M. & Larue, D. K. 2007. Static characterizations of reservoirs: refining the concepts of connectivity and continuity. Petroleum Geoscience, 13, 195–211.

STRATIGRAPHIC COMPARTMENTALIZATION PREDICTION Hovadik, J. M. & Larue, D. K. 2010. Stratigraphic and structural connectivity. In: Jolley, S. J., Fisher, Q. J., Ainsworth, R. B., Vrolijk, P. J. & Delisle, S. (eds) Reservoir Compartmentalization. Geological Society, London, Special Publications, 347, 219– 242. Howell, J. A., Skorstad, A., Macdonald, A., Fordham, A., Flint, S., Fjellvoll, B. & Manzocchi, T. 2008. Sedimentological parameterization of shallow-marine reservoirs. Petroleum Geoscience, 14, 17–34. Jolley, S. J., Barr, D., Walsh, J. J. & Knipe, R. J. 2007. Structurally complex reservoirs: an introduction. In: Jolley, S. J., Barr, D., Walsh, J. J. & Knipe, R. J. (eds) Structurally Complex Reservoirs. Geological Society, London, Special Publications, 292, 1– 24. Jones, A., Doyle, J., Jacobsen, T. & Kjonsvik, D. 1995. Which sub-seismic heterogeneities influence waterflood performance? A case study of a low net-to-gross fluvial reservoir. In: Haan, H. J. (ed.) New Developments in Improved Oil Recovery. Geological Society, London, Special Publications, 84, 5 –18. Keogh, K. J. 2002. Sequence stratigraphy and 3-D modelling of the East Pennine Coalfield, UK: a deterministic and stochastic approach. PhD thesis, University of Liverpool, UK. King, P. R. 1990. Connectivity and conductivity of overlapping sand bodies. In: Buller, A. T., Hjelmeland, O., Kleppe, J., Torsaeter, O. & Aasen, J. O. (eds) North Sea Oil and Gas Reservoirs – II. The Norwegian Institute of Technology, Graham & Trotman, 353–362. Knipe, R. J., Jones, G. & Fisher, Q. J. 1998. Faulting, fault sealing and fluid flow in hydrocarbon reservoirs: an introduction. In: Jones, G., Fisher, Q. J. & Knipe, R. J. (eds) Faulting, Fault Sealing and Fluid Flow in Hydrocarbon Reservoirs. Geological Society, London, Special Publications, 147, vii –xxi. Larue, D. K. & Friedmann, F. 2005. The controversy concerning stratigraphic architecture of channelized reservoirs and recovery by waterflooding. Petroleum Geoscience, 11, 131– 146. Larue, D. K. & Hovadik, J. M. 2006. Connectivity of channelized reservoirs: a modelling approach. Petroleum Geoscience, 12, 291– 308. Larue, D. K. & Legarre, H. 2004. Flow units, connectivity, and reservoir characterization in a wavedominated deltaic reservoir: Meren reservoir, Nigeria. American Association of Petroleum Geologists Bulletin, 88, 303–324. Lasseter, T. J., Waggoner, J. R. & Lake, L. W. 1986. Reservoir heterogeneities and their influence on ultimate recovery. In: Lake, L. W. & Carroll, H. B. (eds) Reservoir Characterization. Academic Press, Orlando, 545–559. Legaretta, L., Uliana, M. A., Larotonda, C. A. & Meconi, G. R. 1993. Approaches to nonmarine sequence stratigraphy – theoretical models and examples from Argentine basins. In: Eschard, R. & Doligez, B. (eds) Subsurface Reservoir Characterization from Outcrop Observations. Proceedings of the 7th Exploration and Production Research Conference, Institute Francais du Petrole, Editions Technip, Paris, 51, 125–145. Manzocchi, T., Walsh, J. J., Tomasso, M., Strand, J., Childs, C. & Haughton, P. D. W. 2007. Static and

217

dynamic connectivity in bed-scale models of faulted and unfaulted turbidites. In: Jolley, S. J., Barr, D., Walsh, J. J. & Knipe, R. J. (eds) Structurally Complex Reservoirs. Geological Society, London, Special Publications, 292, 309–336. Martinsen, R. S. 1994. Stratigraphic compartmentation of reservoir sandstones: examples from the muddy sandstone, powder river basin, Wyoming. In: Ortoleva, P. J. (ed.) Basin Compartments and Seals. American Association of Petroleum Geologists Memoir, Tulsa, 61, 273–296. Manzocchi, T., Carter, J. N. et al. 2008. Sensitivity of the impact of geological uncertainty on production from faulted and unfaulted shallow-marine oil reservoirs: objectives and methods. Petroleum Geoscience, 14, 3 –15. McKie, T., Jolley, S. J. & Kristensen, M. B. 2010. Stratigraphic and structural compartmentalization of dryland fluvial reservoirs: Triassic Heron Cluster, Central North Sea. In: Jolley, S. J., Fisher, Q. J., Ainsworth, R. B., Vrolijk, P. J. & Delisle, S. (eds) Reservoir Compartmentalization. Geological Society, London, Special Publications, 347, 165–198. Mijnssen, F. C. J. 1997. Modelling of sandbody connectivity in the Schooner Field. In: Ziegler, K., Turner, P. & Daines, S. R. (eds) Petroleum Geology of the Southern North Sea: Future Potential. Geological Society, London, Special Publications, 123, 169– 180. Pattison, S. A. J. 1995. Sequence stratigraphic significance of sharp-based lowstand shoreface deposits, Kenilworth Member, Book Cliffs, Utah. American Association of Petroleum Geologists Bulletin, 79, 444– 462. Pettijohn, F. J., Potter, P. E. & Siever, R. 1973. Sand and Sandstone. Springer-Verlag, New York, Heidelberg, Berlin. Plint, A. G. 1988. Sharp-based shoreface sequences and ‘offshore bars’ in the Cardium Formation of Alberta: their relationship to relative changes in sea level. In: Wilgus, C. K., Hastings, B. S., Kendall, C. G. St. C., Posamentier, H. W., Ross, C. A. & Van Wagoner, J. C. (eds) Sea-Level Changes: An Integrated Approach. SEPM Special Publication, Tulsa, 42, 357– 370. Posamentier, H. W. & Allen, G. P. 1999. Siliciclastic Sequence Stratigraphy – Concepts and Applications. SEPM Concepts in Sedimentology and Paleontology 7. Reynolds, A. D. 1999. Dimensions of paralic sandbodies. American Association of Petroleum Geologists Bulletin, 83, 211–229. Schlager, W. 1993. Accommodation and supply – a dual control on stratigraphic sequences. Sedimentary Geology, 86, 11–136. Shanley, K. W. & McCabe, P. J. 1994. Perspectives on the sequence stratigraphy of continental strata. American Association of Petroleum Geologists Bulletin, 78, 544–568. Swift, D. J. P. & Thorne, J. A. 1991. Sedimentation on continental margins, I: a general model for shelf sedimentation. In: Swift, D. J. P., Oertel, G. F., Tillman, R. W. & Thorne, J. A. (eds) Shelf Sand and Sandstone Bodies. International Association of Sedimentologists, Special Publication, Oxford, 14, 3– 31.

218

R. B. AINSWORTH

Taylor, K. G., Gawthorpe, R. L., Curtis, C. D., Marshall, J. D. & Awwiller, D. N. 2000. Carbonate cementation in a sequence-stratigraphic framework: Upper Cretaceous sandstones, Book Cliffs, UtahColorado. Journal of Sedimentary Research, 70, 360– 372. Terwindt, J. H. J. 1981. Origin and sequences of sedimentary structures in inshore mesotidal deposits of the North Sea Basin. In: Nio, S. D., Shuttenhelm, R. J. E. & Van Weering, Tj. C. E. (eds) Holocene Marine Sedimentation in the North Sea Basin. International Association of Sedimentologists, Special Publication, Oxford, 5, 4– 26. Tye, R. S., Bhattacharya, J. P., Lorsong, J. A., Sindelar, S. T., Knock, D. G., Puls, D. D. & Levinson, R. A. 1999. Geology and stratigraphy of Fluvio-deltaic deposits in the Ivishak formation: Applications for development of Prudhoe Bay Field, Alaska. American Association of Petroleum Geologists Bulletin, 83, 1588– 1623. Van Wagoner, J. C., Mitchum, R. M., Jr., Campion, K. M. & Rahmanian, V. D. 1990. Siliciclastic Sequence Stratigraphy in Well Logs, Cores, and Outcrops. American Association of Petroleum Geologists, Methods in Exploration Series, 7, Tulsa. Visser, M. J. 1980. Neap-spring cycles reflected in Holocene subtidal large-scale bedform deposits: A preliminary note. Geology, 8, 543–546.

Walsh, J., Watterson, J., Heath, A., Gillespie, P. A. & Childs, C. 1998. Assessment of the effects of subseismic faults on bulk permeabilities of reservoir sequences. In: Coward, M. P., Daltaban, T. S. & Johnson, H. (eds) Structural Geology in Reservoir Characterization. Geological Society, London, Special Publications, 127, 99–114. Weber, K. J. 1982. Influence of common sedimentary structures on fluid flow in reservoir models. Journal of Petroleum Technology, 34, 665– 672. Weber, K. J. 1986. How heterogeneity affects oil recovery. In: Lake, L. W. & Carroll, H. B. (eds) Reservoir Characterization. Academic Press, Orlando, 487– 544. Weber, K. J. & Van Guens, L. C. 1990. Framework for constructing clastic reservoir simulation models. Journal of Petroleum Technology, 42, 1248– 1297. Wehr, F. L. & Brasher, L. D. 1996. Impact of sequencebased correlation style on reservoir model behaviour, lower Brent Group, North Cormorant Field, UK North Sea. In: Howell, J. A. & Aitken, J. F. (eds) High Resolution Sequence Stratigraphy: Innovations and Applications. Geological Society, London, Special Publications, 104, 115–128. Yoshida, S., Steel, R. J. & Dalrymple, R. W. 2007. Changes in depositional processes – an ingredient in a new generation of sequence stratigraphic models. Journal of Sedimentary Research, 77, 447–460.

Stratigraphic and structural connectivity J. M. HOVADIK1* & D. K. LARUE2 1

Chevron Energy Technology Company, 6001 Bollinger Canyon Rd., San Ramon, CA 94583, USA

2

Chevron Energy Technology Company, 9525 Camino Media, Bakersfield, CA 93311, USA *Corresponding author (e-mail: [email protected]) Abstract: The connectivity of a reservoir to a well-bore represents a fundamental initial condition for drainage of an oil or gas field. The size of the static connected volume is a function of the stratigraphic and structural architecture of the reservoir. The most important stratigraphic factor affecting connectivity is a net-to-gross threshold which determines whether a reservoir is highly or poorly connected. Other stratigraphic factors affecting connectivity are those that impact the reservoir dimensionality (for example, compartmentalizing continuous mudstones or parallel channel deposits) and the size of geobodies relative to the total reservoir size. Structural compartmentalization may cause fault compartments that are too small in volume to support reservoir connectivity: as the size of the geobodies approaches compartment size, connectivity is typically less predictable. Static connected volumes alone do not predict flow performance, but are a component in predicting flow performance. To more completely address predictions of flow performance, dynamic connectivity is sometimes considered. However, dynamic connectivity, which is dependent on fluid type, permeability heterogeneity, time and other factors, confuses connectivity with tortuosity and sweep- and displacement-efficiency and is probably best avoided. Finally a connectivity flow diagram is proposed as a guide to help formulate key questions concerning uncertain reservoir parameters affecting reservoir connectivity.

Connectivity is a fundamental reservoir property that strongly affects the efficiency at which hydrocarbon is recovered. If a part of a reservoir is not connected to a producing well then the hydrocarbon present in that region cannot be produced. For successful secondary recovery using water injection, both producing and injecting wells need to connect to the same reservoir geobody in order to create sweep zones. Connectivity is therefore a necessary condition for reservoir producibility. But connectivity itself is not a guarantee of hydrocarbon recovery. Other reservoir properties also play critical roles. Some, described as static, are determined by the reservoir structure and stratigraphy and include reservoir volumes, permeability, heterogeneity, or reservoir tortuosity. Other reservoir characterizations are described as dynamic properties because of their time dependency. These dynamic properties include reservoir pressure, relative permeability, or fluid properties. In literature, connectivity is sometimes referred to as a dynamic property (Ballin et al. 2002; Renjun & Barton 2006). Dynamic connectivity is discussed in this study, and it is suggested that the concept be replaced by more useful concepts such as tortuosity, continuity, and permeability heterogeneity. In this present paper we first stress the importance of studying reservoir connectivity using

discrete three dimensional numerical models of reservoirs. Outcrop studies do provide insights on what geological features impact connectivity. Local mudstones beds, varying width to thickness ratio of sedimentary deposits, depositional stacking patterns or reservoir net-to-gross ratio are among the stratigraphic features most often referenced in literature (Larue & Legarre 2004; Larue & Hovadik 2006). On the structural side, faults, fault throws, fault density, whether faults are sealing or non-sealing through cementation, shale smears, or the shale gouge ratio, are among the most obvious structural factors influencing reservoir compartmentalization (Bailey et al. 2002; Ainsworth 2006). However to quantitatively determine relationships between the reservoir geology and reservoir connectivity in all three spatial dimensions, one has to adopt an experimental approach using numerical models.

Connectivity uncertainty in reservoir models In a preliminary experiment, two standard modelling workflows are tested to determine their accuracy in predicting reservoir compartments. Here, reservoir compartmentalization is quantified as the


220

J. M. HOVADIK & D. K. LARUE

volume of the largest geobody relative to the total reservoir volume (this is also known as ‘geobody connectivity’ by Larue & Hovadik 2006). A numerical representation of a synthetic reservoir, composed of channelized sand deposits is used as a benchmark to assess the relative performance of two geostatistical simulation algorithms in predicting compartmentalization under various well densities (dynamic data which could define reservoir compartments is not considered in this example). The two algorithms examined here are: Sequential Indicator Simulation (Journel & Isaaks 1984) and Multiple-Point Statistics simulation (Strebelle & Journel 2001). Since the synthetic reservoir that is being inverted is perfectly known, no uncertainty is implied in the input parameters of the simulation algorithms. A perfectly defined variogram best representing spatial continuity in the reservoir is used to constrain the Sequential Simulation Algorithm. Likewise, a training image volume, with analogous geomorphology to the reservoir is used to constrain the Multiple-Point Statistics algorithm. The target net-to-gross ratio for both simulation algorithms is defined on the basis of the varying sampled well data and therefore has a range of values. Reservoir compartmentalization is measured on multiple stochastic realizations by computing the volume fraction of the biggest connected geobody volume. Results shown on Figure 1 reveal that under sparse well conditions (2000, 1500 and 1000 m well spacing) the uncertainty of predicting geobody connectivity is high, and that both algorithms tend to produce models that are too optimistic in terms of connectivity. Given a perfect training image and a perfect variogram model, the Multiple-Point Statistics algorithm seems to perform marginally better then the two-point Sequential Indicator Simulation algorithm at well spacing equal or less then 500 m. However the sole purpose of this experiment is to highlight that under well conditions that are not dense, in green or even mature fields, simple statistical extrapolation may produce reservoir models with wrong connectivity. Therefore an Earth scientist can not simply rely on well information. Instead, informed deterministic decisions should be made in order to explore wide ranging scenarios of reservoir stratigraphy and structure that are likely to control reservoir connectivity.

Definitions and previous work The methodology for studying connectivity presented here intends to catalogue the many intertwined structural and stratigraphic factors that potentially impact connectivity. Among the geological factors studied in the proposed numerical experiments

Fig. 1. Geostatistical inversion of static reservoir connectivity at various well spacing intervals. (a) Map views of actual reservoir and inverted reservoir models using MPS and SIS at 500 and 1000 m well spacing. (b) Net-to-gross cross-plot in inverted models: MPS v. SIS. Net-to-gross is variable due to differences in well sampling of the synthetic reservoir (c) Geobody connectivity in inverted models: MPS v. SIS. (d) Geobody connectivity v. net-to-gross cross-plot for SIS. (e) Geobody connectivity v. net-to-gross cross-plot for MPS. See text for discussion.

are: reservoir net-to-gross ratio, reservoir element width and thickness, parallel channel deposits and orientation, deviation, sinuosity, sheeted reservoir intervals, compensational stacking patterns, continuous mudstone layers, local heterogeneity, volume support, faults, fault offset, sealing faults, fault density, orientation, length. In order to be successful, an experimental method should be able to classify these geological features and propose rules that explain the underlying phenomena that control reservoir connectivity. But before considering these experiments, reservoir connectivity in numerical model needs to be defined quantitatively. Several definitions of connectivity in numerical models have been proposed in previous publications. All definitions refer to geometrical characterizations of the interconnected porous network. These include: global volume descriptions of

STRATIGRAPHIC AND STRUCTURAL CONNECTIVITY

221

Fig. 1. (Continued)

interconnected reservoir flow units, local path length descriptions that are fully contained inside a reservoir, and connectivity functions describing the probability of connectivity for two reservoir locations. These definitions are briefly reviewed in the next four sections.

Bulk reservoir connectivity Bulk reservoir connectivity can be characterized by the ‘connected sand fraction’ and is measured by computing the ratio of the biggest geobody volume to the volume of the grid. As mentioned previously, this provides a measure of reservoir compartmentalization. This definition was previously used by Allen (1978), King (1990), and Hird & Kelkar (1994). Bulk reservoir connectivity can first be explained by percolation theory (King 1990; Hovadik & Larue 2007). Percolation theory describes the behaviour of interconnected clusters in a random and stationary system. Here a reservoir geobody refers to a connected cluster of reservoir flow units. Another definition of bulk reservoir connectivity refers to the probability for a reservoir location of belonging to the biggest ‘percolating’ reservoir cluster and is measured by computing the ratio of the biggest geobody volume to the volume of all reservoir geobodies. This definition was previously referred to as the ‘strength’ P of a percolation cluster (Christensen & Moloney 2005). In a percolation problem, P is a function of net-to-gross and becomes non-zero for net-to-gross values greater than the percolation threshold pc. For

net-to-gross values slightly greater than the percolation threshold, P has a mathematical formulation that is approximated to: P (NTG) (NTG – pc)b. These bulk characterizations of geobody connectivity may be easy to work with and provide some description of the reservoir compartmentalization prior to drilling, but from a practical or business perspective have limited value, as no wells drain the reservoir.

Reservoir to well connectivity In producing reservoirs, reservoir connectivity is both a function of reservoir architecture and well placement (Larue & Hovadik 2006; Hovadik & Larue 2007). Reservoir to well connectivity is defined as the volume fraction of reservoir flow units connected to one or more completed well interval. During the primary depletion phase of producing reservoirs, reservoir to well connectivity refers to the reservoir volume connected to the producing wells, or to the reservoir volume connected to producing wells and to an aquifer. During the secondary phase of enhanced hydrocarbon recovery, when a reservoir is produced using pore voidage replacement, a more useful characterization can refer to the volume fraction of the reservoir connected to both injecting and producing wells.

Connectivity function The connectivity function (Pardo-Iguzquiza & Dowd 2003) is a characterization of reservoir

222


connectivity in which the probability of two connected reservoir locations is plotted as a function of the Euclidian distance between the locations. For net-to-gross values that are lower than the percolation threshold, the connectivity function converges to zero as the distance between the two reservoir locations increases.

Local characterizations of reservoir connectivity The previous connectivity definitions are helpful for describing poorly connected reservoirs at lower net-to-gross values, typically below percolation thresholds. For reservoir net-to-gross above percolation thresholds, bulk reservoir connectivity is typically high, and hence the previous descriptions become less useful. Consequently, refined descriptions have been proposed to achieve better characterizations of reservoir connectivity even at high net-to-gross, once the previous bulk descriptions normally fail to provide useful information. Two local characterizations of reservoir connectivity briefly reviewed here, have been subsequently used in attempts to characterize dynamic aspects of connectivity. The Resistivity Index developed by Hird & Dubrule (1995) was used to estimate the hydrocarbon recovery and is determined by computing the least resistive paths between two locations in the reservoir. Here resistivity was derived from Darcy’s equations and Ohm’s law for a linear

electric current. This index was initially used to predict primary recovery and water breakthrough on 2D cross sections and subsequently on 3D grids (Ballin et al. 2002). Hovadik & Larue (2007) used an implementation of the fast marching algorithm (Sethian 1996, 1999) to compute a set of distance properties originating from the wells and propagating inside the connected reservoir flow units. These path lengths weighted by permeability are analogous to the least resistive paths of Hird & Dubrule (1995), and have been used to determine regions of higher sweep efficiency between injectors and producers, or regions with higher tortuosity that are likely to indicate zones of bypassed oil.

Increasingly complex connectivity studies In order to study the relationships between connectivity and the stratigraphic and structural aspects of reservoirs, an experimental approach is chosen in which reservoir model complexity and the degrees of realism are progressively increased (Fig. 2). First, simple percolation models are used to reveal the primary controls on geobody connectivity, which include: reservoir net-to-gross, its dimensions and volumes. Then aspects of reservoir stratigraphy and continuity are analyzed in variogram-based statistical models. Further geological dimensions are explored by performing numerous sensitivity studies, in which controls on geobody connectivity are understood by varying one stratigraphic or

Fig. 2. A methodology for studying connectivity in reservoirs. Knowledge and rules are derived from connectivity studies on increasingly complex stochastic conceptual models. Rules are then applied on real field studies and outcrop studies.


structural parameter at a time. Additional experimental studies of models describing specific depositional environments help reveal potential outstanding connectivity issues that may degrade well connectivity. Finally, rules derived from these experiments are used in real field studies, where extra effort is used to highlight potential problems causing reservoirs to compartmentalize.

From percolation theory to numerical geology Percolation theory (Christensen & Moloney 2005) proposes to study connectivity in infinite and stationary random systems. Earlier studies (King 1990) have shown that bulk reservoir connectivity in channelized reservoirs may be viewed as in percolation problems, in which the net-to-gross ratio, above a certain percolation threshold, is a key factor contributing to connectivity. Above the percolation threshold, reservoirs are likely to be well connected, and below that same threshold, reservoirs are likely to compartmentalize. This behaviour is best illustrated by the bulk connectivity v. net-to-gross ‘S-curve’ as show on Figures 3 and 4,

223

and previously described by Hovadik & Larue (2007). The value of the net-to-gross threshold at which a random system begins to percolate, depends on its dimensionality. A 2D system, illustrated by Figure 3, percolates at about 60% net-to-gross, whereas for a 3D system (Fig. 4), the threshold is at a lower 30% net-to-gross. King (1990) used percolation theory to predict 2D and 3D connectivity based on geometrical characterizations of sand deposits. Figure 5 describe the 1D, 2D or 3D nature of geobody connectivity and reservoir to well connectivity in three reservoir architectures: amalgamated channel deposits, straight and parallel channel deposits and sheet deposits. In all three reservoir architectures, mud drapes are modelled at the bottom of every channel or sheet deposit element. The sealing or leaking nature of mud drapes is controlled by randomly turning on or off a volume fraction of the mud drape cells. For amalgamated channelized systems, channel deposits eroding into each other tend to connect in all three spatial dimensions, and behave like in 3D percolation problems, with a percolation threshold around 30% net-to-gross (Fig. 5a, b). For perfectly straight and parallel channels, bulk reservoir connectivity (geobody as well as reservoir to

Fig. 3. From Hovadik & Larue 2007. 2D percolation. Connectivity v. net-to-gross curve illustrated by three examples of the biggest connected clusters colored in red at 3 levels of net-to-gross. At 60% net-to-gross the biggest percolating cluster spans across the entire section.

224


Fig. 4. From Hovadik & Larue 2007. 3D percolation. Connectivity v. net-to-gross curve illustrated by three examples of the biggest connected clusters colored in red at 3 levels of net-to-gross. At 30% net-to-gross the biggest percolating cluster spans across the entire cube.

well) occurs in a 2D vertical plane perpendicular to the channel deposits, for net-to-gross values around 60% (Fig. 5c, d). For sheet like reservoirs, where thin reservoir intervals are separated by shale units, geobody connectivity behaves like in a 1D percolation problem, that is, horizontal shale units act as vertical buffers to geobody connectivity, causing the individual sheeted reservoirs to disconnect even at high net-to-gross ratio. However, reservoir to well connectivity in sheeted reservoirs is typically very high because a vertical well penetrating a stack of true infinite horizontal sheets is connected to all the reservoir intervals (Fig. 5e, f). Hence, reservoir to well connectivity, key to reservoir producibility, may in fact diverge from bulk geobody connectivity. Figure 5 also indicate that mud drapes do not significantly affect static connectivity when they cover 75% or less of the bottom surface of channel deposits. If mud drapes are 100% sealing then static connectivity in channelized systems is reduced to 0. By contrast, static connectivity in sheet-like reservoirs seems unaffected by mud drapes, even if they are 100% sealing. However it is important to note that mud drapes may impact reservoirs in other ways such as increase reservoir tortuosity and or decrease dynamic connectivity. Complementary studies are necessary to further

examine the impact of mud drapes on dynamic aspects of reservoir connectivity. Previous studies have also shown that the volume support effect (King 1990; Larue & Hovadik 2006; Hovadik & Larue 2007) also plays an important role. If the reservoir geobodies are large compared to the entire reservoir volume, or fault block size, then reservoir connectivity can either be very high, if the large geobodies do connect, or be very low if the geobodies do not connect. This phenomenon is expressed by a dispersion of the bulk connectivity v. net-to-gross ‘S-curve’ (Fig. 6).

Varying stratigraphic and structural parameters Percolation studies were helpful at identifying three fundamental factors affecting reservoir connectivity. The primary factor is reservoir net-to-gross. Above a net-to-gross threshold reservoirs begin to percolate, and below that threshold reservoirs begin to compartmentalize. The percolation threshold depends on the second primary factor: the spatial dimensions in which the reservoir sand deposits tend to connect. Finally, the third major factor refers to the volume support effect. If the individual


225

Fig. 5. Connectivity dimensionality in channelized and sheeted reservoir. (a) Channels are sinuous and erode one another: connectivity occurs in all 3 dimensions. (b) For sinuous channel deposits, a net-to-gross threshold lower than 30% indicates that connectivity is 3D. 100% sealing mud drapes reduce static connectivity to 0. (c) Channels are straight and parallel: connectivity occurs in a 2D plane. (d) For straight and parallel channels, a net-to-gross percolation threshold around 60% is consistent with 2D connectivity. 100% sealing mud drapes reduce static connectivity to 0. (e) In sheeted reservoir geobody connectivity is 1D, and connectivity to a vertical well is always high. (f) Geobody and well connectivity v. net-to-gross in sheeted reservoirs. 100% sealing mud drapes do not affect static connectivity.

226


Fig. 6. From Hovadik & Larue 2007. Illustration of the volume support effect in a random stationary and isotropic system. (a) and (b) the volume support of the grid is high compared to the small size of the active red cells. (c) and (d) the volumes support decreases and the connectivity v. net-to-gross ‘S-curve’ disperses.

sand deposits are large compared to the volume of the reservoir or fault block size, then the connectivity v. net-to-gross ‘S-curve’ begins to disperse. Most stratigraphic and structural parameters affecting reservoir connectivity do contribute to either one of the two categories above: that is, the dimensionality of the percolation problem, or the volume support effect. A classification is achieved by varying one parameter at a time along with reservoir net-to-gross and observing how these variations affect the connectivity v. net-to-gross ‘S-curve’. An ‘S-Curve’ that shifts laterally indicates that the examined parameter affects the dimensionality of the percolation problem. On the other hand, a scattered ‘S-Curve’ is an indication that the parameter affects volume support. First, variogram-based models are examined (Hovadik & Larue 2007). Figure 7 reveals that the aspect ratio of the variogram tends to shift the geobody connectivity v. net-to-gross ‘S-curve’ to the right as it becomes less isotropic. As expected, this is an indication that variogram anisotropy affects the dimension of the percolation problem.

On the other hand, continuity ranges of the variogram contribute to the volume support effect. Indeed longer variogram ranges create larger reservoir bodies, causing the connectivity v. net-to-gross ‘S-curve’ to disperse. Subsequently, more parameters were studied using highly parameterized Boolean object models. In channelized reservoirs parameters such as widthto-thickness ratio, deviation, and sinuosity have been previously studied (Larue & Hovadik 2006). Table 1 summarizes these experimental results. Other experiments have been undertaken to classify structural parameters such as: the sealing v. non-sealing character of faults for production time-scales, the angle between major faulting axes, and the size of fault blocks v. the size of the sedimentary deposits. Figure 8 illustrates the effects of varying the offset of a vertical non-sealing fault that is oriented perpendicular to the channel deposits. Increasing fault offsets tend to further disconnect a reservoir. The shifting to the right of the geobody connectivity v. net-to-gross ‘S-curve’ is evidence that fault offsets affect the dimensionality


227

Fig. 7. From Hovadik & Larue 2007. Effect of an anisotropic variogram on the connectivity v. net-to-gross ‘S-curve’ in a random stationary and anisotropic system. (a) Isotropic variogram. An anisotropic variogram (b) shifts the ‘S-curve’ to the right (c).

Table 1. Summary table of various stratigraphic factors: their impact on controlling either connectivity dimensionality or connectivity volume support Stratigraphic factors Variogram range Variogram anisotrophy Channel width and thickness Channel width to thickness ratio Channel parallelism Channel deviation Continuous mudstone bed % Local mudstone drapes Channel clustering

Dimensionality

Volume support X

X X X X X X X X

Fig. 8. (a) 10 m fault offset; (b) 100m fault offset; (c) effect of fault offset on geobody connectivity. A large offset affects connectivity dimensionality by shifting the connectivity v. net-to-gross ‘S-curve’ to the right. See text for discussion.

228


Table 2. Table of various structural factors: their impact on controlling either connectivity dimensionality or connectivity volume support Structural factors No. of sealing faults Fault block size Fault offset Fault length

Dimensionality

Volume support X X X X

of the percolation problem. Instead of connectivity occurring everywhere in the 3D reservoir, connectivity is controlled by the area of overlapping reservoirs on the 2D fault plane. As the fault offset increases, the area of overlapping reservoirs becomes smaller, and has a negative effect on volume support. However, increased dispersion of

connectivity is not observed on the ‘S-curve’, possibly because fault offset has in this example a minor effect on volume support. Table 2 is an attempt at classifying these major structural parameters according to their contribution to either the dimensionality of the percolation problem, or the volume support effect.

Modelling stochastic reservoirs in various depositional settings The previous experiments are useful for exploring and classifying various stratigraphic and structural parameter dimensions in a reservoir, and their impact on geobody connectivity. Yet, in order to fully assess reservoir connectivity in real reservoirs, four issues need to be addressed. First, real reservoirs have multiple parameters varying at the same time, and therefore it is necessary to account

Fig. 9. Reservoir to well connectivity in 3 deepwater reservoir systems. (a) Illustrations of a slope valley channel complex, a weakly confined channel complex, and an unconfined channel and sheet complex. Reservoir facies are colour-coded. (b) Parameterizations of reservoir geomorphology in all 3 systems. (c) Reservoir to well connectivity v. net-to-gross ‘S-curves’. In the weakly confined channel complex, the net-to-gross threshold can be as high as 45%. See text for discussion. (d) The Pareto chart indicates that channel drapes, the width to thickness ratio, the thickness of the channel deposits can affect the volume support or the dimensionality and degrade connectivity. The red line is an indication of statistically significance: bars below that line have a statistically insignificant effect on connectivity.


for the complex dependencies between these variables and reservoir connectivity. Second, unlike in percolation problems, reservoir parameters are rarely stationary. In reality, geometrical characterizations of reservoirs vary as sediments are distributed from source to sink, and from axis to margins. Third, unlike in percolation problems, reservoir deposits are not placed fully at random. In fact, various stacking patterns, or specific continuous shale intervals, may be part of a stratigraphic sequence, and have real impact on reservoir connectivity. Finally, since the fundamental condition affecting reservoir producibility is the connectivity

Fig. 9. (Continued)

229

of a reservoir to a well-bore, the focus of this discussion is now shifting from geobody connectivity to reservoir-to-well connectivity. Do these four issues seriously undermine conclusions reached from percolation studies? Or does percolation theory hold true for reservoir models with realistic geology? These concerns can be addressed by studying reservoir to well connectivity in stochastic reservoir models for various depositional environments. The concerns expressed above are best addressed in a series of stochastic models that represent realistic reservoirs in various depositional environments. An experimental design approach is presented here, where low and high parameter ranges are specified according to values published in previous literature (Beauboeuf et al. 2000; Posamentier & Venkatarathnan 2003; Barton et al. 2004). The nonstationary issue is addressed by using probability cubes describing local probabilities of net-to-gross ratio, at any location in the reservoir, as soft conditioning to a Multiple-Point Statistics algorithm. The third issue, regarding specific stratigraphic sequences can be addressed using advanced modelling algorithms, such as Multiple-Point statistics simulation (Strebelle & Journel 2001), object modelling or an Event Based approach (Pyrcz & Strebelle 2008). Figure 9a represents three stochastic reservoirs in three deepwater environments: a slope valley channel complex, a weakly confined channel complex, and an unconfined channel and sheet complex. In all three environments, stratigraphic parameters (Fig. 9b) were varied according to ranges observed on outcrop studies (Beauboeuf et al. 2000; Barton et al. 2004). A Boolean modelling approach was used to create models honouring detailed internal facies architectures, such as reservoir channel axis and margins, channel mud drapes and reservoir sheets. A full factorial design type of experiment has been used to generate models, in all three environments, corresponding to all possible parameter value combinations, with the net-to-gross input parameter ranging from 0.1 to 0.9. Finally, for all models reservoir to well connectivity was computed for a pair of vertical well injectors and producers, and plotted against net-to-gross (Fig. 9c). Several lessons are learned from modelling exercises in various depositional settings, like the one described in Figure 9. Among the three depositional environments studied here, reservoir to well connectivity is best achieved in unconfined sheeted reservoirs, and is worst in weakly confined channelized reservoirs. However, in all three depositional environments, the concept that reservoirs are mostly connected above a percolation threshold, ranging from 30 to 45%, holds true (Fig. 9c). For the slope valley channel complex environment, the 30% net-to-gross threshold is indicative of true 3D

230


connectivity. However, in the weakly confined channel complex, the higher threshold indicates that certain variables do impact connectivity dimensionality and volume support. The Pareto plot on Figure 9d ranks the experimental variables according to their impact on reservoir to well connectivity. Both the Pareto plot and Table 1 reveal that in the weakly confined channel complex, mud drapes and straight and parallel channels affect connectivity dimensionality, whereas width and thickness affect connectivity volume support. The combined effects of these factors yield to higher percolation thresholds around 45%. This indicates that some stratigraphic factors can affect the dimensionality of the percolation threshold, to values between 2D and 3D thresholds (Larue & Hovadik 2006). Below the threshold, a large range of connectivity behaviours does occur. In some cases, reservoirs become quickly compartmentalized as reservoir net-to-gross drops below 30%. In other cases, reservoirs may still be very well connected to injection or production wells, even at net-to-gross values as low as 10%. Under these realistic conditions, the reservoir to well connectivity v. netto-gross plot is no longer shaped in an ‘S-curve’, but more like in a ‘cascade zone’ (Larue & Hovadik 2006). The reason for this ‘smearing out’ of the connectivity ‘S-curve’ has to do with the complex interactions between input model parameters, simultaneously affecting reservoir dimensionality in both ways, as well as affecting volume support. Further experiments are required to address the issue of reservoir net-to-gross spatial nonstationarity. In the experiments described here, net-to-gross non-stationarity is caused by local variations of reservoir confinement and does not appear to significantly affect reservoir to well connectivity; especially when the average net-to-gross of the reservoir is above the 30 to 45% percolation threshold. However, defining a single average value for net-to-gross in a non-stationary reservoir is itself problematic. Non-stationarity can conceivably take infinite forms and therefore can impact either or both the dimensionality of the reservoir or volume support. In addressing the third issue, certain outstanding stratigraphic features become apparent, that if present, strongly degrade reservoir connectivity. The Pareto chart presented in Figure 9d suggests that mudstone drapes can hamper reservoir to well connectivity in a weakly confined channel complex. Certainly, mudstone beds (Larue 2004), if continuous, may compartmentalize a reservoir creating a stack of 2D layered reservoirs with a high percolation threshold for geobody connectivity (Fig. 5e, f). Continuous mudstone drapes covering the base of a channel sand deposit may cause serious compartmentalization effects, although their

continuity has been previously debated (Barton et al. 2004; Larue & Hovadik 2006). Compensational stacking patterns are another outstanding stratigraphic feature, that if present cause sand deposits to avoid each other, leading to high net-to-gross reservoirs with poor reservoir to well connectivity. But as for continuous mudstone beds, the existence of compensational stacking patterns at the reservoir scale is largely conceptual: field examples are required to address their true significance (Larue & Hovadik 2006). Finally, a third outstanding characteristic refers to local reservoir heterogeneities, including: local mudstones beds, or diagenetic alterations of the pore system leading to either higher or lower permeabilities. Although having an effect on the local scale, local heterogeneities typically do not affect global measures of static reservoir to well connectivity. Yet, it is not too early to stress that local heterogeneities do affect dynamic aspects of connectivity, continuity and tortuosity (Larue & Legarre 2004). Insights from previous percolation studies have been reaffirmed in the parameter sensitivity studies. First, the primary stratigraphic factor contributing to reservoir to well connectivity is the net-to-gross ratio. At high to moderate net-to-gross (100% to 30%), reservoirs are typically connected to wells; at low net-to-gross (below 30%) reservoirs may or may not be connected to wells. As in percolation, the connectivity threshold between low and high net-to-gross is determined by the dimensionality of the reservoirs. Reservoirs connecting in a two dimensional plane as in the case of straight and parallel channel deposits, have a net-to-gross threshold defined around 60% net-to-gross. For reservoir architectures made of amalgamated channel deposits eroding one into another, reservoir connectivity occurring in three dimensions yields to a lower net-to-gross threshold defined around 30%. Certain stratigraphic scenarios may present transitioning connectivity dimensions between 3D and 2D, with net-to-gross thresholds between 30 and 60%. Finally the third factor revealed by percolation theory is the notion of volume support effect. Large sedimentary deposits in small volumes produce reservoirs that have an equal probability of being highly or poorly connected. Stochastic, and yet realistic, reservoir models in various depositional settings reveal that when complex interactions occur between multiple stratigraphic and structural parameters, accurately predicting reservoir to well connectivity at lower net-to-gross is a very difficult task. Indeed, below the percolation threshold a wide range of connectivity behaviours occur, from highly connected reservoirs, to very compartmentalized reservoirs. On the other hand, above the threshold, reservoirs are typically connected to wells, unless outstanding


stratigraphic features occur, such as large continuous mudstone beds, or compensational stacking patterns. These features have the potential of seriously degrading reservoir to well connectivity, yet it is debatable how often they appear in real reservoirs. Finally local heterogeneities such as local mudstone beds, or local diagenetic alterations, do not affect reservoir connectivity globally, but may locally affect aspects of reservoir continuity and tortuosity (Larue & Legarre 2004).

Interactions between reservoir stratigraphy and structure and their effects on connectivity At this point in the discussion, aspects of reservoir stratigraphy controlling reservoir to well connectivity have been described independently of reservoir structure. The experiment illustrated in Figure 10 is an attempt at sorting some of the stratigraphic and structural interactions in a faulted channelized reservoir. A series of numerical reservoir models (Fig. 10a) have been generated by varying simultaneously stratigraphic and structural parameters (Fig. 10b). The full combination of all parameter values led to 2160 stochastic reservoir architectures, for which reservoir to well connectivity was computed for a five spot patterns of well injectors and well producers. Statistical ranking of the input parameters indicates that the sealing or leaking nature of the faults has a major impact on reservoir to well connectivity (Fig. 10c). If faults are sealing, reservoir connectivity to injecting and producing wells can be achieved if fault compartments are large enough to accommodate both well types, or if faults are short enough to be contoured by connecting tortuous paths. The first two parameters affecting reservoir to well connectivity are stratigraphic (net-to-gross, and width-to-thickness ratio), followed by the geostatistical seed and the fault block size (Fig. 10d). The implication for reservoir to well connectivity is that sealing faults and larger channel deposits reduce the volume support for connectivity. With weak volume support, the stochastic model realizations (controlled by the seed number) lead to a greater dispersion of the reservoir to well connectivity v. net-to-gross ‘S-curve’ (Fig. 10c). On the other hand, if faults are leaking (Fig. 10e), the reservoir has enough volume support for better connectivity. The tornado diagram (Fig. 10f) indicates that thickness, a function of the width and width to thickness ratio of the channel deposits, and the maximum fault offset, are the two parameters that have the highest impact on connectivity. Indeed, compartmentalization across a leaking fault

231

plane (Fig. 10g) may occur if the offset is large compared to the thickness of the reservoir beds, especially at low net-to-gross. The impact of faults on reservoir compartmentalization depends foremost on their sealing or nonsealing nature, due to the presence or absence of shale smear or cementation along the fault plane. If a fault is sealing, the size of the fault block compartment, relative to size of the reservoir sedimentary deposits, may produce volume support issues. In contrast, if faults are leaking, and the fault throw is small relative to the thickness of the reservoir intervals, connectivity is mostly controlled by stratigraphic factors affecting the dimensionality of the reservoir. If faults are leaking, and the fault throw is large relative to the thickness of the reservoir intervals, then reservoirs units may not overlap across a fault plane, especially at low net-to-gross ratios, causing compartmentalization. Moreover, in the same way that local heterogeneities may impact local aspects of connectivity and tortuosity, leaking faults may also affect connectivity locally.

Problems defining dynamic connectivity Up to this point, the main topic has been the structural and stratigraphic controls affecting static reservoir to well connectivity, where static connectivity is defined as the volume fraction of reservoir flow units connected to a well-bore and measured on numerical reservoir models using global statistics. It has also been briefly mentioned that local heterogeneities and leaking faults may impact dynamic connectivity. Finally, measures attempting to characterize connectivity locally have also been briefly reviewed. Before discussing any details of ‘dynamic connectivity’, it is necessary to clarify the term. The word ‘dynamic’ suggests that connectivity is a reservoir property that is time dependent. Certainly, as a reservoir is being produced, certain parts may be drained sooner than others, and all hydrocarbon may not be recoverable, even if static connectivity reaches 100%. However, a concise definition of dynamic connectivity is yet to be proposed. In previous literature, (Hird & Dubrule 1995; Tang & Ji 2006; Snedden et al. 2007 or Tang 2007) dynamic connectivity has been used as a proxy for reservoir producibility, the reservoir recovery factor as a function of time, or recovery factor as a function of pore volume injected. The term has also been used for wells that appear connected after undergoing pressure tests. Sometimes the culprit for bypassed oil is lack of ‘dynamic connectivity’. In spite of lacking a clear definition for dynamic connectivity, it can be stated that dynamic

232


Fig. 10. Interactions between reservoir stratigraphy and structure and their effects on connectivity. (a) Examples of experimental models. Faults are vertical, have variable, length, density, and offset and are perpendicular to reservoir channel deposits. Sand-bodies (in orange) are connected to well injectors (in blue) and a well producer (in green). (b) Table summarizing the parameterization of reservoir structure and stratigraphy. (c) In the case of sealing faults, small compartments further decrease the volume support for connectivity, resulting in a wider dispersion of the connectivity v. net-to-gross ‘S-curve’. (d) Tornado diagram ranking factors controlling reservoir connectivity in the case of sealing faults. See text for discussion. (e) In the case of large fault offsets, a shift of the connectivity v. net-to-gross ‘S-curve’ to the right is an indication of worsening connectivity due to weaker dimensionality. (f) Tornado diagram ranking factors controlling reservoir connectivity in the case of leaking faults. See text for discussion. (g) Connectivity across a fault plane in the case of a leaking fault. Overlapping sand bodies across the fault plane are marked in red.

connectivity does guarantee static reservoir to well connectivity. However the direct opposite proposition is not true. That is, static reservoir to well connectivity is a necessary but insufficient condition for dynamic connectivity. Indeed, dynamic connectivity is a function of static connectivity and many other factors that also impact reservoir producibility: such as time, reservoir volumes, permeability and heterogeneity, path lengths, pressure gradient, fluid properties, relative permeability, and many other factors. Because of the long list of factors affecting dynamic connectivity, static reservoir to well connectivity is much easier to characterize then dynamic connectivity in terms of stratigraphic and structural reservoir architectures.

Three aspects of stratigraphy and structure are discussed here, all affecting dynamic connectivity, but not static connectivity. First, the concept of reservoir tortuosity is briefly reviewed here, followed by aspects of permeability heterogeneity and its impact on reservoir recovery, and one example of leaking faults creating dynamic compartmentalization.

Reservoir tortuosity Reservoir tortuosity refers to the minimum path length between two points, typically completion and injection zones, in a connected reservoir (Larue & Friedmann 2005). Differences in reservoir


233

Fig. 10. (Continued)

tortuosity may be significant in connected reservoirs at low net-to-gross, or in highly sinuous or deviated channel deposits, or may be caused by local heterogeneities. Reservoir tortuosity can be quantified in numerical characterizations of reservoirs by computing geometrical path lengths that are fully contained within the connected flow units (Hovadik & Larue 2007). Figure 11 describes three reservoirs with identical rock properties that are 100% connected to a producing and injecting well pair. The recovery v. pore volume injected curves show that the sweep efficiencies on the straight and sinuous reservoirs are about the same, despite the two models having very different path lengths between the injecting and producing wells. However, the third reservoir has much lower sweep efficiency owing to early water breakthrough in the straight channel. Consequently, tortuosity and permeability are comparable

properties, in that their mean value affects the rates at which a reservoir can be produced, and their heterogeneity strongly impacts sweep efficiency.

Permeability heterogeneity Mean permeability of a reservoir impacts the rates at which a reservoir can be produced. On the other hand permeability heterogeneity can cause flow fingering and inefficient sweep in the cases in which water or gas is displacing oil. Permeability heterogeneity can be characterized by the global permeability histogram of reservoir rocks. Typically, the coefficient of variation of the permeability distribution is an indication of heterogeneity (Jensen et al. 1987; Jensen & Lake 1988). Previous characterizations of permeability heterogeneity include the Dykstra –Parsons coefficient (Dykstra & Parsons 1950) originally used as an indicator of early water

234


break-through in highly stratified ‘layer cake’ reservoirs, and the Lorenz coefficient, a characterization of sweep efficiency (Jensen & Lake 1991). Permeability heterogeneity can also be expressed in local and regional permeability trends. Figure 12a illustrates a reservoir produced by two well injectors and one well producer. Then permeability is modeled using various local trends and keeping the global permeability histogram constant. In the first model (0D), at any cell location, permeabilities are randomly drawn from the histogram, thus resulting in a ‘salt and pepper’ type of model. For the second permeability model (3D), a variogram was used to create areal continuity but no vertical continuity. In the third permeability model (3Dþ), permeability continuity has been extended vertically. The fourth model (3D þ trend), accounts for trends from higher permeability channel axis to lower permeability channel margins. The next model (3D þ trend þ 50% drapes) extends the permeability trend with partially

sealing mud drapes at the base of the channel. The last model (3D þ trend þ 100% drapes) has all drapes perfectly continuous and impermeable. All models were flow simulated under the exact same well constraints. The differences in the oil recovery v. pore volume injected (Fig. 12b) is directly linked to various degrees of bypassed oil caused by the various types of permeability heterogeneity. The first three models have similar recovery profiles. Adding 3D and 3Dþ heterogeneities did not impact results significantly. However, it is noted that far less water was injected in the ‘salt and pepper’ model. The non-stationary trend created thief zones responsible for early water breakthrough. Tortuosity and compartmentalization due to mud drapes further reduced sweep efficiency. Figure 13 highlights the combined effects of permeability heterogeneity and continuity. A series of models have been built by varying the Dykstra–Parsons coefficient and the variogram ranges used for simulating permeability using a

Fig. 11. Effects of tortuosity on sweep efficiency. (a) Single straight channel produced by pore volume displacement with one water injector (blue) and one producer (green). (b) Meandering channel produced under same conditions as for channel A. (c) Straight and meandering channel produced together. (d) Recovery factor v. pore volume injected for scenarios A, B and C produced under the same pressure constraints at the producers and same pressure constraints at the injectors. See text for discussion.


Sequential Gaussian algorithm. In all models, porosity and water saturation have been kept constant. Figure 13a illustrates water saturation maps computed at an arbitrary 0.6 pore volume injected. Figure 13b plots recovery factor as a function of the Dykstra–Parsons coefficient and the variogram range. If the correlation length is short, then varying the Dykstra–Parsons coefficient does not have a dramatic effect on sweep efficiency. If the Dykstra–Parsons coefficient is low, varying the correlation has also limited effect on sweep efficiency. Worst sweep efficiency is achieved by maximizing

235

both permeability heterogeneity and the correlation lengths.

Faults Faults and fault transmissibility are other factors affecting ‘dynamic connectivity’ and are briefly illustrated here. Fault transmissibility does act like a vertical baffle to flow, and is a convenient parameter often used to calibrate dynamic simulation with production history (Fisher & Jolley 2007). In Figure 14 a confined channelized reservoir is

Fig. 12. Effects of permeability trends on sweep efficiency. 0D: permeability is randomly distributed inside the channel deposits in a ‘salt and pepper’ fashion. 3D: 3D correlations are being introduced to the permeability model. 3Dþ: introduction of a shorter vertical correlation length to the permeability model. 3D þ trend: introduction of non stationarity to the permeability model, the channel axis has higher permeabilities then the channel margins. 3D þ trend þ 50% drapes: impermeable drapes wrap the channel deposits and 50% of the draping surface is leaking. 3D þ trend þ 100% drapes: impermeable drapes wrap the channel deposits and the draping surface has no holes. Well producer is green and two well injectors are blue. (a) Illustration of various trends. (b) Impact on the recovery factors as functions of pore volume injected. See text for discussion.

236


Fig. 13. Combined effects of permeability heterogeneity and continuity. (a) The Dykstra–Parsons’ coefficient ranges from 0.6 to 0.9. The variogram range varies from short to long. 4 well injectors are marked in blue. The well producer is marked in green. Maps of water saturation are displayed at 0.6 pore volume injected. Worst sweep occurs in the case where a high Dykstra– Parsons’ is combined with a long variogram range. (b) Response model of recovery factor at 0.6 pore volume injected as a function of the Dykstra– Parsons’ coefficient (Vdp) and the variogram range. See text for discussion.


237

Fig. 14. Impact of faults on sweep efficiency. (a) Non faulted reservoir. (b) Faulted reservoir. (c) The faulted reservoir has reduced sweep efficiency. (d) The fault with reduced transmissibility impacts reservoir production rate. See text for discussion.

Fig. 15. The Connectivity Flow Diagram is an attempt at summarizing and ordering the major concerns about a reservoir structure and stratigraphy that impact reservoir compartmentalization and that need to be addressed.

238


produced by pore voidage displacement using a down-dip well injector and an up-dip well producer. Three cases are examined: a non-faulted reservoir, a reservoir cut by a perfectly leaking fault, and a reservoir cut by a leaking fault with transmissibility reduced by a factor of 0.1. Results are shown on Figure 14c, d. First, recovery factor as a function of pore volume injected indicates that the presence of a leaking fault does reduce sweep efficiency. The difference in sweep could be linked to differences in tortuosity. However, the fault transmissibility multiplier does not have an impact on sweep efficiency but, rather affects the rate at which the reservoir is produced.

Designing a connectivity flow diagram At this point in the study, an effort has been made to classify stratigraphic and structural parameters affecting static reservoir connectivity. Then, a few more examples of key parameters affecting

‘dynamic connectivity’ have been described. In the next section, the connectivity flow diagram (Fig. 15) is presented, as an attempt to further sort the main geological features affecting reservoir connectivity. The connectivity flow diagram is intended as an aid for earth scientists who are trying to explore potential downsides in reservoir to well connectivity. However, the connectivity flow diagram is not a replacement to reservoir modelling, since it has already been shown that no tool can sort all the complex parameter interdependencies controlling connectivity. Rather, the tool can be used as an assistant to help formulate key questions concerning reservoir connectivity and could be very valuable in the appraisal dataacquisition phase of field development. The connectivity flow diagram is a guide through a series of connectivity checks. If a reservoir passes all the connectivity checks, then connectivity uncertainty is low and is typically 100%. Following the flow connectivity diagram, the first test is related to reservoir net-to-gross. Is reservoir net-to-gross

Fig. 16. Applying the connectivity flow diagram in a sparse well environment where major faults are poorly imaged. (a) Three possible interpretations of major faults. (b) The presence of faults needs to be addressed first. See text for discussion.


less then 30%? Below 30% net-to-gross, connectivity uncertainty is highest and can range from very low to 100%. If reservoir net-to-gross is above 30%, more questions need to be answered. The second issue to be addressed is related to the presence of faults. If faults are not present, then the next topic concerns volume support. But if faults are present, one should investigate whether the faults are sealing or leaking. If faults are

239

sealing, then again, the volume support needs closer examination. However, if faults are leaking, do they present large offsets? Faults with large offsets are likely to create compartmentalization challenges. But if leaking faults have small offsets, then what about volume support or dimensionality effects? A reservoir has enough volume support if the individual reservoir elements are small compared to the actual reservoir or fault

Fig. 17. Applying the connectivity flow diagram in a sparse well environment where major faults are imaged, but their sealing or leaking nature needs to be addressed. (a) Example of relations between the reservoir faults and stratigraphic architecture. (b) Small fault blocks and large sedimentary deposits may limit the volume support needed for good connectivity. See text for discussion.

240


block size. If that is not the case, chances are that the reservoir will experience compartmentalization issues. On the other hand, if volume support does not appear to be a problem, then one should investigate next the dimensionality of the reservoir. If the reservoir presents two dimensional characteristics (for example, continuous mudstone units define thin 2D reservoir compartments), then the percolation threshold can be as high as 60%. If the reservoir connects in all three dimensions, then local heterogeneities are next. A reservoir successfully passing all checks, has a high net-to-gross ratio, is non faulted, or the fault are leaking with small offsets, has enough volume support, connects in all three dimensions, and is absent of local heterogeneities. Such a reservoir is likely to have good

static connectivity. However, static connectivity does not guarantee reservoir producibility. Therefore, at this point in the connectivity flow diagram, many more questions concerning dynamic connectivity need to be raised and addressed. Examples of factors affecting ‘dynamic connectivity’ have briefly been discussed and include: tortuosity, permeability heterogeneity, or fault offset and transmissibility.

Assessing reservoir connectivity using the connectivity flow diagram In this section, the connectivity flow diagram is used to raise compartmentalization concerns in three real oil field examples.

Fig. 18. Applying the Connectivity Flow Diagram in a reservoir under development. (a) Cross section of the reservoir. The colour scale is an indication of pressure in the fault compartments (red is high pressure, green is low pressure). (b) Major connectivity concerns have been cleared. However further assessment of dynamic connectivity is needed. See text for discussion.


In the first example (Fig. 16a), oil has been discovered and no seismic attributes are available to characterize the reservoir. Reservoir net-to-gross in the discovery well and in an appraisal well is greater then 60%. Following the connectivity flow diagram, the first outstanding compartmentalization issue to be addressed is the presence and extent of faults (Fig. 16b). Unfortunately, seismic data cannot be used to image faults. More appraisal wells need to be drilled, and well pressure tests may be performed to investigate whether or not these wells are connected. Initial numerical models can attempt to study the effects of various different reservoir architectures, and their sensitivities to volume support issues. In the second example (Fig. 17a), another field has been appraised by three wells, all indicating that the net-to-gross is above 75%. Seismic images reveal the existence of large faults. According to the connectivity flow diagram, structural compartmentalization created by large sealing faults visible on seismic data, reduce the volume support necessary for the good connectivity of channel and sheet deposits. More information is needed to assess the sealing or leaking nature of faults (Fig. 17b). In the third example (Fig. 18a), a reservoir under development has five producing wells, and three water injecting wells. Reservoir net-to-gross appears to be approximately 55%. Major faults identified on seismic and fault plane profiles appear to be sealing, creating structural compartments that need to be produced individually. However the reservoir appears to have enough volumes support for individual reservoir units to connect in all three dimensions. Globally, static connectivity appears to be good. However, the connectivity flow diagram indicates that local permeability heterogeneity and tortuosity need further examination, because of their potential impact on sweep efficiency (Fig. 18b). Hence, ‘Dynamic connectivity’ can be closely examined by constructing fit-for-purpose numerical models used for dynamic simulations studies.

Conclusions Both the reservoir structure and stratigraphy affect static reservoir to well connectivity. Among all stratigraphic factors controlling connectivity, the reservoir net-to-gross is the first to consider. Between 0 and 30% net-to-gross, stratigraphic compartmentalization is likely, and may result in large parts of a reservoir being under produced. Between 30 and 60% net-to-gross, lack of volume support or dimensionality problems can also yield to significant connectivity issues. Above 60% net-to-gross, reservoir to well connectivity is unlikely to be a problem. Sealing faults seriously hamper reservoir

241

connectivity. If a fault block compartment is small compared to the reservoir geobodies, then lack of volume support further degrades connectivity. If a fault is leaking, connectivity can still be achieved if the fault throw is small compared to the reservoir thickness. Finally, a connectivity flow diagram is proposed as a guide to help formulate key questions concerning reservoir connectivity. Static reservoir to well connectivity is a requirement for a reservoir to be producible. But static connectivity in itself is not a guarantee for reasonable hydrocarbon recovery rates and should not be confused with the term ‘dynamic connectivity’. Global and local characterizations of static reservoir to well connectivity can be used and measured on a static numerical reservoir model. Evaluating ‘dynamic connectivity’ requires a reliable numerical model used for dynamic flow simulations. ‘Dynamic connectivity’ is a function of static connectivity plus many other factors, such as tortuosity, permeability heterogeneity, fault transmissibility, but also time, pressure, and even the trading cost of a barrel of oil. . . We thank the Chevron Energy Technology Company for continued research support over the past years. We are also greatly thankful to Stephen Naruk, John Snedden and Peter Vrolijk for their insightful reviews.

References Ainsworth, 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. Allen, J. R. L. 1978. Studies in fluviatile sedimentation: an exploratory quantitative model for the architecture of avulsion-controlled alluvial suites. Sedimentary Geology, 21, 129–147. Bailey, W. R., Manzocchi, T. et al. 2002. The effects of faults on the 3-D connectivity of reservoir bodies: a case study from the East Pennine Coalfield, UK. Petroleum Geoscience, 8, 263– 277. Ballin, P. R., Solano, R., Hird, K. B. & Volz, R. F. 2002. New reservoir dynamic connectivity measurement for efficient well placement strategy analysis under depletion. Society of Petroleum Engineers, SPE Paper 77375, presented at the SPE Annual Technical Conference and Exhibition, San Antonio, Texas. Barton, M., O’Byrne, C. et al. 2004. Understanding hydrocarbon recovery in deepwater reservoirs: modeling outcrop data in the third dimension. Annual AAPG Meeting, 2007. Available online at: http://www.searchanddiscovery.net/documents/ abstracts/annual2004/Dallas/Barton.htm Beauboeuf, R. T., Rossen, C., Zelt, F., Sullivan, M. D., Mohrig, D. & Jennette, D. C. 2000. Deep-water sandstones, Brushy Canyon Formation West Texas. Field Guide for AAPG Hedberg Field Research

242


Conference, April 15–20, 1999, AAPG Continuing Education Course Note Series #40, 1– 48. Christensen, K. & Moloney, N. R. 2005. Complexity and criticality. Advanced Physics Texts Volume 1. Imperial College Press, London, 15– 29. Dykstra, H. & Parsons, R. L. 1950. The prediction of oil recovery by water flood. In: Dykstra, H. & Parsons, R. L. (eds) Secondary Recovery of Oil in the United States, Principles and Practice. 2nd edn. American Petroleum Institute, New York, 160–174. Fisher, Q. J. & Jolley, S. J. 2007. Treatment of faults in production simulation models. In: Jolley, S. J., Barr, D., Walsh, J. J. & Knipe, R. J. (eds) Structurally Complex Reservoirs. Geological Society, London, Special Publications, 292, 219–233. Hird, K. B. & Dubrule, O. 1995. Quantification of reservoir connectivity for reservoir description applications. Society of Petroleum Engineers, SPE Paper 30571, presented at the SPE Annual Technical Conference and Exhibition, Formation Evaluation and Reservoir Geology, Dallas, Texas. Hird, K. B. & Kelkar, M. G. 1994. Conditional simulation method for reservoir description using spatial and well-performance constraints. SPE Reservoir Engineering, 9, May, 145– 152. Hovadik, J. & Larue, D. K. 2007. Static characterizations of reservoirs: refining the concepts of connectivity and continuity. Petroleum Geoscience, 13, 195–221. Jensen, J. L. & Lake, L. W. 1988. The influence of sample size and permeability distribution on heterogeneity measures. SPE Reservoir Engineering, 3, 629–637. Jensen, J. L. & Lake, L. W. 1991. A review of heterogeneity measures used in reservoir characterization. In Situ, 15, 409– 439. Jensen, J. L., Hinkley, D. V. & Lake, L. W. 1987. A statistical study of reservoir permeability distributions, correlations and averages. SPE Formation Evaluation, 2, 461– 468. Journel, A. G. & Isaaks, E. H. 1984. Conditional indicator simulation: application to a Saskatchewan uranium deposit. Mathematical Geology, 16, 685–718. King, P. R. 1990. The connectivity and conductivity of overlapping sand bodies. In: Buller, A. T., Berg, E., Hjelmeland, O., Kleppe, J., Torsæter, O. & Aasen, J. O. (eds) North Sea Oil and Gas, Reservoirs – II. Graham & Trotman, London, 353– 361. Larue, D. K. 2004. Outcrop and waterflood simulation modeling of the 100-foot channel complex, Texas, and the Ainsa II channel complex, Spain: analogs to multistory and multilateral channelized slope reservoirs. In: Grammer, G. M., Harris, P. M. & Eberli, G. P. (eds) Integration of Outcrop and Modern Analogs in Reservoir Modeling. American Association of Petroleum Geologists Memoir, Tulsa, 80, 337–364.

Larue, D. K. & Friedmann, F. 2005. The controversy concerning stratigraphic architecture of channelized reservoirs and recovery by waterflooding. Petroleum Geoscience, 11, 131– 146. Larue, D. K. & Hovadik, J. 2006. Connectivity of channelized reservoirs: a modelling approach. Petroleum Geoscience, 12, 291– 308. Larue, D. K. & Legarre, H. 2004. Flow units, connectivity, and reservoir characterization in a wavedominated deltaic reservoir: Meren Reservoir, Nigeria. American Association of Petroleum Geologists Bulletin, 88, 303–324. Pardo-Iguzquiza, E. & Dowd, P. A. 2003. CONNEC3D a computer program for connectivity analysis of 3D random set models. Computers & Geosciences, 29, 775–785. Posamentier, H. W. & Venkatarathnan, K. 2003. Seismic geomorphology and stratigraphy of depositional elements in deep-water settings. Journal Sedimentary Research, 73, 367 –388. Pyrcz, M. J. & Strebelle, S. 2008. Event-based geostatistical modeling. In: Ortiz, J. & Emery, X. (eds) Geostatistics Santiago, Springer, Netherlands, 1143– 1148. Renjun, W. & Barton, M. 2006. Computer modeling of internal architecture in deep water reservoirs: a quantitative method to estimate connectivity and performance. AAPG Bulletin, 90, April 9 –12, 2006. Sethian, J. A. 1996. A fast marching level set method for monotonically advancing fronts. Applied Mathematics, 93, 1591–1595. Sethian, J. A. 1999. Level Set Methods and Fast Marching Methods. 2nd edn. Cambridge University Press, Cambridge, UK. Snedden, J. W., Vrolijk, P. J., Sumpter, L. T., Sweet, M. L., Barnes, K. R., White, E. & Farrell, M. E. 2007. Reservoir Connectivity: Definitions, Strategies, and Applications. AAPG Annual Convention and Exhibition, 2007. Available online at: http://www. searchanddiscovery.net/abstracts/html/2008/geo_ bahrain/abstracts/meurer.htm Strebelle, S. B. & Journel, A. G. 2001. Reservoir modeling using multiplepoint statistics. Society of Petroleum Engineers, SPE Paper 71324, presented at the SPE Annual Technical Conference and Exhibition, New Orleans, Louisiana, 30 September–3 October. Tang, H. 2007. Rank Reservoir Connectivity using Dynamic Data. AAPG Annual Convention and Exhibition, 2007. Available online at: http://www. searchanddiscovery.net/abstracts/html/2007/annual/ abstracts/lbTangHong.htm Tang, H. & Ji, H. U. 2006. Incorporation of spatial characteristics into volcanic facies and favorable reservoir prediction. SPE Reservoir Evaluation & Engineering, 9, 565–573.

Fault seal calibration: a brief review G. YIELDING*, P. BRETAN & B. FREEMAN Badley Geoscience Limited, North Beck House, North Beck Lane, Hundleby, Spilsby, Lincolnshire, PE23 5NB, UK *Corresponding author (e-mail: [email protected]) Abstract: Calibration is a necessary step in the workflow for prediction of fault seal because there is no direct way to detect the hydraulic behaviour of a fault at the scale of a hydrocarbon trap. Over the last 20 years two general approaches have been developed: (i) (ii)

Measurement of hydraulic properties of fault-zone samples (lab calibration), then mapping these results onto the appropriate parts of trap-bounding faults. Design of simple algorithms which attempt to capture a salient feature of the fault zone (e.g. CSP, SSF, SGR), then looking at known trap-bounding faults to find a relationship between the algorithm and the presence or capacity of a seal (sub-surface calibration).

Seal capacity is typically described by Hg– air threshold pressure in the lab or static pressure differences in the subsurface (e.g. hydrocarbon buoyancy pressure). In addition to likely interpretation and geometry errors in approaches (i) and (ii), further uncertainty is introduced when converting the calibrated seal strength to potential hydrocarbon column height, because of the variability of subsurface hydrocarbon fluids (interfacial tension). Despite these potential problems, the different methodologies typically agree reasonably well in their predictions for fault-seal capacity. However, this agreement may be largely coincidental and is likely to be a response to the heterogeneity of fault-zone structure (especially at intermediate ‘compositions’ or SGR).

Faults frequently form the side-seals to hydrocarbon reservoir compartments. Identification of faults as apparent side-seals usually comes from recognition of different hydrocarbon contacts across a fault, from measurement of different reservoir porepressures in adjacent compartments, or from poor flow performance in a producing reservoir. Note that all of these techniques relate to fluid behaviour measured in adjacent wells, and not to any measurement of the fault itself. In essence, this is the fault seal ‘problem’ – we cannot directly image the sealing properties of fault zones using seismic reflection data (though direct hydrocarbon indicators such as flat spots may locally help). So instead, we must find something we can measure, and then relate that proxy property to the known geological processes that may cause a fault to seal. The act of checking the predictions at known sealing faults is the process of ‘fault seal calibration’. Before proceeding further, it is worth clarifying the word ‘seal’. A dictionary definition (Oxford Paperback Dictionary 1988) is ‘a substance used to close an opening and prevent air or liquid from flowing through it’. This is the context in which the word is used in exploration and appraisal – we are looking for a lithological seal that has prevented hydrocarbon flow on a geological timescale, so that a commercial accumulation is still present when the trap is drilled. However, in a

production/development context, the term ‘sealing fault’ is often used for a fault that acts as an impairment or baffle to flow on a production time-scale, rather than something that completely prevents flow. The influence of a baffle on flow is dependent upon its permeability and thickness, whereas the static trapping capability is mainly dependent upon the capillary properties (see below). In this contribution, we shall focus on the exploration definition, that is, a seal as a lithology or structure that can hold a hydrocarbon column in place over a geological time-scale.

How do faults seal? There are three fundamental conditions which control the sealing behaviour of a fault in siliciclastic sequences (see Fig. 1): (1) (2) (3)

The juxtaposition of the reservoir against sealing lithologies across the fault. Fault-zone products created by deformation during the fault displacement and subsequent evolution. The current stress state of the fault and its proximity to failure (slip).

The first condition is geometric, and provides a way for a shale (or other sealing lithology) to be


244

G. YIELDING ET AL.

Fig. 1. Cartoon to illustrate the three critical factors for a viable fault seal: juxtaposition seal, fault rock seal, and in-situ stress state.

placed in a position where it can be a side-seal to a reservoir compartment. Accurate subsurface mapping of the fault system is critical to assessment of juxtaposition geometries, or as Tearpock & Bishke (2003) succinctly state, ‘If you want to drill more than your share of dry holes, don’t map faults’. Jolley et al. (2007) provide a powerful example of how simple errors in the fault-network geometry can render a subsurface model useless in terms of its flow properties. Although across-fault juxtaposition relationships have traditionally been assessed using fault-plane maps or Allan diagrams (Allan 1989), a true 3D approach is increasingly employed to represent the 3D topology of potential reservoir connections and separations (e.g. Dee et al. 2005). In places where reservoir –reservoir juxtaposition occurs at the fault, it is possible that the fault zone itself can provide the necessary conditions for a side-seal. Most fault-zone rocks have capillary threshold pressures significantly larger than reservoir rocks (e.g. Fisher & Knipe 2001; Sperrevik et al. 2002). The fault-zone rock-type depends upon the composition of the faulted sequence, and the burial/temperature history during and after faulting. We discuss these aspects in more detail below. The third condition describes the in-situ stress state of the fault zone. Barton et al. (1995) present compelling evidence that in low-permeability host rocks the permeability of a fracture surface is controlled by the stress state – fractures are only permeable when critically stressed, that is, close to failure. Hence a critically-stressed fault may act as leak zone through the reservoir overburden. It is important to note that a critical stress state is a vertical leakage criterion, not a seal criterion: if a fault is not critically stressed then this tells us nothing about whether it may be a side-seal to a compartment. The juxtaposition geometry and/or

fault-zone properties are required to set up the seal. So the three conditions illustrated in Figure 1 may be summarized as (Jones & Hillis 2003): A fault is sealing: † IF the reservoir is juxtaposed against a sealing lithology, † OR fault rock provides a seal, † AND the fault is not critically stressed. The rest of this paper is concerned principally with the second of these three sealing conditions.

Fault-zone processes In this section we very briefly review the formation of fault rock in siliciclastic reservoir sequences. The main factors influencing the nature of the deformation products found in fault zones are: † The composition and rheology of the wall-rocks that are slipping past each other at the fault, in particular their content of fine-grained phyllosilicate clay minerals. † The stress conditions at the time of faulting, which are most strongly controlled by the tectonic setting and the initial depth of burial during faulting. † The maximum temperature experienced by the fault zone after faulting, controlled by the maximum post-faulting burial depth and the geothermal gradient. Figure 2 provides examples of the typical fault rocks produced by different combinations of the first two factors (plotted schematically on the horizontal and vertical axes respectively). When the wall-rocks contain significant beds of clay or shale, then clay smears or shale smears can be produced down the fault plane, depending on the burial depth and degree of consolidation of the

FAULT SEAL CALIBRATION

245

Fig. 2. Schematic plot to illustrate the main fault rock types generated in siliciclastic sequences. The axes represent 2 of the 3 main controls (clay content, and stress conditions during faulting); the third main control is post-faulting temperature history. Photographs from Lindsay et al. (1993), Fristad et al. (1997), Fossen & Gabrielsen (2005), van der Zee & Urai (2005), Fossen et al. (2007).

beds. Clay smears represent a ductile deformation, and are often wedge-shaped with the thickest smear immediately adjacent to the source bed (e.g. Weber et al. 1978; Lehner & Pilaar 1997; Aydin & Eyal 2002; van der Zee & Urai 2005; Eichhubl et al. 2005). Penetration of such features in the subsurface demonstrates their ability to sustain differences in fluid type and pore-pressure (e.g. Færseth 2006). In cases where the sequence was more deeply buried and lithified during faulting, shale smears can be generated by abrasion rather than ductile flow, producing thin shale veneers of approximately constant thickness along the fault plane (Lindsay et al. 1993). Both clay smears and shale smears tend to become disrupted with increasing fault displacement: disruption can occur at any point in the smear but is often near the upthrown or downthrown source bed (Childs et al. 2007). In clay-poor sequences, the principal fault rock types are disaggregation zones and cataclasites (Fisher & Knipe 2001; Sperrevik et al. 2002;

Fossen & Gabrielsen 2005). Disaggregation zones are formed during fault slip at relatively low confining stress, and are characterized by grain reorganization without grain fracturing. Consequently they tend to have similar hydraulic properties to their host sandstones. At higher effective stress (e.g. burial .1 km) cataclastic processes become significant and the resultant grain fragments infill the pore-space resulting in higher capillary threshold pressure and lower permeability. Both disaggregation zones and cataclasites are prone to post-deformation quartz cementation if subjected to temperatures .90 8C (e.g. Fisher et al. 2003), which would be at depths .3 km in typical geothermal gradients. The process of quartz dissolution and re-precipitation might be expected to be more significant in cataclasites because of the greater (and uncoated) surface area of the poorly sorted, comminuted grain fragments. In sediments of intermediate composition (15– 40% phyllosilicate), microfaults are characterized

246

G. YIELDING ET AL.

by a texture termed phyllosilicate-framework fault rock (Fisher & Knipe 1998), or more simply claymatrix gouge or shaly gouge (e.g. Gibson 1998). Deformation-induced mixing of quartz grains and clay matrix occurs, generally without grain fracturing. With increased burial, both chemical and mechanical compaction may result. With such a variety of fault rocks, it is clear that the ‘sealing capacity of a fault’ is not generally a single number, but will vary both along-strike and down-dip on the fault surface. Depending on the disposition of sand-rich or clay-rich beds in the host sequence, and the amount of displacement on the fault, different parts of the fault zone may have very different compositions and fault rocks because of the different beds dragged past each point. Stress conditions during faulting, and temperature history after faulting, are also likely to vary over large (seismically-mappable) fault surfaces. Therefore, to predict the sealing behaviour of the fault we need a methodology to estimate the probable fault-zone composition at each point on the fault surface, and couple this to relevant parameters in the burial (stress/temperature) history.

Fault seal algorithms Fault seal algorithms are simply a way of automating part of the process of predicting the properties of the fault zone. They fall into 2 broad categories (Yielding et al. 1997): smear factors which describe aspects of smearing of clay or shale beds, and gouge ratio which notionally describes the composition of the fault rock. The Clay Smear Potential (CSP) was developed by Shell in the 1980s (Bouvier et al. 1989; Fulljames et al. 1997; Lehner & Pilaar 1997). Based on observations of ductile clay smears, and a model of Newtonian viscous flow, the CSP at any point between the two halves of an offset clay bed is dependent on the square of the clay bed thickness and the distance of the point from the nearest half of the bed (Fig. 3). If multiple clay beds are involved then the result is the sum of their contributions. The CSP does not predict the actual thickness of the clay smear, but rather the general likelihood of clay smear to be developed (high CSP ¼ higher likelihood of smear). The relationship gives greater weight to thicker clay beds, and the distance term relates to the tapering of individual clay smear wedges away from the source bed. Doughty (2003) suggests that a thickness exponent of 1.7 rather than 2 gives a better description of clay smears at outcrop for the Calabacillas Fault, New Mexico. The Shale Smear Factor (SSF) was published by Lindsay et al. (1993) following studies of abrasion-type shale smears at outcrop. SSF is

simply the ratio of throw to bed thickness for a single offset shale bed (Fig. 3). This ratio is clearly a single number that does not vary up/ down the fault plane between the offset halves of the bed, but would vary laterally as fault throw changes. A low SSF corresponds to a higher likelihood of intact smear (see below). Gouge ratio methods make a simple assumption that the wall rocks are, on average, mixed into the fault zone in a uniform way. Therefore the bulk composition of any part of the fault rock will be the same as the bulk composition of the wall rocks that have slipped past that part of the fault. The key compositional component for sealing potential is the clay content, since small grain sizes lead to small pore-throats and therefore to high capillary threshold pressure. The Shale Gouge Ratio (SGR) algorithm (Yielding et al. 1997; Freeman et al. 1998) takes the average clay content of those beds that have slipped past any point (as determined by the fault throw), and treats this as an estimate of upscaled fault-zone composition (Fig. 3). Where SGR is high (.40 –50%) the fault rock is assumed to be dominated by clay smears, where SGR is low (,15 –20%) the fault rock is likely to be disaggregation zone or cataclasites. There is some observational support for this assumption (Yielding 2002; van der Zee & Urai 2005), though there are important scaling issues that are further discussed below. Figure 4 shows how these fault seal attributes can be mapped onto a fault plane in a subsurface model. The example shows one fault surface, mapped in a 3D seismic survey from offshore Gulf of Mexico. Local well data provide information on the lithological sequence, which can be mapped onto each side of the fault (upthrown Vclay distribution shown at top left). The displacement distribution on the fault is shown at bottom left. Combining these two inputs in the SGR equation (Fig. 3) allows us to plot the variation of SGR over the entire mapped fault surface (Fig. 4, right). In the chosen colour scheme, orange and red indicate SGR .50% and therefore probably clay smeared, whereas green indicates SGR ,15% and therefore probably disaggregation zone or cataclasite depending on the burial history. Note the lateral and vertical variation in the SGR prediction, and the greater heterogeneity in the results where the fault displacement is low (lateral and upper tip regions).

Calibration: does it seal or not? All of the above algorithms need to be calibrated in some way because the resultant numbers are not, in themselves, a prediction of seal capacity. The simplest kind of calibration looks for a threshold


247

Fig. 3. The three main fault-seal algorithms. After Yielding et al. (1997), redrawn by Jolley et al. (2007).

number which separates sealing locations from nonsealing locations on the fault surface (or strictly, probably sealing from probably non-sealing). Figure 5 gives some examples of calibration of Shale Smear Factor (SSF) based on lab experiments and outcrop observations. The property that is measured here is the continuity of the smear from upthrown to downthrown side, as seen in a crosssection view. When SSF is small, all smears are continuous; when SSF is large, all smears ultimately become discontinuous. The assumption is that a

continuous smear can act as a seal whereas a discontinuous smear will not (and no other sealing fault rocks are present). From the shape of the curves, one might conclude that when SSF ,4–5 there is a high chance of a continuous smear (and therefore the presence of seal at the fault). Færseth (2006) suggests a critical SSF of 4 from a study of larger faults at outcrop. To apply this critical value to subsurface prediction, a further assumption must be made that along-strike variation would not change the cross-section statistics shown in Figure 5.

Fig. 4. Example of a subsurface fault-seal model. The fault surface was interpreted on 3D seismic reflection data (West Cameron area, Gulf of Mexico). Upper left diagram shows the sand-shale pattern mapped onto the upthrown side of the fault from well curves (Vshale). Lower left diagram shows the displacement pattern mapped on the fault surface (from horizon offsets). Right diagram shows calculated distribution of shale gouge ratio (SGR) (equation in Fig. 3).

248

G. YIELDING ET AL.

Fig. 5. Two calibrations of Shale Smear Factor (SSF). The curves show the observed probability of a smear being continuous from upthrown to downthrown side in cross-section view, at various different values of SSF (throw/thickness, Fig. 3). The blue curve summarizes deformation of laboratory samples (Takahashi 2003); the magenta curve summarizes outcrop observations by Childs et al. (2007).

To calibrate the attribute more directly, it is better to use calculations on mapped subsurface faults that are known from well data to be sealing. Figure 6 shows a calibration of Clay Smear Potential (CSP) using subsurface fault mapping in Shell’s Niger Delta fields (from Fulljames et al. 1997). At low CSP very few reservoir –reservoir juxtapositions are observed to seal, but the proportion of sealing juxtapositions progressively rises as CSP increases, up to a ‘saturation’ value after which the seal probability does not increase further. Unfortunately the calibration was published without numbers on the CSP axis. Jev et al. (1993) state that in another nearby field, CSP . 30 represents a sealing value. A similar subsurface calibration is shown in Figure 7 for SGR. The two curves, representing observations from the Columbus Basin and the Brent Province, imply that when SGR .c. 20% there is a high chance of a fault-zone seal. Both these examples are from sand-shale sequences where the smear-free parts of the fault zones are composed of non-sealing disaggregation-zones (as seen in well cores, e.g. Fisher & Knipe 2001). The correspondence between this SGR threshold of 20%, and the maximum clay content of disaggregation zones (20%) gives some credence to the assumption that the SGR value is representing the fault-zone composition. All of the calibrations shown in Figures 5–7 relate to geological environments where seal at the

Fig. 6. Subsurface calibration of Clay Smear Potential (CSP) published by Fulljames et al. (1997). The data for this calibration were gathered on 91 reservoirs along 10 faults in three different fields (Niger Delta). (No scale on horizontal axis in original figure.)

fault surface is developed by clay/shale smear or phyllosilicate-framework fault rock, and where clay-poor fault rock is disaggregation zone rather than cataclasite. The threshold value of the faultseal attribute can therefore be viewed as an on/off switch between non-sealing areas (disaggregation zones) and well-sealed areas (smears, PFFR). In many exploration contexts, this may be sufficient to test the viability of a fault-bound prospect. However, often a more quantitative result will be needed – specifically, how tall a hydrocarbon column might the fault seal hold back? Also, in cases where cataclasites or cemented fault rocks are present, the nonsmeared parts of a fault plane may also seal. The next section reviews calibrations for these situations.

Calibration: how much does the seal hold? Over the last two decades, two fundamentally different approaches have been advocated to predict the capacity of fault seals. These approaches have often been referred to as deterministic and empirical. † Deterministic approach. Values of capillary threshold pressure and clay content are measured in the lab for fault rock samples local to the prospect (e.g. from nearby cored wells). To assign these various measurements to the appropriate reservoir-reservoir juxtapositions, it is usually necessary to perform an SGR analysis of the fault and then (assuming that SGR equals the clay content of the fault-zone) the measured capillary threshold values can be interpolated onto the fault. † Empirical approach. SGR analysis is performed on known sealing faults, and compared with


Fig. 7. Two subsurface calibrations of Shale Gouge Ratio. Black line summarizes calibration data from 12 reservoirs along 7 faults in Mahogany field, Columbus Basin, offshore Trinidad (from Gibson & Bentham 2003). Red line summarizes data from 29 faults in 15 fields in the Brent Province (from Yielding 2002).

measured hydrocarbon column heights or pressure differences trapped at the fault. The derived SGR – column height relationship is then used predictively on adjacent prospects. Note that SGR is used as a proxy attribute to obtain seal capacity – this approach does not require that SGR corresponds to any real geological property. The deterministic approach is perhaps best exemplified by the work of Sperrevik et al. (2002). Measurements of permeability and Hg–air capillary threshold pressure were made on core samples of microfaults from nine fields in the northern North Sea and mid-Norway shelf. (Fault displacements were in the range millimetres to centimetres.) Results were analysed in terms of fault clay content, depth at time of faulting, and maximum burial depth – these are the three controlling factors described above in the section on Fault-zone processes. Best-fit polynomial equations were derived to express the variation in permeability and threshold pressure in terms of these three variables. Examples of subsets of the database are shown in Figure 8. Threshold pressures increase with faultzone clay content and with maximum burial depth, and at low clay contents they also increase with depth of burial during faulting (cataclasis effect). Assuming that subsurface fault rock clay content is given by SGR, and that the faulting/burial history is known, then the derived equations can be used for prediction of subsurface fault-seal capacity. Because the equations are a best-fit through a scatter of measured capillary pressures, they can be regarded as an average prediction of fault rock seal capacity for a given set of input variables.

249

Two significant issues can be raised with regard to the deterministic approach. Firstly, the measurements of fault rock properties are made on plug samples from microfaults with mm– cm displacements. Such small faults are relatively simple and do not have the complex internal structure of trapbounding faults with .10 m displacement (e.g. Childs et al. 1997; Foxford et al. 1998; Wibberley et al. 2008). Larger faults typically comprise multiple components (e.g. smears, clay-poor fault rock, and intact protolith), rather than uniformly mixed material. It is arguable whether SGR gives a representative estimate of fault-zone composition – within the upscaled ‘average’, there are always likely to be smear components that are relatively more sealing. We return to this point later. A second issue is that lab-measured capillary threshold pressures are obtained in the mercury-air system, whereas obviously the relevant system for fault-seal prediction is hydrocarbon– water. Hg– air values can be converted to hydrocarbon-water values if the interfacial tension for the hydrocarbonwater system is known for the subsurface PT conditions. This information is frequently unknown. Values are known to depend upon fluid type, temperature and pressure, as indicated in Figure 9. At shallow depths the variation with fluid type is considerable, though appears to reduce with increasing depth. Because the Hg– air interfacial tension is much higher than the hydrocarbon–water values shown in Figure 9, hydrocarbon–water threshold pressures will be smaller than measured Hg– air threshold pressures by a factor of 0.08– 0.16 depending on hydrocarbon type and PT conditions, giving a factor of 2 range in the final result. The empirical approach to calibration involves studying a known sealing fault and computing the pressure difference across it at each reservoir – reservoir juxtaposition. Comparison with a faultseal attribute (such as shale gouge ratio) at the same points on the fault can reveal whether there is any relationship between the attribute and the supported pressure difference, see Figure 10. This workflow was introduced by Yielding et al. (1997), and additional examples were compiled and summarized by Yielding (2002) and Bretan et al. (2003). Although a local calibration is always preferred for predictive purposes (to ensure cases with the same geological history), it is instructive to combine different datasets to reveal broader trends in the relationships. Figure 10 shows a global compilation of data from many basins (see figure caption for details), and indicates a general trend of increasing SGR value being able to support increasing across-fault pressure difference (AFPD). Lines represent ‘seal-failure envelopes’, that is, maximum across-fault pressure that can be supported at a specific SGR value, for particular maximum burial

250

G. YIELDING ET AL.

Fig. 8. Examples of laboratory measurements of Hg –Air threshold pressure of fault rock samples from the northern North Sea (Sperrevik et al. 2002). Diagram at left presents measurements on samples buried to 2500–3000 m, whereas that at right shows measurements on samples buried .3600 m. The curves are examples of a model equation fitted to .100 samples, which relates threshold pressure to variations in fault rock clay content, maximum burial depth Zmax, and depth of burial at time of faulting Zf.

depths. Note that the seal-failure envelope for shallow depths (,3 km, blue) exhibits no seal at SGR , 15–20%, corresponding to the simple seal/no-seal threshold discussed in the previous section. Because the envelopes are drawn along the upper edge of all observed data, they represent

Fig. 9. Some published estimates of the variation of hydrocarbon–water interfacial tension with respect to depth (pressure & temperature conditions). The methane & decane curves indicate experimentally-measured trends from Firoozabadi & Ramey (1988). The ‘oil’ values are from Nordga˚rd Bola˚s et al. (2005), constructed from empirical equations of Firoozabadi & Ramey (1988). Arrows show typical industry default values for oil–water (green) and gas–water (red) (d’Onfro, pers. comm., 2007).

a maximum seal capacity for any combination of input data – contrast with the equations based on fault rock sample measurements, which represent an average seal capacity at given conditions. Some of the original data included in the Figure 10 compilation involved across-fault aquifer changes (Bretan et al. 2003). These are not relevant as examples of static capillary trapping, which by definition is a physical process dependent on the presence of two immiscible phases. Figure 11 (left) is a revised compilation, concentrating on true buoyancy pressures measured in the hydrocarbon phase relative to the aquifer pressure at the fault. If aquifer pressures are the same across the fault, then this is a measure of the buoyancy pressure in the trap relative to the water in the fault rock (the sealing lithology). On the other hand, if aquifer pressures differ across the fault, then the plotted buoyancy pressure is the difference between the hydrocarbon phase and the higher-pressure aquifer (see discussion by Underschultz 2007). Cases where the across-fault pressure differences are measured between different aquifers (no hydrocarbon involved) have been omitted. Such data are telling us about the hydrodynamic behaviour of the fault rather than its static trapping capacity (e.g. Grauls et al. 2002; Harris et al. 2002). This is an important distinction: in hydrodynamic systems the pressure drop between aquifers at the fault will depend upon the fault rock permeability and thickness (not threshold pressure) and the overall flow rate. In Figure 11 (left) the dashed lines show a possible revision of the seal-failure envelopes for depths ,3 km (blue) and .3.5 km (green). Compared with


Fig. 10. The empirical approach to faul-seal calibration (from Bretan et al. 2003). The insets at left show how the across-fault pressure difference can be defined at reservoir-reservoir overlaps from data at adjacent wells. The plot at right shows a global compilation of across-fault pressure differences and their relationship to SGR at the same point on the fault surface. Data come from several tens of faults in nine different extensional basins. Clouds of small points correspond to entire reservoir juxtaposition areas. Large points (þ) correspond to ‘trap-critical’ locations that represent the highest pressure difference at a particular value of SGR on that fault. Data points are coloured by maximum burial depth, blue ,3 km, red 3 –3.5 km, green .3.5 km. The dashed lines indicate suggested log-linear fault-seal-failure envelopes for the different depth ranges. 251

252

G. YIELDING ET AL.

Fig. 11. A comparison of deterministic (lab) measurements and empirical calibration of seal capacity as a function of fault rock composition or shale gouge ratio. The cross-plot at left is subset of the subsurface data presented in Figure 10, filtered to retain only true measurements of hydrocarbon buoyancy pressure (Bretan & Yielding 2005; Underschultz 2007). Data points are coloured by maximum burial depth, blue ,3 km, red 3–3.5 km, green .3.5 km. The cross-plot at right is an example of Hg– air core-plug threshold pressures from Sperrevik et al. (2002). The vertical axes of the two plots are aligned assuming that hydrocarbon/water threshold pressures are one tenth of the Hg– air values. The curved lines are possible fault-seal failure envelopes, constructed on the plot of subsurface data (left) but also drawn on the core-plug data (right) – note the consistency of the two approaches.

Figure 10, some of higher pressure differences at high SGR are removed. The data now do not justify a continued increase in trapping potential

Fig. 12. Linear plot of the same data as that shown in the left-hand plot of Figure 11. Data points are coloured by maximum burial depth, blue ,3 km, red 3– 3.5 km, green .3.5 km. Note that linear seal-failure envelopes are consistent with the data distribution.

above about 40 –50% SGR. A possible interpretation is that sealing by clay smears becomes complete over the fault surface at this value, and addition of further clay smear does not enhance the capillary sealing capacity (in the same way that doubling the thickness of a top seal layer does not double its sealing capacity). Also shown in Figure 11 (at right) is an example of the deterministic calibration of Sperrevik et al. (2002); a factor of 0.1 aligns the vertical axes of the two plots, on the assumption that hydrocarbon–water threshold pressures are about one tenth of Hg–air values. The revised seal-failure envelopes from the subsurface plot (at left of figure) are also shown on this plot for reference. It is striking, and perhaps surprising given the earlier discussion, that the subsurface seal-failure lines are broadly consistent with the core-plug measurements. The core-plug measurements display a wide scatter (order-of-magnitude), and if this variability occurs on a subsurface fault then some of the lower threshold-pressure values might be expected to be critical in controlling a particular trap fill. However, the fact that both data types suggest similar trends for fault-zone threshold pressures gives some support to the use of SGR as a proxy for average fault-zone clay content – even though


large trap-bounding faults are expected to be more heterogeneous. This is an important result, and indicates that (for practical purposes) estimates of fault-seal capacity produced by the empirical and deterministic methodologies are not mutually exclusive but rather are alternative possibilities in the range of subsurface uncertainty. The revised empirical seal-failure envelopes shown on Figure 11 are curved, which suggests that a log-linear relationship is not the most appropriate. Figure 12 shows the same data as Figure 11, but plotted with linear axes instead of log-linear. This revised compilation is perhaps better described by linear fault-seal failure envelopes, more similar to those originally suggested in the initial application of this methodology (Yielding et al. 1997). Addition of further datasets in the future will doubtless continue to revise these bounding envelopes. The most promising future avenue for quantifying seal-prediction uncertainty appears to be a greater understanding of fault-zone heterogeneity, particularly at intermediate clay-contents. At high clay content (high SGR, e.g. .50%), both methodologies would interpret the fault rock as being dominated by clay smear, and therefore the capillary properties of clay smears should be representative of the fault surface. At low clay content (low SGR, ,20%), the fault rock is likely to be dominated by disaggregation zones or cataclasites, and again the properties of these rocks are likely to be appropriate for the fault surface. However, at intermediate SGR values (20 –50%) differing interpretations are possible: the deterministic approach would infer that the fault-zone is composed of phyllosilicate-framework fault rocks (PFFR) like those sampled in core-plugs of microfaults, whereas the empirical approach simply uses observed in-situ pressures without necessarily requiring a particular fault rock composition. Observations of large faults at outcrop would generally reveal a heterogeneous assemblage of clay-poor components and clay/shale smears (some possibly discontinuous) rather than a slab of PFFR. So, although the average or upscaled composition of the fault zone may be the same as a PFFR, PFFRs might be only a minor component: the assemblage of smears may provide the control on the faultzone properties. Interestingly, Childs et al. (2007) have produced stochastic models of disrupted impermeable shale smears and then examined the effective fault-zone permeabilities that result. When fault throws are much greater than the bed thicknesses, the effective permeability of the fault zone is equivalent to a simple log-linear relationship between SGR and permeability, mimicking the observed log-linear relationship between fault rock clay content and

253

permeability (Sperrevik et al. 2002). This result cannot be directly applied to threshold pressures, since trapping does not depend on average properties and one weak spot could destroy the seal. However, modelling of trapped column heights (as opposed to threshold pressure) relative to disrupted shale smear geometries suggests that an analogous relationship can arise, with taller columns being trapped as more disrupted smears are entrained in the fault zone (Yielding 2009) If confirmed, this would imply that the apparent consistency of the empirical and deterministic approaches at intermediate SGR values is largely coincidental, and improving the agreement by further work should not be expected.

Conclusions † Fault seal algorithms are required to characterize variation over a mapped subsurface fault. † Each algorithm needs to be calibrated against local data to be meaningful. † Calibration may be qualitative (on/off) or quantitative (pressure differences or column heights). † Published calibrations have involved fault-zone samples in the lab (‘deterministic’), or hydrocarbon traps in the subsurface (‘empirical’). † Surprisingly, lab & in-situ calibrations often agree to within a factor of 2–3. Comparison of the two approaches is perhaps one estimate of the uncertainty in any prediction, given the influence of fault zone heterogeneity on fluid trapping processes. Further outcrop studies of larger (trap-bounding scale) faults may represent the best way to quantify this uncertainty. Many thanks to reviewers Chris Wibberley and Pete D’Onfro for suggestions that improved an earlier version of this manuscript. Conrad Childs is thanked for providing a copy of his Shale Smear modelling paper. Discussions with Quentin Fisher are gratefully acknowledged.

References Allan, U. S. 1989. Model for hydrocarbon migration and entrapment within faulted structures. American Association of Petroleum Geologists Bulletin, 73, 803– 811. Aydin, A. & Eyal, Y. 2002. Anatomy of a normal fault with shale smear: implications for fault seal. American Association of Petroleum Geologists Bulletin, 86, 1367– 1381. Barton, C. A., Zoback, M. D. & Moos, D. 1995. Fluid flow along potentially active faults in crystalline rock. Geology, 23, 683– 686. Bouvier, J. D., Kaars-Sijpesteijn, C. H., Kluesner, D. F., Onyejekwe, C. C. & Van der Pal, R. C. 1989. Three-dimensional seismic interpretation and fault sealing investigations, nun river field, Nigeria.

254

G. YIELDING ET AL.

American Association of Petroleum Geologists Bulletin, 73, 1397– 1414. Bretan, P. & Yielding, G. 2005. Using buoyancy pressure profiles to assess uncertainty in fault seal calibration. In: Boult, P. & Kaldi, J. (eds) Evaluating Fault and Cap Rock Seals. American Association of Petroleum Geologists Hedberg Series, Tulsa, 2, 151– 162. Bretan, P., Yielding, G. & Jones, H. 2003. Using calibrated shale gouge ratio to estimate hydrocarbon column heights. American Association of Petroleum Geologists Bulletin, 87, 397 –413. Childs, C., Watterson, J. & Walsh, J. J. 1996. A model for the structure and development of fault zones. Journal of the Geological Society, London, 153, 337– 340. Childs, C., Watterson, J. & Walsh, J. J. 1997. Complexity in fault zone structure and implications for fault seal prediction. In: Møller-Pedersen, P. & Koestler, A. G. (eds) Hydrocarbon Seals: Importance for Exploration and Production, Norwegian Petroleum Society (NPF) Special Publication, Elsevier Science, Singapore, 7, 61– 72. Childs, C., Walsh, J. J. et al. 2007. Definition of a fault permeability predictor from outcrop studies of a faulted turbidite sequence, Taranaki, New Zealand. In: Jolley, S. J., Barr, D., Walsh, J. J. & Knipe, R. J. (eds) Structurally Complex Reservoirs. Geological Society, London, Special Publications, 292, 235– 258. Dee, S., Freeman, B., Yielding, G., Roberts, A. & Bretan, P. 2005. Best practice in structural geological analysis. First Break, 23, 49– 54. Doughty, 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. Eichhubl, P., D’Onfro, P. S., Aydin, A., Waters, J. & McCarty, 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. Færseth, R. B. 2006. Shale smear along large faults: continuity of smear and the fault seal capacity. Journal of Geological Society, London, 163, 741– 752. Firoozabadi, A. & Ramey, H. J. 1988. Surface tension of a water hydrocarbon system at reservoir conditions. Journal of Canadian Petroleum Technology, 27, 3. Fisher, Q. J. & Knipe, R. J. 1998. Fault sealing processes in siliciclastic sediments. In: Jones, G., Fisher, Q. J. & Knipe, R. J. (eds) Faulting, Fault Sealing and Fluid Flow in Hydrocarbon Reservoirs. Geological Society, London, Special Publications, 147, 117–134. Fisher, Q. J. & Knipe, R. J. 2001. The permeability of faults within siliciclastic petroleum reservoirs of the North Sea and Norwegian Continental Shelf. Marine & Petroleum Geology, 18, 1063–1081. Fisher, Q. J., Casey, N., Harris, S. D. & Knipe, R. J. 2003. Fluid-flow properties of faults in sandstone: the importance of temperature history. Geology, 31, 965– 968. Fossen, H. & Gabrielsen, R. H. 2005. Strukturgeologi. Fagbokforlaget, Bergen, Norway, 1– 369.

Fossen, H., Schultz, R. A., Shipton, Z. K. & Mair, K. 2007. Deformation bands in sandstone: a review. Journal of Geological Society, London, 164, 755– 769. Foxford, K. A., Walsh, J. J., Watterson, J., Garden, I. R., Guscott, S. C. & Burley, S. D. 1998. Structure and content of the Moab Fault Zone, Utah, USA, and its implications for fault seal prediction. In: Jones, G., Fisher, Q. J. & Knipe, R. J. (eds) Faulting, Fault Sealing and Fluid Flow in Hydrocarbon Reservoirs. Geological Society, London, Special Publications, 147, 87– 103. Freeman, B., Yielding, G., Needham, D. T. & Badley, M. E. 1998. Fault seal prediction: the gouge ratio method. In: Coward, M. P., Daltaban, T. S. & Johnson, H. (eds) Structural Geology in Reservoir Characterization. Geological Society, London, Special Publications, 127, 19– 25. Fristad, T., Groth, A., Yielding, G. & Freeman, B. 1997. Quantitative fault seal prediction: a case study from Oseberg Syd. In: Møller-Pedersen, P. & Koestler, A. G. (eds) Hydrocarbon Seals: Importance for Exploration And Production. Norwegian Petroleum Society (NPF) Special Publication, Elsevier Science, Singapore, 7, 107– 124. Fulljames, J. R., Zijerveld, L. J. J. & Franssen, R. C. M. W. 1997. Fault seal processes: systematic analyses of fault seals over geological and production time scales. In: Møller-Pedersen, P. & Koestler, A. G. (eds) Hydrocarbon Seals: Importance for Exploration and Production. Norwegian Petroleum Society (NPF) Special Publication, Elsevier Science, Singapore, 7, 51–59. Gibson, R. G. 1998. Physical character and fluid-flow properties of sandstone-derived fault gouge. In: Coward, M. P., Daltaban, T. S. & Johnson, H. (eds) Structural Geology in Reservoir Characterization. Geological Society, London, Special Publications, 127, 83–97. Gibson, R. G. & Bentham, P. A. 2003. Use of fault-seal analysis in understanding petroleum migration in a complexly faulted anticlinal trap, Columbus Basin, offshore Trinidad. American Association of Petroleum Geologists Bulletin, 87, 465–478. Grauls, D., Pascaud, F. & Rives, T. 2002. Quantitative fault seal assessment in hydrocarbon-compartmentalised structures using fluid pressure data. In: Koestler, A. G. & Hunsdale, R. (eds) Hydrocarbon Seal Quantification. Norwegian Petroleum Society (NPF) Special Publication, Amsterdam, 11, 141– 156. Harris, D., Yielding, G., Levine, P., Maxwell, G., Rose, P. T. & Nell, P. 2002. Using Shale Gouge Ratio (SGR) to model faults as transmissibility barriers in reservoirs: an example from the Strathspey field, North Sea. Petroleum Geoscience, 8, 167– 176. Jev, B. I., Kaars-Sijpersteijn, C. H., Peters, M. P. A. M., Watts, N. L. & Wilkie, J. T. 1993. Akaso Field, Nigeria: use of integrate 3D seismic, fault-slicing, clay smearing and RFT pressure data on fault trapping and dynamic leakage. American Association of Petroleum Geologists Bulletin, 77, 1389– 1404. Jolley, S. J., Dijk, H., Lamens, J. H., Fisher, Q. J., Manzocchi, T., Eikmans, H. & Huang, Y. 2007. Faulting and fault sealing in production simulation models: Brent province, northern North Sea. Petroleum Geoscience, 13, 321 –340.

FAULT SEAL CALIBRATION Jones, R. M. & Hillis, R. R. 2003. An integrated, quantitative approach to assessing fault-seal risk. American Association of Petroleum Geologists Bulletin, 87, 507–524. Lehner, F. K. & Pilaar, W. F. 1997. The emplacement of clay smears in synsedimentary normal faults: inferences from field observations near Frechen, Germany. In: Møller-Pedersen, P. & Koestler, A. G. (eds) Hydrocarbon Seals: Importance for Exploration and Production. Norwegian Petroleum Society (NPF) Special Publication, Elsevier Science, Singapore, 7, 39– 50. Lindsay, N. G., Murphy, F. C., Walsh, J. J. & Watterson, J. 1993. Outcrop studies of shale smear on fault surfaces. In: Flint, S. & Bryant, A. D. (eds) The Geological Modelling of Hydrocarbon Reservoirs and Outcrop. Special Publication International Association of Sedimentologists, Blackwell, Oxford, 15, 113 –123. Nordga˚rd Bola˚s, H. M., Hermanrud, C. & Teige, G. M. G. 2005. Seal capacity estimation from subsurface pore pressures. Basin Research, 17, 583–599. OXFORD PAPERBACK DICTIONARY. 1988. 3rd edn. Oxford University Press, Oxford. Sperrevik, S., Gillespie, P. A., Fisher, Q. J., Halvorsen, T. & Knipe, R. J. 2002. Empirical estimation of fault rock properties. In: Koestler, A. G. & Hunsdale, R. (eds) Hydrocarbon Seal Quantification. Norwegian Petroleum Society (NPF) Special Publication, Amsterdam, 11, 109–125. Takahashi, M. 2003. Permeability change during experimental fault smearing. Journal of Geophysical Research, 108, B5. Tearpock, D. J. & Bishke, R. E. 2003. Applied Subsurface Geological Mapping. 2nd edn. Prentice Hall PTR, New Jersey, 1–822.

255

Underschultz, J. 2007. Hydrodynamics and membrane seal capacity. Geofluids, 7, 148–158. Van der Zee, W. & Urai, J. L. 2005. Processes of fault evolution in a siliciclastic sequence: a case study from Miri, Sarawak, Malaysia. Journal of Structural Geology, 27, 2281– 2300. Weber, K. J., Mandl, G., Pilaar, W. F., Lehner, F. & Precious, R. G. 1978. The role of faults in hydrocarbon migration and trapping in Nigerian growth fault structures. 10th Annual Offshore Technology Conference Proceedings, 4, 2643– 2653. Wibberley, C. A. J., Yielding, G. & Di Toro, G. 2008. Recent advances in the understanding of fault zone internal structure: a review. In: Wibberley, C. A. J., Kurtz, W., Imber, J., Holdsworth, R. E. & Collettini, C. (eds) The Internal Structure of Fault Zones: Implications for Mechanical and Fluid-Flow Properties. Geological Society, London, Special Publications, 299, 5– 33. Yielding, G. 2002. Shale Gouge Ratio – calibration by geohistory. In: Koestler, A. G. & Hunsdale, R. (eds) Hydrocarbon Seal Quantification. Norwegian Petroleum Society (NPF) Special Publications, Amsterdam, 11, 1– 15. Yielding, G. 2009. Using Probabilistic Shale Smear Factor to relate SGR predictions of column height to fault-zone heterogeneity. In: Ligtenberg, H. & Knipe, R. (eds) Fault and Top Seals: from Pore to Basin Scale. European Association of Geoscientists & Engineers (EAGE), 21– 24 September 2009, Montpellier. EAGE, Houten. Paper E05, 93–95. Yielding, G., Freeman, B. & Needham, T. 1997. Quantitative fault seal prediction. American Association of Petroleum Geologists Bulletin, 81, 897– 917.

Cross-fault sealing, baffling and fluid flow in 3D geological models: tools for analysis, visualization and interpretation S. R. FREEMAN, S. D. HARRIS* & R. J. KNIPE Rock Deformation Research Limited, School of Earth Sciences, University of Leeds, Leeds, LS2 9JT, UK *Corresponding author (e-mail: [email protected]) Abstract: The effective computation and visualization of cross-fault sealing or flow, and parameters that infer or control that distribution, is a key step in the production of more reliable exploration and production simulation models. A better understanding of the impact of faultrelated flow or baffling through visualization can lead to the development of more robust and useful geological models that better define the likely range in flow behaviour. A range of visualization tools are available, from the traditional fault plane juxtaposition map to the vector visualization of cross-fault fluid flux. Each tool has its applications and limitations. In this contribution we discuss the application of these different techniques and highlight situations where these are particularly successful. A number of existing visualization approaches will be reviewed and improvements to those techniques are shown. A series of existing property visualization techniques are critiqued, such as the imaging of shale gouge ratio (SGR) and fault transmissibility multipliers (TMs) on the fault faces, both of which are limited in their ability to act as a proxy for cross-fault fluid flux in many circumstances. Fault rock property visualizations, such as hydraulic resistance and fault transmissibility, are presented. More direct and hence more powerful indications of probable cross-fault fluid flux are also described, such as the effective cross-fault transmissibility (ECFT) and the effective cross-fault permeability (ECFP). These static proxies for cross-fault fluid flux are compared against back-calculated and visualized cross-fault fluid flux values derived from either streamline or full flow simulation data. The ECFT is shown to provide a useful and rapid indication of likely fluid flux from the static model; however, the direct imaging of cross-fault fluid flux derived from simulation results allows for a far better understanding of how the faults have contributed to the reservoir flow simulation result. Visualizations of the fault- and flow-related properties: (a) on the fault face; (b) in the grid cells adjacent to the fault face; (c) as vectors; or (d) as fault-wide summations, all provide useful insights for different parts of the reservoir evaluation workflow. This contribution highlights a series of new and efficient techniques to image and hence improve the understanding and modelling of fault sealing in both exploration and production settings.

In this paper we review workflows used for visualizing and analysing properties that control cross-fault fluid flow; these range from traditional fault plane maps to more complex property distribution calculations and the back-analysis of flow simulation data to inform on fault sealing and baffling. During the initial exploration stage rapid assessments of fault juxtapositions (Allan 1989; Knipe 1997) are useful in assessing the likelihood of developing a faulted hydrocarbon trap. Fault plane maps, also known as Allan maps (Fig. 1), can highlight the key juxtapositions present (Allan 1989; Marchal et al. 2003). Often the creation of these maps is a time consuming process, and they can be complex and hence difficult to interpret in all except the simplest structural –stratigraphic configurations (compare Fig. 1b, c). We present different visualization styles and show techniques that enable a more rapid generation and interrogation of the data thus enhance the understanding of this data.

An improved understanding of the subtleties in fault sealing has coincided with the increased complexity of stratigraphic layering in geological models and the increased computing capabilities available to model these scenarios. Previously, stratigraphic layers have been modelled as relatively thick, laterally continuous zones populated with relatively uniform properties (e.g. Fig. 1b). Recently, a larger number of layers, typically tens to hundreds of grid layers, have been used to attempt to model rapidly varying host rock properties within each of the laterally and potentially discontinuous layers (e.g. Fig. 1c). These improvements in modelling resolution (continuous fault property variations, laterally varying stratigraphy and high numbers of modelled layers) have lessened the applicability of simple Allan line maps for predicting the sealing and/or cross-fault fluid flow properties. More refined tools are now required for calculating, visualizing and incorporating fault

From: Jolley, S. J., Fisher, Q. J., Ainsworth, R. B., Vrolijk, P. J. & Delisle, S. J. (eds) Reservoir Compartmentalization. Geological Society, London, Special Publications, 347, 257–282. DOI: 10.1144/SP347.15 0305-8719/10/$15.00 # The Geological Society of London 2010.

258

S. R. FREEMAN ET AL.

Fig. 1. (a) A 3D view of a simple fault-bound prospect with three internal zones. (b) The fault plane map along the main prospect-bounding fault (the solid lines show the upthrown markers and the dashed lines show the downthrown side). Note that at the edges of the data it is straightforward to track the juxtapositions of zones across the fault, but in the centre it becomes complicated even with this simple structural–stratigraphic scenario. (c) The fault plane map showing the detailed layer-cake stratigraphy for the model shown in (b). Determining juxtapositions from this data is challenging and likely to lead to inaccuracies. For numerous laterally discontinuous layers the task of determining juxtapositions becomes even more difficult.

rock flow properties into geological or reservoir models. It is important to note that although these stratigraphic models are being generated at an ever high resolution, a significant uncertainty in that stratigraphic population combined with model geometric and property uncertainties will lead to a compounded uncertainty in the resulting fault seal prediction. In production situations faults have historically been modelled as either sealing or open to

cross-fault fluid flow in reservoir simulations by using transmissibility multipliers (TMs) of 0 or 1, respectively (see Manzocchi et al. 1999). The transmissibility multiplier is the ratio of the faulted to unfaulted cross-fault transmissibilities between adjacent grid blocks across the fault (Knai & Knipe 1998). The direct measurement of fault rocks and inferred fault rock permeabilities from high-resolution well tests around faults indicate that faults are better modelled as membrane

VISUALIZING AND INTERPRETING CROSS-FAULT FLUID FLOW

baffles rather than either fully open or fully retarding to cross-fault fluid flow (e.g. Fisher & Knipe 1998; Sperrevik et al. 2002; Jolley et al. 2007; Manzocchi et al. 2008). The incorporation of these baffles into flow simulation models is usually achieved via the use of TMs (e.g. Manzocchi et al. 1999, 2002; Al-Busafi et al. 2005; Jolley et al. 2007; Zijlstra et al. 2007). These TM values are applied as modifiers to the transmissibility connection data in the reservoir simulation grid such that the permeability of the fault rock and its thickness are taken into account when computing fluid flow across the faults, and they are central to the inclusion of fault rock properties into flow simulators (e.g. Knai & Knipe 1998; Manzocchi et al. 1999, 2002). As such, the TM is the fault property most often visualized within reservoir modelling packages, but, as will be shown later, the visualization of this parameter is generally of little use for understanding and interpreting the behaviour of faults within the simulator. The TM is an abstracted mathematical tool used to conveniently modify the behaviour of the simulator; it is not a physical property, but is defined in terms of physical variables in a form that has no predictive benefits. For example, a low TM value will generally be found at locations of thicker fault rock (large displacement) and a fault permeability that is lower than the permeability of the adjacent grid cells. A high TM value will generally correspond to a thinner fault rock and a fault permeability that is comparable to the permeability of the adjacent grid cells. The dependency of the TM value on the host grid cell size, which will vary significantly through the model, further complicates these relationships. Both the high and low TM values can thus correspond to areas of high or low cross-fault fluid flow. Therefore, it is clearly more useful to visualize and interpret true physical properties that inform on fluid flow. Several steps are typically required to predict fault transmissibility variations that occur along fault zones from the host stratigraphic model (the reservoir model must contain spatial predictions of the host clay, permeability and porosity distributions) and the fault geometry. Under the assumption that the correct geometric model has been produced, one such workflow would be: (a) prediction of the host clay distribution through the stratigraphy against the fault (from well logs, e.g. gamma ray); (b) prediction of the clay distribution in the fault zone from the combined influences of the fault displacement and the lithological stacking sequence (e.g. using shale gouge ratio (SGR) and/ or clay smears); (c) determination of the relationship between fault rock clay and fault rock permeability; (d) interpolation of the fault rock permeability across the fault; (e) prediction of the fault rock thickness (typically from the fault displacement);

259

and (f ) computation of the fault transmissibility multiplier (see Manzocchi et al. 1999). Even for this example single workflow there is currently no consensus on how to determine many of the individual relationships (e.g. Manzocchi et al. 1999, 2008; Sperrevik et al. 2002; Jolley et al. 2007). The majority of published datasets suggest a significant scatter in the nature of the critical parameters (Hull 1988; Childs et al. 1997; Knipe et al. 1997, 1998; Fisher & Knipe 2002; Sperrevik et al. 2002; Jolley et al. 2007; Freeman et al. 2008; Manzocchi et al. 2008). The data also indicate that a range of other factors can influence those relationships; these include burial history (e.g. Fisher & Knipe 2002; Sperrevik et al. 2002; Freeman et al. 2008) and the timing of hydrocarbon migration. With this array of valid but potentially different relationships, it is important to be able to differentiate the most appropriate ones to apply in a given situation, or alternatively to assess what effect selecting different relationships will have on the predicted flow behaviour. Developing a set of visualization tools to rapidly evaluate the effects of the different relationships chosen should lead to more robust interpretations (and simulation results). The relative applicability of the different fault property prediction techniques (e.g. the fault clay prediction, the fault clay to permeability relationship and the prediction of the fault rock thickness variation) is dependent on the reservoir stratigraphy and deformation history (as described in the above publications for each of these techniques). The effective screening of these datasets through visualization and analysis should help to target critical scenarios for simulation and thus more efficiently define the likely range in cross-fault fluid flow. In this contribution we present a series of different techniques that aid in the computation, visualization and interpretation of fault property and cross-fault fluid flow data.

Visualization tools In this section we present a series of visualization tools that begin with relatively simple juxtaposition problems; these tools are more appropriate to primary prospect analysis and/or initial risking situations. More complex compound property visualization tools are then considered, before a final interrogation of flow simulation data within and around fault zones is used to provide a series of new and enhanced visualizations of cross-fault fluid flow. As the degree of complexity increases through the various stages, so does the amount of data required to be able to perform those visualization tasks. Initially, only a seismic marker and a simple concept of the stratigraphy is required. For

260


the later cases an understanding of certain specific fault rock properties is needed, and ultimately an understanding of the fluid types, pressure variations and production strategies is required. The development of these new techniques has been driven by a need to better understand the role and impact of faults within exploration prospects and production simulation models.

Visualizing fault stratigraphic juxtapositions Often, during the initial stages of exploration, only an interpretation of the top reservoir seismic marker exists. The stratigraphy at the target structure or fault is often only poorly constrained by distant wells. At this early stage a rapid means of screening faulted traps is critical. Given the large uncertainty in the stratigraphic architecture, the required technique must allow for rapid evaluations of the impact of varying stratigraphies. Historically, the construction of a fault plane diagram from such datasets was time consuming and only a limited number of visualizations of the different potential scenarios were generated. Figure 2 shows an example of an alternative ‘quick-look’ fault juxtaposition mapping method. The 3D geometry of the top reservoir surface in Figure 2a is used to develop the associated fault juxtaposition map in Figure 2b. In order to generate the fault map at this early stage of interpretation, several assumptions must be made. The primary assumption is that the interpreter can define a polyline on the seismic marker surface that provides an indication of the fault –horizon intersection lines. However, these interpreter polylines do not often coincide with the ‘real’ fault-horizon intersection lines, since seismic data deteriorates in proximity to the fault. Thus interpreter polylines are a proxy for hanging wall and footwall cut-off lines, which are then combined with a user-defined stratigraphy to generate the fault plane map. This tool can utilize multiple seismic markers and incorporate independent layer thickness variations in the hanging wall and footwall stratigraphies. The throw can also be manipulated to account for a degree of seismic, structural and stratigraphic uncertainty. This tool therefore facilitates a rapid assessment of the distribution of different stratigraphic juxtaposition styles along the fault of interest (cf. Fig. 2c–e). This helps in early structural screening to determine the importance of faults within the prospect. However, if the fault juxtapositions are deemed to be critical then the analysis should become more sophisticated, using the greater modelling accuracy available within a 3D model. An alternative approach is to use stochastic stratigraphic population and fault offsets (see James et al. 2004).

Traditionally, juxtapositions are visualized in commercial 3D geological modelling packages by projecting multiple seismic markers onto modelled fault surfaces to define horizon–fault intersection lines (i.e. fault-cut line or juxtaposition maps; e.g. Walsh & Watterson 1987; Allan 1989; Childs et al. 1997). When only a few stratigraphic layers are present and those layers are formed from relatively uniform properties, this is a powerful technique for identifying overlapping reservoir areas or non-overlapping areas on the fault surface (see Figs 1b, 2b & 3a). These types of projection become increasingly complex to interpret and more unreliable where there are complex displacement variations and where an increasing number of layers and/or property variations occur. This effect is compounded where uncertainties surround either throw or property distributions. In order to produce a more interpretable fault plane map the areas of the fault face can be filled with unique colours that relate to the specific juxtaposition type developed. An example is shown in Figure 3b below the unfilled version (Fig. 3a), and in this colour scheme ‘hotter’ colours have generally been applied to areas of higher fault throw (i.e. Zone 1 juxtaposed against Ness 1). This style of simple juxtaposition map can be extended for use in more complex geological situations. The isolation of specific zones (e.g. Fig. 3c) across the fault further enhances the ability to interrogate the data, and applying such filters can rapidly delineate critical juxtapositions and limit the potential for mis-identification. Although this approach can allow gross unit juxtapositions to be interpreted rapidly, there are a number of limitations. The main one is the difficulty of back-relating the juxtaposition areas to the specific stratigraphy that is juxtaposed. A specific zone may be juxtaposed against another specific zone, but unless the stratigraphy within those zones is of a particular type (e.g. high-permeability sands) the knowledge of zonal juxtapositions may be of little relevance. The generalized fault plane windows may contain a wide range of juxtaposition types (e.g. sands against shales, or sands against sands), so that a large number of these very generalized zonal area juxtapositions are not important (particularly for highly heterogeneous stratigraphy). In such circumstances it is only necessary to visualize a few key stratigraphic juxtapositions, and one approach to address this problem is to filter the fault plane map by stratigraphic type. The resulting fault plane map will then be far more useful (see Fig. 4). Figure 4a, b show examples of the stratigraphy against the fault for the upthrown and downthrown blocks, respectively (yellow is sand, green is impure sand and brown is shale). The purple areas on the fault face in Figure 4c represent the key sand on sand juxtapositions along the fault.


261

Fig. 2. (a) Top reservoir seismic marker and (b) the corresponding ‘quick-look’ juxtaposition map (green zone/dotted lines for downthrown side; brown zone/solid lines for upthrown side). Diagrams (c), (d) and (e) show the low, mid and high cases, respectively, for sand thicknesses for the target reservoir modified ‘on-the-fly’ within the software. Note that the ability to dynamically modify the stratigraphy in (c)– (e) allows the range in geological scenarios to be rapidly evaluated. Note also that in each diagram (b)– (e) the layers have been shown as semi-transparent, so that sand on sand windows can be easily identified.

The ability to view specific property overlaps is a powerful additional tool for isolating critical connections. Figure 4d shows the same sand on sand windows as in Figure 4c, but this time by isolating the cells in the upthrown and downthrown side of the fault that have led to those juxtapositions. Both of the plot styles in Figure 4c, d are useful, but the visualization of the juxtaposition sand on sand windows via the cells in the grid adjacent to the faults enables those windows to be more easily related back to the stratigraphic layers that created that juxtaposition. In addition, the view in Figure 4d allows the form of the stratigraphy in the immediate area surrounding the juxtaposition window to also be evaluated for its geological validity. In such a view, the juxtaposition windows developed by surfaces that have been poorly interpolated into faults become apparent, and such inaccuracies can then be corrected and are therefore less likely to be carried forward into either the final geological model, subsequent fault seal assessment or the simulation model. When dealing with stratigraphies that contain significant internal facies variations (e.g. fluvio– deltaic systems, turbidites or mass transport complexes), the applicability of fault-cut line maps is limited. The approach described above of defining specific single property juxtapositions is useful and can be readily extended to allow the simultaneous filtering of numerous parameters, potentially providing greater insight into cross-fault connectivity. Figure 5 shows the juxtaposition of

only a specific facies in one zone against a specific facies in another zone across the fault. In general, the window geometry will be defined by certain sets of parameter criteria and the colour fill of the windows then displays a different property. In this case, the property mapped onto these windows is the harmonic average host permeability of the rocks juxtaposed across the fault, an example of one parameter that controls the fluid flow across the fault zone. This type of multiple parameter property visualization and filtering can be extremely useful for determining the critical parameters that influence cross-fault fluid flow. By combining the visualization of stratigraphic juxtapositions via stratigraphic fault-cut lines, fault connection areas and grid cells adjacent to the fault, we are now able to achieve a much clearer definition of the critical windows and the relationship of those windows to the general stratigraphic architecture, and also to understand the validity of the windows. This provides a robust and efficient interpretation process, in which fault plane maps or 3D juxtaposition zones are used to understand the likely cross-fault flow behaviour of a prospect or field. For these techniques to add the maximum value they need to be rapid to apply, and easy to understand and interpret.

Membrane seal visualization Current visualization tools. It is becoming more common to both compute and display a variety

262


Fig. 3. (a) The horizon–fault intersection lines along a fault (solid lines for upthrown block; dotted lines for downthrown block; two other crossing faults are also partially shown). (b) A colour-filled 3D fault juxtaposition map showing the different reservoir zone overlaps. Note the increased ease of understanding for the colour-filled image (b) v. the traditional line drawing (a). This allows the key areas to be identified more rapidly and with greater confidence. (c) A filtered fault map showing the juxtaposition of the Ness 2 in the footwall against the Tarbert 3 in the hanging wall.

of different predicted fault rock properties – principally in order to estimate the flow properties of faults for flow simulation. Most geological modelling packages now permit the computation of the shale gouge ratio (SGR; Yielding et al. 1997), and this is used either as a direct proxy for sealing (e.g. Yielding 2002) or more commonly as a step toward defining the likely cross-fault permeability, and ultimately the fault TMs for input to flow simulation (e.g. Manzocchi et al. 1999, 2008; Sperrevik et al. 2002; Jolley et al. 2007; Freeman et al. 2008). Fault permeabilities and TMs along the fault surface can then be viewed as values for each cross-fault grid connection (e.g. Dee et al. 2005). Figure 6 shows an example of the clay content

distribution predicted using the SGR algorithm and the corresponding TM data calculated for the cross-fault grid connections, displayed on the fault surfaces. At present, most ‘general’ geological reservoir modelling packages only allow the two fault property parameters of SGR and the related TM to be computed and imaged (although some specialist packages allow more complex algorithm choices, and the export of other fault seal parameters to flow simulation models. There are two major flaws in this approach. Firstly, the implementation of only a small number of specific algorithms is likely to fail to capture the probable range in fault rock properties. The literature documents a number of other widely


263

Fig. 4. Stratigraphy of (a) the upthrown block and (b) the downthrown block against the fault (sand in yellow; impure sand in green; shale in brown) (this and the other fault views are shown with the horizon– fault intersection lines). (c) The sand on sand windows shown on the fault face. (d) A 3D view of the sand on sand windows shown in the cells adjacent to the fault (also shown is one of the horizons in the grid for perspective purposes). Displaying the cells adjacent to the fault provides both a clearer understanding of those units that led to the development of the fault window and a useful quality check that geologically valid architectures have generated the windows present.

accepted techniques (e.g. Lindsay et al. 1993; Lehner & Pilaar 1997) and the need to select data and algorithms that are suited to the specific geology of a specific reservoir and the fluid phases that it contains (e.g. Fisher & Jolley 2007). Secondly, and more importantly for visualization and interpretation purposes, the SGR and TM data in isolation are not directly related to the likely cross-fault fluid flow. Both limitations will be addressed in more detail in the following section and potential solutions to these problems will be suggested. Effective SGR, side-specific SGR, and a comparison of clay mixing and smearing algorithms. The SGR algorithm is a linear mixing model that estimates the fault zone clay distribution that arises as a

result of the mixing of clays from the host stratigraphy within the fault zone (Yielding et al. 1997; Yielding 2002). The originally defined prediction of the SGR value at any point on a fault is by a uniform (weighted by unit thickness) average of the clay contents in the stratigraphies/lithologies that have moved past that point. At any point P on the fault where the displacement is dP, the SGR can therefore be defined as SGR ¼

1 X Vi Dzi dP Slipped interval

where the summation is over all units i with thicknesses Dzi and clay content Vi that have slipped

264


Fig. 5. Fault plane property map showing the result of multiple filtered geometry criteria, and displaying a separate property on the resulting fault windows. The fault plane map shows the juxtaposition of the silt and sand facies of the Ness against the silt and sand facies of the Tarbert. The property shown on the fault windows is the harmonic average of the juxtaposed host permeability.

past that point P (note that here the fault displacement and the thicknesses of the units are measured in a direction down-dip along the fault surface). Two major shortcomings are present when applying the SGR algorithm. The first issue is the assumption that perfect mixing occurs both across and vertically

within the fault zone to produce the fault rock. This assumption is not consistent with outcrop observations (e.g. Shipton et al. 2002; Fredman et al. 2007). The effective shale gouge ratio algorithm (ESGR) applies a further weighting function to the

Fig. 6. (a) The fault clay distribution map defined by the shale gouge ratio (SGR) and (b) the fault transmissibility multiplier (TM) data, shown on the fault faces. Also shown are the horizon–fault intersection lines. This data forms some of the primary input (fault clay content) and output (fault transmissibility multiplier) from fault seal analysis.


(thickness-weighted) averaging and assumes that a greater contribution of the clays in the fault rock is derived from the units closer to the point of interest (see Fig. 7a for a comparison of the SGR and ESGR clay mixing algorithms). Extending the definition of the SGR above, at any point P on the fault we define the ESGR as

ESGR ¼

X Slipped interval

, wi Vi Dzi

X

wi Dzi

Slipped interval

where wi is the weighting factor applied to each unit i and the denominator becomes the displacement at P if each weight is set to 1. The weighting factor wi at a particular location through the slipped interval is defined from a weighting function that depends on the relative distance through that slipped interval from P to the most distal unit. The general form of the weighting function must be specified and the sensitivity of its precise form to the fault seal prediction must be investigated, but the function should generally provide an increased weighting for the most proximal units. For example, the weighting function may have a Gaussian form with maximum value at the point of interest P and that approaches the tail at the most distal point. ESGR overcomes the assumption of perfect vertical and across-fault mixing in the fault zone. A comparison of the difference in clay content prediction between the SGR and ESGR algorithms is shown in Figure 8a, b, respectively, and the stratigraphic and structure model used to generate these plots is that shown in Figure 4. The difference in potential clay content prediction is also shown in Figure 8c. A c. 10% difference in the predicted fault clay content typically implies a predicted fault permeability variation of approximately an order of magnitude (see Jolley et al. 2007). Note that two similar clay prediction algorithms can, in certain stratigraphic and structural situations, imply a considerably different fault clay distribution and hence fault permeability or sealing capacity. The second issue with the SGR algorithm is that it is often computed from the clay content from the hanging wall stratigraphy alone (according to the original SGR definition). Thus, in situations with growth faulting and syn-tectonic sedimentation, the calculation of SGR in this way will not provide a true indication of the average clay content of the host stratigraphy that has moved past that point on the fault because the algorithm’s fundamental assumptions are negated. For the implementation of SGR or ESGR in syn-tectonic sedimentation systems the clay mixing algorithm must be computed independently for both the hanging wall and footwall. To avoid the assumption of perfect

265

lateral mixing through the fault zone these results can then either be retained separately or combined via different functions (i.e. averaging or minimum/ maximum values). The differences between the fault clay prediction algorithms are significant in the example shown in Figure 8. This is due to variations in the host clay content in the stratigraphy. The differences between the fault clay predictions tend to be focused close to stratigraphic interfaces, so this has particular impact where the reservoir units are relatively thin in comparison to any surrounding or inter-collated shales. The range in fault clay prediction between the different algorithms along the reservoir sands is up to c. 20%. Using published fault clay to fault permeability relationships (e.g. Jolley et al. 2007), this increase or decrease in fault clay content equates to a permeability change of in excess of one order of magnitude within the fault rock. For a fault rock of 30 cm thickness (equivalent to a fault displacement of 30 m with a 1:100 fault thickness to displacement ratio) and a host permeability of 1000 mD, a fault permeability reduction from 0.01 to 0.001 mD is equivalent to a transmissibility reduction of around an order of magnitude across a 1000 m section of reservoir (assuming layer-bound flow). The difference between the SGR and ESGR methods for estimating the fault clay content can therefore produce an order of magnitude decrease in the predicted or modelled cross-fault fluid flux, for a constant phase pressure differential. With such large potential implications, it is clearly important to implement a variety of techniques in order to quantify the likely range in cross-fault fluid flux and the resulting impact on the reservoir flow simulation and ultimately the hydrocarbon resource model. Both the SGR and ESGR fault clay prediction algorithms are clay mixing models. A second set of processes are generally agreed to occur in fault zones. These are ductile smears in which shale (clay) beds and sands are sheared and smeared along fault planes (e.g. Aydin & Eyal 2002; Kristensen et al. 2008). Algorithms that describe the redistribution of clay within fault zones due to these processes include the shale smear factor (SSF; Lindsay et al. 1993), clay smear potential (CSP; Lehner & Pilaar 1997) and probabilistic shale smear factor (PSSF; Childs et al. 2007). These different clay smearing algorithms are compared in Figure 7b and defined as follows. The shale smear factor

SSF ¼

Dz t

defines the critical offset Dz that can be achieved before the shale (thickness t) becomes discontinuous.

266


Fig. 7. (a) Conceptual diagrams of the SGR and ESGR clay mixing models. The SGR algorithm estimates the fault zone clay distribution by assuming a uniform mixing of clays from the host stratigraphy into the fault zone. The ESGR algorithm utilizes a weighting function in the clay averaging and assumes that a greater contribution of the clays in the fault zone is derived from the units closer to the point of interest. (b) Conceptual diagrams of the various clay smearing models, namely the shale smear factor (SSF), the clay smear potential (CSP) and the probabilistic shale smear factor (PSSF).


267

Fig. 8. Fault clay mixing predictions using (a) the SGR algorithm and (b) the averaged ESGR algorithm, and (c) a comparison between the fault clay prediction derived from the SGR and ESGR algorithms (SGR minus averaged ESGR). Note that the general observed relationships are that higher fault clay contents relate to lower fault permeabilities and higher static sealing capacities.

The clay smear potential at a point is defined as follows:

CSP ¼

X Shales in slipped interval

ti2 di

and represents the sum of the contributions for each clay unit i of thickness ti that has moved a distance di past the point of interest. The majority of clay

smearing algorithms initiate the clay smear at the clay bed against the fault and shear that clay unit along the fault plane. In contrast, the PSSF algorithm stochastically populates smear patches along the fault surface between the footwall and hanging wall cut-offs. The central assumption is that the smears are continuous for SSF SSFc (SSFc ¼ critical shale smear factor) and the smears are placed at a random location between the source shale layers for SSF . SSFc. The model is based on observations from Lindsay et al. (1993) of a critical shale smear

268


factor related to the displacement to bed thickness ratio that the beds can endure prior to the smear becoming discontinuous. When SSFc is exceeded for a single clay smear, the probability that a gap occurs in that smear at some point between the source layer cut-offs is

capacities; the clay content distribution alone provides only a limited indication of the likely impact on fluid flow. In the following section more direct indicators of flow and/or sealing are discussed.

Cross-fault fluid flow properties t(SSFc 1) 1 dt where d is the fault displacement and t is the apparent source layer thickness along the fault (Childs et al. 2007). The PSSF value at P then defines the probability of having no smear and is obtained from applying the previous equation for all clay layers that have moved past P. If SSF SSFc for any of these layers then PSSF ¼ 0. Continuous shale beds (and sands) provide clay contents in the fault zone that are controlled by the original clay content that was present in the undeformed or unsheared stratigraphy. The impact of including clay smears in the calculation of the fault clay content distribution can be considerable, and in extreme cases it can switch the faults from being modelled as being open to being closed to flow. Figure 9a shows an example of the clay smear distribution generated from an inter-collated sand, impure sand and shale sequence (see Fig. 4). Also shown in Figure 9b is the prediction of the fault clay content distribution resulting from including clay smears in addition to the averaged ESGR clay mixing algorithm (compare this with Fig. 8b). The local variations of 50% in fault clay content in the most extreme areas of the fault equates to a permeability reduction of two orders of magnitude using a typical clay to permeability transform function (e.g. Jolley et al. 2007). A significant natural variability in fault properties is known to occur, and to account for this a series of algorithms have been developed that aim to capture the different processes that arise (e.g. Lindsay et al. 1993; Yielding et al. 1997; Childs et al. 2007). Currently only a limited suite of these algorithms are available in most commercial geological modelling packages, but the varying impact that they can have on the resulting flow simulation is significant (e.g. Freeman et al. 2008). Furthermore, it is not sufficient to only implement the algorithms, as the associated visualization capabilities are also required if the geoscientist or reservoir engineer is to be able to understand or make informed choices on the influence of modifying the spatial variability of these properties. Without the ability to view and analyse the data the system is a ‘black box’ of unknown validity. Fault clay prediction methods are a proxy tool used to infer fault permeabilities, transmissibilities or sealing

Fault permeability, transmissibility and hydraulic resistance Fluid flow in a porous medium (e.g. a rock mass) is proportional to the transmissibility of the medium, and inversely proportional to the viscosity of the fluid, under an applied pressure gradient (Darcy’s law; Darcy 1856). The transmissibility between two points in the medium is defined as the product of the harmonic average medium permeability and the area through which the flow occurs, divided by the distance between the two points. In order to integrate fault rock permeability data into flow simulation models, faults are usually incorporated as modifier values to the inter-block transmissibility applied between the grid cells on either side of the faults; this modifier is known as the transmissibility multiplier (TM) (Knai & Knipe 1998; Manzocchi et al. 1999). To be able to compute this value the host and fault rock permeabilities need to be combined with an estimate of the fault rock thickness and host grid cell sizes of the geocellular reservoir model (e.g. Manzocchi et al. 1999). Fault rock thickness values are typically estimated from the fault displacement (e.g. Hull 1988; Blenkinsop 1989; Evans 1990; Knott 1994; Antonellini & Aydin 1995; Knott et al. 1996; Childs et al. 1997; Manzocchi et al. 1999), but significant uncertainties are inherent in this prediction (Childs et al. 2009). The fault rock thickness estimate can usually be visualized on the fault plane but its simple relationship with displacement provides little additional geological understanding. The fault TM data can also be viewed on the fault surfaces within reservoir modelling packages. This data provides the fundamental control on the cross-fault fluid flow modelled by the reservoir flow simulator and can therefore be useful to interrogate. A TM value of zero indicates that the fault is modelled as a complete seal, whereas a TM value of one indicates that the fault has no impact on fluid flow (e.g. Manzocchi et al. 1999). The TM values typically vary between these ranges, although a TM value of greater than one is possible if the fault rock is more permeable than the host rock at that location. There are difficulties in interpreting TM data in isolation because the TM is intimately associated with the host rock permeability and the grid cell sizes. The result is that both high and low TM values can define zones of high (or low) cross-fault


269

Fig. 9. (a) The fault clay smear distribution (smear factor ¼ 3). (b) The fault clay distribution computed by combining the fault clay smearing algorithm and a fault clay mixing algorithm (averaged ESGR). The fault property data is shown only across the reservoir juxtaposition overlap areas. Combined clay mixing and smearing algorithms often produce local low-clay windows (high permeability) in high-clay (low permeability) zones.

flux. Figure 10 shows an example of the computed SGR and TM data for a single fault situated between an injector and producer pair, from a larger model with many more wells. As this 3D view demonstrates, the relationship between the SGR and TM values and the cross-fault fluid flow predicted by a streamline simulation is complex to interpret. The fluid flow across the fault is observed to be concentrated within the lower-TM zones (Fig. 10b). These areas are associated with the juxtaposition of high-permeability units (e.g. the reservoir sands). The intervening fault rock causes a reduction in the total cross-fault transmissibility and results in the low predicted TM values (c. 0.1 –0.2). Within these general low-TM zones, higher-TM sections tend to concentrate the flow; this is due to locally higher transmissibility fault rocks situated between equivalent high-transmissibility host rocks. The very high TM values with no associated streamlines (low or no flow) are zones of impermeable shale against the fault. Here, although the predicted shale-rich fault rock has a low permeability, this is almost equivalent to the host

shale permeabilities, and so the incorporation of the fault rock has little effect on the cross-fault transmissibility. Both higher and lower TM values can therefore correspond to increasing fluid flow depending on the host rock properties. It is only in circumstances when there is a constant host permeability in the hanging wall and footwall that the TM can be directly related to the impact on crossfault fluid flow; this is a highly unusual situation. A similar discrepancy between cross-fault fluid flow and the SGR is present (Fig. 10a). The fault rock permeability distribution does have a correlation with the fault clay content (normally log– linear with higher clay contents having considerably lower permeabilities), but because the fault rock is often very thin in comparison to the total flow path length the low fault permeability may have only a minor impact on the location and amount of cross-fault flow. Flow through the fault rock itself is controlled by the fault transmissibility, which (per unit area) is a combination of the fault permeability and thickness. Viewing only a proxy for the fault permeability (i.e. the fault clay content

270


Fig. 10. (a) Predicted fault clay content distribution (using the SGR algorithm) shown on the fault face and the streamline simulation results (white lines) showing the high-flow zones between an injector and producer pair. (b) Fault TM values and the same streamline simulation results. Note that neither the fault SGR nor TM values show much correlation with the high-flow zones as indicated by the streamline simulation. In the case of the TM image the high-flow zones occur in the low-TM areas, but low TMs would typically be expected to retard flow. This is a consequence of the varying host permeabilities that are controlling the flow.

distribution) thus limits the interpretability of this parameter. The flow in general will be controlled by the interaction of both the host and the fault transmissibility distributions, something that typically cannot be ascertained from visualizing the fault clay content in isolation. It is important to note that additional uncertainties in fault transmissibility predictions will result from the presence of uncertainty in the definition of the grid geometry, and critically this will influence the cross-fault juxtapositions that are developed. To provide a more useful means of assessing the variation in potential cross-fault fluid flow, the visualization of a set of parameters is required that intuitively relate the displayed property value to its effect on fluid flow. The fault hydraulic resistance, Rf ¼

tf kf

is the ratio of the predicted fault rock thickness, tf, to the fault permeability, kf, and is thus the inverse of the local transmissibility across the fault (i.e. the inverse of the transmissibility per unit area). This provides a measure of the local cross-fault fluid flow resistance created by the fault rock (see Fig. 11a). The fault hydraulic resistance values typically range over several orders of magnitude, with a larger hydraulic resistance value corresponding to a larger flow retardation by the fault rock alone (i.e. neglecting the local host rock juxtapositions across the fault). For example, a fault hydraulic resistance of 1 m/mD may arise from either a fault rock with a thickness of 0.1 m (i.e. a 10 m displacement fault with a thickness to displacement ratio of 1:100) and a permeability of 0.1 mD (e.g. a cataclasite), or a fault rock with a thickness of 0.001 m (i.e. a

10 cm displacement fault with a thickness to displacement ratio of 1:100) and a permeability of 0.001 mD (e.g. a clay smear). Fault hydraulic resistance aids in understanding the impact of the fault rock on the cross-fault fluid flow because it takes into account both the thickness and the permeability of the fault rock without relating it to its neighbouring host cells. In contrast, the TM is in part controlled by the juxtaposed host permeabilities and the size of the grid cells in the hanging wall and footwall. These latter parameters are grid-specific and are therefore only indirectly related to any physical property of the geological model. A second useful parameter is the cross-fault transmissibility (see Fig. 11b) Tf ¼

Akf tf

where A is the cross-fault grid connection area. The cross-fault transmissibility is therefore a direct measure of the potential fluid flux across the grid connection area of the fault for a given pressure gradient and phase, and hence should provide a good indication of potential cross-fault fluid flow. The cross-fault hydraulic resistance and fault transmissibility values are direct results of the mapped or estimated geological distributions and unrelated to any user-specified spatial grid geometry attributes. As such, they should provide consistent transferable parameters that can be compared across and between faults, and also between different fields or prospects. The cross-fault hydraulic resistance is likely to be more useful for investigations in exploration settings, where small leaky windows may have a significant impact on fluid flow, and


271

Fig. 11. Fault plane maps showing examples of (a) the fault hydraulic resistance and (b) the fault transmissibility.

hence hydrocarbon accumulations over geological time-scales. In contrast, the cross-fault transmissibility provides more information on fluid flow in a production environment, where the integrated effect of the cross-fault fluxes over all interfaces is the most important fault-related factor.

Figure 12 shows the fault TM value v. the fault transmissibility for a faulted sand –shale sequence. The inter-grid block transmissibility will have a linear relationship to the cross-fault fluid flow for constant pressure and phase conditions. The figure shows that many orders of magnitude variation are

Fig. 12. Comparison between the transmissibility multiplier and fault transmissibility value for cross-fault connections across a geocellular model. Note that, due to the strong influence of the host permeability on the TM, single TM values can relate to widely varying fault transmissibilities.

272


present in the inter-block transmissibility for similar TM values. This highlights the potential pitfalls of interpreting TM data to infer cross-fault fluid fluxes. Apart from the TM, the above parameters (crossfault hydraulic resistance and cross-fault transmissibility) provide an independent indication of the impact of fault rocks on fluid flow. The true impact on cross-fault flow is, however, a combination of both the pathways to, through and out of the fault zone. One method to assess this is to compute the compound transmissibility of the hanging wall, fault and footwall. A limitation with the usual inter-block transmissibility values is that the user-specified spatial grid geometry, as well as the geology, controls the result. In order to generate a transferable rating scheme, the widths of the hanging wall and footwall host rock used to define the total transmissibility need to be defined and normalized. This new parameter, the normalized crossfault transmissibility (hereafter called the effective cross-fault transmissibility or ECFT), is computed using the harmonic average of the permeabilities of the undeformed footwall adjacent to the fault, the fault rock and the undeformed hanging wall across the fault for a specific width of host wall rock on each side of the fault, and the fault rock thickness defined by the local displacement: ECFT ¼ A ((D tf 1 )=kh1 ) þ (tf 1 =kf 1 ) þ (tf 2 =kf 2 ) þ ((D tf 2 )=kh2 )

where D is the specific width of host rock on each side of the fault, kh1 and kh2 are the permeabilities of the host on sides 1 and 2, respectively, th1 and th2 are the thicknesses of the host on sides 1 and 2, respectively, kf1 and kf2 are the permeabilities of the fault on sides 1 and 2, respectively, and tf1 and tf2 are the thicknesses of the fault on sides 1 and 2, respectively. For a constant phase and pressure differential this value should therefore be proportional to the cross-fault fluid flux, without having to separately consider the host rock permeabilities, grid cell geometries or juxtaposition type. In this contribution we have chosen to define the ECFT by combining the predicted fault rock thickness (calculated from the fault displacement) and the associated fault rock permeability with prescribed 50 m thicknesses (a typical reservoir simulator grid block size) of each of the juxtaposed footwall and hanging wall lithologies (minus the fault rock width on each side) and the associated permeabilities of these lithologies. Figure 13 shows the relationship between the fault transmissibility (fault rock only) and the ECFT (incorporating fault rock and the juxtaposed host rocks). The graph shows how the ECFT (a

better gauge of the likely fluid flow response) changes from being dominated by the host rock permeability to being dominated by the fault rock permeability as the fault rock permeability decreases (in general, the contrast between the host and fault rock transmissibilities controls this dominance). This figure demonstrates that considering the fault transmissibility alone can also be misleading because there is no simple relationship to bulk transmissibility and hence cross-fault fluid flow. Following a similar approach to the ECFT (to remove the influence of the user-specified spatial grid geometry), we define the effective cross-fault permeability (ECFP) to be the length-weighted harmonic average of the host footwall, host hanging wall and fault rock permeabilities: ECFP ¼ 2D : ((D tf1 )=kh1 ) þ (tf1 =kf1 ) þ (tf2 =kf2 ) þ ((D tf 2 )=kh2 )

Again, the host rock length is normalized to a specified representative length scale (in this case 50 m minus the half-width of the fault rock applied on each side). The result is a new parameter that has direct relevance to flow simulation results because it measures the ability of fluids to move through the upstream host rock toward the fault, through the fault rock, and then through the juxtaposed host rock on the downstream side of the fault (as with the ECFT). The parameter also has the advantage that it is intuitive for geologists, geophysicists and reservoir engineers to understand and interpret. The ECFP displays a bulk permeability value that is scaled to a value equivalent to the reservoir grid cell scale (e.g. a 50 m or 100 m grid block size). A value of 40 mD for the ECFP in a grid of 50 mD host permeability is likely to have little effect, whereas an ECFP value of 0.1 mD will have a significant impact. Normalizing the host rock contribution to a specified length scale has the additional advantage that linear changes in the ECFP directly relate to linear changes in flow rate through the fault zone for a given phase, viscosity and pressure differential (as with the ECFT). This relationship is illustrated in Figure 14. In this example the streamline simulation results can be seen to closely follow the locations predicted by the higher ECFT (or ECFP) values. The low ECFT values correspond to the low or no flow zones. Figure 15 shows a series of possible cross-fault flow predictor properties mapped onto an example fault based on the methods discussed above. These images allow a comparison of the locations of likely cross-fault flow areas on the fault when considering: (a) the fault rock specific properties (i.e. the fault rock clay content, permeability and


273

Fig. 13. Fault rock transmissibility v. the ECFT (for one square metre of fault area) of the fault zone and host. The ECFT indicates the ease of flow through the host to the fault, through the fault and then out through the host. The ECFT depends on the specific width D of host rock on each side of the fault; in this example, a prescribed 50 m thickness was used. The fault transmissibility indicates the ease of flow through the fault rock only. Note that at places where the data points follow a horizontal straight line the host permeability is the dominant control on the flow (the juxtaposed host permeabilities kh1 and kh2 are indicated for each of these horizontal asymptotes), and where the points follow a 458 inclined straight line the fault rock controls the flow. The transition between the flow being dominated by the host and fault permeability is very narrow; the location of this transition depends on the relative transmissibilities of the juxtaposed host rocks and the fault rock.

transmissibility; see Fig. 15a –c); (b) the host rock juxtaposed properties (i.e. the harmonic average host permeability; see Fig. 15d); and (c) the combined fault and host rock properties (e.g. the TM, ECFT or ECFP distributions; see Fig. 15e, f ). The

diagram shows how interpreting either the fault rock or the host rock permeabilities alone leads to a significantly different interpretation of the likely cross-fault flux in comparison to using the combined fault and host data.

Fig. 14. An example of ECFT and the streamline simulation data from Figure 10. Note that the majority of streamlines are focused in the high-ECFT zones (hotter fault property colours). The low-ECFT areas are almost devoid of flow. The streamlines shown are limited to this particular injector– producer pair and are clipped by the time of flight. The high-ECFT areas not showing streamlines are associated with either longer flight time flow paths or flow between different well pairings.

274


Fig. 15. A comparison of different parameters to indicate likely cross-fault flow zones. Fault rock specific properties: (a) fault clay distribution estimated via the SGR algorithm; (b) fault permeability from the SGR using the Manzocchi et al. (1999) fault clay to permeability transform; and (c) fault transmissibility based on the fault permeability in (b). The fault transmissibility provides the most direct indicator of flow potential through the fault rock alone. Host rock specific property: (d) juxtaposed harmonic host permeability. Combined host and fault rock properties: (e) transmissibility multipliers; and (f) ECFP (normalized to 50 m host cells).


The analysis of cross-fault fluid flow using flow simulation data The techniques described so far facilitate the visualization of predicted fault rock properties and the visualization of the probable responses of crossfault fluid flow to those properties. These static fault properties that are used to infer the flow response assume a uniform or buoyancy-driven pressure field. Production techniques may raise the pressure at injectors and/or the drawing down at producers, thereby negating the previous assumption. To improve the prediction or visualization of cross-fault fluid flow the impact of production and injection needs to be taken into account. The majority of fault seal analysis techniques are only aimed at predicting inputs for flow simulation, and do not analyse in any detail the results of that flow simulation to better understand how the cross-fault fluid flow has contributed to the general hydrocarbon distribution within the model. By back-analysing the resulting cross-fault fluid flux developed during the simulation process a better understanding of the system can be gained and this can provide a more informed quality assessment of the result, as well as input for the calibration and validation of fault seal workflows. The majority of flow simulation results provide relatively little direct data with which to understand the impact of faults on the fluid flow (apart from at specific places, e.g. where phase boundaries are influenced by faults). The flow simulation results (e.g. phase pressure, saturation and mobility) tend

275

to be single average values per grid cell, but unfortunately related vector (directional) data is often not available for interrogation. This simulation data does, however, have the potential to provide a detailed understanding of how the faults have influenced fluid flow at different times through the simulation. If the resulting data are resolved across the faults at any particular time step then the local crossfault fluid flow component can be ascertained. With these results, the degree and importance of crossfault fluid flow can be defined at any particular time during the simulation run. Figure 16 shows the normalized cross-fault fluid flux derived from a streamline simulation model for a single fluid phase using the instantaneous pressure field at a certain time through the simulation. The direct visualization and analysis of the cross-fault fluid flux developed during a flow simulation should help to answer two questions relating to the geological model. Firstly, are the predicted flow and no flow locations along the faults geologically realistic? Secondly, what impact does changing the fault property prediction algorithm have on the observed cross-fault fluid flow? These questions have previously been difficult to answer, but are fundamental in controlling and assessing the resulting flow simulation prediction. A more routine interrogation of the detail involved in the flow simulation in relation to faults is a natural progression of the drive toward increasingly sophisticated simulation models. This workflow can be readily extended to incorporate geometric (e.g. fault throw) uncertainty, but this involves additional

Fig. 16. Fault face colour-coded for the back-computed normalized cross-fault fluid flux (‘hotter’ colours denote higher cross-fault fluid fluxes) derived from the pressure field generated from a streamline simulation. Also shown are the streamlines between one of the injector and producer pairs in the simulation model. Note that, as expected, the streamlines pass through the high cross-fault flux zones. Back-computing the cross-fault fluid flux and visualizing this property on the fault plane allows a direct imaging of the role that the fault plays in controlling flow in the reservoir.

276


visualization challenges due to the change in crossfault juxtaposition architecture.

Data integration: utilizing vector fields for fault seal and cross-fault flow analysis One of the current shortcomings in typical geological visualization packages is the inability to visualize both the detailed fault property data and the larger-scale field geometry simultaneously. This is primarily due to the often near-vertical nature of faults and the near-horizontal nature of the horizon topography, combined with the very high densities of property data that typically vary most rapidly in the fault dip direction. One of the most common visualization methods in the broader field of computational fluid dynamics is the use of vector fields for the visual analysis of property data or flow results. Figure 17 shows an example of a vector field representing the cross-fault harmonic average host permeability, which is often one of the principal controls on cross-fault fluid flow. The vectors have been drawn normal to the fault surface (oriented

Fig. 17. The cross-fault harmonic average host permeability represented by vectors (drawn normal to the fault surface and oriented in the local dip direction of the fault). In this example, the low-throw faults or the tip zones show high juxtaposed harmonic host permeabilities and this is likely to correspond to high-flow zones, whereas higher-throw zones correspond to juxtapositions with lower permeability host units. Also shown is the base reservoir topography. Using vectors to show flow or fault properties allows that data to be more easily visualized and understood in the context of the overall reservoir geometry and production configuration.

vertically upwards) and their magnitude represents the harmonic average of the juxtaposed host permeabilities. The vectors (e.g. Fig. 17) provide a means of integrating a detailed understanding of the fault in the context of the geocellular model and nearby well control. Similarly, such vector field plots can be used to provide a rapid means of visualizing the cross-fault fluid flow distribution in the model at specific time steps through the flow simulation.

Comparing the predictive capabilities of fault zone properties In this paper a series of different techniques have been presented that either enhance those currently in use within the industry, or have been recently developed to aid our understanding of fault seal and cross-fault fluid flow issues. Comparative studies are required in order to assess how well these techniques, and the fault permeability and flow property measures introduced, predict the flow behaviour in the subsurface. Comparative parameter datasets have only become available following the development of techniques to back-analyse the flow simulation results with respect to cross-fault fluid flux. Figure 18 shows various comparisons between four different parameter predictions and the crossfault fluid flux derived from the reservoir simulation. Limited correlation exists between the resulting cross-fault fluid flow and either the predicted fault rock clay content or the corresponding fault permeability (Fig. 18a, b). A reservoir simulation is itself only a model of the physical processes operating in a reservoir that attempts to match or predict reservoir flow parameters – it is not a truth. Thus, ‘correlations’ derived from a model are subject to the extent to which the ‘model’ is valid. This is not surprising because these parameters, although controlling the flow across the fault rock in isolation, do not capture any of the host rock flow properties. Similarly, Figure 18c demonstrates the poor relationship between the harmonic-averaged juxtaposed host permeabilities and the cross-fault fluid flux. Considered in isolation, the separate fault and host rock permeabilities will show good correlations with the crossfault fluid flow for certain geometric and property configurations. The accuracy of the correlation will be controlled by the relative contributions of the fault and host transmissibility to the bulk transmissibility of the entire system. This relationship is characterized in Figure 13. The ECFT described above is a parameter that takes into account both the host and the fault rock properties and shows a better correlation with the observed cross-fault fluid flux (Fig. 18d). The scatter in this relationship


277

is primarily a function of the cross-fault pressure differential developed along the faults. A direct correlation between the cross-fault fluid flux and the ECFT should be present for a single-phase fluid subject to a constant pressure differential. However, in the fluid flow simulation used to produce Figure 18, the cross-fault pressure difference varied continuously across the fault, thereby causing the observed scatter. When additionally the local pressure difference is considered, the data in Figure 18d show highly clustered corridors that highlight the good correlation between the observed and predicted cross-fault fluid flow for a certain constant pressure difference. The general result is that neither the fault nor the host rock property relationships should be used in isolation to define the likely cross-fault fluid flux. Variations in this relationship are due to lateral and vertical cross-fault pressure changes. In the absence of flow simulation data, the ECFT (or equivalent permeability) should provide an accurate assessment of the probable cross-fault fluid flux throughout the model, assuming that multiphase fluid flow effects can be neglected. An additional and useful step prior to full field simulation is the application of a streamline simulation and then the back-analysis of the observed cross-fault fluid flow. This should provide a good estimate of the pressure field and hence provide a method to compare the geologically-predicted cross-fault fluid flux with the cross-fault fluid flow data provided by a full-field production simulation. The ECFT and ECFP parameters can therefore provide a useful predictive tool that can help screen the geological reservoir and production simulation models prior to and during initial or full-field simulation. The computation of these properties is fast and can also be applied ‘on-the-fly’ (for the static properties). These parameters can therefore aid in the quality control and optimization of geological scenarios used in flow simulations by reservoir engineers. The application of this technology should therefore help to enhance integrated reservoir modelling practice, and improve the quality of the models being taken forward. Fig. 18. Predicted static fault property values v. observed normalized cross-fault fluid flux defined via flow simulation. A consistent straight line would indicate a direct correlation between the modelled cross-fault flow and the approximation parameter. Deviations away from the line show the level of accuracy of the prediction. (a) The predicted fault clay content v. the associated observed cross-fault fluid flux. Very little correlation is present, with very low clay contents both relating to very high and very low cross-fault flux values. (b) The predicted fault permeability v. the observed cross-fault fluid flux. Again little correlation exists, with high cross-fault fluxes observed across low, moderate and high-permeability fault rocks. (c) The harmonic host permeability (or transmissibility) v. the observed cross-fault fluid flux. A clear clustering of the data is present, but this only demonstrates that the flow occurs across zones with the juxtaposition of high-permeability sands against high-permeability sands. Within that style of juxtaposition no further correlation is evident. (d) The predicted ECFT v. the observed crossfault fluid flux. A good correlation is evident. The principal reason for the scatter in the data is due to the variation in pressure across the fault. The ECFT should provide a linear correlation with flux for a constant pressure differential.

278


Screening different static and dynamic scenarios The fault seal analysis process contains a wide variety of uncertainties and potential workflows (Fisher & Jolley 2007). Determining which approach is suitable for determining flow within a given faulted reservoir is often quite complex. Properties such as the ECFT are helpful because they provide a potential proxy for fluid flow that is quick to compute

from the static model. Visualizations such as crossfault vectors can provide a more global context to better understand how these different parameters affect flow; however, visualizations or computations that generate comparable scenarios are required. Figure 19 shows a simple example of the use of a pie chart plot that allows the comparison of numerous geological sealing scenarios against one another. The bulk cross-fault permeability for each cross-fault connection has been summed for each

Fig. 19. Top reservoir structure map with pie charts showing the bulk cross-fault permeability (juxtaposed host rock and fault rock permeabilities) per fault (and located near the centre of each fault). The red segments show the bulk permeability (area-weighted average cross-fault permeability) when fault rock is included in the calculation and the blue segments show the bulk permeability calculated by neglecting the fault permeability. The size of the pie charts indicates the total cross-fault permeability, so that faults with larger pie charts are more likely to have a higher cross-fault fluid flux. Note that the pie charts with small red segments show that incorporating fault rock dramatically reduces the cross-fault permeability across that structure. The plot allows a reservoir scale perspective and a rapid understanding of how different scenarios will impact on the potential cross-fault flow, and highlights where the critical faults are in the system.


fault segment, based on the two scenarios of including (red pie segments) or excluding (blue pie segments) the predicted fault properties. The overall size of the pie chart for each fault indicates the total cross-fault permeability. Such plots allow the relative impact of different fault sealing scenarios on potential reservoir flow to be compared and evaluated on a field-wide scale whilst retaining the geological context. Such plots are particularly useful, for example, for analysing variations in cross-fault facies juxtaposition areas for varying facies models. Engineering requirements to make the modelled fault system more or less open to flow in order to honour production history data can hence be more easily constrained by visualizing the geological parameters that should control that system.

Summary Developing tools to analyse and visualize crossfault fluid flow is a natural progression in the drive toward improved exploration, prospect evaluation and reservoir simulation. As more fault property data becomes available, a wider range in structural processes, physical properties and flow behaviour are being defined (Yielding et al. 1997; Manzocchi et al. 2000, 2002; Fisher & Knipe 2002; Sperrevik et al. 2002; Childs et al. 2007). Thus, it is becoming increasingly clear that using a best-fit single fault property model that can be uniformly applied to all faults in all reservoirs is unrealistic (see Fisher & Jolley 2007 for discussion). Proxy or direct crossfault fluid flow visualization tools allow informed choices to be made on which fault parameter relationships are appropriate for any given reservoir. In favourable circumstances, observable parameters such as time-lapse seismic and well tracer data may help to constrain the choice of potential scenarios. The geocellular visualization of fault properties within the grid adjacent to the faults, on the faulted grid connections (i.e. the fault face), or as vectors across the fault, can provide a greater understanding of the potential for cross-fault flow and how that could vary within a modelled reservoir. Using a combination of different techniques a variety of subsurface datasets can be visualized simultaneously with fault zone properties. This can help in developing a better understanding of the inter-relationships between different properties (e.g. cross-fault flow, well distribution and reservoir architecture), and also helps in the validation of the model geometry that underpins compartmentalization studies. These visualization tools are equally applicable to exploration and production environments. In the exploration environment the rapid effective screening (e.g. via fault plane property maps of facies juxtapositions) is equally relevant but often on a different scale to the production situation.

279

In addition to juxtaposition and traditional fault membrane properties, we have introduced and visualized parameters that can be considered to be proxies for cross-fault fluid flow (e.g. ECFT and ECFP). The direct analysis of cross-fault fluid flux from flow simulation data allows this technology to bridge the gap between geoscientists and reservoir engineers, and vice versa. These forms of data have historically been overlooked during fault seal studies but are the next logical step in improving the accuracy and speed of these analyses. Thus, in this paper, we have shown that combinations of techniques and datasets, along with the re-computing of data, informs more directly on specific cross-fault fluid flow. This has the potential to improve understanding and therefore improves the prediction of hydrocarbon flow across fault zones or hydrocarbon sealing behind faults. The hydraulic resistance and the transmissibility of faults provide direct and easily interpretable measures of the fault rock impedance or flux potential for a given pressure differential, phase and viscosity. These parameters are independent of the user-specified spatial grid geometry and so should represent transferable parameters with which to assess the likely fluid flow across different parts of faults, different faults, different fault geometries, different model vintages and between different fields. This is in contrast to the fault TM, which is the fault property commonly used and visualized in the reservoir modelling packages. This property is useful as a mechanism to integrate fault retardation into the simulator, but is often difficult to interpret relative to the likely or observed fluid flow across faults. The ECFT and associated ECFP are parameters that provide more intuitive, direct proxies to inform on the cross-fault fluid flux of faults. These parameters take into account the host rocks and fault rocks, and are normalized for a specified length scale. This results in a set of parameters that are linearly associated with the fluid flux across the fault zone, for a single phase fluid at a given pressure differential and viscosity, and their values are directly comparable between different faults and grids. Back-computing and visualizing the results of a flow simulation in comparison to the static fault property predictions is an important stage for the validation of integrated reservoir modelling scenarios. In this contribution we have back-calculated the cross-fault fluid flux developed during a synthetic simulation, and displayed this data back in its geological context (here as fault face values and cross-fault vectors). This method allows the relative contributions of the faults to reservoir compartmentalization to be constrained, and the models updated accordingly, since the critical areas of the

280


model are highlighted and the associated geological parameters can then be revised. This should lead to a more geologically robust and rapid means of attaining reservoir and production simulation models that honour all of the subsurface and production data. As with all other areas of reservoir modelling, numerous potentially valid scenarios are possible for the fault seal prediction. Techniques are therefore required to effectively screen the impact of the different approaches to the problem. The screening technique needs to occur at the scale of the overall system. By generating fault-wide summations, the impact of different scenarios can be seen at the reservoir scale and the impact of choosing between the different scenarios is easier to discern.

Conclusions The analysis and visualization techniques outlined in this paper will, if utilized appropriately, allow for a more robust quality control and an improved understanding of the role of fault sealing in geological and flow simulation models that are being used to identify, evaluate and manage the extraction of hydrocarbon reserves. † Developing tools that analyse or visualize the fault zone geology in a way that directly relates to the flow simulation process should facilitate a closer collaboration between geologists, geophysicists and reservoir engineers involved in integrated reservoir modelling. † At the exploration stage, rapid ‘quick-look’ fault plane maps provide an effective means of de-risking prospects. At the evaluation stage, more detailed juxtaposition maps are useful. Traditional line drawings of stratigraphic juxtapositions are complex to interpret, whereas the colour-filled, property-specific filtering of fault plane maps is more informative and intuitive to use. † Highlighting host and fault rock parameter windows across faults, particularly when visualized as cells adjacent to the fault, allows for the rapid identification of key areas of the geological or simulation grid that will impact on the resulting flow simulation model. † Implementing the range in different prediction algorithms and testing the probable impacts of those models is important if the variability in geological properties is to be understood. † Fault transmissibility and hydraulic resistance provide transferable rating schemes to visualize and analyse the likely cross-fault fluid flow predicted through a fault rock.

† The effective cross-fault transmissibility (ECFT) and associated permeability (ECFP) provide transferable parameters that can be used to infer likely cross-fault fluid flux from static models. † The back-analysis of flow simulation data to highlight the predicted cross-fault fluid flux provides a method to constrain fault parameter calculations in geological reservoir and production simulation models. This simulation data provides the most direct indication of the likely effects of faults on hydrocarbon distribution and flow. † The ECFT prediction shows a strong correlation with the resulting cross-fault fluid flux, especially when considering the variable pressure differences across the faults. In the examples shown, the ECFT provided the most accurate prediction of the cross-fault fluid flux observed via simulation. This property therefore appears to provide a useful proxy to infer cross-fault fluid flow that can be generated from the static geological model prior to flow simulation. † Cross-fault fluid flux computed from either streamline or full simulation data provides a more intuitive method of visualizing the impact of faults on hydrocarbon fluid distributions. This information can be used in conjunction with knowledge of the uncertainties in the seal analysis to generate new models that better honour both the geology and the dynamic data. Different viable geological scenarios can be defined that more appropriately capture the likely range in fault seal contribution to reservoir compartmentalization. Some potential techniques are outlined here that might help visualize the impact of these scenarios and the uncertainties associated with them. However, additional techniques are now required that provide data at the scale relevant to the simulation, that is, on a field-wide scale. The goal for the future is to develop techniques that rapidly define the most viable geological models whilst honouring dynamic data. Extending the analysis to incorporate geometric uncertainty is key to developing a holistic understanding. First of all, we thank the reviewers (Fred Dula and Stephen Dee) and the Lead Editor (Steve Jolley) for their insightful comments that have helped to improve the focus of this paper. We also thank all of the members of RDR Ltd that have been involved in the development process of the fault property tools and workflows presented in this paper, particularly Raoul Treverton, Will Bradbury, Philip Jones, Kevin Wood, Rick Berry, Nikki McCabe and Viki O’Connor. The visualization techniques presented here have been developed within RDR and are implemented within PetrelTM (www.slb.com).


References Al-Busafi, B., Fisher, Q. J. & Harris, S. D. 2005. The importance of incorporating the multi-phase flow properties and thickness of fault rocks into production simulation models. Marine and Petroleum Geology, 22, 365–374. Allan, U. S. 1989. Model for hydrocarbon migration and entrapment within faulted structures. American Association of Petroleum Geology Bulletin, 73, 803–811. Antonellini, M. & Aydin, A. 1995. Effect of faulting on fluid flow in porous sandstones: geometric properties. American Association of Petroleum Geology Bulletin, 79, 642–671. Aydin, A. & Eyal, Y. 2002. Anatomy of a normal fault with shale smear: implications for fault seal. American Association of Petroleum Geology Bulletin, 86, 1367–1381. Blenkinsop, T. G. 1989. Thickness–displacement relationships for deformation zones: discussion. Journal of Structural Geology, 11, 1051– 1053. Childs, C., Walsh, J. J. & Watterson, J. 1997. Complexity in fault zone structure and implications for seal prediction. In: Møller-Pedersen, P. & Koestler, A. G. (eds) Hydrocarbon Seals: Importance for Exploration and Production. Norwegian Petroleum Society (NPF), Special Publication 7. Elsevier, Amsterdam, 61–72. Childs, C., Walsh, J. J. et al. 2007. Definition of a fault permeability predictor from outcrop studies of a faulted turbidite sequence, Taranaki, New Zealand. In: Jolley, S. J., Barr, D., Walsh, J. J. & Knipe, R. J. (eds) Structurally Complex Reservoirs. Geological Society, London, Special Publications, 292, 235–258. Childs, C., Manzocchi, T., Walsh, J. J., Bonson, C. G., Nicol, A. & Scho¨pfer, M. P. J. 2009. A geometric model of fault zone and fault rock thickness variations. Journal of Structural Geology, 31, 117– 127. Darcy, H. 1856. Les Fontaines Publiques De La Ville De Dijon (“The Public Fountains of the Town of Dijon”). Dalmont, Paris. Dee, S. J., Freeman, B., Yielding, G., Roberts, A. & Bretan, P. 2005. Best practice in structural geological analysis. First Break, 23, April 2005, 49–54. 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 Publication, 292, 259–270. Evans, J. P. 1990. Thickness–displacement relationships for deformation zone. Journal of Structural Geology, 12, 1061– 1065. Fisher, Q. J. & Jolley, S. J. 2007. Treatment of faults in production simulation models. In: Jolley, S. J., Barr, D., Walsh, J. J. & Knipe, R. J. (eds) Structurally Complex Reservoirs. Geological Society, London, Special Publications, 292, 219– 233. Fisher, Q. J. & Knipe, R. J. 1998. Fault sealing processes in siliciclastic sediments. In: Jones, G., Knipe, R. J. & Fisher, Q. J. (eds) Faulting, Fault Sealing and Fluid Flow in Hydrocarbon Reservoirs. Geological Society, London, Special Publications, 147, 117– 134.

281

Fisher, Q. J. & Knipe, R. J. 2002. The permeability of faults within siliciclastic petroleum reservoirs of the North Sea and Norwegian continental shelf. Marine and Petroleum Geology, 18, 1063–1081. Fredman, N., Tveranger, J., Semshaug, S., Braathen, A. & Sverdrup, E. 2007. Sensitivity of fluid flow to fault core architecture and petrophysical properties of fault rocks in siliciclastic reservoirs: a synthetic fault model study. Petroleum Geoscience, 13, 305–320. Freeman, S. R., Harris, S. D. & Knipe, R. J. 2008. Fault seal mapping – incorporating geometric and property uncertainty. In: Robinson, A., Griffiths, P., Price, S., Hegre, J. & Muggeridge, A. (eds) The Future of Geological Modelling in Hydrocarbon Development. Geological Society, London, Special Publications, 309, 5– 38. Hull, J. 1988. Thickness– displacement relationships for deformation zone. Journal of Structural Geology, 10, 431 –435. James, W. R., Fairchild, L. H., Nakayama, G. P., Hippler, S. J. & Vrolijk, P. J. 2004. Fault-seal analysis using a stochastic multifault approach. American Association of Petroleum Geologists Bulletin, 88, 885– 904. Jolley, S. J., Dijk, H., Lamens, J. H., Fisher, Q. J., Manzocchi, T. & Eikmans, H. 2007. Faulting and fault sealing in production simulation models: Brent Province, northern North Sea. Petroleum Geoscience, 13, 321 –340. Knai, T. A. & Knipe, R. J. 1998. The impact of faults on fluid flow in the Heidrun Field. In: Jones, G., Knipe, R. J. & Fisher, Q. J. (eds) Faulting, Fault Sealing and Fluid Flow in Hydrocarbon Reservoirs. Geological Society, London, Special Publications, 147, 269– 282. Knipe, R. J. 1997. Juxtaposition and seal diagrams to help analyze fault seals in hydrocarbon reservoirs. American Association of Petroleum Geologists Bulletin, 81, 187– 195. Knipe, R. J., Fisher, Q. J. et al. 1997. Fault seal analysis: successful methodologies, application and future direction. In: Møller-Pedersen, P. & Koestler, A. G. (eds) Hydrocarbon Seals: Importance for Exploration and Production. Norwegian Petroleum Society (NPF), Special Publication 7, Elsevier, Amsterdam, 15– 40. Knipe, R. J., Jones, G. & Fisher, Q. J. 1998. Faulting, fault seal and fluid flow in hydrocarbon reservoirs: an introduction. In: Jones, G., Knipe, R. J. & Fisher, Q. J. (eds) Faulting, Fault Sealing and Fluid Flow in Hydrocarbon Reservoirs. Geological Society, London, Special Publications, 147, vii– xxi. Knott, S. D. 1994. Fault zone thickness v. displacement in the Permo-Triassic sandstone of NW England. Journal of the Geological Society, London, 151, 17– 25. Knott, S. D., Beach, A., Brockbank, P. J., Lawson Brown, J., McCallum, J. E. & Welbon, A. I. 1996. Spatial and mechanical controls on normal fault populations. Journal of Structural Geology, 18, 359– 372. Kristensen, M. B., Childs, C. J. & Korstgard, J. A. 2008. The 3D geometry of small-scale relay zones between normal faults in soft sediments. Journal of Structural Geology, 30, 257–272. Lehner, F. K. & Pilaar, W. F. 1997. The emplacement of clay smears in syn-sedimentary normal faults:

282


inferences from field observations near Frechen, Germany. In: Møller-Pedersen, P. & Koestler, A. G. (eds) Hydrocarbon Seals: Importance for Exploration and Production. Norwegian Petroleum Society (NPF), Special Publication 7, Elsevier, Amsterdam, 39–50. Lindsay, N. G., Murphy, F. C., Walsh, J. J. & Watterson, J. 1993. Outcrop studies of shale smears on fault surfaces. In: Flint, S. T. & Bryant, A. D. (eds) The Geological Modelling of Hydrocarbon Reservoirs and Outcrop. International Association of Sedimentologists, Oxford, Special Publication, 15, 113– 123. Manzocchi, T., Walsh, J. J., Nell, P. & Yielding, G. 1999. Fault transmissibility multipliers for flow simulation models. Petroleum Geoscience, 5, 53– 63. Manzocchi, T., Heath, A. E., Walsh, J. J. & Childs, C. 2000. Fault-rock capillary pressure: extending fault seal concepts to production simulation. Norwegian Petroleum Society conference on hydrocarbon seal quantification, Stavanger, Norway. Extended abstracts. Norwegian Petroleum Society (NPF), 101 –106. Manzocchi, T., Heath, A. E., Walsh, J. J. & Childs, C. 2002. The representation of two phase fault-rock properties in flow simulation models. Petroleum Geoscience, 8, 263–277. Manzocchi, T., Heath, A. E., Palananthakumar, B., Childs, C. & Walsh, J. J. 2008. Faults in conventional flow simulation models: a consideration of representational assumptions and geological uncertainties. Petroleum Geoscience, 14, 91–110. Marchal, D., Guiraud, M. & Rives, T. 2003. Geometric and morphologic evolution of normal fault planes and

traces from 2D to 4D data. Journal of Structural Geology, 25, 135–158. Shipton, Z. K., Evans, J. P., Robeson, K. R., Forster, C. B. & Snelgrove, S. 2002. Structural heterogeneity and permeability in faulted eolian sandstone: implications for subsurface modelling of faults. American Association of Petroleum Geology Bulletin, 86, 863–883. Sperrevik, S., Gillespie, P. A., Fisher, Q. J., Halvorsen, T. & Knipe, R. J. 2002. Empirical estimation of fault rock properties. In: Koestler, A. G. & Hunsdale, R. (eds) Hydrocarbon Seals Quantification. Norwegian Petroleum Society (NPF), Special Publication 11, Elsevier, Amsterdam, 109– 125. Walsh, J. J. & Watterson, J. 1987. Distributions of cumulative displacement and seismic slip on a single normal fault surface. Journal of Structural Geology, 9, 1039– 1046. Yielding, G. 2002. Shale gouge ratio – calibration by geohistory. In: Koestler, A. G. & Hunsdale, R. (eds) Hydrocarbon Seals Quantification. Norwegian Petroleum Society (NPF), Special Publication 11, Elsevier, Amsterdam, 1– 15. Yielding, G., Freeman, B. & Needham, D. T. 1997. Quantitative fault seal prediction. American Association of Petroleum Geology Bulletin, 81, 897– 917. Zijlstra, E. B., Reemst, P. H. M. & Fisher, Q. J. 2007. Incorporation of fault properties into production simulation models of Permian reservoirs from the southern North Sea. In: Jolley, S. J., Barr, D., Walsh, J. J. & Knipe, R. J. (eds) Structurally Complex Reservoirs. Geological Society, London, Special Publications, 292, 295 –308.

An uncertainty modelling workflow for structurally compartmentalized reservoirs A. D. IRVING1,2*, E. CHAVANNE3, V. FAURE4, P. BUFFET4 & E. BARBER4 1

Total E&P UK, Geoscience Research Centre, Crawpeel Road, Altens Industrial Estate, Aberdeen, AB12 3FG, UK

2

Fault Analysis Group, School of Geological Sciences, University College Dublin, Belfield, Dublin 4, Ireland

3

Total SA, Geoscience Technologies, CSTJF, Avenue Larribau, 64018 Pau, France

4

Total E&P UK, Crawpeel Road, Altens Industrial Estate, Aberdeen, AB12 3FG, UK *Corresponding author (e-mail: [email protected]) Abstract: Despite their significance, structural parameters are sometimes neglected in assessments of uncertainty on connected volumes and forecast production for compartmentalized reservoirs. A workflow is proposed for modelling multiple realizations of fault geometry and properties using 3D geomodelling software. Geometrical parameters that may be simulated include fault and horizon shape and location, fault displacement and fault pattern, while property variables include fault permeability, thickness and clay smears. Realizations are ranked by estimated connected volume, with selected models being exported for numerical flow simulation. Experimental design is used to assess sensitivity of forecast production and pressure to different parameters. The workflow is illustrated using a North Sea reservoir, in which structural heterogeneities cause considerable uncertainty on connected volumes, with implications for history matching and infill well planning. Fault geometry and permeability were the most important properties for all studied responses, however their relative significance could vary between early and late field life. A number of improvements are proposed, chiefly in the areas of connected volume estimation, handling of uncertain grid geometries and calculation of stress- or saturation-dependent fault permeabilities. Finally, the method can be integrated with conventional sedimentary and petrophysical uncertainties to investigate interactions and relative sensitivities with regard to structural parameters.

Faults affect fluid flow in sedimentary basins on multiple length and time-scales. Over a time period of thousands to millions of years, faults can act as vertical hydrocarbon migration pathways (Weber et al. 1978). Faults may also impede lateral fluid movement, ultimately providing a trapping mechanism for significant hydrocarbon accumulations (Smith 1966, 1980; Watts 1987; Harding & Tuminas 1989; Knott 1993; Gibson 1994; Berg & Avery 1995) or acting as baffles to economic oil and gas production (Bentley & Barry 1991; Jev et al. 1993; Harris et al. 2002). Faults affect lateral fluid flow by juxtaposing different sedimentary layers (Allan 1989; Knipe 1997) or due to the petrophysical properties of the fault zone materials themselves, which result from processes such as particulate flow and grain crushing (cataclasis); entrainment of fine-grained material into the fault zone and preferential diagenesis and mineral reprecipitation (Yielding et al. 1997). Sampling of subsurface information is limited to spatially extensive but low resolution seismic

data and sparse, high resolution well penetrations. Large (greater than c. 10 m displacement) faults are usually avoided as a drilling hazard (Cerveny et al. 2004) and very rarely cored (Hippler 1997; Aarland & Skjerven 1998; Knai & Knipe 1998), while typical seismic surveys are designed to image sub-horizontal reflectors rather than faults (van der Poel & Cassell 1989; Worthington & Hudson 2000). This lack of knowledge must therefore be filled by making inferences from observed responses (inverse modelling) or by employing some kind of conceptual forward model to predict fault rock properties from available information. These tasks have been greatly facilitated by methods developed in the past decade, which allow fault sealing behaviour in siliciclastic sequences to be more realistically quantified and represented in geocellular reservoir models (Lindsay et al. 1993; Yielding et al. 1997; Manzocchi et al. 1999). Algorithms such as Shale Gouge Ratio (SGR), Shale Smear Factor (SSF) and Clay Smear Potential (CSP) provide a quantitative prediction of the


284

A. D. IRVING ET AL.

amount of fine-grained (phyllosilicate) material contained within a fault zone. Published empirical transformations (Manzocchi et al. 1999; Sperrevik et al. 2002) provide estimates of fault rock permeability, which may be readily included in numerical reservoir simulations through the use of single-phase transmissibility multipliers on cell connections. Yielding (2002) reported the successful application of this approach to modelling intrareservoir faults in the Scott field, and similar methodologies have resulted in good history matches for the Snorre (Sverdrup et al. 2003) and Heidrun fields (Knai & Knipe 1998), in the latter case using fault properties derived from small cored faults. Fault properties derived from core analysis have also been successfully employed for fields in the UKCS Brent Province (Jolley et al. 2007); however the authors attached greater importance to rigorous structural interpretation and quality control of reservoir simulation grids to ensure correct reservoir ‘plumbing’ across and around faults. The grid type most commonly used in both geological modelling and reservoir simulation is the stratigraphic grid, where the top and base reservoir surfaces are connected by linear pillars aligned parallel to faults. Internal reservoir layering is defined by sub-parallel intermediate surfaces; together with vertical sections between pillars, these define the boundaries of each grid cell. The popularity of these grids stems from the ability to model geological properties and perform flow simulation using a shared model. Stratigraphic grids are, however, a compromise between the needs of these different tasks, which may necessitate simplification of the structural interpretation, for example, modification or removal of some fault surfaces in order to build a coherent grid. Within the stratigraphic grid, geostatistical algorithms are used to interpolate and/or simulate properties interpreted from well data. As wells only sample a small fraction of the reservoir volume, it is impossible to produce a ‘true’ representation of the subsurface. Instead, geostatistical simulation algorithms aim to produce multiple equally probable realizations that are globally accurate; reproducing the interpreted property continuity while honouring all available data. Advances in both Earth modelling software and computer power have led to the widespread application of such stochastic techniques. In field appraisal and development studies, it is now standard practice to generate hundreds or thousands of equally probable combinations of uncertain input parameters such as sedimentary facies, porosity and permeability. The resulting distributions of connected volumes provide estimates of uncertainty which can aid in decision making and economic forecasting. Save for a few exceptions in the literature (England & Townsend 1998; Ottesen et al.

2005; Rivenæs et al. 2005), uncertainties in fault geometry and properties are typically excluded from such modelling workflows, despite their potential significance for production forecasts (e.g. Lia et al. 1997; Damsleth et al. 1998). Overlooking these factors could lead to sub-optimal or even failed field developments and associated unnecessary costs. To avoid such pitfalls, uncertainties in fault geometry and properties should be modelled and their influence on connected volumes assessed as for sedimentary and petrophysical properties. The aim of this paper is to outline a workflow for modelling uncertainty on connected hydrocarbon volumes and forecast production in reservoirs that are compartmentalized by faults. Two categories of uncertainty are defined: geometrical and property uncertainties. The shape and location of fault and horizon surfaces are the principal geometrical factors, which arise from uncertainties in the acquisition, processing and interpretation of reflection seismic data. Property uncertainties include: fault permeability; thickness; and clay distribution. These parameters are estimated using appropriate analogue databases and empirical functions relating them to available model properties such as fault throw and SGR. Uncertainties can be defined on the constants and coefficients used in these empirical functions and on the underlying spatial continuity of the properties. The defined geometrical and property modelling parameters are then combined using stochastic simulation to generate hundreds of equally probable realizations of reservoir geometry and properties. Realizations may be ranked according to connected hydrocarbon volumes, and selected models exported for numerical fluid flow simulation. The interpretation and modelling of geometrical and property uncertainties will be first described. The workflow will then be illustrated using a field case from the UK sector of the North Sea. Results will be presented as distributions of connected volumes and forecast cumulative production; sensitivity of these outputs to different factors will be analysed using experimental design techniques. Finally, we will discuss the impact of these results on this field case and propose some extensions and improvements to the workflow.

Geometrical uncertainties The first step in the method is the interpretation of seismic data to build a coherent three dimensional network of faults and horizons. Geometrical uncertainties arising from seismic data are estimated and then modelled using a proprietary software package for stochastic simulation of reservoir structure. The processes for uncertainty estimation and stochastic simulation are described respectively by Thore

FAULT UNCERTAINTY MODELLING

et al. (2002) and Lecour et al. (2001) and summarized below. Uncertainties can be considered on six phases of the seismic ‘chain’, namely acquisition, preprocessing, stacking, migration, interpretation and time-to-depth conversion. Each source of uncertainty is described as a correlated vector field consisting of uncertainty direction, magnitude and correlation length at each point of a given surface. All surfaces are considered to be probabilistic, in each case located somewhere within an uncertainty volume surrounding the ‘best estimate’ interpretation. Faults are parameterized as objects with a central ‘backbone’ curve connecting secondary horizontal ‘generator’ curves, along which uncertainty values are stored at each node. Uncertainty vectors are specified at a few locations along the fault and interpolated where no information is provided. Horizons are described by triangulated surfaces, again with uncertainty values stored at each node. Finally, a set of constraints describes the geometric connections between faults and horizons, and an additional scalar property describes the correlation coefficient between horizons, in order to prevent undesired horizon crossing. Generation of surface realizations is performed using probability field (P-field) simulation, which consists of adding a spatially correlated random variable to the ‘best estimate’ interpretation. Using a constant random number for the random field will produce a bulk lateral shift of the fault; while using different random numbers at the top and base of the backbone will cause the fault’s dip to vary. Fault geometries are preserved by using a long correlation length along the horizontal axis and a monotone function (i.e. no inversion of fault curvature) in the Z-axis. After simulation, a postprocessing geometrical optimization is used to ensure each realization obeys the prior fault-horizon constraints. Stratigraphic grids are deformed by linking the stochastic horizon surfaces to their equivalent grid layers. Common applications of this method include assessing uncertainty on estimates of gross rock volume (GRV); optimizing well locations and using structure as an uncertain parameter in history matching. In this paper, we aim to assess uncertainty on estimates of connected volume and forecast production to fault position and displacement. Modelling uncertainty on fault displacement is achieved by simulating independent random variables (P-fields) on the hanging wall and footwall horizon surfaces. In order to ensure that faults do not reverse their sense of slip, uncertainty on either side of the fault is limited to a maximum of half the initial displacement. If both surfaces move in opposite directions, the maximum possible displacement is twice the initial value, and the

285

minimum possible displacement is zero. A long correlation length can be used to simulate variations in ‘global’ fault displacement arising from seismic interpretation uncertainty. Local heterogeneities in fault zone structure (drag, relays, segmentation) can be modelled using a shorter range property; or the two properties may both be modelled.

Fault property uncertainties Having defined a method for simulating geometrical uncertainties, variability in fault properties must be considered. As noted above, faults in reservoirs are rarely sampled by wells; hence their properties must be estimated using available parameters. We use an approach similar to that proposed by Manzocchi et al. (1999), whereby we estimate fault rock thickness from fault displacement and fault permeability from clay content, which is itself approximated using conceptual models such as SGR and/or SSF. To simulate multiple realizations of fault properties, we define the coefficients and constants used in the empirical functions as uncertain variables, whose distributions are estimated from analysis of analogue databases. Here we describe in turn the data available on fault rock permeability and thickness, the data analyses undertaken and the property modelling process.

Fault permeability data The permeability data used in this study are from small faults in North Sea Middle Jurassic reservoirs (Fisher & Knipe 2001; Sperrevik et al. 2002), supplemented with core and outcrop data from various basins (Gibson 1998) and additional unpublished core data. The data of Gibson (1998) comprise 32 outcrop and core samples from sandstones in various depositional and tectonic settings. Faults are classified according to type (normal, reverse or strike slip) and displacement magnitude: micro (,10 cm); meso (10 cm-10 m) and macro (.10 m). From microstructural examination, fault rock type is further divided into clay-matrix gouge zones and deformation bands (cataclastic, solution or complex deformation bands). Permeability is measured perpendicular to the fault surface using Klinkenberg-corrected gas permeametry under a confining pressure of 20.7 MPa. There is no apparent systematic difference between permeability values measured at different scales; however there are only 2 data points for faults with throw greater than 10 m. Nine samples were removed from the database as either host rock clay content or fault permeability was not recorded, leaving 23 samples for further analysis, of which 12 have depth data recorded. For all other data in this study, permeability is measured perpendicular to the fault surface using

286

A. D. IRVING ET AL.

water permeametry (minimum measured value c. 0.0001 mD) under a confining pressure of 0.8 MPa. Further details of the experimental procedure and interpretation are given in Fisher & Knipe (1998, 2001) and Sperrevik et al. (2002). For a total of 193 samples, data are recorded for depth, host rock clay volume and host and fault rock porosity, permeability and capillary entry pressure. Host rock clay volume (Vh) is estimated from microscopic analysis of thin sections. As displacement is small, fault rock clay content (Vf) is usually taken to equal Vh, except for clay smears which are assumed to have Vf ¼ 60%. Samples are classified based on microstructural analysis into cataclasites, clay smears and phyllosilicate framework fault rocks. Thirty-one samples described as disaggregation zones, which lack discrete slip surfaces and instead represent deformation by grain reorganization, are excluded as we consider them to be unrepresentative of the fluid flow properties of the reservoir scale fault rocks we wish to estimate. This leaves 162 samples for further analysis, making a total of 185 samples with the data of Gibson (1998) (Fig. 1). An exponential function is applied to all samples in order to compare measured permeability values at a reference (zero) effective stress, assuming hydrostatic compaction and relatively low stress sensitivity to permeability (David et al. 1994). Reference permeability is hereafter referred to as log ko. The same function can subsequently be used to correct reference values to in situ stress conditions for a reservoir of interest.

zone of more or less deformed material (fault rock) bounded by two or more discrete slip surfaces, or the thickness of the slip surface itself where no fault rock is developed. Fault displacement and thickness data are available for 933 samples of fault rock ranging from mm-scale microfaults to crustal scale strike slip faults with several hundred kilometres offset. Consistent with published observations of displacement and thickness (e.g. Hull 1988; Knott et al. 1996; Sperrevik et al. 2002) we observe an approximately power-law relationship with a scaling exponent close to 1 over this size range (Fig. 2), which is interpreted to reflect the processes (abrasion, grain fragmentation) responsible for fault rock development with progressive displacement. A linear regression between log throw (D) and log thickness (tf) for all samples gives the following expression: Log tf ¼ c Log D þ e

ð1Þ

where c ¼ 0.866, standard error ¼ 0.014; e ¼ 21.62, standard error ¼ 0.02.

Data analysis

To estimate the effects of faults on reservoir fluid flow, we are primarily concerned by the central

Through deformation-induced mixing (Fisher & Knipe 2001), fault permeability in siliciclastic rocks is commonly believed to be negatively correlated with clay content. This is the basis for the empirical transformations of Manzocchi et al. (1999) and Sperrevik et al. (2002) and is borne out by the experimental data of Crawford et al. (2002). Permeability reduction during faulting in impure sandstones occurs by grain rotation and repacking, hence platy clay minerals are more likely to develop interconnected microtextures that reduce permeability. Fault permeability may

Fig. 1. Log permeability v. clay content for 185 fault samples, showing the ‘base case’ regression (blue curve); simulated functions (shaded area) and high/low functions for sensitivity analysis (green/red curves).

Fig. 2. Logs thickness v. log displacement for 933 fault rock samples, showing the ‘base case’ regression (blue curve); simulated functions (shaded area) and high/low functions for sensitivity analysis (green/red curves).

Fault zone thickness


additionally show a negative relationship with maximum burial depth due to the effects of mechanical and thermal compaction, growth of diagenetic minerals and enhanced quartz precipitation (e.g. Fisher & Knipe 2001). A linear regression between log permeability and Vf for all samples (n ¼ 185) gives the following expression: Log ko ¼ aVf þ b

ð2Þ

where Log ko ¼ log permeability at reference pressure Po; a ¼ 20.046, standard error ¼ 0.005; b ¼ 21, standard error ¼ 0.15. We assume that there is no further permeability decrease at values of Vf greater than 60%; this approximates the form of functions derived from laboratory deformation experiments where clay content can be more accurately controlled (e.g. Crawford et al. 2002) and prevents excessively low (,1025 mD) permeability values from being output (Fig. 1). For the data with a depth measurement recorded, a weak negative correlation is observed with fault permeability. However, this adds little predictive power to the existing regression and it is unknown if samples share a common depth datum. Furthermore, including a second term in the permeability equation introduces additional complications later in the modelling process; hence only clay content is kept as a significant parameter. This equation can be used to compute fault permeability from a value of Vf estimated using a proxy property such as shale gouge ratio (Yielding et al. 1997). This is typically performed on a single realization of the reservoir model, using a single SGR-fault permeability function. Instead of regarding this function as fixed, we define the slope and intercept as uncertain variables, which take Gaussian distributions defined by the mean and standard error of a and b respectively. We assume that the input data are representative of the (larger) faults we wish to model and that a geometric average is appropriate to upscale fault permeability from core to reservoir grid scale. Use of a single permeability function throughout the model implies that all the faults developed under similar physicochemical conditions; multiple functions could be used for reservoirs thought to have undergone several deformation events. Several authors have proposed a relationship between fault displacement and permeability. Antonellini & Aydin (1994) reported that deformation bands in aeolian sandstone displayed a large (1–2 orders of magnitude) permeability decrease associated with the development of discrete slip surfaces. As this occurs at a displacement of c. 1 m, faults smaller than this are unlikely to be

287

included in earth models, however this may in part account for the large scatter seen in permeability values for small faults in clean sands. Crawford et al. (2002) noted a 1–2 orders of magnitude permeability decrease in sheared mixtures of quartz and kaolinite, particularly for Vclay ,40%. Although total experimental slip displacement is only 3.5 mm, maximum shear strain is c. 20; hence the authors argue that the gouge fabric developed is comparable to that found in natural seismic scale faults. Most of the permeability decrease takes place during initial displacement (shear strain 1). Small faults in the datasets described above may accommodate similar strains, although the data needed to estimate this are unavailable. In response to applied shear strains of up to c. 10, Morrow et al. (1984) reported 1 order of magnitude permeability decrease for some fault gouge samples but little change in others. If absolute fault permeability values may be comparable between the core and reservoir grid block scales, a difference in distribution is likely due to the effect of sample support, whereby measured variability decreases with increasing sample size. Core plug permeability measurements, as shown in Figure 1, should be inherently more variable than ‘bulk’ permeability values estimated from, for example, well tests. Upscaling statistics from one scale to the other requires knowledge of the spatial variability of fault permeability; however any such information is highly speculative. Anecdotal outcrop evidence (Antonellini & Aydin 1994; Faulkner & Rutter 1998; Foxford et al. 1998) and the results of modelling studies (Fairley et al. 2003; Zhurina 2003) suggest that fault permeability has a correlation length (variogram range) that is larger than a core plug (c. 10 cm) but smaller than a typical simulation grid block (50 –100 m). To make this assumption obviates the need to model fault permeability using a correlated random field, however this could be easily achieved using the workflow outlined below. Even if no explicit correlation model is specified, as SGR is a simple average of laterally distributed clay content over a vertical distance equal to fault throw, it already contains implied vertical (¼ throw) and lateral (¼ Vclay property) correlation structures. This further supports the method of defining uncertainties on the slope and intercept of the SGR-permeability function rather than modelling the distribution of residual values. Upscaling of the mean and standard error of the SGR-permeability function may be required depending on the assumed flow paths (and arrangement of heterogeneities) within the fault zone as being either random or ‘in parallel’, necessitating the use of the geometric (Sperrevik et al. 2002) or arithmetic (Manzocchi et al. 1999) average respectively.

288

A. D. IRVING ET AL.

We assume that the geometric average is appropriate, hence Equation 2 can be used directly in the reservoir grid, however an arithmetic (or even harmonic) average could be used instead.

Stochastic model setup The process of simulating multiple realizations is based on a conventional stochastic geological property modelling workflow normally used to generate parameters such as facies, porosity and permeability (Caers 2005). Each parameter can be modelled as a constant; sampled randomly from a distribution; interpolated or simulated using various geostatistical algorithms, or defined as a function of another property. In this workflow, we use the values from the regression analyses above to define probability distributions for the coefficients and constants in the equations used to estimate fault permeability and thickness. Functions are defined to compute various desired parameters such as fault throw, SGR, SSF, fault permeability and thickness. Finally, instead of representing faults using transmissibility multipliers, we compute a thicknessweighted harmonic average permeability (K*), which is stored in cells adjacent to the fault surface: Dx,y K ¼ [(Dx,y tf )=Kx,y ] þ [tf =Kf ]

ð3Þ

where Dx,y ¼ cell size in x or y grid axis respectively; tf ¼ fault thickness; Kx,y ¼ grid cell permeability in x or y grid axis respectively; Kf ¼ fault permeability. Inclusion of this average permeability in a flow simulation model is equivalent to applying a multiplier to the permeability of the upstream grid cell (Manzocchi et al. 1999, p. 57, Eq. 7; p. 59, Fig. 7j). While a less precise solution, it avoids the need to compute non-neighbour connections for each geometrical realization, which can be excessively time-consuming. Furthermore, the averaged permeability captures the effects of both

sedimentary and fault heterogeneities, so can be used to make rapid estimates of connected hydrocarbon volumes through application of a threshold value. Comparative simulations show that such an averaged permeability property gives a very similar flow simulation response to transmissibility multipliers; however care is required in models where well perforations are close to or within faulted cells. The workflow is shown schematically in Figure 3 and summarized as follows: (1) (2) (3) (4) (5) (6) (7)

Define input PDFs. Simulate grid geometry; compute fault displacement and SGR. Simulate required uncertain properties; compute SSF, fault permeability and thickness. Compute average permeability, apply threshold, estimate connected volume. Repeat steps 2–4 until desired number of realizations attained. Export realizations for fluid flow simulation. Perform sensitivity analysis using experimental design.

Case-study The workflow described above has been tested on a producing field located in the Brent province of the UK Northern North Sea. The reservoir comprises a horst block structure with dip closure at its southern margin, with a central fault extending the length of the structure and several smaller faults (denoted ‘north’, ‘east’ and ‘west’) which truncate against the bounding faults (Fig. 4). Juxtaposition against downthrown reservoir sands defines an apparent hydrocarbon spill point to the north. Gas and condensate are mainly accumulated in sands of the Tarbert and Balta formations, with lesser accumulations in the Etive and upper Ness formations. The Etive formation is interpreted as upper shoreface deposits prograding to the North. A

Fig. 3. Schematic flowchart illustrating the workflow used to generate stochastic realizations of fault geometry and properties.


Fig. 4. Top structure depth map showing the bounding fault surfaces; traces of the central, north, east and west faults; the single producer well (red line) and the initial gas–water contact (blue contour). Line X– X0 indicates the position of the seismic line shown in Figure 6.

transition from the upper Etive to Ness is characterized by tidal and shoreface deposits mixed with bay-lagoon facies (shales and coal). The Ness lithostratigraphical unit represents the deposits of a delta plain environment of floodplain shales, channels (meandering, braided and distributary) and crevasse splays. The Tarbert formation is composed of tidal and shoreface deposits overlain by transgressive Balta sands. Together, the Balta and uppermost Tarbert make up c. 30 m of reservoir sand, with a further 20 m in the lower Tarbert, separated by c. 20 m of non-reservoir shale or siltstone. A displacement of c. 30 m is required to completely offset the upper reservoir units. While these layers would then be partly juxtaposed against the lower Tarbert sands, the intervening shale could conceivably form a baffle or barrier to fluid flow through smearing and/or formation of low permeability clay gouge (Fig. 5). An almost complete Balta-Tarbert succession was cored near the field crest, permitting not only sedimentological and petrophysical studies but also the identification of a number of small deformation features similar to those found within the analogue database described above. Core taken from a well drilled in an adjacent fault block also contained numerous small faults, which were additionally interpreted on borehole image logs. Fault and horizon surfaces interpreted from seismic data were used to build a stratigraphic grid with c. 1 30 000 cells (c. 92 000 active in the flow

Fig. 5. Stratigraphic sequence summarizing the lithology, petrophysical data and interpreted vertical permeability barriers (red lines).

289

290

A. D. IRVING ET AL.

simulator). Average cell sizes are c. 50 m in the I and J directions and 3 to 20 m vertically. A combination of object modelling, sequential indicator simulation and deterministic interpretation was used to model 15 sedimentary facies. Porosity, permeability, Vclay and water saturation were modelled per facies using sequential Gaussian simulation, with NTG being defined from a cut-off applied to porosity. Enhanced diagenesis with depth observed in regional wells was modelled using decline trends applied to both porosity and permeability. Vertical permeability was modelled using different kv/kh transformations per facies. The different modelling steps were combined in a property modelling workflow to generate multiple realizations, with one selected as the operational model for use in numerical fluid flow simulation and reserves estimation. The model was used in a conventional black oil numerical flow simulator without modification, save for defining a number of vertical permeability barriers interpreted from well tests (Fig. 5). Functions were also defined for relative permeability and vertical well flow performance. At the time of this study, the field had approximately two years of production history, with a single well producing gas and condensate by natural depletion. In the simulation model, the well was constrained by historical gas production rate, with bottom hole pressure (BHP) being the main history matching parameter. Production forecasts were based on a gas rate target, with defined minima for export pressure and economic flow rate. In order to match the observed BHP trend, all intra-reservoir faults were modelled as sealing to fluid flow, except for the northern part of the central fault, which was assigned a constant low

transmissibility multiplier. Production forecasts using such a model have implications for a possible infill well in the second panel. The goals of applying a structural uncertainty workflow to this field were: (1) to assess the model’s sensitivity to different parameters; and (2) to search for a match to production history using a geologically plausible combination of parameters.

Geometrical uncertainties Due to the depth of the reservoir (.3500 m), vertical seismic resolution is limited to c. 30 m. Fault offsets are interpreted at the top Dunlin (base reservoir) level and on a strong positive reflector which corresponds to a coal layer near the top of the Lower Tarbert succession. The overlying Base Cretaceous Unconformity (BCU) reflector is also interpreted throughout, although most faults truncate against it (Fig. 6). Although considerable uncertainty exists in the various seismic processing and interpretation stages described above (notably migration and depth conversion), horizons are regarded as ‘fixed’ in order to focus on the influence of fault lateral position and throw. Furthermore, the lateral position of internal faults is also fixed, as shifting these surfaces may cause changes in layer juxtapositions, rendering impossible the separation of fault position and fault offset effects on connected volumes and forecast production. Uncertainty on lateral position was interpreted for the bounding faults from the width of the zone of seismic noise within which each fault could reasonably be picked (e.g. Fig. 6). The standard deviation for P-field simulation was defined as one half of this width in either direction. Maximum

Fig. 6. Seismic section (along line X–X0 shown in Fig. 4), showing the Base Cretaceous Unconformity (BCU), Top Balta and Top Dunlin horizons. ‘Base case’ (blue); high (green) and low (red) positions are shown for the Top Balta and bounding fault surfaces.


fault throw uncertainty on either side of the fault was set to half the initial displacement. If both surfaces move in opposite directions, the maximum possible displacement is twice the initial value, and the minimum possible displacement is zero. To better assess the relative importance of fault position and throw, three geometrical simulations were run: one in which only fault throw was simulated; one in which only fault position was varied, and one in which both fault throw and fault position were varied. In each case, 200 realizations were simulated, making a total of 600 possible reservoir grid geometries.

Property uncertainties In order to focus on the effects of fault uncertainties, grid parameters such as NTG, porosity, permeability and Sw were regarded as ‘fixed’. Uncertainties were defined on the slope and intercept of the equations used to estimate fault permeability and thickness (‘KfSlope’; ‘KfIntercept’; ‘tfSlope’; and ‘tfIntercept’) according to the regression analyses described above (Eqs 1 & 2). In order to account for the possibility of clay smear as a sealing mechanism, uncertainties were defined on two additional parameters. The value of Vclay above which cells are interpreted as being shale (‘ShaleCutoff’) was defined as a Gaussian variable with a mean of 0.5 and standard deviation of 0.08. The value of shale smear factor (SSF) above which clay smears become discontinuous (‘SmearContinuity’) was defined as a uniform variable with a minimum of 0 and maximum of 7. All faulted cells containing shale smears were assigned a Vclay value of 60%, giving a minimum fault permeability value of between 1023 and 1025 mD depending on the simulated values for KfSlope and KfIntercept (Fig. 1). To analyse the relative importance of different geometrical and property uncertainties, three parameter sets were defined for each set of geometrical realizations; one in which the only uncertainty was geometrical (sets 1, 4 and 7); one with additional uncertainty on fault permeability and thickness (sets 2, 5 and 8), and one containing uncertain clay smears (sets 3, 6 and 9). In total, 1800 realizations were simulated in this way: (1) Fault throw variable, n ¼ 200. (2) Fault throw, permeability, thickness variable, n ¼ 200. (3) Fault throw, permeability, thickness, clay smears variable, n ¼ 200. (4) Fault position variable, n ¼ 200. (5) Fault position, permeability, thickness variable, n ¼ 200. (6) Fault position, permeability, thickness, clay smears variable, n ¼ 200.

291

Fault throw, position variable, n ¼ 200. Fault throw, position, permeability, thickness variable, n ¼ 200. (9) Fault throw, position, permeability, thickness, clay smears variable, n ¼ 200. The final parameter to be defined is the threshold value applied to the average permeability K* in order to estimate connected volumes. As with other input parameters, this may be a constant, sampled from a distribution, interpolated, simulated or derived from another property. In this case, a constant isotropic threshold of 1 mD was used; any cells with a lower K* value are assumed to be disconnected from the producer well. This coincides with the general 1 mD cut-off used when modelling the effects of diagenesis on grid permeability with increasing burial depth. However, it is an arbitrary value that reflects the particular grid and fault permeabilities, fault thickness and cell sizes in this model; a similar value would not necessarily be appropriate for different cases. Moreover, it takes no account that the predominant fluid type is gas; which is likely to remain mobile at cell permeability values of less than 1 mD. This should be borne in mind when comparing estimated connected volumes with production forecasts based on numerical flow simulation. (7) (8)

Results The main heterogeneity is the central fault that divides the gas-bearing portion of the model into two panels; grid permeabilities generally exceed 1 mD and do not vary between realizations. The volume of gas connected to the single producer well may comprise only the eastern panel or, depending on the central fault’s throw and permeability, both panels. The gross rock volume (GRV) of these panels is modified in parameter sets 4– 9 by varying the position of the bounding faults. Additional complexity is introduced by the vertical permeability barriers shown in Figure 5, which separate the perforated zones of the well from gas accumulations in the stratigraphically lower Ness and Etive formations, as well as limiting the contribution from the western panel for different geometrical and fault property combinations. A control model was defined with ‘base case’ geometry; fault properties estimated using the mean values (slope and intercept) from the data analysis and containing no clay smears. All reported volumes are normalized as a ratio of the total forecast gas production (FGPT) for this model, obtained from numerical flow simulation. Figure 7 shows plots of computed connected gas volume (CGIIP) against realization number; summary statistics on CGIIP for each of the nine parameter sets are given in Table 1.

292

A. D. IRVING ET AL.

Fig. 7. Connected gas volume per realization for (a) parameter sets 1 (fault throw uncertainty only); 2 (throw þ fault permeability & thickness) and 3 (throw þ fault permeability & thickness þ clay smears); (b) parameter sets 4 (fault position uncertainty only); 5 (position þ fault permeability & thickness) and 6 (position þ fault permeability & thickness þ clay smears) and (c) parameter sets 7 (throw & position uncertainties only); 8 (throw & position þ fault permeability & thickness) and 9 (throw & position þ fault permeability & thickness þ clay smears).

The differences in CGIIP between the first three parameter sets are slight, suggesting that fault throw is the most influential variable for these models (Fig. 7a). For parameter set 4, where fault throw is fixed but the position of the bounding faults may vary, none of the realizations connect across to the western panel, hence variation is small (Fig. 7b).

A greater spread of results is observed when fault permeability, thickness and clay smears are considered as uncertain parameters (sets 5 and 6), in which case about a quarter of the models connect to both panels. Varying both fault position and throw (sets 7–9) produces a result that compounds the two parameters’ individual effects (Fig. 7c).

Table 1. Summary statistics on connected gas in place for the nine simulated parameter sets CGIIP

Mean St Dev Min Max

Parameter set 1

2

3

4

5

6

7

8

9

0.9 0.28 0.63 1.41

0.96 0.3 0.63 1.48

0.94 0.3 0.63 1.48

0.61 0.05 0.54 0.72

0.75 0.27 0.54 1.4

0.73 0.23 0.54 1.4

0.95 0.3 0.54 1.44

0.99 0.31 0.54 1.51

0.97 0.31 0.54 1.45


Fig. 8. Cumulative distribution of connected gas volume for parameter sets 1 (fault throw uncertainty only); 2 (throw þ fault permeability & thickness); 3 (throw þ fault permeability & thickness þ clay smears); 4 (fault position uncertainty only); 5 (position þ fault permeability & thickness); 6 (position þ fault permeability & thickness þ clay smears); 7 (throw & position uncertainties only); 8 (throw & position þ fault permeability & thickness) and 9 (throw & position þ fault permeability & thickness þ clay smears). Blue dots indicate quantiles selected from parameter set 9 for fluid flow simulation.

For all geometries, adding shale smears (blue curves) is seen to reduce the mean connected gas volume, but has a limited effect on standard deviation. Figure 8 summarizes these observations in the form of cumulative distributions of connected volume for each parameter set.

Numerical simulations To investigate the model’s dynamic sensitivity to fault heterogeneities, and to assess the usefulness of connected gas volume as a predictor of forecast gas production, realizations were sampled from parameter set 9, which includes uncertainty on fault throw and position, fault permeability, fault thickness and clay smears. Realizations were exported per decile of connected volume from Q0 to Q100, making a total of 11 simulations (Fig. 8). Numerical fluid flow simulations were performed on the exported realizations with the same settings as for the operational model. From a crossplot of the two responses (Fig. 9, blue crosses), it appears that CGIIP is, at best, an approximate estimator of forecast gas production. Small connected gas volumes tend to underestimate total gas production, and large connected gas volumes overestimate total gas production. This can be understood by imagining a realization with only a single cell above the 1 mD threshold. Estimated connected volume will be high (both panels),

293

Fig. 9. Connected gas volume against forecast total gas production for the ‘base case’ (blue diamond); 11 quantiles sampled from parameter set 9 (blue dots) and 32 planned experiments used in the sensitivity analysis (orange dots).

while across-fault flow rate will be comparatively low. By contrast, a realization with many cells just below the 1 mD threshold would have a lower connected volume (eastern panel only) but a much greater simulated across fault flow rate. A plot of P/Z, (pressure divided by gas compressibility factor), against cumulative production gives an estimate of connected gas in place and ultimate recovery for reservoirs producing by natural depletion. This is shown in Figure 10, (here computed using mostly flowing, rather than shut-in, BHP values). The observed data (black crosses) plot towards the lower end of the range of simulated realizations (shaded area); the model computed using the ‘base case’ geometry and properties (blue curve), which lies at the centre of this range, overestimates true connected gas in place by approximately one quarter.

Sensitivity analysis To more rigorously analyse the model’s sensitivity to individual parameters, a two-level fractional factorial experimental design was employed. The eight uncertain parameter values were systematically varied between ‘low’ and ‘high’ levels in a planned exploration of the uncertainty space. A full factorial design, evaluating all possible combinations, would require 28 ¼ 256 simulations. In order to reduce the computational time, a resolution IV screening design was selected with the generating fraction (in Box notation) I ¼ ABCF ¼ ABDG ¼ ACDEH, requiring 32 experiments. This allowed estimation of the effects of all eight principal factors and thirteen two-factor interactions. Due to the independent simulation of fault throw and fault position used to generate grid geometries,

294

A. D. IRVING ET AL.

Fig. 10. P/Z against total gas production for observed data (black crosses); traditional history matched model (black curve); ‘base case’ (blue curve); simulated quantiles (shaded area) and maximum/minimum cases from the sensitivity analysis (green/red curves). The orange curve represents the best model from the sensitivity analysis, showing the match to the later pressure build-up.

it was not possible to sample realizations directly from parameter set 7 for use in the designed experiments. Instead, geometrical simulations of fault position were first rerun for two input grids with ‘high’ and ‘low’ throw on the central fault. From each of these models, two realizations were selected to represent ‘high’ and ‘low’ GRV, resulting in four discrete grids representing the ‘end member’ geometries. Parameter values for the six fault property factors were taken from the minimum and maximum values output from the stochastic modelling stage of the workflow; these are listed in Table 2. Three responses were chosen to estimate sensitivity coefficients for different parameters: CGIIP (estimated from the permeability cut-off prior to Table 2. Low, high and ‘base case’ factor input values for the sensitivity analysis

Throw (m) nGRV ShaleCutoff SmearContinuity KfSlope KfIntercept tfSlope tfIntercept

Low

High

Base

48 0.89 0.25 7 20.07 21.33 0.9 21.55

24 1.18 0.6 1 20.03 20.65 0.83 21.68

38 1 0.5 4 20.05 21 0.87 21.62

simulation); FGPT and simulated well bottom hole pressure (BHP) after two years of production. The latter response is the main history matching parameter for the operational model. All responses were normalized: CGIIP and FGPT to the ‘base case’ model total production and BHP to the base case model pressure after two years. For each response, sensitivities are estimated using a linear approximation of the form: y ¼ a0 þ a1 x1 þ a2 x2 þ a3 x3 þ þ an xn

ð4Þ

where y ¼ response; a0 ¼ constant; a12n ¼ coefficients 1 to n; x12n ¼ factors 1 to n (and their interactions). The calculated sensitivity coefficients for all single factors and significant (at the 95% confidence interval) two-factor interactions are given in Figure 11; positive sensitivities (factor increase leads to response increase) are shown in blue and negative sensitivities (factor increase leads to response decrease) in red. For CGIIP, the most sensitive parameter is fault throw, followed by gross rock volume (¼ fault position), fault permeability slope, fault permeability intercept and various two factor interactions (Fig. 11a). A similar rank of sensitivities is observed for BHP after 2 years of production, which is also sensitive to individual clay smear parameters


295

in this case. A model with lower average SGR would likely be more sensitive to the value of fault permeability intercept. Flowing P/Z against FGPT for the maximum and minimum experiments is shown in Figure 10 (green and red curves respectively). As expected from a systematic, planned exploration of the uncertainty space, there is a greater spread of values than for the realizations sampled from the CDF of connected volume (shaded area). Four of the models give a pressure response within a few percent of the observed data. These models all have the ‘high throw’ geometry, while the closest model with the ‘low throw’ geometry predicts a BHP 25% greater than observed. The four best-matched models also share the low value for fault permeability slope but have variable values of fault permeability intercept. Three of the models have the ‘large GRV’ structure. Forecast total gas production for the four models varies from 71 to 95% of the forecast total production for the ‘base case’ model, with three and a half years’ difference in predicted field life between the smallest and largest of the four models. Subsequently, additional pressure data were obtained during a shutdown of the field for scheduled maintenance. BHP was seen to increase at a constant rate for two months, indicating probable gas flow through the central fault from the western panel. Only one of the four best models is able to correctly reproduce the observed pressure build-up (Fig. 10, orange curve). This result helps to constrain fault properties in this field and in adjacent untested fault blocks, and to evaluate a possible infill well in the western fault panel.

Discussion Model sensitivity to fault throw and position

Fig. 11. Estimated sensitivity coefficients for (a) connected gas in place; (b) BHP after 730 days of production and (c) forecast total gas production; blue bars indicate positive correlation between factor and response, red bars indicate negative correlation.

(Fig. 11b). GRV is the most influential factor for FGPT (Fig. 11c), which reflects both the economic gas rate limit used to stop the simulation runs and the flow of gas around and through faults late in simulated field life. The greater sensitivity to fault permeability slope can be explained by the relatively high SGR values (average c. 35%) occurring

In this study, we have only altered the position of the boundary faults, as shifting internal reservoir faults causes changes in layer juxtapositions, particularly for dipping horizons and shallow-dipping faults, as in this case. This limits the uncertainty on the size of the first panel and removes some interactions from the model, for example, a small panel with less sealing central fault might give a similar result to a larger panel with a more sealing fault. In this case, we have taken this decision to reduce ambiguity in interpreting sensitivities; however it could be argued that changes in juxtapositions are not only acceptable but desirable, if they preserve realistic geometries that could be interpreted from seismic data. For a study in which the objective was not to separate the effects of individual geometrical parameters but simply to assess the overall uncertainty on connected volumes and

296

A. D. IRVING ET AL.

reserves, the position of internal faults could be varied along with that of the boundary faults.

Fault permeability computation Although all permeability samples were corrected to zero effective stress prior to data analysis, no correction was applied to estimated grid fault permeability values. Under initial reservoir conditions, the maximum reduction in permeability would be less than half an order of magnitude, however this effect would likely increase as the reservoir pressure declines during production. While initial permeabilities could be simply modified to reflect the effective stress conditions, these would require to be updated several times during flow simulation to reflect the decreasing reservoir fluid pressure. Alternatively, this behaviour could be approximated using simulator keywords for rock compressibility applied to the region of faulted grid cells. An overall decrease in fault permeability would shift the simulated production and BHP profiles downwards and increase the number of models able to match production history. All fault permeabilities used in this study are absolute single-phase values applied to all fluid phases in the simulator. Two-phase properties could be modelled using pseudo relative permeability functions (Manzocchi et al. 2002) modified to work with averaged grid permeabilities rather than transmissibility multipliers. If the central fault zone were water-filled (Rivenæs & Dart 2002), relative permeability to gas would be far lower than the single phase permeabilities used in this case. Even if it were mainly saturated with hydrocarbons, end point relative gas permeability for the fault zone material might be two orders of magnitude lower than that of the reservoir material (Fisher 2005). For a synthetic model with a similar configuration to this case, Al-Busafi et al. (2005) observed an interaction between fault thickness and sensitivity to multiphase properties. For a constant fault thickness of 0.3 ft, single phase and multiphase fault properties gave a similar cumulative gas production, but results were very different for a 3 ft thick fault zone. In this case, estimated thickness for the central fault varies over a similar range (c. 0.1 to 1 m), hence for higher displacement, thicker parts of the fault, modelling multiphase fault properties might cause a difference in forecast gas production.

Geometrical uncertainties As noted above, the seismic quality leaves considerable ambiguity in fault and horizon interpretation. As well as stochastically varying fault position and throw, alternative structural interpretations are

possible. Even with the existing fault pattern, additional near-seismic resolution faults could be included in the grid. Either of these methods would require building additional parallel models and repeating the study. Addition of internal faults to the existing model would likely reduce overall recovery factors and connected volumes, even for faults with relatively high permeability, low thickness and discontinuous clay smears. Faults with a displacement below seismic resolution could be modelled using stochastic object-based (Boolean) simulation, with their flow effects being captured as modifiers to cell permeability within the existing grid geometry. Modelling of discrete fault offsets in this manner would be more complex and would probably require stochastic simulation of triangulated surfaces. For the studied field case, we do not think that sub-seismic faults would have a significant impact on reservoir fluid flow, although they might be important in other situations. In this study we have purposely regarded the overall position of the input horizons as fixed in order to focus on fault-related parameters. In reality, with control from only a single well penetration there is obviously considerable uncertainty arising from seismic processing, depth conversion and interpretation. A more complete assessment of uncertainty could be achieved by including horizon position as a variable in the geometrical simulations.

Connected volume estimation One of the main drawbacks of the current method is the use of a cut-off on averaged permeability to estimate connected hydrocarbon volumes, as this has been shown to be inaccurate in this case (e.g. Fig. 9 and accompanying discussion). An improved connectivity estimate would take into account the area and properties of the juxtaposed fault ‘windows’, as well as the distance from in place volumes to injector or producer wells. However, the method must also be sufficiently fast to compute that it can be applied to all realizations generated using the workflow. An acceptable compromise between speed and accuracy might be found with streamline simulation, where pressure is computed in the three dimensional grid, as per conventional finite difference reservoir simulation, but the fluid flow equations are solved instead on one dimensional streamlines. To estimate connected volumes, it would not be necessary to run a full forward simulation if there were no planned changes to the well pattern or operating conditions (i.e. rate and/or pressure constraints). Instead, a cut-off could be applied on the initial streamline pattern using a combination of drainage time to producer wells and


time-of-flight from any injector wells. Values for these cut-offs would need to be chosen in the light of reservoir engineering experience and tuned to the results of full-field simulation for the particular case. The option of using streamlines in this way is attractive as the geomodelling tool used to implement our workflow already contains a link to a commercial streamline simulator. An improved estimation of connected volumes would give a more accurate probability distribution from which to sample realizations for use in full dynamic simulations and would help to better quantify the effect of heterogeneities on forecast production. This is the subject of ongoing research and development.

Integrated uncertainty modelling Finally, as the original purpose of the stochastic modelling tool is to generate multiple realizations of grid properties, these can easily be included in our workflow. This would allow investigation of the relative sensitivities of the model to sedimentary uncertainties (facies, Vclay, porosity, NTG, Sw etc.) and of their interactions with fault properties. Such an integrated approach can also be used to generate ensembles of realizations incorporating fault properties for use in assisted history matching schemes.

Conclusions We have outlined a workflow for modelling multiple realizations of fault geometry and properties using 3D geomodelling software. The workflow has been illustrated using a producing North Sea field in which structural uncertainties are believed to have an effect on connected volumes. The main conclusions are as follows: (1)

(2)

(3)

Connected volume estimated from a cut-off on averaged permeability provides an approximate prediction of recoverable reserves. Accuracy could be improved by taking account of time/distance to the producer well, for example, using streamline simulation. For connected volume and well bottom hole pressure after two years, fault throw and fault permeability are the most important factors. For forecast total production, fault position and fault permeability are the most important factors. This difference is due to the minimum economic rate constraint applied to the producer well and to the flow of gas through and around low permeability faults late in field life. The slope of the function relating fault permeability to clay content is more important

(4)

(5)

297

than the intercept, due to the relatively high average SGR values in this model. Observed well pressures are at the lower end of the range of values predicted by probabilistic modelling. A better match to the observed data could be obtained by modifying fault permeabilities as a function of effective stress or fluid saturation. A planned exploration of the uncertainty space using experimental design covers a wider range of possible pressure profiles. Four models are able to match the observed pressure after two years, but only one correctly reproduces a subsequent pressure build-up. This constrains fault properties and can be used to evaluate a possible infill well.

This work was performed as part of A. Irving’s PhD study at University College Dublin. Total E&P UK is thanked for sponsorship and for permission to publish this work. Don Medwedeff and an anonymous reviewer are thanked for their thoughtful comments, which considerably improved the paper.

References Aarland, R. K. & Skjerven, J. 1998. Fault and fracture characteristics of a major fault zone in the northern North Sea: analysis of 3D seismic and oriented cores in the Brage Field (Block 31/4). In: Coward, M. P., Daltaban, T. S. & Johnson, H. (eds) Structural Geology in Reservoir Characterisation. Geological Society, London, Special Publications, 127, 209–229. Al-Busafi, B., Fisher, Q. J. & Harris, 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. Allan, U. S. 1989. Model for hydrocarbon migration and entrapment within faulted structures. American Association of Petroleum Geologists Bulletin, 73, 803– 811. Antonellini, M. & Aydin, A. 1994. Effect of faulting on fluid flow in porous sandstones: petrophysical properties. American Association of Petroleum Geologists Bulletin, 78, 355 –377. Bentley, M. R. & Barry, J. J. 1991. Representation of fault sealing in a reservoir simulation: Cormorant Block IV, UK North Sea. Society of Petroleum Engineers, SPE 22667, 119 –126. Berg, R. & Avery, A. H. 1995. Sealing properties of tertiary growth faults, Texas Gulf Coast. American Association of Petroleum Geologists Bulletin, 79, 375– 393. Caers, J. 2005. Petroleum Geostatistics. Society of Petroleum Engineers, Richardson, Texas. Cerveny, K., Davies, R., Dudley, G., Fox, R., Kaufman, P., Knipe, R. J. & Krantz, R. 2004. Reducing uncertainty with fault-Seal analysis. Oilfield Review, Winter 2004, 38–51. Crawford, B. R., Myers, R. D., Woronow, A., Faulkner, D. R. & Rutter, E. H. 2002. Porositypermeability relationships in clay-bearing fault gouge.

298

A. D. IRVING ET AL.

Society of Petroleum Engineers, SPE/ISRM 78214, 1–13. Damsleth, E., Sangolt, V. & Aamodt, G. 1998. Subseismic faults can seriously affect fluid flow in the Njord Field off Western Norway – a stochastic fault modelling case-study. Society of Petroleum Engineers, SPE 49024, 295–304. David, C., Wong, T.-F., Zhu, W. & Zhang, J. 1994. Laboratory measurement of compaction induced permeability change in porous rocks: implications for the generation and maintenance of pore pressure excess in the crust. Pure and Applied Geophysics, 143, 425– 456. England, W. A. & Townsend, C. 1998. The effects of faulting on production from a shallow marine reservoir – a study of the relative importance of fault parameters. Society of Petroleum Engineers, SPE Paper 49023, 279– 294. Fairley, J., Heffner, J. & Hinds, J. 2003. Geostatistical evaluation of permeability in an active fault zone. Geophysical Research Letters, 30, 18, 1962, doi: 10.1029/2003GL018064. Faulkner, D. R. & Rutter, E. H. 1998. The gas permeability of clay-bearing fault gouge at 20 8C. In: Jones, G., Fisher, Q. J. & Knipe, R. J. (eds) Faulting, Fault Sealing & Fluid Flow in Hydrocarbon Reservoirs. Geological Society, London, Special Publications, 147, 147–156. Fisher, Q. J. 2005. Recent advances in Fault Seal Analysis as an Aid to Reservoir Characterisation and Production Simulation Modelling. Society of Petroleum Engineers, SPE 94460, 1– 8. Fisher, Q. J. & Knipe, R. J. 1998. Fault sealing processes in siliciclastic sediments. In: Jones, G., Fisher, Q. J. & Knipe, R. J. (eds) Faulting and Fault Sealing in Hydrocarbon Reservoirs. Geological Society, London, Special Publications, 147, 117–134. Fisher, Q. J. & Knipe, 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. Foxford, K. A., Walsh, J. J., Watterson, J., Garden, I. R., Guscott, S. C. & Burley, S. D. 1998. Structure and content of the Moab fault zone, Utah, USA, and its implications for fault seal prediction. In: Jones, G., Fisher, Q. J. & Knipe, R. J. (eds) Faulting and Fault Sealing in Hydrocarbon Reservoirs. Geological Society, London, Special Publications, 147, 87– 103. Gibson, R. G. 1994. Fault-Zone seals in Siliciclastic Strata of the Columbus Basin, Offshore Trinidad. American Association of Petroleum Geologists Bulletin, 78, 1372–1385. Gibson, R. G. 1998. Physical character and fluid-flow properties of sandstone-derived fault zones. In: Coward, M. P., Daltaban, T. S. & Johnson, H. (eds) Structural Geology in Reservoir Characterisation. Geological Society, London, Special Publications, 127, 83– 98. Harding, T. P. & Tuminas, A. C. 1989. Structural interpretation of hydrocarbon traps sealed by basement normal block faults at stable flank of foredeep basins and at rift basins. American Association of Petroleum Geologists Bulletin, 73, 812 –840.

Harris, D., Yielding, G., Levine, P., Maxwell, G., Rose, P. T. & Nell, P. 2002. Using Shale Gouge Ratio (SGR) to model faults as transmissibility barriers in reservoirs: an example from the Strathspey Field, North Sea. Petroleum Geoscience, 8, 167– 176. Hippler, S. J. 1997. Microstructures and Diagenesis in North Sea Fault Zones: implications for Fault-Seal potential and Fault-Migration rates. In: Surdam, R. C. (ed.) Seals, Traps and the Petroleum System. AAPG Memoir, Tulsa, 67, 85–101. Hull, J. 1988. Thickness– displacement relationships for deformation zones. Journal of Structural Geology, 10, 431– 435. Jev, B. I., Kaars-Sijpesteijn, C. H., Peters, M. P. A. M., Watts, N. L. & Wilkie, J. T. 1993. Akaso field, Nigeria: use of integrated 3D seismic, fault slicing, clay smearing, and RFT pressure data on fault trapping and dynamic leakage. American Association of Petroleum Geologists Bulletin, 77, 1389– 1404. Jolley, S. J., Dijk, H., Lamens, J. H., Fisher, Q. J., Manzocchi, T., Eikmans, H. & Huang, Y. 2007. Faulting and fault sealing in production simulation models: Brent Province, northern North Sea. Petroleum Geoscience, 13, 321 –340. Knai, T. A. & Knipe, R. J. 1998. The impact of faults on fluid flow in the Heidrun Field. In: Jones, G., Fisher, Q. J. & Knipe, R. J. (eds) Faulting and Fault Sealing in Hydrocarbon Reservoirs. Geological Society, London, Special Publications, 147, 269– 282. Knipe, R. J. 1997. Juxtaposition and seal diagrams to help analyze fault seals in hydrocarbon reservoirs. American Association of Petroleum Geologists Bulletin, 81, 187– 195. Knott, S. D. 1993. Fault seal analysis in the North Sea. American Association of Petroleum Geologists Bulletin, 77, 778– 792. Knott, S. D., Beach, A., Brockbank, P. J., Lawson brown, J., McCallum, J. E. & Welbon, A. I. 1996. Spatial and mechanical controls on normal fault populations. Journal of Structural Geology, 18, 359–372. Lecour, M., Cognot, R., Duvinage, I., Thore, P. & Dulac, J.-C. 2001. Modelling of stochastic faults and fault networks in a structural uncertainty study. Petroleum Geoscience, 7, S31–S42. Lia, O., Omre, H., Tjelmeland, H., Holden, L. & Egeland, T. 1997. Uncertainties in reservoir production forecasts. American Association of Petroleum Geologists Bulletin, 81, 775–802. Lindsay, N. G., Murphy, F. C., Walsh, J. J. & Watterson, J. 1993. Outcrop studies of shale smears on fault surfaces. In: Flint, S. T. & Bryant, A. D. (eds) The Geological Modelling of Hydrocarbon Reservoirs and Outcrop. International Association of Sedimentology, Special Publications, Blackwell, Oxford, 15, 113–123. Manzocchi, T., Walsh, J. J., Nell, P. & Yielding, G. 1999. Fault transmissibility multipliers for flow simulation models. Petroleum Geoscience, 5, 53–63. Manzocchi, T., Heath, A. E., Walsh, J. J. & Childs, C. 2002. The representation of two-phase fault-rock properties in flow simulation models. Petroleum Geoscience, 8, 119– 132.

FAULT UNCERTAINTY MODELLING Morrow, C. A., Shi, L. Q. & Byerlee, J. D. 1984. Permeability of fault gouge under confining pressure and shear stress. Journal of Geophysical Research, 89, 3193–3200. Ottesen, S., Townsend, C. & Øverland, K. M. 2005. Investigating the effect of varying fault geometry and transmissibility on recovery: using a new workflow for structural uncertainty modelling in a clastic reservoir. In: Boult, P. & Kaldi, J. (eds) Evaluating Fault and Cap Rock Seals. American Association of Petroleum Geologists, Hedberg Series, Tulsa, 2, 125–136. Rivenæs, J. C. & Dart, C. 2002. Reservoir compartmentalization by water-saturated faults – Is evaluation possible with today’s tools? In: Koestler, A. G. & Hunsdale, R. (eds) Hydrocarbon Seal Quantification. Norsk Petroleumsforenig, Special Publications, Elsevier, Amsterdam, 11, 187–201. Rivenæs, J. C., Otterlei, C., Zachariassen, E., Dart, C. & Sjøholm, J. 2005. A 3D stochastic model integrating depth, fault and property uncertainty for planning robust wells, Njord Field, offshore Norway. Petroleum Geoscience, 11, 57–65. Smith, D. A. 1966. Theoretical consideration of sealing and non-sealing faults. American Association of Petroleum Geologists Bulletin, 50, 363–374. Smith, D. A. 1980. Sealing and Nonsealing faults in Louisiana gulf coast salt basin. American Association of Petroleum Geologists Bulletin, 64, 145– 172. Sperrevik, S., Gillespie, P. A., Fisher, Q. J., Halvorsen, T. & Knipe, R. J. 2002. Empirical estimation of fault rock properties. In: Koestler, A. G. & Hunsdale, R. (eds) Hydrocarbon Seal Quantification. Norsk Petroleumsforenig, Special Publications, Elsevier, Amsterdam, 11, 109–125.

299

Sverdrup, E., Helgesen, J. & Vold, J. 2002. Sealing properties of faults and their influence on wateralternating-gas injection efficiency in the Snorre field, northern North Sea. American Association of Petroleum Geologists Bulletin, 87, 1437–1458. Thore, P., Shtuka, A., Lecour, M., Ait-Ettajer, T. & Cognot, R. 2002. Structural uncertainties: determination, management, and applications. Geophysics, 67, 840 –852. Van der Poel, N. J. & Cassell, B. R. 1989. Borehole seismic surveys for fault delineation in the Dutch North Sea. Geophysics, 54, 1091–1100. Watts, N. L. 1987. Theoretical aspects of cap-rock and fault seals for single and two phase hydrocarbon columns. Marine & Petroleum Geology, 4, 274 –307. Weber, K. J., Mandl, G., Pilaar, W. F., Lehner, F. & Precious, G. 1978. The role of faults in hydrocarbon migration and trapping in Nigerian growth fault structures. Paper OTC 3356 presented at the 10th Offshore Technology Conference, Houston, Texas, USA. Worthington, M. H. & Hudson, J. A. 2000. Fault properties from seismic Q. Geophysical Journal International, 143, 937–944. Yielding, G. 2002. Shale Gouge Ratio – calibration by geohistory. In: Koestler, A. G. & Hunsdale, R. (eds) Hydrocarbon Seal Quantification. Norsk Petroleumsforenig, Special Publications, Elsevier, Amsterdam, 11, 1 –15. Yielding, G., Freeman, B. & Needham, D. T. 1997. Quantitative fault seal prediction. American Association of Petroleum Geologists Bulletin, 81, 897– 917. Zhurina, E. N. 2003. Forward and inverse numerical modelling of fluid flow in a faulted reservoir: inference of spatial distribution of the fault transmissibility. PhD thesis, Texas A&M University, USA.

Geological factors effecting compartmentalization of Rotliegend gas fields in the Netherlands F. F. N. VAN HULTEN Energie Beheer Nederland B.V., P.O. Box 6500, 6401JH – Heerlen, The Netherlands (e-mail: [email protected]) Abstract: Since the discovery of the Groningen field, fifty years ago, more than 250 gas fields have been discovered in the Netherlands. A study of most of these fields shows that connected volumes are often smaller than expected from volumetric evaluation. Seismic uncertainty can often be argued as an explanation for the discrepancy, but there may be also a geological explanation. Most gas fields are found in the Permian Rotliegend and good drainage is expected because the reservoirs are relatively thick and homogeneous. Minor faulting is often thought to cause drainage problems, but this study shows it is likely in a number of cases. Twenty three Rotliegend fields with connectivity problems have been studied in part of the Dutch offshore. This analysis outlines a number of regions in the Rotliegend fairway that share a risk for fault seal and therefore reduced connected volumes. Fault seal analysis is important to understand compartmentalization but does not explain all discrepancies. In a number of regions, fault seal analysis is unsuccessful because cataclastic sand-to-sand sealing strike–slip faults are difficult to detect. Stratigraphic compartmentalization is seen to play a role in some well defined areas of the basin. A better understanding of the connectivity problems can highlight areas attractive for appraisal and near field exploration.

Most gas producing fields in the Netherlands are found in reservoirs formed by aeolian and alluvial sands of the Permian Upper Rotliegend Group (Fig. 1). First gas discoveries in the Netherlands were made over 60 years ago in shallower Zechstein carbonate reservoirs. It was after the discovery of the giant Groningen gas field (Fig. 2) 50 years ago that the ‘Rotliegend play’ gave an enormous impulse to gas exploration in what became known as Southern Permian Basin area (Glennie 1998). Energie Beheer Nederland B.V. (EBN) was founded to represent the national interest in management of these new gas reserves and today participates in over 250 gas fields for the Dutch state. More than 50% of these gas fields are producing from the Rotliegend. EBN has therefore had opportunity to study the compartmentalization problem across wider production acreage than most operating companies have available to them. Fields producing from these desert sands are present over a wide east –west corridor (Fig. 3) called the Rotliegend fairway (Van Wijhe et al. 1980; Glennie 1998; Wong et al. 2007). Triassic Bunter, Permian Zechstein, Upper Carboniferous, Cretaceous and Tertiary formations (Fig. 1) which also produce gas are less numerous. Rotliegend gas fields frequently show a discrepancy between static Initial Gas in Place (IGIP) volumes compared with the connected volumes as established from material balance calculations on production (e.g. Frikken & Stark 1993; van der Molen et al. 2003; Zijlstra et al. 2007). A good understanding of these discrepancies is important

since potential remedies like infill or horizontal wells can be very expensive, but very profitable if successful. The amount of gas that is connected to a well is an important economic factor, and varies from good gas fields, like in the Groningen field, with more than 10 109 m3 connected per well to 1–4 109 m3 in most offshore gas fields. Problematic fields or parts thereof have connected volumes of less than 1 109 m3, and a large number of gas fields have one or more poorly drained compartment. A better understanding of the mechanisms responsible for this connectivity problem is important if we are to adequately predict the level of compartmentalization and identify areas of unproduced gas. Much of gas production related infrastructure in the country is aging, so there is a concern that significant amounts of gas may be left behind when fields and production facilities are abandoned. This paper provides a short review of factors influencing and/or controlling degradation and compartmentalization of Rotliegend reservoirs. This review is based on basic subsurface and production data, and anecdotal evidence gathered from operating companies across the Netherlands over the past 30 –40 years.

Production discrepancy and ‘small field behaviour’ In the initial development phase of the Rotliegend gas fields (1970–1980), compartmentalization was


302

F. F. N. VAN HULTEN

the Dutch offshore, volumetric discrepancies began to be more clearly identified in several offshore fields. At that time, at least 20 Rotliegend fields (Van Hulten 1996) showed unexplained discrepancies between static and dynamic volumes. A few examples of gas field compartmentalization were documented in the literature during that period. For example, ‘small field behaviour’ is described in several fields in particular the structural and stratigraphic compartments in the K15-FG field (Frikken & Stark 1993; Frikken 1996a); and compartmentalization in the L13-FE field (Frikken 1996b). Another published example of a gas field with recognized permeability barriers was the Ameland field (Crouch et al. 1996). However, there was no clear explanation for the discrepancy between static and production volumes experienced over a larger region. In discussions with the various operators in the country it was sometimes difficult to extrapolate literature observations and the anecdotal experience of the asset teams operating these fields, to neighbouring fields and other fields across the region. Seismic imaging and interpretation uncertainties of many field descriptions also tended to obscure field-field comparisons. Given these issues and EBN’s unique access to field data from all the operators in the region, this provided the rationale for the study summarized in this paper.

Compartmentalization in Rotliegend gas fields Stratigraphic setting

Fig. 1. Tectono-stratigraphical chart for the Netherlands. The tectonic phases show the timing of the Alpine inversions (De Jager 2007).

not recognized as an important issue. Structural definition was often coarse and many infill wells were drilled at a later stage. For example, based on the information gathered after the discovery, such as pulse tests, the Groningen field development design did not plan for compartmentalization (Udink 1968). Interference of flow by faults (Van Rossum 1975) led to an adjustment of the Groningen development plan. Based on the information available it was not believed that faults could be complete permeability barriers. The second largest gas field, Annerveen (see Fig. 2), was also not thought to have transmissibility problems across faults (Veenhof 1996). However, in the early 1990s, with acquisition of improved 3D seismic in

The Rotliegend prospectivity is found in the extensively drilled east –west fairway which stretches from offshore United Kingdom (UK) into Northern Germany (Fig. 3). The play area is mainly defined by the presence of good Rotliegend reservoir sand, good Zechstein salt seal and the presence of source rocks (Van Wijhe et al. 1980; Wong et al. 2007). There are some new discoveries at the fringes of this corridor, but the play has matured over the last ten years, with most new finds within the fairway, best described as near field wildcats. The prospective aeolian and fluviatile facies belt is situated along the south side of the Southern Permian basin (Van Wijhe et al. 1980; Verdier 1996; Glennie 1998; Wong et al. 2007). The reservoir sands of the Rotliegend Group are called the Slochteren Formation (Fig. 4) and are known in the UK as the Leman sands, and in Germany as the Havel and Elbe subgroups (Van Adrichem Boogaert & Kouwe 1997). At the fringes of the aeolian facies belt, the stratigraphic unit thickens northwards; however the net sand thickness of the Slochteren Formation, becomes less with the

COMPARTMENTALIZATION OF DUTCH GAS FIELDS

303

Fig. 2. Index map with the onshore and offshore area of the Netherlands. Rotliegend gas fields discussed in the text and the major Basin areas (DCG, Dutch Central Graben; BF, Broad Fourteens Basin) are indicated. A major high where the Zechstein and part of the Rotliegend has been eroded is the Texel-IJsselmeer High (TIJH). On the map the position of the cross section of Figure 4 and the outline of map 9 has been indicated.

formation displaying more shale layers in what is called the Rotliegend feather edge. The Rotliegend reservoir depth is typically between 2000– 4700 m (Figs 5 & 10). It is good reservoir sand, with

generally high net-to-gross (N/G) and good porosity (average 15– 20%, ranging from 5.5 –26.9%), considering its burial depth should have induced significant diagenetic degradation. This unusual

Fig. 3. The Rotliegend play ‘fairway’ with the northern limits of the Lower and Upper Slochteren sands (Figs 1 & 4). The sands are deposited along the southern basin margin of the Southern Permian Basin. In this west– east corridor, which stretches from the UK into Germany, all Rotliegend gas fields of the Netherlands can be found. Rotliegend gas fields are orange. On the map the northern limit of the Upper Slochteren (Fig. 4) has been shown (red line) that outlines the narrow facies belt, where the Upper Slochteren is very thin and stratigraphic compartmentalization can be expected. The southern limit of the fairway is the absence of top seal, the Zechstein salt.

304

F. F. N. VAN HULTEN

Fig. 4. Stratigraphic SW–NE cross section indicating the various facies of the Upper Rotliegend Group. It illustrates the depositional edge of the Upper Slochteren. These sands are shaling out to the north as can be seen in in the transition of L10 to L11-A5.

porosity preservation is partly due to the dominant well sorted aeolian facies. Fluvial and sabkha facies, which are present in parts of the area, display poorer reservoir quality. Most fluvial facies contain a relatively high percentage of well rounded quartz grains indicating a long preceding aeolian transport (Almon 1981). Almon describes

0 Rotliegendes

Depth (m)

–1000 L.Cret. Bunter Zechstein

–2000

Rotliegend sandstones as dominantly fine to medium mature to sub mature sandstones. Monocrystalline quartz grains are the dominant grain type. About 20% is polycrystalline. The composition of the primary sandstones is on average quartz (85%), feldspar (4–6%), rock fragments (4–7%). There is an increase of feldspar content to the top of the Slochteren Formation. The thickness of the reservoir sands is generally over 50 m and often exceeds 200 m. However, typical column height in Dutch gas fields is usually below 100 m and rarely exceeds 300 m (Fig. 6). Therefore in almost all cases only the top part of the sequence is important for reservoir studies.

Carbonif.

Structural setting Base Zechstein seismic marker

–3000

–4000

–5000

0

250 N = 250

Fig. 5. Gas fields in the Netherlands sorted by depth. Rotliegend fields display a wide depth range from 2000– 4700 m. Most Rotliegend fields are situated at a depth between 3000–4000 m.

Most if not all Rotliegend traps are structural and generally consist of a dipping fault block (Fig. 7a) with clearly defined spill points (see e.g. Corona 2005). This means that beside the top reservoir structure map and gas–water contact (GWC), two or more faults define the dimensions of the field. Juxtaposition seals describe most reservoir seals in Rotliegend gas fields. This is analogous to the

COMPARTMENTALIZATION OF DUTCH GAS FIELDS 400 Rotliegendes L.Cret. Bunter Zechstein Carbonif.

Column Height (m)

300

200

100

0 0

250

N = 250

Fig. 6. Gas column height of 250 gas fields in the Netherlands. A typical column is below 100 m and columns rarely exceed 300 m. Most Rotliegend fields are dip closed fault blocks. The limited column creates risk for compartmentalization because small faults can seal parts of the reservoir.

(a)

Major faults GW

C

Subtle lineaments

G WC GW

C

1

(b)

2 3

305

majority of structural traps in the northern North Sea (e.g. Knott 1993). Structural definition of the faults is of great importance to understand the Rotliegend fields. The seismic marker that is generally used to define the faults is the geophysical interpretation of the Base Zechstein (¼ Top Upper Rotliegend Group). This most intensively studied structural horizon in the Netherlands, provides a bright coherent seismic image when not dimmed or masked by noise caused by salt and other complexities in the overburden. The underburden is usually poorly imaged by seismic. Base Zechstein structural interpretations typically show a rhomboid pattern of intersecting faults (e.g. Oudmayer & De Jager 1993; Frikken 1996a; De Jager 2007), with a dominant family of NW –SE faults intersected by NNE –SSW faults. Towards the northern limit of the Rotliegend fairway, a more north–south fault trend is found. These trends are fairly consistent throughout the prospective corridor, even in major Jurassic rift basins like the Broad Fourteens Basin in the South and the Dutch Central Graben in the North (Van Adrichem Boogaert & Kouwe 1997). Many faults at Rotliegend level appear to be Triassic–Tertiary features which nucleated on Mid Palaeozoic or even Caledonian NW–SE striking fault patterns which were reworked during the Variscan orogeny (De Jager 2007). The Mesozoic rifting phase established the general extensional normal fault pattern. Thus, most fault arrays at Top Rotliegend level are ‘inherited’ displaying few clearly new-formed structures, with the general exception of easy to identify ‘Alpine’ pop up structures (e.g. Glennie 1998; Barr 2007; De Jager 2007). Present day stress is NNW–SSE (N1358), but pre-historic stress due to compression in the Alpine orogenic period was more north–south or NNE–SSW (Frikken 1996a; Gauthier et al. 2000). The evidence for strike–slip faults as suggested by Frikken & Stark (1993) is not very well documented. Work using seismic surface illumination techniques, suggests that north–south and NE–SW (N458) transfer faults exist with minimal displacements that are almost sub-seismic (Geiss et al. 2009).

Compartmentalization

Fig. 7. (a) 3D visualization of the K15-FG field. The field displays major faults and a number of subtle lineaments that are small faults suspected to seal. The small faults have been interpreted as shear zones or sealing strike-slip faults (see Frikken 1996a). (b) Diagram of a typical sand-to-sand seal of a Rotliegend gas trap in the Netherlands with two major faults (number 1, 3) and a subseismic fault (2). The expression of the subtle lineaments of Figure 7a is shown as subseismic fault (2).

Compartmentalization is a key issue for planning of infill, appraisal and near field exploration drilling. The discrepancy between static IGIP, compared with the connected volumes as established from material balance calculation, assumes an interconnected volume in one single trap. Many fields contain a number of rhomboid shaped fault blocks (Fig. 7a), positioned next to each other with sand to sand contact. In a trap with a number of such blocks, it was common practice during the 1970– 1980s to drill at least one development well

F. F. N. VAN HULTEN

Factors explaining poor connectivity There is still an active debate within the literature on the mechanisms responsible for causing barriers and

baffles at production or trap level. Corona (2005) and Corona et al. (2010) suggest that better geological studies, in particular on fault juxtaposition, can explain most reduced volumes, considering the uncertainties carried in many volumetric calculations; whilst Zijlstra et al. (2007) in a refining of earlier observations of Frikken (1996a) provide an explanation related to capillary rise of water within (low permeability) quartz-cemented cataclastic fault zones, being impermeable to gas. In other fields there is a debate to what extent stratigraphic compartmentalization causes production barriers to the gas.

(a)

30.0 Rotliegendes 25.0

IGIPmb (×109 m3)

in the major intra-field blocks. Even on 2D coverage it was clear that most of the intra field faults that separate the various rhomboids could create some kind of transmissibility barrier. At the time, it seemed reasonable to assume tank-like production behaviour and this turned out to be the case in many Rotliegend fault blocks. However, improved seismic imaging and volumetric estimation of the IGIP, together with better pressure information gathered later in production life of the fields, caused various operators to realize that wells did not always have effective communication with the entire reservoir within the fault blocks. Sometimes new seismic acquisition shows additional minor faults or visualizes subtle lineaments that identify compartment boundaries not recognized before. There is good static and dynamic IGIP information in almost every gas field in which EBN participates. The dynamic information, derived from p/Z plots, is based on comparing average gas field pressure against cumulative gas production. Zijlstra et al. (2007) give some relevant examples. Ideally the static and dynamic volumes are the same in a single block. In a typical Rotliegend reservoir block without aquifer support, the p/Z data should plot as a straight line. This is often the case. The accuracy of the dynamic volume estimates (i.e. material balance calculations) is further aided by the fact that Rotliegend gas fields generally display minor water movement (i.e. rarely have an active aquifer). A straight line is often not observed on a p/Z plot plot from fields that have gas volumes located behind flow baffles. After an initial rapid pressure decline the (slow) gas behind the baffles flows in, so that later in field life, the dynamic volume catches up with initial static gas volume. For compartmentalized reservoirs the dynamic volume is always smaller than the static IGIP. A comparison of static and dynamic IGIP of Rotliegend gas fields in the Netherlands (Fig. 8) shows, that connected volumes in 10–20% of the gas fields are smaller than expected from volumetric evaluation. A number of fields initially had lower dynamic volumes but infill drilling reduced volumetric discrepancies (Fig. 8a). These fields do not show up in the Figure 8a plot, but are shown on the maps as compartmentalized. For this study, a large part of the offshore K and L license blocks has been selected for a more detailed analysis. For 23 gas fields from various operators in the Dutch offshore K and L area, fields with connectivity problems have been plotted. The approximate position of these fields is given in Figure 9a.

L. Cretaceous Bunter Zechstein

20.0

Carboniferous 15.0

10.0

5.0

0 0

5.0

10.0

15.0

20.0

25.0

30.0

IGIPvol (×109 m3)

(b)

20.0 IGIP 1995 15.0

IGIPmb (×109 m3)

306

IGIP 2004

10.0

5.0

0.0 0.0

5.0

10.0

15.0

20.0

IGIPvol (×109 m3)

Fig. 8. (a) Comparison of volumetric (static) Initial Gas In Place (IGIP) and volumes derived from material balance calculation (dynamic). The dynamic data is generally derived from p/Z plots and is based on comparing average gas field pressure against cumulative gas production. Ideally the static and dynamic volumes are the same. The comparison shows a number of fields with less connected volumes than calculated volumetrically. Some fields with connectivity problems do not show a discrepancy on this plot because remedial action like infill drilling has reduced the discrepancy. (b) Comparison of the volumes in 2005 used in the plot of Figure 8a with a previous survey about 10 years earlier (Van Hulten 1996), reveals that in a number of fields part of the discrepancy has been reduced in time, because of new data, revisions, infill drilling or in the accounting by discounting certain parts or blocks of the field.


307

Fig. 9. (a, b) Study area, part of K and L Quads of the Netherlands, alsoknown as the Central Offshore Saddle. The depth map of the Base Zechstein (Wong et al. 2007) gives an indication of the position of the Rotliegend gas fields (red dots) in relation to the basin areas. DCG, Dutch Central Graben; BF, Broad Fourteens Basin. The big white dots indicate fields with connectivity problems. They plot in a narrow zone. Area is situated between two important Jurassic rift basins the Broad Fourteens Basin and the Dutch Central Graben.

0 Rotliegend fields Fields in impaired areas

Depth (m)

–1000

–2000

–3000

–4000

–5000 7.0

9.0

11.0

13.0 15.0 Porosity (%)

17.0

19.0

21.0

Fig. 10. Average porosity of Rotliegend fields plotted against depth. The porosity of the reservoirs is good considering the average depth. The porosity of the sands in the producing fields ranges from about 7% to over 20%. Most fields have porosities between 11– 15%. No simple relation with depth exists. Reservoir quality is impaired in a number of regions like the inverted basins or an areas called D2, close to the TexelIJsselmeer High. Those fields are indicated with a dark triangle.

Static volumetric uncertainty is a key factor in the assessment of compartments. Despite very good coverage from 3D seismic data over almost all producing fields, and application of modern analytical techniques, there are many sources of residual uncertainty within subsurface studies and volumetric calculations. Due to the depth of the Rotliegend between 2000–4700 m and the complexity of its overburden (in particular the complex Zechstein salt configurations), the structure of some fields is poorly imaged or confused by noise within the seismic data (Wong et al. 2007). In addition to volumetric uncertainty caused by seismic time–depth conversion problems, field volumes may be wrongly estimated due to an uncertain position of the Free Water Level (FWL) or the related saturation –height calculation. The latter uncertainties sometimes reduce the volume by more than 10% in particular in low porosity reservoirs. Porosity or N/G calculations for Slochteren gas fields are generally more reliable. The magnitude of the inter-block connectivity problems in many gas fields, however, can be severe enough to make it clear that there is more to the observed discrepancies than a volumetric uncertainty. Indeed, infill drilling has found

308

F. F. N. VAN HULTEN

virtually undepleted compartments in or near fields, relatively late in production life. Three main mechanisms have been found to control Rotliegend compartmentalization and reservoir degradation: (a) faulting and fault sealing (the most significant mechanism); (b) stratigraphic compartmentalization (less important); and (c) reservoir impairment due to diagenetic alteration (also enhances sealing capability of faults).

100

75

C 50

25

0 1

0.8 0.6 0.4 0.2 Discrepancy IGIP volumetrics/IGIP material balance

0

C = Connectivity = 1-(Dn [Net/Gross])

Fault seal

Dn = Normalized throw (Knott 1993)

There is general agreement in the literature and anecdotally during EBN discussions with field operators, that faults are the most important cause of compartmentalization in the Rotliegend. Faults are the most important cause for heterogeneity, in typical Rotliegend reservoirs, capturing the gas at trap level and sealing intra-trap compartments on production. Parameters such as fault strike and throw, reservoir thickness, depth, N/G, porosity and net sand connectivity were plotted against seal performance to define trends by Knott (1993). However, Knott found that fault juxtaposition diagrams did not explain the sealing capacity of all faults, since they revealed large sand-to-sand contact windows. In a more recent study of the Lauwerszee Trough ‘Golden Lane’, Corona (2005) found that 25% (7 of the 30 fault traps) he described are filled deeper than the fault juxtaposition leak point and require sand-to-sand fault seal. Some confusion exists in the literature on fault seal, because explorationists, production geologists and reservoir engineers often define ‘sealing’ of faults differently. Furthermore, faults within the Rotliegend are frequently baffles that only act as total barriers to flow during certain periods of the production, and their sealing behaviour may evolve gradually or change abruptly during the production life-cycle of a field. Zijlstra et al. (2007) give examples of fields with compartments that contribute in a late stage of production presumably when a certain cross-fault threshold pressure has been reached during production pressure draw-down. Fields, that contribute in a late stage with such ‘slow gas’ behaviour, are generally considered to be fault compartmentalized. Some of the late production has been explained anecdotally as high pressure differentials, caused by depletion, overcoming flow barriers. The popular perception is that increases in the rate of cross-fault flow are either caused by mechanical failure of the fault, or by a change in gas saturation increasing the effective gas permeability of the fault rock, however, this phenomenon is still poorly understood. Apparent large sand to sand fault contact areas are known from several fault compartmentalized gas fields in the Dutch Rotliegend. Knott (1993)

Fig. 11. Connectivity (C) v. discrepancy between static and dynamic volumes (IGIPv/IGIPmb). Knott (1993) suggests that incorporating N/G in the connectivity estimation can better explain sealing sand-to-sand juxtaposition. An analysis with the formulas proposed in his papers has been preformed for the Dutch gas fields. This calculation of the connectivity with for N/G corrected throw did not yield a good correlation with the discrepancy.

suggests that incorporation of N/G in the connectivity estimation can help explain sealing sandto-sand juxtaposition. His N/G corrected connectivity analysis has been performed for Dutch gas fields in this study. However calculation of the connectivity with N/G corrected throw did not significantly improve results (Fig. 11). Continuous shale beds are rare in many Dutch compartmentalized reservoirs. Thus, it is clear that sealing sand-to-sand fault contacts truly exist and even large fault throw does not correlate effectively to fault compartmentalization. Indeed, many faults that are suspected of being sealing are at the very limit of seismic throw resolution, suggesting that only minimal fault throw is required to form a barrier to fluid flow. This is consistent with the notion that it is the effective stress and temperature history of Rotliegend reservoirs during deformation, which controls fault sealing and not the fault throw itself (Fisher & Knipe 1998, 2001; Fisher et al. 2003).

Cataclasis Cataclasis (grain fracturing and breakage) is induced by increased stress conditions. The process reduces grain size and grain sorting, thus collapsing porosity; and cataclastic fault rocks become more susceptible to quartz cementation due to their extremely high reactive quartz surface area (Fisher & Knipe 1998, 2001). Fisher & Knipe (2001) found that quartz cemented cataclastic fault gouges are particularly common in Rotliegend reservoirs, and that these low permeability fault gouges have a high sealing capacity, and thus


represent barriers or baffles to fluid flow. However, only a few cores have been collected across known seismic-scale sealing faults within Rotliegend gas fields (that have been shared within the public domain). These cores contain zones of deformation bands (e.g. Fig. 12), that have lower porosity and permeability than the surrounding undeformed sandstone. Intense fracturing of the sand grains causes porosity collapse, even though the displacement estimated for each deformation band is typically only a few mm, with the cumulative displacement of a zone of deformation bands being less than 0.5 m (e.g. Aydin & Johnson 1978, 1983). Larger displacements in Aydin and Johnson’s classification are called slip surfaces. It is difficult to distinguish between a deformation band zone and a slip surface in Rotliegend core, because they look very similar. Fractured grains and collapsed pores create a zone with small grains and poor

5

Permeability (mD) 10 15 20

25

1 2 3 4 5 6 7 8 9

Measurement point

Distance between measurement point 2 cm At point 6 leakage between rock and probe

10 11 12

Fig. 12. Example of a cataclastic fault in the Rotliegend core of the well Bierum-1 (provided by K. Weber and W. Nieuwenhuijs).

309

sorting typical for a cataclastic fault. The rock properties of the displacement zone are therefore significantly different compared to the other parts of the reservoir. In the core of the Bierum-1 well, significant permeability reduction is observed from 25 mD to basically zero in the deformation zone (Fig. 12). A few other examples of Rotliegend cataclastic fault rocks have been described in the literature (Fisher & Knipe 1998). Also examples of the German Rotliegend (Mauthe 2003) have similar deformation bands (their paper also provides photomicrographs of the fault rocks). Another example was discussed at a Royal Geological and Mining Society of the Netherlands meeting December 15, 1993: W. H. Nieuwenhuijs presented a core study of the Grijpskerk-2 exploration well (drilled in 1992), that had cut a sealing fault between two compartments of the gas field having 10 m difference of the FWL. Here the fault zone was also composed of crushed grains (cataclasite). In the K15-FG Rotliegend gas field Frikken (1996a) found ‘fracture fill’ consisting of fine-grained rock flour (ultracataclasite). A more recent example of cataclasis in a sealing fault in L10-4, a few km east from the L10-A field (Fig. 2) was presented on 5 June 2008 by R. H. B. Rijkers at an EBN-TNO workshop in Utrecht on rifting systems and hydrocarbon exploration in the Netherlands (convened by F. F. N. van Hulten and J. E. Lutgert). Examples such as these support and calibrate the cataclastic fault seal mechanisms described and constrained by core analysis, microstructural observations and production data elsewhere in the literature (e.g. Zijlstra et al. 2007). Sand-to-sand fault seals (Fig. 7b), also known as cross seals (Smith 1966; Schowalter 1979; Watts 1987), are related to the high capillary entry pressure of the fault gouge itself. The (membrane) seal will trap a hydrocarbon column until buoyancy pressure exceeds the capillary entry pressure of the seal. More specifically, it has been argued that whilst gas displaces water and fills the reservoir, the higher capillary entry pressure of the fault rocks means that they remain below the critical gas saturation and are therefore impermeable to gas above the FWL – until the buoyancy force at the top of the gas column overcomes the entry pressure of the wet fault rock (Fisher et al. 2001; Zijlstra et al. 2007). Where capillary rise of water within the fault zone spans the height between the FWL and the top seal (especially in the flanks of a field where the top seal is closer to the FWL) – the fault may completely seal-off a compartment at or close to ‘virgin’ pressure until very late in the production life-cycle, despite production pressure draw-down within neighbouring fault blocks (e.g. the Leman field, Zijlstra et al. 2007).

310

F. F. N. VAN HULTEN

The magnitude and direction of the present day effective stress on the fault plane may also exert an influence on the sealing capacity of faults – since ‘critically stressed’ structures may be prone to geologically recent fault reactivation. The suspected sealing of NE –SW striking faults (Frikken 1996b; Geiss et al. 2009) might be explained by present day NW–SE stress and somewhat less from the more north–south pre-historical stress. Most Dutch offshore gas fields that show signs of compartmentalization, plot in a relatively small region sometimes called the Central Offshore Saddle. This area lies between two major Jurassic rift basins; the Broad Fourteens Basin in the south and the Dutch Central Graben in the North. It is assumed that Alpine deformation is concentrated in this transfer zone between the two basins, thus causing the fault pattern there (Fig. 9b), and apparently modifying the fault sealing properties of reactivated faults. However, more work has to be done to determine whether the faults changed by the late inversion movements and where these faults can be expected to map out.

Clay smear and cement precipitation Clay smear (Lindsay et al. 1993; Yielding et al. 1997; Fisher & Jolley 2007) and the precipitation of cements from the Zechstein have also been suggested to explain sealing of sand-to-sand juxtapositions in some reservoirs. Corona (2005) suggests mineral precipitates derived from Zechstein fluids might have caused sand-to-sand sealing in the Saaxum gas field in the Lauwerzee Trough, as was suggested in the Jupiter fields in the UK (Leveille et al. 1997). This is a possibility but lacks support from petrophysical data or core examples. Indeed, carbonate cements are often very patchy in faults seen in cores, and some workers now believe that salt precipitates in particular (e.g. those seen in the Jupiter study samples), are likely to have resulted from human activity collecting, handling, and preserving the core (Q. J. Fisher, pers. comm. 2008). Clay smear mechanisms are also not supported by well logs or core observations. Given the burial depth of the Rotliegend, the timing of deformation and the general absence of clay, structural and stratigraphic logging, petrographic and petrophysical analyses of core samples consistently point to cataclastic fault seal mechanisms.

Stratigraphic barriers When N/G values come close to 40–50%, shales may create flow baffles or isolated stratigraphic compartments where the shale beds are laterally persistent as for example can be found in the northern fringe of the Upper Slochteren of the Ameland

field (Crouch et al. 1996). The Upper Slochteren sands thin systematically towards the north, where they are interbedded with shale layers in the transition to desert lake facies. This causes the formation to be prone to stratigraphic compartmentalization, and can (partly) explain connectivity problems within a relatively narrow and well defined east –west belt in the Netherlands where the Upper Slochteren starts to shale out towards the north (Fig. 3). Thus, the compartmentalization of the Ameland gas field is caused by stratigraphic layering in combination with faulting (Crouch et al. 1996). Broader stratigraphic compartmentalization is also known between the Lower and Upper Slochteren across the 10 m thick Ameland shale, which often acts as a pressure barrier between these two sand members (see Fig. 4). In areas with thick Slochteren sands a hypothesis was put forward by Frikken & Stark (1993) and Frikken (1996b) that facies differences that caused an interbedding of permeable and tight reservoir could explain areas with connectivity problems. Based on information of a Circumferential Borehole Imaging Log a detailed analysis of the facies was obtained. The Slochteren there is characterized by relative thin aeolian dune- and sheet-sands with good reservoir quality v. water laid sands where drainage is fully dependent on cross-flow, inside the reservoir, into depleted aeolian sands. This stratigraphic compartmentalization could also partly explain the volumetric discrepancy of K15-FG.

Diagenesis The presence of authigenic minerals can occlude pore space and thus degrade reservoir properties. It may enhance the above mentioned fault sealing and stratigraphic compartmentalization. These diagenetic effects are generally associated with specific facies belts within the Rotliegend (Fig. 13). Early diagenesis and burial are important factors that can alter reservoir quality. In the Rotliegend Slochteren sands during the early diagenetic processes, dolomite, anhydrite and authigenic quartz are precipitated. The mechanism that can explain the growth of these minerals is shallow groundwater diagenesis caused by evaporation and CO2 degassing (Amthor & Okkerman 1998). The risk of this type of reservoir impairment is greatest in the distal facies of the Upper Slochteren, deposited close to the desert lake (Fig. 3). With the increased temperature during burial the reservoir underwent extensive diagenesis. Kaolinite and illite are the most frequently formed clay minerals in the Rotliegend. With increased burial, grain contact quartz dissolution also causes authigenic quartz cementation and pore space occlusion. Permeability destruction by authigenic growth of


311

Fig. 13. Areas with increased risk for reservoir impairment because of autogenic clays. BF, Broad Fourteens Basin area that has been strongly inverted reservoirs are prone to diagenetic alteration because of deep burial. TIJH, Texel-IJsselmeer High, a palaeohigh where Zechstein cover has been eroded. D2 is an area close to the Texel-IJsselmeer High and the Zuidwal (Jurassic) Volcano. The D2 area shows many Rotliegend reservoirs with high diagenetic illite percentages. The shale-out of the Upper Slochteren (Fig. 3) as a third area that shows diagenetically altered reservoir.

quartz is particularly severe within cataclastic fault rocks, where they were formed at 90 8C or higher temperatures (Fisher et al. 2000, 2003). This has the most obvious impact on reservoir connectivity as explained above. Growth of kaolinite and illite in particular, also has a detrimental impact on reservoir quality (e.g. L13 block, Frikken & Stark 1993). In the Rotliegend fairway a number of regions can

be outlined that show reduced reservoir quality due to the presence of diagenetic clays. There is a general decrease of porosity with depth (Fig. 10). This is partly due to chemical and mechanical compaction and partly caused by the formation of authigenic minerals. Several areas can be distinguished where this kind of reservoir impairment can be found over larger areas.

312

F. F. N. VAN HULTEN

Regions with plugged pore space are essentially limited to three areas of the Rotliegend fairway: † In the Slochteren feather edge to the north, the wadi facies is associated with diagenetic anhydrite and carbonate minerals. In combination with faulting, this can increase the risk of compartmentalization (Frikken & Stark 1993). Areas with dominant sabkha facies have a similar association, and can be found in the Upper Slochteren in the transition zone where it shales out to the north (Fig. 3). Given the close association between facies and enhanced diagenetic mineral growth, it is sometimes hard to distinguish diagenetic impacts from the stratigraphic compartmentalization that is also associated with this facies belt. † In the inverted Broad Fourteens Basin the Rotliegendes has been uplifted 1000– 2000 m (Wong et al. 2007). These areas show reservoir impairment due to the presence of diagenetic illite, formed at pre-inversion depths. Given the combination of poor reservoir quality and poor charge here, there are not many Rotliegend gas fields in this part of the Southern Permian Basin. † Close to the Texel-IJsselmeer High, there is an area which does not have Zechstein salt cover. A number of Rotliegend gas fields in this area lie in the northern corner of the high, and characteristically have severely reduced permeability due to the presence of illite (e.g. L13-FD field, Frikken 1996b). In EBN discussions with operating companies, two basic suggestions have been made to explain the illite in this area. Firstly, that meteoric water accessed the reservoir from shallower horizons during the Triassic or Jurassic (due to the absence of the Zechstein). However, this does not fit with the observation that no serious illite impairment is seen further to the south of the Texel-IJsselmeer High in wells that reached the Rotliegend in the onshore Friesland province. Indeed, meteoric water might be expected to remove K-feldspar during shallow burial, thus making the fields less susceptible to illitization during deep burial. Secondly, it has been suggested that the illitization is related to hydrothermal flow from the nearby Jurassic Zuidwal volcanic centre (Van Adrichem Boogaert & Kouwe 1997). Clearly, more work is required to fully explain the illitization in this area.

Other gas producing formations EBN participates in more than 50 gas fields found in Bunter sands of the Lower and Upper Germanic Trias Group (Fig. 1), the second important gas producing reservoir zone, to which some of the

suggestions from this Rotliegend study can be applied. Compartmentalization is noticeable in a number of these fields. Geographically the Triassic Bunter is more restricted to the West Netherlands Basin, the larger Dutch Central Graben area and the Lower Saxony Basin (Wong et al. 2007). The Bunter sands (Fig. 8a) are depositionally comparable to the Rotliegend (they share a similar desertlike setting), but there are marked differences in structural setting, given the Bunter overlies the Zechstein salt (Fig. 1), and is often overlain by Triassic Ro¨t salt. Bunter reservoirs in the northern offshore have serious ‘salt plugging’ problems not generally known from the Rotliegend reservoirs (Dronkert & Remmelts 1996). In some regions, high and low permeability reservoir is ‘layered’ such that the Bunter is prone to stratigraphic compartmentalization (d’Engelbronner & Haak 1993), and some areas have significant diagenetic alteration (e.g. inverted parts of the West Netherlands Basin), which can impair the drainage of a reservoir block. Other producing reservoirs which lie within the Carboniferous Limburg Group sandstones, Lower Cretaceous and Tertiary sandstones contain too few gas fields to allow the identification of regional compartmentalization trends. However, in general terms, it has been observed that: (a) the interbedded sand and shale nature of the fluvial reservoir architecture in the Carboniferous sandstones tends to generate stratigraphic compartmentalization; and (b) Cretaceous and Tertiary reservoirs carry greater risk of soft sediment deformation and clay-smear fault sealing. Compartmentalization in the carbonates of the Permian Zechstein Group can be very serious (Fig. 8a), although the mechanisms responsible for this and the nature of the natural fracture systems have not been clearly established.

Conclusions The aging infrastructure and the rapidly approaching abandonment of many Rotliegend gas fields, has generated some urgency in the Netherlands around improving our understanding of the volumes of gas present in unswept compartments. There is a strong consensus amongst industry and academic workers, that there are three major impacts on reservoir connectivity within the Rotliegend: (1) Fault seal is the most significant cause of this compartmentalization. Fault juxtaposition of reservoir sands and non-reservoir lithologies (e.g. salt, shale) provide most of the fault seals. However, sand-to-sand fault seals have been observed in up to 25% of cases. The mechanism that creates these particular fault


(2)

(3)

barriers or baffles appears to be porositypermeability collapse within cataclastic fault rocks, enhanced by quartz cementation occluding the porosity. The resulting impermeability to gas is caused by capillary rise and sub-critical gas saturations within the fault zones. Shale smearing is not observed and not expected in the Rotliegend. Stratigraphic compartmentalization is the second factor. Parts of the Upper Slochteren become shale-prone within a narrow aeoliandesert lake facies transition belt within the Rotliegend prospective belt. Reservoirs here are susceptible to compartmentalization across laterally persistent shale, often in combination with fault sealing. Diagenetic impairment of reservoir quality is the third mechanism identified which impacts gas connectivity within Rotliegend reservoirs.

Mapping of the fields with connectivity problems in the Netherlands shows that these three seal and impairment mechanisms tend to be spatially restricted. Most Rotliegend fields with compartmentalization problems plot within a corridor between the Jurassic rifting basins, the Broad Fourteens basin and the Dutch Central Graben. However, in all of the compartmentalized Rotliegend fields examined by EBN in the Netherlands, gas connectivity can be explained by the three mechanisms described in this paper. This includes poor connectivity during production and barriers to gas migration during trapping. The author thanks Energie Beheer Nederland B.V. for permission to publish this paper. Koen Weber and Willem Nieuwenhuijs are acknowledged for their comments and for providing a photograph of a sealing fault of Bierum-1. Quentin Fisher and Steve Jolley are thanked for their constructive review and editing of the manuscript.

References Almon, W. R. 1981. Depositional environment and diagenesis of Permian Rotliegendes sandstones in the Dutch sector of the southern North Sea. In: Longstaffe, F. J. (ed.) Clays and the Resource Geologist. Mineralogical Association of Canada Short Course. Co-op Press, Edmonton, Chapter 7, 119– 147. Amthor, J. E. & Okkerman, J. A. 1998. Influence of early Diagenesis on reservoir quality of Rotliegende Sandstones, Northern Netherlands. American Association of Petroleum Geologists Bulletin, 82, 2246–2265. Aydin, A. & Johnson, A. M. 1978. Development of faults as zones of deformation bands and as slip surfaces in sandstone. Pure and Applied Geophysics, 116, 931–942. Aydin, A. & Johnson, A. M. 1983. Analysis of faulting in porous sandstones. Journal of Structural Geology, 5, 19– 31.

313

Barr, D. 2007. Conductive faults and sealing fractures in the West Sole gas fields, southern North Sea. In: Jolley, S. J., Barr, D., Walsh, J. J. & Knipe, R. J. (eds) Structurally Complex Reservoirs. Geological Society, London, Special Publications, 292, 431– 451. Corona, F. V. 2005. Fault trap analysis of the Permian Rotliegend gas play, Lauwerszee Trough, NE Netherlands. In: Dore´, A. G. & Vining, B. A. (eds) Petroleum Geology: North-West Europe and Global Perspectives. Proceedings of the 6th Petroleum Geology Conference, Geological Society, London, 327– 335. Corona, F. V., Davis, J. S., Hippler, S. J. & Vrolijk, P. J. 2010. Multi-fault analysis scorecard: testing the stochastic approach in fault seal prediction. In: Jolley, S. J., Fisher, Q. J., Ainsworth, R. B., Vrolijk, P. J. & Delisle, S. D. (eds) Reservoir Compartmentalization. Geological Society, London, Special Publications, 347, 317–332. Crouch, S. V., Baumgartner, W. E. L., Houlleberghs, E. J. M. J. & Walzebuck, J. P. 1996. Development of a tight gas reservoir by a multiple fracced horizontal well: Ameland-204, the Netherlands. In: Rondeel, H. E., Batjes, D. A. J. & Nieuwenhuijs, W. H. (eds) Geology of Gas and Oil Under the Netherlands. Royal Geological and Mining Society of the Netherlands, Kluwer Academic Publishers, Dordrecht, 93–102. Dronkert, H. & Remmelts, G. 1996. Influence of salt structures on reservoir rocks in Block L2, Dutch continental shelf. In: Rondeel, H. E., Batjes, D. A. J. & Nieuwenhuijs, W. H. (eds) Geology of Gas and Oil Under the Netherlands. Royal Geological and Mining Society of the Netherlands, Kluwer Academic Publishers, Dordrecht, 159– 166. De Jager, J. 2007. Geological development. In: Wong, T. E., Batjes, D. A. J. & De Jager, J. (eds) Geology of the Netherlands. Royal Netherlands Academy of Arts and Sciences, Amsterdam, 5–26. D’ Engelbronner, W. & Haak, A. M. 1993. Waalwijk field history: impact of compartementalization or layered depletion on reservoir management. American association of petroleum geologists, international conference & exhibition, October 17– 20, The Hague, the Netherlands, abstract. American Association of Petroleum Geologists Bulletin, 77, 1618. Fisher, Q. J. & Jolley, S. J. 2007. Treatment of faults in production simulation models. In: Jolley, S. J., Barr, D., Walsh, J. J. & Knipe, R. J. (eds) Structurally Complex Reservoirs. Geological Society, London, Special Publications, 292, 219–333. Fisher, Q. J. & Knipe, R. J. 1998. Fault sealing processes in siliclastic sediments. In: Jones, G., Fisher, Q. J. & Knipe, R. J. (eds) Faulting, Fault Sealing and Fluid Flow in Hydrocarbon Reservoirs. Geological Society, London, Special Publications, 147, 117–134. Fisher, Q. J. & Knipe, 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. Fisher, Q. J., Knipe, R. J. & Worden, R. 2000. The microstructure of deformed and undeformed sandstones

314

F. F. N. VAN HULTEN

from the North Sea: its implications for the origin of quartz cement. In: Worden, R. H. & Morad, S. (eds) Quartz Cementation in Sandstones. International Association of Sedimentologists, Special Publication, Wiley-Blackwell, Oxford, 29, 129 –146. Fisher, Q. J., Harris, S. D., McAllister, E., Knipe, R. J. & Bolton, A. J. 2001. Hydrocarbon flow across sealing faults: theoretical constraints. Marine and Petroleum Geology, 18, 251– 257. Fisher, Q. J., Casey, M., Harris, S. D. & Knipe, R. J. 2003. The fluid flow properties of faults in sandstone: the importance of temperature history. Geology, 31, 965– 968. Frikken, H. W. 1996a. CBIL logs: vital for evaluating disappointing well and reservoir performance, K15-FG field, central offshore Netherlands. In: Rondeel, H. E., Batjes, D. A. J. & Nieuwenhuijs, W. H. (eds) Geology of Gas and Oil Under the Netherlands. Royal Geological and Mining Society of the Netherlands, Kluwer Academic Publishers, Dordrecht, 103– 114. Frikken, H. W. 1996b. Sub-horizontal drilling: remedy for underperforming Rotliegend gasfields, L13 block, central offshore Netherlands. In: Rondeel, H. E., Batjes, D. A. J. & Nieuwenhuijs, W. H. (eds) Geology of Gas and Oil Under the Netherlands. Royal Geological and Mining Society of the Netherlands, Kluwer Academic Publishers, Dordrecht, 115– 124. Frikken, H. W. & Stark, J. B. 1993. Character and performance of small Rotliegend gas reservoirs, Central Offshore Netherlands. In: Aasen, J. O., Buller, A. T., Hjelmeland, O., Holt, R. M., Kleppe, J. & Torsæter, O. (eds) North Sea Oil and Gas Reservoirs – III, Proceedings of the 3rd North Sea Oil and Gas Reservoirs Conference, Trondheim, Norway, Kluwer, Dordrecht, 41– 50. Gauthier, B. D. M., Franssen, R. C. W. M. & Drei, S. 2000. Fracture networks in Rotliegend gas reservoirs of the reservoirs of the Dutch offshore: implication for reservoir behaviour. Geologie en Mijnbouw/ Netherlands Journal of Geosciences, 79, 45– 57. Geiss, B., Kremer, Y., Van Koppen, J. K. J. & Bertotti, G. 2009. Field compartmentalisation by subtle transfer faulting an example from blocks K4/K5 offshore Netherlands. European Association of Geoscientists and Engineers, 71th Conference and Technical Exhibition, June 8 –11, Amsterdam, the Netherlands, extended abstracts W023, 5. Glennie, K. W. (ed.) 1998. Petroleum Geology of the North Sea, Basic Concepts and Recent Advances. Blackwell Science Ltd, London. Knott, S. D. 1993. Fault seal analysis in the North Sea. American Association of Petroleum Geologists Bulletin, 77, 778–792. Leveille, G. P., Knipe, R. et al. 1997. Compartmentalization of Rotliegendes gas reservoirs by sealing faults, Jupiter Fields area, southern North Sea. In: Ziegler, K., Turner, P. & Daines, S. R. (eds) Petroleum Geology of the Southern North Sea: Future Potential. The Geological Society of London, Special Publication, 123, 87– 104. Lindsay, N. G., Murphy, F. C., Walsh, J. J. & Watterson, J. 1993. Outcrop studies of shale smear

on fault surfaces. In: Flint, S. T. & Bryant, A. D. (eds) The Geological Modelling of Hydrocarbon Reservoirs and Outcrop Analogues. International Association of Sedimentologists, Special Publication, Wiley-Blackwell, Oxford, 15, 113– 123. Mauthe, G. 2003. Kompartmentbildende Verwerfungen infolge Kataklase (Rotliegend, NW-Deutschland). Erdo¨l Erdgas Kohle, 119, 12–17. Oudmayer, B. C. & De Jager, J. 1993. Fault reactivation and oblique-slip in the Southern North Sea. In: Parker, J. R. (ed.) Petroleum Geology of Northwest Europe. Proceedings of the 4th Conference. Geological Society, London, 2, 1281– 1290. Schowalter, T. T. 1979. Mechanics of secondary hydrocarbon migration and entrapment. American Association of Petroleum Geologists Bulletin, 63, 723– 760. Smith, D. A. 1966. Theoretical considerations of sealing and non-sealing. American Association of Petroleum Geologists Bulletin, 50, 145–172. Udink, H. G. 1968. Reservoir behaviour and field development. Symposium Groningen gas field, March 15–16, Groningen, Verhandelingen Koninklijk Nederlands Geologisch en Mijnbouwkundig Genootschap, Geol. Serie, 25, 35–42. Van Adrichem Boogaert, H. A. & Kouwe, W. F. P. [comp.] 1997. Stratigraphic nomenclature of the Netherlands, revision and update by RGD and NOGEPA. Mededelingen Rijks Geologische Dienst, nieuwe serie, 50 (a – j). Van der Molen, I., Zijlstra, E. B., Okkerman, J. A. & Reemst, P. H. M. 2003. Compartmentalisation in Rotliegend gas fields, examples from offshore and onshore the Netherlands. In: Konstanty, J. J. C., Grauls, D. & Gelder, E. N. (conv). Fault and Top Seals: What do we know and where do we go? European Association of Geoscientists and Engineers, Montpellier, Paper-28. Van Hulten, F. F. N. 1996. Compartmentalized gas reservoirs of the Netherlands. In: Longacre, S., Katz, B., Slatt, R. & Bowman, M. (conv.). Compartmentalized Reservoirs: Their Detection, Characterization and Management, AAPG/EAGE Research Symposium, October 20–23, The Woodlands, Texas, abstract. AAPG, Tulsa. Van Rossum, B. 1975. Aspects of the geology and appraisal/development of the Groningen Gasfield. Erdo¨l-Erdgas Zeitschrift, 91, 254– 256. Van Wijhe, D. H., Lutz, M. & Kaasschieter, J. P. H. 1980. The Rotliegend in the Netherlands and its gas accumulations. Geologie en Mijnbouw, 59, 3– 24. Veenhof, E. N. 1996. Geological aspects of the Annerveen gas field, the Netherlands. In: Rondeel, H. E., Batjes, D. A. J. & Nieuwenhuijs, W. H. (eds) Geology of Gas and Oil Under the Netherlands. Royal Geological and Mining Society of the Netherlands, Kluwer Academic Publishers, Dordrecht, 79–92. Verdier, J. P. 1996. The Rotliegend sedimentation history of the southern North Sea and adjacent countries. In: Rondeel, H. E., Batjes, D. A. J. & Nieuwenhuijs, W. H. (eds) Geology of Gas and Oil Under the Netherlands. Royal Geological and Mining Society of the Netherlands, Kluwer Academic Publishers, Dordrecht, 45–56.

COMPARTMENTALIZATION OF DUTCH GAS FIELDS Watts, N. L. 1987. Theoretical aspects of cap-rock and fault seals for single- and two-phase hydrocarbon columns. Marine and Petroleum Geology, 4, 274–307. Wong, T. E., Batjes, D. A. J. & De Jager, J. (eds) 2007. Geology of the Netherlands. Royal Netherlands Academy of Arts and Sciences, Amsterdam, 354. Yielding, G., Freeman, B. & Needham, D. T. 1997. Quantitative Fault Seal prediction. American

315

Association of Petroleum Geologists Bulletin, 81, 897– 917. Zijlstra, E. B., Reemst, P. H. M. & Fisher, Q. J. 2007. Incorporation of fault properties into production simulation models of Permian reservoirs from the southern North Sea. In: Jolley, S. J., Barr, D., Walsh, J. J. & Knipe, R. J. (eds) Structurally Complex Reservoirs. Geological Society, London, Special Publication, 292, 295–308.

Multi-fault analysis scorecard: testing the stochastic approach in fault seal prediction F. V. CORONA1*, J. S. DAVIS2, S. J. HIPPLER2 & P. J. VROLIJK2 1

ExxonMobil Production Deutschland GmbH, Riethorst 12, 30659 Hannover, Germany

2

ExxonMobil Upstream Research Company, P.O. Box 2189, Houston, Texas 77252, USA *Corresponding author (e-mail: [email protected]) Abstract: Multi-fault analysis is an ExxonMobil stochastic tool for analysing the impact and sensitivities of stratigraphic uncertainty and variability on cross-fault leakage of hydrocarbons in faulted traps. This juxtaposition-based method provides quantitative prediction of hydrocarbon contact levels through a complex system of structural spills and juxtaposition leak points in traps with stacked reservoir systems and one or more faults. Validation of the Multi-fault analysis technology was carried out by comparing pre-drill predictions to post-drill results from 41 faulted exploration prospects drilled from 1994–2001. Of the 41 prospects, 29 were valid tests in which we made 22 successful predictions. Of the 22 successful outcomes, 11 were discoveries and 11 were dry wells. Some of the dry wells were drilled assuming the presence of sealing fault-zone material to trap hydrocarbons despite a Multi-fault analysis failure prediction. The seven Multifault failures comprise four predicted successes that were failures and three predicted failures that were successes. Most of the Multi-fault prediction failures can be attributed to data quality and uncertainty; however, some may be associated with sealing fault-zone material. Other considerations in fault seal analysis (i.e. dip leak along faults and sealing fault zone materials), model input uncertainties, and using drill-well learnings are also discussed.

Fault seal analysis has been used since the 1960s to predict the impact of faults on the flow or storage of fluids in hydrocarbon reservoirs. Initial attempts at fault seal analysis concentrated on defining pressure or hydrocarbon column height differences across faults and using those observations to infer the sealing behaviour of faults (e.g. Smith 1966). Such studies used cross-sections constructed at high angles to the strikes of faults to understand the connections of permeable units across the faults. If disparate pressures or hydrocarbon fluid contacts were observed where a cross section provided evidence of a potential sand-to-sand juxtaposition, then it was assumed that the fault zone itself sealed in order to isolate the sands on each side of the fault from each other. However, the cross-section approach gives only a one-dimensional look at a fault (i.e. a single profile view along the fault), and the sand-to-sand juxtapositions across it, and as such, failed to define possible juxtapositions everywhere along the strike length of a fault. As well, early attempts at fault seal analysis were hampered by map quality and accuracy. In the 1960s and 1970s, depth-structure maps were constructed from variable quality two-dimensional seismic data and/or from well penetration data (formation or sequence tops), making derivative analyses, such as fault seal analysis, of suspect quality.

With the development by Allan (1989) of a method for making fault plane profiles (so-called Allan diagrams), sand-to-sand juxtapositions along the length of faults could be defined in much better detail. Fault plane profiles are cross-sections drawn parallel to the strike of a fault onto which the stratigraphic cutoffs from both the hanging wall and footwall are projected. The fault plane profile, therefore, allows a geologist to examine at what depths sand-to-sand cross-fault juxtapositions occur, the areal extent of such juxtapositions, and how hydrocarbon fluid distributions or flow behaviours relate to the juxtapositions. The use of fault plane profiles became an industry standard practice in the 1990s. During the 1990s, fault plane profiles were commonly combined with an estimate of what type of fault zone materials were likely to be found along the fault at sand-to-sand juxtapositions, which in turn was accompanied by an estimate of the capillary sealing capacity of those fault zone materials (e.g. Yielding et al. 1997; Freeman et al. 1998; Yielding 2002; Bretan et al. 2003). From the understanding of where sand-to-sand juxtapositions existed along a fault and the capillary sealing capacity and permeability of the fault zone materials, predictions of fault behaviour impacting both exploration and production time-scale fluid flow were made (e.g. trapped hydrocarbon column


318

F. V. CORONA ET AL.

heights, fault transmissibility; e.g. Walsh et al. 1998; Manzocchi et al. 1999; Wehr et al. 2000). The implementation of fault seal analysis by using fault plane profiles and, optionally, predictions of fault zone materials, typically is on a fault-by-fault basis. The fault plane profiles are made using a deterministic stratigraphic model constructed from well logs or interpretation of seismic data. Commonly, the stratigraphic model is coarse, often at the formation or reservoir group level, and generally ignores the possibility of coarse-grained intervals in the seal packages of the model (i.e. seal is homogenous shale). Through the 1980s and early 1990s, ExxonMobil found that this approach yielded poor predictive capability. In its place a new methodology, Multi-fault analysis, was developed that allows simultaneous analysis of sand-tosand connections across multiple faults, includes use of high resolution deterministic and stochastic stratigraphic models, and implements a careful book-keeping system to track fault juxtaposition and spill controls on fill (summarized below; see James et al. 2004 for a detailed discussion of Multifault analysis). This paper presents results of a 1994–2001 validation study of ExxonMobil’s stochastic Multi-fault analysis technology as applied to geological timescale fault seal analysis. Three key issues recognized as critical to the success of the validation study are the objective use of drill-well learnings to evaluate geological time-scale fault seal, the importance of structural and stratigraphic uncertainty in fault-seal analysis, and consideration of alternative, admissible geological interpretations. The validation study shows that the stochastic approach returns a 76% success rate for predictions and, thus, we believe validate the Multi-fault approach. The validation study also shows that hydrocarbon fill in faulted traps in many basins worldwide appears to be controlled primarily by cross-fault juxtapositions and that, although faultzone capillary seals do occur and can produce important seals in hydrocarbon traps, they are few and difficult to predict.

Multi-fault analysis Multi-fault analysis is a stochastic approach for analysing faulted traps (James et al. 2004; Fig. 1). This technology was developed to address the impact of stratigraphic uncertainty and variability on the depths and locations of cross-fault, sand-to-sand juxtapositions. Stratigraphic variability is simulated by stochastically constructing 1D stratigraphic columns that satisfy inputs of mean seal/leak bed thicknesses and proportion (i.e. net-to-gross). Structural uncertainty is addressed through variation of

the input data from depth maps or faulted framework models. Moreover and as important, this approach allows us to evaluate the effect of juxtaposition on multiple faults simultaneously. The major assumption of this method is that all sand-to-sand fault juxtapositions allow hydrocarbons to leak across faults on a geological time-scale. The process of hydrocarbon fill and spill in traps can be very complicated, even in simple structures (James et al. 2004). In traps with numerous faults and stacked reservoirs, it is extremely difficult to use traditional methods, such as one-at-a-time faultplane profiles (Allan 1989; Badley et al. 1991). The difficulty is not the construction of a single fault plane profile, rather it is in the devising of a method to keep track of all the structural spills and juxtaposition leak points that ultimately may control hydrocarbon contact levels in structurally and stratigraphically complex traps. Multi-fault analysis provides the computerized methodology required to track fill and spill at structural spills and juxtaposition leak points of these complex traps, and to determine resulting fluid contact levels within these traps for multiple stochastic stratigraphic model realizations. Multi-fault analysis yields quantitative prediction of hydrocarbon contact levels through the complex system of structural spills and juxtaposition leak points in faulted traps using stochastic stratigraphy and variable structural models (Fig. 1). The results are used as direct input to assessment risking and sizing and/or to rank and grade hydrocarbon prospects and leads. Multi-fault analysis allows the user to quickly distinguish big, less risky prospects from small, more risky leads, and identify key controls on trap fill (e.g. a high change in throw segment along a trap-bounding fault that may constrain the fill, or small crestal faults that may prevent stacked pay to form; see James et al. 2004). In post-drill analyses, Multi-fault analysis is used to validate or determine the main control on trap fill, whether it is fault juxtaposition, enhancement by fault-zone materials, or other leak mechanisms (e.g. dip leak along faults, top seal constraints, fault-related fracturing, etc.). We will explore later the benefits of using post-drill analysis and drill-well learnings to evaluate fault seal and controls on trap fill. Multi-fault analysis includes some limitations and simplifying assumptions for the benefit of computer implementation. Beside the assumption of sand-to-sand fault juxtapositions as hydrocarbon leak windows, the stratigraphic models are laterally continuous and uniform, and the structure vertically uniform. Other assumptions include no dip leak along fault zones, no sealing fault gouge, and fully hydrocarbon-charged traps. Multi-fault analysis does not account for hydrocarbon migration

MULTI-FAULT ANALYSIS SCORECARD

319

Fig. 1. Illustrative diagram of the Multi-fault analysis approach (modified after James et al. 2004). Using structural depth maps and stochastic stratigraphic models for a faulted prospect or lead, Multi-fault analysis yields a chance of success of finding hydrocarbons for the prospect or lead and range of hydrocarbon column heights that can be directly incorporated into prospect/lead assessment.

shadows, or the dynamic effects of production. The program provides estimates of total hydrocarbon column height, and does not distinguish oil from gas. Multi-fault analysis provides a prediction based on the static geological time-scale (i.e. thousands to millions of years v. months to years in the production time-scale). The advantage of these limitations and assumptions is that they allow complex models to be generated rapidly and modified quickly for sensitivity analysis (James et al. 2004). More complex structure and stratigraphy (e.g. growth faults, channelized reservoirs, unconformities, etc.) can be realized with intelligent application of Multi-fault analysis to minimize the impact of the above simplifications. Although Multi-fault analysis does not account for dip leak along fault zones or additional hydrocarbon column supported by capillary fault

gouge, our observations in post-drill analyses (see below) indicate that these factors commonly are secondary compared to cross-fault juxtaposition. However, these variables can be considered later in the assessment process if necessary. Moreover, we recommend a broader integrated trap analysis approach and workflow as summarized by Corona (2005).

Multi-fault analysis validation 1994 – 2001 After some early success and experience applying Multi-fault analysis in our exploration program, we undertook a validation study to document the degree of success of this new technology. Thus, from 1994 through 2001, we carried out pre-drill predictions followed by post-drill comparison of

320

F. V. CORONA ET AL.

Table 1. General location of prospects incorporated in the ExxonMobil Multi-fault analysis (MFA) validation study from 1994 to 2001. Of the 41 prospect applications, 29 were considered valid tests of the technology in which 22 were successful predictions for a 76% success rate. Note that these 41 prospects represent a small portion of ExxonMobil exploration drilling program during this time period Global area

MFA successes

MFA failures

Non-tests

Sucess rate

North America South America Sub-Saharan Africa Far East Middle East Europe

13 2 5 1 1 0

4 0 2 1 0 0

8 3 1 0 0 0

76.5% 100.0% 71.4% 50.0% 100.0% N/A

Total

22

7

12

75.9%

41 faulted exploration prospects in many parts of the world (Table 1). This number represents a small portion of ExxonMobil exploration drilling program during this time period. Of the 41 prospects drilled, 29 were considered valid tests of the technology. Of those 29 valid tests, 22 were successful predictions, yielding a 76% success rate (Fig. 2). The 12 non-tests were deemed as such for reasons such as lack of reservoir, trap, top seal, and/or charge. A Multi-fault analysis success is defined as an outcome in accord with the predicted chance of success (i.e. a low chance of

success prediction followed by a dry well, a high chance of success prediction followed by a discovery), and discovered hydrocarbon columns within the predicted range of column height. Of the 22 successful predictions, 11 were discoveries and 11 were dry wells. Some of the dry wells were drilled on the assumption of sealing fault-zone material to trap economic hydrocarbon volumes despite a Multifault analysis failure prediction. The seven Multifault failures comprise four predicted successes that were failures and three predicted failures that were successes. Most of the Multi-fault prediction

Fig. 2. Chart showing the scorecard of the Multi-fault analysis (MFA) validation study from 1994 to 2001. Of the 41 prospect applications, 29 were considered valid tests of the technology in which 22 were successful predictions for a 76% success rate. Refer to the text for discussion.


321

failures can be attributed to either flawed input (data quality, limitations, or uncertainty) or failure of the technology (i.e. inputs seem reasonable but the assumption of fault juxtaposition controlling hydrocarbon fill did not provide a correct prediction). As such, some of the predicted failures that were successes may be associated with sealing fault-zone material (to be discussed in the next section). In conclusion, the Multi-fault analysis stochastic approach yielded a high success rate based on both drill-well successes and failures and the range of predicted hydrocarbon column heights of these tests and, thus, we believe validates this technology. This validation and numerous applications since imply that hydrocarbon fill in faulted traps in many basins worldwide appear to be controlled primarily by cross-fault juxtaposition (also see James et al. 2004). Other potential fault seal factors such as dip leak along fault zones or fault-zone capillary seals may be secondary compared to fault juxtaposition. The Multi-fault analysis approach addresses the first-order questions of fault seal analysis (i.e. the geometry of permeable connections and the uncertainty upon which those assessments are based), and is an integral component of our workflow for evaluation of faulted exploration prospects and near-field wildcats.

Other considerations in fault seal analysis There are two major fault seal analysis considerations not accounted for in stochastic Multi-fault analysis: (1) dip leak along fault zones (instead of across fault zones); and (2) additional hydrocarbon column supported by capillary sealing fault-zone material (Fig. 3). Dip leak of hydrocarbons may occur in areas along faults where the effective differential stresses and mean stress cause the ratio of shear to normal stress to exceed the governing frictional failure criterion (e.g. Barton et al. 1995; Sibson 1995; Finkbeiner et al. 1997, 2001; Wiprut & Zoback 2002; Wilkins & Naruk 2007). In this model, fault-zone frictional failure is accompanied by dilation and hydrocarbon leak or fluid flow along the fault. Where this process is suspected to be important based on post-drill analyses or other indications, it can be accounted for with a separate analysis (James et al. 2004). Fault-zone material can produce important geological time-scale capillary seals in hydrocarbon traps, and clear examples have been recognized in our post-drill analyses (e.g. Corona 2005). Numerous other authors have made similar observations and interpretations along faults in basins worldwide (e.g. Weber & Daukoru 1975; Weber et al. 1978; Smith 1980; Bouvier et al. 1989; Nybakken 1991; Jev et al. 1993; Knott 1993; Gibson 1994;

Fig. 3. Schematic profiles showing two major fault seal analysis considerations not accounted for in stochastic Multi-fault analysis: (a) dip leak along fault zones; and (b) additional hydrocarbon column supported by capillary seal material in the fault zone (i.e. CBP, capillary break-through pressure). Yellow represents reservoir and green the associated hydrocarbons; green and blue bars in the well traces represent hydrocarbons and water, respectively, encountered in the hypothetical wells; brown bar along the fault trace represent capillary seal material in the fault zone.

Fristad et al. 1997; Leveille et al. 1997a, b; Yielding et al. 1997; Alexander & Handschy 1998; Davies et al. 2003; Gibson & Bentham 2003). Moreover, several authors have published methodologies and algorithms to model fault-zone material or gouge and to predict the sealing potential of such material (e.g. Weber et al. 1978; Lindsay et al. 1993; Antonellini & Aydin 1994; Fulljames et al. 1997; Knipe 1997; Knipe et al. 1997; Yielding et al. 1997; Fisher & Knipe 1998; Freeman et al. 1998; Yielding 2002; Bretan et al. 2003). We accept examples of fault-zone capillary seals only after alternative, plausible geological interpretations that may allow for juxtaposition seal are explored (e.g. Corona 2005). Our observation is that many apparent fault zone seals are the result of poor quality structural and stratigraphic input

322

F. V. CORONA ET AL.

models to juxtaposition analysis (Corona 2005), as well as past failures to recognize simple geometric controls on contacts, such as breakovers (Vrolijk et al. 2005), that can yield contact distributions reminiscent of sealing fault zones. We have observed that small errors or uncertainties in structure and stratigraphy may have large impacts on the prediction of juxtapositions across faults, leading to incorrect fault trap and seal analysis results (James et al. 2004). Greater benefit is gained by in questioning and testing our structural and stratigraphic interpretations and their impact on fault-juxtaposition seal (to be discussed in the next section) than implementation of predictions of the existence and sealing capacity of fault zone materials. It is therefore imperative that we recognize and capture these uncertainties, and use high quality structural and stratigraphic interpretations in our fault trap and seal analyses. Our internal post-drill analyses clearly show that capillary sealing fault zones do occur in faulted traps in basins worldwide. However, in more than 25 years of regional to sub-regional fault-seal studies (e.g. Gulf of Mexico, Niger Delta, Congo Basin, North Sea, Gippsland Basin, Malay Basin), our attempts to develop a robust methodology for predicting the locations and seal capacity of sealing fault zones have mostly failed. Applying both proprietary and widely used industry methodologies and algorithms, such as shale gouge and shale smear analyses, we consistently failed to predict the existence or non-existence of, the location of, and the additional hydrocarbon column heights held behind, sealed sand-to-sand juxtaposition windows. While fault zone materials likely exist in faults associated with many or most faulted traps, the high success rate of juxtaposition analysis, and the high failure rate of fault zone material analysis, is most likely explained by continuity, or lack of continuity, of the sealing material. Field exposures of faults indicate that fault rock development is extremely variable and that the associated fault rock material usually contains holes that would permit significant leakage over the geological time-scale (Wehr et al. 2000; Aydin & Eyal 2002; Doughty 2003). In these cases, hydrocarbon fill would be controlled by the shallowest hole of the critical sand-on-sand juxtaposition window, and the location of the hole is almost certainly random as observed and concluded in our regional fault-seal studies. In a recent study Dee et al. (2007) suggested that a deterministic fault seal approach incorporating capillary fault gouge as an important sealing mechanism results in a similar post-drill analysis for the Ling Gu to the one published by James et al. (2004) using Multi-fault analysis. This outcome may be significant or parsimonious. One important element in a trap structure such as this one is the

complex fault-juxtaposition ensemble that creates a large number of juxtaposition leak points. It was for this reason that Multi-fault analysis was created: the computer is much better adapted to long, tedious bookkeeping than is a human. Our experience over the years is that even if we hand-edit individual leaks identified by the Multifault program, there are multiple, redundant leaks at a similar elevation, sometimes on the same fault and sometimes on other faults, that all need to be closed before we achieve a significantly different trap fill outcome. Although Dee et al. (2007) identify a weak capillary link on the fault analysed, the controlling leak may actually occur elsewhere. Dee et al. (2007) further suggested that the choice of well location on the Ling Gu structure and the resulting small gas columns inferred by the Multi-fault approach may miss a possible favorable sequence of up-dip gas accumulations inferred by the deterministic fault seal approach. If such a situation was common, then we would expect to see a systematic bias in our post-drill results, something which, so far, has yet to reveal itself. Numerous Multi-fault post-drill analyses over the last 10–15 years fail to consistently predict smaller hydrocarbon columns than are observed, as might be expected if fault-zone materials act as significant capillary seals (also noted in James et al. 2004). Therefore, we conclude from these post-drill analyses that fault juxtaposition is the primary control on hydrocarbon fill in faulted traps on the geological time-scale and that fault-zone seals should be considered as a secondary control. Anecdotally, fault-zone seals in our global database may account for anywhere from 10% to as much as 30% of a fault-trap play in a given basin. In contrast, we believe that fault-zone materials play a significant role in forming transmissibility baffles to hydrocarbon flow during the production time-scale (i.e. months to years to 10s years), and that the degree of fault transmissibility can be estimated based on fault-zone processes, environment of formation, and deformation history (Wehr et al. 2000; Myers et al. 2007).

Model input uncertainties Uncertainty in fault seal analysis grows from imperfect data and interpretation of those data. In particular, stratigraphic and structural input models, which are the basis for any fault seal analysis, are based on the imperfect data and associated interpretations. The primary impact of a flawed structural and stratigraphic interpretation is inaccurate juxtaposition of reservoir and sealing layers across faults. Inaccurate prediction of juxtapositions leads to erroneous interpretation and prediction of hydrocarbon distributions, fluid pressures, and flow behaviour


in the subsurface. These, in turn, result in over- or under-assessment of discovered resources, poorly designed field development plans, and misinterpretation of production data leading to poor reservoir management in producing fields. Traditional fault seal analysis starts with predictions of cross-fault stratigraphic juxtapositions by combining structural geometry with stratigraphic architecture. In our experience, starting with carefully constructed structural and stratigraphic models to evaluate juxtapositions of reservoir and seal units across faults reduces, or eliminates, the need to introduce higher uncertainty predictions regarding the presence, distribution, and quality of capillary sealing material in a fault zone. Dee et al. (2007) discuss the sensitivity of the Multi-fault analysis to choice of Vshale cutoff applied to discretize a stratigraphic column into leak and seal components and proposed that a higher Vshale cutoff of 0.5 would have led to substantially different hydrocarbon column interpretations. However, all types of fault-seal analysis are subject to this initial definition of what is a leak or seal bed. Multifault analysis relies only on the choice of Vshale cutoff as a parameter to be calibrated, and that calibration step was done during the development of the technology. Alternative Vshale cutoffs are evaluated

323

to learn of sensitivity to this parameter, but only in rare instances where independent calibration is undertaken is a different base-case value of Vshale cutoff applied. If there were a large variability in Vshale cutoff for leak/seal behaviour, we would expect either random or systematic error in our predrill predictions, something which is undocumented in this study. Because the deterministic fault seal methodology (e.g. Yielding 2002) employs a number of parameters that have undergone calibration over more than a decade, the Multi-fault analysis approach is inherently more reproducible (i.e. independent of operator judgement). Further commonly observed phenomena, such as offset hydrocarbon contacts and different fluid pressures across faults, may lead to hypotheses about sealing materials in fault zones (Fig. 4). Often we find that within the range of structural uncertainty and assumptions, juxtaposition models can be built to satisfy the observations and are testable within the seismic data from which they ultimately are derived (James et al. 2004; Corona 2005). Furthermore, recognition that common geometric elements associated with juxtaposition windows (e.g. fault-juxtaposition spill at the top of, and breakover at the base of, a juxtaposition window; Vrolijk et al. 2005) can yield both offset

Fig. 4. Schematic profiles illustrating potential scenarios related to structural and stratigraphic uncertainties. When well and seismic interpretations yield offset contacts across a fault (a), it is commonly assumed to be associated with sealing fault-zone material. However, within the range of structural and stratigraphic uncertainty, there may be possible alternative and permissible interpretations that explain this offset contact by juxtaposition (b, structural uncertainty; c, stratigraphic uncertainty). Nevertheless, the interpretation of sealing fault-zone material cannot be excluded (d). See Figure 3 for explanation of pattern items; in addition, brown represents seal lithologies and the blocky arrow represents a fault-related spill.

324

F. V. CORONA ET AL.

Fig. 5. Schematic profile illustrating common geometric elements associated with fault-juxtaposition windows: spill at the top and breakover at the base of the juxtaposition window (Vrolijk et al. 2005). Yellow, green, and light red represent reservoir and associated oil and gas, respectively; red, green and blue bars in the well traces represent gas, oil and water, respectively, encountered in the hypothetical wells. Such different contacts across a fault have led to the interpretation of sealing fault-zone materials; however, such contacts also may be related to reservoir-trap geometry and fluid separation at spills and/or breakovers that require sand-on-sand juxtaposition to be non-sealing to hydrocarbons and fluid flow (e.g. oil is connected and in pressure communication across the illustrated fault, as indicated by the wide double-headed green arrow).

contacts and different fluid pressure regimes across faults without the need for ad-hoc prediction of fault zone sealing materials (Fig. 5). An advantage of the juxtaposition approach to fault seal analysis, as Multi-fault analysis, is that it frequently guides revised geological interpretations that can be tested by examination of seismic data and drilling. That is, the methodology compels us to consider alternative geological interpretations. The distribution and quality of fault-zone materials are rarely directly tested. However, a fault-zone seal interpretation based on the imperfect stratigraphic and structural input models in post-drill analysis cannot be excluded from scenario consideration (see Fig. 4). To help reduce the uncertainties on fault-seal or trap-fill controls, and to develop criteria for evaluating, risking, and sizing faulted traps, we suggest objective analysis and use of all drillwell learnings within a basin or field area.

Using drill-well learnings In this section, we explore an example of a regional drill-well learnings study that was used to identify and understand the geological controls on trap fill and the potential for fault-zone seals along normal

faults in the Lauwerszee Trough, NE Netherlands (Fig. 6; Corona 2005). The key points emphasized in this section are the objective use of drill-well learnings and the importance of uncertainty in structural (and stratigraphic) interpretation and mapping. Careful and impartial consideration of these elements is essential to derive valid and reliable fault trap analysis results and criteria to be used in evaluating these traps. A limitation of this section is the absence of the defining data (e.g. well logs, 3D seismic data, field reports, etc.) that reside at Nederlandse Aardolie Maatschappij B. V. (NAM), Shell International, Shell UK, and ExxonMobil and are not shown for proprietary reasons. Corona (2005) analysed 30 drilled faulted prospects in the Lower Permian Rotliegend gas play of the Lauwerszee Trough in the onshore area of northeastern Netherlands (Fig. 6). The fault trap play is composed of the prolific, high net-to-gross sandstones (.85%) of the Slochteren Formation in which gas in entrapped in tilted fault blocks and sourced from the underlying Carboniferous Coal Measures. The overlying Upper Permian Zechstein evaporitic rocks provide the top and lateral seal for the fault traps. Structural depth maps used in this study are based on the interpretation of good quality pre-stack, depth-migrated, 3D seismic data (not shown for proprietary reasons). Map uncertainties exist in places related to time-to-depth conversion of the varying lithologies and associated velocities in the Upper Permian Zechstein evaporite interval. Of the 30 examined drilled prospects in the study area, 22 are associated with hydrocarbon accumulations, yielding an exploration success rate of 73% (Fig. 7). Almost all the faulted traps are underfilled with respect to synclinal spill, and most traps appear to be filled to or near to a fault-juxtaposition spill. Moreover, associated hydrocarbon–water contacts become shallower from one end of the trough to the other through these spills, defining the direction of hydrocarbon fill and spill through the area (Fig. 8). The drill-well learnings from these drilled faulted prospects are key in understanding the geological controls on trap fill and in the evaluation and prediction of geological time-scale fault seal. The results also allow the formulation of a fault trap analysis workflow that can be applied to undrilled prospects and leads in the region. Trap fill, based on the analysis of existing structural depth maps and hydrocarbon–water contact data, is interpreted to be controlled primarily by fault juxtaposition for 22 of the 30 drilled faulted prospects (Figs 7 & 8). These include seven dry or uneconomic drill wells. The hydrocarbon column heights of the remaining 15 (juxtapositioncontrolled) fault-trap discoveries range from 60 to 190 m with an average column height of 115 m.


325

Fig. 6. Location map of the Southern North Sea and the Netherlands showing Mesozoic basins, major faults, and structural features including the Lauwerszee Trough in NE Netherlands (from Corona 2005; modified after Rondeel et al. 1996).

Fig. 7. Chart showing trap closure (bars) and associated hydrocarbon fill of 22 discoveries of 30 drilled faulted prospects in the regional study area. Green represents hydrocarbon fill to the fault-juxtaposition spill, orange is the fill below the fault-juxtaposition spill requiring sand-on-sand fault seal, and blue is the water leg to the mapped synclinal spill. Field A and Field B identified on the chart are discussed in the text. Average trap fill in this region is about 50% of the available structural closure (modified after Corona 2005).

326

F. V. CORONA ET AL.

Fig. 8. Map of Rotliegend fault traps in the main part of the Lauwerszee Trough, NE Netherlands, showing gas– water contacts, fault-block juxtaposition spill points, and potential fault-zone seal segments (modified after Corona 2005). The associated hydrocarbon–water contacts become shallower from NW to SE through these spill points, defining the direction of hydrocarbon fill and spill through the area.

Average trap fill is around 50% of the available closure height (from structural crest to synclinal spill). Seven other fault-trap hydrocarbon accumulations appear to be filled deeper than the faultjuxtaposition spill as mapped and require sand-on-sand fault seal (Fig. 7). Retained fault-seal hydrocarbon columns of these accumulations in addition to the fault-juxtaposition-seal column are relatively small, mostly less than 50 m with one exception of about 90 m. Two examples of potential fault-zone sealed traps, designated as Field A and B (Fig. 7), are described in more detailed below highlighting the key elements that lead to the identification criteria of such fault seals in this region.

Field A fault-zone seal Field A lies near the central part of the study area and is typical of the faulted trap play in this basin (Corona 2005). The hydrocarbon accumulation is bound to the south and east by normal faults that juxtapose it against the regional seal interval (Fig. 9). However, along the south-bounding fault system as mapped, a possible 50 m, sand-on-sand, fault-zone seal was identified along a short (c. 600 m) lowside fault segment that juxtaposes the reservoir against itself.

The fault-plane profile along the south-bounding fault segment at Field A shows the geometry of the sand-on-sand, fault-juxtaposed hydrocarbon accumulation (Fig. 10). The hydrocarbon-bearing interval is c. 95% net-to-gross clean sandstone with good porosity. As mapped and shown in this profile, sand-on-sand juxtaposition geometry is associated with a short fault length of about 600 m and a relatively small juxtaposition area of c. 23 000 m2. This observation can also be made for two other discoveries in this basin in which the key fault-seal trap segments are generally less than 1 km in length and the associated sand-on-sand juxtaposition area is less than 25 000 m2. Most of the 22 juxtaposition-controlled fault traps are associated with longer fault lengths and larger sand-on-sand juxtaposition areas (Fig. 8). An explanation for the apparent sand-on-sand fault seal at Field A, as well as the two other discoveries referenced above, may be the formation of cataclastic or cemented cataclastic fault rocks derived from the clean, porous, and permeable sandstone (shale gouge is not a factor since there is little shale in the reservoir interval). Such fault rocks in this area have been observed in core, and laboratory capillary pressure measurements of these fault rocks indicate values comparable to typical seal


Fig. 9. Top reservoir structural depth map of Field A in the regional study area (modified after Corona 2005). Estimated hydrocarbon– water contact (HCWC) from well-log and pressure data is about 4080 mss (+5 m); hydrocarbon fill is shown in green. The south-bounding fault in red is the key structural seal element shown in the fault-plane profile in Figure 10.

327

lithologies (e.g. Antonellini & Aydin 1994; Knipe 1997). However, these measurements do not address the continuity of these fault rock materials where holes may occur anywhere and allow hydrocarbons to leak during the geological time-scale, as previously noted (Wehr et al. 2000; Aydin & Eyal 2002; Doughty 2003). Therefore, the faultjuxtaposition geometric considerations may be significant in discerning potential sand-on-sand sealing fault segments in that small juxtaposition area and short fault length may increase the chance of a continuous fault-zone seal. A closer look at the structural depth map (Fig. 11) reveals a possible alternative interpretation to explain the trap fill at Field A. The key sealing south-bounding fault system has mapping uncertainties including fault throw polarity changes, complex faulting, inconsistent cross-fault structure, and structural contours that may close and not reach the bounding fault. Minor changes in the map and/or horizon interpretation may eliminate the sand-on-sand juxtaposition and the need for a faultzone seal, and explain trap fill by fault juxtaposition (a revised permissible map to show this alternative scenario is not shown as we did not have access to the 3D seismic data for reference). This is an example where the interpreted fault-zone capillary seal is made suspect by structural uncertainty. This does not exclude, however, the fault-zone capillary seal scenario (Fig. 4).

Fig. 10. Fault-plane profile along the south-bounding fault of Field A in the regional study area (modified after Corona 2005). Estimated hydrocarbon–water contact (HCWC) is about 4080 mss and the hydrocarbon fill is shown in green.

328

F. V. CORONA ET AL.

Fig. 11. Enlargement of the southern part of the top reservoir structural depth map of Field A highlighting potential mapping uncertainties: (A) fault throw polarity change; (B) complex faulting and inconsistent cross-fault structure; and (C) structural contours that may close and not reach the bounding fault.

Field B fault-zone seal Field B is located about ten kilometers to the E–SE from Field A in the same play. The faulted trap is bound to the NE by the footwall of a normal fault along which the main reservoir sandstone is juxtaposed against the regional seal. To the SE the trap

is bound by the hanging wall of a normal fault which juxtaposes the reservoir against itself (Fig. 12; Corona 2005). As mapped, the entire SE-bounding fault juxtaposes a 90 m hydrocarbon column at Field B against water-bearing sandstone to the SE. The fault-plane profile along the SE-bounding fault shows the extent of the sand-on-sand, faultjuxtaposed hydrocarbon accumulation (Fig. 13). The hydrocarbon-bearing interval is similar to Field A: about 95% net-to-gross clean sandstone with good and better porosity (again, shale smear should not be a factor). In contrast to the observations and conclusions made at Field A, the associated juxtaposition area is much larger, c. 150 000 m2, with a longer fault length of about 3 km. The fault-juxtaposition geometric relationship of small juxtaposition area (,25 000 m2) and short fault length (,1 km) appears to be contradicted by the presumable lowside fault-seal boundary at Field B. Inspection of the 3D seismic data along the SE-bounding fault at Field B indicates that the fault actually consists of a complex narrow fault zone (seismic data not shown for proprietary reasons). Deflections of the structural contours and the kink in the fault trace on the structural depth map may reflect the occurrence of this narrow fault zone (Fig. 12). Similar features have been observed throughout the Southern North Sea in

Fig. 12. Top reservoir structural depth map of Field B in the regional study area (modified after Corona 2005). Estimated hydrocarbon–water contact (HCWC) from well-log and pressure data is about 3645 mss (+5 m); hydrocarbon fill is shown in green. The SE-bounding fault in red is the key structural seal element shown in the fault-plane profile in Figure 13.


Fig. 13. Fault-plane profile along the SE-bounding fault of Field B in the regional study area (modified after Corona 2005). Estimated hydrocarbon– water contact (HCWC) is about 3645 mss and the hydrocarbon fill is shown in green.

both the UK and Netherlands sectors, and have been attributed to oblique shear during Tertiary inversion (Leveille et al. 1997a). In places, these features appear to form sealing fault-zone elements that may be attributed to fault-zone cataclasis during shearing (Leveille et al. 1997a), or to anhydrite cements that may have been derived from pore fluids originating within the overlying Zechstein evaporites and moved downward into the Rotliegend sands during fault dilation events (Leveille et al. 1997b). Other drilled prospects in the area with similar bounding fault zones, however, are not associated with fault-zone capillary seals. Moreover, Field B appears to be a one-of-the-kind occurrence in the study area and, thus, is not a reliable identification criterion for potential fault-zone capillary seals. Nevertheless, narrow fault zones in the greater regional area do appear to form sealing fault-zone elements (see above), but these occurrence are few and difficult to predict, both for chance of occurrence and column height potential (Corona 2005). A more thorough discussion of this study with other examples can be examined in Corona (2005).

Regional fault trap evaluation workflow Based on the analysis, observations, and conclusions derived from the above study, Corona (2005) developed an integrated technical workflow for the evaluation of faulted traps in this area. This workflow comprises three main components, each in turn consisting of a series of work steps: (1) structural interpretation and mapping; (2) trap and seal evaluation; and (3) sizing and risking. Though the components and associated work steps are shown in a numerical order, iteration is expected between the processes and the work steps within each

329

process in order to test preliminary structural maps for fault seal and assessment implications. This type of integrated approach is used with local modification for all faulted traps in any basin (Corona 2005). Structural interpretation and mapping consists of four steps: (1) build faulted framework structural maps; (2) quality check and interrogate fault interpretation; (3) evaluate structural uncertainty; and (4) consider/construct alternative models (Corona 2005). High quality structural maps from the detailed mapping and integration of both horizon and fault surfaces (i.e. faulted framework structural maps) are essential for valid and reliable fault trap and seal analysis. As noted, small errors or uncertainties in the structural maps may have large impacts on fault trap and seal analysis (e.g. determining chance of success, column heights, and trap size). Completion of the recommended structural interpretation and mapping work steps provide better evaluation of trap size, higher confidence in trap closure, and potential decrease in fault seal uncertainty. While this process results in a higher map confidence, it does not necessarily result in a higher chance of trap closure or trap seal adequacy because the risk severity may increase or decrease with completion of these steps. Trap and seal evaluation consists of three steps: (1) define trap and lateral/top seals; (2) evaluate seal controls using historical drilling results; and (3) analyse fault juxtaposition and other potential seal controls on hydrocarbon accumulation (Corona 2005). In the study area, reservoir sands are contained in tilted fault blocks that define the trap and an overlying regional seal provides the top and lateral seal for the faulted traps. Juxtaposition analysis as well as the analysis of other potential controls on seal should begin with the construction of basic fault-plane profiles along key trap-forming fault segments (e.g. Figs 10 & 13). Fault-plane profiles are also excellent tools to verify cross-fault correlation, to quality check the structural interpretation, and to identify areas of high juxtaposition risk (Needham et al. 1996). However, such profiles have limitations in accounting for stratigraphic variability and stratigraphic or structural uncertainties, determining fill-and-spill architecture of a complexly faulted trap, and treating other mechanisms of fault seal (other than identifying potential fault-zone seals). Multi-fault analysis is an ideal approach to address the first two limitations (James et al. 2004). Based on the regional study presented above, sizing and risking of the faulted traps in that play should be based primarily on fault juxtaposition. Additional but limited hydrocarbon column may occur below the fault juxtaposition spill provided that one or more of the fault-seal criteria described

330

F. V. CORONA ET AL.

above are present (e.g. occurrence of a small juxtaposition area and short fault segment, and possibly the occurrence of a narrow fault zone). This additional column potential should be considered in the high-side assessment case, not as an assessment base case, and should be given a low chance of occurrence (e.g. ,25%). In addition, objective interrogation of the fault interpretation and evaluation of structural and mapping uncertainties, particularly along key trapping fault segments, should be fundamental steps of the fault-trap analysis workflow in this and any area.

Application of Multi-fault analysis to field development and production problems As described by James et al. (2004), the purpose of the Multi-fault approach is to help differentiate the good, mediocre, and poor prospects in undrilled, faulted structure. The focus of the tool is in exploration, but the concentration on geometric uncertainties early in the process helps accelerate the discussion of those same issues during field development and production. The accepted practice after field appraisal is to evaluate the impact of faults on subsurface flow by building one or more deterministic stratigraphic and structural framework(s), evaluating the geometric cross-fault connections that arise from those models and the impact of various fault gouge types on cross-fault flow (e.g. Fisher & Jolley 2007; Myers et al. 2007). These and other recent studies emphasize the primary importance of the fault-juxtaposition framework, a message wholly consistent with the Multi-fault approach. Thus, even though Multi-fault analysis is rarely used in deterministic field development and production problems, its early emphasis on uncertainties in fault juxtaposition geometries helps set the stage for subsequent analyses.

Summary and conclusions Stochastic Multi-fault analysis yielded 22 successful fault-seal predictions of 29 valid tests analysed from 1994 to 2001. This translates to a success rate of 76% and, thus, we believe validates this technology. We now use this methodology on all faulted exploration prospects and near-field wildcats. From our worldwide analysis, hydrocarbon fill in faulted traps appear to be controlled primarily by cross-fault juxtaposition. We acknowledge that fault-zone capillary seals do occur, and can produce important seals in hydrocarbon traps, but these occurrences are few and commonly difficult to predict, both for chance of occurrence and column height potential. Objective use of drill-well learnings is key in understanding geological controls on trap fill and in the

evaluation and prediction of geological time-scale fault seal. In the evaluation of fault-zone capillary seals, potential examples are accepted only after alternative, plausible geological interpretations that may allow for juxtaposition seal are explored. We find greater benefit in questioning and testing our structural and stratigraphic interpretations and their impact on fault-juxtaposition seal. As noted, small errors or uncertainties in structure and stratigraphy may have large impacts on the analysis of fault trap and seal; therefore, it is important to consider these structural and stratigraphic uncertainties in fault-seal analysis. We believe the Multi-fault analysis approach addresses the first-order questions of fault-seal analysis, and the methodology compels us to consider alternative, permissible geological interpretations. The authors would like to thank ExxonMobil Exploration Company and ExxonMobil Upstream Research Company for permission to publish the concepts and results of the Multi-fault analysis tool presented in this paper. We also would like to thank A. N. Bruijn (NAM) and Dave Wilkin (ExxonMobil International Limited, Leatherhead, UK) for permission to publish the Lauwerszee Trough examples presented in this paper, and Herald Ligtenberg (NAM) and Lars-Peter Meier (ExxonMobil International, Leatherhead, UK) who helped expedite this permission. The text was greatly improved by helpful comments by the Geological Society of London, Petroleum Group editorial committee members, particularly Steve Jolley (Shell), Wayne Bailey (Woodside Energy Ltd.), and Sylvie Delisle (Total E&P UK). Randi Lewis and Ilsa Kerscher (ExxonMobil Exploration Company), C. C. Wielchowsky and Ellen Meurer (ExxonMobil Upstream Research Company), C. W. Kiven (ExxonMobil Production Company), and V. M. Corona also provided reviews and their time and dedication are greatly appreciated. The authors also are indebted to our many colleagues, past and present, at ExxonMobil who have applied Multifault analysis and have provided plenty of input and discussion for improvement of this technique and in fault-seal analysis in general.

References Alexander, L. L. & Handschy, J. W. 1998. Fluid flow in a faulted reservoir system: fault trap analysis for the block 330 field in Eugene Island, South addition, offshore Louisiana. American Association of Petroleum Geologists Bulletin, 82, 387–411. Allan, U. S. 1989. Model for hydrocarbon migration and entrapment within faulted structures. American Association of Petroleum Geologists Bulletin, 73, 803– 811. Antonellini, M. & Aydin, A. 1994. Effect of faulting on fluid flow in porous sandstones: petrophysical properties. American Association of Petroleum Geologists Bulletin, 78, 355– 377. Aydin, A. & Eyal, Y. 2002. Anatomy of a normal fault with shale smear: implications for fault seal. American

MULTI-FAULT ANALYSIS SCORECARD Association of Petroleum Geologists Bulletin, 86, 1367–1381. Badley, M. E., Freeman, B., Roberts, A. M., Thatcher, J. S., Walsh, J., Watterson, J. & Yielding, G. 1991. Fault interpretation during seismic interpretation and reservoir evaluation. In: The Integration of Geology, Geophysics, Petrophysics and Petroleum Engineering in Reservoir Delineation, Description and Management. American Association of Petroleum Geologists, Tulsa, Special Volume, SP26, 224–241. Barton, C. A., Zoback, M. D. & Moos, D. 1995. Fluid flow along potentially active faults in crystalline rock. Geology, 23, 683 –686. Bouvier, J. D., Sijpesteijn, K., Kleusner, D. F., Onyejekwe, C. C. & Van der Pal, R. C. 1989. Threedimensional seismic interpretation and fault sealing investigations, Nun River field, Nigeria. American Association of Petroleum Geologists Bulletin, 73, 1397–1414. Bretan, P., Yielding, G. & Jones, H. 2003. Using calibrated shale gouge ratio to estimate hydrocarbon column heights. American Association of Petroleum Geologists Bulletin, 87, 397–413. Corona, F. V. 2005. Fault trap analysis of the Permian Rotliegend gas play, Lauwerszee Trough, NE Netherlands. In: Dore´, A. G. & Vining, B. A. (eds) Petroleum Geology: North-West Europe and Global Perspectives – Proceedings of the 6th Petroleum Geology Conference. Geological Society, London, 327–335. Davies, R. K., An, L., Jones, P., Mathis, A. & Cornette, C. 2003. Fault-seal analysis South Marsh Island 36 field, Gulf of Mexico. American Association of Petroleum Geologists Bulletin, 87, 479–491. Dee, S. J., Yielding, G., Freeman, B. & Bretan, P. 2007. A comparison between deterministic and stochastic fault seal techniques. In: Jolley, S. J., Knipe, R. J., Barr, D. & Walsh, J. J. (eds) Structurally Complex Reservoirs. Geological Society, London, Special Publications, 292, 259– 270. Doughty, 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. Finkbeiner, T., Barton, C. A. & Zoback, 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. Finkbeiner, T., Zoback, M., Flemings, P. & Stump, B. 2001. Stress, pore pressure, and dynamically constrained hydrocarbon columns in the South Eugene Island 330 field, northern Gulf of Mexico. American Association of Petroleum Geologists Bulletin, 85, 1007–1031. Fisher, Q. J. & Jolley, S. J. 2007. Treatment of faults in production simulation models. In: Jolley, S. J., Knipe, R. J., Barr, D. & Walsh, J. J. (eds) Structurally Complex Reservoirs. Geological Society, London, Special Publications, 292, 219–233. Fisher, Q. J. & Knipe, R. J. 1998. Fault sealing processes in siliciclastic sediments. In: Fisher, Q. J. & Knipe, R. J. (eds) Faulting, Fault Sealing and Fluid Flow

331

in Hydrocarbon Reservoirs. Geological Society, London, Special Publications, 147, 117– 134. Freeman, B., Yielding, G., Needham, D. T. & Badley, M. E. 1998. Fault seal prediction: the gouge ratio method. In: Coward, M. P., Daltaban, T. S. & Johnson, H. (eds) Structural Geology in Reservoir Characterization. Geological Society, London, Special Publications, 127, 19– 25. Fristad, T., Groth, A., Yielding, G. & Freeman, B. 1997. Quantitative fault seal prediction: a case study from Oseberg Syd. In: Møller-Pederson, P. & Koestler, A. G. (eds) Hydrocarbon Seals: Importance for Exploration and Production. Norsk Petroleumsforenig, Amsterdam, Special Publications, 7, 107– 124. Fulljames, J. R., Zijerveld, L. J. J. & Franssen, R. C. M. W. 1997. Fault seal processes: systematic analysis of fault seals over geological and production time scales. In: Møller-Pederson, P. & Koestler, A. G. (eds) Hydrocarbon Seals: Importance for Exploration and Production. Norsk Petroleumsforenig, Amsterdam, Special Publications, 7, 51– 60. Gibson, R. G. 1994. Fault-zone seals in siliciclastic strata of the Columbus basin, offshore Trinidad. Association of Petroleum Geologists Bulletin, 78, 1372– 1385. Gibson, R. G. & Bentham, P. A. 2003. Use of fault-seal analysis in understanding petroleum migration in a complexly faulted anticlinal trap, Columbus basin, offshore Trinidad. American Association of Petroleum Geologists Bulletin, 87, 465– 478. James, W. R., Fairchild, L. H., Nakayama, G. P., Hippler, S. J. & Vrolijk, P. J. 2004. Fault-seal analysis using a stochastic multifault approach. American Association of Petroleum Geologists Bulletin, 88, 885– 904. Jev, B. I., Kaars-Sijpesteijn, C. H., Peters, M. P. A. M., Watts, N. L. & Wilkie, J. T. 1993. Akaso field, Nigeria: use of integrated 3-D seismic, fault slicing, clay smearing, and RFT pressure data on fault trapping and dynamic leakage. American Association of Petroleum Geologists Bulletin, 77, 1389–1404. Knipe, R. J. 1997. Juxtaposition and seal diagrams to help analyze fault seals in hydrocarbon reservoirs. American Association of Petroleum Geologists Bulletin, 81, 187 –195. Knipe, R. J., Fisher, Q. J. et al. 1997. Fault seal analysis: successful methodologies, application and future directions. In: Møller-Pederson, P. & Koestler, A. G. (eds) Hydrocarbon Seals: Importance for Exploration and Production. Norsk Petroleumsforenig, Amsterdam, Special Publications, 7, 15– 40. Knott, S. D. 1993. Fault seal analysis in the North Sea. American Association of Petroleum Geologists Bulletin, 77, 778 –792. Leveille, G. P., Knipe, R. et al. 1997a. Compartmentalization of Rotliegendes gas reservoirs by sealing faults, Jupiter field area, southern North Sea. In: Ziegler, K., Turner, P. & Daines, S. R. (eds) Petroleum Geology of the Southern North Sea: Future Potential. Geological Society, London, Special Publications, 123, 87– 104. Leveille, G. P., Primmer, T. J., Dudley, G., Ellis, D. & Allinson, G. J. 1997b. Diagenetic controls on reservoir quality in Permian Rotliegendes sandstones, Jupiter Fields area, southern North Sea. In: Ziegler, K.,

332

F. V. CORONA ET AL.

Turner, P. & Daines, S. R. (eds) Petroleum Geology of the Southern North Sea: Future Potential. Geological Society, London, Special Publications, 123, 105– 122. Lindsay, N. G., Murphy, F. C., Walsh, J. J. & Watterson, J. 1993. Outcrop studies of shale smears on fault surfaces. In: Flint, S. T. & Bryant, A. D. (eds) The Geological Modelling of Hydrocarbon Reservoirs and Outcrop. International Association of Sedimentologists, Special Publications, Wiley-Blackwell, Oxford, 15, 113–123. Manzocchi, T., Walsh, J. J., Nell, P. & Yielding, G. 1999. Fault transmissibility multipliers for flow simulation models. Petroleum Geoscience, 5, 53– 63. Myers, R. D., Allgood, A., Hjellbakk, A., Vrolijk, P. & Briedis, N. 2007. Testing fault transmissibility predictions in a structurally dominated reservoir: Ringhorne Field, Norway. In: Jolley, S. J., Knipe, R. J., Barr, D. & Walsh, J. J. (eds) Structurally Complex Reservoirs. Geological Society, London, Special Publications, 292, 271–294. Needham, D. T., Yielding, G. & Freeman, B. 1996. Analysis of fault geometry and displacement patterns. In: Buchanan, P. G. & Nieuwland, D. A. (eds) Modern Developments in Structural Interpretation, Validation and Modelling. Geological Society, London, Special Publications, 99, 189– 199. Nybakken, S. 1991. Sealing fault traps – an exploration concept in a mature petroleum province: Tampen Spur, northern North Sea. First Break, 9, 209–222. Rondeel, H. R., Batjes, D. A. J. & Nieuwenhuijs, W. H. 1996. Synopsis: Petroleum geology of the Netherlands – 1993. In: Rondeel, H. R., Batjes, D. A. J. & Nieuwenhuijs, W. H. (eds) Geology of Gas and Oil Under the Netherlands. Kluwer, Dordrecht, S3– S20. Sibson, R. H. 1995. Selective fault reactivation during basin inversion: potential for fluid redistribution through fault-valve action. In: Buchanan, J. G. & Buchanan, P. G. (eds) Basin Inversion. Geological Society, London, Special Publications, 88, 3 –19. Smith, D. A. 1966. Theoretical considerations of sealing and non-sealing faults. American Association of Petroleum Geologists Bulletin, 50, 363 –374. Smith, D. A. 1980. Sealing and nonsealing faults in Louisiana gulf Coast salt basin. American Association of Petroleum Geologists Bulletin, 64, 145–172. Vrolijk, P., James, B., Myers, R., Maynard, J., Sumpter, L. & Sweet, M. 2005. Reservoir

connectivity analysis – defining reservoir connections and plumbing. Society of Petroleum Engineers, Paper SPE 93577, 1 –23. Walsh, J. J., Watterson, J., Heath, A. E. & Childs, C. 1998. Representation and scaling of faults in fluid flow models. Petroleum Geoscience, 4, 241– 251. Weber, K. J. & Daukoru, E. 1975. Petroleum geology of the Niger Delta. Proceedings of the Ninth World Petroleum Congress, Volume 2, 209– 221. [Reproduced in AAPG Treatise of Petroleum Geology Reprint Series, No. 6, Traps and Seals I, Structural/Fault-Seal and Hydrodynamic Traps. Compiled by Foster, N. H. & Beaumont, E. A., 1988, 67–79.] Weber, K. J., Mandl, G., Pilaar, W. F., Lehner, F. & Precious, R. G. 1978. The role of faults in hydrocarbon migration and trapping in Nigerian growth fault structures. Tenth Offshore Technology Conference Proceedings, Houston, Paper OTC 3356, 2643– 2653. [Reproduced in AAPG Treatise of Petroleum Geology Reprint Series, No. 6, Traps and Seals I, Structural/Fault-Seal and Hydrodynamic Traps. Compiled by Foster, N. H. & Beaumont, E. A., 1988, 81–91.] Wehr, F. L., Fairchild, L. H., Hudec, M. R., Shafto, R. K., Shea, W. T. & White, J. P. 2000. Fault seal: contrasts between the exploration and production problem. In: Mello, M. R. & Katz, B. J. (eds) Petroleum Systems of South Atlantic Margin. American Association of Petroleum Geologists Memoirs, Tulsa, 73, Chapter 10, 121– 132. Wilkins, S. J. & Naruk, S. J. 2007. Quantitative analysis of slip-induced dilation with application to fault seal. American Association of Petroleum Geologists Bulletin, 91, 97–113. Wiprut, D. & Zoback, M. D. 2002. Fault reactivation, leakage potential, and hydrocarbon column heights in the northern North Sea. In: Koestler, A. G. & Hunsdale, R. (eds) Hydrocarbon Seal Quantification. Norsk Petroleumsforenig, Special Publications, Elsevier, Amsterdam, 11, 203 –219. Yielding, G. 2002. Shale Gouge Ratio – calibration by geohistory. In: Koestler, A. G. & Hunsdale, R. (eds) Hydrocarbon Seal Quantification. Norsk Petroleumsforenig, Special Publications, Elsevier, Amsterdam, 11, 1– 15. Yielding, G., Freeman, B. & Needham, D. T. 1997. Quantitative fault seal prediction. American Association of Petroleum Geologists Bulletin, 81, 897– 917.

Reservoir connectivity analysis of a complex combination trap: Terra Nova Field, Jeanne d’Arc Basin, Newfoundland, Canada F. W. RICHARDS1, P. J. VROLIJK2*, J. D. GORDON3 & B. R. MILLER3,4 1

ExxonMobil Canada Limited, 1701 Hollis St., P.O.Box 517, Halifax, Nova Scotia, B3J 3M8, Canada

2

ExxonMobil Upstream Research Company, PO Box 2189, Houston, Texas 77252, USA 3

Petro-Canada Limited, Scotia Centre, Suite 201, 235 Water Street, St. John’s, NL, A1C 1B6, Canada

4

Petro-Canada – USA, 999 – 18th Street, Suite 600, Denver, Colorado, 80202, USA *Corresponding author (e-mail: [email protected]) Abstract: Terra Nova Oil Field is formed by a complex, structural, stratigraphic and diagenetic combination trap. About 1 billion barrels Stock-Tank Oil-In-Place (STOOIP) are trapped in a faulted plunging anticline, reservoired in stacked, braided fluvial sandstones which are of varying quality and extent due to onlap, cementation and facies transitions. Terra Nova Field is unusual when compared to the other three major hydrocarbon fields in the Jeanne d’Arc Basin in that it lacks a gas leg, and hydrocarbons are confined to the deepest of four major reservoir intervals, the Jeanne d’Arc sandstones. In this reservoir interval, the highest part of the plunging anticline is wet while the lowest part has the deepest fluid contact and longest oil column. Aquifer pressures vary between hydrostatic and significantly overpressured, but are far below leak-off. This apparently unusual distribution of fluids, pressures and contacts can be understood by systematically describing the many compartments within the overall structure, defining the connections between compartments, and building a ‘connectivity diagram’ which ensures a disciplined analysis of the fluid system. This type of approach, used historically by the Operator, Petro-Canada, and formalized as ‘Reservoir Connectivity Analysis’ (RCA) at ExxonMobil, is effective and necessary at multiple scales: basin, field, producing fault block and individual sandstone, and continues to be important at Terra Nova eight years into development and production.

The objective of this paper is to illustrate how fluid and pressure distribution in a highly compartmentalized, complex combination trap can be understood using a conceptually simple, systematic, three component approach to compartment identification and connectivity. This approach, formalized as ‘Reservoir Connectivity Analysis’ (Vrolijk et al. 2005), has had significant utility in delineating, developing and producing Terra Nova oil field, offshore Newfoundland. The initial component of Reservoir Connectivity Analysis (RCA) – compartment description – is achieved by systematically inspecting and integrating conventional interpretations of fluids, pressures, and the geometry of reservoirs and seals. This component is illustrated at Terra Nova by interpreted seismic traverses from the 1997 3D survey, depth structure maps, a pressure –depth plot and a 2D traverse from the 2006 geological model of the field. Fluids, pressures and compartment boundaries are then summarized on a 3D perspective view from the geological model.

The second component of RCA – connection identification – is similarly achieved by systematically considering geometry, buoyancy and the capacity of rocks to seal by mechanical and capillary mechanisms. The results of this component – interpreted connections, spills, breakovers and leaks – are then added to the annotated perspective view of the field. The third component of RCA – constructing a connectivity diagram – is unique to RCA and can become very complex. The connectivity diagram is an attempt to map all compartments and all connections schematically, from fluid entry to fluid exit. The connectivity diagram forces a disciplined, thorough and often insightful consideration of the whole system, identifying both inconsistencies in interpretation as well as possibilities that might otherwise have been overlooked. At Terra Nova developing a comprehensive understanding of the field (and the whole plunging anticline within which the field is trapped) has required the RCA process to be applied at multiple


334

F. W. RICHARDS ET AL.

scales: from basinal scale, through large pressure compartments (c. 10–30 sq km) separated by major faults (offsets generally exceeding gross reservoir thickness), to individual producing fault blocks (c. 1 –6 sq km) separated by minor faults (offsets generally less than gross reservoir thickness), to individual sandstones within producing fault blocks. An RCA diagram of the greater Terra Nova area is exhibited here, at the sub-regional to large pressure compartment scale. This is followed by an example where connectivity analysis within and between producing fault blocks has had significant operational impact on development well planning and placement. Finally, an example (with connectivity diagram) is presented where connectivity at the finest scale, between individual sandstones, has to be considered to explain fluid distribution in a field appraisal well. Terra Nova Field is one of 18 significant hydrocarbon discoveries in the Jeanne d’Arc Basin, underlying the Grand Banks of Newfoundland, about 300 km offshore St. John’s (Figs 1 & 2a). Four of these significant discoveries exceed 1 billion oil equivalent barrels (original, in-place). Three fields are producing oil: Hibernia since 17th November 1997, Terra Nova since 20th January 2002 and White Rose since 12th November 2005. An agreement to proceed with the Hebron offshore development project was signed by the province and industry partners on 20th August 2008. In addition to its utility at Terra Nova, RCA has been an excellent reservoir management tool at Hibernia Field and led directly to identification of over 200 million barrels of additional reserves, 25 years

after discovery and 7 years into production. Overviews are provided by Magoon et al. (2005), who synthesize the Egret–Hibernia petroleum system in the Jeanne d’Arc Basin; and by Haugen et al. (2007), who provide a summary of the geoscience and petroleum engineering of the Terra Nova field. Terra Nova is unlike the three other major fields in the Jeanne d’Arc Basin in two respects: no gas leg is present, and significant oil is present in only one of four major reservoir intervals. In order to understand the absence of gas (and whether this should be anticipated in undrilled or nearby fault blocks), it is necessary to look at source distribution, maturity and charge in relation to reservoir connectivity at a basin scale. Connectivity in the trap as a whole has to be considered in order to understand why three of four major reservoir intervals are wet, and this understanding is necessary to assess near-field or bypassed potential. Fluid contact and aquifer pressure data indicate five major compartments at the producing reservoir level at Terra Nova (each containing multiple production scale fault blocks). These major pressure compartments can be rationalized by understanding connectivity at the major bounding faults which change from juxtaposition seals (i.e. cross-fault juxtaposition of reservoir to non-reservoir) to juxtaposition connections (i.e. cross-fault juxtaposition of reservoir to reservoir) with change in throw. At production fault-block scale (producer – injector well pair scale), lateral and vertical connectivity of individual sandstone units within and between blocks is the focus. Vertical connectivity

Fig. 1. Location of major oil fields, offshore Newfoundland, Eastern Canada.

TERRA NOVA, RCA

335

Fig. 2. Jeanne d’Arc Basin. (a) Significant oil and gas discoveries. The four major oil fields in the basin are highlighted (.1 billion oil equivalent barrels in-place). (b) Major structural elements. The Jeanne d’Arc Basin is a Mesozoic rift basin, an asymmetric graben, bounded by the dominant Murre and Mercury Faults to the west and the Voyager Fault system to the east. It is segmented by the late trans-basin fault trend that includes the Nautilus and Trinity Faults. Plunging anticlines at Hibernia and in the Terra Nova area indicated. The Hebron Complex occurs where northward extension of the Terra Nova anticline intersects horsts and tilted fault blocks in the trans-basin fault trend.

can be controlled internally, via incision, onlap, internal faulting or cementation, but it often results from ‘stairstepping’ of fluids to and from adjacent fault blocks within the same major compartment via juxtaposition connections.

Perched waters are present at Terra Nova in several parts of the field occurring either very locally, in individual sandstones within single production scale fault blocks, or are common to multiple sandstones within groups of two or more

336


production scale fault blocks. Perched waters exist locally within otherwise hydrocarbon filled traps when hydrocarbons filling from top down cannot displace water from local, isolated, structural depressions mapped at base-seal level. Within a given trap, perched waters are in pressure communication with adjacent regional aquifers via a contiguous hydrocarbon column, are at higher pressure than adjacent regional aquifers (relative to elevation), and have hydrocarbon–water contacts that are shallower than the contacts with adjacent regional water (hence the term ‘perched’). However, beyond the trap area regional aquifers, in particular hydrostatic aquifers, commonly extend to elevations considerably shallower than locally perched waters within a hydrocarbon trap. Contacts associated with perched fluids are controlled by ‘breakover’: the spill of the pressure-driven, denser fluid, like water over a dam (Vrolijk et al. 2005).

Methodology and basin overview

Two further steps are intrinsic and are also illustrated. It is common to refine initial interpretations subsequent to RCA, particularly near faults where the seismic image is often poorest. In seismic timeto-depth conversion, a controlling saddle or crossfault leak point is often fine-tuned to the associated fluid contact depth known more precisely and accurately from pressure or log data. RCA models offer non-unique interpretations of fluid distribution and pressures, particularly in undrilled fault blocks. This forms a natural basis for scenario analysis: the identification of multiple permissible scenarios, which are then volumetrically assessed, risked and aggregated to facilitate business decisions. An important aspect of RCA is documentation. At each business stage, there is usually speculation as to how fluids are trapped or escape through a system, but often there is no formal vehicle for retaining these ideas and alternate hypotheses. The connectivity diagram (or annotated montage) in RCA provides this implicitly.

RCA methodology

Jeanne d’Arc basin: structural overview

Reservoir Connectivity Analysis is a systematic, rigorous process defining reservoir plumbing and fluid distribution, which integrates and reconciles fluid types, fluid pressures, contacts, and geological interpretations (Vrolijk et al. 2005). RCA is based on well known physical principles and types of seals and connections between reservoir compartments: buoyancy, spill and breakover (controlled by folding, faulting, stratigraphy and diagenesis), and mechanical and capillary seal capacity. At Terra Nova, RCA has involved technical collaboration between the Operator, Petro-Canada, and ExxonMobil as well as independent analyses. In Vrolijk et al. (2005), three basic components in RCA are defined and described in detail. The goal is to map how hydrocarbons displaced by buoyancy from any compartment will migrate to the system exit point.

The Jeanne d’Arc Basin (Fig. 2b) is an asymmetric rift graben formed by two initial phases of extension beginning in Late Triassic (NW–SE extension) and Late Jurassic (west –east extension) times. The dominant, listric Murre Fault to the west and the Voyager Fault system to the east define the basin margins. The late, trans-basin fault trend, which includes the Nautilus and Trinity Faults, results dominantly from a third, Aptian –Albian, phase of extension (SW –NE). The trans-basin fault trend is important structurally and commercially in forming hydrocarbon traps, and also in relation to source rock maturity in controlling gas distribution within the basin. Two of three major oil fields in the south part of the Jeanne d’Arc Basin – Hibernia, and the Hebron complex – occur where the trans-basin fault trend intersects major plunging anticlines, which result from the interplay of stresses associated with the extensional episodes described above, and locally some salt movement at depth. The Hibernia anticline plunges south from the Nautilus Fault forming a consistent high-side trap at multiple stratigraphic levels in Hibernia Field. Fluid fill is dependent on juxtaposition against a thick shale section to the north across the Nautilus Fault. The Terra Nova anticline has low-side fault dependency at its south end against the basin bounding Voyager Fault, where it is ineffective as a hydrocarbon trap, and the anticline plunges to the north. Hydrocarbons are trapped in a broad, relatively low relief, faulted, dip reversal at King’s Cove in one shallow reservoir (the oil is heavily biodegraded,

(1) (2) (3)

Describe reservoir compartments. Define connections between compartments. Build an RCA model.

RCA models are built at various scales, various levels of detail, and are presented in different ways. The process is simple in principle; in practice it often becomes complex and revealing. In this paper we show a connectivity diagram at the major reservoir interval/major compartment scale and a second at the individual sandstone/production fault block scale. Map montages or 3D perspective views shown here can be very effective at each stage in this process, when annotated with fluids, pressures, connections, and barriers.

TERRA NOVA, RCA

requiring down hole pumps to recover at the surface) and further north at Terra Nova in the deepest of four reservoir intervals. At the Hebron horst where the plunging Terra Nova anticline orthogonally intersects the trans-basin fault trend, hydrocarbons are, like Hibernia Field, trapped consistently at multiple stratigraphic levels. Several additional hydrocarbon traps are formed within the trans-basin fault trend, most relevant to this discussion is a gas accumulation, West Bonne Bay, only 7 km north of Terra Nova discovered in 2005 and Springdale where there are minor gas and oil accumulations, at relatively shallow depths, on the rift shoulder east of Terra Nova. Fields to the north of the trans-basin fault trend, the largest of which is White Rose, are not discussed here, but pressure data from these fields are presented in addressing fluid and overpressure distribution in the basin as a whole.

Jeanne d’Arc basin: stratigraphic overview The Terra Nova area is underlain by thick shelf carbonates of the Rankin Formation (Fig. 3), within

337

which thick, organic, calcareous mudstones of the Egret Member (Kimmeridgian) form a mature, prolific, Type 2 source rock, (Grant & McAlpine, 1990). Above the Rankin Formation, at a broad scale, three reservoir/topseal pairs are present within the syn-rift section of the Jeanne d’Arc Basin (Fig. 3a, b).The Jeanne d’Arc Sandstones/Fortune Bay Shale couplet contains significant oil at Terra Nova. The Catalina/Hibernia Sandstones are overlain by a thick White Rose Shale topseal, and the Ben Nevis Avalon Sandstones by Nautilus Shale, but both reservoir intervals are wet at Terra Nova apart from some minor shows. In the post-rift section, Tertiary shales and Late Cretaceous and Early Tertiary sandstones trap hydrocarbons locally, but also act as conduits for hydrocarbons to exit the basin. A stratigraphic chart (Fig. 3c) illustrates that the basin margins (up-dip to the south, west and east) are much sandier than the basin centre (to the north). If stratigraphic connectivity is enhanced by cross-fault juxtaposition connectivity, then it is likely that hydrostatic (or near hydrostatic) pore

Fig. 3. Stratigraphy: Terra Nova/Jeanne d’Arc Basin. (a) Summary: seals, reservoirs & source rock. (b) Representative Terra Nova well. Gamma Ray and Density logs are displayed. Sandstones are yellow, shales grey, and carbonates blue. Porous sandstones are indicated in red. (c) Stratigraphic chart. A prolific source interval, the Egret Member of the Rankin Formation is overlain by three reservoir-topseal pairs in the syn-rift section. Post-rift, Late Cretaceous and Tertiary reservoirs, overlain by shale topseals, trap hydrocarbons locally and also act as exit conduits.

338


pressures prevail at the basin margins. Towards the basin centre (and at depth) geometric connectivity will likely diminish, and the higher pressures associated with capillary and mechanical connections should be more common.

In terms of reservoir connectivity, these stratigraphic and diagenetic observations provide potential trap mechanisms on the south and north sides of the field.

Terra Nova field: reservoir overview

RCA workflow: application at Terra Nova field

At the reservoir scale, Terra Nova is illustrated by a south–north traverse from the Operator’s 2006 geological model (Fig. 4a). Eight well penetrations are shown in seven production scale fault blocks in the East Flank of the field. Reservoir properties are annotated at three wells. Perched waters illustrated in Figure 4a are discussed later in the context of the major pressure compartments within the field. The location of this cross section is shown in Figure 4b where Terra Nova Field is illustrated schematically. The major (pressure) compartments are shown and annotated: the wet West Flank, oil and water in the Graben, East Flank, Far East Central and Far East South. Two additional field areas are annotated: the ‘attic’ and the ‘horst’. Areas with uncertain fluid fill are uncoloured. Gas injection, water injection, oil production and exploration/appraisal wells are shown. The Jeanne d’Arc reservoirs at Terra Nova have historically been divided into the Beothuk Member, the ‘E’ Sandstone, that is separated by a consistent through-going shale from the underlying Terra Nova Member, which comprises the more amalgamated ‘D’ and ‘C’ sandstones. A total of six reservoirs are defined in the current reservoir model, five of which are interpreted as incised valley fill, dominantly high quality braided fluvial sandstones. The lower of two D sandstone intervals, the ‘Da’, contains similar high quality reservoir to the incised valley fill reservoirs but transitions laterally into poor reservoir quality, low energy, marginal marine facies. The depositional orientations of these reservoir sandstones vary, with provenance either from the south or west, but in aggregate there is a west–east fairway exemplified by the G-90 1 well in Figure 4. This well has over 100 m net reservoir and a high net-to-gross ratio of 0.7. This changes quickly to the north (G-90 4 well) moving out of the stacked incised valley trend, and this lateral degradation likely forms part of the trap at Terra Nova. To the south, the Jeanne d’Arc reservoirs progressively lap onto the Rankin Formation, accompanied by early dolomite cementation, which is attributed to mixing of marine and magnesium-rich terrestrial waters derived from the Rankin carbonates. The I-97 well in Figure 4, with 14 m net reservoir and a NTG of 0.42, illustrates this point.

Basin scale RCA: ‘Gas Shadow’ at Terra Nova In order to understand the absence of a gas leg at Terra Nova, it is necessary to consider source rock distribution and maturity (Grant & McAlpine 1990) in relation to the structural configuration of the basin, represented in Figure 5a by depth structure at Base Hibernia Formation (Base Berriasian). Overmature Egret source rock occurs exclusively in the north part of the basin, north of the trans-basin fault trend. As no known gas accumulations occur south of this trend (Fig. 5a) it is reasonable to speculate that Terra Nova is in a gas migration shadow. Gas generated in the deep, north part of the basin likely migrates upwards and outwards and is in a sense ‘deflected’ at the transbasin fault trend. Inspection of seismic and pressure data suggests gas migration initially occurs by mechanical and capillary leak and then follows geometrically connected pore space, moving stratigraphically up-section via cross fault juxtapositions. At Hibernia field, RCA by Rod Myers (internal study, ExxonMobil) led to the conclusion that the large gas cap at the crest of the Hibernia sandstone reservoir leaks by capillary seal-failure, and through-going gas has preferentially removed light oil resulting in tar mats down-dip. This capillary gas leakage at Hibernia field has been characterized as a natural ‘gas separator’ on the west side of the basin. Pressure data from the Geological Survey of Canada ‘BASIN’ database (available online at the GSC website) are displayed in Figure 5b relative to a hydrostatic line, a leak-off trend based on leak-off tests in the same data base, and a lithostatic gradient defined with a density of 2.3 gm/cc. Mechanical seal-failure due to opening of pre-existing fractures typically occurs at a lower pressure than failure measured by leak-off tests, which often have to reach sufficient pressure to first create fractures. It is important to note that capillary sealfailure can also occur, potentially at pressures significantly below leak-off measurements. Also of interest is that the leak-off trend is not linear, inferring a change in the magnitude of the principle stresses with depth. A comprehensive evaluation of stresses and the causes of elevated pore pressures is beyond the scope of this paper. Rather, we

TERRA NOVA, RCA

339

Fig. 4. (a) Reservoir scale traverse, Terra Nova East Flank from the Operator’s 2006 geological model. Colour scale represents modelled porosity values in Jeanne d’Arc sandstones. Gamma ray logs are shown at three wells (blue). Five wells are shown as trajectories only (black). Perched aquifers and potential ‘breakover’ points are annotated. Arrows indicate the direction of lateral displacement of water as the trap filled. (b) Schematic field map at Top D reservoir. The major field-scale reservoir compartments are annotated (West Flank, Graben, East Flank, Far East Central and Far East South). Areas above interpreted oil– water contacts are green, below are blue. Grey indicates undrilled blocks where two or more fluid interpretations are possible. Solid black circles are exploration and appraisal wells which were drilled pre-development except for the Far East South well. Black trajectories are oil producers, blue are water injectors and red are gas injectors. Other wells are grey.

340


Fig. 5. (a) Jeanne d’Arc Basin: Structure at Base Hibernia Formation (Base Berriasian). Red, rift margins; blue, basin centre). Source rock maturity (Grant & McAlpine 1990) is indicated. Pink arrows indicate interpreted direction of gas migration out of overmature area. Known gas accumulations are shown in red. (b) Pressure–Depth plot shows wells colour-coded and referenced to map. Wells south of the trans-basin faults (yellow) are at, or near, hydrostatic pressure. In the trans-basin trend (blue) two populations are evident, one at or near hydrostatic, a second approaching leak-off. North of the trans-basin trend (red) pressures are elevated, commonly approaching leak-off. (Source of pressure data: online, GSC ‘Basin’ database). Empirically Terra Nova occurs in a gas shadow separated from the overmature part of the basin by the trans-basin faults. Pressure and seismic data indicate mechanical leak out of overmature source and adjacent poor quality sandstones followed by geometric leak at reservoir juxtapositions. At Hibernia Field there is evidence, including tar mats, of capillary gas leak (R. Myers, ExxonMobil, internal communication).

TERRA NOVA, RCA

document the boundaries of excess pore pressure and relate them to the distribution of gas in the basin. We interpret the transition to nearhydrostatic pressure at the threshold of reservoir connectivity (i.e. the transition of a seal-dominated to reservoir-dominated stratigraphic section) and similarly interpret a gas migration pathway to the seafloor at that point. Pressure data from wells south of the trans-basin fault trend (yellow in Fig. 5 and cross-referenced between the pressure–depth plot and map) are at, or close to, hydrostatic pressure and support the notion that in the proximal sandier southern part of the basin there is a high level of geometric connectivity. Data from wells within the trans-basin fault trend (blue in Fig. 5), form two populations, one near hydrostatic pressure, again indicating geometric connectivity, and a second near the leak-off trend. The higher pressure population largely comprises tests in the Rankin Formation (carbonates and mature source rock) and overlying poor reservoir quality Jeanne d’Arc sandstones. North of the trans-basin trend (red wells in Fig. 5) pressures are generally well above hydrostatic pressure reflecting increasing restriction to fluid movement in more distal sediments and increasing propensity at higher pressures for capillary or mechanical seal-failure.

Sub-basin scale RCA Analysis via Ben Nevis Sandstone structure map and south– north seismic traverse. The Terra Nova anticline is illustrated (Fig. 6a) at Top Ben Nevis sandstone level, mapped in time structure from 2D and 3D seismic data. This is the youngest of the three major syn-rift reservoir intervals. The anticline plunges north from the low side of the Voyager Fault system near the South Brook well through King’s Cove and Terra Nova fields to the Hebron Complex (seismic traverse, Fig. 6b). The effectiveness of this anticline as a trap is compromised at its south end by cross-fault and stratigraphic juxtapositions of Ben Nevis and Hibernia sandstones, as well as subcrop of Hibernia sandstone to the post-rift section at the base Cenomanian unconformity. Seal potential at this unconformity is limited due to the presence of post-rift sandstones in the late Cretaceous and early Tertiary section and potentially by low bed-seal capacity. Disrupted Tertiary reflections are interpreted as an indication of fluid escape to the sea bed. Onlap of the Jeanne d’Arc sandstones (and incision into the Rankin Formation) is also annotated on this seismic line and with an appropriate orthogonal configuration a trap is possible. In the Hebron Complex, oil and some gas is trapped structurally (volumetrically mostly in the Hebron horst) and spills south. Heavy oils (12–188 API) were tested at King’s Cove where a

341

local low relief dip reversal associated with minor faulting is mapped. Minor hydrocarbon columns are trapped at Springdale to the east of Terra Nova, in a high-side fault dependent trap on the rift shoulder: oil in the Ben Nevis sandstone and gas in minor post-rift sandstone. Analysis via Hibernia Sandstone structure map and west –east Seismic Traverse. In the Hibernia Formation, the Terra Nova anticline has a similar configuration and high propensity for leak as at the shallower Ben Nevis level. However, the Hibernia Fm. is more block faulted than the younger Ben Nevis Fm., and the same major compartments are evident as interpreted in the deeper Jeanne d’Arc sandstones. This can be seen both in map view and the west –east seismic line shown in Figure 7. These major compartments are: the West Flank, Graben, East Flank, Far East Central and Far East South. The Hibernia sandstones in this map area contain structurally controlled oil and gas accumulations in the trans-basin fault trend. In the Upper Hibernia Fm. (Fig. 7a) an oil accumulation at Hebron spills to the south, and at West Bonne Bay a gas accumulation spills at a saddle to the north then leaks via fault juxtaposition to reservoirs in the Ben Nevis – Avalon Fm. There are, in addition, small accumulations in the Lower Hibernia sandstone below interbedded thin sandstones and shales of the Middle Hibernia (which is both a poor reservoir and a poor seal). South of the trans-basin trend, there appears to be too much self-juxtaposition (i.e. cross-fault reservoir-to-reservoir juxtaposition where reservoirs are within the same stratigraphic interval) and juxtaposition to Ben Nevis – Avalon sandstones to yield major accumulations in the Hibernia Fm. sandstones. The west –east seismic line in Figure 7b also illustrates a key trap mechanism at Jeanne d’Arc level. This reservoir interval is encased between Fortune Bay Shale above and Rankin carbonate/ Egret shale below. These non-reservoir intervals commonly form juxtaposition seals to the Jeanne d’Arc reservoirs at the north–south faults that separate the major compartments at Terra Nova (indicated by red and brown bars in Fig. 7b). Mature Egret source rock (indicated in dark green in Fig. 7b) directly juxtaposes Jeanne d’Arc reservoirs over significant distances at these north–south faults.

Field scale RCA of the Jeanne d’Arc sandstones Compartment identification and description: major pressure compartments and production scale fault blocks. The Jeanne d’Arc reservoir interval at Terra Nova is shown in perspective view in Figure 8, from the Operator’s 2006 geological

342


Fig. 6. (a) Terra Nova anticline, time structure at Top Ben Nevis (Albian) mapped via 3D and 2D seismic data. Hachured green and red areas are mapped oil and gas accumulations. Green circles indicate heavy oil tests (below 208 API) at King’s Cove and Springdale. Green arrows indicate interpreted hydrocarbon spill and migration. Terra Nova development wells are indicated in blue. (b) South–north 2D seismic profile illustrating interpreted up-dip stratigraphic juxtaposition of Hibernia, Ben Nevis and post-rift sandstones with ensuing low trap potential. Deeper, and further north, onlap of Jeanne d’Arc reservoirs is highlighted.

TERRA NOVA, RCA

343

Fig. 7. (a) Terra Nova Anticline, time structure at Top Hibernia (Berriasian), mapped via 3D and 2D seismic data. Hachured green and red areas are mapped oil and gas accumulations. Green and red arrows indicate interpreted hydrocarbon spill and migration. Terra Nova development wells are indicated in blue. WF, West Flank; G, Graben; EF, East Flank; FEC, Far East Central; FES, Far East South. (b) West– east 2D seismic profile illustrating Hibernia Formation reservoir-to-reservoir juxtapositions across bounding faults between major compartments. In contrast, at Jeanne d’Arc reservoir level, reservoir to non-reservoir juxtapositions are more common at major faults: brown bars indicate Jeanne d’Arc sandstone to Fortune Bay Shale juxtaposition and red bars indicate Jeanne d’Arc sandstone to Rankin carbonate/Egret source rock juxtaposition. Ben Nevis– Avalon interval in light green, Hibernia in yellow, Jeanne d’Arc orange, Egret dark green.

344


Fig. 8. Terra Nova Field: Jeanne d’Arc D & C reservoirs – perspective view from north (from the Operator’s 2006 geological model). Major compartments and interpreted barriers. Aquifers are colour-coded in blue by excess pressure (above hydrostatic). Oils are colour-coded in green by API. Contacts are annotated in call-outs. Undrilled blocks in the Far East are white. Interpreted barriers are indicated in pink, red and brown and annotated. Exploration and appraisal wells are in grey. Production wells are coloured coded by fluid: red, gas injection; green, oil production; blue, water injection. (Note that there is considerable uncertainty in the 23585 m contact interpreted in the Far East South).

model. The majority of west– east faults in the field downthrow to the north, and a perspective view from the north provides the clearest single 3D view of the field. The field is on the low side of the Voyager Fault system and the high side of the Trinity Fault. The major compartments in the Jeanne d’Arc sandstones at Terra Nova are identified by integrating fluid and pressure data (Fig. 9) with barriers identified and mapped via seismic and well data (Figs 6 & 7). Fluid observations are captured in Figure 8 by colour coding aquifers and oil columns of differing pressures and API gravity, and by annotating fluid contacts. Coloured lines represent juxtaposition and diagenetic barriers. In this way the major (pressure) compartments are identified and described. Although pressures are defined at individual well penetrations, careful geological interpretations of connectivity between production scale fault blocks (via juxtaposition and ‘stairstepping’) are used to aggregate production scale fault blocks into pressure-defined compartments (i.e. defined by a single, original oil pressure–depth line). Where there is significant

uncertainty in extrapolating fluid and pressure interpretations, fault blocks are coloured white in Figure 8. In conjunction with the summary pressure –depth plot, this annotated perspective view summarizes the first component of RCA at the major compartment level: identifying and describing the major compartments. The West Flank of Terra Nova is separated from the Graben by a south –north fault which, over much of its length, has throw which exceeds gross reservoir thickness. The West Flank is wet, at near hydrostatic pressure. The Graben contains 348 API oil up-dip, to the south, and a common aquifer with the West Flank down-dip, to the north. The Graben and East Flank are treated as separate major compartments for production purposes based on the observation that the intervening fault offset appears consistently to exceed gross reservoir thickness. However, the oil accumulations in these areas are of the same API gravity and sit on a common pressure –depth line, and from these observations it is inferred that these major fault blocks are locally in communication and form a single pressure compartment in the oil leg.

TERRA NOVA, RCA

345

Fig. 9. Terra Nova Jeanne d’Arc Sandstones – Summary of Pressure Depth Data. The hydrostatic aquifer line is derived from pressure data in the Jeanne d’Arc and shallower reservoirs. Note that considerable uncertainty exists in the 23585 m FWL (free-water-level) interpreted in the Far East South.

Within the Graben and East Flank at least four separate perched water accumulations have been encountered in wells, and two of these are indicated in Figure 8a in the East Flank (at excess pressure c. 900kPa and with contacts at 23351 and 23338 m). These perched waters occur where water has not been displaced by hydrocarbons from isolated structural depressions mapped at the baseseal, and are further illustrated in profile view in Figure 4. Two perched waters drilled in the Graben occur very locally in single sandstones that are not visible in Figure 8. The Graben and East Flank are an excellent illustration of the difficulty in evaluating connectivity or compartmentalization by considering hydrocarbon contacts alone. To do so here would mask the fact of a large connected oil column in original pressure communication. The Far East of Terra Nova is at different oil and aquifer pressures from the East Flank, and these major compartments are separated by faults that exceed gross reservoir thickness. The Far East Central and Far East South compartments have different pressures and fluids (at face value from well data), and are separated, in the base case interpretation, by a diagenetic barrier. The ‘attic’ area annotated in Figure 8 is an undrilled, narrow, elevated set of small fault blocks that is interpreted to be in fluid and pressure continuity with the East Flank (small to zero fault offsets at its north end). The ‘horst’ area, west and NW of the Far East South is also undrilled and is coloured white in Figure 8.

The Far East Central compartment comprises a group of fault blocks with aquifer pressure about 7000 kPa above hydrostatic and oil at 378 API. These fluids are extrapolated from five well penetrations in three blocks to immediately adjacent undrilled blocks across faults where offset is considerably less than gross reservoir thickness. These fault blocks, as well as adjacent fault blocks to the east and north (coloured white in Figure 8 because connectivity is less certain), are potentially isolated by juxtaposition to tight Rankin Formation to the east and west, and by the diagenetic barrier interpreted to the south. There is some uncertainty in interpreting the northerly limits of the Far East Central and East Flank compartments. Rapid northward reservoir degradation, observed at development wells in the East Flank and Graben, provides reasonable evidence to suggest a stratigraphic (and partly diagenetic) trap south of the Trinity Fault, and this is the simplest explanation. However, low-moderate quality Jeanne d’Arc reservoir is present north of the Trinity Fault at the North Trinity H-71 well (35 m net, NTG 0.15, average porosity 12%). Therefore, it is possible that connected Jeanne d’Arc sandstone porosity in the Far East Central compartment extends north to the Trinity Fault, and based on seismic correlations south from the H-71 well may partly juxtapose Lower Hibernia sandstone (of considerably lower pressure, Fig. 9) across the fault (in addition to juxtaposition against Fortune Bay shale). This observation introduces several alternate possibilities. Firstly, seismic correlations may be

346


incorrect due to a considerable fault shadow. Secondly, the Lower Hibernia sandstone might itself trap oil, between the fault and the H-71 well. Further possibilities are that due to ‘fault rocks’ (formed through smear, cataclasis or cementation) a low transmissibility dynamic connection or a capillary seal has formed, maintaining a pressure differential between these juxtaposed sandstones. A recent well (I-66), drilled in the Far East South fault block encountered fluids at considerably lower pressures than the Far East Central compartment. Oils on two different pressure–depth lines and intervening water at hydrostatic pressure were observed in this well. PVT and geochemical analyses indicate that these oils are of differing API gravities (and are both different from other oils in Terra Nova) although there it is some uncertainty in the PVT analysis due to contamination by oil based mud. The obvious inference is that the Far East Central and Far South areas constitute different pressure compartments and, based on observed cementation in both the I-66 well and the nearby well to the NE, a diagenetic barrier is invoked. In this scenario, fluids, pressures and contacts in the Far East South are explained by potential connections to the west, east or south (next section).

However, an alternative interpretation is that the Far East Central and Far East South form a single compartment that requires a different configuration of cemented reservoir. In this case, fluids and pressures observed in I-66 are explained by differential pressure depletion due to production from the Far East Central compartment.

Connection definition Having described the major compartments based on fluid data and interpreted barriers, the next component in RCA is to define and evaluate potential connections between compartments and identify potential exit routes for hydrocarbons. This component is illustrated in Figure 10 where connections (absence of barriers), spills, breakovers, capillary leaks and some alternate connection hypotheses are added to Figure 8. There are two potential exit paths out of the West Flank. Leak at the up-dip, South-Western onlap edge is likely in principle, as the basin margin rises in this direction. In addition, up-dip and south of the geological model in Figure 10, some local trapping and spill of oil is possible due to irregular incision of Jeanne d’Arc sandstones into the

Fig. 10. Terra Nova Field: Jeanne d’Arc D & C reservoirs – perspective view from the north (from the Operator’s 2006 geological model). Potential connections between major compartments: Red/white arrows indicate potential juxtaposition connections between Jeanne d’Arc reservoirs and other formations. Solid red arrows indicate Jeanne d’Arc self-juxtaposition connections. The black arrow indicates a possible route by which the Far East South might link to the East Flank producing area, via self-juxtapositions.

TERRA NOVA, RCA

Rankin Carbonate (Enachescu et al. 1994). This concept appears to have been unsuccessfully tested at the King’s Cove A-26 well, which encountered about 14 m of porous wet Jeanne d’Arc reservoir. A second possibility is that in an up-dip southerly position, Jeanne d’Arc sandstones leak via crossfault juxtaposition to Hibernia sandstones in the Graben. The throw of the bounding fault between the West Flank and Graben increases southwards, and seismic data indicate that cross-fault juxtaposition of these sandstones may occur before the Jeanne d’Arc sandstones fully lap out onto the Rankin Fm. to the south. The observation that Jeanne d’Arc aquifer pressure in the West Flank is elevated about 350 kPa relative to a regional hydrostatic pressure line (established at multiple stratigraphic levels) requires some discussion. It is likely that ongoing fluid influx due to hydrocarbon charge slightly exceeds exit, preventing equilibration to hydrostatic pressure. To the SW, where Jeanne d’Arc sandstones in the West Flank appear structurally open to the rapidly rising, shallow southern termination of the Jeanne d’Arc Basin it is possible that convoluted, low permeability, stratigraphic connections prevent pressure equilibration. To the SE, where Jeanne d’Arc sandstones potentially juxtapose hydrostatically pressured Hibernia sandstones, a low transmissibility connection may provide a similar explanation, or, it is possible that an oil column trapped by capillary seal contributes to maintaining this pressure differential. Pressures in the Jeanne d’Arc on the West Flank were measured with strain gauges, whereas the hydrostatic line is established with quartz gauge measurements, but the 350 kPa difference is too large to explain as measurement error in strain gauge data, typically +100 to 175 kPa (Dewan, 1983). Oil in the Jeanne d’Arc sandstones in the Graben at Terra Nova is interpreted to spill to the West Flank at the top of a self-juxtaposition window which develops as the intervening fault diminishes in offset from south to north, eventually tipping out. However, nearby wells indicate variable reservoir thickness and quality making it difficult to pinpoint the spill precisely. Oils in the Graben and East Flank have similar compositions and API gravities and lie on a common pressure–depth line. From these observations it is inferred that oil accumulations in the East Flank and Graben are at least partially connected. The most likely place for this connection to occur is where the intervening fault splits into two faults, such that a single throw that exceeds gross reservoir thickness (but is less than the thickness of the topseal) becomes two throws, each less than gross reservoir thickness, thus creating juxtaposition connections. The lowest point of these

347

windows controls breakover of perched water from the north part of the East Flank to the Graben. As at the controlling connection in the Graben, nearby well data indicate reservoir thinning and quality degradation making it difficult to be precise (or emphatic) in this interpretation. In this type of situation, where there is uncertainty in connectivity, two approaches can be employed. The first approach, called ‘stochastic multifault analysis’ (James et al. 2004), treats the problem probabilistically, acknowledging uncertainty in both stratigraphy and fault throw. The second, deterministic approach, employed here, requires some judgement in how areally confined, potentially discontinuous sandstones are incorporated. At Terra Nova, the E sandstone (Beothuk Member) is separated by a consistent shale from the more amalgamated D/C stratigraphic interval (Terra Nova Member). The E sandstone is of limited areal extent, and production data have consistently shown poor internal connectivity. So at the major compartment level the E sandstone is at a first pass disregarded, but it then has to be considered when more detail might be necessary to explain fluid observations (or to predict possible scenarios). Thus, a locally well-developed E sandstone might contribute to this oil connection between the Graben and East Flank. We often arrive at similar non-unique but geologically permissible interpretations in an RCA study. The fluid pressure and composition data suggest a geological connection must exit, but the geological details of that connection and its implications for reservoir-scale, production fluid flow are beyond the resolution of the available data to define those connections. Perched waters in the East Flank are illustrated in Figure 4. These perched waters are controlled by breakover points. Because it is the base reservoir surface that is important in defining breakover points these connections are not visible in Figure 10, but are illustrated in Figure 4. The Far East Central compartment comprises a group of fault blocks at the centre of which is a faulted four-way dip closure (the ‘central culmination’) containing a contiguous oil column in the Jeanne d’Arc E and D/C reservoirs (coloured green in Fig. 10). The central culmination oil accumulation is underlain by a contiguous aquifer at about 7000 kPa excess pressure (at least 4000 kPa greater than any adjacent pressure compartment). It is coloured dark blue in Figure 10 and potentially extends into fault blocks to the east, coloured white. Also coloured white are parts of the NW and NE fault blocks within the Far East Central compartment that are above the elevation of the free-water-level (FWL) in the central culmination. For the purposes of this discussion these

348


fault blocks are interpreted to form part of the Far East Central compartment, at least as defined by a common aquifer. Given the barriers already described around the Far East Central compartment and the abnormally high excess pressure observed in both oil and water, it is possible to speculate that the compartment is entirely sealed. Anomalously high excess pressure in the Far East Central compartment is attributed to ongoing influx of oil from adjacent mature Rankin source rock combined with inhibition to fluid flow out of the compartment. If the initial volume of the aquifer and the current volume of oil are estimated, and reasonable rock and water compressibility are assigned, then the observed excess pressure can be accounted for with no expulsion of water from the compartment (using reverse material balance type calculations). A different view is based on the observation that, at the top of the main D/C reservoir interval, the mapped FWL in the central culmination is at the elevation of several cross fault spill points to the north or east (all at very similar elevations). This observation is either coincidental or reflects fill and spill of oil out of the D/C sandstones in the central culmination (and requires that oil in the E sandstone is essentially attic oil with no further connection). Without considering pressure data, it would be simple to invoke fill and spill of oil out of D/C sandstones in the central culmination to the fault blocks to the north, followed by up-dip migration and leak at the juxtaposition of Jeanne d’Arc and Lower Hibernia sandstones at the Trinity Fault. However, because of the 4000 kPa pressure differential between these sandstones it is necessary to invoke a dynamic seal caused by degraded, low permeability reservoir – which could occur either as a consequence of ‘fault rocks’ at the juxtaposition window or due to northward degradation associated with cementation and/or facies changes (as observed in the East Flank). Two dynamic mechanisms are plausible: slow Darcy flow through low transmissibility sandstones (with varying balance of oil influx and outflux), or a capillary seal which initially allowed expulsion of water, but subsequently became sealed because oil could not displace water from fine capillaries. In either case, leak of oil across the Trinity Fault must occur above the level of the FWL in the central culmination or the whole central compartment should have backfilled below the level of the observed FWL (unless just enough oil has been generated to fill the central culmination but not the fault dependent closures in the blocks to the north). Although we conjecture a number of different, possible controls on the total oil column trapped in the Far East Central fault blocks, it appears likely

that the primary control on oil distribution in the series of fault blocks that comprise the Far East Central (i.e. characterized by anomalously high aquifer pressures) includes geometric fault juxtaposition connections. Results from the five wells drilled into this area are consistent with simple fillspill relationships defined across mappable reservoir connections along faults. The Far East South pressure compartment comprises at least a single fault block, tested by the I-66 well in 2007. Pressure continuity probably extends to adjacent blocks to the west and NW based on seismically interpreted reservoir presence and juxtapositions. Four potential controlling connections have been identified, and these form the basis for scenario analysis of this part of Terra Nova Field. (Scenario analysis is highlighted as key part of the RCA workflow but not discussed in more detail here). The Far East South area has necessitated connectivity analysis at the individual sandstone level, and this is described in a later section. Connection between the Far East South and either the Far East Central producing area or East Flank producing area is permissible if the observed pressure and fluid distribution at I-66 is considered due to differential pressure depletion and if some of the observations at the well are considered unreliable due to contamination of PVT data by oil based mud. Connection to the Far East Central would be consistent with pre-drill predictions and would not require any type of permeability barrier between the two areas. However, material balance calculations based on Far East Central production show that the size of the connected permeable rock volume is inconsistent with the pore volume required to connect to the I-66 area in the Far East South, which is several kilometres away. Connection to the East Flank is considered possible via a tortuous path through numerous fault blocks in the ‘horst’ and ‘attic’ areas (the horst is the white block immediately to the west of the Far East South in Figs 8 & 10 while the ‘attic’ is the elevated area at the extreme east of the East Flank). Note that material-balance calculations do not help in this situation because of the uncertainty due to the presence of multiple connected fault blocks with multiple production and injection wells. Two alternative interpretations exist based on simple juxtaposition and require no pressure depletion due to production: leak to the footwall of the Voyager Fault (fractures or low permeability sandstones are possible), or a spill from the Far East South to the ‘horst’ and then to the Hibernia sandstone in the Graben. Seismically, the D/C sandstone and Hibernia sandstone are very close, but not quite at juxtaposition, in this area but local development of a permeable E sandstone would achieve this

TERRA NOVA, RCA

(and a nearby high delta throw fault would allow connection between E and D/C sandstones). Integration of compartments and connections-field scale connectivity diagram. A critical step in a connectivity analysis is to follow the implications of a spill or leak interpretation made from one compartment into all subsequent reservoir compartments until a system exit leak point (a pressure sink) is reached. An example of this process is illustrated in Figure 11a –e. The connectivity diagram documents the analysis and tests for consistency and completeness. It reveals errors in logic, for example, oil spilling out of a compartment and through a series of subsequent compartments, ultimately reaching a compartment independently interpreted to spill back in the direction just traveled. Moreover, the analysis is a means to recognize and identify where assumptions and interpretations have been made that could be subsequently explored in a sensitivity analysis. The connectivity diagram represents all the relevant trapping elements, including structural, stratigraphic, and diagenetic compartments, as rectangular boxes (Fig. 11a). Further information about fluid types and pressures encountered in compartments are further annotated on the diagram. The challenge in this part of the analysis is to include a sufficient level of detail to account for all connections that are responsible for controlling gas and oil fill limits while at the same time avoiding so much detail that the connectivity framework becomes intractable. This framework then serves to help evaluate connections between compartments. In Figure 11a oil entry points in the Jeanne d’Arc reservoir from the Egret source rock are defined based on seismic interpretations. Lateral and vertical connections between mature Egret source rock and Jeanne d’Arc sandstones are present in each major compartment in Terra Nova Field with the possible exception of the Far East South, although this might be accomplished via splays in the Voyager Fault system which are beyond the detail of this diagram. Exit routes out of the East Flank and Graben to the West Flank are shown in Figure 11b. The crossfault link between East Flank and Graben is indicated and the common oil column in these compartments spills from the Graben via simple cross-fault juxtaposition to the West Flank. Two potential exit routes out of the West Flank are identified. Route A leaks up-dip in the Jeanne d’Arc sandstones to the SW onlap edge which rises to the south end of the Jeanne d’Arc Basin. Route B invokes cross-fault leak (or a capillary leak) from Jeanne d’Arc sandstones in the West Flank to Hibernia sandstones in the Graben followed by cross fault leak to

349

Ben Nevis sandstones and subcrop to sandstones in the post-rift section. Topseal-failure interpreted seismically (Fig. 6b) is annotated, as well as up-dip leak to the SW mapped seismically in the post-rift section. Four potential exits routes are possible for the oil column encountered in the Far East South (Fig. 11c): (C) a tortuous series of juxtaposition connections via the ‘horst and attic’ to the East Flank Jeanne d’Arc producing area, (D) a connection with the Far East Central area producing area, forming a contiguous pressure compartment, (E) spill to the ‘horst’ then cross fault leak to Hibernia sandstone in the deepest part of the East Flank, and (F) leak to possible permeable rocks in the footwall of the Voyager fault. Route (G) represents stairstepping up-section along the fault between the Far East South and rift flank blocks in the Springdale area (dashed purple lines). This route has potential to fill a known small oil accumulation in the Ben Nevis reservoir at Springdale, but access to mature source is more likely from the north. It is possible that there is no simple geometric leak out of the Far East Central compartment at current pressures. Capillary seal-failure is indicated notionally in Figure 11d (Route H) and could occur through the topseal, via a lateral stratigraphic, diagenetic or ‘fault rock’ seal, or even via baseseal. A more specific exit route (I), across the Trinity Fault to juxtaposed Lower Hibernia sandstones could occur via a low transmissibility dynamic connection or via capillary seal-failure. If leak occurs across the Trinity Fault, then there is potential for oil to stair-step via shallower Hibernia sandstones to the Ben Nevis sandstones on the north side of the Trinity Fault and then up-dip to Hebron. The full connectivity diagram appears in Figure 11e and incorporates two connections from the trans-basin fault trend: oil spill from Hebron and inferred migration of gas from the Upper Hibernia sandstone at West Bonne Bay to the post-rift section at Springdale. Regional mapping indicates that migration of gas from the overmature parts of the basin into the Upper Hibernia sandstone at West Bonne Bay is unobstructed by juxtaposition seals at faults, and subsequent spill and juxtaposition leak to Ben Nevis Avalon sandstones has been interpreted seismically. However, a further connection to post-rift sandstones at Springdale has not been located seismically but is suspected from the proximity of these gas accumulations.

Production fault block scale RCA The producing fault blocks and wells at Terra Nova are illustrated in Figure 12a. Fifteen oil producers,

350


Fig. 11. (a) Connectivity diagram between Rankin/Egret source rock and Jeanne d’Arc reservoirs (via inspection of 3D seismic data). Grey indicates mature source rock, white immature source in rift flanks. Compartments are colour-coded by fluid fill (red, gas; green, oil; blue, water); aquifer excess pressures (relative to a hydrostatic reference), oil gravities, and FWLs are indicated. BN, Ben Nevis; JdA, Jeanne d’Arc. Solid lines indicate lateral juxtaposition at faults. Dashed lines are potential vertical connections. (b) East Flank, Graben and West Flank Connectivity. Exit out of the Graben and East Flank via cross-fault spill to the West Flank is shown. Two exit routes, A and B, out of the West Flank are illustrated and annotated. Each ultimately leaks oil to post-rift section and out of the Jeanne d’Arc Basin. (c) Far East South connectivity diagram. Internal connection within the Far East South is indicated, and four potential exit routes. Route C is a tortuous series of juxtaposition connections via the ‘horst and attic’ to the East Flank producing area (from which exit is shown in b). Route D invokes a contiguous, diagenetically controlled compartment with the Far East Central compartment (exit from which is summarized in d). Route E invokes cross-fault leak to the Hibernia

TERRA NOVA, RCA

351

Fig. 11. (Continued) sandstone in the East Flank (via inferred locally developed E sandstone) from where multiple cross-fault exit routes are available. Route F occurs if there is cross-fault leak to permeable sandstones or permeable fractured carbonates in the footwall of the Voyager Fault and is potentially followed by stairstepping via Route G to an oil accumulation at Springdale and leak to the post-rift section across the Trinity Fault. (d) Far East Central connectivity diagram. Leak by capillary seal-failure (Route H) is shown notionally. This could occur via top, lateral or base seal. Route I invokes a capillary leak through degraded reservoirs or fault rocks to Lower Hibernia sandstones across the Trinity Fault, or via a low transmissibility dynamic connection. Subsequent stairstepping up section is indicated. (e) Completed connectivity diagram at major compartment scale for Terra Nova. Oil spill from Hebron and gas migration from West Bonne Bay have been added to each of the connectivity interpretations for the West Flank/ Graben/East Flank, Far East South and Far East Central described individually in Figure 11b–d.

352


Fig. 11. Continued.

eight water injectors and three (solution) gas injectors have been drilled from four drill centers. In reviewing the field after five years of production, Haugen et al. (2007) comment that, ‘initial concerns of barriers at sub-seismic faults have not

materialized,’ and, ‘material balance and history matching indicate extensive cross-fault connectivity.’ Two examples of this high level of connectivity that affect reservoir management are illustrated here.

Fig. 12. (a) Schematic map of Terra Nova production fault blocks and wells. Red areas are gas flood blocks, blue areas are water flood. Wells with small solid symbols have been drilled, small open symbols notionally indicate planned wells. Oil producers are green, gas injectors red and water injectors blue. Other wells are black. Red and blue arrows indicate injector-producer pairs. Drill centres are circled in yellow. Key wells referenced here are circled in blue, and green. The I-66 recent exploration well is circled in dashed red. (b) 3D juxtaposition diagram (‘Allan’ diagram) used to estimate cross-fault fluid flow and optimize well planning. Reservoir grid cells are open on the south (high) side of the fault, colour-filled on the north (low) side. A gamma ray log is shown at the PG1 well. (c) Seismic traverse illustrating a seismic interpretation refinement, (two faults instead of, previously, one) which achieves reservoir connectivity necessary to history match pressures.

TERRA NOVA, RCA

In the NW fault block in the Graben (WIG3 well, Fig. 12a, b) cross-fault connectivity to adjacent blocks was predicted pre-drill via inspection of 3D seismic data and the Operator’s geological model and construction of a 3D juxtaposition diagram (‘Allan diagram’) from the model. On this basis a water injector (WIG3) was drilled into the NW block and a complementary oil producer within the block was deliberately deferred. Subsequent pressure measurements and production rates at producing wells in adjacent fault blocks (PG1 and PG5) confirmed successful injection and cross-fault sweep of oil – enabling a considerable cost saving of one production well. The injector well (WIG3) and one of the producers (PG1) are shown in Figure 12b together with the 3D Allan diagram at the intervening fault. Reservoir grid cells on the high side of the fault are shown in outline and in colour-fill on the low side. Reservoir juxtapositions are outlined, and these areas were measured and calibrated against known cross-fault production elsewhere in the field. Figure 12c illustrates a simple iteration between seismic interpretation, RCA and history matching. Initially, a single fault was interpreted between the Graben and South Graben, with no reservoir juxtaposition or cross-fault connection. History matching indicates that these areas are connected and close inspection of the seismic data indicates that two faults can be interpreted with reservoir connections via an intermediate relay ramp.

Individual sandstone scale RCA In most production-scale fault blocks at Terra Nova, there is sufficient reservoir connectivity that only one oil accumulation and/or one water accumulation is evident. In the I-66 well in the Far East South, drilled in 2007, two oil accumulations at different pressures and with different API gravities were observed, separated by water at hydrostatic pressure (Fig. 13). The simpler, and favoured interpretation developed, allows an arrangement of fluids and pressures seen at the I-66 well that is explained geometrically with minor interpretation adjustments. Alternatively, these observations are satisfied with an interpretation of differential pressure depletion of an original, single oil accumulation with a deep aquifer, unpenetrated by the well. This interpretation requires contamination of PVT data to explain compositional differences between the oils and perched water within the reservoirs. The I-66 well provides an example of two key aspects of RCA: as a basis for scenario analysis and as a driver for improving interpretations beyond the inherent resolution of the data so that observed fluid and pressure arrangements are physically plausible.

353

The I-66 well tested multiple reservoirs within a lowside, fault-dependent, three-way dip closure at the Voyager Fault (the FES1 compartment). Structure at top Jeanne d’Arc reservoir level is shown in Figure 13a. Shallow targets in the Hibernia, Ben Nevis and other formations were wet (apart from one very minor show), interpreted to be either due to juxtaposition with permeable sandstones across the Voyager Fault or the presence of a hydrocarbon migration shadow (adjacent source rock to the east, in the footwall, is likely of low maturity). Jeanne d’Arc reservoirs contain oil, interpreted to result from a locally effective fault trap against impermeable Rankin Formation across the Voyager Fault. A shallow FWL at about 23415 m is interpreted in the D and UC sandstones from pressure–depth data and fluid samples (Fig. 13b). This observation is rationalized by a minor modification to the predrill structure map as shown in Figure 13a. This modification results in interpretation of a controlling saddle spill point on the SW side of the structure, to the H5S compartment to the west. A further interpretation is that a small near-crestal antithetic fault recognized seismically (Fig. 13c) has allowed leak of oil through the intermediate UC reservoirs in the well, which are of poor quality due to cementation and thus retain low oil saturation. There is seismic evidence that the deepest, LC, reservoir (with 368 API oil) fails to be offset by this crestal fault and laps onto Rankin to the west. The FWL in this reservoir, via pressure– depth data, is at about 23585 m. This FWL is interpreted to be controlled by saddle spill to the NE. Beyond this saddle, there is a small faulted culmination (with the FES1 N compartment on its west and the FEE2 compartment on its east). Juxtapositions at this fault allow upward stairstepping to a small D sandstone culmination – which fills, and then spills back to the SW to the FES1 compartment. This interpretation is systematically annotated and described in a detailed, local connectivity diagram at the individual reservoir scale, Figure 13d.

Conclusions In a prolific petroleum system, such as in the Jeanne d’Arc Basin, where all play elements are present and effective in abundance, reservoir connectivity is the primary control on fluid, pressure and resource distribution. It follows that there is greater value in a systematic rather than ad-hoc approach to reservoir connectivity. Reservoir Connectivity Analysis provides a rigorous, systematic approach. Compartments and the connections between compartments are identified and described by methodically integrating fluid pressures, contacts

354


Fig. 13. (a) Depth Structure map at top Jeanne d’Arc reservoir in the Far East South of Terra Nova. Four compartments are labelled: FES1, FES1 N, FEE2, H5S. The I-66 exploration well was drilled in 2007 and the fluid/pressure distribution encountered can be rationalized with the minor modifications to mapped structure indicated: a minor crestal fault and adjustment of the SW saddle. (b) I-66 well: well logs, and pressure–depth plot. Fluids and FWLs are interpreted from log, pressure and sample data. (c) Seismic traverse along I-66 borehole. A small crestal fault is interpreted in the UC and D sandstones. This is interpreted to have allowed upward stairstepping of oil through the marginal reservoir quality UC sandstone, which retains some oil saturation based on log interpretation. (d) Connectivity diagram. Interpreted oil and water are shown in green and blue, degraded reservoir in brown. One of several possible scenarios is presented here. Oil entering the deep FEE2 compartment leaks up-dip to either the FES1 N or FES1 compartments (annotated as connections 1 & 2). Connection 3 indicates spill to the north from the LC sandstone in the FES1 compartment into the FES1 N compartment at the intervening NE saddle (at approx. 23585 m). The high delta throw fault between FEE2 and FES1 N allows oil to stairstep up to the D sandstone (connection 4) in FES1 N, which fills south to the NE saddle and spills south back to FES1 (connection 6). The D sandstone in FES1 fills to the SW saddle at about 23415 m and then spills west to the H5S compartment (connection 7), The H5S compartment D sandstone fills and spills at a saddle on its west side at connection 8 – locally the system exit. Connection 5 occurs internally at the small crestal fault identified within the FES1 compartment and allows stairstepping of oil up through the UC sandstones to the D sandstone.

and geological interpretations with knowledge of the types of compartments and connections that are physically possible. Following iteration with initial geological interpretations, a connectivity model(s) is constructed which first, in itself, tests

for consistency and completeness, and then provides predictions within compartments and in undrilled compartments. Reservoir connectivity analysis is performed at multiple scales and levels of detail at all stages

TERRA NOVA, RCA

of the exploration, development and production cycle. Even at the development and production stage it is important to consider connectivity and compartments from basin scale to individual sandstone scale. Terra Nova field and the Terra Nova anticline provide an excellent example of this approach to better understand fluid and resource distribution. Basin-scale connectivity has to be considered in understanding and predicting fluid type (e.g. gas or oil) in step-out blocks. Despite a large structural rollover (at the tens of kilometres scale) the anticline as a whole is an inconsistent trap due to lack of dip closure at the low side of the Voyager Fault system and a high level of connectivity between blockfaulted, thick, high reservoir quality, sandstones at multiple stratigraphic levels. This is understood by methodically describing compartments and identifying the connections between them. At Jeanne d’Arc reservoir level, additional mechanisms – facies changes and pinch-out due to onlap and cementation – reduce the inherent connectivity of the system creating a commercial-scale hydrocarbon trap. RCA enables understanding of the major pressure compartments at Terra Nova, as well as at the connectivity at the production fault block scale where it is key to reservoir management, development planning and well optimization. The operator and partners for permission to publish: PetroCanada, ExxonMobil Canada, StatoilHydro, Husky, Murphy, Mosbacher and Chevron Canada. Geoscience and reservoir engineering colleagues who contributed to the ideas, discussions and products shown here, notably: Eric Albrechtsons, Jim Costello, Larry Wilcox, Irene Kelly and Geoff Minielly at Petro-Canada; Charlie Wierstra, Stacy Anderson, Ed Shaw, Michelle Lund, Denise Hodder, and John Eastwood at ExxonMobil Canada; Bill James (now retired), Rod Myers, Steve Davis and Lee Esch at ExxonMobil Upstream Research; Bill

355

Devlin at ExxonMobil Development Company; Einar Haugen at StatoilHydro; John Andrews, Iain Sinclair and Judith McIntyre at Husky; Javed Iqbal and Richard Dingwall at Mosbacher. The Canada Newfoundland and Labrador Offshore Petroleum Board (CNLOPB) and the Geological Survey of Canada (GSC) for their excellent databases and literature, much of which is available online.

References Dewan, J. T. 1983. Essentials of Modern Open-Hole Log Interpretation, PennWell, Tulsa. Enachescu, M. E., Harding, S. C. & Emery, D. J. 1994. Three-dimensional seismic imaging of a Jurassic paleodrainage, system, geology, earth sciences and environmental factors; Proceedings of the 1994 Offshore Technology Conference, Houston, TX, 26, 179– 191. Grant, A. C. & McAlpine, K. D. 1990. The continental margin around Newfoundland; Chapter 6. In: Keen, M. J. & Wiiliams, G. L. (eds) Geology of the Continental Margins of Eastern Canada (Geology of Canada), Geological Survey of Canada, 2, 239– 292. Haugen, E., Costello, J., Wilcox, L., Albrechtsons, E. & Kelly, I. 2007. Reservoir management challenges of the Terra Nova Field: lessons learned after 5 years of production. Society of Petroleum Engineers, SPE Paper 109587. James, W. R., Fairchild, L. H., Nakayama, G. P., Hippler, S. J. & Vrolijk, P. J. 2004. Fault-seal analysis using a stochastic multifault approach. American Association of Petroleum Geologists Bulletin, 88, 885– 904. Magoon, L. B., Hudson, T. L. & Peters, K. E. 2005. Egret-Hibernia(!), a significant petroleum system, northern brand banks area, offshore eastern Canada. American Association of Petroleum Geologists Bulletin, 89, 1203–1237. Vrolijk, P., James, B., Myers, R., Maynard, J., Sumpter, L. & Sweet, M. 2005. Reservoir connectivity analysis – defining reservoir connections and plumbing. Society of Petroleum Engineers Middle East Oil Show and Conference, SPE Paper 93577.

Index Page numbers in italic denote figures. Page numbers in bold denote tables. Aalenian, intra, unconformity 170 accommodation: coarse sediment supply, and compartmentalization potential 203, 204, 205– 208, 210 –215 alkylbenzene carbon analysis 12, 13, 61 Allan maps see fault plane profiles amplitude v. offset analysis 43, 46, 47 appraisal see field appraisal aquifer, hydrodynamic, Azeri field 109 –111 Auger Blue reservoirs 56– 69 4D seismic surveys 60–63 Blue O Massive 57, 58, 59, 60, 61– 63 geochemical fingerprinting 63–68 Blue O Stray 57, 58, 59, 60, 63 geochemical fingerprinting 66–68 cross-fault drainage 62–63, 64 geochemical fingerprinting 61, 63–68 geological setting and production history 57–58 Lower Blue O2 57, 58, 59, 60, 61– 63 geochemical fingerprinting 63–66 pulsed neutron logging 57, 58, 60, 61–63 stratigraphic communication 63, 65, 66– 68 Avalon Formation, reservoir connectivity analysis 337, 341 Azeri field 103–111, 104 hydrodynamic aquifer 109– 111 oil–water contact 109– 111 Pereriv B reservoir pressure 107 –109 baffles, membrane permeability 73, 258–259, 309 Balta Formation 288– 289 barium sulphate, precipitation 75 barriers, flow 26 inter sand-body 200, 202, 203, 205, 209, 210 inter-parasequence 200, 202, 203, 205, 208, 209 intra sand-body 200, 202, 206– 208, 209, 210 Nelson field 73, 80–82 N’Kossa field 150– 153 development plans 157–162 Pierce field 129– 131 Rotliegend gas fields 306– 312 Schiehallion field 92– 93, 94, 96–101 Skagerrak Formation 168, 186, 188–189 Base Cretaceous Unconformity 290 Ben Nevis Sandstone, reservoir connectivity analysis 337, 341, 342 bioturbation, Skagerrak 172, 173, 174 Brent Group fields drainage cells 72– 73 hydraulic units 72 Brent province reservoir, uncertainty modelling 288–297 Broad Fourteens Basin 311, 312 bubble point 33, 44–45, 46 bulk rock volume, uncertainty 15 Bunter sandstone 168, 169, 172, 192, 312

Canada, Terra Nova field, reservoir connectivity analysis 333 –355 capillary sealing 321– 322 failure 352 cataclasis, Rotliegend gas fields 308– 310 cataclasites 245, 253, 309 Central Graben 116, 125 Skagerrak depositional systems 167–168, 172–176 Central Offshore Saddle 307, 310 channel systems connectivity 223–224, 225 modelling 229–231 Forties Sandstone 114– 115, 121– 123, 126, 130 Nelson field 71, 73, 82 Skagerrak 172, 174–175 Chirag field 103, 104, 106 chloride ion concentration, Nelson field 75, 76–84 clay minerals Rotliegend gas fields 310–312 Skagerrak fault zones 179– 184, 192 Clay Smear Potential (CSP) 182, 184, 246, 247, 248, 265– 267, 266, 291 clay smears see also shale smears 179–184, 180, 184, 244–245, 265–268, 291– 292 compartmentalization challenges and impact 9– 20 definition 1–2 phase-behaviour modelling 47 pressure analysis 48– 51 seismic analysis 43, 46, 47 depositional systems 166, 200– 215 dynamic 1, 5, 219 fault see fault seal fluid phase composition 43–53 Heron Cluster fields 177– 194 identification 25– 26 impact on recovery 1, 27–31 turbidite oilfields 29– 31 inter sand-body 200, 202, 203, 205, 209, 210, 211 inter-parasequence 200, 202, 203, 205, 208, 209, 210, 211 intra sand-body 200, 202, 206–208, 209, 211 Pierce field 126–131 prediction 9, 200–215 role of accommodation: coarse sediment supply 203, 204 Rotliegend gas fields 301–313 Schiehallion field 91–101 static 1, 5, 55, 219 stratigraphic marginal marine reservoirs 199– 215 hierarchy 200, 202, 203 –208, 209 predictive matrix 208, 209 predictive tools 210– 214 compartmentalization complexity score 27– 28 complexity appraisal 16, 17, 30 index 27– 28

358 Congo, Republic of, N’Kossa field 133– 162 connections, Terra Nova field 346–348 connectivity 219–241 Azeri field 106 bulk reservoir 221 dynamic 219 definition 231–238 local characterizations 222 reservoir analysis, Terra Nova field 333 –355 reservoir modelling increasingly complex 222– 231 uncertainty 219– 220 reservoir stratigraphy and structure 231, 232 reservoir to well 221 Rotliegend gas fields 306–312 static 231– 232 connectivity flow diagram 237, 238– 241 Terra Nova field 333, 348–352 connectivity function 221–222 core data 11– 12 Cretaceous, base, unconformity 290 critical point 44– 45, 46 cut-off efficiency 28 data integration 10, 71–72 deformation, Skagerrak 170– 172, 179 deposition and compartmentalization 166 fluvial 172–176, 189–192, 200, 208, 209 marginal marine systems 200–215 Triassic, central North Sea 167– 168, 172 –176 dew point curve 44–45, 46 diagenesis N’Kossa field 146–148 Rotliegend gas fields 310–312 Skagerrak 179 diapirs, salt Pierce field 113, 115, 117, 119–122, 125–126, 130 see also halokinesis diffusion modelling 37–38 molecular 33– 34, 37, 38 dip leak 321 disaggregation zones 245, 248, 253 dolomite, N’Kossa field 135, 142, 143, 144, 146– 148, 149 drainage cells 72–73 Nelson field 73– 86, 74, 84, 85 drainage chart, Nelson field 84, 86 drainage efficiency factor 28– 29, 30, 31 Eastern Channel, Nelson field 71, 76 produced water 77– 78, 79, 81, 83 effective cross-fault permeability 272 effective cross-fault transmissibility 272 effective Shale Gouge Ratio (ESGR) 264– 265, 266 efficiency factors 28–29 Egret field 166, 168, 171, 173, 176, 184, 189 dynamic behaviour 184– 185 Egret source rock 351 Ekofisk Formation 116, 117–118, 177 electrofacies, N’Kossa field 149, 154 enhanced oil recovery, injection data 12 erosion, Jurassic, Central North Sea 170–172

INDEX Etive Formation 288– 289 extension Central North Sea 170– 171 N’Kossa field 136, 138, 139 fault plane profiles 257, 258, 260, 271, 317, 318 fault seal 5, 165–166, 178–184, 231, 232– 233, 243–253, 257–280, 308–310, 317–330 algorithms 5, 246– 248, 247, 263–268, 266 analysis 317 capillary sealing 321– 322 dip leak 321 multi-fault 318–330 success rate (1994–2001) 319–321 uncertainty 322–324 Azeri field 106, 111 calibration 246–253 deterministic approach 248–249, 252–253 empirical approach 249– 253 fault reactivation seal 244 fault-rock seal 244 –253 Heron Cluster fields 181– 184 N’Kossa field 154 Pierce field 125–129 Rotliegend gas fields 308– 310 stratigraphic layering, visualization 257, 260–261 visualization tools 259– 268 fault zone thickness, modelling 286 faults fluid flow Auger Blue reservoirs 62–63, 64, 66 flow simulation data analysis 275– 279 properties predictive capabilities 276 –277 uncertainty modelling 285– 288, 291– 297 visualization 268– 274 Heron Cluster fields 176– 177, 192 permeability 178–181 sealing potential 181– 184 hydraulic resistance 270– 272, 279 juxtaposition 257, 323– 324 sand-to-sand 317, 318 seal 244 visualization 260– 261, 263 leaking 231, 232– 233 prediction 11– 12 and secondary recovery 12 transmissibility 235, 237, 238, 268, 270–274, 276–277, 279 multipliers 258– 259, 262– 263, 264, 268–274 Ferron Sandstone 208 field appraisal 1, 9, 11, 15–16, 17 Azeri field 103, 107 turbidite oilfields 29–31 fluid flow 113 cross-fault 5 Auger Blue reservoir 62–63, 64 properties predictive capabilities 276 –277 uncertainty modelling 285– 288, 291– 297 visualization 261, 262, 268–274 simulation data analysis 275–279 see also fluids, reservoir, cross-fault drainage

INDEX debris 120 depositional systems 166 gravity 120 hybrid, Pierce Field 120 sealing faults 165– 166 sedimentological control heterogeneities 200, 202, 206– 207, 219, 230 N’Kossa field 150– 153 Pierce field 129 –131 fluids, reservoir critical 45 cross-fault drainage 62–63, 64, 66 see also fluid flow, cross-fault density differences 43–53 gravity segregation 46, 47, 49, 51 formation pressure testing 92 gravitational overturning 36– 37, 38 integrated description 12– 13 integrated toolkit 10– 11 mixing 31– 39 unified diffusive model 37– 39 near-critical 45– 46, 47, 51 phase composition 43–45 dual phase 45, 46, 50 single phase 45, 46, 51 pressure dissipation 35– 36, 37 properties, variation 26, 31– 33 stratigraphic communication 63, 66–68, 231, 232 fluvial facies, Skagerrak Formation 168, 170, 172–176 forensic analysis 16, 20 formation pressure testing 92 formation volume factor 51 Forties Fan system 71 Forties Montrose High 71, 72, 84, 171, 176 Forties Sandstone 113, 114– 118 depositional environment 120 –123, 127 Lower 118– 122 Upper 120, 122–123, 127 fractures, prediction 11– 12 free-water level Pereriv B reservoir 109 Rotliegend gas fields 307 Terra Nova field 347–348, 354 Gannet/Guillemot fan 114 gas dry 45, 46 injection, N’Kossa field 157–158, 159, 161 wet 45 gas cap 43, 46, 47, 50 gas condensates 44, 45, 46, 46, 50–51 Blue O Massive reservoir 63, 65–66 gas fields, Rotliegend, compartmentalization 301 –313 gas migration shadow, Terra Nova field 339, 340 gas–oil ratio 30, 33, 43 Auger Blue reservoirs 66 Schiehallion field 96 West African fields 47, 48, 51 geochemistry Pierce field connectivity 123–125, 127, 128 produced water, Nelson field 74–86 time-lapse 55– 56 Auger Blue reservoirs 61, 63–68 Germanic Trias Group, gas fields 312

359

Groningen field 301–302 Gulf of Mexico Auger Blue reservoirs 56–69 field complexity 16, 17, 19–20 Gunashli field 103, 104, 106 halokinesis, Central North Sea 170 Hebron field 334, 335, 336–337 Heron Cluster fields 165– 194 compartmentalization 177–194 by shales 185– 186, 188– 192 dryland fluvial facies 172 –176, 189 –190 biogenically disrupted sheet deposits 172– 173 channel belt deposits 172 channel-fill deposits 174– 175 floodplain deposits 173, 175 lake and palustrine deposits 173–174, 175 playa deposits 175 terminal splay complexes 175 dynamic behaviour 184–185 faults 176 –177, 192 permeability 178– 181 sealing potential 181–184 reservoir barriers 168 structural framework 176–177 Triassic depositional systems 167– 168 Heron field 166, 168, 171, 176, 189 dynamic behaviour 184, 186 Heron Shale 168, 172, 173, 174, 176, 188, 189 heterogeneity, permeability 200, 202, 206–207, 219, 230, 233–235, 236 Hibernia field 334, 335, 336, 339 Hibernia Sandstones, reservoir connectivity analysis 337, 339, 341, 342, 343, 345, 348, 351– 352 hydraulic units 72 hydrocarbons, composition 33–34, 37 hydrostatic gradient, Azeri field 109 infill targets, Schiehallion field 96–101 injection N’Kossa field 157–162 Pierce field 129–130 Schiehallion field 96–101 secondary recovery 12 and micro-seismic events 14 J-Ridge 168, 169 Jackson field, Australia, water shut-off 80 Jeanne d’Arc Basin 334, 335, 336 –337, 340 stratigraphy 337 Terra Nova field, reservoir connectivity analysis 339 –355 Jeanne d’Arc sandstones, reservoir connectivity analysis 337, 341, 344, 345, 346–348, 351 Judy Member sandstone 169, 170, 172, 186 Julius Mudstone 168, 169, 170, 172, 189 Keuper sandstone 169 lake deposits, Skagerrak 173 –174, 175 Last Chance Sand 71 produced connate water 77, 78, 79, 83 Last Chance Shale 71, 73 Lauwerszee Trough, faulted traps 324– 330

360 layering, stratigraphic, models 257–258 limestone, N’Kossa field 142, 143, 144–145, 146, 151 Lista Formation 116, 117, 118, 119 Loeme Salt 133, 136, 138, 140 Lower Forties Sandstone 118 –122 Marnock Shale 170, 175, 178, 189, 190 Maureen Formation 116, 117, 118 membrane seal 73, 258–259, 309 visualization 261 –268 micro-seismicity, surveillance 13–14 Modular Formation Dynamics Tester 10, 18, 33, 35 mud drapes 223–224, 230 mudstone, Skagerrak Formation 168, 169, 170, 172, 173–174 Multiple Point Statistics algorithm 220, 229 Muschelkalk 169, 170 Nelson field 71– 86, 72 chloride ion concentration 75, 76– 84 compartmentalization 82, 84 see also Nelson field, drainage cells data integration 71–72 drainage cells 73–86, 74, 84, 85 drainage chart 84, 86 flow barriers 73 macroforms 73, 82 produced water chemistry analysis 74–86 Ness Formation 288 –289 Netherlands, Rotliegend gas fields, compartmentalization 301– 313 Newfoundland, Terra Nova field, reservoir connectivity analysis 333–355 Nigeria, deepwater fields, fluid phase composition 44, 45–53 N’Kossa field 133–162, 134 connectivity problems 135–136, 150– 158 development plans 157–162 diagenesis 146–148 dynamic synthesis 153–157 extension 136, 138, 139 rafting 140 reservoir barriers 150– 157 reservoir characteristics 149– 150 sedimentology 141–148 oil black 45, 46, 47, 51–52 Blue O2 reservoir 63, 65–66 mixing 31–34 volatile 45, 46, 46 Blue O2 reservoir 63, 65–66 volume, uncertainty 15, 16 oil– water contact, Pereriv B reservoir 109– 111 optical fluid analyzer 57–58 overturning, gravitational 36– 37, 38 Panther Tongue deposit 207–208 parasequences, compartmentalization 200, 202, 203, 205, 208, 209 Peng-Robinson EOS 47 percolation theory 221, 223– 224, 226– 228, 230 Pereriv B reservoir 103, 105, 106 pressure analysis 107–109

INDEX permeability cross-fault 268 –270, 272–273, 274, 276– 279 modelling 285–288, 291, 296 heterogeneity 200, 202, 206–207, 219, 230, 233–235, 236 phase-behaviour modelling 47 phyllosilicate-framework fault rock 245, 246, 248, 253, 267–268 Skagerrak 170, 179–181, 184 Pierce field 113, 115– 131 aquifer support 130–131 compartmentalization 126– 131 depositional environment 120–123 fault seal potential 125–129 fluid flow, sedimentological control 129– 131 geochemical fingerprinting 123– 125, 127, 128 Lower Forties Sandstone 118– 122 salt diapirs 113, 115, 117, 119–122, 125–126, 130 seismic stratigraphy 117– 123 stratigraphy 116–117 Upper Forties Sandstone 122–123 playa deposits, Skagerrak Formation 168, 169, 170, 175, 190 Pointe Noire Marl 137, 139 porescale displacement efficiency factor 28 pressure, fluid 35–36, 37, 38 pressure analysis compartmentalization definition 48– 51 Pereriv B reservoir 107 –109 pressure testing, formation, Schiehallion field 92 pressure– volume– temperature analysis 12, 13 Auger Blue reservoirs 59, 63 Nigerian deepwater fluid 44–45, 47, 48, 49–51 probabilistic shale smear factor (PSSF) 265– 268, 266 production data 12 in performance prediction 16, 18–20 production logging tools, Schiehallion field 92, 96 pulsed neutron logs, Auger Blue O Massive 57, 58, 60, 61– 63 quartz, authigenic, Rotliegend gas fields 310 –311 rafting, N’Kossa field 140 Rankin Formation 337, 338, 339, 347 recovery factor 27–28 recovery, hydrocarbon efficiency factors 28– 29 impact of compartmentalization 27– 31 secondary 12 repeat formation testers 33, 35 reservoirs complexity, appraisal 16, 17 connectivity 219 –241 analysis 6, 333, 336 Terra Nova field 339– 355 bulk 221 function 221– 222 local characterizations 222, 230 modelling channel systems 229– 231 increasingly complex 222–231 uncertainty 219–220 net to gross 223– 224, 226– 228, 230 stratigraphy and structure 231, 232

INDEX to producing wells 221, 230 volume support effect 224, 226, 230 dryland fluvial, Heron Cluster fields 166–194 fluid properties 26, 31–33 forensic analysis 16, 20 heterogeneity 200, 202, 206–207, 219 integrated toolkits 10– 11, 18–19, 20 management 16, 18–20 marginal marine, stratigraphic compartmentalization 199 –215 modelling integrated 4D, Schiehallion field 94 uncertainty, workflow 284–297 structure, description 11–12 tortuosity 224, 231, 232– 233, 234 resistance, hydraulic, faults 270–272, 279 Resistivity Index 222 Rotliegend fault seal analysis 324– 329 gas field compartmentalization 301–313 connectivity factors 306–312 IGIP 301, 305 –306 stratigraphy 302 –304, 310 Heron Cluster fields 170, 176– 177 sabkha, N’Kossa field 144, 146, 148, 150 salinity, formation water, Nelson field 76– 84 salt Heron Cluster fields 171, 176 N’Kossa field 133, 136, 138, 140 Pierce field 113, 115, 117, 119–122, 125–126, 130 see also halokinesis; Zechstein halite sand-bodies, compartmentalization 200, 202, 203, 205–208, 209, 210 sandstone N’Kossa field 135, 142, 143, 144, 149 Skagerrak Formation 168, 169, 170, 172, 174–175 Schiehallion field 89, 90 compartmentalization 91–101 4D integrated reservoir modelling 94 field surveillance 91–96 full field model 93–94 infill targets 96–101 Seagull field 171, 177 compartmentalization 185, 186, 188 oil fingerprinting 186, 187, 188 sediment supply, and compartmentalization 203, 205– 208, 210 –215 segregation, gravity 46, 47, 49, 51 seismic amplitude, Azeri field 110 seismic amplitude v. offset data 43 seismic analysis 4D Auger Blue reservoirs 60–63 Schiehallion field 93, 94–96, 100 and fluid compartmentalization 43, 46, 47 Sele Formation 116, 117–118 Sendji Carbonate 133, 134, 136– 137 diagenesis 146–148 lithofacies 148– 149 reservoir barriers 150–157 reservoir characteristics 149–150 sedimentology 141– 148 stratigraphy 140– 141

361

transgressive– regressive sequences 140– 141, 145–146, 147, 150 –151 Sequential Indicator Simulation algorithm 220 shale Nelson field 80– 82 Skagerrak Formation 173– 176, 182– 186, 188– 192 see also Heron Shale; Marnock Shale Shale Gouge Ratio (SGR) 126, 181, 182, 246 –249, 247, 262, 263– 267, 266, 287 Shale Smear Factor (SSF) 246, 247, 265, 266, 291 shale smears see also clay smears 179–184, 180, 184, 244– 245, 265– 268, 291– 292 sheet deposits connectivity 223–224, 225 Skagerrak 172– 173 siltstone N’Kossa field 142, 143, 145 Skagerrak 173 Skagerrak Formation 166, 193 compartmentalization by shales 185– 186, 188– 192 dryland fluvial facies 168, 170, 172 –176 fault seal potential 181–184 fault zone permeability 178–181 preservation 170– 173 reservoir barriers 168 sedimentology 172–176 stratigraphy 168–170 Skua field 166, 168, 171, 176 –177, 178 compartmentalization 185, 186 oil fingerprinting 186, 187, 188 shale–fault interaction 189 Slochteren Formation 302– 304, 307, 310, 312, 324 Smith Bank mudstone 168, 169, 170, 171, 172 South Caspian Basin, ACG fields 103, 104, 106 splay complexes, Skagerrak 175, 190, 191 Sr isotope ratios 34, 37 Heron Cluster fields 185, 188 STOOIP uncertainty 15, 16 turbidite oilfields 29– 30, 31 stochastic multi-fault analysis 6, 317 –330 stratigraphy and compartmentalization dryland fluvial 166, 172– 176, 189– 190 marginal marine 200–215 Rotliegend gas fields 310 and connectivity 63, 66–68, 231, 232 Jeanne d’Arc Basin 337 sequence, as predictive tool 210, 211 Skagerrak Formation 168– 170 Sunrise field 210, 212, 213 sweep efficiency 28– 29, 30, 233, 234, 235, 237 systems tracts and compartmentalization 200, 201, 203, 204, 205, 206 see also transgressive –regressive sequences Tarbert Formation 288–289, 290 tectonics, Triassic– Jurassic, and Skagerrak preservation 170–173 Terra Nova field 334– 336, 335, 338–339 gas migration shadow 339, 340 reservoir connectivity analysis 336, 339– 355 Beothuk Member 347 connection definition 346–348

362 Terra Nova field (Continued) connectivity diagram 333, 348, 349 –351, 351– 352 East Flank 344–345, 347, 351 Far East 344– 347, 351 Graben 344, 345, 347, 351, 353 individual sandstone scale 354 production fault blocks 352– 354 Trinity Fault 345, 348, 352 Voyager Fault 348, 351, 354 West Flank 344, 346– 347, 351 reservoir overview 338– 339 Texel-IJsselmeer High 303, 311, 312 toolkits, integrated 10– 11, 18–19, 20 tools, visualization 259–268 tortuosity 224, 231, 232–233, 234 transgressive– regressive sequences, Sendji Carbonate 140– 141, 145– 146, 147, 150–151 transmissibility 235, 237, 238, 268 cross-fault 270– 274, 276– 277, 279 multipliers 258–259, 262–263, 264, 268–274 traps, multi-fault analysis 318– 330 Triassic, depositional systems, central North Sea 167–168, 169, 171– 176 Triassic–Jurassic, deformation and erosion, central North Sea 170–172, 179 Trinity Fault 345, 348, 352 turbidites Auger Blue reservoirs 57 Nelson field, compartmentalization 73, 82 offshore oilfield, compartmentalization 30– 31 onshore oilfield, compartmentalization 29–30 Pierce field 114, 117–123 Schiehallion field compartmentalization 89– 101

INDEX geometrical, modelling 284– 285, 290– 291, 296 management 14–16, 17, 19– 20 modelling, workflow 284–297 Brent province reservoir 288–297 unconformity base Cretaceous 290 intra-Aalenian 170 uplift, thermal, Jurassic, Central North Sea 170 Upper Forties Sandstone 122–123, 126, 127 vector fields, cross-fault flow analysis 276 Voyager Fault 348, 351 water formation, mixing 34, 37 injection 12 and micro-seismicity 14 N’Kossa field 157 –158, 159, 160, 161 –162 Pierce field 129–130 Schiehallion oil field 96– 101 perched, Terra Nova field 335– 336, 338, 344, 347, 354 produced connate Nelson field 74–86 shut-off 75, 77, 80, 81 water chemistry analysis 74–86 see also free-water level well log data, as predictive tool 210, 212 well testing extended 19, 26, 30 Schiehallion field 92 Western Channel, Nelson field 71 drainage chart 86 produced water 77–78, 79, 81, 83 xylene, Auger Blue reservoirs 66–67, 68

uncertainty fault properties, modelling 285– 288, 291 fault seal analysis 322– 324

Zechstein halite 170, 171, 176, 305, 312, 324 Pierce salt diapirs 126

Reservoir compartmentalization – the segregation of a petroleum accumulation into a number of individual fluid/pressure compartments – controls the volume of moveable oil or gas that might be connected to any given well drilled in a field, and consequently impacts ‘booking’ of reserves and operational profitability. This is a general feature of modern exploration and production portfolios, and has driven major developments in geoscience, engineering and related technology. Given that compartmentalization is a consequence of many factors, an integrated subsurface approach is required to better understand and predict compartmentalization behaviour, and to minimize the risk of it occurring unexpectedly. This volume reviews our current understanding and ability to model compartmentalization. It highlights the necessity for effective specialist discipline integration, and the value of learning from operational experience in: detection and monitoring of compartmentalization; stratigraphic and mixed-mode compartmentalization; and faultdominated compartmentalization.of continental collision and growth zones.

Reservoir Compartmentalization

Reservoir Compartmentalization

Reservoir sandstones

The Reservoir

The Reservoir

The Reservoir

Reservoir Geomechanics

Advanced Reservoir Engineering

Reservoir Engineering Handbook

Reservoir Capacity and Yield

Essentials of reservoir engineering

Basic Applied Reservoir Simulation

Reservoir Engineering Handbook

BASICS RESERVOIR ENGINEERING

Reservoir Engineering Handbook

Reservoir Engineering Handbook

Petroleum Reservoir Simulation

Reservoir Flood Estimation

Geostatistical Reservoir Modeling

Advanced reservoir engineering

Reservoir Engineering Handbook

Carbonate Reservoir Characterization

Reservoir Stimulation , 3rd Edition

Geostatistical Reservoir Modeling

Reservoir capacity and yield

Structural geology in reservoir characterization

Petroleum Reservoir Engineering Physical Properties

Principles of Applied Reservoir Simulation

Geothermal Reservoir Engineering, Second Edition

Working Guide to Reservoir Engineering

Quantitative Methods in Reservoir Engineering

Reservoir Compartmentalization