Norwegian Petroleum Society (NPF), Special Publication No. 6
Quantification and Prediction of Hydrocarbon Resources Proceedings of the Norwegian Petroleum Society Conference, 6-8 December 1993, Stavanger, Norway
Further titles in the series:
1. R.M. Larsen, H. Brekke, B.T. Larsen and E. Talleraas (Editors) STRUCTURAL AND TECTONIC MODELLING AND ITS APPLICATION TO PETROLEUM GEOLOGY- Proceedings of Norwegian Petroleum Society Workshop, 18-20 October 1989, Stavanger, Norway 2. T.O. Vorren, E. Bergsager, Q.A. DahI-Stamnes, E. Holter, B. Johansen, E. Lie and T.B. Lund (Editors) ARCTIC GEOLOGY AND PETROLEUM POTENTIAL- Proceedings of the Norwegian Petroleum Society Conference, 15-17 August 1990, Tromso, Norway 3. A.G. Dor~ et al. (Editors) BASIN MODELLING" ADVANCES AND APPLICATIONS- Proceedings of the Norwegian Petroleum Society Conference, 13-15 March 1991, Stavanger, Norway 4. S. Hanslien (Editor) PETROLEUM" EXPLORATION AND EXPLOITATION IN NORWAYProceedings of the Norwegian Petroleum Society Conference, 9-11 December 1991, Stavanger, Norway
5. R.J. Steel, V.L. Felt, E.P. Johannesson and C. Mathieu (Editors) SEQUENCE STRATIGRAPHY ON THE NORTHWEST EUROPEAN MARGIN Proceedings of the Norwegian Petroleum Society Conference, 1-3 February, 1993, Stavanger, Norway
Norwegian Petroleum Society (NPF), Special Publication No. 6
Quantification and Prediction of Hydrocarbon Resources Proceedings of the Norwegian Petroleum Society Conference, 6-8 December 1993, Stavanger, Norway
Edited by
A.G. Dore
Statofl UK Ltd, Swan Gardens 10, Piccadilly, London W1V OLJ, UK
and
R. Sinding-Larsen
Department of Geology and Mineral Resources Engineering, The Norwegian Institute of Technology, N-7034 Trondheim, Norway
ELSEVIER Amsterdam - Lausanne - New York-
O x f o r d - S h a n n o n - T o k y o 1996
ELSEVIER SCIENCE B.V. Sara Burgerhartstraat 25 P.O. Box 211, 1000 AE Amsterdam, The Netherlands
ISBN 0-444-82496-0 91996 Elsevier Science B.V. All rights reserved. No part of this publication may be reproduced, stored in a retrieval system or transmitted in any form or by any means, electronic, mechanical, photocopying, recording or otherwise, without the prior written permission of the publisher, Elsevier Science B.V., Copyright and Permissions Department, P.O. Box 521, 1000 AM Amsterdam, The Netherlands. Special regulations for readers in the USA--This publication has been registered with the Copyright Clearance Center Inc. (CCC), 222 Rosewood Drive, Danvers, MA 01923, USA. Information can be obtained from the CCC about conditions under which photocopies of parts of this publication may be made in the USA. All other copyright questions, including photocopying outside the USA, should be referred to the copyright owner, Elsevier Science B.V., unless otherwise specified. No responsibility is assumed by the publisher for any injury and/or damage to persons or property as a matter of products liability, negligence or otherwise, or from any use or operation of any methods, products, instructions or ideas contained in the material herein. This book is printed on acid-free paper. Printed in The Netherlands
Preface The international conference "Quantification and Prediction of Hydrocarbon Resources", held in Stavanger in December 1993, was something of an experiment. Unlike other conferences arranged by the Geological and Geophysical wing of NPF, which are mainly aimed at a G&G audience, this conference was advertised to a wider group of disciplines. Why? Because the issue of resource quantification is the confluence at which the upstream petroleum disciplines (geology, geophysics, engineering and economics) meet. This meeting of minds, now a matter of routine in the workplace, can be directly attributed to the economic dictates of the last decade. The oil price shocks of the mid-1980s created major enigmas for those concerned with petroleum resources and their prediction. On the one hand, it was now clear that, contrary to the views of virtually all of the pundits of previous decades, the "big pump" in the Middle East was going to be supplying low-cost oil to the rest of the world for some time to come. On the other hand, petroleum (and particularly oil) was still as finite a resource as it ever had been. How could these two apparently contradictory situations be reconciled? And what were the consequences for resource management and economic prediction? Despite the gloomier forecasts, the upstream oil industry did not go to the wall. Instead, threatened with extinction it evolved to adapt to its new economic environment. In a world apparently awash with oil, new low-cost solutions became a necessity in established operating areas, while new technology evolved rapidly so that exploration could continue in difficult plays or frontier areas. For much of the workforce, however, the gravy train was derailed forever. Sections of the industry were cut to the bone, and those who survived found that little remained of the freewheeling individualism of the early 80s. Increasingly, the emphasis moved towards teamwork, involving the integration of disciplines and with the express purpose of realising short-term value from assets. Hydrocarbon discoveries and exploration prospects were no longer viewed as a series of separate calculated risks, but were instead regarded as an economic portfolio, to be manipulated and optimised in the same way as stocks and shares. Risk analysis and portfolio management, essentially the province of the engineers and economists before the mid-80s, became second nature to the geoscientists too. Given this developing climate, a conference was considered timely to assess the newly emerging approaches to resource quantification and prediction. The scope was kept deliberately broad, ranging from the macro (global) to micro (field and prospect) level. This book essentially follows the same themes of the conference, and is divided into six parts. Worm Hydrocarbon Resources (Part I), contains three papers which, although very different in style and emphasis, are concerned with the same basic issues - the amount of petroleum in the world and its longevity. Part II tackles the thorny question of Definition and Classification of Petroleum Resources. Although at first glance a semantic exercise, closer examination soon shows the importance of this issue. Put quite simply, resources change according to how they are defined, and meaningful debate on resources is therefore impossible without terms of reference. Part III, Assessment of Undiscovered Hydrocarbon Resources examines resource forecasting on a basin or province level from both a national and a company perspective. This is followed by a section on the general topic of Risk Analysis (Part IV), with emphasis on the play and prospect level. A wide-ranging selection of papers covers ground from formal risking techniques, through probabilistic assessment of volumes, to the control of specific input variables via basin modelling. Part V, Management of Discovered Resources, describes efforts to optimise the performance of fields through their life-cycles and the status of enhanced oil recovery techniques. Although economic considerations underpin all of the papers in this volume, Part VI, The Economic Interface contains a series of papers with expressly economic themes. They serve to illustrate the
Vl
Preface
importance of economic planning in allocating financial resources to exploration and production projects at company level, at national level and in the banking sector. As editors, we readily admit that the broad scope of this volume precludes an exhaustive coverage. Each of the sections could, in fact, quite easily be the subject of a separate book. Accordingly, some sections will be of interest to geoscientists, some to economists, and others to engineers and managers. On the other hand, we hope that by casting this net wide we have at least obtained an overview of thought processes currently prevalent in the industry and academia on the subject of Quantification and Prediction. These thought processes range from the integration of geological information and the application of methods to estimate undiscovered resources, to the modelling of the full cycle from exploration to development. Although hydrocarbon exploration has often been examined as an economic activity and described in terms of productivity, there have been few attempts at integration of theory, methods and models for the economic valuation of hydrocarbon resources. We strongly believe that continued close cooperation between geoscientists, engineers and economists is needed to instigate improvements in the methodologies, and to generate a new breed of explorationist that can use these integrated techniques effectively. We would like to warmly thank Elizabeth Holter and Karin Haugnaes (NPF) and Annette Leeuwendal (Elsevier) for their help and patience as this book came together. Special thanks go to the numerous unpaid workers - the authors and referees - who managed to borrow time from their busy schedules to help bring this volume into existence. A.G. Dor6 and R. Sinding-Larsen 1996
VII
List of Contributors A.E. ABBOTT
Department of Geological Sciences, University of South Carolina, Columbia, SC 29208, USA
K.A. ABRAHAMSEN
Statoil, N-4035, Stavanger, Norway
K. ASHTON
Phillips Petroleum Company UK Ltd., 35, Guildford Road, Woking, Surrey GU22 7QT, UK
T.J. BEARDALL
ERC Tigress Ltd., Chapel House, Liston Road, Marlow, Bucks, SL7 1XJ, UK Present address: T.J. Beardall and Associates, Ltd., 2 Silver Lane, West Challow Wantage, Oxon OX12 9TX, UK
H. BREKKE
Exploration Department, Norwegian Petroleum Directorate, P.O. Box 600, N-4001 Stavanger, Norway
C.J. CAMPBELL
c/o Petroconsultants S.A., P.O. Box 152, 1258 Perly, Geneva, Switzerland
S. CAO
Department of Geological Sciences, University of South Carolina, Columbia, SC 29208, USA
Z. CHEN
Department of Geology and Mineral Resources Engineering, The Norwegian Institute of Technology, N- 7034, Trondheim, Norway
B. DAHL
Norsk Hydro Research Centre, Bergen, Norway Present address: Dept. of Geology, University of Bergen, Norway
E. DAMSLETH
Norsk Hydro, P.O. Box 200, 1321 Stabekk, Norway
A.G. DORt~
Statoil UK Ltd, Swan Gardens, 10 Piccadilly, London W1V OHL, UK
D.W. DORN-LOPEZ
Conoco Norway Inc., P.O. Box 488, N-4001 Stavanger, Norway
B.A. DUFF
Fina Exploration Norway, Skogstostraen 37, P.O. Box 4055 Tasta, N-4004 Stavanger, Norway Present address: Fina Exploration & Production, Rue de l'Industrie 52, B-1040 Brussels, Belgium
N. FULLER
Phillips Petroleum Company UK Ltd., 35, Guildford Road, Woking, Surrey GU22 7QT, UK
S. GRANT
BP Norge UA, P.O. Box 197, Forusbeen 35, 4033 Forus, Norway
H.H. HALDORSEN
Norsk Hydro a.s., P.O. Box 200, N-1321 Stabekk, Norway
D. HALL
Fina Exploration Norway, Skogstostraen 37, P.O. Box 4055 Tasta, N-4004 Stavanger, Norway Present address: Fina Exploration & Production, Rue de l'Industrie 52, B-1040 Brussels, Belgium
C. HERMANRUD
Statoil, Postuttak; 7004 Trondheim, Norway
R.G. HEYWOOD
Phillips Petroleum Company UK Ltd., 35, Guildford Road, Woking, Surrey GU22 7QT, UK
List of Contributors
VIII R.W. HOLT
Phillips Petroleum Company UK Ltd., 35, Guildford Road, Woking, Surrey GU22 7QT, UK
C. JOURDAN
Statoil, P.O. Box 300, 4001 Stavanger, Norway
J.-E. KALHEIM
Exploration Department, Norwegian Petroleum Directorate, P.O. Box 600, N-4001 Stavanger, Norway
G.M. KAUFMAN
Sloan School of Management, MIT, Room E53-375, Cambridge, MA 02142, USA
B.L. KING
Phillips Petroleum Company UK Ltd., 35, Guildford Road, Woking, Surrey GU22 7QT, UK
K.R. KNUDSEN
Norwegian Petroleum Directorate, Postbox 600, 4001 Stavanger, Norway
W. KROKSTAD
IKU Petroleum Research, N-7034 Trondheim, Norway
I. LERCHE
Department of Geological Sciences, University of South Carolina, Columbia, SC 29208, USA
K. LINDBO
Statoil, N-4035, Stavanger, Norway
E. MAYORGA-ALBA
The World Bank, 1818 H Street NW, Washington, DC 20433, USA
E MCGAUGHRIN
Phillips Petroleum Company UK Ltd., 35, Guildford Road, Woking, Surrey GU22 7QT, UK
I. MEISINGSET
Norsk Hydro Exploration and Production, Oslo, Norway
R.G. MILLER
BP Exploration Operating Company Limited, 4/5 Long Walk, Stockley Park, Uxbridge, Middlesex UBll 1BP, UK
N. MILTON
BP Norge UA, P.O. Box 197, Forusbeen 35, 4033 Forus, Norway
S.J. MORBEY
Strategic Exploration, WEBG, Amoco Exploration and Production, Houston, USA
S. NORDAHL
Statoil, P.O. Box 300, 4001 Stavanger, Norway
K.O. SANDVIK
IKU Petroleum Research, N-7034 Trondheim, Norway
R. SINDING-LARSEN
Department of Geology and Mineral Resources Engineering, The Norwegian Institute of Technology, N-7034, Trondheim, Norway
S. SMITH
The World Bank, 1818 H Street NW, Washington, DC 20433, USA
J.H. SNOW
Conoco Norway Inc., P O. Box 488, N-4001 Stavanger, Norway
J.J.G. STOSUR
U.S. Department of Energy, Fossil Energy-33, Germantown, Washington, D. C. 20586, USA
~. SYLTA
IKU Petroleum Research, N-7034 Trondheim, Norway
M. THOMPSON
BP Norge UA, P.O. Box 197, Forusbeen 35, 4033 Forus, Norway
E. TORHEIM
Elf Petroleum Norge a.s., P.O. Box 168, N-4001 Stavanger, Norway
J. VOLLSET
Statoil, N-4035, Stavanger, Norway
S. WHITE
Phillips Petroleum Company UK Ltd., 35, Guildford Road, Woking, Surrey GU22 7QT, UK
J. WILKINSON
Esso UK, 94-98 Victoria St., London SWIE 5JW, UK
E.V. ZAKHAROV
VNIIGAS International p. Razvilka, Leninsky raion, Moskovskaya oblast, 142717, Russia
IX
Contents Preface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . List of Contributors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
V VII
I. World Hydrocarbon Resources World oil: reserves, production, politics and prices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . C.J. Campbell
1
Gas in the 21st century: a world-wide perspective . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . I. Lerche
21
Estimating global oil resources and their duration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . R.G. Miller
43
II. Definition and Classification of Hydrocarbon Resources The world of reserve definitions - - can there be one set for everyone? . . . . . . . . . . . . . . . . . T.J. Beardall
57
Resource classifications and their usefulness in the resource management of an oil c o m p a n y . . K.A. Abrahamsen, K. Lindbo and J. Vollset
63
Reserve and resource definition: dealing with uncertainty . . . . . . . . . . . . . . . . . . . . . . . . . . J. Wilkinson
71
The Norwegian Petroleum Directorate's Resource Classification System . . . . . . . . . . . . . . . . K.R. Knudsen
77
III. Assessment of Undiscovered Hydrocarbon Resources A method for the statistical assessment of total undiscovered resources in an area . . . . . . . . . E. Damsleth The Norwegian Petroleum Directorate's assessment of the undiscovered resources of the Norwegian Continental s h e l f - background and methods . . . . . . . . . . . . . . . . . . . . . . H. Brekke and J.-E. Kalheim Cross-validation of resource estimates from discovery process modelling and volumetric accumulation modelling: example from the Lower and Middle Jurassic play of the Halten Terrace, offshore Norway . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . R. Sinding-Larsen and Z. Chen The Russian method for prediction of hydrocarbon resources of continental shelves, with examples from the Barents Sea . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . K.O. Sandvik and E.V. Zakharov Offshore Brazil: analysis of a successful strategy for reserve and production growth . . . . . . . . S.J. Morbey
83
91
105
115
123
X
Contents
IV. Risk Analysis Risk analysis: from prospect to exploration portfolio and back . . . . . . . . . . . . . . . . . . . . . . . G.M. Kaufman
135
Risk analysis and full-cycle probabilistic modelling of prospects: a prototype system developed for the Norwegian shelf . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . J.H. Snow, A.G. Dor6 and D.W. Dorn-Lopez
153
Play fairway analysis and risk mapping: an example using the Middle Jurassic Brent Group in the northern North Sea . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . S. Grant, N. Milton and M. Thompson
167
A model-based approach to evaluation of exploration opportunities . . . . . . . . . . . . . . . . . . . B.A. Duff and D. Hall
183
Risk and probability in resource assessment as functions of parameter uncertainty in basin analysis exploration models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . S. Cao, A.E. Abbott and I. Lerche
199
Risk assessment using volumetrics from secondary migration modelling: assessing uncertainties in source rock yields and trapped hydrocarbons . . . . . . . . . . . . . . . . . . . . . . . . . . W. Krokstad and O. Sylta
219
Prospect resource assessment using an integrated system of basin simulation and geological mapping software: examples from the North Sea . . . . . . . . . . . . . . . . . . . . . . . . . . . . B. Dahl and I. Meisingset
237
V, Management of Discovered Resources Enhanced oil recovery ~ the international perspective . . . . . . . . . . . . . . . . . . . . . . . . . . . . J.J.G. Stosur
253
Nessie: a process analysis of a genetic North Sea field life cycle . . . . . . . . . . . . . . . . . . . . . . P. McGaughrin, K. Ashton, N. Fuller, R.G. Heywood, R.W. Holt, B.L. King and S. White
261
Changing perceptions of a gas field during its life cycle: a Frigg field case study . . . . . . . . . . E. Torheim
273
Vl. Hydrocarbon Resources: the Economic Interface Choosing between rocks, hard places and a lot more: the economic interface . . . . . . . . . . . . . H.H. Haldorsen The usefulness of resource analysis in national economic planning ~ Norwegian Shelf . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . J.-E. Kalheim and H. Brekke
291
examples from the 313
Evaluation of undrilled prospects m sensitivity to economic and geological factors . . . . . . . . C. Hermanrud, K. Abrahamsen, J. Vollset, S. Nordahl and C. Jourdan
325
The World Bank's financial support to the petroleum sector in developing countries . . . . . . . . E. Mayorga-Alba and S. Smith
339
References index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
345
Subject index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
349
World oil: reserves, production, politics and prices C.J. Campbell
The assessment of the world's oil endowment is a sensitive subject with political implications and many vested interests. Some published reserve numbers are spurious, and lax definition has led to misconceptions. Study of the world's oil endowment involves the following elements: Cumulative Production (718 Gb; Gb = billion (109) barrels; numbers based on Oil and Gas Journal reserve data, updated for end 1993); Reserves (Reported less "political" u 722 Gb); Total Discovered (1440 Gb); Undiscovered (210 Gb); Remaining (932 Gb); Ultimate (1650 Gb). The rounded numbers in parentheses refer to conventional oil recoverable with today's technology and economics, excluding NGL, heavy oil, small fields etc., as of end 1993. Discovery (<8 Gb/y) and Depletion Rates (2.3%) are critical. Methods for assessing the Ultimate value include: intuitive; delphic; published trend; discovery pattern; and distribution pattern. An "ideal" resource constrained depletion model shows production rising in each country to the midpoint of depletion (half Ultimate) and falling at the then depletion rate, with a few Middle East countries acting as swing producers under alternative scenarios of world demand. It provides a valuable yardstick by which to understand the trends of actual production. It shows that the swing producers now providing almost 27% of the world's supply will pass 30% in 1995-1998, which is likely to herald another oil shock and the onset of a chronic new condition of declining supply as the world passes its midpoint in 1998. Some explanations for the current low prices in the face of these imminent storm clouds are suggested. Norway, through this conference, is invited to take the lead in promoting better disclosure of the world's discovered oil, as an essential first step to credible prediction of what remains. The situation is serious, and the political and economic implications are colossal.
Introduction Daniel Yergin's book, The Prize, is a brilliantly written and researched tour de force that recounts the oil industry's growth from its birth in the middle of the last century to its present key role in providing cheap energy, the lifeblood of the world's economy. It is the history of a period w h e n almost half of the world's oil e n d o w m e n t was produced. N o prizes are h o w e v e r offered for studies of the second half of depletion which is a m u c h less popular subject: indeed it is almost a taboo. The C o n s u m e r Society has been led to believe that there are no problems that engineers and economists cannot solve. It claims a God-given right to burn up energy any way it pleases, and failing to grasp the resource constraints, finds energy conservation as unpalatable as prohibition. It tends to reject warnings o f impending oil shortage as d o o m s d a y talk. N o one, least of all his bankers, wanted to hear that the disgraced tycoon Robert M a x w e l l was falsifying the accounts. The world's oil accounts are in an equally parlous condition, and this volume is an excellent opportunity to put the record straight, to improve the accounting rules and to throw a little light on the subject before it is too late. It picks Up
a theme which was m u c h debated at the time of the oil price shocks of the 1970s but has faded f r o m the public mind as the threat of shortage s e e m e d to recede. This c o m p l a c e n c y is due for a shock as the spectre of pending supply shortfall returns, this time in a m u c h more serious form. The distinguished A m e r i c a n scientist M. King Hubbert rightly observed almost forty years ago that the production of any finite resource starts at zero, rises to one or more peaks and ends at zero. That m u c h is known, oil being undeniably a finite resource. A hot air balloon with a given a m o u n t of fuel can use its resources either to rise quickly to great height before p l u m m e t i n g b a c k to earth or to cover more ground at a lower altitude. It is the same with oil production. The first step is to ask h o w m u c h fuel the balloon has or, to c o m e to the point, how m u c h oil the world has. That prompts the question of what is meant by the term oil. Is it the a m o u n t in the ground? Is it the fraction that can be recovered with present day technology, facilities and e c o n o m i c s ? Should it include what might be recovered by unconventional means in the far future? Should gas be included? W h a t about condensate? W h a t about natural gas liquids? W h a t about oil in gasfields, like Troll? W h a t about the contribution of small and very small fields?
Quantification and Prediction of Petroleum Resources edited by A.G. Dor6 and R. Sinding-Larsen. NPF Special Publication 6, pp. 1-20, Elsevier, Amsterdam. 9 Norwegian Petroleum Society (NPF), 1996.
2
C.J. Campbell
What about oil in extremely hostile environments, such as the Siberian Arctic, or very deep water? What about thousands of infill wells on Ghawar, the world's largest field?
Definition Fig. 1 illustrates a classification of oil from discovered to undiscovered. The complex and important question of definition is well covered by DeSorcy et al. (1993) and Grace et al. (1993), and for the purposes of this global review it is sufficient to define oil as "cheap oil", producible with foreseeable technology at prices not more than about double those of today, say up to $40 a barrel. It is of course not at all that simple since timing, recovery factors and investment are all involved. The tail end depletion of large fields can go on for a very long time and eventually recover a higher percentage of the oil-in-place than is indicated in earlier estimates. Furthermore, infill drilling which may be carried out given the incentive and investment can add reserves as well as raise production. There are, for these reasons, grey areas between what are termed Proved and Probable Reserves, between Probable Re-
serves and Undiscovered oil, and between economic, sub-economic and non-economic oil, but these uncertainties do not mask the general picture of resource constraint which is now clear. There is a beguiling notion that new technology will inevitably extract more and more of the oilin-place, increasing the reserves and progressively delaying the pending shortage. It is important therefore to keep a sense of proportion and realize that the amount of work involved in applying enhanced recovery techniques globally so as to make a real impact would be truly prodigious. Unconventional hydrocarbons will have their own depletion pattern and become more important later in the depletion cycle but they do not qualify as "cheap oil", the substance considered here.
Determining ultimate recovery There are five ways to attack the problem of the world's oil endowment, however defined: the I n t u i t i v e ~ ask an expert for an intelligent guess; the D e l p h i c ~ ask several experts;
Fig. 1. Classification of depletion stages.
World oil: reserves, production, politics and prices
3
Fig. 2. Publishedestimatesof ultimaterecovery. the R e c o r d ~ study the trend or scatter of past estimates; the D i s c o v e r y P a t t e r n ~ plot discovery against wells drilled; see how the curve flattens and points at the Ultimate Recovery; and the D i s t r i b u t i o n P a t t e r n ~ look at distribution models based on what has been found using fractal and other statistical techniques. If the researcher does not have access to experts to ask for their intuitive or delphic views, he can look at the record of their published estimates as given in Fig. 2. The optimist can take the top end of the range and propose 2600-3000 billion barrels (Townes, 1993). The realist may prefer the trend, recognizing that modem estimates have the advantage of vastly improved geophysical and geochemical technology and a much larger database from worldwide drilling. The trend, at least until recently, points to 1600-1800 billion barrels (Campbell, 1991, 1992, 1993). If the experts are distrusted as being too subjective, Discovery and Distribution Patterns can be investigated. Fig. 3 shows the Discovery Pattern of Norway,
a country that evidently has some potential left. The picture is less promising in Denmark, where most of the oil was found by the first 70 exploration wells, with little to show for the second 70 wells (Fig. 4). Fig. 5 is a regional DiscoveryPattern of Latin America, based on confidential Petroconsultant information (Campbell, 1993). Each cusp is a new exploration play. Overall the plot tends to be hyperbolic with the asymptote corresponding to the Ultimate value. Many variants of this approach have been tried, some using statistical techniques. They essentially relate results to effort, normally showing declining returns (Haun, 1975). The information to construct Discovery Patterns for most countries is not in the public domain, and access to the Petroconsultant database would be needed. For reserve and production data there is an annual December issue of the Oil and Gas Journal in which it publishes a world compilation based on official information. It is by all means a valiant effort that is widely reproduced, but it is confusing to find that World Oil publishes a different set of numbers. Another source is the BP Statistical Review of World
4
C.J. Campbell
NORWAY Discovery Pattern 100%
1992
-o
80%
-ca
0~
60%
o~..~
1979
I O 40% /
.r0~
i i
20%
j
0% 0
,
100
200
300
400
500
600
Exploration wells Fig. 3. Discovery pattern of Norway.
DENMARK Discovery Pattern 100%
t
80%
1992
60% ~
0 40%
O9 r 20%
1936 0%
1966
M" 100
50
150
Exploration wells Fig. 4. Discovery pattern of Denmark.
Energy, coming from a reputable oil company, but in the small print it contains a disclaimer saying that it is simply a compilation from other sources, and does not necessarily represent the company's own assessment. Although Petroconsultants data are needed to plot individual field discoveries, information on giant
fields, which hold about 70% of all discovered oil, has been published (Carmalt and St. John, 1986; Roadifer, 1986). It can be used with limitations for distribution studies. Laherrere et al. (1993) have explained the application of fractal techniques to the assessment of oil resources, proposing the use of the parabolic fractal
World oil: reserves, production, politics and prices
5
LATIN AMERICA Discovery P a t t e r n 100%
m "~3
1~ I i
80%
i i I
j,
60% ,,-4
)
1973
O 40%
20%
00/0 6
8
14
Thousands
Exploration wells Fig. 5. Discovery pattern of Latin America.
as an improvement on earlier techniques. They have tested the method by plotting, for example, the populations of the larger towns of the world on a log-log format with size against rank. The resulting parabola gives a world population close to the official number. The parabolic fractal evidently characterizes one of the fundamental laws of distribution, applicable equally to matter in the universe as to oilfields. It takes only three control points to build a parabola. Since the larger oilfields are found first they provide the basis to plot the parabola which defines the distribution of all the fields to be found in the basin down to whatever economic cutoff is chosen. As long ago as 1976, Folinsbee, applying the less satisfactory linear fractal to giant field data, computed that the world's Ultimate Recovery was 1800 billion barrels, which is still a good number. Subtracting the Total Discovered from the Ultimate gives the Undiscovered. The Total Discovered is made up of two ingredients, namely what has been produced, Cumulative Production, and what remains to be produced from discovered fields, Reserves: Cumulative Production + Reserves - Total Discovered
Suspect reserves Published data on what has been produced seem fairly sound, but examination of the reported reserves soon reveals what appear to be anomalies. It is natural to assume that published reserve estimates are based
on proper engineering studies, and yet the apparent anomalies seem to indicate otherwise. It turns out that the determination and declaration of reserves are two different things. The determination is indeed a serious technical business, but the declaration is all too often political. This is illustrated by the initial reports of a recent discovery in Colombia. One of the partners in the venture, an independent company whose shares react strongly, reported a ten billion barrel discovery; the second, whose chairman was desperate for good news, reported three billion; and the third, whose technical people like to keep something in hand to feed their management in lean years, reported one billion. This example is nothing compared with what countries can do. For them oil is revenue, and revenue is political power and prestige: power for the State Oil Company; power for the Minister of Oil; power for the politician; and power versus other governments or OPEC. It is a particularly good form of political power since it is hidden underground where no one can check it: something like the missiles of the nuclear umbrella. One way to identify "political" reserves is to track the historical record. Table 1 lists the Reported Reserves of the main producing countries during the 1980s (Oil and Gas Journal). Up to 1982, the reported reserves evolved in a reasonable way that more or less matched what was known of the main fields, most of which had been found long before. Then in 1983, Iraq announced an 11 billion barrel increase, although only one field, now with only eight wells on
6
C.J. Campbell
Table 1 Suspect reported reserves (Gb)
1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990
Abu Dhabi
Dubai
Iran
Iraq
Kuwait
Saudi Arabia
Venezuela
28 29 31 31 30 31 31 31 92? 92 92
1.4 1.4 1.4 1.3 1.4 1.4 1.4 1.4 4.0? 4 4
58 58 57 55 51 49 48 49 93? 93 93
31 30 30 41? 43 45 44 47 100? 100 100
65 66 65 64 64 90? 90 92 92 92 92
163 165 165 163 166 169 169 167 170 170 258?
18 18 20 22 25 26 26 25 56? 58 59
9 = suspect increase.
it, was reported. Perhaps it was a valid increase, but it was somewhat suspect. Two years later Kuwait, Iraq's neighbour ~ evidently not wanting to be outdone announced a colossal 26 billion barrel increase. There was apparently no corresponding discovery. Kuwait may have questioned the wisdom of its imaginative accounting when Iraq invaded a few years later and set fire, not to the accounts ~ that might have been a good idea ~ but to the wells themselves. This was nothing compared with what followed in 1988, when Abu Dhabi, Dubai, Iran, Iraq, and Venezuela retaliated by announcing staggering increases made even more remarkable for being simultaneous, not only in the Gulf but in Venezuela on the other side of the world. If there was a technical explanation it is truly remarkable that the revision should have applied to all these many reservoirs at exactly the same time. It is noteworthy too that the Neutral Zone, which is owned by Kuwait and Saudi Arabia, announced no such increase m perhaps the two politicians could not agree on a number or had no motive to exaggerate. It is also significant that Saudi Arabia, which had the least need to exaggerate having the largest genuine reserves, did not increase then, although two years later it too felt obliged to join the bandwagon with a huge increase to maintain its authority. Incidentally, population statistics in several OPEC countries are equally suspect because they were also used as a basis for production quota. Then there is the case of Mexico which wanted to borrow money using reserves as collateral. They have now confessed to exaggerating recovery (Los Angeles Times, 1991). Lastly there are the increasing number of countries that keep on reporting the same number year after year, which in 1993 comprised: Abu Dhabi, Bahrain, Dubai, Iran, Iraq, Kuwait, Neutral Zone, Qatar, Ras al Khaimar, Sharjah, Syria, Yemen, Albania, Bulgaria, former Soviet Union, Czechoslovakia, Romania, Serbia, France, Greece, Spain, Brunei, China, Taiwan, Indonesia, Myanmar, Viet Nam, Algeria, An-
gola, Benin, Cameroon, Congo, Eq. Guinea, Gabon, Ghana, Libya, Nigeria, Tunisia, Za'ire, Argentina, Chile, Colombia, Cuba, Guatemala and Peru. There is evidently a political reluctance to report reductions although such are inevitable as production eats into reserves unless exactly matched by new discoveries or reserve revision. Reported ranges also need to be viewed with caution: the low estimate is commonly the best one, with the rest being exponential speculation. As the U.S. Energy Information Administration comments in its 1993 report, it is difficult to determine the precise size of the "political" reserves without access to a detailed database of individual fields, but an indication was published by Ivanhoe and Leckie in 1993. They added up the initial reserves of 41,000 oilfields worldwide, partly from the Petroconsultants database, and came to a total of almost 1400 Gb for the total oil discovered. It compares with the 1700 Gb if the official figures as reported by the Oil and Gas Journal are accepted at face value. The missing 300 seem to be "political" reserves (Laherrere et al., 1994). The issue is not clear-cut, however, because the suspect reserves are in effect an application of higher than previously justified recovery factors on which there is some room for legitimate technical debate. Perhaps some of the excessive reserves claimed by the Middle East countries can indeed be eventually achieved by close infill drilling and in other ways towards the end of the depletion cycle, assuming that the incentive and investment are available. They do not however qualify as "cheap oil" in the sense we use the term. It is useful to record every country's Reported Reserves, adding a probable element: Positive Probable for what is likely but unsure and Negative Probable for what is reported but unlikely (Table 2). In viewing the delphic estimates of Ultimate Recovery, such as published for example by the U.S. Geological Survey (Masters, 1991), it is necessary not only to deduct
Worm oil: reserves, production, politics and prices the political reserves from past reported discovery, but also to discount their indirect influence on estimates of the Undiscovered. Taking these factors into account, a world Ultimate of 1650 Gb as proposed by Campbell (1991, 1992, 1993) still seems acceptable, increasing to about 1800 Gb if gas-related oil (condensate, NGL and oil-legs in gasfields) is included, and depending where the economic cut-off for small and very small fields is placed. The 1650 Gb number is used here to be on sure ground for the purposes of describing the Depletion Model. The details in Table 2 have been revised from earlier published versions (Campbell, 1993) on the basis of new input and 1993 Oil and Gas Journal data. To the end of 1993, the Cumulative Production was 718 Gb, and the non-political Reserves, as here assessed, are 722 Gb giving a total discovered of 1440 Gb. It is slightly higher than the Ivanhoe and Leckie number of 1400 Gb, mainly because of increased reserves now reported in the former Soviet Union, but it remains in the same order of magnitude. With a 1650 Gb Ultimate that means 210 Gb are yet-to-find. Cumulative Production 4- Reserves = Discovered U l t i m a t e - Discovered = Undiscovered
The undiscovered "yet-to-find" The Petroconsultants database suggests that the average annual discovery (including NGL) has been about 8 Gb over the past 10 years and is falling. It is stressed that upward reserve revision is not new discovery, as is sometimes implied, but simply a correction of a previous under-estimate. Failure to backdate revision to the associated discovery gives a misleadingly optimistic impression of the depletion pattern. The Discovery Rate will almost certainly continue to fall because the pool of giants has almost dried up, at any rate outside the Middle East. New prolific provinces are not being found. Evidently most of what remains to be found will come from ever smaller fields in the established oil patches. So far, the world has drilled some 640,000 wildcats to find almost 90% of what is there. (Again, there are difficulties with definition: in the U.S.A. appraisal and step-out wells are commonly classed as exploration wells). It may take as many again or more to pull in the balance. Most of the viable larger finds are likely to be made over the next decade. Drilling for very small fields could continue almost indefinitely, although it is difficult to visualize circumstances in which a huge number of wells are drilled if they yield only rare successes and those small. Much depends on oil price and tax. Exploration is already heavily subsidized for many tax-paying companies by being
taken as a deduction against high marginal tax on very profitable past flush production. So it is hard, taking a general worldwide view, to imagine a greater economic incentive. Governments are unlikely to allow huge windfall profits if prices rise greatly, nor allow excessive exploration write-offs if the results do not justify them. Furthermore, once the flush production from the large fields is finished, high operating costs on the small fields may consume most of the available write-off. Small fields were often found when looking for large fields and are a much less viable exploration objective in their own right. It also depends on who is doing the exploration. Much of the Undiscovered lies in countries where the state has a dominant or exclusive position. Some of them have priorities, like buying guns, above drilling wildcats.
Depleting the remaining "yet-to-produce" If the Undiscovered is somehow brought in, it means that 932 Gb of oil remain yet to produce. (The term Remaining here includes the Undiscovered, and is not to be confused with the tautologous usage of Remaining Reserves: by definition, all reserves are remaining.) U l t i m a t e - Cumulative Production = Remaining In examining patterns of depletion it is instructive to look first at an example of an individual field such as Statfjord, the largest in the North Sea. Production started in 1979 and rose quickly to a plateau in 1986 but will soon begin to fall rapidly to near exhaustion in the early years of the next century (Fig. 6). Efforts to maintain plateau production may be successful, but if so the pending decline will simply be that much steeper. The current depletion rate (percentage of Remaining produced each year) is 19%. Fig. 6 identifies the Midpoint of Depletion, namely when half the Ultimate had been produced, and illustrates a hypothetical Midpoint Depletion curve at a rate of 12%, a technique discussed further below. The plateau is characteristic of an offshore field with the limitations of the platform. Otherwise production normally reaches a peak, unless it is arbitrarily held back for some reason. High depletion rates are efficient under traditional economics, but run counter to the new "sustainable" economics as are now being proposed (Hawken, 1993). A country's production profile is no more than the composite of its fields. The large fields are normally found first. It takes some time to develop them. so that there is commonly about a 20-year lag between the Midpoint of Discovery and the Midpoint of Depletion, which more or less coincides with peak production. Late stage small discoveries make little impact.
Table 2 Distribution of oil. 07/02/94 Summary data sheet, world, reference date 1993, sorted by UlLimate (Gb. billion barrels) Country
Production
Cumulative
Reserves
(kb/d' 1993) A
B
production C
reported D
S. Arabia F.S.U. U.S.A. Iran Iraq Venczucla Kuwait Abu Dhabi Mexico China Libya Nigeria Canada Indonesia U.K. Norway Algeria E ~ptY Brasil India Neutral Zn Oman Qatar Colombia Malasia Argentina Romania Australia Yemen Ecuador Syria Angola Dubai Brunei 'X'rinidad Tunisia Peru Gabon Germany Sharjah Congo
probable (+I-) total E F
Discovered
Undiscovered
Ultimate
Remaining
G
H
I
J
Und. of ult. (%)
Rem. of ult. (%)
Prod. of Disc. in Dep. Dep. disc. (9%) giants (%,) rate (9%) midpoint
Italy Turkey Denmark Yugoslavia Cameroon Viet Nam Papua Hungary Bahrain France Austria Netherlands Albania Philippines Chilc
Pakistan Thailand Bolivia Subtoval Other World Region A
M. East Gulf Eurasia N. America L. America Africa East W. Europe M. Europe (other) Other World F.S.U. = former Soviet Union. "Other" co~nprisescountries: (a) with <51)0 G b ult.; (b) not in production; (c) anomalous production due to politics; and unforeseen discoveries and revisions anywhere. ' =
former Yugoslavia.
10
C.J. Campbell
Fig. 7. Production plot of Germany.
Fig. 7 is a plot of Germany's production. It is a very mature country which has been thoroughly explored. Even the optimists must concede that it has found most of its oil. Future production is extrapolated at the current Depletion Rate, which is not very different from the Midpoint Depletion Rate of
1973. The United States (Fig. 8) is another mature country that is beginning to show the same pattern, although the late Alaskan discovery distorts the pattern somewhat (Phipps, 1993). Alaska is at a much less mature stage of exploration than the rest of the country, and for this purpose could be treated as a
Worm oil: reserves, production, politics and prices
11
Fig. 9. Production plot of Norway.
separate entity. Norway (Fig. 9) has still to reach its peak, but afterwards production will fall rapidly. These examples put into question the oft-quoted statements that the world has enough oil to sustain current production for 50 years or some such number (see for example World Oil, 1994). That notion is based simply on a calculation of to-day's produc-
tion versus reserves; namely the traditional Reserve to Production (R/P) ratio. (Actually on the number preferred here it is 33 years rather than 50.) But, apart from that discrepancy, the procedure totally ignores the natural depletion curve and will not happen in fact unless new discovery can continuously replace what is being produced. Today only about one-third
12
C.J. Campbell
of what the world produces is being replaced by new discovery, and the trend is falling. To think that it can sustain current production for 33 years, let alone 50, is sheer folly. R/P ratios become increasingly misleading as resource constraints begin to bite, and should be abandoned in favour of Depletion Rates.
A depletion model The background discussed above provides a basis for modelling how the world's remaining oil should be produced if proper attention to resource constraints were to be applied. In essence, it provides that production in each country rises to the Midpoint of Depletion after which it falls at the then Depletion Rate. The current Depletion Rate is applied to countries past their Midpoint, which in most cases is in fact not greatly different from the Midpoint Rate. With most countries, except the swing producers discussed below, close to Midpoint (Fig. 10), it is now too late to contemplate curbing production prior to Midpoint, desirable as that would have been. The model has much in common with the remarkably prescient predictions of Hubbert (1956). We now look the inflexion, foreseen by Hubbert's bell-shaped curve, full in the face. It is, so-to-speak, an ideal or optimal depletion model, and there is clearly no hope that the real world is ready to accept such a discipline. Even so, it is very important to have such a notional profile
in mind if the implications of the actual short-term production trends are to be understood in depletion terms. Thus, if actual short-term production rises above the Model Curve it simply means that the subsequent decline will be steeper. The model in fact is bound to increasingly reflect the actual production profile, for there is an inevitable cross-over between the Model and Actual curves as early discrepancy is compensated for in a model with a finite total. Table 3 is a spreadsheet of the base case scenario showing the production profile derived from the model. A short term fix will make the future crisis worse; there is nothing new about that. A fundamental inflexion as the rising production of the past gives way to the declining production of the future is imminent, although we are too close to see it clearly without the benefit of a resource based depletion model.
Swing Producers Table 2 lists the distribution of oil by country, and indicates the relationships of some of the parameters. About half of the Remaining oil lies in six Middle East countries which are still at a very early stage of depletion. They accordingly have the ability at least in resource terms to perform a swing role-making up the difference between world demand and what the other countries can supply. As illustrated in Fig. 11, their share of world supply inexorably rises until resource
Fig. 10. Yearsfrom depletion midpoint.
Worm oil: reserves, production, politics and prices
13
Fig. 11. Share of production from swing producers and price.
constraints eventually force them also into retreat. To determine the Swing production it is necessary to consider alternative scenarios of world demand (Figs. 12 and 13). It is likely that most non-swing countries are now producing at close to optimal rates with the highest feasible recovery, whereas there may indeed be scope to increase recovery in the swing countries. It is often claimed that the higher recovery is the answer to the pending supply shortfall, but if most of the potential for higher recovery is in the swing countries, it will simply make their position even stronger in the world market. To rely on the Middle East having more oil, is like saying that there is more water in the ocean. The former Soviet Union is another important element in supply but its role is uncertain. From a resource standpoint it has the capacity to increase production, but if and when it does so, much will probably be absorbed by growing domestic demand, leaving the Middle East with its essential swing role. Incidentally, Ulmishek and Masters (1993) give higher resource numbers for the former Soviet Union than accepted in Table 3.
Low price politics make the future crisis worse The base case scenario is that world demand rises at one percent a year until the Swing Producers supply 30% of the world's needs. By the end of 1993 their share had already risen to almost 27%
from the 1986 low of 16%. The 30% threshold is taken to herald price rises enough to curb further increases, giving a flat production scenario from then until the World Midpoint of Depletion, whereupon production declines at the then depletion rate. The other scenarios considered are: High case m 2% annual increase to World Midpoint; Low case ~ no increase to World Midpoint; Swing case -as the Base case except the plateau lasts not until World Midpoint but until the Swing Producers reach their Midpoint, when they would be producing about 60% of the world's supply ~ not a likely scenario. Accordingly, the next oil price shock under the base case scenario can be expected in 1996 or thereabouts. The scale of the shock depends on a range of political factors that are hard to predict. It will come sooner if demand increases above one percent, as some forecasts predict on the basis of the expanding economy of China and the Far East (International Energy Agency, 1994) or if the Swing Producers fail to find the investment to fulfil their role: a distinct possibility. It is estimated that the OPEC countries will require $50 billion to simply maintain current production to 2000 and most of them, even Saudi Arabia, are already in debt ~ the latter even seeking deferment of debt repayment on defense contracts. It will come later if demand stays flat, or if exports from the former Soviet Union rise rapidly. Spare capacity
14
C.J. Campbell
Table 3 Spreadsheet of the Model Production Profile. World scenario "B", Ref. Date 1993, 07/02/94 Cum. prod Reserves reported probable Discovered Undiscovered Remaining Ultimate Date
Pre-1930 1930 1931 1932 1933 1934 1935 1936 1937 1938 1939 1940 1941 1942 1943 1944 1945 1946 1947 1948 1949 1950 1951 1952 1953 1954 1955 1956 1957 1958 1959 1960 1961 1962 1963 1964 1965 1966 1967 1968 1969 1970 1971 1972 1973 1974 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984
717.94 722.29 999.13 - 276.84 1440.24 209.77 932.06 1650.00
Midpoint amount date years Giants amount % last
Production (kb/d)
(ab/y)
3712 4017 4190 4379 4646 4855 5066 5258 5418 5697 5860 6070 5734 6169 7076 7101 7518 8253 9324 9251 10343 11620 12232 13163 13563 15243 16585 17462 17625 19258 20927 22365 24276 26001 28257 30088 32679 35122 38500 41229 45088 48167 49612 55671 57206 54001 57306 59505 59816 62638 59636 55930 53092 53034 54242
18.867 1.355 1.466 1.529 1.598 1.696 1.772 1.849 1.919 1.978 2.079 2.139 2.216 2.093 2.252 2.583 2.592 2.744 3.012 3.403 3.377 3.775 4.241 4.465 4.804 4.950 5.564 6.054 6.374 6.433 7.029 7.638 8.163 8.861 9.490 10.314 10.982 11.928 12.820 14.053 15.049 16.457 17.581 18.108 20.320 20.880 19.710 20.917 21.719 21.833 22.863 21.767 20.414 19.379 19.357 19.798
Production 1st peak 2nd peak 3-yr trnd Dep. rate current midpoint diff.
825.00 1998 23 898.0 62% 1993
1973 2015 -1% 2.28% 2.66% 0.02
Remaining (Gb)
Dep. rate (%)
Giant fields (Gb)
1631.138 1629.784 1628.317 1626.788 1625.190 1623.494 1621.722 1619.873 1617.954 1615.976 1613.897 1611.758 1609.542 1607.449 1605.198 1602.615 1600.023 1597.279 1594.267 1590.863 1587.487 1583.711 1579.470 1575.006 1570.201 1565.251 1559.687 1553.633 1547.260 1540.827 1533.797 1526.159 1517.996 1509.135 1499.645 1489.331 1478.349 1466.421 1453.601 1439.549 1424.500 1408.043 1390.462 1372.354 1352.034 1331.154 1311.443 1290.527 1268.807 1246.975 1224.112 1202.345 1181.930 1162.551 1143.194 1123.396
0.08 0.09 0.09 0.10 0.10 0.11 0.11 0.12 0.12 0.13 0.13 0.14 0.13 0.14 0.16 0.16 0.17 0.19 0.21 0.21 0.24 0.27 0.28 0.31 0.32 0.36 0.39 0.41 0.42 0.46 0.50 0.53 0.58 0.63 0.69 0.74 0.81 0.87 0.97 1.05 1.16 1.25 1.30 1.48 1.54 1.48 1.59 1.68 1.72 1.83 1.78 1.70 1.64 1.67 1.73
73.706 103.895 1.741 3.558 0.958 0.760 0.866 4.023 1.960 72.012 0.000 6.490 15.800 0.630 0.000 5.651 9.610 2.000 1.029 105.035 2.225 0.000 36.100 0.600 16.500 6.010 17.200 17.060 28.600 35.819 20.550 4.820 29.315 22.900 2.790 36.800 33.044 11.250 21.225 11.150 8.291 11.472 8.138 5.565 7.552 9.365 6.139 33.600 6.550 12.795 37.139 5.800 1.405 2.519 2.700 3.900
Notes
World oil: reserves, production, politics and prices
15
Table 3 (continued) Date
1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022 2023 2024 2025 2026 2027 2028 2029 2030 2031 2032 2033 2034 2035 2036 2037 2038 2039 2040 2041 2042 2043 2044 2045 2046 2047 2048 2049 2050
Production (kb/d)
(Gb/y)
53483 55938 55745 57738 59395 60604 60129 60003 59618 60214 60816 61425 61425 61425 59784 58195 56648 55143 53677 52250 50861 49510 48194 46913 45666 44452 43270 42120 41001 39911 38850 37817 36812 35834 34881 33954 33052 32173 31318 30486 29675 28886 28119 27371 26644 25936 25246 24575 23922 23286 22667 22065 21478 20907 20352 19811 19284 18771 18273 17787 17314 16854 16406 15970 15545 15132
19.521 20.417 20.347 21.074 21.679 22.121 21.947 21.901 21.761 21.978 22.198 22.420 22.420 22.420 21.821 21.241 20.677 20.127 19.592 19.071 18.564 18.071 17.591 17.123 16.668 16.225 15.794 15.374 14.965 14.567 14.180 13.803 13.436 13.079 12.732 12.393 12.064 11.743 11.431 11.127 10.831 10.544 10.263 9.991 9.725 9.466 9.215 8.970 8.732 8.499 8.274 8.054 7.840 7.631 7.428 7.231 7.039 6.852 6.669 6.492 6.320 6.152 5.988 5.829 5.674 5.523
Remaining (Gb) 1103.874 1083.457 1063.110 1042.036 1020.357 998.236 976.289 954.388 932.627 910.649 888.451 866.031 843.611 821.191 799.369 778.128 757.451 737.324 717.732 698.661 680.096 662.025 644.435 627.312 610.644 594.419 578.625 563.251 548.286 533.718 519.538 505.735 492.298 479.219 466.487 454.094 442.030 430.287 418.856 407.729 396.897 386.354 376.090 366.100 356.375 346.908 337.694 328.724 319.992 311.493 303.219 295.166 287.326 279.695 272.267 265.036 257.997 251.145 244.476 237.984 231.664 225.512 219.524 213.695 208.021 202.498
Dep. rate (%) 1.74 1.85 1.88 1.98 2.08 2.17 2.20 2.24 2.28 2.36 2.44 2.52 2.59 2.66 2.66 2.66 2.66 2.66 2.66 2.66 2.66 2.66 2.66 2.66 2.66 2.66 2.66 2.66 2.66 2.66 2.66 2.66 2.66 2.66 2.66 2.66 2.66 2.66 2.66 2.66 2.66 2.66 2.66 2.66 2.66 2.66 2.66 2.66 2.66 2.66 2.66 2.66 2.66 2.66 2.66 2.66 2.66 2.66 2.66 2.66 2.66 2.66 2.66 2.66 2.66 2.66
Giant fields (Gb) 7.501 4.802 0.000 7.520 0.700 1.500 2.600 0.000 0.500 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000
Notes
50% produced
16
C.J. Campbell
Fig. 13. Scenariosof world production 1970-2020. is estimated at no more than one million barrels per day (Economist, 1993). The model does not aim to predict short term production as such, a dubious undertaking given so many imponderable factors, but it provides a resource-
based yardstick against which production trends can be compared. That a supply crisis is looming can hardly be doubted. The world's automobile population is rising faster than its people, already exploding (Fig. 14). Since
World oil: reserves, production, politics and prices
17
WORLD POPULATION People & Automobiles 2000
1500
~0 1000
500
1950
1960
1970
1980
1990
2000
2010
2020
Date
+
People
.. .....Automobiles
1950 = I00 After W . H . O . , U . N . E . P .
Fig. 14. World population of people and automobiles.
both feed directly or indirectly on cheap oil, the depletion model carries a warning message that should not be lightly ignored. Given such heavy clouds on the not-too-distant horizon it is surprising that oil prices are so low ~ at any rate they were at the time of writing (the situation could change overnight). There are several possible explanations for this seemingly anomalous situation: (1) Few people understand or believe the resource constraint. The economics of producing oil are different from those of finding it, and very different from those of sustainability. Some economists treat reserves as if they are a costly inventory to be held as low as possible, and do not realise that high prices do not necessarily find more oil even if stimulating exploration. Traditional economic theory discounts the future and depreciates the past, and is not applicable to the depletion of a finite resource. The unreasonably optimistic statements of some explorers are not helpful either, nor are the blandly reassuring statements by the oil companies. Many of the professional international explorers who do have a realistic understanding of the resource constraints are not free to speak. (2) The price of oil is now largely set by the futures market which is a short term speculative market with no strategic role or objective, and the transactions are paper transactions. The reentry of Iraqi production on the ending of the embargo overhangs the market in the speculators' minds.
(3) Saudi Arabia, or more exactly the House of Saud, which already effectively controls the price of oil, may feel that loss of revenue is a price worth paying if it damages its political adversaries more than it harms itself. It may also remember only too clearly the ravages of the 1985 "glut" and, perhaps not fully appreciating the global resource situation from its very privileged standpoint, wants to tread carefully to avoid a repetition, unlikely as that is. According to Aburish (1994), its ruling family has a major stake in the West with more to gain from low oil price than the country as such. (4) The United States has evidently maintained a grip on Saudi Arabia having a long commitment to its present government. The quid pro quo for defense has been low oil price, critical to the world's largest importer. The ending of the Iraq embargo is another critical issue in holding prices down. The agreement that is implicit in its ending may give the Swing Producers more ground to cooperate than at present, which contrary to what the futures market thinks may herald higher prices. But meanwhile the producers are pumping at maximum levels to secure the highest possible benchmark before negotiating cut-backs to accommodate Iraq. Schuler (1991) comments on the short-sightedness of a Western oil policy aimed at holding down prices, tacitly confirming that such a policy exists. It has even been suggested that the Gulf War itself may have been directly or indirectly contrived by the West
18
C.J. Campbell
Fig. 15. Gasoline prices. to secure control of oil which however far-fetched may contain a germ of truth (see Aburish, 1994). But on the other hand such a conspiracy theory implies an extraordinarily successful implementation, which is perhaps implausible. The new U.S. administration appears to be more concerned with domestic issues, not currently even having an ambassador in Riyadh, yet prices fell even further. So it is necessary to search for alternative explanations. Perhaps the Gulf war was simply the consequence of over-arming Iraq as a counter to the perceived threat of Iran. Saudi Arabia has become accustomed to living exclusively on oil revenues, but demands on such revenues have grown over the years m especially with the perceived need for new armaments. Currently the country is spending $45 billion a year, yet earning no more than about $20 billion from oil revenues (Hoagland, 1994). With this desperate imbalance it may not feel able to cut production to raise the price. It is like asking a starving farmer to practice crop rotation: he has not the time. The world's economy is already fragile and can ill-tolerate an increase in the price of oil, such an important ingredient to economic growth. But it walks on increasingly thin ice if it prolongs low oil prices until the resource constraints begin to bite in earnest. It takes ten years to build a nuclear power station and longer to adjust to a low energy, more sustainable life style. Surely, it would be better to begin to pay a price for oil that comes closer to its replacement or
substitution cost instead burning up the "capital" of past exploration which cannot be replaced. It is like speeding downhill with no insurance. To speak of an impending oil shock is an oversimplification. The oil shocks of the 1970s were the consequence of political events in the Middle East, and were short-lived. Next time, it will be not so much a shock but the onset of a permanent chronic condition, where the consumers will have to curb demand, whatever temporary shocks may occur. Clearly higher taxes on fuel should be applied to soften the impact of unfolding shortfall. Nowhere is this more urgent than in the United States, which is consuming an excessive share of the world's oil (Fig. 15), now importing 50% of its needs. Ironically, Europe which is a less profligate user of energy, although bad enough, may find it harder to reduce demand because there is less slack in the system. The moves in this direction being taken on the pretext of the environment, which is more politically palatable than resource constraint, are greatly to be welcomed. Traffic jams, crowded airports and well stocked supermarkets illustrate how much the world has come to depend on cheap oil. Curbing demand will be neither popular nor easy, although Hawken (1993) offers promising proposals by which consumeristic attitudes may be changed, as does Gorbachev through the Green Cross movement (Greer, 1994).
World oil: reserves, production, politics and prices
A call to arms This conference and this volume of its proceedings have a critical role to play. Before anyone will accept the argument for a changed attitude to energy consumption it is first essential to demonstrate clearly that there is a good reason. That means better definition of reserves and greater clarity on many of the topics already discussed. Depletion Rate should replace Reserve to Production Ratios (R/P) as a better measure of future supply. All of these elements are too often clouded in the public mind. But above all, ways need to be found for technical expertise to penetrate the political smoke screen that surrounds the size of the world's already discovered reserves and the recovery factors upon which they are based. It is impossible to predict what remains without first knowing accurately what has been found so far. It is not so much a geological or engineering challenge - - that expertise is already in place - - but a political challenge to open the files and really see what is there. Norway sets a splendid example with the openness of its Oil Directorate and perhaps she can also take a lead in this challenge which is of critical importance for mankind in the 21st century. It will be a very different world, and grasping the greater hope, it may be a better one, but there is no time to lose in facing up to the adjustment that the end of cheap oil will impose on everyone.
Acknowledgements The valuable contributions of L.F. Ivanhoe and J.H. Laherrere are gratefully acknowledged, as is the permission from Petroconsultants to use certain of its proprietary material.
Note added in proof For updated assessments, see: - C a m p b e l l , C.J., 1996. The status of world oil depletion at the end of 1995. Energy Exploration & Exploitation, 14(1), 63-81. - C a m p b e l l , C.J. and Laherr6re, J.H., 1995. The world's oil supply 1930-2050. Report. 3 vols. Pertroconsultants, Geneva.
References Aburish, S.K., 1994. The Rise, Corruption and Coming Fall of the House of Saud. Bloomsbury, London, 226 pp. Abraham, K.S., 1993. ICEED shapes pursestring attitudes, too. World Oil, June 1993, p. 31 (report on ICEED conference). Barry, R.A., 1993. The Management of International Oil Operations. PennWell Books, Tulsa, Okla. Bee, A.C., 1991. Long-run industry performance in exploration outside North America. Am. Assoc. Pet. Geol., 75(8): 1405 (abstr.).
19 British Petroleum Co., BP Statistical Review of World Energy. Published annually by BP London. Browne, E.J.E, 1991. Upstream oil in the 1990s: the prospects for a new world order. Oxford Energy Seminar, Sept. 1991, The British Petroleum Company, London. Carmalt, S.W. and St. John, B., 1986. Future petroleum provinces of the world. Am. Assoc. Pet. Geol. Mem., 40. Campbell, C.J., 1991. The Golden Century of Oil 1950-2050: The Depletion of a Resource. Kluwer Academic Publishers, Dordrecht, 345 pp. Campbell, C.J., 1992. The depletion of oil. Mar. Pet. Geol., 9: 666671. Campbell, C.J., 1993. The depletion of the world's oil. Rev. Pet. Techn., 383: 5-12. Campbell, C.J., 1994. The Depletion Model: a resource constrained yardstick for production forecasting. Rep., Petroconsultants S.A., Geneva. Cleveland, C.J., 1992. Yield per effort for additions to crude oil reserves in the Lower 48 United States 1946-1989. Am. Assoc. Pet. Geol., 76(7): 948-958. Davies, E, 1994. Oil supply and demand in the 1990s. World Pet. Congr., Topic 16. Demaison, G. and Huizinga, B.J., 1991. Genetic classification of petroleum systems. Am. Assoc. Pet. Geol., 75: 1626-1643. DeSorcy, G.J., Warne, G.A., Ashton, B.R., Campbell, G.R., Collyer, D.R., Drury, J., Lang, R.V., Robertson, W.D., Robinson, J.G. and Tutt, D.W., 1993. Definitions and guidelines for classification of oil and gas reserves. J. Can. Pet. Technol., 32(5): 0-21. Economist, 1993. A shocking speculation about the price of oil. The Economist, Sept. 15, pp. 87-88. The Economist, 1994. Power to the people - - a survey of energy. June 18. Energy Economics, 1993. The 1996 oil shock? Energy Economics, May, 139/17. Energy Information Administration, 1993. International oil and gas exploration and development 1991. U.S. Department of Energy, Washington. Folinsbee, R., 1977. World view from Alph to Zipf. Geol. Soc. Am., 88: 897-907. Fuller, J.G.C., 1993. The oil industry today. Br. Assoc. Adv. Sci. Proc., pre-print. Grace, J.D., Caldwell, R.H. and Heather, D.I., 1993. Comparative reserve definitions: U.S.A., Europe and the former Soviet Union. J. Pet. Technol. Sept., pp. 866-872. Greer, C., 1994. My challenge to the World: an interview with Mikhail Gorbachev. Times-Picayune Parade, N. Orleans, Jan. 4. Halbouty, M.T., 1970. Geology of giant petroleum fields. Am. Assoc. Pet. Geol. Mem., 14. Haun, J.D. (Editor), 1975. Methods of estimating the volume of undiscovered oil and gas resources. Am. Assoc. Pet. Geol. Stud. Geol., 1. Hawken, E, 1993. The Ecology of Commerce - - A Declaration of Sustainability. Harper Business, New York, N.Y., 250 pp. Hoagland, J., 1994. The U.S.-Saudi line is off the hook. Int. Herald Tribune, Jan. 31. Hubbert, M.K., 1956. Nuclear energy and the fossil fuels. Am. Pet. Inst. Drilling and Production Practice. Proc. Spring Meeting, San Antonio, Texas, pp. 7-25 (essential reading) Hubbert, M.K., 1982. Technique of prediction as applied to the production of oil and gas. Natl. Bur. Stand., Spec. Publ., 631: 16-141. International Energy Agency, 1994. World Energy Outlook. Ismail, I.A.H., 1944. Untapped reserves, world demand spur production expansion. Oil Gas J., May 2, pp. 95-102 Ivanhoe, L.F., 1976. Evaluating prospective basins. Reprint of Articles in Oil Gas J. of Dec. 6, 13 and 20, 1976. Ivanhoe, L.F., 1986. Oil discovery index rates and projected discoveries of the free world. In: Oil and Gas Assessment. Am. Assoc. Pet. Geol. Stud. Geol., 21: 159-178.
20
C.J. Campbell
Ivanhoe, L.E, 1988. Future crude oil supply and prices. Oil Gas J., July 25, pp. 111-112. Ivanhoe, L.E and Leckie, G.G., 1983, Global oil, gas fields, sizes tallied, analyzed. Oil Gas J., Feb. 15, pp. 87-91. Klemme, H.D., 1983. Field size distribution related to basin characteristics. Oil Gas J., Dec. 25, pp. 169-176. Klemme, H.D. and Ulmishek, G.F., 1991. Effective petroleum source rocks of the world: stratigraphic, distribution and controlling depositions factors. Am. Assoc. Pet. Geol., 75(12): 1908-1951. Laherrere, J.H., Perrodon, A. and Demaison, G., 1993. A new approach for estimating the world's undiscovered oil potential. Rept. Petroconsultants S.A., Geneva. Laherrere, J.H., Perrodon, A. and Demaison, G., 1994. Published figures and political reserves. World Oil, Jan., p. 33. Los Angeles Times, 1991. Mexico lied about Proven Oil Reserves, report says. Dec. 10. Martin, A.J., 1985. In: T. Niblock and R. Lawless (Editors), Prediction of Strategic Reserves in Prospect for the World Oil Industry. Univ. of Durham, pp. 16-39. Masters, C.D., 1987. Global oil assessments and the search for non-OPEC oil. OPEC Rev., Summer 1987, pp. 153-169. Masters, C.D., 1991. World resources of crude oil and natural gas. Review and Forecast Paper, Topic 25, Proc. World. Pet. Congr., Buenos Aires, 1991, pp. 1-14. Masters, C.D., 1993. U.S. Geological Survey petroleum resource assessment procedures. Am. Assoc. Pet. Geol., 77(3) 452-453 (with other relevant references). Masters, C.D., Attanasi, E.D. and Root, D.H., 1994. In: World petroleum assessment and analysis (eds John Wiley & Sons). Proc. 14th World Petroleum Congress, Stavanger, 2, 529-542. Miller, R.G., 1992. The global oil system: the relationship between oil generation, loss, half-life and the world crude oil resource. Am. Assoc. Pet. Geol., 76(4) 489-500.
C.J. CAMPBELL
Nehring, R., 1979, The outlook for conventional petroleum resources. Paper P-6413, Rand Corp., 21 pp. Oil and Gas Journal. World Production Reports, December each year. Perrodon, A., 1992. Petroleum systems, models and applications. J. Pet. Geol., 15(3): 319-326 Petroconsultants S.A., 1993. World Production and Reserve Statistics; oil and gas 1992. Petroconsultants, London. Phipps, S.C., 1993. Declining oil giants, significant contributors to U.S. production. Oil Gas J., Oct. 4. Roadifer, R.E., 1986. Size distribution of world's largest oil, tar accumulations. Oil Gas J., Feb. 26, pp. 93-98 Schuler, G.H.M., 1991. A history lesson: oil and munitions are an explosive mix. Oil Gas J., Nov. 18. Solomon, C., 1993. The hunt for oil. Wall Street Journal Aug. 25. Takin, M., 1993. OPEC, Japan and the Middle East. OPEC Bull., 4(2): 17-34 (March-April). Tomatite, T., 1994. World oil perspectives and outlook for supplydemand in Asia-Pacific region. World Petroleum Congr. Townes, H.L., 1993. The hydrocarbon era, world population growth and oil use - - a continuing geological challenge. Am. Assoc. Pet. Geol., 77(5): 723-730. Ulmishek, G.F., Charpentier, R.R. and Barton, C.C., 1993. The global oil system: the relationship between oil generation, loss, half-life and the world crude oil resource: discussion. Am. Assoc. Pet. Geol., 77(5): 896-899. Ulmishek, G.E and Masters, C.D., 1993. Oil, gas resources estimated in the former Soviet Union. Oil Gas J., Dec. 13, pp. 5962. World Oil, 1994. World not running out of oil. March issue, p. 9 Yergin, D., 1991. The Prize: the Epic Quest for Oil, Money and Power. Simon and Schuster, New York, 877 pp.
c/o Petroconsultants S.A., P.O. Box 152, 1258 Perly, Geneva, Switzerland
21
Gas in the 21st century: a world-wide perspective I. Lerche
Current availability of gas and current distribution of gas are examined in relation to potential gas discoveries, their likely world-wide distributions, and the present-day markets for gas. Current demands for gas in relation to future likely demands are also evaluated in respect of traditional gas uses. Strategies for evaluating likely new disciplines for gas usage are related to government policies and national security interests. The economic concerns of well-head price, pipelines, and liquification are considered in relation to evolving technologies for extracting known reserves, predicting the discoverability of unknown fields, and setting trends for exploration beyond the year 2000. The relation of gas to oil is also considered. The up-shot of all of these factors, evaluated in an integrated manner, would seem to imply that, through about the first third of the next century, adequate proven and potential supplies of gas will be available; by the middle of next century there are likely to be local to regional shortages of both reserves of gas as well as supplies of gas in relation to domestic and industrial consumption. By the last third of next century it would appear that gas will not be a dominant energy source in relation to oil, solar, nuclear fusion or other alternatives. Continued industrialization of nations will shorten these lifetime estimates as well as raise significantly the price of gas on a competitive market demand basis.
Introduction As a clean-burning fuel, relative to coal or oil, gas has a prominent place in providing both industrial and residential heat, either directly or by conversion to other transportable forms of energy such as steam and electricity. Perhaps of concern in this theater alone are the questions of currently available reserves, 'their locations and their distribution, and the rates of consumption. On a broader scale, there are the interests of determining the potential reserves of gas likely to be available in the future, the locations of such reserves, and the likely future demands for the reserves. In addition, the concerns of individual nations' interests play political and strategic roles in constraining or accelerating development and transportation of gas. This review is concerned with several dominant components to the perception of gas usage in the 21st century: Current reserves, their distribution and consumption; Potential reserves and their likely distribution; Prices, costs and pipelines; Current and future uses; Political, social and economic factors. A n o t e on units. Gas is usually measured in units of cubic feet at STP in North America. However, not all countries follow the same standards. Below are
given approximate conversion factors and definitions of units: 1 Mcf -- 1 thousand cubic feet; 1 MMcf = 106 cf; 1 Bcf = 1 billion cubic feet (_= 10 9 cf); 1 Tcf = 1 trillion cubic feet (= 1012 cf); 1 Tcf ~ 26 million tonnes of oil equivalent (mtoe or MMtoe); 1 Bcm = 1 billion cubic meters (= 10 9 m 3 ) ; 1 Tcm = 1 trillion cubic meters (_= 1012 m 3 ) ; 40 Tcf ~ 1 Tcm; 1 Tcm ~ 720 mtoe; Burnability equivalence: 1 Tcf = 1.05 Quad BTU
Current reserves, their distribution and consumption The proved gas reserves of the world increased from about 70 Tcm in 1978 to about 110 Tcm at the end of 1989 at a roughly linear rate with time of 4 Tcm/yr, suggesting that current world proved reserves stand at about 130 Tcm (~ 3500 Tcf). Current world consumption of gas is about 2.5 Tcm/yr for a worldwide average lifetime of about 60 years to depletion of proved reserves. Detailed (and precise!) figures for each nation are difficult to come by because: (i) there is often reluctance by nations to be specific about their economic assets (and occasionally nations just plain do not
Quantification and Prediction of Petroleum Resources edited by A.G. Dor6 and R. Sinding-Larsen. NPF Special Publication 6, pp. 21-41, Elsevier, Amsterdam. 9 Norwegian Petroleum Society (NPF), 1996.
22
L Lerche
know how much gas they have); (ii) there is a very difficult relationship between price of gas and availability ~ a field which might be productive at a price of, say, $3/Mcf may be unproductive at $1.50/Mcf and so may be declared sub-commercial; (iii) presumably, the more advanced the technology, the greater the ability to assess more accurately the proven reserves, so that the evaluation of reserves in less developed nations is not a precise indicator of the actual reserves present (and may be either an over-estimate or an underestimate depending on how the estimate is made); (iv) there are often "political" adjustments made to released reserve figures, which may have nothing at all to do with actual reserves, and are presented to influence a geo-political point. Accordingly, the determination of current reserve pictures and current usage pictures must be viewed somewhat circumspectly but must, at the least, be viewed. The current world picture of proven reserve estimates needs to be more finely divided than the lumped estimate given above. Of interest are the questions of where the reserves are located, what is the current reserve to production ratio (lifetime) for each locale, where are the reserves being used, what is the growth in consumption, and where are the main transport movements of gas, so that one can
determine who depends on whom for supplies, and so who must be politically, socially and/or economically swayed by the give-and-take of the market place.
Spatial distribution of current reserves The data determining reserves and production figures for each country or group of countries have been taken from a variety of sources, not all of which are consistent with each other nor, occasionally, are they internally consistent. When discordances are present, averages were used to assess values. Drawn on Fig. 1 are assessments of current proved reserves for different blocs. Perhaps the most obvious point is that the Commonwealth of Independent States (CIS) and the mid-East together have about 90 Tcm out of a world estimate of 130 Tcm i.e. about 2/3 of current proved reserves (Fig. 1). By way of contrast, the combined Americas appear to have about 8 Tcm, just over 6% of the world proved reserves. It is certain that the proved reserves picture must be continually updated, because in 1970 the total world production was around 800 mboe/yr (for an estimated reserve lifetime then of 40 years), whereas at the present production of 2000 mboe/yr the world lifetime is estimated at 60 years. Clearly the on-going
Fig. 1. Rough estimates of current proven reserves around the world, with a total of about 130 Tcm.
Gas in the 21st century: a worM-wide perspective
23
Fig. 2. Current world production estimates of gas by region. The total is 2000 mtoe/yr.
picture is one of greater amounts of proved reserves being added over the last two decades than are being consumed. The potential discoverable and producible gas reserves are then the factors determining the limiting lifetime.
Spatial distribution of production Estimates of production figures from different sources seem to show larger relative fluctuations than do proved reserve figures. Presumably, then, what should be looked at are the relative magnitudes of production estimates. Drawn in Fig. 2 are the average estimates of current production (in mtoe/yr) taken from the sources. The dominant factor from Fig. 2 is that the CIS and North America account roughly equally for twothirds of the production, while Latin America, the mid-East, Western Europe and Asia account roughly equally for 25% of world production.
Lifetime estimates Irrespective of how or by whom the produced gas is used, estimates can be made of time to depletion
of current proven reserves at current production rates. Some care has to be exercised in interpreting such values because of the high degree of variation in both production rates and the reserve estimates, plus the fact that the near doubling of the world's proven gas reserves (from 70 Tcm in 1970 to 130 Tcm today) suggests current reserves are still increasing at about 4 Tcm/yr, so that a running average lifetime estimate should be made. Indeed, based on the last twenty years estimates (40 year world lifetime in 1970, 60 year world lifetime in 1989) one could draw the conclusion that the lifetime to depletion is increasing at about 1 yr/yr. But as consumption of gas rises, this rate of increase of lifetime must eventually slow and finally decrease. We take up the problem of future predictions later. For the moment we note that a naive division of current reserves by current production gives lifetime estimates as drawn on Fig. 3. The simple inference from Fig. 3 is that North America and Western Europe will not be doing too well by the first decade of the 21st century while, by mid- to late century, the dominant players are likely to be the CIS and the mid-East, based solely on estimates of current reserves and current production.
24
L Lerche
Fig. 3. Dividing current proven reserves by current production rates gives an estimate of years of reserves left. For the world as a whole the estimate is 60 years.
Spatial distribution of consumption Fig. 4 shows the main consumption areas for gas. It is, perhaps, not surprising that the colder Northern Hemisphere nations (CIS, Eastern Europe, Western Europe and North America) consume about 80% of the world's production, with the warmer parts of the globe using correspondingly less (about 20%). Note that the 350 million people of Western Europe are considerably more frugal (by a factor of 2) in their consumption of gas than the 280 million people of the North American continent; while the cold continental climate of the CIS and eastern Europe permits the 250 million people there to eke out an existence only by expending considerably more gas (~36%) to stay marginally warm.
Growth in consumption In the last two decades there have been significant shifts around the world in consumption of gas. Europe and North America have decreased their combined consumption by about 50%, while the CIS and Eastern Europe have increased by 160%, mid-East consumption has doubled as has Asian consumption, while Latin America and Africa have hardly changed
their consumption. Part of the decrease in the North American market has been due to the price of gas. For instance, in 1977 the USA consumed nearly 23 Tcf while in 1986 only 16 Tcf was consumed ~ a nearly 30% reduction. There is a strong tie-in of high oil price in the late 70's, and low oil price in the midto late 80's with the demand for, and price of, gas. This variation in price may be directly tied to the use of gas rather than oil. At the $1.05/Mcf price of gas in the mid-1980's versus the $10/bbl for oil, it was cheaper (BTU by BTU) to buy oil rather than gas - - which, of course, had a feedback effect on the price of gas, lowering the price to be competitive to oil but ending up below finding cost occasionally. Thus the drop in gas usage is directly tied to the increase in oil usage and the drop in oil price ~ and so it goes. While the interrelation of oil usage to gas usage would make for a fascinating discourse such a discussion would take us too far afield in this article.
Main transport movements Because gas is a gas, its transport modes tend to be restricted to pipelines and, when liquified as LNG (Liquid Natural Gas), to container transport by ship. Because of the liquifaction costs, unless there are
Gas in the 21st century: a world-wide perspective
25
Fig. 4. World consumption estimates of natural gas as a percentage of the total of 2000 mboe/yr.
compelling reasons to ship to an island nation devoid of major gas resources (such as Japan), the dominant supply mode is by pipeline. Trade movements of gas then tend to be restricted to continental land masses, so that no gas comes from the Russian platform to the Americas, for example. Fig. 5 provides a description of the main intertrade movements of gas. About 40-50 Bcm per year comes from Canada to the USA by pipeline; about 100 Bcm by pipeline from the CIS to Europe, and is roughly equally split between Western and Eastern Europe; while between 30-40 Bcm of LNG is shipped from Indonesia, Malaysia and southern Asia to Japan 1.
Potential reserves and their likely distribution Major difficulties exist in producing "hard" estimates for potential reserves. Indeed even the phrase "potential reserves" seems to be verging on providing strong emotive reaction. For instance, the Enron 1The BP Statistical Review shows a marked shift from 1989 to 1994 in suppliers of LNG to Japan with the mid-East fraction rising to about 80-90% from a 10-20% estimate five years ago.
Corporation (1993) has written "Resources, on the other hand, not only include proved reserves, but also probable reserves, possible reserves, undiscovered reserves and any substance from which hydrocarbons can be derived (with improved technology or more favorable economics)". Enron go on to argue that it is the lifetime estimate from the ratio of resources to production rate "that should be used for long range planning for the supply of fossil fuels". Perhaps so, but it would seem less than prudent to anticipate that all such resource likelihoods would come to pass as predicted. Truly, a resource is something one has, not that which one hopes to have. But in favor of a future projected resource-based estimate is its guiding nature of what needs to be done for survival. For example, current proven reserves divided by current production in North America give a lifetime estimate to depletion of about 9-14 years depending on how the estimate is made. This lifetime tells one that if nothing is done then currently proven reserves will be depleted in a decade or so. Two options are possible: cut consumption and/or find more reserves, both of which increase the lifetime estimate. Clearly one has to assess the likelihood of increasing reserves beyond those currently proven, and then formulate
26
I. Lerche
Fig. 5. Main trade movements for gas at the present day. Separate pipelines supply Europe from the C.I.S., and the U.S.A. from Canada and Mexico. Ships supply Japan with LNG dominantly from Indonesia and the Malay Archipelago.
and execute strategies to convert the likelihoods into actualities. This sense of future prediction is the only one that makes any logical sense and one which exploration organizations can follow effectively. In the goal to find more reserves, it is the type of reserve conditions and categories anticipated that one assesses in attempts to estimate where effort should be maximally spent to improve a potential reserve base. In good conscience one should not be mixing potential reserves from gas reservoir development in known producing fields (a technical matter) with possible reserves in unknown fields or from speculative sources (such as gas shales); both of the latter are not only exploration problems but also problems on which technical feasibility and cost need to be done. The parallel is to argue that the USA has a virtually unlimited supply of oil because of the massive oil shale deposits, so that the reserve base is enormous. Perhaps so, but the cost of liquid oil hydrocarbons is around $10-20/bbl, whereas break-even price for oil from oil shale is estimated at around $50-100/bbl. And the technical extraction plants are not available in situ on a commercial scale for oil shale. It is certainly true that oil shale is a proven resource that could be tapped, but it is not a reserve that will be tapped in the immediate future. One must distinguish those reserves of gas that are likely to be discoverable and producible within the lifetime determined from the ratio of current proven reserves to production rate from those that will not, or cannot, be produced in such a time frame. Only in this way is it possible to assess the likelihood of gas production and gas usage with time.
Unfortunately: (i) figures available do not always make clear the distinction between the various categories of potential reserves; (ii) there is often a political "enhancement" of estimated reserves which has nothing to do with reality; and (iii) there is often a political statement made by omission, where all possible resources are reported to government authorities as one figure, without explaining that the figures presented do not represent what is available, only what one would like to have available. Wishfulfillment may be far from the wish itself. It cannot be emphasized too strongly that this particular point must be borne in mind when assessing the potential gas reserve categories reported in this section. For example, the Potential Gas Agency of the USA has provided the different potential gas resource estimates given in Table 1, so that a clear breakdown is available of who made what estimate and when, how the estimates differ internally and, most importantly, how they vary by category. For instance, in the Conventional Resources category of Table 1 for both the Lower 48 States and Alaska, the Undiscovered Fields are typically estimated as providing between 2-3 times as much gas as will enhanced recovery of the Existing Fields category, where copious data are available. But one should note that the estimates of potential reserves of the Existing Fields vary by a factor of two themselves. Clearly, even when copious, high quality data are available there are still rather large uncertainties in any estimate. In the Unconventional Resources category, Tight Sands, Coal Bed Methane and Shale Gas are considered as providing the dominant reserves. But the
Gas in the 21st century: a world-wide perspective
27
Table 1 Reported potential gas resource estimates (Tcf) (adjusted to 12/31/92) Resource category (area)
DOE 1988
NRA (USGS/MMS) 1989
NES NES NPC NPC reference access + adv. current advanced 1990 1990 1992 1992
ENRON 1993
GRI current 1993
GRI advanced 1993
Potential gas committee 1993
Conventional resources (Lower 48 States) Existing fields a Undiscovered fields
186 353
0 324
108 256
196 327
149 375
168 413
+ +
209 608
209 667
167 493
28 93
28 75
33 **
33 75
30 76
30 84
** **
** **
** **
37 157
** ** ** **
230 50 15 **
383 90 30 **
232 62 37 15
349 98 57 15
+ + + +
156 81 48 0
241 103 127 43
+ 90 + **
**
**
**
**
37
57
**
**
**
57
660 259 798 919
427 ** 324 427
397 295 659 692
631 503 1026 1134
630 383 870 1013
695 576 1100 1271
+ + 1145 1145
817 285 1102 1102
876 514 1390 1390
854 147 750 1001
Conventional resources (Alaska) Existing fields Undiscovered fields
Unconventional resources (Lower 48 States) Tight sands Coalbed methane Shale gas Other
180 48 31 **
Unconventional resources (Alaska) Coalbed methane Total Total Total Total
conventional resources unconventional resources Lower 48 States United States
** -- Not assessed; + = not reported as separate category. a Includes 25 T c f of gas reported as proved reserves prior to 1988.
estimates are widely varying. For example the colunto labeled "GRI Advanced 1993" would suggest about 510 Tcf potential reserves in the Unconventional Resources category for the Lower 48, whereas the Potential Gas Committee reports a combined estimate of only 90 Tcf, a factor of 5 disparity. It is useless to quote three and four figure precision (as done in Table 1) when accuracy is hardly better than factors of 2 to 5 in any and all categories. So what do we infer for the potential gas reserves of the USA from the estimates of Table 1 allowing for the obvious uncertainties? Remember also that the estimates are openly available through Freedom of Information, so that one can peruse not only the figures of Table 1 but also the data on which they were based and the methods of estimation. Thus, some appreciation is to hand of the manner in which the results were arrived at. In other cases in this section, we will not have even this luxury and will have to accept the figures available at face value m a naive position but an inevitable situation. What we have done is not to give yet another definitive figure, rather we have followed the path of assessing a probabilistic interpretation allowing, therefore, for some sway in the values. The probabilistic method is well used in exploration economic evaluations of plays and prospects, and there is no need to repeat the method here (see e.g. Lerche, 1992). We will quote estimates as follows: From Table 1, the "Total United States" category is determined as
nn-5O0 Tcf. which is to be read as there is a 2/3 '"+200 chance that the potential total reserves will be less than 900 Tcf, a 90% chance of less than 1100 Tcf (900+200), and a 90% chance of greater than 400 Tcf (900-500). For each category one can then provide such an assessment based on the information provided in Table 1. Thus: potential reserves in Existing Fields (Lower 48) 60 Tcf; in Undiscovered Fields are estimated at 160_, 0._~ (Lower 48) at 3 5 0 + ~ Tcf, and so on. This way of reporting results not only provides a risk-weighted assessment, but also provides an idea of the volatility of a result and its sensitivity to the data and method on which the estimate is based. A combined Existing and Undiscovered Fields estimate then yields the proba~nn-14~ Tcf, (20+~0 Tcm). At a usbilistic assessment .~,.,.+260 age rate of 20 Tcf/year, the Lower 48 States probabilistic lifetime against the combined Fields assessment is 25+~ 3 years, and against the "Total United States" the -25 probabilistic assessment is 45+10 years. A worst case scenario would suggest about 18 years to ultimate depletion, a best case scenario would suggest 70 years. It would seem less than prudent to assume that the best case scenario will prevail. But, either way, somewhere around the middle third of the 21st century is about the best that can be hoped for from USA reserves versus usage rates in the region of 20 Tcf/yr. In the case of mainland China, detailed data are not as abundantly available as in the case of North America. Further, data available directly from Chi-
L Lerche
28
Fig. 6. Main on-shore sedimentary basins in China.
nese sources are different than data supplied to, and obtained from, western companies who are actively exploring in China. Fig. 6 provides a sketch of the main sedimentary basins in China with current (early 1994) concession blocks. Fig. 7 provides a map of China divided into several sections with superposed values on each section. Fig. 7 gives the minimum and maximum estimates of expected total gas reserves (in units of Tcm), while the numbers below the range estimate are the expected gas reserves from coal. The data used on the map of Fig. 7 were taken directly from available Chinese sources. The most optimistic estimates would suggest around 23 Tcm of total discoverable gas reserves, while the minimum estimates would indicate 17 Tcm of discoverable gas, with about 10-12 Tcm arising from coal. Current estimates of proven gas reserves are around 1.3-2 Tcm. As China industrializes more, and the demand for gas increases, presumably the consumption rate will also increase. It is not clear what sort of lifetime estimate will be relevant in China but, if consumption use rises by a factor ten then the lifetime against c u r r e n t proven
reserves would drop to about 10-20 years, while if even half the estimated gas reserves, other than from coal (i.e. about 3-6 Tcm), were to eventuate then the lifetime against a factor ten increase in consumption would be in the range 20-80 years. At a more likely increase in consumption by a factor 5, the expected future lifetime is 40-160 years. It would appear that China will be self-sustaining in gas reserves and production throughout most of the 21st century. In terms of current and potential reserves the two biggest players in the gas game over the next century are likely to be the CIS and the mid-East. In the case of the CIS about 80% of current gas reserves are within the Russian Federation (Fig. 8). The CIS total proven gas reserves are about 60 Tcm. But the Barents Sea, Kara Sea, East Siberian Sea, Sakhalin region, East Siberia basin, Lena-Tunguska Basin, Amur region, the North Caspian Basin are all sedimentary basin areas in which little to zero exploration has been done. Indeed, even in the explored basins, current western technology is already claimed to have increased estimated proven reserves by about 10-20% since the Balkanization of the U.S.S.R.
Gas in the 21st century: a worM-wide perspective
Fig. 7. Estimated minimum and maximum potential gas reserves by region in China. The figures above the lines represent total potential reserve estimates (minimum and maximum). The figures below the line represent potential gas reserves expected from coal sources.
Based solely on area alone, as derived from Fig. 8, one could estimate an increase in potential reserves of around 30-50%, suggesting a minimum potential gas reserve figure of about 80-90 Tcm. If one were to apply the factor of 2-3 to undiscovered reserves in unexplored basins as used in the U.S. estimates (cf. Table 1) and, coincidentally, the same range of uncertainty as estimated by the Chinese for their basins, then numbers of order 120 Tcm to 180 Tcm become rough maximum potential gas reserve estimates for the CIS. One might hazard the estimate 1 ~n+4~ -v_30 Tcm as the ultimate gas reserve for the CIS, implying a lifetime estimate to depletion (at current rates of production) f~f~+50 yrs. It would appear that the CIS of around 1,,,.,_25 will be self-sustaining in gas through most of the next century and possibly, but more speculatively, even into the 22nd century, at current rates of production. The mid-East has current proven gas reserves of around 40 Tcm (assuming that the figu,'es reported represent actuality and are not too colored by political enhancement). Even if all of the last three year's reporting were solely "political reserves", the midEast would still contain proven reserves of about 2835 Tcm. By any standards, this amount is a worldclass figure. The distribution of current gas reserves is extremely uneven in the mid-East, with Iran having about 45%, Abu Dhabi, Qatar and Saudi Arabia having around 15% each, and Iraq being at about
29
10%. Reserves have been found steadily and have increased by nearly a factor two in the last decade. The mid-East alone could supply the world's current needs for around 10-15 years, or its own needs for 350 yrs ! There appears not to be any diminution in the rate of finding of gas reserves in the mid-East so that one might hazard a reserve picture of around 80-120 Tcm at a rough maximum and about 40-80 Tcm at a rough minimum. The mid-East will do quite well over the next century in gas (and oil, of course). African gas reserves are dominated by Algeria (5 Tcm) and Nigeria (4 Tcm), while the Caribbean region (including Venezuela and Trinidad) has current reserves of around 4 Tcm, as do Indonesia and Malaysia. On a world stage over the next century even doubling or tripling the estimates of current reserves will not make these nations major players in the game ~ although on local scales they may indeed be dominant satraps. Mexico and Canada combined have proven reserves of 3-4 Tcm (with Mexico about 80% of the total), so that, again, even a factor 3 increase raises the amount to only 9-12 Tcm. Norway, the UK and the Netherlands currently supply the bulk of the gas to Western Europe, but even here the total estimated maximum reserves are only around 8-10 Tcm. The dominance of the CIS, mid-East, and possibly China, over the next century is not impacted. A sketch of potential reserves (and their ranges) by region is given in Fig. 9. It is possible, perhaps even likely, that each and every estimate of potential reserves is incorrect. Based on the bewildering variety of estimates in Table 1, where copious data are available, it would indeed be a miracle if the values and ranges reported in Fig. 9 did not need to be re-assessed as the years go by. But the point is that the prime positions occupied by the CIS and mid-East are exceptionally difficult to change their proven reserves and potential reserves are just so enormous that even fairly uncertain ranges do not topple the CIS and the mid-East from their positions of dominance. The same cannot be said of smaller reserve-valued regions, where jockeying for positions of dominance is especially sensitive to estimates made ~ some of which are less than appropriate (see Table 1). On the broad-scale of the next century, even if no further increase in consumption were to take place, the world would require 260 Tcm of proven reserves to last until the year 2100. Fig. 8 implies that such an amount is going to be extremely difficult to find unless both the mid-East
30 and CIS proven producible reserves are doubled or tripled. It makes sense to plan for some other way to keep warm by the latter half of the 21 st century.
Prices, costs and pipelines Prices and costs In the estimates above of potential reserves no consideration was given to the cost of producing the reserves, the price of selling the reserves, the transportability of reserves, or the effects of future technology as a determinant of supply. This section considers these points. Perhaps of immediate concern is the question of production cost versus selling cost (well-head price). This ratio is not available for each and every country but the U.S. statistics, presumably being controlled by "spot" price of gas, possibly provide an indication of the relative ratio. Given in Fig. 10 are the ratio of well-head price (per Mcf) to the finding cost (per Mcf) versus time. The straight line is of ratio 2 so that a $1 per Mcf finding cost is translated into a $2 per Mcf well-head price. While it is the case that there have been fairly large fluctuations around the line of constant ratio 2 in Fig. 10 it is a fact that, over the last decade to two decades, the selling price of gas (well-head price) has been approximately twice the finding cost price, no matter where in the world the estimate is made. The general problem of estimating potential econ o m i c reserves can then be cast in this framework (note the appearance of the word "economic"). For instance the Enron Corporation (1993) has remarked "The USGS estimates that there are 5,000 Tcf of gas in place in the greater Green River Basin of Wyoming alone. While only a small fraction of this resource is now considered economically recoverable, new technology may someday permit increased recovery. The USGS also reports enormous volumes of gas hydrate off our shores. This is gas captured in the lattice structure of water under certain conditions of temperature and pressure. At some later date, as with the diffuse Green River gas, these hydrates may become economically recoverable." The problem with this perception, which epitomizes all aspects of future predictions, is that neither is the technology available for the foreseeable future that could help convert this fascinating position to one of reality, nor is there any idea of the cost of such technology relative to the then price for gas maybe it will be cheaper and more beneficial in terms of deliverable BTUs to burn the paper on which the money is printed! Realism and pragmatic attitudes to those efforts which are attainable, and the cost of
L Lerche
attainment, must surely be divorced from platitudes of pious hope, as with the above quotation. So what do we have for future estimates of costs? In terms of fixed 1993 dollars the cost of development of potential gas reserves is given in Fig. 11. At some level, in between High Risk to Super Speculative, one decides that other forms of energy (oil, oil shale, nuclear fission or fusion, solar, ...) are cheaper and no development of the ludicrous options (in fixed dollars) takes place. It would be foolhardy to estimate for any other contingency plan! Effectively, such fiscal considerations limit gas reserves in competition with other forms of energy (would one be prepared to pay $10/Mcf for gas when oil is $15/bbl?). Somewhere, as one drifts over from the categories of conventional gas to unconventional gas (perhaps at coalbed methane, or tight sands, or, more distantly, tight shales or even hydrates) the economics wipes out completely the majority of likely non-conventional sources for gas as dominant contributions for the foreseeable future. It becomes, then, less than responsible to include such "resources" as major components for the 21 st century. It would seem that price alone will limit gas dominantly to the conventional arena with, perhaps, a small addition to allow for unconventional contributions. Viewed from this perspective China becomes less of a major player in the 21 st century and the world supply becomes even more controlled by conventional proven and potential reserves of the CIS and mid-East.
Pipelines The question of deliverability of product to the demand market locales becomes of concem in the next century. Within the North American and Latin American land mass, the pipeline infrastructure exists to supply that continental-scale contiguous land-mass. Likewise the pipeline supply system in westem Europe and the CIS forms an integrated network which, at least from western Siberia (see Fig. 8), can be used to supply the northern Asia and European contiguous land-mass. Presumably a continuation to the mainly undeveloped Eastern Siberia fields, and to the north, is not a great difficulty. But supply networks from the East Siberian Sea or from Sakhalin would, probably, be better developed through China or Japan than through to the distant markets of Europe. It will be fascinating to see how such marketing networks develop over the next half century driven both by supply and demand. While no major infra-structure pipeline exists throughout the whole of China, although the eastern seaboard of China is poised more beneficially in
This Page Intentionally Left Blank
This Page Intentionally Left Blank
Gas in the 21st century: a world-wide perspective
this regard, major finds in the Tarim Basin of, say, 3-8 Tcm would quickly create an internal pipeline network. It would seem that the avowed goal of the Chinese people and leadership to become world-wide competitive would help foster the infra-structure development once a major find at the level of a Tcm or so is made m although this economic perspective has to be modified by nationalistic fervor and, in this respect at least, the Chinese public position remains inscrutable. In the case of the mid-East, with its estimated 1000 Billion bbls of oil reserves, there is an extremely well-developed regional pipeline system for local distribution of oil and for delivery of oil to ships. But if the world is to rely significantly in the mid-21st century on gas from the mid-East, then either a gas pipeline system has to be constructed to Europe, or LNG facilities and ship transport of LNG have to be massively developed further over the next decade or so. Without either (or both) of these developments the mid-East gas will be too far from demand market locales to be of much use.
Current and future uses Approximately 55% of the U.S. market for natural gas is for residential heating, with industry picking up the remaining 45% both in terms of direct heating of industrial plants and for heat supply to particular processes (e.g. firing of converters in the steel industry, liquification of glass in the plate and glass industry, etc.) as well as to steam generation in electrical power plants. In the world as a whole, about 60-80% of the gas supply is currently used for residential heating with the remainder going to industrial usage. Due to conservation efforts and increased efficiency of currently available furnaces, the residential market for gas consumption is likely to remain relatively static over the next few decades. The main likelihood of an increase in demand will come only as conversion to gas heating from coal-heating or woodheating in less developed countries is forced upon consumers by progressively more stringent Clean Air Acts. But such moves seem well into the future in a large number of global theaters, so that there is only a low probability of residential increase overall. In the industrial arena the biggest potential for increasing demand for natural gas would seem likely to come from electricity generation concerns. Advanced gas-firing generating units are now less capital intensive than most alternatives and can be fabricated faster. The operational costs of gas-firing equipment will obviously depend on distance from a pipeline, on need for electricity and so on. But there
35
are pragmatic possibilities for: (i) gas conversion of incomplete nuclear plants or shut-down nuclear plants; (ii) gas conversion of coal or oil-fired plants to cut pollution; (iii) adding efficient gas-fired new power plants as alternatives to nuclear, coal or oilfired plants; (iv) upgrading already existing gas-fired plants to be more efficient. The difficulties with current and future uses of gas as a heating device for electrical and industrial concerns are related to: (i) future price fluctuations; (ii) future customer conviction that a supply of gas will be available when needed; (iii) that no cheaper alternatives with guaranteed supplies of fuel are more steadily and readily available. In some cases the alternatives are extremely limited. For instance, Japan lacks a significant reserve base of any form of natural fuels. The main options for Japan are: (i) external, i.e. buy oil and LNG from other countries; but this option not only ships currency abroad but also makes Japan almost completely beholden to outside interests; (ii) internal, i.e. develop nuclear fission power plants of both the conventional and breeder form so that some measure of internal control of power supply is available. Japan exercises both options but has a declared goal of becoming energy self-sufficient, which means building large numbers of nuclear power plants, and Japan has undertaken this venture. In other nations, the options are highly variable because of large proven reserve bases of coal, oil, wood, etc., as well as of gas. The options in such cases are competitively driven by the base price per unit of energy generation from the different sources. Future uses of gas through the 21st century would then seem to depend rather heavily on three factors: (a) a burgeoning global demand for more power; (b) a desire to provide clean sources of residential heating; (c) regulations throughout the world to ensure a cleaner air than is currently the case. Without all three of these factors in place, presumably humanity will continue to use the cheapest form of fuel available, irrespective of environmental consequences; economic considerations will dominate.
Political, social and economic considerations Oil and gas are the mainstays of industrialized development of the last century and are likely to remain dominant in the framework of modem civilization for some time to come. There has always been a considerable financial risk attached to the exploration for, and discovery of, hydrocarbons in commercial amounts. This risk factor has led, in turn, to a global political and economic impact of countries with proven, producible reserves.
36 The structure of the petroleum industry has been changing the last couple of decades and the change shows no signs of abating, nor perhaps should it. The management of opportunities, created by the uncertain and variable global economic conditions, is a relatively healthy factor, requiting on-going evaluation and re-evaluation of the raison d'etre of a corporation, and of the projects, capital outlay, and rate of return to which the corporation has commitments. Prior to the 1960's the Seven Sisters exercised a benevolent type of despotic control on the world's hydrocarbon exploration and production. The rise of OPEC to world prominence throughout the decade of the 1960s, and its oil embargo against the USA and Japan in 1973, caused a painful realization that the era of benign colonialistic control had ended and that a more equitable distribution and re-alignment of hydrocarbon-related profits was not only required but mandated. Indeed, in the USA throughout the mid-1970s, there was a strong popular belief that, to control one's own destiny and not be beholden to others for raw energy needs in the form of hydrocarbons, an absolute need existed to develop a secure supply of domestic reserves with the attendant rider of conservation and the specter of potential nationalization. This need seems to have faded from memory in that the USA now imports around 50% of its oil needs versus the 30-40% imported at the time of the embargo. But Japan has remembered this lesson well and is actively constructing nuclear power plants to lessen its dependence on foreign energy imports. In the era since the mid-1970's the situation has again changed. Partly as a result of the high price of oil in the seventies of around $25-30 per barrel, active exploration burgeoned with the consequence that hydrocarbons have been found in many nonOPEC countries, so that the control on supply once exercised by OPEC has waned somewhat. The price per barrel, however, is still very much determined by OPEC, given the enormous reserves under OPEC's control. The "flooding" of the world with oil in the mid-1980's drove the price down to the $10 per barrel region, leading to economic devastation for those countries that had previously borrowed heavily to develop their own oil industries. But this very flexibility of price has led to suggestions for developments of alternative fuel and energy sources, to substitutions where possible, and to shifting strategy assessments of nations in relation to each other if they are to survive and even grow in directions they perceive as appropriate and beneficial. From a market perspective, the demands and projected needs of individual nations for gas and oil
L Lerche
are relatively well determined if the nations are to maintain their industrialized societies at the levels currently in operation and with inferred assessments for growth and expansion. The superposement of an emotional, rather than a rational, response to some global event rather quickly leads to a panic crisis atmosphere, impacting the price of hydrocarbons. Presumably such responses will always be with us, so that the fluctuations in oil and gas prices will likely continue. Indeed, arguments can be made for not attempting to set up a steady supply/demand type of market equilibrium. As the fates of nations vary, as their perceptions of the future change, and as goals of individual nations expand and contract, their demands for oil and gas must surely change. So a steady-state equilibrium market is inherently an unstable entity as well as being expensive to maintain in terms of the price and supply supports that would be required. There is the further point that the domestic replacement of proven reserves by exploration success has to be offset against the cost of extraction, and by the cost of buying already proven foreign reserves at a price perhaps lower than the finding and extraction costs. The problem is always one of time: the project time between exploration success and on-line production can be a decade or more, while the market response price is virtually instantaneous. A balance has to be struck between immediate goals, with little cloudy perception of the future, and long range projects, with a higher uncertainty of conditions at the time hydrocarbon production comes on stream. The relative degree of success with long range goal achievement can be enhanced, or diminished, by the constraints of national policy in relation to the degree of dependence that a nation is prepared to accept on foreign oil and gas imports; and by the rate at which such a policy changes. Clearly the impact here is one of the cost of carrying out a long range goal versus the political stability of the policies of the government versus the perception of the population as to the need for such an independence m and their willingness to pay the price. The economic health of a nation or corporation has a lot to do with the ability to explore successfully for, and then produce economically, oil and gas reserves. The capital required is usually borrowed, leading to an increase in debt-to-asset ratio. In turn the ability to borrow capital is tied to the value of a nation's currency: a declining dollar makes it that much more difficult to accept national controls on the ability to market oil and gas ~ suggesting that a free-enterprise system may, in fact, be the ultimate method by which supply and demand are self-determinedly controlled.
Gas in the 21st century: a world-wide perspective
37
Maximum Potential Gas Reserves ( T c m )
CIS Mid-East China Africa Mexico & Canada W. Europe Indonesia & Malaysia USA
0
I
I
10
20
I
I
30 40
I
I
50
60 70
I
I
I
I
I
I
I
I
I
I
I
I
I
I
80 90 100 110 120 130 140 150 160 170 180 190 200 Tcm
Fig. 9. Potential world gas reserves by region with minimum and maximum ranges given.
Well-head Price/Finding Cost vs. Year
~
m
2 -q
m
0 1970
1974
I
I
I
I
1978
1982
1986
1990
Fig. 10. Ratio of well-head price (per Mcf) to finding cost price (per Mcf) over the last two decades. The ratio is around 2, suggesting a long term stable sales cost to finding cost ratio.
It seems difficult to guard against inflation, which influences both the cost and final value of an exploration project, nor does it seem possible to predict the rate of inflation on a continuous basis, despite the many attempts to construct theories which will do so. This uncertainty in the inflation factor has a role to play in assessing the economic worth of exploration but can be allowed for to some extent by using probabilistic estimates of uncertainty. As is not uncommon, recession and/or depression
are often tied to the declining value of a nation's currency. Unemployment rises as do interest rates, so that a greater percentage of a declining currency is needed to buy a barrel of oil or a Tcm of gas. The ability to pay the foreign price demanded again underlies the health of a nation. Default on debt interest and/or nationalization of foreign interests and investments, as measures of self-survival in a nation, are sure signs of financial ill-health. These problems can have a significant role in determining the ability of a nation or a corporation to explore domestically for hydrocarbons with or without foreign investment. The incentives underlying the search for hydrocarbons are roughly three in number: on the one hand there are the social, economic and industrial needs of a nation; at the same time there is the incentive of a corporation to be profitable; and on the other hand there are the personal incentives of individuals in the oil industry and governments, which range from financial wealth to power, and are as diverse as the individuals who are involved. From a geopolitical perspective, the needs, requirements, goals, strategies, and philosophies of nations, and groups of nations, also impact the oil and gas exploration game. Strategies that have been employed have ranged from boycott to austerity and rationing, to physical intervention, to global "flooding" by overproduction. There seems no reason to believe that such strategies will not continue to be used as possible options depending on prevailing local, regional and global conditions. The pervasive presence of OPEC as a world eco-
38
L Lerche U.S.A. Lower 48 States: Reserves Versus Cost, from the Sublime to the Ridiculous Potential Reserves (mcf). 10,0007
?
Gas
Hydrates? 5000?
?
?
? ? ?
?9
.?
"?
? ? .?
? ? ? ? ?
7
.? Super Speculative ?
_
?
9
~
Tight Shales?
2000 1200
Total Estimated
I
I
~;
Speculative~
?
9
High Cost Possible
.?
?
.
?
/////////___/\\
/ /
7
?
?
960
;asZab,eCost" " I / / /
9
" ~ ~
800
,..t//
Produced 1
2
3
4
\\ 5
6
7
8
9
I
I
10
11
Cost ($/Mcf), Present Day Technology Fig. 11. Estimated costs of development of potential gas reserves.
I
Likely Events
Gas 9 Supplies Adequate; eTechnologyAids Finding and Recovery; Costs 9 (fixed $) reasonable (_<$2-3/acf).
J J J
J J
Gas 9 Supplies startto drop; 9 to: Oil, Coal, Nuclear, Solar, Fusion; 9 and C.I.S. retain copious reserves; Costs 9 (fixed $) become high (___$3-4/Mcf).
I
I
2000
] eNew sources of
I 9 used only I energy developed when so close I J (Fusion, Solar ....) J to field no to replace gas other energy J heating; J source competes; J 9 on CIS 9 most of world I and mid-East for I reliance on J j solar, fusion and j J their gas reserves; derived energy 9 for gas sources takes J (fixed $) too J over from all high compared fossil fuels; I to alternative I 9 of Hydrocarbon I s~ I Age.
I
I
I
I I
I I
I I
I
I
I
I
I
I
I
I
I I
I
I
I
I
2030
2050
2070
2100
F
Year Fig. 12. The 21st century from a gas perspective.
nomic power since the 1960's argues for its continued viability into the future. Perhaps OPEC does have internal disagreements between its members, and perhaps individual members occasionally violate their agreed-upon production quotas, but OPEC has wielded major authority for 30 years and, with the mid-East still controlling a dominant fraction of the world's oil and gas supply, there seems little
doubt that OPEC will continue to play a vital role in geopolitics. The dominance of OPEC during the next century is assured. In the Western Hemisphere Western Europe, Japan and, to a lesser extent, North America sit in relatively uneasy compromise positions, beholden to foreign imports and so responsive to external pressures. In the cases of Western Europe and Japan there is a very
Gas in the 21st century: a worM-wide perspective
limited choice of immediate alternatives because their needs far outstrip their proven, potential or even possible reserves. Nuclear energy, as in the case of France and now Japan, has been one such alternative employed in order to maintain a nation's energy. In the case of North America, there is sufficient gas in proven and potential reserves to last through the next 20-30 years, but a compromise position is still taken. At the current world price it is just too expensive to explore domestically on a massive scale. The uneasy compromise here would seem, on the long term, to be between paying out now for foreign oil and gas, and so retaining domestic reserves for some future time when foreign supplies dwindle, and recognizing that the foreign indebtedness of North America is increasing daily. In the end this indebtedness may be reflected in the recycling of foreign petrodollars back into North America in the form of outright purchasing of oil companies and of domestic oil and gas fields, both proven and potential an uneasy compromise indeed. In the Communist bloc radical changes over the last four or five years have led to an increasing interaction with western technological capability for oil and gas finding. The removal of Soviet control on Eastern Europe, the break-up of the Soviet bloc into independent states, and the avowed aim to thrust the CIS economies into a free market type of system, seem to herald the opening-up of Eastern Europe and the ex-Soviet bloc to massive exploration possibilities. But guarantees of the satisfaction of competing national and corporate goals must be worked out, as must the ability to inculcate a strategy for exploration and development that neither starves a nation into economic servitude to outside investment nor, on the other hand, runs helter-skelter out of control to satisfy instantaneous demand for domestic production at the expense of long-range well-being of the nation. The direction that China intends to take regarding oil and gas exploration in a geopolitical sense remains inscrutable. In the Third World nations a serious problem exists: these nations are, in the main, too poor to explore and develop for major reserves on their own and so require massive outside assistance to encourage such development. And yet the national requirements for adequate domestic supplies would seem to preclude significant corporate investment, because the corporations require some form of profitability payable in oil, gas, cash or barter. Without other resources to trade for the costs of exploration, Third World nations are caught in a vicious bind. Clearly some form of long term altruistic assistance is needed to develop the untapped potential of such countries. From the perspective of global geological statistics and technological ability, adequate methods exist to
39 explore for hydrocarbons in sedimentary basins and the capability of new methods has been and continues to be continually tested against their use in finding oil and gas in exploration projects. In addition, most of the sedimentary basins of the world are known, and the types of targets to aim for with low risk are also known. From purely geological and technical capability viewpoints there is no fundamental difficulty. The main problem, as usual, is capital. Also, with the rising nationalistic fervor of individual nations, foreign corporations would do well to remember that investments made do n o t produce an immediate equity position in terms of hydrocarbon ownership. A more appropriate philosophy/strategy might be to consider that technology, capability, and willingness to negotiate, lead to a c c e s s to a fraction of production as a joint agreement decision rather than as a fundamental right. As for the future, crystal balls are notoriously murky and unreliable. Nevertheless, several factors stand out. Unless there is to be a major upheaval of the current industrialized civilization, it is highly likely that hydrocarbons and petroleum products will be major factors for the next few decades. Development of any other energy resource of comparable value requires massive capital investment for a considerable period of time at either governmental level (e.g. the long-term nuclear fusion project) or within the private sector. It is difficult to commit massive continued funding within the private sector without some relatively secure guarantee of a pay-out that recoups investment and produces an ongoing profit. Within the oil industry arena, it seems that both onshore and shallow water (less than 600 ft. deep) plays, with the addition of enhanced recovery techniques, will remain the archetype of conventional strategies, yielding significant oil and gas, but likely not leading to many major "giant" discoveries. More interesting from a frontier exploration viewpoint is the likely major build-up of strategies for Arctic basins and also sub-salt sheet and deep-water plays as in the offshore Gulf of Mexico. The prices to be paid here are the difficulties of the terrains and environments, thereby escalating exploration costs; and in the Arctic basins there is the problem of transport to demand market locales. In the intermediate future, heavy oils, tar sands and oil shales together provide nearly inexhaustible supplies of oil. Advantages are that these intermediate future sources of oil are transportable as solids or semi-solid sludges, and so can be cleanly delivered to needed areas with less likelihood of environmental contamination from spillage. The disadvantage of these forms of oil equivalents is the difficulty of extraction of oil. While not technologically a problem,
40 the extraction is expensive. Current estimates would set the price at around $50-100 per barrel equivalent, making these proven reserves uneconomic at the present time. In the case of gas, the estimates for exploitation of unconventional sources (coalbed methane, tight shales, etc.,) set the cost at around $38/Mcf, making these speculative reserves uneconomic over the next 20-30 years. As an alternative energy source, coal has often been touted as an intermediate solution to energy needs, in terms of providing hydrocarbon liquids and gases by liquifaction, cokificaton and gasification. The current difficulties with such proposals are tied to economic cost, to solid waste product disposal, and to the copious amounts of water needed for gasification, which water would need to be cleaned and recycled, adding to the cost. Atomic energy, in the form of nuclear fission plants, has long been with us. Here the associated problems have to do with emotional, environmental and economic issues versus developing technology. The disposal of potent radioactive waste, with incredibly long half-life, becomes a serious concern; as does the possibility of accidental leakage or partial melt down (Windscale, Three Mile Island, Chernobyl) creating environmental havoc over a continental scale. The long term future of nuclear fission plants would seem to be limited, but replacement in the form of nominally cleaner fusion plants seems to be equally distant. While other forms of energy equivalents are being developed (solar energy, tidal control, windmills, etc.) their efficiencies are still low. Also, collection area or volume is considerable, and long term storage and distribution of the energies involved raise problems of their own whose ramifications, potentials for benefit and for disaster, and technical execution in a cost efficient manner, have yet to be seriously developed. Hydrocarbon exploration is a subject that will persist for some considerable time to come.
Conclusion From a world-wide perspective gas will be a dominant energy source for the first third of the next century; by the middle of the 21st century major depletions of reserves will start to impact the dominance of gas as an on-going readily available source of energy; by the last third of the 21st century Russia and the mid-East are likely to be the only significant gas suppliers left to the rest of the world (Fig. 12). It would appear that planning for, and development of, alternate energy sources should be seriously underway around the world by the mid-21st century at the very latest, else approximately one-third to
I. Lerche
one-half of the world's population will feel the four Horsemen of the Apocalypse breathing coldly and heavily over their shoulders.
Acknowledgments Many individuals and organizations provided data, information and help in connection with this perspective. Tony Gorodi of the Gas Research Institute and Tom Woods of the same organization are, perhaps, foremost amongst those who have provided papers, position statements, resource outlooks, future projections, and so on. Other individuals who deserve recognition for their contributions are: Ashley Abbott, Song Cao, Kenneth Petersen, Tony Dor6, Bob Megill, Joe Warren, Marty Perlmutter, Brett Mudford, Chuck Shearer, Ren6 Thomsen, and my secretary, Donna Black. Organizations that provided information are: Enron Corporation, B P, Potential Gas Agency, C.I.A., U.S. Committee on Natural Gas of the National Petroleum Council, U.K. Department of Energy, Energy Statistics Sourcebook, U.S. Department of the Interior, and Oil and Gas Journal.
References AAPG Explorer, Issues of August 1993 (pp. 12-13), November 1993 (pp. 24-25), December 1993 (pp. 22-23). BP Review of World Gas, August 1992, BP, Britannic House, London. BP Statistical Review of World Energy, 1991, BP, Britannic House, London. Central Intelligence Agency, 1993. Energy in The Newly Independent States of Eurasia. Energy Exploration and Exploitation, 1992, Volume 10, Number 2: Special Issue of Selected Papers from the 10th CERI International Oil and Gas Markets Conference, Calgary. Energy Statistics Source Book, 1992, Pen Well Publications, Houston. Enron Corporation, 1993. The Outlook for Natural Gas, 16 pp. Exploration for Coal, Gas and Oil in Onshore and Offshore China, 1993, A P.R.C. Document (in Chinese). Keegan, W., 1985. Britain Without Oil. Penguin Books, Harmonsdworth, 128 pp. Klemme, H.D. and Ulmishek, G.F., 1991. Effective petroleum source rocks of the world: stratigraphic distribution and controlling depositional factors. AAPG Bull., 75:1809-1851. Lerche, I., 1992. Oil Exploration: Basin Analysis and Economics. Academic Press, San Diego, 178 pp. Minerals Management Service, 1990. Alaska Update: September 1988-January 1990. Publication MMS 90-0012, OCS Information Program. Oil and Gas Journal Energy Databases for 1981 through 1992. Oil and Gas Journal. Porter, E.D., 1992. U.S. Petroleum Supply: History, Prospects and Policy Implications. Research Study 64 of American Petroleum Institute. Potential Gas Agency Report, 1993. A Comparison of Estimates of Ultimately Recoverable Quantities of Natural Gas in the United States, Colorado School of Mines, 27 pp. Richardson, EH. (Chairman), 1992. The Potential for Natural Gas. Committee on Natural Gas of the National Petroleum Council, Report to the Secretary of Energy, USA.
Gas in the 21st century: a world-wide perspective Time Magazine, Issue of 20 Aug. 1990. Why the U.S. Is Vulnerable. U.K. Dept. of Energy Brown Book, 1992. USGS-MMS, 1989. Estimates of Undiscovered Conventional Oil and Gas Resources in the United States - - A Part of the Nation's Energy Environment. Woods, T.J., 1993a. How Ultimate is Ultimate Gas Recovery? Paper presented at AI ChE 1993 Spring National Meeting, Houston, Texas.
I. LERCHE
41 Woods, T.J., 1993b. The Long-Term Trends in the U.S. Gas Supply and Prices: 1993 Edition of the GRI Baseline Projection of U.S. Energy Supply and Demand to 2010. Woods, T.J., 1993c. Looking for the Frontier in the Lower-48. Preprint. World Petroleum Trends, 1989-90. Petroconsultants, Switzerland. World Energy Conference Survey of Energy Resources, Crude Oil and Natural Gas, 1988.
Dept. of Geological Sciences, University of South Carolina, Columbia, SC 29208, USA
This Page Intentionally Left Blank
43
Estimating global oil resources and their duration Richard G. Miller
Oil is continually being created and destroyed. The amount that exists at any instant is the global resource. With certain geological assumptions, this volume can be linked to the flux rate (input or output) and an age function: Half-life • system flux rate = In 2 • system size The geological assumptions concern the long-term stability and equilibrium of the system in question, and a simple model of exponential oil destruction. This equation can be applied either to the entire global oil resource or to the oil in economically viable reservoirs. The relationship is also applicable to some other resource systems. Miller (1992) calculated that economically reservoired oil has a half-life of about 29 million years. The global oil generation rate is poorly constrained but seems to be about 2.7 million barrels/year, which is at least compatible with what is known of oil loss by seepage. Perhaps 0.8 million barrels of oil seep annually from reservoirs, which defines the minimum flux rate. These numbers suggest an ultimate global reserve of recoverable conventional oil of 5 trillion barrels, or more than twice most conventional estimates. If the conventional estimates are correct, then some aspect of the model and of our understanding is clearly flawed. Whatever the volume of undiscovered reserves may be, there are some striking aspects of oil supply and duration that should influence long-term planning in the oil exploration and exploitation industry. For example: - For each recoverable barrel conventionally assumed to remain to be found, the industry has abandoned 10-20 as unrecoverable in known fields. - Even a doubling of the conventional reserves will not increase their duration if consumption rises by 2% annually. - The present oil glut is in sharp contrast to the oil-supply fears of 20 years ago, but the conventional view of a 2000 billion barrel ultimate oil reserve has not changed in that time. If that view is correct, then we have only been finding the predicted remaining oil reserves faster than we expected; the long-term prospects for the industry remain essentially unchanged.
Introduction Twenty years ago there was a general mood of foreboding about an impending exhaustion of oil supplies as demand started to exceed supply. This fear seemed to be borne out by the oil price shocks of 1974 and 1981, although the shocks were prophetically dismissed by some (see e.g. Odell, 1991). Arguably the fundamentals were correctly seen in 1973. Today's reported finding rate of 8 billion b/y is only about one third of the consumption rate (Campbell, 1996). Admittedly the reported global reserves have grown by more than 60% to 1000 billion barrels since 1973 (British Petroleum Company, 1993), but these increases include both real improvements in recovery efficiency and probably a degree of politically induced optimism (Campbell, 1996). Certainly the typical view of the ultimate global conventional oil resource has changed little in the past twenty years. Nevertheless the prediction of continual price rises was wrong; the real price of oil today is as low as in 1973 despite oil consumption being 20% more than
in 1973, having risen by 1.25% p.a. over the past ten years. This paper first considers one component of these issues, namely the amount of conventional oil remaining to be discovered, and the possibility raised by a recent, unusual model that this amount may be much larger than generally thought. The paper then notes some simple but fundamental observations concerning how long this reserve may take to be found and how long it may last, and the resulting implications for the oil industry.
The global oil reserve Planners need a view of what the ultimate, conventional, economic, global oil reserve will be. For exploration company planners this view dictates when they should adopt their preferred end-game strategy, whether that be to maintain conventional oil and gas reserves by exploration, to buy proven reserves, to move into unconventional hydrocarbons, or to leave the game. For political planners this view dictates
Quantification and Prediction of Petroleum Resources edited by A.G. Dor6 and R. Sinding-Larsen. NPF Special Publication 6, pp. 43-56, Elsevier, Amsterdam. 9 Norwegian Petroleum Society (NPF), 1996.
R. G. Miller
44 how short a time they have to coax society into change before society is forcibly pushed. Miller (1992) proposed a new way to estimate the ultimate global oil reserve. The fundamental conclusion reached then was that there does not seem to be enough conventional oil to support observed oil seepage. The amount of oil required to support seepage was unsettlingly high, but refuting the model proved difficult (Ulmishek et al., 1993; Miller, 1993). The little new evidence that has since come to light is neither strongly for nor against the hypothesis. The first part of this paper reviews the model and some new observations.
Past estimates and methodologies In the 1930s, typical estimates of the total global recoverable crude oil (including past production) were around 350 billion barrels. Subsequent estimates rose steadily, to reach and stabilise at around 2 trillion (2.1012) barrels for the past twenty years or more (Fig. 1). About 80% of this 2 trillion barrels has already been discovered; indeed, 30-40% of it has already been burned (Fig. 2). Most estimates of undiscovered remaining recoverable conventional oil are in the range 250-400 billion barrels. Most rigorous estimates of the amount of undiscovered oil fall into one of two varieties. The geological analogue methods assess unexplored basins by analogy with explored ones. Mathematical methods, typified by the rate of result analogues, extrapolate the declining returns of exploration per unit of time, cost or effort. Hubbert (1966), for example, recognised that rates of discovery and production are initially low, rise to a maximum, and then fall away in similar bell-shaped curves. The difference between the cumulative discoveries and production is
4000 T
9 9
3500
Billion Barrels
9 tit 9
25005 2000 9
15004" 1000
t
500.
0 1940
9 n"
9
9 99
9
m 9
~ 1950
1960
1970
1980
Fig. 2. Two estimates of the composition of the global oil reserve. These estimates are no better or worse than typical estimates at the time (after Miller, 1992).
the reserve volume, which Hubbert proposed is predictable once the rate of discovery starts to decrease. The Hubbert model has worked well so far in the well-explored onshore USA. Such methods all assume that some aspect of geology is completely understood. For example, many geological analogues assume that mainland USA is geologically representative of the world, and assume that all US oil has been discovered or can be reasonably predicted. Both assumptions are at least debatable. Although about 80% of the world's exploration wells have been drilled in the USA, unsuspected oil is still being discovered there. Similarly the mathematical extrapolations may correctly predict how much more oil remains to be discovered in familiar plays, but they cannot predict the outcome of novel and untested plays. The effect of the recovery factor in reserves estimation is frequently under-valued. Twenty years ago this factor was around 30% at best for most fields, and the global average was probably under 20%. Now recovery can be 45% for many fields (e.g. Anonymous, 1993), and the global average must consequently have risen significantly. This rise should be reflected by a proportionate rise in the global ultimate reserve estimates in Fig. 1, but curiously it is not apparent at all. The world's 1600 billion barrels of discovered recoverable reserves mean that our industry has probably found, but written off as unrecoverable, another 5000 billion barrels. Compare this to the 250-400 billion barrels of recoverable oil conventionally regarded as yet-to-find. It is safe to predict that global recovery factors will at least double before the game is played out, and that this increase will probably have more effect upon the ultimate global oil reserve than any new discoveries.
1990
Year of estimate Fig. 1. Published estimates of the ultimate global oil reserve. These estimates reached 2000 barrels by 1960 and have generally stayed at that level since (after Miller, 1992).
New estimate and methodology Miller (1992) described a completely different method of estimating global oil reserves which mod-
45
Estimating global oil resources and their duration
els the earth as a single, dynamic oil-generating system. In this section, oil means conventional oil and excludes heavy oil, tar sand and shale oil. Seepage refers to oil that reaches the surface, but "seepage losses" are more loosely extended to include oxidation within reservoirs. Oil reserves are proven and exist in identifiable fields. Resources include reserves, undiscovered deposits and currently uneconomic deposits. Yet-to-find oil is the economically extractable oil which remains to be discovered. The global expelled oil system contains all the expelled oil that existed before human exploitation. At any moment this system has a fixed content and input and output rates. The reservoired oil system is the part of the global expelled oil system which consists of the total original oil-in-place in all fields, known and undiscovered, which would be included in global oil reserves using current criteria. These two isolated systems have fixed sizes which Miller's model can estimate. However, the recoverable portion is not fixed but variable, as it depends upon the variable global recovery factor.
The model Both of the oil systems defined above are dynamic. Oil is continually being generated and expelled from source roqks, and it is continually being destroyed. Destruction occurs partly by burial and thermal cracking, but probably principally by seepage to surface and thence oxidation. Between expulsion and destruction, all the oil resides somewhere within the global expelled oil system (Fig. 3). Any oil which migrates into a commercially viable field also becomes part of the reservoired oil system. Given some simplifying assumptions, the volume of the entire reservoired oil system is mathematically related to the mean half-life of the oil it contains and the flux rate at which oil enters and leaves. The half-life can be estimated from the distribution of generation ages of the world's reservoired oil. The flux rates are very uncertain but they can be constrained. The minimum global volume of reservoired oil can therefore be estimated from the oil's minimum half-life and flux rate. The recoverable portion can then be estimated using global recovery and shrinkage factors. A similar model should apply to gas, but we know neither the generation age nor the ratio of thermogenic to biogenic gas for many large deposits. This model avoids the problems of analogue models, but it requires values for several new parameters. It also makes two major geological assumptions, namely (1) equilibrium of the reservoired oil system
Fig. 3. The suggested global oil system. The global expelled oil system contains all existing oil. Included within this is the reservoired oil system, containing all the oil in economically exploitable reservoirs, discovered or otherwise. The recoverable reserve is some fraction of the reservoired oil system which depends upon the recovery factor (adapted from Miller, 1992).
over the most recent few half-lives, and (2) exponential destruction of oil. These assumptions will be clarified before the parameters are considered.
Equilibrium The model assumes that the global expelled oil system and the reservoired oil system are each in equilibrium, and have been so for perhaps 100 Ma. The model also assumes that the volume of potential traps, i.e. the capacity of the reservoired oil system, has not changed significantly. The global rate of oil generation then equals the rate of oil loss, and the system has held much the same amount of oil for a long time. The global rate of oil generation has probably varied during the Phanerozoic, due to long-term changes in biological evolution, plate tectonics or climate. Source-rock deposition has certainly been erratic on the 50-100 Ma time-scale, although such fluctuations have been relatively minor for the past 150 Ma (Klemme and Ulmishek, 1991). If all other factors were unchanged, oil generation rates would presumably reflect the fluctuating source rock deposition rate. However, oil generation rates probably show a relatively smoothed distribution in time because of the generally long but highly variable time required for maturation. The global seepage rate might
46 also be erratic, particularly at times and places of significant tectonism such as present-day California. There is no obvious reason for the reservoired oil system to be out of equilibrium today. Overall equilibrium over the million-year time-scale is not geologically unreasonable, and is compatible with the age distribution of oil (discussed below). The world can probably generate around 2000 billion barrels of oil every million years. Any significant or long-term imbalance would therefore quickly lead to a global glut, or to complete depletion which would make the coincidence of human society and oil supplies unlikely.
Exponential loss The second assumption is that oil follows a natural decay law. Oil destruction is assumed to be sufficiently random, on a global basis, to be treated statistically. This paper assumes that global oil destruction is exponential with time, analogous to radioactive decay, and that oil therefore has a half-life. This is not to say that all oil-fields and basins have the same half-life; the half-life is only applicable to the entire global oil system. In such a model, 50% of the world's existing oil was created during the last half-life, and almost 90% was created in the last three half-lives (Fig. 4). The oil in different basins probably has significantly different half-lives. Macgregor (1993) has shown that the frequency of seeps, and by implication the cumulative rate of seepage, is strongly dependent upon the type of petroleum basin. Since 66% of the world's known oil reserves lie in one place, the Middle East, the assumption of a world-wide, long-term
R. G. Miller
half-life must be viewed with caution. However, the assumption still fits the observations.
Oil half-life equation With these assumptions, Miller (1992) showed that the volume of oil in the reservoired oil system is related to the system's half-life and the system flux rate by the equation: Half-life x system flux rate -- In 2 x system size where In 2 ( - natural logarithm of 2) -- 0.693. The same equation will apply to the global expelled oil system but with different half-life, flux rate and system size parameters. The minimum amount of oil-in-place in reservoirs is therefore defined by the minimum values of the system flux rate and of the half-life. The flux rate is the mean global reservoir filling or loss rate. These rates are not directly measurable, but they can be constrained by the global oil generation rate, and the rate and distribution of oil seepage. The half-life can be estimated from global reservoir age and size data. These parameters are discussed below. The ensuing calculations and estimates are not yet amenable to a rigorous statistical treatment. The numerous parameters used have rarely if ever been estimated before, and certainly none has a known probability distribution. Estimating a probability distribution would only add a further layer of uncertainty, whilst spuriously suggesting that the answer is now less uncertain than before. Probability distributions will only come with far better data than we presently have. Miller (1992) detailed the complete calculations, so that the effect of different parameter values can be tested, and also suggested practical maximum and minimum values for the various parameters.
Fig. 4. The hypotheticalage structure of an oil systemwith a fixed input rate and an exponential loss rate. 50% of existing oil was created during the last half-life (after Miller, 1992).
Estimating global oil resources and their duration
47
Fig. 5. A model of the global sedimentary mass (after Miller, 1992).
Global oil generation rate Estimating the global annual oil generation rate requires several new mean global parameters for sediment organic geochemistry and burial rates. The selection of these values is described elsewhere (Miller, 1992).
Global source-rock organic carbon characteristics The quantity of organic (biogenic) carbon in existence is about 1022 g, plus or minus 50% (from data in: Welte, 1970; Hunt, 1972; Tissot and Welte, 1978; Javoy et al., 1982). About 15% of this carbon forms graphite in meta-sediments, 85% forms kerogen, and 0.1% lies in the biomass, the sea, soil and all known fossil fuels (Welte, 1970; Degens and Ittekot, 1987). This discussion excludes unconventional methane in gas hydrates, which may contain 1018-102~ g of carbon (Kvenvolden, 1988). As the sedimentary mass weighs around 2.5.1024 g, its mean TOC is around 0.4%, although other estimates can be as low as 0.15% (Miller, 1992). The volume of global source rocks and the mean concentration and chemistry of their kerogen are not directly known. This paper assumes that 5% of the sedimentary mass contains at least 1% TOC and has a mean TOC of 1.6% prior to maturation, therefore containing 20% of all kerogen. This kerogen is assumed to have an average expellable yield of 25% hydrocarbon, which is 50% oil.
Global maturation The statistical distribution of global sediment depth is modelled here as a triangle with its apex (the maxi-
mum depth) at 10 km. Threshold depths are set at 2.5 km for the onset of oil generation, 4 km for oil cracking and gas generation, and 5.5 km for over-maturation (Fig. 5). This model implies that 44% of sediment is immature, 20% lies within the oil window, 16% is in the gas and oil-cracking window, and 20% is overmature. These values correspond to relatively high heat flow; lower global heat-flow leads to greater threshold depths and a smaller potential oil yield. Oil generated from source rocks now eroded or metamorphosed is assumed unlikely to still exist. About 80% of the calculated total oil yield comes from rocks which are now below the thermal cracking threshold. If 90% of this oil escaped cracking, then 350 trillion barrels of oil are estimated to have been generated and expelled from the existing global sedimentary mass. A minimum value of 7 trillion barrels can be estimated by assuming 100% cracking efficiency, but this quantity is less than the known oil resource. The estimated maximum is over 1,300 trillion barrels. Wilson et al. (1973) estimated that around 200 trillion barrels of oil had been generated, based on estimates of global recoverable oil reserves. Kvenvolden and Harbaugh (1983) believed this value to be too high, and only accepted that it could not be less than 7.6 trillion barrels.
Global oil generation and expulsion rate The model assumes that sediment loss by erosion or metamorphism is proportional to the size of each maturity division. For each 100 mass units of new sediment deposited, 56 units cross the Oil Generation Threshold (OGT, Fig. 5), 36 units cross the Gas Generation Threshold (GGT, Fig. 5) and 20 units
48 become overmature. The long-term average global sedimentation rate is assumed to be 1016 g/y, the pre-human erosion rate (Blatt et al., 1980). These assumptions suggest that maturing sediments generate and expel about 2.7 million barrels of oil annually. Over 2 trillion barrels of oil could be emplaced in one million years. If all source rocks could reach maturation and expel all their petroleum, then the maximum generation rate would still be less than 8 million b/y. This could only be exceeded if source facies are richer than suggested.
Global oil loss rate Oil eventually undergoes natural destruction, either by burial and thermal cracking or by seepage and oxidation. Seepage is probably the dominant process (and if not, then even more oil proves to be required to support seepage). Oil seepage includes direct migration of oil from its source kitchen to the surface, as well as loss from leaking reservoirs. At equilibrium, the global seepage rate cannot exceed the global generation rate. Understanding and quantifying global oil seepage is difficult. Many, and perhaps most, small seeps are invisible; the offshore oil seeps are hidden by the sea, and onshore oil seeps do not generally rise above the water-table. These invisible "microseeps" can only be detected with sensitive sampling and analytical techniques. The flow rate of even a large seep is difficult to measure, and the source of seepage (leaking reservoir or active generation and direct migration from the source) is usually speculative. The following observations can be made: (1) Wilson et al. (1973, 1974), extrapolating from published data on seep flow rates, estimated that natural global oil seepage into the sea is between 1.5 and 45 million barrels annually. 1.5 million b/y is used here as a minimum value for all oil seepage world-wide. (2) Most seepage from reservoirs presumably takes place onshore, where 80% of known oil reserves are located. Similarly, most onshore seepage is probably from leaking reservoirs, because onshore basins tend to be static or inverting. Such basins are unable to generate oil and are prone to fracturing. Offshore seepage may be dominated by direct migration to surface. The ratio of total onshore to offshore seepage rates is unknown. (3) Given a 30~ oil generation window and a typical heating rate due to subsidence of I~ a "typical" oil-field with 1 billion barrels in place accumulates at an average rate of 33 b/y (but obviously much faster at the peak of generation). Seepage rates in excess of several hundred b/y over a single
R. G. Miller
drainage area might suggest a leaking reservoir rather than active generation. (4) If offshore seepage is at least 1.5 million b/y (Wilson et al., 1973, 1974), then the global total oil loss rate (including cracking) cannot reasonably be much less than the 2.7 million b/y global oil generation and expulsion rate estimated earlier. A rate of 2.7 million b/y represents just 18,000 barrels per basin per year for the world's 150 oil-productive basins. For comparison, the Coal Oil Point seep off California flows at 33,000 b/y (Kvenvolden and Harbaugh, 1983). If, as noted, there does not seem to be enough oil to support oil seepage, then perhaps the global seepage rate has been over-estimated. Is even 1.5 million b/y an over-estimate? Global seepage data are very limited, but as explorers increasingly search for seeps as indicators of a functional sourcing system, more and more natural seeps are being found worldwide. Clarke and Cleverly (1991) amassed a data-base of several thousand visible (mostly onshore) seeps, but only found flow-rate data for thirty. The median flow rate for these thirty was 300 barrels/year. MacDonald et al. (1993) have recently published an independent estimate of regional offshore seepage. From observations of natural oil slicks they calculate that oil seepage in the Gulf of Mexico totals at least 120,000 b/y. This region holds 2-3% of the known (pre-exploitation) global reserves. Simple extrapolation would suggest seepage of 4 - 6 million barrels annually from the global reservoired oil system. This result is consistent with the model outlined here, although the Gulf is a very leaky basin, and Macgregor's (1993) conclusions on the variable leakiness of different basin types are a warning about relying on such extrapolations.
Oil half-life Miller (1992) estimated the half-life of the reservoired oil system directly from volume and generation-age data for well over half the known global oil reserves. He calculated the age ranges of the youngest 50% of the world's oil, the next youngest 25% and so on. In the model each such fraction represents a half-life. The three most recent half-lives are all very similar, between 25.8 and 31.3 Ma (average 29 Ma), as required by the assumption of time-dependent exponential loss (Fig. 6). The three half-lives represent 87% of the data-base oils. Data for older half-lives are statistically distorted by a cluster of Permian oil-fields in the US and the USSR. D.S. Macgregor (pers. commun., 1993) has accumulated independent data on the generation age and total oil-in-place (rather than recoverable reserves)
49
Estimating global oil resources and their duration
Global recovery and shrinkage factors Miller (1992) assumed a conservative global oil recovery factor of 15%, but it now seems more realistic to use 20%, which changes some of the conclusions reported in that study. The higher value makes seepage even harder to sustain by the known global oil reserve. The global shrinkage factor used here is 20%.
The volume of the reservoired off system
Fig. 6. The past nine half-lives of the reservoired oil system. Reserves totaling 1.15 trillion barrels were used in the calculations. The mean of the first three half-lives (87% of world oil) is 28.6 Ma (after Miller, 1992).
for the world's giant fields, both conventional and unconventional. His total of 6300 billion barrels includes over 1000 billion barrels from the Canadian Athabasca tar sands. The data do not show a smooth exponential decay curve (Fig. 7) but are bimodal, with a Late Cretaceous peak due to the tar sands. Tar sands were specifically excluded by Miller (1992) as non-conventional oil. Allowing for this difference, Macgregor's data at least confirm that 29 Ma is a plausible half-life.
The half-life equation given earlier can now be applied to the reservoired oil system (Fig. 3). The system volume will only include reservoired conventional oil because the half-life value was derived from data restricted to economic, conventional oil reservoirs. The flux rate also applies only to mobile oil. If current estimates of around 2 trillion barrels of ultimately recoverable oil are accepted, then the reservoired oil system contained over 12 trillion barrels of oil in place prior to exploitation. With a 29 Ma half-life, the calculated filling rate (and natural loss rate) would be 0.3 million b/y. A reservoir filling rate of 0.3 million b/y seems initially plausible, with the remaining 2.4 million barrels of annual oil generation migrating directly to surface or forming uneconomic deposits elsewhere in the global expelled oil system. However, 0.3 million b/y seems too low for the global loss rate from oil reservoirs. Even if no oil was lost by thermal cracking, this loss rate from reservoirs is equivalent to only 2000 b/y per productive petroliferous basin, or to just 9 seeps world-wide of the size of the Californian Coal Oil Point seep. Even if global seepage is only 1.5 million b/y, then 80% of all seepage would be due to petroleum generation and migration straight to
Fig. 7. A plot of volume of oil-in-place against generation age for the world's giant fields, including oil sands, from the data of Macgregor (pers. commun.). There is no obvious pattern of steady generation and exponential decay (compare with Fig. 5). It is not clear why the data should be so different from those used in Fig. 7.
50 surface, with 20% coming from leaking reservoirs. Also, if most onshore seepage comes from leaking reservoirs (as suggested earlier), then onshore seepage could be only a fraction of the rate of offshore seepage. This does not accord with most explorers' experience. Not all expelled oil enters reservoirs, and not all seepage comes from reservoirs; some oil migrates directly to the surface after generation and expulsion. The rest enters the global expelled oil system, and may enter the reservoired oil system. Calculating the global volume of reservoired oil requires an estimate of how much oil enters and leaves reservoirs. The input and output must balance, and losses must be divided between seepage and cracking. The rates suggested in Fig. 3 are somewhat arbitrary but fit our present understanding. The global oil generation rate has been set at 2.7 million b/y as estimated earlier. The global seepage has been set slightly above the minimum at 2.0 million b/y, leaving 0.7 million b/y to be lost by thermal cracking. Most cracking losses will occur from small accumulations outside the reservoired oil system, which have low buoyancy and thus less potential to migrate upwards and seep. Of the total seepage, 40% (0.8 million b/y) seems a reasonable minimum for seepage from reservoirs (economic and otherwise). A smaller proportion would result in a lower calculated volume for the reservoired oil system, but there is currently no evidence that direct oil migration from active source kitchens makes up over half the world's oil seepage. Most onshore seepage is ascribed to leaking reservoirs for good geological reasons, such as simple observation or basin inversion, and there is no evidence that onshore seepage is significantly less than offshore seepage. However, the assumption of 40% for this parameter will remain a very weak constraint until we have more data on the global rate of seepage from reservoirs. These constraints mean that 1.5 million b/y enter the global expelled oil system while another 1.2 million b/y migrate directly to the surface. 0.8 Million b/y seep from the global expelled oil system and 0.7 million b/y are cracked. Most of the seepage loss (0.7 million b/y) and a little of the cracking loss (0.1 million b/y) from the global expelled oil system is assigned to the large reservoir system. Consequently, 0.8 million b/y enters the reservoired oil system while 0.7 million b/y remain elsewhere within the global system. With a half-life of 29 Ma, the total oil in place in the reservoired oil system is 33 trillion barrels. With 20% recovery and 20% shrinkage, these numbers imply over 5 trillion barrels of recoverable oil.
R. G. Miller
Discussion of the model This analysis suggests that an ultimate reserve of 2 trillion barrels of conventional recoverable oil cannot support known oil seepage. As Miller (1992) discussed, the discrepancy could be resolved by any of the following scenarios: (1) Global seepage rates are very much lower than the Wilson et al. (1973, 1974) minimum estimate of 1.5 million b/y, and generation rates have been similarly over-estimated by Miller (1992). (2) Seepage from oil reservoirs is only 0.3 million b/y and occurs predominantly onshore; most offshore seepage is due to direct generation and migration to surface and is about 2.4 million b/y, slightly above the minimum estimate of Wilson et al. (op. cit.). (3) Seepage from oil reservoirs significantly exceeds the filling rate and the reservoired oil system is (temporarily) considerably out of equilibrium. (4) The global recoverable oil reserve is much greater than typical estimates made by analogue modelling. (5) The half-life has been over-estimated by a factor of at least 2. Miller (1992) suggested that option 4 above, which then involved a 100% increase in the global recoverable reserve to about 4000 billion barrels, was the most likely solution to the problem. The higher recovery factor now preferred in this paper implies an even larger ultimate recoverable reserve of 5000 billion barrels, and the problem has become even more intractable. A 150% increase in the ultimate reserve is really not plausible, yet all the data are internally consistent. The "lost oil" does not reside in tar sands, which represent either unconventional (super-heavy) oil reservoirs or seeped oil that has not yet been completely oxidised. The half-life and filling rate parameters used in the half-life equation only apply to conventional oil, and so the calculated system volume also only applies to conventional oil. Rather than picking a single answer, it seems better to look for a combination of solutions to the problem. Options (1) and (5) above are still regarded here as highly unlikely, but some combination of the remainder may hold the key. As well as questioning the conventional view of the world's oil resource, we have to start considering whether the oil system is seriously out of balance, and we need better observational data to assess the true seepage rate from oil reservoirs. The "missing" oil, if real, need not be related to active seepage areas. It might lie within unexplored basins, although very rich new basins are now generally thought to have "no places to hide" (U1mishek et al., 1993). Alternatively the missing oil
Estimating global oil resources and their duration
may lie in unforeseen plays in known petroliferous basins. One hypothetical example of a novel play will illustrate this option. The phenomenon of simultaneous reservoir filling and diagenetic cementation has been observed in several North Sea fields (Emery et al., 1993). In these fields parts of the reservoir were diagenetically sealed by quartz cement during oil emplacement in other parts. The exact cause is unclear; the injection of oil must forcibly displace the existing pore water to different pressure-temperature regimes, but it is not evident that this effect by itself could deposit the amount of quartz found. Perhaps a stable pore water column is destabilised and undergoes a major density-driven overturn, which greatly increases the water and solute flux; or perhaps new aqueous fluids are introduced with the hydrocarbons. In any event, if diagenetic quartz was deposited as a result of oil emplacement within a carrier bed with no trap, the quartz might cement and seal the reservoir up-dip of the advancing oil. This might produce a diagenetically sealed and trapped oil pool. Such a trap would be almost impossible to predict and extremely subtle, if visible at all, on seismic traces.
The duration of oil supplies The first section of this paper considered the size of the conventional oil resource. How long will that resource satisfy demand? How long can the exploration industry continue in its present form as the exploration targets run out? What will the future market for professional oil explorers be? Similar questions are genetic to all finite natural resource based industries. There are some surprisingly robust generalities that can be drawn despite the uncertain assumptions, estimates and models that under-pin such predictions. Finite a n d replaceable resources In a general way resources can be classified as either finite or replaceable. Finite resources, for example, include Precambrian banded iron formation and Rembrandt portraits. There will never be any more of these materials than there are today, and if we consume (or destroy) them the supply will run out. "Rocks" and "paintings', however, are replaceable resources that are continually created. They can be consumed indefinitely if the use rate is below the natural replacement rate, and as long as the exploited materials return to their natural cycles. Iron ore, trees and dead organic matter can constantly be replaced to make steel, paper or oil at some sustainable level. Fossil energy resources are all replaceable to some extent, for as long as the sun shines. The classification of oil as a finite resource only reflects the
51
fact that society uses it faster than it is replaced. Oil is being consumed about 8500 times faster than its apparent replacement rate of 2.7 million barrels per year. This rate of global oil generation could theoretically support the manufacture of around 200 disposable polystyrene coffee cups each second indefinitely (1.8 g cups and a 3% yield of ethylene from oil for polystyrene manufacture). A brief review of the available data on resource ages and volumes suggest that coal is being used up about 500,000 times faster than it forms (10,000 fly?); thermogenic gas is being used up perhaps 30,000 times faster than it forms (2 billion cf/y?); uranium of course is not being replenished at all; and, as an example of a non-energy commodity, iron ore is being used up at least 600 times faster than it forms (< 1.1 million t/y). All fossil fuels together are probably replaced at around 20.1012 BTU (thermal) per year, which after conversion to electricity and back to heat could heat just 20 baths per second indefinitely. Global energy reserves Oil is still the principal energy source of choice, followed by coal, then gas, nuclear and hydroelectric energy (Fig. 8). Since 1988 the use of oil has increased, but nuclear power use has grown even faster, while coal use has declined (Fig. 9). Existing reserves of conventional oil are around 1.1012 barrels. The ultimate remaining reserve of conventional oil is generally put at about 1.4.1012 barrels (although this paper argues that it may actually be much greater). Unconventional oil has been estimated at a further 2.9.1012 barrels, for a total of 4.3-1012 barrels. Global oil usage is around 23.109 barrels/year, and grew at 1.3% per year from 1982 to 1992 (Fig. 9). The peak rate of increase was 3%, reached in 1986; economic recession reduced the growth rate to nearly zero during 1991 and 1992. Reported oil discovery rates are around 8 billion b/y (Campbell, 1996). However, reported reserves have been essentially unchanged since 1989, implying a replacement rate of about 23 billion b/y (British Petroleum, 1993). The discrepancy is partly due to reserves growth (including improved production technology) but also to political optimism (Campbell, 1996). Campbell's data suggest that the 250 billion barrels of yet-to-find would all be discovered in 30 years if current discovery rates continued unchanged. For comparison with the oil data, the conventional global gas reserve is put at around 5000 TCF. The estimated ultimate conventional yet-to-find is hard to estimate, and the unconventional resource is frankly speculative but seriously large. Estimates of the global methane held in hydrate (clathrate)
52
R. G. Miller
Fig. 9. The annual changes in the use of various energy resources. In absolute quantities, the growth in gas use has outstripped that of any other fuel since 1987. In percentage terms, however, nuclear power has been the growth energy resource over the past five and ten years. Only coal has shown a significant reduction in use.
form range through three orders of magnitude, from 100,000 to 100,000,000 TCF (Kvenvolden, 1988). Annual global gas usage is around 72 TCE Annual gas consumption growth has averaged almost 4% since 1982, despite dropping to only 2.5% since 1988.
The insidious growth rate One might naively conclude that the world has about 60 years' supply of conventional oil at today's consumption rates, so there will be no serious shortfall in our lifetimes. This prognosis ignores many possible factors that may change the 60-year horizon for oil, ranging from new technology to political interference. The effects of two such factors are to some degree predictable, namely growth in consumption and the phenomenon of deceleration.
Resources rarely last as long as we hope. Suppose there is a million years' supply of something at present rates of consumption. If consumption of this resource increases at 3% per year, it will run out after just 349 years. Global oil use in fact increased by 3% in 1986. At this growth rate a thousand year supply would actually run out after 117 years, and a hundred year supply would run out after 47 years. Small but persistent increases in consumption have a seemingly disproportionate effect upon supplies. The 1400 billion barrels of recoverable conventional oil generally thought to remain represent 60 year's supply at current consumption rates. An increase in oil consumption of 1% p.a. would require around a 30% increase in the reserve in order to last for the same time (Fig. 10). A 2% p.a. growth rate over 60 years would require about double the oil reserve. Oil consumption growth averaged over 2%
Estimating global oil resources and their duration
53
Fig. 10. The duration of oil supplies varies according to the consumption growth rate and the ultimate reserves. This diagram does not take deceleration into account. The black bars show the calculated duration of the remaining conventional ultimate reserve. Stacked on these are bars representing successively larger possible reserves. A 30% increase in reserves is cancelled by an extra 1% per year on the growth rate, and a 100% increase is cancelled by a 2% annual rise in consumption.
during the boom years of 1986-1990. Doubling the ultimate global oil reserves might actually have little effect upon their predicted duration. In trying to foresee the duration of oil supplies, neither the amount of oil nor the rate of its future use are known, but it is clear which number is most important. Of course, unexpected conservation developments may serve to stretch the duration of the oil resource. As deceleration becomes apparent, there will be an increasing financial incentive for individuals, companies and countries to improve their energy efficiency. A Western-style economic affluence is probably supportable at lower levels of energy consumption. For example, Fig. 11 compares 1988 data for the national gross domestic product (GDP) with the energy consumption of various countries. Japan was then the best country at converting energy to dollars, with twice the efficiency of the United Kingdom. By this
measure, Japan was the greenest country in the world. Japan, of course, has strong incentives to be energy efficient, having rather few oil and gas resources. Fig. 11 shows that the USA was remarkably inefficient, a measure of its wealth and energy endowment, while the USSR was almost equally inefficient, reflecting its subsidised fuel costs and low-grade technology.
Deceleration The exhaustion of oil may seem comfortably distant, but society will not use oil without restraint until the day it runs out. Similarly the oil exploration industry will not discover conventional oil uniformly until it is all found, which might only take 30 years if the general view of the global resource is correct. These limits are not cliffs but slopes. There will be
Fig. 11. The efficient use of energy. Countries apart from U.K. are shown in order of decreasing 1988 gross domestic product (GDP). The black bars simply show how many dollars are produced per unit of energy. Japan clearly generates most dollars for its energy consumption.
54 a long slow deceleration, where consumption falls, prices rise, and discoveries generally become smaller and fewer despite increasing efforts. Global ultimate reserves may represent decades of future supply, but society and the oil industry will be forced into changing long before then. Deceleration in the oil exploration business is already evident. For example, reported global oil reserves peaked in 1989, and if political exaggeration is discounted, these reserves are now almost certainly dwindling. The proven finding rate is only 35% of the consumption rate. The world's last great land rush, in the CIS, is now almost over; beyond that, some explorers believe that there are no more unexplored basins with the potential for billion barrel fields. The industry is increasingly returning to the mature basins (of course the US has continually re-visited explored basins, but taking place now is the first concerted review of a mature basin, the North Sea, with promises of a new future of increasing output based on lower costs and better technology. This is only equivalent to draining the oil tank faster and more thoroughly: it does not change the volume of the tank, and it will not postpone the empty tank for ever). Finally, conservation has arrived as a consistent political force, even within the oil industry. Future symptoms of deceleration will probably include the return of continuous price volatility when supplies match demand. When small changes in oil supply cause magnified changes in price, political power holdings will inevitably change within and among the producing countries, companies and workforces. The current relatively free market approach to oil exploration and production may become inappropriate. There will probably be more serious attempts to reduce society's dependence on natural oil, by technological change and alternative fuels like bio-diesel. Efficient exploration and exploitation will increasingly require the best technology and expertise from the multinational companies. However, despite the increasing acreage available, competition for the diminishing choice tracts will be fierce, and specialist exploration capability could become a buyer's market.
R. G. Miller
Some developments seem inevitable. Overcapacity has currently driven prices down to a point where oil is virtually a fixed-margin commodity (i.e. the price of every barrel of oil is little more than the price of the cheapest, essentially the global production cost plus a small fixed profit margin for efficient operators. If an extra barrel is needed, it costs no more than the rest). It is not clear how long such excess capacity will continue. We may have supplies for two generations at likely consumption rates, but only if we will pay much more for them. The Middle East, with 100 year's worth of reserves at current production, must inevitably become a more important controlling player simply because its proportion of global remaining reserves will rise. When demand exceeds supply, host governments will be in increasingly powerful political positions. For a while the market-place might remove the need for a producer's cartel, until the market finally fails to meet the demands of the industrialised world; a cartel might then have to ration oil supplies to increasingly anxious consumers. Oil explorers will find the rules changing. For explorers, the last frontier rush and the first concerted return to the mature basins show how the exploration environment is irreversibly changing. At current discovery rates, the conventional yet-to-find recoverable oil will be found by the year 2023. In practice, of course, the discovery rate will drop long before then. Those who accept this conventional view need a plan, and the visible onset of deceleration suggests that there is not much time to change their company's vision and strategy. Among the options are: Focussing on gas as an oil replacement; undiscovered conventional reserves may not be well constrained or necessarily large, but the resource is certainly enormous. - Returning to the 5000 billion barrels of oil which we will leave behind in reservoirs as unrecoverable. Focussing on the 3000 billion barrels of unconventional oil resources such as tar sands, where the potential resource is even larger. exploration contractors to national governments. Leaving the game. These options carry different risks and rewards. The first three options all carry a bill for research and development, particularly the achievement of a quantum leap in recovery factors. This science and technology will struggle for funding in today's financial climate. Those who believe that there is significantly more conventional oil yet to be found will need a different plan. This oil will not be "conventional" in the sense of lying in conventional plays. It is easy to speculate -
-
- B e c o m i n g
-
Oil exploration and consumption: some futures As consumers, we are probably relatively safe in terms of future oil supplies, although our children may not be. From the global perspective, adequate oil supplies will last for around two generations, barring unforeseen (and generally catastrophic) developments.
55
Estimating global oil resources and their duration
on hypothetical new play types, but they will not be easy to find, or they would have been found already. The required research will again be unpopular in a time of increasing cost pressures.
gentle with his criticism. I thank reviewer Ian Lerche. I am also grateful to Tony Dor6 and the Norwegian Petroleum Society for this opportunity to air these ideas.
Conclusions
References
Oil exploration is a natural resource industry whose resource base of undiscovered fields is visibly declining. Although the size of this remaining resource is uncertain, it is clear that changes are inevitable and imminent. Few of the fundamentals of the oil industry have changed much since the scares of 20 years ago. A sense of complacency and global oil over-supply have translated into price depression and consequent difficulties in justifying and financing exploration. It seems unlikely that over-supply can last more than a few years. This paper has illustrated a range of views concerning the volume of the global ultimate oil resource, and some inescapable facts about its duration. This range leads to a variety of options for the oil exploration industry. Amid this uncertainty there are still some conclusions that seem likely to be valid for some time to come. - The world's reported proven oil reserves have never been greater. This observation is compatible with the unchanged general perception over the past thirty years of an ultimate reserve of 2 trillion barrels. - If the conventional view that 250-400 billion barrels of conventional reserves remain to be found is correct, then many players must soon leave the exploration game. At present discovery rates this oil will have all been found within just 30 years. In reality a long period of deceleration has started, in which exploration success will decline. - If much more than 400 billion barrels of oil remain undiscovered, then exploration success will go to those who understand why we have not found it so far.
-Conservation will do far more to increase the longevity of oil reserves than the discovery of anything except improbably enormous new reserves. - Perhaps explorers need to look harder at producing the 5000 billion barrels presently regarded as unrecoverable. The target is far larger, and we know where it is. This process is already becoming part of our culture. However, it will not change the ultimate size of the resource upon which we depend.
Acknowledgements I thank BP for permission to publish this paper, and I emphasize that the views expressed here are entirely my own and not those of BP. Duncan Macgregor has been generous with his thoughts and data and
Anonymous, 1993. Engineers and geologists redefine economic limits of recovery. J. Pet. Technol., 45(10): 928-929. Blatt, H., Middleton, G. and Murray, R.C., 1980. Origin of Sedimentary Rocks. Prentice-Hall, Englewood Cliffs, N.J., 782 pp. British Petroleum Company, 1993. BP Statistical Review of World Energy. British Petroleum Co. Ltd., 37 pp. Campbell, C.J., 1996. World oil-reserves, production, politics and prices. In: A.G. Dor6 and R. Sinding-Larsen (Editor), Quantification and Prediction of Hydrocarbon Resources. Norwegian Petroleum Society (NPF), Special Publication 6, Elsevier, Amsterdam, pp. 1-20 (this volume). Clarke, R.H. and Cleverly, R.W., 1991. Petroleum seepage and post-accumulation migration. In: W.A. England and A.J. Fleet (Editors), Petroleum Migration. Special Publication Geological Society London No. 59, pp. 265-271. Degens, E.T., and V. Ittekot, 1987. The carbon cycle-tracking the path of organic particles from sea to sediment. In: J. Brooks and A.J. Fleet (Editors), Marine Petroleum Source Rocks. Special Publication Geological Society London No. 26, pp. 121-135. Emery, D., Smalley, P.C. and Oxtoby, N.H., 1993. Synchronous oil migration and cementation in sandstone reservoirs demonstrated by quantitative description of diagenesis. Philos. Trans. R. Soc. London, Ser. A, 344:115-125. Hubbert, M.K., 1966. History of petroleum geology and its bearing upon present and future exploration. Am. Assoc. Pet. Geol. Bull., 50:2504-2518. Hunt, J.M., 1972. Distribution of carbon in crust of earth. Am. Assoc. Pet. Geol. Bull., 56: 2273-2277. Javoy, M, Pineau, F. and All~gre, C.J., 1982. Carbon geodynamic cycle. Nature, 300:171-173. Klemme, H.D. and Ulmishek, G.E, 1991. Effective petroleum source rocks of the world: stratigraphic distribution and controlling deposition factors. Am. Assoc. Pet. Geol. Bull., 75: 1809-1851. Kvenvolden, K.A., 1988. Methane hydrate m a major reservoir of carbon in the shallow geosphere? Chem. Geol., 71: 41-51. Kvenvolden, K.A. and Harbaugh, J.W., 1983. Reassessment of the rates at which oil from natural sources enters the marine environment. Mar. Environ. Res., 10: 223-243. MacDonald, I.R., Guinasso, N.L. Jr., Ackleson, S.G., Amos, J.E, Duckworth, R., Sassen, R. and Brooks, J.M., 1993. Natural oil slicks in the Gulf of Mexico are visible from space. J. Geophys. Res., 98-C9: 16351-16364. Macgregor, D.S., 1993. Relationships between seepage, tectonics, and subsurface petroleum reserves. Mar. Pet. Geol., 10:606-619. Miller, R.G., 1992. The global oil system: the relationship between oil generation, loss, half-life, and the world crude oil resource. Am. Assoc. Pet. Geol. Bull., 76: 489-500. Miller, R.G., 1993. The global oil system: the relationship between oil generation, loss, half-life, and the world crude oil resource: reply. Am. Assoc. Pet. Geol. Bull., 77: 900-902. Odell, P.R., 1991. Global and regional energy supplies: recent fictions and fallacies revisited. Energy Exploration and Exploitation, 9: 237-258. Tissot, B.P. and Welte, D.H., 1978. Petroleum Formation and Occurrence. Springer-Verlag, Berlin, 538 pp. Ulmishek, G.E, Charpentier, R.R. and Barton, C.C., 1993. The global oil system: the relationship between oil generation, loss, half-life, and the world crude oil resource: discussion. Am. Assoc. Pet. Geol. Bull., 77: 896-899. Welte, D.H., 1970. Organischer Kohlenstoff und die Entwicklung der Photosynthese auf der Erde. Naturwissenschaften, 57:17-23.
R. G. Miller
56 Wilson, R.D., Monaghan, EH., Osanik, A., Price, L.C. and Rogers, M.A., 1973. Estimate of annual input of petroleum to the marine environment from natural marine seepage. Trans. Gulf Coast
R.G. MILLER
Assoc. Geol. Soc., 23: 182-193. Wilson, R.D., Monaghan, EH., Osanik, A., Price, L.C. and Rogers, M.A., 1974. Natural marine oil seepage. Science, 184: 857-865.
BP Exploration Operating Company Limited, 4/5 Long Walk, Stockley Park, Uxbridge, Middlesex UBll 1BP, UK
57
T h e w o r l d of reserve definitions - - can there be one set for e v e r y o n e ? T.J. Beardall
Within the oil and gas industry, the search for a consistent world-wide set of definitions for reserve quantities has been an elusive goal for many years. This paper discusses some of the historical and technical issues which separate the industry into the "deterministic" and "probabilistic" camps. It also reviews the attempts by the Society of Petroleum Engineers to promote its definitions as a world-wide standard and summarises feed-back from the membership. Finally, the paper tries to look to the future and offers some thoughts on how an industry wide set of reserve definitions might be achieved.
Introduction
reserves and they represent only a small part of the total resource base as illustrated by Fig. 1. In this paper I will concentrate on definitions of reserves; the remainder of the resource base will be the subject of other papers in this volume. Reserves numbers are used by the numerous stakeholders in the oil and gas industry. They are needed for diverse purposes such as field development plan-
The petroleum resource base can be classified in a number of ways but all classifications recognise both geological and economic uncertainties. Within a classification system, some part of the resource base will have been discovered and is economically recoverable. These estimated quantities are called
THE WORLD OF RESERVES DEFINITIONS RESERVES AND RESOURCES
Resources
I
I
Discovered resources
Undiscovered resources
or
or
initial volumes in place
future initial volumes in place
I Initial reserves
Unrecoverable volumes
I Future initial reserves
I Future unrecoverable volumes
I
- currently uneconomic volumes - residual unrecoverable volumes
Fig. 1. Classification of resources and reserves.
Quantification and Prediction of Petroleum Resources edited by A.G. Dorfi and R. Sinding-Larsen. NPF Special Publication 6, pp. 57-62, Elsevier, Amsterdam. 9 Norwegian Petroleum Society (NPF), 1996.
58
T.J. Beardall
ning, company strategic planning, loan financing, financial reporting, mergers, acquisitions and government reporting. They are used by reservoir engineers and geoscientists, economists, accountants, financial analysts and civil servants to name but a few, each with his or her perspective on what a reserve number should represent. To add to this diversity of opinion, there are different oil industry cultures around the world that have resulted in different concepts of what reserves numbers should represent. For instance, there is little correlation between the Middle East view centred on OPEC quotas and the requirements of the United Stated Securities and Exchange Commission (SEC). Against this background, a number of organisations such as the World Petroleum Congress (WPC) and the Society of Petroleum Engineers (SPE) have tried to develop and promote reserve definitions that can be used as an industry standard on a worldwide basis. In particular, the SPE published revised definitions in 1987. Given the SPE's large international membership, the SPE Board hoped that, if the revised definitions were acceptable within the SPE, they would quickly become the industry standard. In practice, the definitions were not accepted by the world-wide membership and the industry is still a long way from achieving a consensus. The main purpose of this paper is to discuss the SPE consultation process and the subsequent attempt to modify the reserves definitions before looking to a way forward. However, it is important for the reader to understand some of the issues concerning the history and basis of the main reserve definitions used in the industry. These are also discussed.
North Sea and other frontier areas. The application of probabilistic definitions was originally developed within Shell. They were first used publicly in the United Kingdom in 1982 during the privatisation of British Gas. Within the probabilistic definitions the word Proven was used to describe the 90% probability that the reserves would be larger than the number quoted. Corresponding probabilities of 50% and 10% were used to describe Proven + Probable and Proven 4- Probable + Possible categories. Although the words used are virtually identical to the deterministic definitions, they describe different quantities and concepts and this has led to confusion and misunderstanding ever since (D.R. Keith et al., 1986). In 1984, the WPC published a similar classification system to the SPE, and then in 1987 the SPE published a revised set of definitions which formally introduced Probable and Possible reserves categories: Probable being "less certain" than Proved and Possible being "less certain" than Probable. Proved reserves were to be estimated using "current prices and costs" with no escalation for inflation. Based on comments, particularly from the European sections, that they could not use the revised definitions, the SPE set up a task force and then proposed further revisions in 1991. The main changes were the introduction of confidence levels to try and quantify what is meant by "reasonable certainty" and the use of "specified" rather than "current" prices and costs. Finally in 1993, the Canadian Institute of Mining, Metallurgy and Petroleum has published a set of definitions for use in Canada which have many similarities to the proposed 1991 SPE revisions.
History of reserves definitions
Comparison of probabilistic and deterministic definitions
In the late 1930s, the American Petroleum Institute started to classify reserves and these were reported in its annual bulletins using the term Proved reserves. In 1965, the SPE published its first definition of reserves and this used the words "reasonable certainty" to describe Proved reserves. In 1979, the SEC issued its own similar definitions and shortly thereafter in 1981, the SPE updated its definitions to reflect the improvements in technology that had occurred over the period. However, the wording and concepts within both the SEC and SPE definitions were firmly based on single onshore wells. Also in this period, engineers had started to use the terms Probable and Possible to describe reserves which could not be classified as Proved. Meanwhile, it had been recognised that some of the restrictions within the definitions were not appropriate for the development of large offshore fields in the
Although there are many institutions promoting different reserve definitions, most can be divided into two camps: those using a deterministic approach based on the SPE and SEC definitions, and those based on probabilistic concepts. As mentioned previously, the deterministic definitions have been developed using single onshore wells as the primary quantification unit. Estimates are typically made using either decline curve analysis or relatively simple volumetric calculations where the uncertainty is mainly in the drainage area of the well and the likelihood of offset locations being productive. All other parameters tend to be best estimates although there are limitations on the use of hydrocarbon contacts. Finally, to reflect the incremental investments required, the Proved reserves are classified into developed and undeveloped categories.
The w o r m o f reserve definitions ~ can there be one set f o r everyone ?
Reserves estimated on a field scale are often built up from single well values so that Proved reserves tend to grow as more wells are drilled. Once a field is fully developed all the reserves are considered Proved and any uncertainty in reservoir performance is rarely quantified. Conversely, probabilistic definitions have been developed to describe large accumulations, often offshore, where the primary quantification unit is on a field scale. The whole approach is to recognise and quantify uncertainty for all parameters and then define an overall probability distribution for the field, based either on volumetric calculations or reservoir performance. The Proven reserves are always at a constant confidence level (usually 90%), while the Proven + Probable reserves are at a 50% confidence level. For practical purposes, the Proven + Probable reserves represent a most likely or best estimate for the field reserves at all stages of development. Finally, the reserves at all confidence levels are classified into Commercial, Potentially Commercial and Technical categories dependent on their economic and approval status. Economics are based on forecast prices and costs.
To help illustrate how the two sets of reserve definitions are applied, Fig. 2 shows a hypothetical field with two fault blocks. Block A to the east is fully developed. It has been on production for some time using a peripheral waterflood. Block B to the west is
59
undrilled but it is believed that the fault separating the two blocks probably seals. The engineer has made a deterministic best estimate of 50 MMstb (million stock tank barrels) recoverable from Block A and 10 MMstb recoverable from Block B assuming the two blocks have a common oil water contact. Under deterministic definitions, all of Block A would be considered Proved, i.e. 50 MMstb. Block B, because it is undrilled, would be classified as Probable. Additional Possible reserves might be considered based on a deeper contact in Block B but this scenario will often be ignored. Under probabilistic definitions, there will be both volumetric and recovery factor uncertainty in Block A. Block B is even more uncertain; it might not be oil bearing, it could be in communication with Block A or it could have a significantly deeper oil water contact. All these possibilities would be included in the probability distribution illustrated in Fig. 2. As a result, the reserves at the 90% confidence level would be only 35 MMstb with a corresponding upside of 70 MMstb at the 10% confidence level. In this case, the 50% confidence level or Proven + Probable reserves coincide with the deterministic Proved estimate of 50 MMstb. However, it needs to be recognised that the two estimates might not represent the same "barrels". For example, an actual recovery of 50 MMstb could be made up of a low recovery factor in Block A offset by additional reserves in Block B.
Fig. 2. An example of the application of deterministic and probabilistic reserve definitions.
T.J. Beardall
60
The other point to note is that the deterministic Proved § Probable estimate lies between the 50% and 10% confidence levels. In this case, it corresponds to about the 30% confidence level, so the overall range of the deterministic reserve estimates is much narrower than in the probabilistic case.
The SPE process The revised SPE definitions were published in 1987. In response to the SPE Board's desire to promulgate the definitions world-wide, a European Reserves Definitions Committee with representatives from a large number of oil companies and government institutions was set up. Many of the discussions were complex but in essence the committee concluded that the SPE definitions could not be used in a European context for two main reasons: lack of definition of what is meant by "reasonable certainty", and the specification of current prices and costs. These findings were reported to the SPE Board and as a result a task force was set up to try and address the concerns raised by the European committee and others. This task force recommended that numerical confidence levels should be associated with the words Proved (80 to 90%), Proved + Probable (50%) and Proven -t- Probable + Possible (10 to 20%). The wording was changed to emphasise that these confidence levels were not necessarily probabilities but a measure of the confidence the estimator had in the reserve estimate, as illustrated in Fig. 3. The task force also
recommended that the use of current prices and costs should be changed to specified economic conditions with the rider that current prices and costs were the industry standard. The proposed revisions were published for comment in 1991. Table 1 shows a brief summary of the responses by geographical area and the level of agreement with the two main issues. As can be seen, the proposals were generally widely accepted outside the USA and there was reasonable support from within the USA. However, many of the concerns in the USA were directed at the requirements of the SEC and this was particularly evident from the responses of the large group of US based consultants. The consultants and other respondees argued that the revised definitions were in conflict with the SEC. In particular, they objected to the use of specified rather than current prices and costs, and recommended that the SPE definitions should not be changed until the SEC changed. Table 1 SPE 1991 proposed divisions: summary of responses Area
Number of comments
Agreement with confidence levels (%)
Agreement with specified economics (%)
USA Canada Europe Rest of the world
86 2 18 9
35 100 89 67
24 100 94 45
US consultants
18
16
22
Fig. 3. Use of confidence levels - - SPE proposals.
The world of reserve definitions - - c a n there be one set for everyone?
The detailed responses also confirmed some earlier observations that for fields with good production history, estimators were using their best estimate as the Proved reserve: corresponding to Proven 4- Probable reserves under probabilistic terminology. This was not the case at the earlier stages of field development where Proved reserves tended to agree more closely with the 90% confidence level. It was therefore clear from the responses that the term "reasonable certainty" corresponds to a varying confidence level which is dependent on the stage of development of the field. This varying confidence level appears to be the biggest stumbling block in the way of bringing the two types of definitions together. One of the other issues of debate was that under probabilistic definitions there is an obvious problem with aggregation of reserves. If probabilistic estimates are added arithmetically then, other than at the 50% probability level, the probabilities are not preserved and the lower and upper bounds therefore become increasing pessimistic or optimistic respectively. Theoretically therefore, providing that the reasons for the uncertainty are not common to all fields, reserve estimates should be added probabilistically to preserve the correct confidence levels. This creates presentation problems with tables that "do not add up". This point was used to argue against using probabilistic definitions. However in reality, the same problems occur with deterministic definitions. If the confidence levels vary from one field to another depending on the development status, i.e. the estimator is adding a 90% confidence level estimate with a 50% confidence level estimate then the problem is at least as bad as for probabilistic estimates reserves. The estimator certainly has no idea what the resultant confidence level might be. The other major concern was that the SEC and, as a consequence, the SPE does not allow the disclosure or routine reporting of Probable and Possible reserves. Much of the potential resource base is hidden and therefore there is a tendency to call a reserve Proved to give it value. By specifying that Proved reserves had to correspond to an 80% confidence level there would undoubtedly be reserve write downs which could only be compensated for by disclosing Probable reserves. This is against the SEC rules. Following the membership feedback, some further modifications were made in an attempt to resolve some of the issues and address the membership concerns. The modifications also tried to further distance the definitions from probabilistic concepts to combat the ignorance and misconceptions which were evident from the initial responses. However, after a further restricted round of comment, the SPE Board has decided that there is insufficient consensus to move
61
forward and the modifications have been shelved for the time being.
The future As mentioned in the introduction, there are different cultures throughout the oil industry and the debate on reserves definitions has highlighted many areas of ignorance and misunderstanding on all sides. Before attempting to revise the definitions again, the SPE intends to try to reduce the level of misunderstanding. A number of vehicles are being considered such as a Forum series, panel discussions and conference papers. However, a number of key issues will also have to be resolved, particularly the dominant influence of the SEC reporting requirements. One way forward may be for the SEC and the industry to adopt a "Canadian style" system. The Standing Committee of the Canadian Institute of Mining, Metallurgy and Petroleum (CIMMP) has recently published a report (DeSorcy et al., 1993) detailing proposed reserve definitions for use in Canada. The report also gives some simple guidelines as to how the definitions should be applied. In Canada, reporting of Probable reserves is allowed, as is the use of specified prices and costs. The CIMMP defines Proved reserves at the 80% probability level while Probable reserves are those between 40 and 80% probability. While the system can be criticised on a number of minor points it is a good attempt to fit both the deterministic and probabilistic camps together without needing to run Monte Carlo simulations every time. As well as Proved, Probable and Possible reserves, the CIMMP specifies that Expected reserves can be calculated using: Expected reserves = Proved + (Pb x Probable) + (Ps x Possible) where Pb = probability of recovering the probable reserves (80 to 40%), and P~ = probability of recovering the possible reserves (40 to 10%). The expected reserves can be aggregated to give a likely out-turn for a large number of reservoirs. The Canadian system is very similar to the proposed revisions to the SPE definitions and therefore suffers from many of the same problems in gaining acceptance. However, if we are to achieve a worldwide system something similar is needed, and the author believes the Canadian system should be used as a model for further debate.
Conclusions The major difficulties with bringing deterministic and probabilistic definitions into a common system
T.J. Beardall
62
centre on two main issues. Firstly, the term "reasonable certainty" used to define Proved reserves under a deterministic system has a changing probability which is dependent on development status or production history. Secondly, the restrictions imposed by the SEC in forcing the use of current prices and costs and the non-disclosure of Probable and Possible reserves means that a major part of the resource base is hidden. This conflicts with practice in many other parts of the world where full disclosure is encouraged. The future has to be to define a constant confidence level for Proved and then to allow disclosure of other categories of reserves. Therefore either the SPE has to break the SEC chain or the SEC itself will need to change. This process will require time, effort and education. It is possible to have a single set of reserve definitions and the Canadian system may be a way forward. There can be one set for everyone ~ but not yet.
T.J. BEARDALL
References DeSorcy, G.J., Warne, G.A., Ashton, B.R., Campbell, G.R., Collyer, D.R., Drury, J., Lang, R.V., Robertson, W.D., Robinson, J.G. and Tutt, D.W., 1993. Definitions and guidelines for classification of oil and gas reserves. J. Can. Pet. Technol., May, pp. 10-21. Eleventh World Petroleum Congress, 1984. 1933 and 1983 Study group report, classification and nomenclature systems for petroleum and petroleum reserves. Keith, D.R., Wilson, D.C. and Gorsuch, D.P., 1986. Reserve definitions - - an attempt at consistency. Society of Petroleum Engineers European Offshore Petroleum Conference, London, October, SPE Paper 15865. Society of Petroleum Engineers, 1965. Definitions of proved reserves for property evaluation. J. Pet. Technol., July, p. 815. Society of Petroleum Engineers, 1981. Proved reserves definitions. J. Pet. Technol., Nov., pp. 2113-2114. Society of Petroleum Engineers, 1987. Definitions for oil and gas reserves. J. Pet. Technol., May, pp. 577-588. Society of Petroleum Engineers, 1991. SPE considering changes for reserves definitions. J. Pet. Technol., June, pp. 708-709.
ERC Tigress Ltd., Chapel House, Liston Road, Marlow, Bucks, SL7 1XJ, UK Present address: T.J. Beardall and Associates, Ltd., 2 Silver Lane, West Challow Wantage, Oxon 0X12 9TX, UK
63
Resource classifications and their usefulness in the r e s o u r c e m a n a g e m e n t of an oil c o m p a n y Kjell A. Abrahamsen, Kjell Lindbo and Jan Vollset
Statoil has developed a new resource classification system to support investment decisions and the management of petroleum assets. To achieve this goal, resource definitions have been related to the value chain from early exploration to production, and to economics. Each resource class is associated with a typical business action. Both undiscovered and discovered resources are included, and the discovered, undeveloped petroleum assets have had a major focus due to their importance in resource management. The use of the classification system increases the awareness of the quality and consistency of the resource calculations and of the booking of discovered resources. A method of analysing prospects and finds in their geological segments (fault blocks, etc.) has been chosen to achieve a fairly consistent reserve assessment which reduces the possibility of surprises. The estimates of resources are made using probability distributions. In this way the summation of resources by the use of Monte Carlo simulation is a fast operation and the variance can be estimated. The proven reserves, which we define as a low value with a high degree of certainty (90-percentile), are easy to calculate both within individual fields and for aggregated reserves of many fields.
Introduction In the past Statoil has used the resource classification system developed by McKelvey, 1975. However, Statoil's transition from a company mainly investing on the Norwegian Shelf to an international oil company, and the persistent low oil prices resulted in a decision to improve the management of our inventory of prospects and discoveries. A new resource classification system was one element in this process. The following specification was made: - The system should establish a company standard for internal use. - The system should include both undiscovered and discovered resources. -Fractiles from probabilistic distributions should define proved, probable and possible reserves. - To avoid surprises a more conservative booking of discovered resources should be introduced. - The system should be based on the profitability of prospects, finds and fields. - The categories should be related to various business actions. - The system should be suitable for target setting and performance measurement. - The system should be easily implemented into Statoil's resource database. Rules should be formulated for translating the -
categories into the standard reserves classes of external institutions like SPE and WPC. From the large number of publications on resource or reserves classifications, we focused on two documents which try to establish a standard for universal adoption (Martfnez et al., 1987; SPEE, 1988), one document which presents a Norwegian contribution (Ormaasen et al., 1991), and one review paper (Grace et al., 1993). The Society of Petroleum Engineers (SPE) reserves classification (SPEE, 1988) deals with reserves (proved, probable and possible) only, and does not include undiscovered resources or uneconomical discoveries. Reserves are defined as petroleum anticipated to be commercially recoverable from known accumulations under current economic conditions and established operating practices. Reserves are classified as proved if they can be estimated with reasonable certainty. Proved reserves are limited by lowest known structural occurrence of hydrocarbons in the absence of engineering data in areas delineated by drilling. The SPE system is a deterministic system but it recognises that probabilistic methods can be used. The World Petroleum Congress (WPC) classification system (Martfnez et al., 1987) includes both discovered and undiscovered petroleum (called "undiscovered potential recovery"). Reserves are defined as the recoverable portion of petroleum deposits. The
Quantification and Prediction of Petroleum Resources edited by A.G. Dor6 and R. Sinding-Larsen. NPF Special Publication 6, pp. 63-70, Elsevier, Amsterdam. 9 Norwegian Petroleum Society (NPF), 1996.
64 definition of proven reserves is rather similar to that of SPE. The unproven reserves are required to be economically recoverable. However, the effect of possible future improvements of economic conditions and new technological developments are allowed when allocating petroleum deposits to the unproved reserves categories. WPC recommends the use of probabilistic methods. Classifying the potential recovery of an oil field by the proven, probable and possible concepts alone is considered to be too imprecise for the decisionmaking process when major investments are considered. WPC therefore suggests the use of a probability distribution in which a low value corresponds with the proven reserves, an intermediate value represents the expectation of proved plus unproven reserves and a high value indicates the upper limit. NPD's aim is to provide a resource classification that supports the government's planning of the future activities on the Norwegian Continental Shelf (Ormaasen et al., 1991). Reserves are defined as that part of the recoverable resources that can be sold at acceptable economic conditions from fields declared commercial by the owners. The undiscovered resources include both those that are mapped and those which are unmapped (speculative). The term "resources" is mainly used to mean recoverable resources. NPD uses the terms minimum, most probable and maximum resources, and in our terminology relate them to 90, 50 and 10 percentiles, respectively, when a probability distribution is available. NPD deliberately avoids what it calls the "dubious" terms proven, probable and possible. Grace et al. (1993) compare the SPE and WPC reserve definitions. The SPE definitions are most applicable in mature situations and tend to be conservative. The probable and possible categories are vaguely defined, which limits their use in immature field evaluations. The background for this situation is the focus on legal and regulatory environments, small tracts and a strong emphasis on performance projections (decline curves). One result is a significant reserve growth through positive revisions. The WPC definition, however, is based on reservoir-level treatment and brackets all potential reserves in a reservoir, and thus facilitates the identification of under-recovered situations. The more precise definitions used by WPC allow increased use of provenplus-probable definitions in such areas as investment planning. Grace et al. (1993) also discuss the lack of enthusiasm among US engineers for the probabilistic reserves definitions. This may be related to unwillingness to accept Monte Carlo techniques, unwillingness to depend on volumetric estimates, concep-
K.A. Abrahamsen, K. Lindbo and J. Vollset
tual difficulty with applying a recovery factor range, difficulty in applying a reservoir-level analysis to per-well evaluations by some process of allocation, and difficulty in recognizing the relevance of the technique to mature, established production situations. None of the resource classification systems discussed is particulary useful for an international oil company. Undiscovered resources and discovered resources with uncertain economics are not categorized in a way that support the decision-makers in the process of bringing resources quickly along the value chain to production. As a result Statoil made a new classification system.
The classification system The main groups in the classification system are speculative resources (postulated prospects), hypothetical resources (mapped prospects) and discovered resources. These groups are further subdivided on the basis of economics (Fig. 1).
Speculative resources Speculative resources are resources in non-mapped or poorly mapped, potential petroleum deposits. They are used to assess the petroleum potential in a basin or play, and are the basis for deciding on whether or not to apply for acreage in an area. These resources are not further subdivided. Theoretical traps and leads (potential prospects) are grouped together.
Hypothetical resources Hypothetical resources are resources in prospects. A prospect is a trap that may contain a hydrocarbon deposit, and is mapped in three dimensions. Prospects are divided by means of the expected net present value or E(NPV) concept. The E(NPV) includes the risk of drilling a dry well. The definition of E(NPV) is: E(NPV) = Pdiscovery X NPVdiscovery if- Pdry X NPVdry Fig. 2 is an illustration of how to use E(NPV). We can separate prospects into three groups depending on their outcome: (1) Profitable. These prospects have a positive E(NPV), and are commercial in the long run. (2) Potentially profitable. These prospects have a negative E(NPV), but a positive net present value after discovery. Drilling these prospects will result in a loss over a period of time. If the probability of discovery is increased by reducing uncertainty, the
Resource classifications and their usefulness in the resource management of an oil company
65
Fig. 1. Major subdivisions in the Statoil classification system.
split into several categories, e.g. parts of a gas field may be developed, while other parts await development. R e s o u r c e s and reserves
Reserves are a part of the discovered resources. Reserves are discovered resources expected to be recovered profitably. This only includes categories R1 through R6. Definitions of resource categories Fig. 2. A prospect with negative expected net present value E(NPV) and positive NPV given discovery, and which is consequently classified as potentially profitable.
resources can be transferred to the profitable category. (3) Non-profitable. Includes prospects with a negative net present value even after discovery. These are prospects that have a very small chance of ever becoming profitable. Discovered resources Discovered resources are initially subdivided according to economic criteria. They are further divided according to the maturity of the resource (how much work has been done) or according to the reason for the economic state (Fig. 3). The main and associated phase of a discovered resource may belong to different categories. Sometimes the same phase of a discovered resource must also be
The classification system is illustrated in Figs. 1 and 3. U1 Theoretical traps and leads. Resources in unmapped or poorly mapped, potential petroleum deposits. U2 Prospects, non-profitable. Resources in prospects with a negative net present value given discovery. U3 Prospects, potentially profitable. Resources in prospects with a positive net present value given discovery, yet with a negative E(NPV). U4 Prospects, profitable. Resources in prospects with a positive E(NPV). D1 Non-profitable, isolated. Non-profitable, discovered resources, that are too small for an independent profitable development. There is no adjacent discovery or potential discovery that makes a joint development profitable. D2 Non-profitable, restrictions regarding price, mar-
ket, transportation technology or development technology. Non-profitable, discovered resources, typically large, where the economic outcome is the result of the petroleum price and/or lack of
K.A. Abrahamsen, K. Lindbo and J. Vollset
66
Fig. 3. Determination of the resource categories for discovered resources.
market and/or lack of transportation/development technology. D3 Potential joint development. Non-profitable discovered resources, potentially profitable by way of a joint development with adjacent resources. The adjacent resources may be discovered or non-discovered. Adjacent non-discovered resources may be in other segments in the same prospect or in new prospects. D4 Recovery uncertain. Resources with known petroleum volumes present in the reservoir, but a very uncertain recovery factor. An economically reliable appraisal cannot be made.
D5 Fields with potential for increased recovery. Resources connected to non-sanctioned, yet technically feasible methods to increase recovery in petroleum accumulations from the categories "fields, late appraisal stage" to "fields, producing" (R3 to R6). The profitability of a potential implementation is not clarified. R1 Fields, long-term perspective. Reserves in fields that are being evaluated for development. The time for production start-up is not determined, or is estimated to be several years ahead (tentatively more than 10 years from now). The resources are expected to be profitable. Still there may be major
Resource classifications and their usefulness in the resource management of an oil company
R2
R3
R4
R5
R6
uncertainties regarding geology, technology and recovery methods (including methods to increase recovery). Fields, short-term perspective. Reserves in fields that are evaluated for development. The time to production start-up is estimated to be short (tentatively within 10 years). The resources are expected to be profitable. There may still be major uncertainties regarding geology, technology and recovery methods (including methods to increase recovery). In Norway, gas management studies and the work of the Gas Negotiation Committee (GFU) are important in the process of identifying gas fields belonging to this category or R1. Fields, late appraisal stage. Reserves in fields that are expected to be recovered profitably. A decision to prepare a plan for field development and operation is made. The plan for field development and operation may be in preparation or completed. The category also includes resources with a completed plan waiting for sanction by the required authorities. Fields, development sanctioned. Reserves in fields where a plan for field development and operation is sanctioned by all required authorities. If recovered, the resources are expected to be profitable. Fields, developed, non-producing. Reserves in developed (production facilities are in place) fields that are expected to be recovered profitably, but are not yet producing. If not all of the field is developed, only the developed part of the resource is classified under R5. Fields, producing. Reserves in fields where production has started, regardless of whether the field is completely developed or not. Fields with test production or early production are not included here. If not all of the field is developed, only the developed part of the resource is classified in R6.
The use of segments
The segment is the smallest geological unit of a petroleum accumulation used in the resource classification. All categories are either based directly on the segment or they represent a summation of resources in several segments. Before drilling, a segment is the part of a prospect that will be tested by one single well, that is, the area to which the result of the well can be extrapolated with a reasonable degree of certainty. A prospect consists of one or more segments. After a successful well, the segment is that part of the petroleum accumulation which with a reasonable degree of certainty can be said to contain hydrocarbons (Fig. 4).
67
Fig. 4. The method of prospect segmentation.
Resources should not be classified as discovered until there is a reasonable degree of certainty. This certainty only exists within the segment. If a prospect is not correctly divided into segments, all resources in the prospect will wrongly be classified as discovered in case of discovery. When the next segment is tested at a later stage, the resource estimates can be dramatically reduced if the well is dry. In other words, the booked resource figure was too optimistic, and the subsequent well resulted in an unpleasant surprise. The advantages of introducing the segment concept are: - Smaller resource fluctuations due to more conservative estimations, and improved book-keeping of resources. Segments may be used to retain the uncertainty of the resource estimates for discoveries and fields. - The use of the concept "segment well" will clarify which wells supply the company with "new" petroleum. -
68 - The use of the segment concept makes after-drill reviews easier. - The segment concept helps planning the number of wells required to prove and delineate a (potential) discovery. Unconfirmed segments and prospects will statistically be identical. They both carry a risk of being dry and there is an uncertainty regarding petroleum volume. Thus, there is no principal difference between a prospect and an unconfirmed segment beyond the fact that a segment is defined as part of a prospect or a field. A well only tests one segment in each prospect. The resources in this segment will (in case of discovery) be transferred from "hypothetical" to "discovered". The resources in the remaining non-drilled segments of the prospect will still be classified as hypothetical. When a prospect is drilled, the segments will gradually be confirmed. As long as resources in confirmed segments are too small for an independent development, but adjacent resources make a joint development possible, the resources are classified under D3potential joint development. The adjacent resources may be unconfirmed segments in the same prospect. However, decisions about the drilling of a prospect or a segment in field are based on the assessment of the entire prospect or field, and not only the resources in a single segment.
Statistical description of uncertainty In its classification system, Statoil has chosen to report the 90 percentile, the expected value and the 10 percentile of recoverable resources. - The 90 percentile is the resource volume for which there is a 90% certainty that the final outcome will be larger than or equal to this volume (low estimate). - The expected value (mean) is the expected final recoverable resource volume. - The 10 percentile is the resource volume of which there is only 10% certainty that the final outcome will be larger than or equal to this volume (high estimate). The expected value is a measure of "the centre" of a distribution. Discovered resources are included in the classification system with the expectation value of recoverable resources. Non-discovered resources are included with the expectation value of risked recoverable resources. Proven, probable and possible reserves are defined from the 90 percentile, expectation value and the 10 percentile of the reserve distribution. This is shown in Fig. 5.
K.A. Abrahamsen, K. Lindbo and J. Vollset
Aggregation of resources The resources are aggregated in a computer system using a Monte-Carlo simulation technique. It is possible to derive a distribution curve on all levels of the aggregation. The MC-simulation retains a statistically correct measure of the spread on the aggregated resource. One result is that proven reserves in a group of fields are not the same as the algebraic sum of the proven reserves in each field. In the example in Fig. 5 the algebraic sum of the 90-percentiles is 251 compared to the probabilistic sum that equals 265. The algebraic summation is a less attractive method for aggregation of resources. The method is in accordance with a deterministic tradition. It will not handle uncertainty in a consistent manner. With algebraic summation there is no longer a 90% probability that "proven reserves" or more are recovered. The proven reserves definition is restrictive. It is most likely that more than the proven reserves from a field will be produced rather than less. When looking at a group of fields it will be even less likely that all the fields will produce only their proven reserves or less. The probabilistic addition of reserves credits this fact by linking the definitions to levels of uncertainty. However the drawback is that the "extra proved reserves" due to probabilistic addition cannot be allocated to specific fields.
Potential use of the system Statoil's new resource classification system is developed mainly for internal use. SPE discourage the public disclosure of probable and possible resources because of the perceived potential misconception by the public. Statoil, as well as the rest of the industry, focuses on the booked resources (i.e. proven and sometimes probable reserves) in its annual reports. However, undrilled possible reserves, finds with uncertain profitability, and prospects are the future of most companies, and are of great interest for short and long term planning. The successful application of our classification system is greatly enhanced by a digital database that is regularly updated and a Monte Carlo program that calculates means and percentiles from the resource distributions. The resources on each level are assigned to specific resource categories, by building a hierarchy from segment via reservoir (pool) to prospect (pool complex). Resources in different fields can be added stochastically, and the total resources by category can be calculated for any geographical region. By defining change classes it is possible to trace the yearly changes in each class and why the change happened.
Resource classifications and their usefulness in the resource management of an oil company
69
Fig. 5. Statoil's reserve concept versus proven, probable and possible.
In this way we have a very specific overview of our resource portfolio, and can keep track of any changes. The system can also be used for target setting for and performance measurement of each operating unit. It is also possible to check quickly if the balance between the different volumes in each class is satisfactory. If not, the appropriate actions can be made to restore the balance. Also, by recording costs related to the movements of resources between different classes, the unit cost of moving resources from one class to a higher valued class can be calculated.
Conclusions Statoil's new classification system has the following advantages" - It gives a better description of fields and prospects with uncertain economics.
- It is based on the value chain from unmapped prospects to production. - The categories are linked to typical business actions. - T h e importance of economics in the system makes it useful for target setting and performance measurements. - T h e probabilistic distribution of resources within a segment, reservoir, prospect or field makes a statistically correct summation possible. - The use of the segment concept provides a better separation between areas with tested oil and areas carrying a risk of being water-bearing, and is the basis for classifying reserves as proved.
70
K.A. Abrahamsen, K. Lindbo and J. Vollset
Acknowledgements The authors thank Den norske stats oljeselskap a.s. (Statoil) for the permission to publish this paper.
References Grace, J.D., Caldwell, R.H. and Heather, D.I., 1993. Comparative reserves definitions: U.S.A., Europe, and the Former Soviet Union. J. Pet. Technol., 45(9): 866-872. Martfnez, A.R., Ion, D.C., DeSorcy, G.J., Bekker, H. and Smith, S.,
K.A. ABRAHAMSEN K. LINDBO J. VOLLSET
Statoil, N-4035, Stavanger, Norway Statoil, N-4035, Stavanger, Norway Statoil, N-4035, Stavanger, Norway
1987. Study Group Report: Classification and nomenclature systems for petroleum and petroleum reserves. Proc. 12th World Petroleum Congress, pp. 259-274. McKelvey, V.E., 1975. Concepts of reserves and resources. In: J.D. Haun (Editor), Methods of Estimating the Volume of Undiscovered Oil and Gas Resources. AAPG Stud. Geol., 1: 1114. Ormaasen, E., Knudsen, K.R. and Berli, K., 1991. Oljedirektoratets ressursklassifikasjonssystem. NPD, Contribution 31, 15 pp. SPEE, 1988. Guidelines for application of the definition for oil and gas reserves. Society of Professional Evaluation Engineers, Houston, Monograph 1, 69 pp.
71
Reserve and resource definition: dealing with uncertainty John Wilkinson
Effective use of a chosen reserves classification system requires that both the presently established reserves and future additions to the resource base be recognized. Quantification of the total resource base can be accomplished by recognizing the uncertainties associated with both technical factors, such as in-place volumes and recovery factors, and with the economic factors associated with pricing and project implementation. Either deterministic or probabilistic evaluation of the many variables can be employed to develop the best reserves estimate for a field. Reserve estimators must develop an appropriate methodology to allow appraisal of limited and highly uncertain data sources. They are often faced with the challenge of generating a conservative estimate to satisfy investment criteria and also a potential resource estimate that recognizes the upside associated both with advances in technology and improvement in product prices. Categorization of the reserve base generally falls into three subdivisions. Proved (developed and undeveloped) reserves are of most interest to the investment community while Probable reserves are the focus of the field development team. Together these two categories constitute the reserve that is expected to be produced under current technology and economic constraints. The Static (or Possible) category is commonly of most interest to those in R&D and new technology who are charged with development of new approaches to the profitable extraction of marginal reserves. This category may also be under further appraisal while awaiting either sales contracts or environmental permits to be approved.
Reserves versus resource
this s c h e m e , w e c a n r e p r e s e n t h y d r o c a r b o n assets as e i t h e r R e s e r v e s or R e s o u r c e s (Fig. 1).
A s w i t h a n y d i s c i p l i n e , a c e r t a i n set o f c o n v e n t i o n s
B y definition, a r e s e r v e m u s t h a v e b e e n d i s c o v e r e d
and terminology has developed around hydrocarbon
by a well penetration and have a reasonable certainty
s c h e m e s ( G r a c e et al., 1993). A rep-
o f b e i n g e c o n o m i c a l l y r e c o v e r a b l e in t h e future. T h e
resentation of reserves classification was developed
r e s o u r c e base, h o w e v e r c a n i n c l u d e t h o s e h y d r o c a r -
b y M c K e l v e y o f the U n i t e d States G e o l o g i c a l S u r v e y
b o n s that are e i t h e r u n d i s c o v e r e d or are c l e a r l y s u b
(USGS)
e c o n o m i c for the n e a r future. If w e c o u l d d r a w an
classification
in 1972. In a s i m p l i f i e d r e p r e s e n t a t i o n
of
Fig. 1. Reserves vs resource.
Quantification and Prediction of Petroleum Resources edited by A.G. Dor6 and R. Sinding-Larsen. NPF Special Publication 6, pp. 71-76, Elsevier, Amsterdam. 9 Norwegian Petroleum Society (NPF), 1996.
J. Wilkinson
72
analogy with gold assets, we could say that the upper left comer of the box represents gold bars in a Swiss bank while the lower fight comer recognizes the gold that is dissolved in ocean water. Clearly there is a much larger volume of gold as a resource than as a reserve, but the probability of ever realizing an economic gain from the resource is much lower. Within the framework of the McKelvey Box, a variety of subclassification practices have developed in different parts of the world and within different oil and gas companies. Whatever the nomenclature of areas enclosed by the subdivisions, it is important that both technical and economic risk be incorporated in the guidelines and practices employed for classification.
Reservoir sampling limitations One of the constant appeals by the geoscientist and reservoir engineer is for more data. It is a characteristic of the oil and gas industry and especially the offshore sector that major investments must be made based on a limited and largely qualitative set of reservoir data. In the example shown, a reservoir with a gross rock volume of 1 billion cubic metres has been evaluated by a total of 10 wellbores that have been cored, logged, pressure tested and tied to 3D seismic. As Fig. 2 shows, the fraction of the reservoir that is sampled is extremely small and the types of data obtained are typically qualitative in nature. While significant advances in the technology of measurement and interpretation of reservoir data have been realized over the past decades, the small fraction of a reservoir that can be sampled has led to development of various stochastic and geostatistical technologies to fill in the missing data between wells. Both limited data and limitations in computing power still result in both geological and reservoir flow models being extremely coarse representations of the reservoir. Thus we have a situation where early reserve estimates are often subject to significant adjustments as
a result of internal reservoir geometries that are not detectable at the scale of sampling realized during field development. These internal flow paths are often only detectable by continuous surveillance of well behaviour during the production phase. This need to continuously upgrade reserve estimates during the life of the field requires that a reserves classification scheme be able to accommodate and document changes at a reasonable frequency.
Technical uncertainty Uncertainty analysis of the technical aspects of reserves determination can be achieved in a number of ways. The most common approaches involve either sensitivity analysis of geological and reservoir flow models (by making deterministic assumptions and then recalculating the results), or alternatively, a probabilistic assessment can be made by defining a range of possible outcomes for several variables and then running a Monte Carlo analysis. In either situation it is important that the principal unknowns be identified and assessed to determine a reasonable range of possibilities that makes physical sense. For example, the deepest occurrence of oil could not be set below the spill point of the reservoir, and the combined flow rates of the wells could not exceed the facility's inlet capacity. Some key variables as shown in Fig. 3 will include factors that influence both the in-place estimates and the recovery factor. For each individual reservoir the result may or may not be sensitive to a given factor. If computer models of both the geology and flow behaviour are available then it is possible to generate sensitivity analyses by systematically varying a single variable at a time to develop a wide range of scenarios. In a situation where data is sparse, it may make more sense and be more convenient to generate a probabilistic analyses. In the future, application of geostatistical techniques combined with powerful computers will allow
Fig. 2. Reservoir sampling example.
Reserve and resource definition: dealing with uncertainty
73
Fig. 3. Technical uncertainties.
generation of multiple realizations of the geological and flow models, to produce a fully integrated confidence analysis of the recoverable hydrocarbons.
Commercial uncertainty The assessment of commercial factors and how they might impact reserves is one of the most difficult aspects of the reserves process (Fig. 4). The lesson of the 1980s is that product pricing is both uncertain and highly volatile, and that to a large degree falling prices have more than offset advances made in technology. In a stable pricing environment the technology advances would in all likelihood have improved the profitability of field development. More recently, we have seen that fiscal stability and exchange rates are playing a significant role
in the overall economics of projects. This is driven by the large investments that are made in several currencies during the development phase and, in the case of European gas sales, the fact that product revenues are also received in several currencies. Development costs are a significant factor since they occur during the early life of a field. Plateau rates are often determined by optimizing of the investments required to build a processing facility and then by drilling of wells to meet the inlet capacity. The initial rate of production will then in turn affect the recoverable reserves realized within given license period. Later in the field life the impact of tariffs, government take and license expiry becomes more significant in that the minimum economic rate for the field and/or loss of the license will become the determin-
Fig. 4. Commercial uncertainties.
74
J. Wilkinson
ing factors in estimating the ultimate recovery. This analysis should also consider the possibility of prolonging field life by using existing infrastructure, or conversely the need to shorten field life because of increasingly strict environmental regulations. Late life reserve estimates can also be impacted by decisions to delay abandonment for either cost or environmental permitting reasons. As time progresses in the life of a field, the technical uncertainty of reserves estimation is reduced but the commercial uncertainty increases as production rates drop and profitability is diminished.
Example of reserves estimating For the purposes of illustrating the application of reserves estimating logic, three reservoirs have been chosen. Each reservoir has been penetrated by a discovery well so the illustration (Fig. 5) will focus on classification of reserves versus resources. Borrowing from a classical Western movie, I have chosen to call these reservoirs The Good, the Bad and the Ugly. The primary target is the Good reservoir which contains a thick oil column overlain by a small gas cap, possibly including a small water column down dip. Delineation wells have been tested and evaluated as commercial to produce from this zone. This reservoir will be the first one developed by production wells to be drilled as part of the upcoming area development. The secondary zone, called here the Bad zone, has tested oil at low but commercial rates and will be penetrated by most wellbores going down to the Good zone. It is not anticipated, however, that the wells will be completed in the Bad zone until after depletion of the Good zone.
The Ugly zone has been penetrated by one well but has proved to be overpressured and to contain corrosive gas, so it was not tested. The dimensions of the reservoir are uncertain but it is anticipated that the field could be producible with future technology improvements and future development of gas pipelines in the area.
Plan of development Since there is no immediate gas market in the region and aquifer support of these reservoirs appears to be limited, it has been decided to re-inject produced gas as a means of maintaining reservoir pressure in the Good zone. Wells will initially be drilled updip from the oil-water contact (OWC) so that near vertical wells can be drilled first from the platform. Therefore the exact position of the OWC will remain unknown for the first several years. It is anticipated that either water or water alternating gas (WAG) injection from downdip locations will commence after the necessary injection wells are completed on the downdip flank. For the Bad reservoir, a waterflood project is planned but will only be implemented after depletion of the Good reservoir makes wellbores and facilities available. Even though this development does not look attractive at today's prices, it is anticipated that by the time of development the economics will look attractive. The Ugly zone presents far greater challenges. A new generation of metallurgy and equipment will be required on the platform to handle production from these wells and an inert gas scrubber and disposal system will have to be designed. It is not clear that this can be done either physically or economically
Fig 5. Reserves base example.
Reserve and resource definition: dealing with uncertainty
when gas sales infrastructure reaches the area. For now, it is not planned to drill the Good formation wells deeper to the Ugly formation because the added drilling and casing costs can not be justified. An overall factor in this plan might be that the production license expires within 15 years so production from the secondary Bad zone may not be completed and development of the Ugly zone would have to be contingent on license extension. It is not clear that extension can be obtained in this case.
/5
in the downdip region. On the commercial side, a significant upside is the procurement of profitable gas markets. In the case of the other two reservoirs, technical uncertainty is significant. This will most likely be reduced by wells penetrating the Bad zone but commercial uncertainty will remain. The Ugly zone has hope for future development but requires significant new developments in technology.
Categorization
Uncertainty Early in the life of the field uncertainty will exist with respect to the expected recovery efficiencies that might ultimately be realized through the life of the field. Since these recoveries are dependent on the field development plan, number of wells, etc., it is necessary to anticipate future work programs and technologies that might be applied. It is also important to anticipate the economic climate and expenses that the field will experience at the end of field life since this will impact the point at which a field is abandoned. Abandonment timing may not be coincident with the technical and economic limits of production because of either environmental needs or because the facility is later used for processing of production from other prospects. For the Good reservoir, key technical and commercial uncertainties are listed in Fig. 6. All these parameters suggest that at the beginning of field life a conservative reserve might be booked to proved, but considerable upside exists for both downdip oil and future sales of gas. Enhanced recovery potential might also be very good, if a gravity stable miscible gas drive can be performed in the region of the updip oil, or alternatively a WAG project can be operated
The subdivision of discovered reserves into categories serves several purposes. The naming convention used here is that employed by Esso but is similar to names used in other systems. Identification of reserves that are producible from existing wells and facilities versus those that require additional activity or investment to be producible helps focus the organization on future work programs. The classification system also creates an incentive to transfer reserves from static and probable into proved. A common practice is to depreciate investments on a unit of production basis. This basis is tied to Proved reserves by considering the remaining undepreciated investment divided by the remaining booked reserve. Thus the higher the proved reserve, the lower the unit costs (non-cash cost per unit of production). Total non-cash costs for the year are then calculated by multiplying the unit cost by produced volume. This system is designed so that revenues and non-cash costs rise and fall together as production changes. For the Good reservoir, oil that has been demonstrated to be recoverable by existing well completions can be classified as proven oil (Fig. 7 and 8). Oil below the lowest well point could potentially be called
Fig. 6. Reserves base example.
76
J. Wilkinson
Good Fig. 7. Reserves base uncertainties. Technical: Early stage of development (2 of 12 wells completed and producing); OWC not yet intersected by well but has possible DHI from seismic; recovery factor of updip wells near Gas Cap expected to be lower. Commercial: No immediate market for gas produced (requires re-injection); only first stage of develpment funded at this time.
Good Fig. 8. Reserves base categorization. Proved developed: Oil estimated to be produced by completed wells. Proved undeveloped: Other oil down to base of wells and up to GOC. Probable: Oil from base of wells to estimated OWC. Static: Gas except for fuel gas.
proven if a strong DHI from seismic was mapped or other reliable information (shown to be accurate by drilling) is available about the OWC. A more prudent course, however, will be to classify this oil as probable until further data and experience are obtained. Since it is assumed that no market exists for gas at this time, all but the fuel gas volumes should be kept as Static until a commercial market is developed. The Bad zone would primarily be classified as Probable (likely to be developed) while the Ugly zone would be considered Static.
Conclusions Reserves and Resource classification is an important part of a company's business practice. The assets and future prospects of the business are to large extent tied to petroleum that lies thousands of meters below the surface. Since the physical properties of oil and gas reservoirs can only be measured in an imprecise manner, and since we can only have an estimate of the future worth of production at the time that investments are made, the reserves estimator must of necessity be able to rationalize and estimate both technical and economic uncertainty. The factors that enter into the uncertainty analysis, should be periodically reviewed and tested against either new reservoir data or updated economic outlooks to ensure that
J. WILKINSON
Esso UK, 94-98 Victoria St., London SWIE 5JW, UK
both the appropriate volumes and classifications are applied to the booked reserves. A single recipe or formula cannot readily be applied to this problem, since each discovery or prospect will have unique attributes and economic constraints that will impact the way in which it is reflected on the books. Also, different product streams from a single reservoir might carry different classifications based on the presence or absence of profitable sales options and/or environmental constraints on shipping, disposal or re-injection. Within most countries and companies a set of conventions are in place to aid in use of acceptable reserves booking practices. The essential ingredient that must applied is judgement that is based on time and experience working with the reserves booking process. No system or set of rules can cover all situations, so that application of human experience and judgement is essential.
References Grace, J.D., Caldwell, R.H. and Heather, D.I., 1993. Comparative reserves definitions: USA, Europe, and the former Soviet Union. J. Pet. Technol., Sep., pp. 866-872. McKelvey, V.E., 1975. Concepts of reserves and resources. In: John D. Haun (Editor), Methods of Estimating the Volume of Undiscovered Oil and Gas Resources. AAPG, Stud. Geol., 1:1114. Internal studies and documentation, Esso Norge AS, Forus.
77
The Norwegian Petroleum Directorate's Resource
Classification System Kjell Reidar Knudsen
The Norwegian Petroleum Directorate (NPD) has the responsibility of keeping record of all petroleum resources on the Norwegian Continental Shelf. Its classification system is appropriate to the NPD's needs and the situation on the Norwegian Continental Shelf. The NPD Classification System was established in 1985, and it has been revised twice, in 1991 and late 1993. The NPD system is somewhat different, compared to the USGS-System and the recently revised SPE Resource Classification System. These differences are briefly explained. The main part of the paper is devoted to the NPD definitions of resource categories, the naming convention, the units and how particular discoveries and the total resource base are presented from the NPD's side. The software system "PROFF" is also explained. "PROFF" is a database system developed for the NPD to perform statistics and presentations of resource estimates. The paper also gives references to publications on resource figures from Norway.
Purpose There are several reasons for establishing a resource classification system within an organization. In general terms we can say it is a tool for communication, one way to tell about the resources in which a company or a nation has interest. There are two main purposes, which in turn will require different ways of reporting resource figures:
Financial reporting Current Federal securities law regulations require U.S. oil and gas producing companies registered with the Securities and Exchange Commission to include in their annual reports to shareholders estimates of the amount and value of their proven reserves in the ground. The purpose is to give the public m and particularly investors and potential investors m a better understanding of each company's worth and a better means of comparing one company with another. Such systems often focus on "proven reserves". It is debatable how precisely such figures represent the value of the company, bearing in mind that production will occur over a long period in which significant changes may occur: in oil prices, in production cost and production capabilities, and in competitive sources of energy. Confusion may also arise from the exclusion of other, less certain classes of oil and gas resources (probable resources, static resources and undiscov-
ered potential), for which formal reporting is not required. The United States Geological Survey (USGS) system established in 1972 (Securities and Exchange Commission, 1981) and the Society of Petroleum Engineers (SPE) system established in 1987 (Cronquist et al., 1987) are good examples of systems of this kind. Such systems tend to be conservative, resulting in considerable reserve growth through positive revisions. They are most suitable for mature situations, such as in the USA. These are often "deterministic" systems, meaning that one single figure is calculated for each reserve category.
Investment decisions Other systems, such as that of the World Petroleum Congress (WPC) (Martfnez et al., 1984), are more focused on probable or possible reserves, which in many cases may turn out to be more important than the proved reserves. For example, when a new field is discovered, several years may be required before the field size, its producing characteristics and its recoverable reserves are sufficiently defined for it to be considered commercial. Nevertheless, these values should be considered when making, for instance, investment decisions. Such systems are also often based on an probabilistic approach, i.e. the definitions attempt to quantify uncertainty and provide a measure of upside potential. Such systems use uncertainty as a criterion for defin-
Quantification and Prediction of Petroleum Resources edited by A.G. Dor6 and R. Sinding-Larsen. NPF Special Publication 6, pp. 77-81, Elsevier, Amsterdam. 9 Norwegian Petroleum Society (NPF), 1996.
78
K.R. Knudsen
ing reserve categories. The definitions are here based on reservoir-level treatment (instead of well-level) and the bracketing of all potential reserves in a reservoir. Application of this approach is particulary useful for unconventional plays and where quantification of the upside potential is needed to support high-cost development programs (e.g. in the North Sea).
location, geological age and other factors are useful. The NPD Resource Classification System is also designed to be an important tool for such analysis. Most known to our external contacts is, however, the resource status given in the NPD annual report (NPD, 1993b)and the Royal Ministry of Industry and Energy's Fact Sheet (MIE, 1993).
Purpose of the NPD Resource Classification System
NPD's System - - classification of resources according to maturity
One of the main tasks of the Norwegian Petroleum Directorate (NPD) is to keep an overall account of the petroleum resources on the Norwegian Continental Shelf. This account forms the basis for the authorities' planning of future activities; the total provides an estimate of the resource base, and each field or group of discoveries and prospects provide input for decisions on new development. As part of the evaluation of new play models, or areas, well organized statistics of discovery size,
The NPD's Classification System comprises both petroleum resources in place and recoverable resources. The resources are divided into undiscovered and discovered resources. Only the values for recoverable resources are normally published (see Table 1). Undiscovered resources
The estimates of undiscovered resources include both mapped and unmapped prospects. The calcula-
79
The Norwegian Petroleum Directorate's Resource Classification System Table 2 Form used internally for estimates of undiscovered resources on individual areas or prospects
DATE:
PROSPECT:
UNDISCOVERED UNMAPPED
MAPPED expected value
min.
max.
HCPV
OIL 106Sm 3 STOOIP/GOIP
rain.
exp.
GAS 109Sm 3 max.
rain.
exp.
max. T.O.E. 106 min.
exp.
,
max.
RECOVERABLE
tions for unmapped prospects are carried out using the area within each play model. In addition, the NPD carries out estimates of the mapped resources in undrilled prospects, which are mapped in areas with sufficient geological control. However, for external publications, the NPD only presents the total volumes of undiscovered resources. These are distributed between North Se.a, the Norwegian Sea and the Barents Sea. The figures are given in recoverable volumes calculated in tonnes of oil equivalent (toe). In some reports a distribution between oil and gas is given. The total undiscovered resources also include resources in areas where there is very little or no regional geological control. For internal purposes, the NPD keeps account of hydrocarbon pore volume (HCPV), resources in place (STOOIP/GOIP) and recoverable resources for oil and gas indicated at minimum, expected and maximum estimates, together with estimates for each prospect and area (see Table 2). The basis for minimum, expected and maximum estimates is an overall evaluation provided by the mapping documentation for the individual prospect. Discovery probability and volume distribution is also documented. In particular reports, the NPD also publishes a distribution for total undiscovered resources in order to describe the uncertainty in these estimates (NPD, 1993a). Discovered resources
The discovered resources fall into two categories" discoveries and fields.
Discoveries This category comprises newly made discoveries, which are therefore currently under evaluation, and discoveries which in the current situation are not deemed to be commercial. In practice this will include confirmed and production tested resources in structures which the licensees have not yet declared to be commercial. Discoveries often include compartments which may be isolated from the main discovery and which consequently cannot be claimed to be confirmed by drilling. The extent of the actual "discovery" and the definition of what is to remain an undrilled prospect, has to be decided in each case and is a matter of discretion. This must, however, be clearly set out in the documentation from the mapping and resource evaluation work constituting the basis for the estimate. For internal purposes, the NPD keeps account of hydrocarbon pore volume (HCPV), resources in place (STOOIP/GOIP) and recoverable resources for oil and gas, indicating minimum, expected and maximum estimates for each discovery (see Table 3). The probability estimate for the various categories is based on a total evaluation, which shall be defined in the mapping documentation for the individual discovery. The volume estimates shall be documented, and if the estimate is based on Monte Carlo simulation, a 95 and 5% probability for the maximum and minimum estimate, respectively, should be used. For external purposes, the NPD normally publishes an overview of the various discoveries and their expected value of recoverable volumes of oil (including NGL) and gas (NPD, 1993c).
K.R. Knudsen
80 Table 3 Form used internally for estimates of resources on individual discoveries FIELD/DISCOVERY: DATE:
DISCOVERED MAPPED expected value
marl.
max.
HCPV
OIL 106Sm 3 min.
STOOIP/GOIP
GAS 109Sm 3
exp.
max.
min.
exp.
CURRENT TECHNOLOGY RECOVERABLE IMPROVED RECOVERY OIL
GAS
NGL
exp. - expected value
I Accumulated production per date: ...........................
Fields Three categories of field are recognised by the NPD: Fields planned to be developed (i.e. the field has been declared commercial), fields decided to be developed (i.e. permission to develop has been granted) and fields in production.
Fields planned to be developed. This category includes all fields which the operator has declared commercial, but where the plan for development and operation (PDO) has not yet been approved by the authorities. Fields decided to be developed. This category includes all fields where an approved plan for development and operation (PDO) exists, but where regular production has not yet commenced. Fields in production. This category includes all producing fields. A production permit must be obtained from the authorities before production can commence. Fields often include compartments which may be isolated from the main field and which have not been confirmed by drilling, or have not been decided to be developed. The classification system for fields indicates total resources within the area outlined in the plan for development and operation (fields decided to be developed or in production) or by the declaration of commerciality (fields planned to be developed).
Resources associated with a field, in reservoirs either above or below or in nearby structures, may be given a separate discovery or prospect name and may be referred to in the classification system either as "discovered" or "undiscovered" resources depending on whether or not they have been confirmed by drilling. For internal purposes, the NPD keeps account of hydrocarbon pore volume (HCPV), resources in place, oil, associated and free gas (STOOIP/GOIP) and recoverable resources and reserves for oil, gas and NGL. The various estimates are based on a total evaluation, which is defined in the mapping documentation for the individual field. With regard to fields in production, the volumes and production forecasts reported by the operators most frequently form the basis for the NPD's account of resources. For external purposes, the NPD publishes an overview of the various fields and the expected value of originally recoverable volumes of oil, gas and NGL. In the reserves account the original volumes are stated. A separate system is used to keep account of produced volumes of oil, gas and NGL, as well as injection of gas.
Naming conventions In connection with the organization of the Resource Classification System, it has been deemed necessary to lay down rules for designation and naming
The Norwegian Petroleum Directorate's Resource Classification System
of the petroleum deposits on the Norwegian Continental Shelf (NPD, 1993c). These rules are based on the guidelines for naming of petroleum deposits issued by the Royal Ministry of Industry and Energy, 11 October 1984 (MIE, 1984). The conventions are shown below for various stages of exploration and development.
Prospect A prospect is a mapped potential discovery. If there are possibilities for discoveries at several stratigraphic levels, each of these, if mapped prior to drilling, will be a prospect. Names for the prospects may be chosen by the licensees themselves. Each prospect is given its own designation. The designation contains quadrant and block number, plus a name of the operator's own choice. Example" 9/2 Epsilon.
Discoveries The nomenclature for new discoveries is the discovery well name plus a name chosen by the operator. It is recommended that the operator chooses the same name as the prospect name. It may also be useful to include the geological age in the name at this stage. For existing discoveries the discovery well designation plus, if applicable, an unofficial or an approved name will continue to be used. Example: 9/2-1 Gamma, 24/5-1 Jura, 15/5-1 Dagny.
Fields Names of fields are approved by the Norwegian Petroleum Directorate when the Plan for Development and Operation (PDO) is submitted. Application for approval of name may, however, be submitted earlier if the discovery is well mapped. The latest time of application for approval of field name is consequently when the Plan for Development and Operation (PDO) is submitted. The application may also contain a request for the new discovery to be included in a field which already has an approved name. The application for approval of field name contains information on which discoveries are to be included in the field and on whether the proposed name has been approved in advance by Norsk Spr~krgtd (The Norwegian Language Council). A map showing the field boundaries is included in this application.
K.R. KNUDSEN
81
PROFF database system (prospect, field and discovery database) To keep a record of all resources on the Norwegian Continental Shelf defined using the above system, the NPD has developed a database system called PROFE The system runs on a Norsk Data, Sibas database with Unique report and user interface. However, a prototype on a Unix/Sybase platform is now in the testing phase. The database is linked to our corporate database called ILGI (geo-information database system), where all discovery and field names are consistency checked, and from which attributes relevant to these can be retrieved. This is useful when reports are requested of groups of fields and prospects within a specified water depth, company name, geographic area or other user-specified criterion. The PROFF database also keeps track of the discovery wells and the date of discovery, and subsequent delineation wells which influence the resource estimates. This makes it possible to refer the resource estimates at any time back to the date of the first discovery well, a useful procedure for statistical play analysis. The database also keeps track of dates and history of all changes made to each prospect, discovery and field. Either NPD's official figures or alternative estimates (e.g. operator's estimate) can be reported.
References Cronquist, Ch. et al., 1987. SPE Task Force on Reserve Definitions. J. Pet. Technol., May, pp. 576-578. Martfnez, A.R., Ion, D.C., DeSorcy, G.J., Dekker, H. and Smith, S., 1984. Classification and nomenclature systems for petroleum and petroleum reserves SGR1. l lth World Petroleum Congress, London, 1984, pp. 325-339. Norwegian Petroleum Directorate, 1993a. Description of the Norwegian Petroleum Directorate's Resource Classification System. NPD Contribution No. 37. Norwegian Petroleum Directorate, 1993b. Norwegian Petroleum Directorate Annual Report 1992. ISBN 82-7257-388-1, pp. 78-89. Norwegian Petroleum Directorate, 1993c. Petroleum Resources, Norwegian Continental Shelf. NPD report number ISBN 82-7257394-6. Securities and Exchange Commission, 1981. Reserve definitions as shown in Bowne and Co. Inc. Pamphlet Dated March 1981. Regulations S-X, rule 4-10. The Royal Ministry of Industry and Energy, 1993. Norwegian Petroleum Activity Fact Sheet 93. The Royal Ministry of Oil and Energy, 1984. New regulations for naming of petroleum deposits on the Norwegian Continental Shelf. Letter with ref. OED 84/1021-RB/KS.
NorwegianPetroleum Directorate, Postbox 600, 4001 Stavanger, Norway
This Page Intentionally Left Blank
83
A method for the statistical assessment of total undiscovered
resources
in a n a r e a
Eivind Damsleth
The paper considers the situation where a (possibly large) number of potential hydrocarbon-bearing prospects have been identified within an area. Each prospect can be characterized by a chance factor (probability of discovery) and by a conditional probability distribution for the resources (or the reserves), given that a discovery is made. The aim is to compute a probability distribution for the total number of discoveries in the area, and the associated resources/reserves. These distributions provide crucial input to company policy. If all the prospects are assumed to be statistically independent, this is a fairly straightforward analysis. However, in most cases it is more realistic to model dependencies between the prospects, so that a discovery in one prospect changes the chance factor for other prospects in the same geographical area or within similar depositional environments. A model and a method have been developed to cope with this problem. The foundations of the technique and its implementation is presented, and illustrated by artificial but realistic examples.
Introduction An oil company will at any time have a large number of potential hydrocarbon bearing prospects under consideration. These may be prospects within the company's licence areas, prospects in open areas, or prospects in other licences which may be of interest for potential farm-in. In any case, each prospect is uniquely identified, usually from interpretation of regional seismic surveys. One definition classifies these kind of prospects as "prospective resources". The prospects have not been drilled. Thus, for each prospect there is definitely a (usually large) element of uncertainty as to whether it is hydrocarbon bearing at all. Even if there are hydrocarbons present in the prospect, there is additional uncertainty related to the actual volume in place - not to mention the recoverable reserves. Any evaluation and comparison of such prospects or groups of prospects must take these uncertainties into account. A correct assessment of such prospect will be important in several situations. If the prospects are within the company's licence areas, they constitute part of the company's underground fortune, and the evaluation of the prospects and their potential will directly influence the company's stock value. - If the prospects are in an open area offered for -
bids in one of the rounds on the Norwegian Shelf, say, evaluation of the prospects in different areas provides key input to the strategy in the bidding process. - I f the prospects are in an area offered for farmin, the correct assessment of the prospects and their uncertainty provides the key input in deciding the right price. In this paper a very simple and natural model for the uncertainty related to one particular prospect is presented. The question of statistical dependency between prospects is then addressed. This may have significant effect on the joint evaluation of a number of prospects within an area. A technique to take this dependence into account is then proposed. The method is exemplified with a real, but unfortunately anonymous, data set from Norsk Hydro. The mathematical details are given in an appendix. This topic has been gathering considerable interest for a number of years, i.e., the play concept, where the hit probability for a number of prospects and their interrelationship is expressed using the probability of a common source, reservoir rock, traps, etc. Harbaugh (1984) gives an overview of a number of applicable techniques, and Nederlof (1994) discusses the topic within a probabilistic framework. The aim of the present paper is mainly to present a new, operational technique, and to provide its mathematical/statistical foundation.
Quantification and Prediction of Petroleum Resources edited by A.G. Dor~ and R. Sinding-Larsen. NPF Special Publication 6, pp. 83-90, Elsevier, Amsterdam. 9 Norwegian Petroleum Society (NPF), 1996.
E. Damsleth
84
The model
Model for single prospects Each prospect is characterized by: (1) The chance factor, i.e. the probability of finding hydrocarbons in the prospect. (2) The probability distribution of the potential resources and/or reserves, g i v e n that a d i s c o v e r y is m a d e . For simplicity, we have chosen to use normal (Gaussian) distributions with different means and standard deviations to describe this uncertainty. Note that the distribution is c o n d i t i o n a l . The normal distribution allows for negative values. However, with the actual means and standard deviations that are applicable for the conditional prospect volumes, the probability of negative hydrocarbon volumes can be neglected for all practical purposes. Although this assumption may be unrealistic, the end results are very insensitive to the actual shape of the distributions. The use of normal distributions for the conditional volumes is not crucial for the proposed technique. Any other shape, e.g. log-normal, can easily be implemented. The u n c o n d i t i o n a l distribution of the hydrocarbons in place will then have a spike at 0, representing a dry prospect, while the rest of the probability mass will be distributed according to the prescribed normal distribution. It is common to have several prospect models for the same physical prospect. For example, if a discovery is made, the prospect may be oil- or gas-bearing. The expected volumes and their standard deviations may differ substantially in the two cases. In other situations, there may be different geological interpretations giving rise to different volumes in place. The multi-prospect models are handled in two steps. In step one, an individual chance factor qi, i = 1 . . . . . s is assigned to each of the s prospects models. Each prospect model will also have its associated volume distribution. These are, in turn, assumed to be mutually exclusive, so that the total probability of discovery for that particular physical prospect is given by p = ql + - " + qs.
which of the prospects in the group has actually been drilled. For such a prospect group we assign a factor )~ which describes the degree of co-variation between the prospects, that is how much a hit or a miss in one prospect will affect the chance factors for the other prospects in the group. This ;k-factor is formally defined in Appendix 1. For now, it suffices to say that it will be a number between 0 and 1, where )~ = 0 means that the prospects are statistically independent, while ~. = 1 implies maximum dependence between the prospects in the group. In this respect, )~ closely resembles the ordinary correlation coefficient.
Model for the co-variation between prospects in different groups Prospects which belong to different groups can usually be regarded as statistically independent. When, occasionally, there is also a need to model dependence between prospect groups, this can be done by another "between-group" )~-factor describing the co-variation between one typical prospect in one group and one typical prospect in the other. This "between-groups" )~-factor will normally be less than the "within-group" factors.
Necessary input data Let the total number of prospects in the analysis be N. The N prospects can be grouped into M groups, with n l, n2, ... , nM prospects in each group. n 1+" 99+ n M = N. Further, assume that the prospects are numbered continuously, so that the prospects with numbers 1. . . . . n l belong to group no. 1, the prospects numbered n l + 1 , . . . , n l -Jr-n2 belong to group no. 2, etc. until the prospects numbered N n M + 1 . . . . , N , which belong to group no. M. The following input data will be required: For each prospect i, i -- 1 . . . . , n, the user must provide: - The chance factor (hit probability) pi - The number of scenarios or prospect models si for the prospect For each scenario k, k = 1 . . . . , si within the prospect one must input: (1) The probability qik for the scenario. The sum of the probabilities for each single scenario equals the total hit probability for the prospect (qil + ' " + qi~ = Pi). (2) Whether the scenario describes an oil- or a gas case. (3) The mean J~ik and (4) the standard deviation ai~ in the normal distribution for the volume in the scenario, conditional on that particular scenario being found. - G r o u p identification, that is an identification of the group the prospect belongs to. -
Model for prospects within the same group We assume that prospects can be grouped together in "natural" groups. The grouping may be based on geographical vicinity, a common depositional environment, a common source, etc. The important feature of such a group is "statistical symmetry". This means that if one prospect is drilled and a discovery is made, the chance factors of the other prospects in the group will be affected in the same way, no matter
~3b
A method f o r the statistical assessment of total undiscovered resources in an area
For each group j, j -- 1. . . . . M of prospects the user must provide: - The common co-variation factor/~j between discoveries on the individual prospects in the group, as described in Appendix 1./~.j 0 implies that discoveries on the individual prospects occur statistically independent (the grouping is then for convenience only). )~j -- 1 implies maximum co-variation between discoveries on the different prospects within the group. Each )~j refers only to hits or misses for the individual prospects, and says nothing about a potential correlation between the conditional volumes, given discoveries. Such dependence could also easily be included in the model. However, practically all the uncertainty is caused by the "hit or miss" phenomenon. Thus, for simplicity we have chosen not to incorporate volume dependence at the present stage. For all pairs of prospects groups where some "between-group" dependence is required, the user must also provide the common "between-group" )~factor. For all pairs of groups where such a )~-factor is not explicitly given, the co-variation is assumed to be 0. -
-
A North Sea case study
The prospects and their grouping As an example, let us consider data for a particular part of the North Sea. The data are real, but since this kind of information is regarded as highly sensitive
from a management perspective, the actual data area cannot be revealed. There are 159 identified prospects in the area, comprising from only one up to five different scenarios or prospect models. The probability of discovery varies from 0.4% to 75%, and the expected volumes (given discovery) range from 0.01 to 185 million TOE (tons of oil equivalent), with standard deviations mostly below 4-30%. The 159 prospects are grouped into 23 groups, where most of the groups contain 4-10 prospects. The ;.-factor for the different groups varies from 0 to 0.8, and is mostly in the 0.4-0.6 range. Prospects from different groups are regarded as statistically independent- there is no "between-groups" correlation.
Results Table 1 gives a summary of the distributions for the number of discoveries and for the oil- and gas volumes. In the table, the results are broken down into different resource classes, where (somewhat simplified) class A contains the most probable prospects while class E contains the prospects with very low probability. In the table, the columns labelled Ps0, Ps0 and P20 give the (upper) 80-, 50- and 20-% percentile, respectively. The last column, labelled % > 0, gives the probability of having at least one discovery from the given class within the area. Note from Tables 1 and 2 that the P80 and P20 for the Total are significantly closer to the mean than
Table 1 Oil discoveries and volumes Class
A-class B-class C-class D-class E-class Total
Volume (106 TOE)
Number of oil discoveries Mean
St. dv.
P80
Pso
P20
Mean
% > 0
St. dv.
P80
P50
P20
4.08 5.37 5.64 0.55 0.85
1.57 2.31 2.45 0.76 1.04
3 3 4 0 0
4 5 5 0 1
6 7 8 1 2
4.9 60.6 55.7 7.5 17.6
2.7 38.8 41.6 19.0 41.3
2.1 25.2 22.6 0.0 0.0
5.1 56.2 43.5 0.0 1.4
7.5 92.2 87.0 10.4 20.2
98 99 100 41 53
16.48
4.78
12
16
20
146.2
76.2
82.7
133.4
201.6
100
Table 2 Gas discoveries and volumes Class
Volume (106 TOE)
Number of oil discoveries Mean
St. dv.
Ps0
Ps0
P20
Mean
St. dv.
A-class B-class C-class D-class E-class
0.50 3.02 3.37 0.34 0.45
0.50 1.44 1.88 0.58 0.71
0 2 2 0 0
1 3 3 0 0
1 4 5 1 1
1.2 67.7 22.7 2.8 19.0
1.3 60.2 19.9 5.6 102.5
Total
7.68
2.73
5
8
10
113.3
121.3
% > 0 Ps0
Ps0
P20
0.0 9.3 5.2 0.0 0.0
0.0 44.4 17.1 0.0 0.0
2.6 133.0 38.0 6.1 8.8
50 97 96 29 34
34.2
84.8
171.0
100
86
E. Damsleth
Fig. 2. Distribution of the oil and gas volumes for no correlation (~. = 0, left), base-case correlation (centre) and maximum correlation (X = 1, right).
the sum of the Ps0s and P20s for the five classes. The central parts of Figs. 1 and 2 show the same distributions graphically, though only for the oil and gas total.
Effect of the normal assumption To investigate the effect of the assumption of a normal distribution for the prospect volumes (conditional, given discovery), we repeated the analysis above without volume uncertainty. Thus, for each prospect where a discovery was made, the volume was taken as/z - the mean of the normal distributions used above. For all practical purposes, the results were indistinguishable from the above.
Effect of the correlation The )~-values for the different prospect groups were defined by the geologists to the best of their ability. To study the importance of these correlations, and also to get some idea of the sensitivity of the results on the choice of )~s, we performed a sensitivity study.
First we put all the ~.s to zero, removing all the co-variation between the prospects. The resulting distributions for the total number of discoveries and for the volumes are shown in the left-hand parts of Figs. 1 and 2 respectively, where the base-case distributions are also shown, for comparison. The reduced variability for the )~ = 0 case is clear, though not dramatic. Secondly we put all the ~.-values to 1, giving maximum co-variation between the prospects within each group. The different groups were still treated as independent. The right hand parts of Figs. 1 and 2 show the results in this case. The increased uncertainty in the number of discoveries and in the total volumes is striking. Tables 3 (a and b) and 4 (a and b) summarize the results with regards to mean, standard deviation and P80 for the oil and gas respectively. As can be seen, the difference between the base-case and the "no-correlation"-case is not that dramatic, although making the erroneous )~ = 0 assumption will overestimate the resources by 10-20%, when quantified by the P80. This is not very surprising, as most of the )~-
A method f o r the statistical assessment o f total undiscovered resources in an area
87
Table 3a Effect of the correlation on the number of oil discoveries Class
Mean No
A-class B-class C-class D-class E-class Total
St. dv. Base
Full
No
Base
Full
No
Base
A-class B-class C-class D-class E-class Total
No
Base
Full
4.08 5.37 5.64 0.55 0.85
4.10 5.40 5.56 0.55 0.86
1.13 1.89 2.17 0.72 0.91
1.57 2.31 2.45 0.76 1.04
2.51 3.62 3.84 0.90 1.51
3 4 4 0 0
3 3 4 0 0
0 2 2 0 0
100 100 100 43 59
98 99 100 41 53
78 90 97 36 37
16.55
16.48
16.46
3.32
4.78
8.54
14
12
9
100
100
100
ASCII-file, which can easily be input into any statistical analysis or report generating software.
Effect of the correlation on the oil volume (10 6 TOE) Mean No
Full
4.08 5.43 5.63 0.55 0.87
Table 3b
Class
%>0
P80
Base
4.9 61.6 55.6 7.5 18.0
4.9 60.6 55.7 7.5 17.6
147.6
146.2
St. dv.
P80
No
Base Full
No
2.4 36.4 40.5 16.9 41.9
2.7 38.8 41.6 19.0 41.3
3.7 2.5 2.1 55.4 28.4 25.2 46.7 23.9 22.6 25.3 0.0 0.0 45.6 0.0 0.0
0.0 3.7 15.2 0.0 0.0
147.3 69.7 76.2
111.0 90.4 82.7
49.8
Full 4.9 61.0 55.4 8.1 17.9
Base Full
values used in the model were of only moderate size. On the other hand, as can be seen from the tables, overestimating the correlations can lead to a dramatic increase in the variability, and thus reduction in P80. In this particular example, P80 is reduced by around 40%, compared to the base-case. Thus, a correct assessment of the co-variation between the different prospects is very important in the process of estimation and quantification of the total resources. In this particular example, a 50% change in the resources, as quantified by the P80, can easily be obtained just by playing around with the correlations!
Implementation The input data are stored in the Norsk Hydro resource data base, and updated when needed, at least annually. The actual statistical computations are performed by Monte Carlo simulation, in which a large number of potential outcomes are simulated and the necessary statistics calculated. The results presented here are all based on 10000 repetitions in the Monte Carlo simulation. A fairly large number of repetitions is required to ensure that even prospects with a very small probability of discovery will be realized at least a few times. Standard report-generating tools are used in the presentation of the results. All the results from the Monte Carlo simulations are stored on a standard
Conclusions Norsk Hydro has developed and implemented a technique to describe efficiently the potential statistical dependence between undrilled prospects, and compute probability distributions for the total number of discoveries and for the total resources. It is very important to take the statistical dependency correctly into account in the evaluations of the total resources. Ignoring the dependence will give too little spread and thus too optimistic values for e.g. P80, while simple addition of the P80 values for each prospect will give a much too pessimistic result. The method and the programs are used routinely in Norsk Hydro's resource evaluation and reporting.
Appendix 1" Generating correlated discoveries
Mathematical setting A prospect group contains n prospects. The chance factor (probability of discovery) for the ith prospect is given by pi, i = 1 . . . . . n. The co-variation, or correlation, between two prospects in the group is closely related to the conditional probability of making a discovery in one prospect, given that a discovery is made in the other. Let F/--
1 if discovery in prospect i 0 if not .
i -. . 1,.
.
n
(1)
Usually the co-variation will be positive, so that Pr(Fi = IIFj = 1) > Pi
>
Pr(Fi = IIFj - - 0 )
(2)
The correlation coefficients between discoveries in prospects i and j are given by" er(Fi = 1 A Fj = 1 ) - Pi Pj Pij =
v/Pi(1 - P i ) P j (1 - P2)
_- Pr(Fi - 0A Fj - 0 ) - (1 - pi)(1 - P2) v/Pi(1 - p i ) p j ( 1 - p j )
(3)
E. Damsleth
88 Table 4a Effect of the correlation on the number of gasdiscoveries Class
Mean
St. dv.
P80
% > 0
No
Base
Full
No
Base
Full
No
Base
Full
A-class B-class C-class D-class E-class
0.50 3.03 3.38 0.33 0.47
0.50 3.02 3.37 0.34 0.45
0.49 3.00 3.36 0.33 0.47
0.50 1.44 1.69 0.56 0.67
0.50 1.44 1.88 0.58 0.71
0.50 1.61 2.98 0.62 1.00
0 2 2 0 0
0 2 2 0 0
0 2 1 0 0
50 97 97 29 38
50 97 96 29 34
49 96 83 26 26
Total
7.71
7.68
7.65
2.45
2.73
4.35
6
5
4
100
100
99
Table 4b
No A-class 1.2 B-class 68.2 C-class 23.0 D-class 2.7 E-class 19.1 Total
114.2
St. dv. Base 1.2 67.7 22.7 2.8 19.0 113.3
Full
No
P80 Base
Full
1.2 1.3 1.3 1.3 68.0 60.2 60.2 60.3 23.2 18.5 19.9 31.3 2.6 5.6 5.6 5.7 19.5 101.8 102.5 105.4 114.5
119.4 121.3
No
Base
Full
(1) Generate U1, U2 . . . . . U~ as standard multinormally distributed variables with correlation matrix C given by"
Effect of the correlation on the gas volume (106 TOE) Mean
No
Base Full
0.0 9.4 6.6 0.0 0.0
0.0 9.3 5.2 0.0 0.0
0.0 9.5 0.8 0.0 0.0
131.0 39.3 34.2 22.2
Throughout, we use the notation Pr(event) to denote the probability of event. Ideally, the co-variation should be the same between all the prospects in the prospect group. Unfortunately, this is not easy to achieve in practice. In a situation with standard multi-normally distributed variables, correlation and the correlation coefficient are well defined concepts, which do not depend on the means and standard deviations in the marginal distributions. In a binary situation, this is no longer the case. In the Gaussian case the simultaneous distribution for all the n variables will be uniquely given by the means, standard deviations and the correlation matrix. In the binary case, on the other hand, the simultaneous distribution is only uniquely given when Pr(F1 = fl A F2 = f2 A . - . ~ F~ = f~) is specified for all the 2 ~ possible combinations of fi - 0, 1 and i = 1 . . . . . n. In addition, the correlation coefficient between two binary variables can no longer vary in the full [ - 1 , 1] interval. The actual region of variation depends on the marginal probabilities of discovery. The maximum correlation between two prospects is given by:
maxp - min (~/pi(1- pj) ~ (1- pi)pj ) (1 - pi)pj' pi(1- pj)
(4)
Suggested solution Our solution is a variation of a technique proposed by Emrich and Piedmonte (1991)"
c
-
1
)~ ...
.
.
)~
.
.
9
(5)
9
)~ )~ ...
1
where ~, is the required co-variation factor. To ensure that C is non-negative definite, - 1/ (n - 1) _< ~, < 1. (2) Let
Fi-
1
0
i f ~ ( U i ) <_Pi otherwise .
i.
.
1, .
"
//
(6)
where ~(.) is the cumulative standard normal distribution function. This gives all the prospects the correct chance factor, and the prospects will be "correlated" in such a way that the correlation coefficient between any two prospects, say i and j , will be approximately given by" corr(Fi,
Fj) ~ ~.
and max corr(Fi,
max corr(Fi,
Fj),
)~ >_ 0
(7)
Fj) is given by (4).
Finding the ~-factors To find the co-variation coefficient ~. within a group one can start by choosing two "typical" prospects within the group, say i and j. Let pi and pj be the marginal probabilities of discovery for the two prospects. Further, denote the conditional probabilities for the same two prospects by Pll~ and P010 respectively, where: Pill -- Pr(discovery in prospect i I discovery in prospect j ) P010 - Pr(no discovery in prospect i I no discovery in prospect j ) Using standard formulae: pl _< Pljl _< min (P~~., 1)
and 1 - pl _< polo -< min
1-
pj'
(8)
89
A method for the statistical assessment of total undiscovered resources in an area
Hopefully, the user can quantify either P l l l or P010. The correlation between the two prospects is then given by:
trix C which can be block-divided as follows:
C -P _
V/
pipj (p,,l _1 ) (1 - pi)(1 - p j ) pi
/(1-pi)(1-pj)(piPj
lP~176
1)
(10)
and the ~.-factor for the two prospects can by found from ~. = P/Pmax, where p is given by (10) and Pmax by (4). The same value for ~. will then be used for all pairs of projects within the group. Example: Consider two prospects, 1 and 2, with probability of discovery pl - 0.1 and p2 - - 0.2. From (9), 0.1 < Pill < 0.5. The user specifies Pill = 0.3, so that a discovery in prospect 2 triples the probability of discovery in prospect 1. Using (10), these figures give p = 0.33, and (4) gives / ) m a x = 0.67. Thus, ~. = 0.33/0.67 = 0.5.
Illustration To illustrate the above procedure, and to demonstrate the effect of the ~.-factor, let us look at a group with five prospects, with probabilities of discovery equal to 0.1, 0.3, 0.5, 0.7 and 0.9, respectively. The maximum correlations between discoveries, using (4), is in this case given in the correlation matrix in Table A1. For )~ = 0, 0.25, 0.5, 0.75 and 1.0, Table A2 shows: (a) Probabilities of discovery for the five prospects. First unconditional, then conditional, given that 0, 1, 2, 3 and 4 of the other prospects resulted in a discovery. (b) The correlation matrix for (F1, F2, F3, F4, Fs). The tables are computed from Monte Carlo simulations with 5000 repetitions.
Simultaneous generation of discoveries for several groups The technique presented above can also be used to generate discoveries/non-discoveries for different prospect groups, when there may be co-variation both within and between the groups. The approach is then: (1) Generate U1, U2 . . . . . UN as standard multinormally distributed variables with correlation maTable A 1
Cll
C12
...
C1M
C21
C22
...
C2M
.
.
.
.
CM1
CM2
...
CMM
(11)
Every CjK, j ~ k represents a nj x nk correlation matrix describing the co-variation between discovery in groups j and k. Cjk is given by:
Cjk --
i~,j k
~jk
9 9 9 )~j k
~.jk
~.jk
...
)~jk
9
i
~176
~
~jk
~,jk
9 9 9 ~jk
j ~ k
(12)
where i~,jk is the required co-variation coefficient between (any) prospects in groups j and k. If prospects in the two groups can be regarded as independent, so that there is no co-variation, Cjk contains only zeros. Every C j j, j = 1 . . . . . M represents the n j • n j correlation matrix which describes the co-variation within group no. j, as described earlier. Cjj is given by:
Cjj
--
1
~.j
...
~,j
~j
1
...
/~,j
.
.
..
9
~j
~,j
(13)
. 9
. ..
.
1
where /~.j is the required co-variation coefficient. If there is no c-variation within the group, then Cjj is the identity matrix with ones on the main diagonal and zeros elsewhere. (2) As before, let Fi-
1 0
if~(Ui)
.
.
i.
1, .
. n
(14)
As before, q~(.) denotes the standard normal cumulative distribution. Then all the prospects will have the (marginally) correct chance factor, and the correlation coefficient between two prospects, say i and j, will be approximately given by:
corr(Fi, Fj)
~kl" max corr(Fi, Fj), ~.kl >_0, k r ~.~. max corr(Fi, Fj), ~.~ > 0 , k = l
(15)
with max corr(F/, Fj) given by (4). The ~.lk-values for prospects from different groups can be found in the same way as the within-group X-values.
Maximum correlations between discoveries 1
0.51 1
0.33 0.66 1
0.22 0.43 0.66 1
0.11 0.22 0.33 0.51 1
Simultaneous generation of volumes The volumes for each prospect, given discovery, can be related to one of several possible scenarios or prospect-models, where the different scenarios cover
E. Damsleth
90
Table A2 Conditional probabilities and correlations for different X-values E nr.
cond.
0 disc.
1 disc.
2 disc.
3 disc.
4 disc.
1
2
~. = 0 . 0 0 1 2 3 4 5
0.10 0.30 0.50 0.70 0.90
0.05 0.19 0.49 0.66 0.89
0.09 0.32 0.51 0.70 0.90
0.11 0.31 0.50 0.69 0.90
0.10 0.28 0.47 0.67 0.88
0.16 0.38 0.56 0.80 0.91
1.00
0.02 1.00
)~ = 0.25 1 2 3 4 5
0.10 0.30 0.50 0.70 0.90
0.03 0.09 0.23 0.42 0.78
0.03 0.18 0.37 0.62 0.89
0.06 0.26 0.49 0.73 0.93
0.13 0.40 0.65 0.83 0.97
0.18 0.50 0.72 0.88 0.97
1.00
X = 0.50 1 2 3 4 5
0.10 0.30 0.50 0.70 0.90
0.01 0.03 0.09 0.29 0.72
0.01 0.09 0.26 0.55 0.89
0.04 0.21 0.49 0.77 0.97
0.11 0.45 0.75 0.91 0.97
0.26 0.70 0.88 0.99 1.00
1.00
)~ = 0.75 1 2 3 4 5
0.10 0.30 0.50 0.70 0.90
0.00 0.00 0.03 0.14 0.66
0.00 0.03 0.15 0.53 0.93
0.01 0.14 0.50 0.87 0.99
0.06 0.47 0.85 0.98 1.00
0.32 0.87 0.98 1.00 1.00
1.00
0.10 0.30 0.50 0.70 0.90
0.00 0.00 0.00 0.00 0.67
0.00 0.00 0.00 0.49 1.00
0.00 0.00 0.51 1.00 1.00
0.00 0.50 1.00 1.00 1.00
0.35 1.00 1.00 1.00 1.00
1.00
3
4
5
0.03 -0.02
0.02 0.00 -0.01 1.00
0.00 0.00 -0.01 0.00
1.00
1.00 0.12 1.00
0.12 0.16
1.00
0.25 1.00
0.21 0.33
1.00
0.10 0.13 0.16 1.00
0.06 0.11 0.11 0.14 1.00
0.18 0.28 0.31 1.00
0.10 0.17 0.24 0.26
1.00 0.40 1.00
0.30 0.48 1.00
0.21 0.38 0.49 1.00
0.11 0.22 0.31 0.41
1.00
X = 1.00
1 2 3 4 5
different degrees of filling, whether the discovery is mainly oil or mainly gas, etc. Each scenario has an associated probability. The scenarios within one prospect are mutually exclusive, and there is no covariation between scenarios in different prospects. For each prospect and for each scenario within the prospect, the volume (in TOE) is assumed to be normally distributed with a given mean and standard deviation. In the present model, all the volumes are regarded as statistically independent. There are no technical problems in introducing dependency between the volumes as well, but we have decided to keep the model as simple as possible at the present stage. For a prospect where a "discovery" has been made as a result of the algorithm for generating discoveries we described above, the volumes are generated according to the following procedure:
E. DAMSLETH
Norsk Hydro, P.O.Box 200, 1321 Stabekk, Norway
0.51 1.00
0.33 0.66 1.00
0.22 0.43 0.66
1.00
0.11 0.22 0.33 0.51
1.00
(1) Based on the conditional probability distribution for the different scenarios, given that a discovery is made, one scenario is chosen. It is recorded whether the scenario describes an oil- or a gas-case. (2) The volume is drawn from the normal distribution with the parameters according to the specific scenario.
References Emrich, L.J. and Piedmonte, M.R., 1991. A method for generating high-dimensional multivariate binary variates. Am. Statistician, 45: 302-304.
Harbaugh, J.W., 1984. Quantitative estimation of petroleum prospect outcome probabilitieswan overview of procedures. Mar. Pet. Geol., 1: 298-312.
Nederlof, M.H., 1994. Comparing probabilistic predictions with outcomes in petroleum exploration prospect appraisal. Nonrenewable Resources. 3:183-189.
91
The Norwegian Petroleum Directorate's assessment of the undiscovered resources of the Norwegian Continental shelf--- background and methods Harald Brekke and Jan Erik Kalheim
In February 1993 the Norwegian Petroleum Directorate presented their recent assessment of the undiscovered hydrocarbon resources of the Norwegian shelf. The assessment is based on play analysis. Much effort was put into defining the different plays, in mapping the geographical limits of each play, and compiling data on their critical volume parameters. Each volume parameter was defined as a size distribution with a minimum, expected and maximum value. These parameters were used as input in the computation of the total resources in each play. The basic assumption behind this computation is that the relative field size distribution for the finite number of accumulations in each play may be fitted to a lognormal distribution. The estimate of the total undiscovered resources in each play is given as a probability distribution and is presented as a set of a low (P95), a mean and a high (Ps) estimate. The estimated distribution for each play was aggregated by histogram summation to give the total probability distribution for the undiscovered resources of the whole of the Norwegian shelf. The computations were performed with the computer program FASPUM which was developed by R.A. Crovelli and H. Balay at the United States Geological Survey. Examples from selected plays are given. A major concern in the analysis was the calibration of input data and the risk factor associated with each parameter. We found the number of drillable prospects within each play to be the most difficult parameter to assess. At the same time, this parameter has a great impact on the final result. The risk analysis on both play level and prospect level was done according to the NPD internal standard procedure. The plays fall into two different categories; those confirmed by discoveries and those that are conceptual and not yet confirmed by discoveries. The risk associated with the non-confirmed plays is difficult to assess because it requires a subjective calibration against the risk-level in the confirmed plays. The calibration problems and the risk analysis are discussed.
Introduction In February 1993 the Norwegian Petroleum Directorate (NPD) released a report on its revised and current assessment of the total recoverable petroleum resources on the whole of the Norwegian shelf, including both opened and unopened acreage (Fig. 1) (NPD, 1993). In addition to the annual update of the discovered resources, the report includes the new 1992 assessment of the undiscovered resources to replace the 1988 assessment. The report concludes that the total recoverable resources are distributed as follows: Discovered recoverable resources: Potential for improved oil recovery: Undiscovered recoverable resources"
5.6 btoe 0.5 btoe 3.7 btoe
Total
9.8 btoe
The geographical distribution of these resources is shown in Fig. 2. The 1993 report presents in
some detail the results of the assessment of the undiscovered resources (Fig. 3), but is very brief on how NPD arrived at these numbers. The aim of the present paper is to describe the methods and procedures behind the estimates.
Basic decisions
Appraisal method A number of methods have been applied to estimate undiscovered petroleum resources through the years. An excellent introduction to these is given by Dolton (1984) who groups them into five major classes: (1) extrapolation of discovery rates; (2) areal and volumetric yield methods; (3) geochemical material-balance analyses; (4) prospect and play analyses; and (5) direct assessments. A basic condition for the NPD has been that, within the allotted time frame, as much as possible of the geological and geophysical data from the Norwegian shelf should be incorporated in the assessment
Quantification and Prediction of Petroleum Resources edited by A.G. Dor6 and R. Sinding-Larsen. NPF Special Publication 6, pp. 91-103, Elsevier, Amsterdam. 9 Norwegian Petroleum Society (NPF), 1996.
92
H. Brekke and J.E. Kalheim
Fig. 1. Simplified map showing the areal extent of sedimentary rocks and the regulatory status of the Norwegian shelf acreage.
The NPD's assessment of the undiscovered resources of the Norwegian Continental shelf
Fig. 2. The geographical distribution of the petroleum resources on the Norwegian Shelf.
93
H. Brekke and J.E. Kalheim
94
Fig. 3. The current estimates of the undiscovered petroleum resources on the Norwegian shelf.
of the undiscovered resources. Therefore an appraisal method was required that would allow the use of a variety and wealth of data, but could also be applied in areas with a relative lack of data. The NPD also wanted a geological framework that would make the estimates easily updatable and make them fit into the structure of the NPD resource database. To achieve these goals, NPD found an approach based on play analysis to be the best alternative. The NPD has adopted the following play definition: "A play is defined by a set of geological factors and attributes that, when combined, give the conditions necessary for the accumulation of hydrocarbons. All prospects and discoveries within a play share the same set of necessary attributes and are hence distinguishable from prospects and discoveries belonging to other plays." The geological attributes required to define a play are the following: (1) existence of a mature source rock in the possible drainage area(s) of the traps;
(2) existence of sealed traps formed prior to the end of hydrocarbon migration; (3) existence of reservoir rock. Based on these principles the first task for the NPD assessors was to define and describe the possible plays of each basin in terms of the specific nature of the three attributes (i.e. a specification of the source rock formation and migration area, reservoir rock formation, and type of trapping mechanism). Each play was then assumed to constitute a unique set of attributes and could thus be regarded as a statistically separate entity. The resources in each play were then calculated and the total resources in an area or basin was given by the aggregation of the resources in all the plays.
Statistical model A play comprises a population of deposits with a given size probability distribution. In order to calculate the undiscovered resources in a play that comprises at least some unmapped prospects, one
The NPD's assessment o f the undiscovered resources o f the Norwegian Continental shelf
must assume a realistic deposit size distribution for the whole play. Following work by Arps and Roberts (1958), the shape of the lognormal distribution has been a popular choice to fit such size distributions. However, it has been pointed out that this assumption may not be correct in all details and in all cases (see Kaufman, 1993, for a discussion). Nevertheless, Crovelli (1986, 1987) has shown that within the range of variations encountered at the level of the individual basin, a fit to a lognormal curve is a reasonable assumption. Deviations are seen only at the extreme tails of the curve. Besides being a good approximation for the shape of the deposit size distribution, the lognormal distribution also has some favourable theoretical properties (Crovelli, 1984):
SINGLE
95
- the product of many independent random variables is a lognormal distribution; - t h e product of independent lognormal random variables is itself lognormal; the shape of a lognormal distribution is geometrically easy to construct. The NPD decided to adopt the assumption that the deposit size distribution in the individual play may be approximated by the shape of a lognormal distribution (Fig. 4). This model requires that the plays be defined and constrained so well that the size distribution of the prospects in each play consists of a single population. If, for instance, a play includes several trap types, there is a risk that the deposit size distribution may consist of a mixed population giving unpredictable shapes to the curves (Fig. 4). -
POPULATION
10C I0000-
~9 80. 1000~
60.
.,,
9 100-
>
~ 4o_
10-
i
20_
!
|
|
10
100
1000
a)
i
99.999
10000
VOLUME
b)
9i0
510 liO ~
.1
0 i
Cumulative Probability %
MIXED POPULATION
IOOOO~
8O 1000-
ioo
9 100-
>
"~ 40
10-
2O
1
c)
10
100
VOLUME
1000
10000
'i
99"19 919 90
d)
510
ll0
I 0I 1 .1
Cumulative Probability %
Fig. 4. (a) Cumulative lognormal distribution plot, logarithmic volume scale (x-axis) vs. normal probability scale (y-axis). (b) Same distribution as in (a), plotted on lognormal probability paper. Gaussian probability scale on x-axismakes the distribution plot as a straight line. (c) and (d) show the effect of a mixed population on the shape of a lognormal distribution curve plotted as in (a) and (b).
96
H. Brekke and J.E. Kalheim
Definition and mapping of plays The NPD approach was to try to avoid defining play models with mixed size populations and at the same time keep the total number of plays at a minimum. A total of 44 plays were defined for the whole shelf area, of which 23 have been confirmed by at least one discovery and 21 are still unconfirmed (Table 1). The plays were defined on the basis of stratigraphic columns and parameter maps showing the geographical distribution and quality of potential reservoirs, mature source rock, migration routes and structural and depositional trends. The stratigraphic columns and parameter maps were based on borehole data and seismic interpretation. Fig. 5 shows the stratigraphic positions of the reservoirs of the defined plays. Play Table 1 Number of plays on the Norwegian shelf Confirmed
Unconfirmed
Total
North Sea Norwegian Sea Barents Sea
12 7 4
1 12 8
13 19 12
Total
23
21
44
0
40
80
120
160 Km
summary maps (White, 1988) were constructed from the parameter distribution maps: the play area was defined by the area in which the overlap of all parameter maps shows favourable conditions for the play (Fig. 6). Along with the maps, each play was systematically described in terms of critical geological factors, geography and references (Table 2). Ideally, a play should be defined such that all the prospects and discoveries belonging to the play constitute a geologically homogeneous group (see White, 1980). This is necessary both to ensure a functional and sensible nomenclature of plays and to avoid operating with mixed deposit size populations. However, to decide whether geological differences between prospects are so subtle that they may be ignored or so significant that the prospects must be split into two or more plays, is not a straightforward exercise. Among assessors one finds both splitters and lumpers, as in other branches of geology. Furthermore, there is a general trend that in the early stages of exploration an increase in the data available leads to a need for splitting a play into two or more plays. This reflects the fact that when facing a frontier area with little data, one is constrained to construct simple play models that must be applied to very large areas (for example
0
80
160
:;)40 Km
Fig. 5. (a) Play area of the mature, confirmed play NJI,JM-I in the northern North Sea. This play consists of a reservoir of Lower to Middle Jurassic sandstones in traps made by rotated fault blocks. Well known fields in this play are the Statfjord, Gullfaks and Oseberg Fields. (b) Play area of the unconfirmed, frontier play BCL-1 in the Barents Sea. This play consists of a reservoir of Lower Carboniferous clastics in traps consisting of rotated fault blocks. No discoveries have yet been made.
97
The NPD's assessment of the undiscovered resources of the Norwegian Continental shelf
Fig. 6. Stratigraphic columns showing the stratigraphic positions of the reservoirs of the plays defined on the Norwegian shelf, both confirmed and unconfirmed. Examples of representative discoveries and fields are indicated for the confirmed plays
Table 2 Two examples of play descriptions Name
Reservoir facies
Seal level
BCL-1 Continental/ Lower fluvial/ Carbonif. shallow marine sandstone
NJL, JM-1
Fluvial/ deltaic/ shallow marine sandstones
Lunde Fm. Statfjord Fm. Cook Fm. Brent Grp
Trap
facies
level
Marine shales
Carbonif.
Source
Source
level
area
Rotated Devonian/ fault Carbonif. blocks + stratigraphic elements
Local basins
Play area
Reference
firmed Finnmaek Platf. No Bjarmeland Platf. Loppa High Sentralbanken H. Gardarbanken H. SCrkapp Basin Olga Basin Kong Karls Platf.
Viking Grp Viking Graben North Sea Shallow/ Dunlin Grp Rotated Middle Sogn Graben between open Viking Grp fault Jurassic Tampen Spur 60~ and 62~ marine blocks + shales stratigraphic elements
Fig. 6b). It is obvious that such early frontier play models will prove to be more complicated as information from drilling increases. However, it is impossible to know "up front" in which direction models will be changed through time, and one has to rely upon the early "lump" models to have an idea of the resource potential in frontier areas. This means that in many
Con-
Yes
7128/6-1 shallow drilling 1988
Gullfaks Statfjord Oseberg Snorre Veslefrikk Brage
cases, where little data is available, the NPD estimates will undoubtedly be based on simplified models with mixed deposit size populations. To try to avoid this by splitting a large frontier model into every conceivable model, based on scanty data or mere mind constructions, will only lead to other problems. Basing an estimate on the aggregation of many models
H. Brekke and J.E. Kalheim
98
leads to a narrower uncertainty range in the estimate (given that the new "split" models are regarded as independent entities) and gives the impression that the risk is reduced. To counteract this, one has to estimate the degree of interdependence between all of the "split" models. Because of the lack of data, this will in most cases be mere guesswork which will probably be even more meaningless than the results from a large "lump" model with a mixed population. So one is left with the compromise between accepting "lump" models and mixed populations or a number of more specific models with highly speculative descriptions and estimates of interdependencies. The problem of a mixed population of a "lump" model is at least a simple concept to keep in mind, whereas all the guesswork in the details of the "split" models is much less transparent when the time comes to take decisions. The NPD therefore decided to base their estimates in frontier areas on "lump" models.
Table 3 Input data for FASPUM computation program m risk parameters and volume parameters Evaluator Date Evaluated
Net rock volume The product of the parameters "area of closure" and "reservoir thickness" (Table 3) gives the net reservoir rock volume of the prospects. The NPD assessors found that applying values for the reservoir thickness or the vertical closure does not give a correct estimate of the reservoir volumes because
Probabilityof Favorable or Prensent
Attribute
Comments
HydrocarbonSource Timing Migration Potential ReservoirFacies
,~ _~" ~ ~:
Marginal Play Probability TrappingMechanism Effective Porosity(>3%)
~ 8.~ A- ~
HydrocarbonAccumulation ConditionalDeposit Probability ReservoirLithology Hydrocad:)on
~
'~,ractiles .Attribute 9 ~ , . .
Sand Carbonate Ga~ Oil
Sand
Probability of equal to or greater than 95 75 50 25
5
0
Area of Closure(Km^2)
Computation As a computation tool the NPD chose the computer program "Fast Appraisal System for Petroleum Universal Metric" (FASPUM) developed by the United States Geological Survey (see Crovelli and Balay, 1988). The reasons for this choice were that the system is based on a lognormal distribution approximation, the computations are quick and the system is user friendly and may be applied to plays of very differing data base levels. FASPUM computes the total oil and gas resources in each play based on certain critical geological volume parameters (see Table 3). Each parameter is entered as seven fractiles of a probability distribution in order to reflect the range of uncertainty and variety of the parameter values. Much effort was put into picking realistic values and range of uncertainty for the different parameters in each play. In general the range of uncertainty was increased for plays with a restricted data base relative to mature plays with a substantial data base. The two most difficult parameters to assess were the size distribution of the traps (net rock volume) and the number of remaining drillable prospects. Details on the methods for selection of the volume parameters are given below.
Play Name
ReservoirThickness /vertical closure(meters)
Ratio of area and net res. ~olullle
EffectivePorosity% Trap Fill(%) ReservoirDepth(m) HC Saturation(%) ! 1
No.of drillable prospects (a playcharacteristic)
the three dimensional shape of the trap is then not taken into consideration. An alternative method was developed in which the reservoir thickness/vertical closure is replaced by the value of the ratio between the net reservoir rock volume and the area of closure. This ratio is characteristic of the separate trap types. The method is based on the assumption that both the area of closure and the net rock volume of the deposits in a play may be fitted to a lognormal distribution curve. Lognormal plots of the area of closure and net rock volume were constructed for all plays by using lognormal probability paper (Fig. 7). The value of the largest known prospect or discovery in the play was plotted at 0.1% cumulative probability and values matching a reasonable economic minimum size were plotted at 95% cumulative probability. A reasonable economic minimum size was set to 1590.106 m 3 net rock volume. This procedure allows for the possibility (though with a low probability) of making a discovery that is larger than the presently known discovery or prospect mapped/assumed in each play. At the same time it allows for a tail of numerous, subeconomic and uneconomic deposits. The main problem in this plotting method is to de-
99
The NPD's assessment of the undiscovered resources of the Norwegian Continental shelf
PLAY
- NJL,
PLAY
JM-I
- BCL-
1
b)
a)
'1
STATFJORD
u~ 4
IO,92 -
0,1
1
5 10
25
50
75
90 95
99
99,9
FREQUENCY
0,1 0,1
1
5 10
25
50
75
90 95
99
99,9
FREOUENCY
Fig. 7. Lognormal distribution plots of "area of closure" and "net rock volume" for traps in two plays: (a) NJL,JM-1 in the northern North Sea (see Fig. 6a)" and (b) BCL-1 in the Barents Sea (see Fig. 6b).
cide the minimum size and at which cumulative probability to plot it. The minimum sizes were set according to rough estimates of minimum economic sizes in the different regions of the shelf. However, in many cases this minimum size (plotted at 95% percentile) had to be adjusted so that the curves could be calibrated to give reasonable median values (50% percentile). "Reasonable median values" were based on the available statistics on discoveries taking into consideration the decline in mean deposit sizes with time (plotted at less than 50%). This implied relatively lower mean sizes for the remaining deposits in mature plays with many discoveries than in immature plays with few discoveries and in unconfirmed plays with no discoveries. In immature and unconfirmed plays analogs were used as a guideline where possible. The area and volume plots (e.g. Fig. 7) were used to find the values for the seven percentiles to be entered in the FASPUM table (Table 3) under the parameters "Area of closure" (derived directly) and "Reservoir thickness" (as the rock volume vs. area of closure ratio).
Effective porosity The probability distribution for reservoir porosity was based on well data or analog models. Lateral variations were predicted on the basis of depth maps and facies maps.
Trap fill This parameter was difficult to set. It is a fact that, in many cases traps in areas of late uplift, in areas with a marginally mature source rock, or in areas with long migration distances are not filled to spill point. Experience has shown us that in the Barents Sea most traps are underfilled, and the NPD assessors took this into consideration there and in all areas with a history of late uplift. However, in most other cases there is no systematic or statistically significant empirical basis to predict this parameter, and the portion of underfilled traps was estimated according to a subjective estimate of the effect of assumed degree of source maturity and migration distances.
H. Brekke and J.E. Kalheim
100
Reservoir depth
analogs. The less problematic of these plays seemed to be those with fault dependent traps, because analogs were easy to find and there was generally a reliable database for predicting lateral variations due to variations in faulting intensity. The prospect density of plays with stratigraphic traps, however, was not easy to predict. Such traps are very difficult to map, and even in proven plays it is not easy to say how many prospects may remain. In addition, it was speculated that different trap types perhaps have their own, characteristic density range so that there may be a systematic difference between structural and stratigraphic traps. This possibility would render the use of prospect densities for structural traps as a model for stratigraphic traps highly speculative. Nevertheless, the NPD assessors assumed, based on experience, that as a rule stratigraphic plays have a lower prospect density than structural plays.
The frequency distribution of the deposits was based on depth maps of the relevant reservoir levels. This information and the given "oil floor depth" is used by FASPUM to compute the portion of oil that may have been cracked to gas.
Hydrocarbon saturation The range of HC saturation was based on well data and general experience with the different types of reservoirs.
Number of drillable prospects The probability distribution of the remaining, drillable prospects was based on an assumed range in prospect density multiplied by the total area of each play. The main effort here was put into deciding the prospect densities. These were based on locally derived empirical data and analogs. In the mature models (like NJL,JM-1, Table 2, Fig. 6a) the prospect density was based on the proven density of mapped prospects, discoveries and dry wildcats within the most thoroughly explored subareas of the play. In plays with no or very few mapped prospects, estimates of prospect densities had to be based on
PVT parameters In addition to the volume parameters FASPUM requires some PVT data as input. These parameters are assumed to be depth dependent. The NPD assessors thought it quite pointless to try to predict how these parameters vary laterally over large areas, and it was decided to use linear depth functions and standard values common to all plays (Table 4).
Table 4 Input data for FASPUM computation program - - reservoir parameters
Geological variables Four types of mathematical functions (1) Zones linear function:
A 9depth + B Maximum of 4 zones with 3 transition depths (m) A 9exp(B 9 depth) A 9depth ** B A 9ln(B 9 depth)
(2) Exponential function: (3) Power function: (4) Logarithmic function:
For each of the five geological variables below, select one type of function and assign values for the parameters A and B Pe: T: Rs: Bo: Z:
Original reservoir pressure (bar) Reservoir temperature (K) Gas-oil ratio (m3/t) Oil formation volume factor (no units) Gas compressibility factor (no units)
Variable Pe T Rs Bo Z
Function
Parameters
Linear Linear Linear Linear Linear
A 0.1 0.035 0.079 0 0.0000992
Oil floor depth (m): Oil recovery factor (%): Gas recovery factor (%):
4800 40 75
B 0 277 0 1.25 0.67
D
A
B
D
A
B
D
A
B
101
The NPD's assessment of the undiscovered resources of the Norwegian Continental shelf
Risk analysis The FASPUM approach is based on a conditional probability model where the risk analysis is split into two levels; the play level and the prospect level. In the FASPUM scheme the chance of success at the play level and prospect level are termed "the marginal play probability" and "conditional deposit probability" respectively (Table 3). (NPD prefers to use the terms "play chance" and "prospect chance"). At the play level one assesses the risk of the regional parameters that determine whether the play is favourable (i.e. will lead to a discovery). At the prospect level one assesses the parameters that determine the chance of success in each single prospect if the play chance equals 1 (i.e. if all regional parameters are favourable). The chance of success for each of the parameters on both play level and prospect level is assigned a value between 1 and 0, and the final chance of success, Pf, equals the product of all the risk parameter values. In the FASPUM scheme the critical parameters at play level are identified as the chance for the existence of mature source rock, migration, reservoir rock and favourable timing of trap formation; the critical parameters at prospect level are identified as the chance of effective trapping, effective porosity and hydrocarbon accumulation. The NPD had to modify this scheme to make it fit with the well established NPD internal procedure for prospect risk analysis. The NPD prospect risk analysis is based on four main risk parameters, P1 to P4, some of which are split into two sub-parameters: P(reservoir) = P(reservoir rock) x P(porosity) P1 Pla Plb P(trap) = P(mappability) x P(trap quality) P2 P2a P2b P(source & accumulation) = P3 P(retention) P4
P(quality & volume) x P(migration) P3a P3b
The NPD standard defines a set of status-categories for each parameters, with a corresponding range interval for a likely chance of success of the parameter. The task of the assessor is then to identify the correct status-category for each 15arameter and subjectively pick the most likely chance of success within the stated range. This process is always checked by a group of experienced assessors to ensure a repeatable result. The product of all parameters then equals the chance of success, Pf (i.e. the chance of finding movable hydrocarbons, not necessarily commercial volumes): P f = P1 x P2 x P3 x P4
Pm (COMMON)
1)
Pf = Pla x Plb x P2a x P2b x P3a x P3b x P4
//
Pp ( I N D I V I D U A L )
2)
Pf
= P l a x P2b x P3a x P l b x P 2 a x P 3 b x P 4
Pm
Pp
Pm = 1
CONFIRMED
Pm < 1
UNCONFIRMED
Fig. 8. Structure of the risk assessment in the NPD. The NDP standard for assessing the chance of success of a prospect is given in 1). Adjusting this standard to fit the assessment of the chance of success in play analysis is done by separating the regional parameters that are common to all prospects in a play model, Pm, from the parameters that vary between the individual prospects, Pp. Pm is the chance of success at play model level ("play chance"), Pp is the chance of success at prospect level ("prospect chance"), and Pf is the final chance of success for making discoveries among all the prospects in a play. Plays may be classified into two groups: those that are confirmed by discoveries and hence Pm = 1, and those that are not confirmed and hence 0 < Pm < 1.
In this scheme an assessor who estimates the chance of success for a single prospect, estimates the risk factors common to the play as a whole, together with the factors that vary between the different prospects. To modify this scheme to fit the assessment of risks in a play analysis, the NPD assessors simply identified the common, critical factors and assigned these to the play level and the remaining parameters were kept at prospect level (Fig. 8). This led to the following scheme for the play analysis: At play level: Pm = P la x P2b x P3a (i.e. reservoir x trap quality x source quality and volume) In P la one estimates the chance of finding the stated reservoir rock across the whole of the play area. In P2b one considers the general quality of the defined trapping mechanism and estimates the average chance for such traps to seal hydrocarbons. In P3a one estimates the chance that there will be a sufficient volume of mature source rock within the
102 drainage area to fill all prospects in the play. At prospect level: Pp = Plb x P2a x P3b x P4 (i.e. porosity x mapping quality x migration x retention) In P lb one estimates the chance of finding the minimum porosity in all prospects. In P2a one considers the data quality and coverage and estimates the chance for being able to accurately map the trap type of the play. In P3b one estimates chance of having an effective migration into all prospects in the play area. In P4 one considers the geological factors that may affect the sealing in the traps after accumulation and estimates the chance that they stay sealed until the present. The plays are divided into two categories: those that are confirmed by discoveries and those that are still conceptual (Figs. 8 and 6). The two categories are conceptually quite different when it comes to the risk analysis. In the confirmed plays the discoveries make up a basis for calibration for estimating the chance of success, and the assessor relates his "statistical" thinking to a number of prospects which are local analogs to the discoveries. The assessor knows that the play works, and a large number of remaining prospects gives a sound basis for thinking in terms of frequencies and probabilities. This calibration based on rate of success cannot be used on the play level, simply because the number of plays is not large enough to make frequency counts. Only a few of the unconfirmed plays have known, confirmed direct analogs in time and space on the Norwegian shelf. For most of the plays analogs were required from other parts of the world or from other parts of the stratigraphic column (for example Jurassic sand in rotated, Jurassic fault blocks may be an analog for Cretaceous sand in Cretaceous rotated fault blocks). The restricted number of plays, however, makes it meaningless to base estimates of play chance on a frequency count of success vs. failure, even on a world wide database on playtypes. Instead the NPD assessors based their estimation of play chance very much on their assessment of the critical regional parameters (Pla, P2b and P3a). In many cases only one of these parameters gives the dominant risk, while geological and geophysical data may indicate or even prove that the other parameters are favourable. In general in the NPD scheme, the less that is known about a play, the higher the risk. In frontier areas where the existence of the most uncertain parameter is based on weak indications, regional geology and speculative palaeogeographical models, the play chance was estimated to be generally less than 0.05. This is the case for more than half of the 21 unconfirmed plays in the analysis. In the cases
H. Brekke and J.E. Kalheim
where good, direct analogues exist locally, and the exploration of the new play only implies one step out from the confirmed play, the play chance was estimated to be between 0.1 and 0.25. Estimating the play chance of an unconfirmed play thus implies subjective assessments to a large degree. However, through the common, regional parameters (Pla, P2b and P3a) there is a link between the well established NPD standard for estimating the chance of success of a prospect and estimating the play chance of an unconfirmed play. By applying the NPD prospect standard on the regional parameters of the unconfirmed plays, we found that we were able to arrive at repeatable values for the play chances. We therefore believe that the NPD is able to make play chance assessments that are consistent over time, similar to the assessments of chance of success for prospects. The play analysis also showed that the assessors were able to make repeatable assessments of the total chance of success for both unconfirmed and confirmed plays. However, one cannot be certain that the relative differences in total risk level between the confirmed and unconfirmed plays are correct in absolute terms. Based on the estimated play chances for the 21 unconfirmed plays in the NPD analysis it is possible to arrive at an estimate for the chance of success for at least one of them to be confirmed. Given five independent plays ml to m5, one may estimate the chance, Pf, for at least one play to be successful in the following way: Pf= 1(1 Pro1)(1 - Pm2)(1 -- Pm3)(1 - Pm4)(1 - Pro5) Estimated in this way, the chance of success for at least one of the 21 unconfirmed plays in the NPD analysis is 0.7. Even considering the uncertainties in the estimation of the play chances, this shows that it is very likely that at least one new, independent play will be confirmed on the Norwegian shelf in the future. -
Summary and conclusions Play analysis combined with a lognormal size distribution model is a flexible assessment method that makes it possible to use all available data in an optimal way. Ideally, the method requires that the individual plays be defined such that the size distribution of the prospects of the play consists of a single statistical population. However, in areas with very little data it is probably better to make use of simple, regional "lump" models that are highly likely to comprise several statistical populations than to try to split such models into more refined models which cannot be substantiated. Such "split" models give a false impression of reduced risk.
The NPD's assessment of the undiscovered resources of the Norwegian Continental shelf
The two most difficult parameters to assess were the size distribution of the traps (net rock volume) and the number of drillable prospects. The NPD based its size distribution on the assumption that the area of closure and the net rock volume of the prospects in a play may be fitted to a lognormal distribution curve. In that way the problem was reduced to finding curves that would include both the maximum possible prospect of the play and a reasonable median size which could be calibrated by maps and discovery statistics. The number of drillable prospects in a play was estimated from prospect densities based on statistics and analogs. In the risk analysis the geological risk factors were split into two levels: the play level and the prospect level. The play level includes the regional factors that are the critical factors common to all prospects in a play, while the prospect level includes the factors that vary from prospect to prospect. Prospect chance (prospect level), and therefore chance of success of confirmed plays, was calibrated by analogs and frequency counts. Play chance (play level) of unconfirmed plays cannot be calibrated by frequency counts, and was based on the risk assessment of the individual regional factors which indirectly may be calibrated by confirmed analog plays. In many cases only one of the regional factors gives the dominant risk of the play. The NPD internal procedure for risk analysis is
H. BREKKE J.E. KALHEIM
103
seen to give repeatable results for both play chance and total chance of success. References Arps, J.K. and Roberts, T.G., 1958. Economics of drilling for Cretaceous oil and gas on the East Flank of the Denver-Julesberg Basin. AAPG Bull., 42(11): 2549-2566. Crovelli, R.A., 1984. Procedures for petroleum resource assessment used by the U.S. Geological Survey - - statistical and probabilistic methodology. In: C.D. Masters (Editor), Petroleum Resource Assessment. IUGS Publ., 17: 24-38. Crovelli, R.A., 1986. A comparison of analytical and simulation methods for petroleum play analysis and aggregation. U.S. Geol. Surv. Open-File Rep. 86-79, 21 pp. Crovelli, R.A., 1987. Probability theory versus simulation of petroleum potential in play analysis. In: S.L. Albin and C.M. Harris (Editors), Statistical and Computational Issues in Probability Modelling, 1. Ann. Oper. Res., 8:363-381. Crovelli, R.A. and Balay, R.H., 1988. FASPUM metric version: analytic petroleum resource appraisal microcomputer programs for play analysis using a reservoir-engineering model. U.S. Geol. Surv. Open-File Rep. 87-414, 14 pp. Dolton, G.L., 1984. Basin assessment methods and approaches in the U.S. Geological Survey. In: C.D. Masters (Editor), Petroleum Resource Assessment. IUGS Publ., 17: 4-23. Kaufman, G.M., 1993. Statistical issues in the assessment of undiscovered oil and gas resources. Energ. J., 14(1): 183-215. NPD, 1993. Petroleum resources E Norwegian Continental Shelf. External report, Norwegian Petroleum Directorate, 40 pp. White, D.A., 1980. Assessing oil and gas plays in facies-cycle wedges. AAPG Bull., 64(8): 1158-1178. White, D.A., 1988. Oil and gas play maps in exploration and assessment. AAPG Bull., 72(8): 944-949.
ExplorationDepartment, Norwegian Petroleum Directorate, P.O. Box 600, N-4001 Stavanger, Norway Exploration Department, Norwegian Petroleum Directorate, P.O. Box 600, N-4001 Stavanger, Norway
This Page Intentionally Left Blank
105
Cross-validation of resource estimates from discovery
process modelling and volumetric accumulation modelling: example from the Lower and Middle Jurassic play of the Halten Terrace, offshore Norway Richard Sinding-Larsen and Zhuoheng Chen
Volumetric accumulation modelling attempts to capture the essentials of the hydrocarbon generation, entrapment and retention processes leading to a family of accumulations in a petroleum play. Discovery process modelling uses the size distribution of discovered hydrocarbon accumulations and the efficiency of the exploration effort to estimate the number and sizes of undiscovered accumulations in a play. Direct incorporation of exploration data and geological judgement is a prerequisite for the volumetric modelling, whereas accurately representing exploration efficiency through the discovery sequence is a prerequisite for discovery process modelling. How to achieve a valid play definition is a problem common to both methods. Cross-validation of the two methods applied to the same play can sharpen the focus of each method if used individually, as well as capitalize on the strong points of both methods if used in an integrated approach. These issues are exemplified through a case study focussing on the assessment of the undiscovered potential of the pre-rift Lower and Middle Jurassic play of the Halten Terrace, Mid-Norway continental shelf.
Introduction Many methods and models for petroleum resource assessment have been developed using different geological perspectives. Most hydrocarbon assessment approaches in current use can be classified by a simple dichotomy consisting of those that rely on modelling the natural processes of deposit formation and those that rely on modelling the process of exploration and subsequent discovery of hydrocarbon accumulations. Each method requires a specific level of geologic knowledge or degree of exploration information, and provides in many cases compatible types of results (e.g. aggregate estimates and field size distributions). The natural hydrocarbon accumulation process is modelled in this study by a volumetric approach leading to estimates of the underlying field size distribution and to a distribution of the total number of fields in a petroleum play. Because each step in this methodology is associated with genetic parameters, each step is also interpretable in terms of specific geological processes. However, many factors affect the reliability of the estimation results. Volumetric variables may, for example, be partially correlated. Ignoring these correlations may cause biases in the
e s t i m a t e s o f the field size distribution and the resulting play potential. A discovery process model estimates the play potential by modelling the interaction between the natural processes leading to hydrocarbon accumulation and the results of exploration drilling. Bias due to non-random sampling from the underlying field size distribution is modelled by specifying that the discovery probability is proportional to size. Problems related to the application of this method often come from an inadequate play definition or from the fact that insufficient information is provided by the discovery sequence to estimate the underlying parent field size distribution. The application of both methods to the same play permits us to sharpen our focus on the limitation of each method if used individually. The credibility of the assessment will however be greatly improved if results from individual approaches are in agreement. Any difference in the assessment results, if significant, will challenge the assumptions of the individual methods or cast insight into the geological concepts used. In this paper, the aforementioned cross-validation approach is illustrated with assessment of the play potential of the pre-rift Halten Terrace play on the Mid-Norway continental shelf.
Quantification and Prediction of Petroleum Resources edited by A.G. Dor6 and R. Sinding-Larsen. NPF Special Publication 6, pp. 105-114, Elsevier, Amsterdam. 9 Norwegian Petroleum Society (NPF), 1996.
106
R. Sinding-Larsen and Z. Chen
Geological setting The Halten Terrace play comprises discoveries and drillable prospects in the pre-rift sequence of Early to Middle Jurassic age and has been the focus of exploration since 1980. The major reservoir units are the Lower-Middle Jurassic shallow marine sands. The petroleum accumulations are sourced by two major units, a gas/condensate-prone Upper TriassicLower Jurassic coal unit, and a Upper Jurassic oilprone black shale. A top-seal is provided by the Upper Jurassic and/or Lower Cretaceous impermeable shales. The major trap styles are similar to the traps in the North Sea, including simple extensional traps consisting of rotated fault blocks, horsts, and the combination of these two geometries formed during the Late Jurassic-Early Cretaceous Cimmerian rifting phase. Salt movement is also believed to have contributed to the formation of structural traps (Jackson and Hastings, 1987; Koch and Heum, 1995; Blystad et al., 1995). Fig. 1 is a play map of the pre-rift play of the Halten Terrace, offshore Mid-Norway. The database used in this study is mainly built upon the information provided by Kalheim (1989) and has been updated from recent publications (Norwegian Petroleum Directorate, 1992; Ehrenberg, 1990; Fagerland, 1990; Koch and Heum, 1995). Table 1 lists the field and prospect data and the estimated recoverable reserves. Reserves for each field were
estimated by different organizations and differ one from the other. Three different estimates are shown in Table 1. The assessment from a discovery process model may vary if different estimates of reserves are used. A hypothetical order of discovery is included in Table 1 and used to test the sensitivity of the Norwegian licensing procedure on the discovery process modelling results.
Volumetric approach The volumetric accumulation modelling technique used is based on a statistical play analysis method developed by United States Geological Survey (USGS) for their 1989 assessment of undiscovered petroleum resources in the United States. This method has recently been used by the Norwegian Petroleum Directorate (NPD) for their assessment of petroleum resources on the Norwegian Continental Shelf (NPD, 1993) and a description of the methodology is provided by Brekke and Kalheim (1996). The size of any given field in the population is determined by an appropriate variant of the reservoir engineer's field size equation. The solution of the equation gives the volume of recoverable hydrocarbons for the field. In order to get the distribution of all possible field sizes in the play, it is necessary to input each variable as a frequency distribution describing both the range of possibilities and probabilities for all possible accu-
Fig. 1. Pre-rift play of Halten Terrace, Mid-Norway continental shelf (based on Kalheim, 1989).
107
Cross-validation of resource estimates from discovery process modelling and volumetric accumulation modelling Table 1 Field and prospect data (from NPD, 1992 and Scenario 1.3, table 3 of Kalheim, 1989)
Midgard Tryihans S. Tyrihans N. SmCrbukk
Heidrun SmCrbukk S. Mikkel Njord
Trestakk 6507/08-4 P-04 P- 13 P-37 P-48 P-42
N.R.V. (km 3)
C1.A. (km 2)
N.Th (m)
N/G
Res. I a ( 106 )
Res. II a ( 106 )
Res. III a ( 106 )
2.25 0.77 1.04 8.38 3.59 2.18 0.51 2.01 9 9 1.20 0.77 4.60 1.96 0.44
47 21 13 120 32.5 25 10 25 9 9 24 8 80 25 5
48 36 80 70 110 87 51 80 9 9 50 96 58 78 88
0.9 0.85 0.9 0.55 0.80 0.45 0.85 0.6 9 9 0.95 0.75 0.45 0.55 0.55
101 15.5 18.9 125.3 109.4 49.7 21.1 36.0 3.9 20.4 dry dry dry dry dry
124.1 15.5 18.9 108.7 116.0 42.0 14.7 30.6 3.9 11.2 dry dry dry dry dry
112.0 17.0 20.0 102.0 164.0 46.0 20.6 43.0 8.2 20.0 dry dry dry dry dry
Hypoth. order 3 4 9 2
1 6 5 8 7 10
a Reserves are measured in t.o.e. (recoverable). N.R.V = net rock volume; C1.A = closure area; N.Th and N/G -- net thickness and net/gross ratio; Res. I -- estimated reserves by NPD (NPD, 1992); Res. II = estimated reserves by industry (unpublished); Res. III = estimated reserves by Statoil (Kock and Heum, 1992); Hypoth. order = hypothetical order of discovery only influenced by exploration information (J.E. Kalheim, 1993, pers. commun., 1993).
Table 2 Input parameters and estimating results from the volumetric approach Geologic variables
F 100
F95
F75
F50
F25
F05
F0
Closure (km 2) Net thickn. (m) Porosity (%) Trap fill (%) HC sat. (%)
4.00 10.0 10.0 50.0 50.0 20
6.00 12.0 12.0 55.0 52.0 21
12.0 25.0 13.0 60.0 60.0 22
19.0 40.0 16.0 75.0 70.0 25
29.0 60.0 19.0 90.0 78.0 30
55.0 110.0 23.0 95.0 85.0 35
110.0 150.0 26.0 100.0 90.0 40
Nr. of prosp.
Success ratio = 0.273; recovery factor for oil = 0.35; recovery for gas = 0.65 Probability for oil = 0.25; probability for gas = 0.75 Expected play potential for oil: 33.4.106 ton 110.6-109 m 3 Expected play potential for gas: Total play hydrocarbon potential: 144.0.106 t.o.e.
mulations in the play. Data reflecting the geologist's opinion on the range of values and their probabilities are specified for each variable (Table 2). The distribution of field sizes is computed by the multiplication of a recovery factor with the frequency distributions of closure, thickness, porosity, trap fill, and hydrocarbon saturation, and divided by a formation volume factor. Field size -
C.RF.
A . T. ~ . TF.
FVF
Snc
(1)
where C is a constant equal to 0.84 for 10 6 metric tons of oil and 1.0 for 10 9 m 3 of non-associated gas, R F is recovery factor (%), A is area of closure (km), T is net reservoir thickness (m), ~ is porosity (%), SHc is hydrocarbon saturation (%), T F is trap fill (%), and F V F is the formation volume factor (no unit). The distribution of the number of fields is computed by multiplying the estimated distribution of
the number of prospects with the anticipated success ratio. The multiplication of the volumetric input expressed as distributions to obtain the field size distribution is an approach that can be readily understood, but can be flawed, sometimes seriously (Lee et al., 1990) because of the interdependency between variables. If the co-variances are believed to be nonnegligible and are unknown, they will have to be estimated either by analogy or prior knowledge.
Volumetric modelling results In a preliminary assessment, all geological variables (Table 2) were assumed to be independent and correlations were ignored. The recovery factors are assumed to be equal for all the fields in the play and all resource figures are expressed as recoverable
R. Sinding-Larsen and Z Chen
108
I00 i
.*
i
:
:
!~,
!
,
!
i
i
'
90 80
70 60 O
g O
50 ..
40 i
i
.
:!:
i
:
::i
i
:
:
i
i
i
ii
i
i
:
!i
"g:l
30 t.)
i
20
i
-
,
!
'
10
.
i
i
l00
101
'
.
~i i -
.
.
_
i
102
103
Field size (10^6 TOE, recoverable) Fig. 2. Field size distributions estimated from the volumetric approach: remaining (REM) (/z = 2.5, cr2 a 2 = 1.07), and observed (DISC).
resources. The expected remaining oil potential was calculated equal to 30.10 6 tons with 95% chance of having more than 4.106 tons and a 5% chance of having more than 94-10 6 tons. The expected remaining gas potential was calculated to be 100.109 m 3 with a 95% chance of having more than 32-109 m 3 and a 5% chance of having more than 220.109 m 3. In terms of hydrocarbons, the expected undiscovered recoverable play potential was calculated to be 130.106 t.o.e. (tons of oil equivalents) which represents the total resources coming from seven undiscovered accumulations greater than a minimum economic cut-off of size of 4-106 t.o.e. The assumption of independence of the geological variables is challenged by the analysis of the input data which suggests the presence of correlations among several geological variables. Table 3 shows a matrix of covariances estimated from well data. The covariance (0.0824) between the logarithmic net reservoir thickness and HC saturation in this play Table 3 Covariance matrix
Porosity Net thickn. HC sat. Closure
Porosity
Net thickn.
HC sat.
Closure
0.1014 0.0213 0.0549 0.0
0.0213 0.7203 0.0824 -0.065
0.0549 0.0824 0.0754 0.0
0.0 -0.065 0.0 0.7838
aZ = ~ a z + 2 y'~aij,
i :/: j
Mean = exp(# + 0.5o-2) Variance = exp(2/z + o'2)[exp(a 2) - 1]
_
_
0.94), parent (PAR) (# = 3.1,
represents a significant contribution to the overall covariance, but its impact on the volumetric assessment is small because the positive covariances are being compensated by negative covariance. Because of this compensating effect, the expected remaining play potential now calculated equal to 144.106 t.o.e., is only about 10% larger than the previous estimate obtained when dependencies were ignored. The estimated mean size of the remaining fields is now calculated equal to 21-106 t.o.e. Fig. 2 shows the final estimated remaining field size distribution (REM) and the underlying parent size distribution (PAR) computed as the aggregate of REM with the ten discoveries (DISC).
Discovery modelling The attributes of a deposit and its surrounding environment which influence the deposit's probability of being discovered can, in most cases, be interpreted as arguments of a function that may usefully be defined as a deposit's "magnitude". To date, most models have taken deposit size (area, rock volume, forms of oil equivalent) as a measure of magnitude, but the relationship between such simple definitions of magnitude imply that complexity must be represented in the next generation of models. Identification of attributes that determine magnitude is a first step toward estimating the functional relation between magnitude and probability of discovery. A second step is to determine how definition of magnitude
Cross-validation of resource estimates from discovery process modelling and volumetric accumulation modelling
should change with changes in geologic environment, technology, economics and resources remaining to be discovered (Kaufman et al., 1988). Discovery modelling is based on an understanding of the discovery process and requires a set of assumptions. The following two features characterize a discovery process model: (1) a discovery process involves sampling from a frequency distribution of field sizes without replacement; and (2) that the order of discovery is governed by sampling proportional to magnitude and without replacement. Magnitude is defined here as field size measured as tons of oil equivalent, raised to a power ft. The successive sampling scheme conforms to empirically based industry experience that the discovery sequence is biased towards finding the larger fields early in the exploration process. Assuming N fields in a play with associated sizes Y(yl, y2,... , YN), and j - 1 fields that have been discovered in the order Yl first, y2 second and so forth, the probability that the field labelled j will be the next discovery conditioned on the j - 1 discoveries is" t~ Yj P(yj IY) - ~ (2) Yj + Yj+I -+-""-k Y~N where fl is called the bias parameter or discoverability parameter. The lognormal discovery process model (LDP) used in this study is developed by the Geological Survey of Canada (GSC) and is based on the successive sampling model of Kaufman et al. (1975). Although the size distribution can be in any form, lognormal is recognized as a frequently used model and its use facilitates the statistical estimation. Details of the model are given in Lee and Wang (1985, 1986). Using a discovery sequence, the LDP model provides /z and 0 "2, the parameters of the estimated lognormal parent population, N, the total number of deposits in the play, and fl, the discoverability parameter. If the fl exponent is zero, the probability of discovering any field is proportional only to the number of fields. If the fl factor is one, the probability of discovery is directly proportional to field size. The fl parameter is influenced by many factors, such as licensing sequence, water depth in the offshore, availability of infrastructure and increase in the ability with time to seismically discriminate certain types of prospects. Discovery modelling is applicable to semi-mature and mature plays where the discovery process can be characterized by a time sequence of discovered oil and gas magnitudes. The data required are the magnitudes of field sizes and their order of discovery. The geological analysis that allows the organization of this information into an adequate definition of a petroleum play is, however, critical for obtaining reasonable answers.
109
Discovery modelling results The discovery sequence of ten discoveries is shown in Table 1. Using the NPD discovery sequence, global maximum likelihood estimates of fl, /z and 0-2 (discoverability, mean and variance) and the log maximum likelihood for different numbers of deposits, N, can be obtained as shown in Table 4. The most plausible value for N is 12 which coincides with the minimum value used as the range of possible input values, but N = 12, is not consistent with the number of prospects and the success ratio. The failure to obtain a reasonable maximum likelihood estimate of N may indicate that the discovery sequence does not contain enough information for determining N. In this case other methods or additional data such as the geological control provided by the prospect list are needed for a more reliable determination of N. However, the use of maximum likelihood analysis may still be a useful tool in judging what is the most plausible value of N (Lee and Wang, 1985). The application of a non-parametric discovery process model, the geo-anchored method (Chen, 1993), indicates that the discoverability parameter fl may lie in the interval of 1.1-1.2. Accepting the choice of the discoverability parameter fl = 1.2, marginal maximum likelihood estimates of/z, a 2, and the marginal likelihood values for different values of N can be calculated by the LDP model as shown in Table 5. The marginal estimates have now a maximum log likelihood value at N = 18 which is consistent with the number of prospects and the success ratio of the play. The corresponding marginal estimates of the parameters for the underlying field size distribution Table 4 Maximum likelihood estimates of/z, 0.2 and /3 and log-likelihood values (ML) for different N's (LDP) N
/~
/2
6 .2
ML
12.0000 13.0000 14.0000 15.0000 16.0000 17.0000 18.0000 19.0000 20.0000 22.0000 24.0000 26.0000 28.0000 30.0000 32.0000 34.0000
0.9223 0.9954 1.0492 1.0919 1.1273 1.1571 1.1840 1.2071 1.2280 1.2641 1.2946 1.3213 1.3437 1.3641 1.3826 1.3990
3.2310 3.1240 3.0290 2.9430 2.8640 2.7910 2.7240 2.6610 2.6010 2.4930 2.3960 2.3070 2.2270 2.1520 2.0830 2.0190
1.2290 1.3290 1.4110 1.4800 1.5400 1.5930 1.6420 1.6850 1.7250 1.7970 1.8590 1.9140 1.9630 2.0070 2.0470 2.0840
-58.4552 -58.4872 -58.5250 -58.5611 -58.5941 -58.6240 -58.6509 -58.6752 -58.6974 -58.7361 -58.7690 -58.7972 -58.8219 -58.8437 -58.8631 -58.8805
N = the number of fields"/3 = the discoverability index;/z and 0 .2 are the estimated parameters for a lognormal size distribution.
110
R. Sinding-Larsen and Z. Chen
100
90 . . . . . . i .....i....!.i.ii.i-i . . . . . ~ 9i.i
i l !ii!!!i LoP
iiii!iiii i ilii !iiii!i84
80
-S
70
o
50
~
40
!i ! !!!ii!
3o
~
20 10 0 I0-
101
I0 o
10 2
IO s
10 4
Field size distribution (10^6 BOE, recoverable)
Fig. 3. Lognormal field size distributions. "RD" estimated under assumption of random sampling (/z = 3.47, 0.2 = 1.17)" "LDP" estimated by the lognormal discovery process model of CGS with N = 18 and 13 = 1.2 (/z = 2.81, 0.2 = 1.552).
Table 5 Marginal log-likelihood and estimates of/z and 0.2 for/3 = 1.2 and different N's N
/2
~.2
log-likelihood
12 14 16 18 20 22 24 26 28 30
3.238 3.062 2.923 2.810 2.716 2.636 2.568 2.508 2.455 2.409
1.258 1.384 1.467 1.525 1.567 1.598 1.621 1.639 1.652 1.662
-54.2303 -54.0913 -54.0447 -54.0305 -54.0307 -54.0382 -54.0496 -54.0633 -54.0783 -54.0939
N - - the number of fields;/3 = the discoverability index; # and are the estimated parameters for a lognormal size distribution.
0 .2
are measured in units of 10 6 tons of oil equivalents, /2 - 2.8 and 6 2 = 1.5 (Table 5). A/3 value of 1.2 suggests that field size has exerted a strong influence on the order of discovery, as can be inferred from Fig. 3 where the parent lognormal distribution estimated by assuming a random sampling (RD) is compared to the lognormal distribution estimated by the LDP model discussed previously. The separation between these two distributions reflects the magnitude of/~ and indicates that there is a strong tendency for large fields to be discovered at an early stage. Using the parameters of the LDP curve in Fig. 3, /2 - 2.8, ~.2 __ 1.5 and N -- 18 (Table 5), the expected total resources in the play are calculated to be 630.10 6 t.o.e. (tons of oil equivalent). Subtracting the discoveries
gives an expected remaining play potential of 130.10 6 t.o.e., corresponding to a mean remaining field size of 18.10 6 t.o.e. The lognormal discovery process model used on simulated data shows that the LDP model may slightly underestimate the field size distribution when /~ is large (Chen and Sinding-Larsen, 1994). The estimates of the field size distribution from the LDP model in this study may therefore be regarded as somewhat pessimistic. It may be desirable to test the sensitivity of/3 to change in the order of observations, because the number of discoveries are small. The reserves of fields estimated by different organizations are significantly different in some cases (Table 1). On the other hand, the licensing process may also cause problems in the application of discovery process modelling. However, a sensitivity study of the two factors (licensing process and differing field estimates) has been done by using the discovery sequence with different estimates of the reserves of fields and re-ordering the discovery sequence according to the hypothetical order in Table 1. The test shows that in the case of this play no significant impact on the assessment of the resource potential can be observed.
Integrated approach The discovery process model employed here uses only the magnitudes in a discovery sequence. It is therefore desirable to incorporate as much as possible of the geological information available to constrain the estimates. On the other hand, from our experience,
Cross-validation of resource estimates from discovery process modelling and volumetric accumulation modelling
111
of some individual discoveries are known or inferred. Searching the two parameters /z and 0-2 from the range, 2.8 < /z < 3.1, 1.2 < 0-2 < 1.5, defined by PAR in Fig. 2 and LDP in Fig. 3, one can obtain many combinations between the outer ranges of these two parameters that describe candidates for the "true" parent log-normal field size distribution. Such combinations can be ranked according to how well they reflect the degree to which means of predicted fields fit discovered fields. Fig. 5 shows one match that accommodates both the size of discoveries made in the play and the estimate of N -- 17 from the volumetric approach. The figure indicates that the first four largest fields have been discovered with the seven undiscovered fields in the 5th, 7th and 12th-16th rank. This match was examined along with others and is consistent with our geological interpretation that not more than one undiscovered field is expected to be greater than Njord (36.106 t.o.e.). The same matching technique was applied to the estimates provided by the LDP model with N - 18 and conditioned on the same set of geological arguments (the first four largest discoveries are found and no more than one remaining field should have a size larger than the Njord). Fig. 6 shows the matched field sizes according to rank representing the estimated sizes of individual remaining fields. This matching procedure can be regarded as a feedback process which allows estimates of field-size distribution parameters and the total number of fields in place to be successfully refined. It may provide an
the field size distribution derived by the volumetric approach often leads to an overestimation because of a bias towards the more favorable situations in the geological observations. One way of comparing these assessments is to construct the field sizes by rank from the estimated parameters/z and 0 -2 derived by the two methods. This relation is used in a matching procedure that allows for cross-calibration through the introduction of judgmental information about the geology.
Field size by rank distribution A major advantage of using the play analysis approach for hydrocarbon assessment is its ability to produce estimates of expected field sizes that should exist within the play. Given a field size distribution shape and specific numerical parameters for the number of fields, order statistics can be used to generate the expected field sizes of the play (Taylor et al., 1991). A rank plot (Fig. 4) is made of the expected field sizes arranged in their order of discovery conditioned on N, the total number of fields. Individual boxes encompass a 90% range in the predicted size of each field. The lower limit represents a 95% probability that the field is at least that large and the upper limit indicates that there is only a 5% probability of the field being larger than that value.
Matching process The ability to produce field rank plots using order statistics has another important application if the size
Pre-rift play, Halten Terrace
10 2 9
::::::::7':7~.
"" ~: i: : ; : : : : ~ : . : : : i : ~ : : : ~ : : .
~.~..:..:.~:::~i::~::.:~.~:;::~:~:::~:~:::::::~::~::~::~::::::;~:z
r
9
.........................
!ii,
iiiiiii
!
I0 ~
Y. 10o
. 10-~
0
2
4
6
8
.
i.
10
12
.
i. 14
16
18
Size rank Fig. 4. Field size distribution constrained to N = 17 estimated by the volumetric approacn with the 5-95% intervals (plus sign represents mean value).
R. Sinding-Larsen and Z. Chen
112
Pre-rift play, Halten Terrace 10s
::::1:::::11:::::::::::::::::::::::::::::::: ............... .
.
::::::::: :::::::::::::::::::::::::::::::::::::::::::::
! ................................
.
.
.
.
.
.
.
. . . . . . . . . . . . .
.
.
:
.
.
! ...............
.
.
.
.
.
.
.
.
.
.
.
s
.
: .
.
9
102
.
. . . . . . . . . . . . . . . . . . . . . .
i . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ;
i :
..............
; . . . . . . . . . . . . . . . . . .
~
:
i
.
. .
.
.
.
.
.
.
.
!
.
.
.
! . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
. . . . . . . . . . . . . . . . . .
.
:
.
.
.
.
.
.
.
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . .
.
.
.
: :
.
. . . . . . . . . . . . . . . . . . . . . . . . . . . .
;
. . . .
~.
.
.
i . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . :
.
~
9
i ...........
.
.
.
.
,
. . . . . . .
i . . . . . . . . . . . . . .
i .................
~ ....
.. :..:.::.
; . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
; . . . . . . . . . . . . . . . ;. . . . . . . . . . .
; ;
. . . . . . . . . . . . . .
~!
.. .. .. .. .. .. . . . . .
:
', . ~ o , ~
............
9
. . . . . . . . . . . . . . . . . . . . . .
. * . : .
o,
; !
9
i ...........
.... ..
. . . . . . . . . . . 9 .
" ............ . . . . . .
~,
; ;
.
. . . . . . . . ; . . . . . . . . . . . . . . . . . . . . . . . . .
9 .............
. . . . . .
:
:!:i;
..,
!
.
.
.
.
: .....
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
~ .............
~ . . . . . . . . . . . . . . : . . . . . . . . . . . .
9................
~: ............... . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . .
. . . . .
:
_
9 ....
.
.
ii:i :!! .
.
"
"
i
:
.
.
i
:
:
.
. . . . . . . . . i . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
~
.
.
? :
.
.
i.
. . . . . . . . . . . . . . . .
~
; ;
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
0
_iii:ii::i
~ ..............
! . . . . . . . . . . . . . . . . . . . . . . .
. . . . .
. . . . . . . . . . . . .
. . . . ::...: . . . . . . . . . . . : : : : : : : : .... : : : . : : : : : =
; . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . .
; _;
!
:
. . . . . . . . . . . . . . .
-............
O~
:1:1:::i:1:::::::::::::::::::::::::::::::::::::::::::::::::::::::
! ............................... .
.
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . .
. . . . . . .'. ....... ,, . . . . . . . . . . . . . . . . . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
i
. . . . . . . . .
:
.
:
:
10 ~ :::':::::::::i :" :::::"!'": .... . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
:.
. . . . . .
: :
. . . . . . . . . .
:':::~:
':~
...... : ....
:-:-:.1.....: :
:..-: ......
,: . . . . . . . . . . . ~...:..: ========================================== : . . . . . . . . . . . . . . . . . . , ......... : ............
:
.
. . . . . . . . . .
!
.
9
;
.
.
.
.
.
.
.
.
.
.
.
.
!
!
.
!
. . . . . . . . . .
: . . . . . . . . .
:
.
! . . . . . . . . . . . .
.
..
;
.
.
.
.. .
.
.
.
.
.
.
.
.
.
.
.
.
.
;
.
.
i .
9
i
"
i
2
9
:
~
t
6
8
,.
~
4
. ,
.
.
.
.
.
:
lfl-I
. . . . . . . . . . . . .
;
. .
:
0
.... :::: : . . . . . .
.
:.
. . . . . . . . . . . . . . . . . . . .
.
~
~
I0
i
12
i
14
16
18
20
Size rank Fig. 5. Individual field size by rank constrained to the discovery record and N = 17. The lognormal parameters are estimated by the volumetric approach. The empty boxes represent fields undiscovered and the uncertainty is expressed as 25th and 75th upper percentiles. The dots are the discoveries.
Pre-rift play, Halten Terrace 103
~!!:!'i::!!!!!!!!!" :::!!:!:i!: ::'!::!!i!i'!::'! :! :i "1 :..:!!!!.i!!1!:!~.1.~.~i!!:!!!!i.1~!..i!.!!!!!i!!!~!!!!!:!!!1~!!!~!~!!!~!~:!!!!.:i~!1~::!!~:!!.!!!!!:!!~!~ .
.
.
.
.
.
.
.
.
.
.
.
. .
.
.
.
.
.
.
.
.
.
.
.
.
. .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
. .
.
"
.
:
.
.
.
.
.
..
.
.
.
.
.
.
.
.
.
. .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
::
.!.
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ......................................
:!:i
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
:. . . . . . . . . . . . . . . . . . . . . . . .
.
:.
. . . . . . . . . . . .
!
..
i0 2
;
": .
.
.
.
.
.
.
.
.
.
9
:::
: .! : i:::!!!!.?:.!:
......
'
.
.
.
.
.
.
.
9
. .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
. .
. . .
.
.
; . . . . . . .
.
.
.
" .
" -
:
9
_
.
. . . . . . . . . . . . . . . .......................................
::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::
; . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ~ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
:::...::...::.::...: ..... :.'..t33:.,
.
.
:
..
i
.
":.
-.
.:-:
:...
: ....................................... :. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. ....... : ...... ..i
....
:
..... ::
;
i . . . . . . ; . . . . . .
.:::::.i-:
::
:
0 v
i
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
;
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
: . . . .
y. 100
:::::::..:.::::::::..::.::::..:...: .
9
.
.
.
:.'
.
::"
.
. :
.
.
....
.
.
.
.
.:'..i
.
.
.
.
.
.
.
.
.
IO-t
0
.
.
.
.
.
.
.
.
.
.
"
::::::::::::::::::::::::::::::::::::::::::::::::::
: ::':':!:!!.i
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
'
'
.!" ;
.
'
"
" .
.
.
.
.
.
;
.
.
.
.
.
.
.
.
.
.
.
. . . . . . . . . . . . . . . . . . . . . . . .
.
"."::: }:!":i'!! : :: i::
:..:
.
:i
. . . . . . . . . . . . . . . .
.
.....
:
..
: .
5
" .
.
.
.
.
10
" .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
:..
.
.
_
: .....
. . . . . . . . . . . . . . . . . . . .
i ....
...,i
~
.
.
.
.
.
.
.
.
.
.
.
15
.
.
.
.
.
.
.
.
.
.
.
.
i
. . . .
.
20
Size rank Fig. 6. Individual field size by rank constrained to the discovery record and N = 18, the pre-rift play of Halten Terrace. The lognormal parameters are estimated by the LDP model. The empty boxes represent fields undiscovered and the uncertainty is expressed as 25th and 75th upper perecentiles. The dots are the discoveries.
alternative to the estimation of N (number of fields) for cases in which one is totally dependent on expert judgment. A convenient way of portraying this procedure is to represent different candidates for the parent distribution in a / z and r 2 plot. In Fig. 7 the final estimates for/z and cr2, marked F, representing our
"best" estimate of the parent distribution, lies between the estimates by each method individually. The total resources in the play can now be represented as the sum of all the discovered fields plus all the mean values of the "empty boxes" shown in Fig. 4. Once the assessor is confident that a successful "match" has been made, he has effectively
Cross-validation of resource estimates from discovery process modelling and volumetric accumulation modelling
11 3
1.6
1.4
. . . . . . . . . . . . . . . .
D .Q
1.2 . . . . . . . . . . . . . .
~r
:. . . . . . . . . . i F
:
R
:
9
!
.................
o
V ~ ra~
I
. . . . . . . . . . . . . . . . . . . .
.o . . . . . . . . . . . . .
: .........
: :
0.8 . . . . . . : . . . . . . . . . . . . . i
. . . . . . .
,
. . . . . .
: 0.6
-.: . . . . . |
2.6
:-
!
i. . . . .
. . . . . . . . . . . . .
2.8
3
3.2
i . . . .
|
3.4
!
3.6
4
3.8
M u
Fig. 7. Estimated lognormal parameter and/.t and 0 -2 pairs. R = estimated directly from discoveries; V = volumetric estimates before cross validation; D -- discovery process model estimate; F = estimates of discovery process model and volumetric approach after cross validation.
contributed additional information. That is, the ranges (uncertainty) of the largest four fields, as well as other "matched" fields, can now be replaced with a single value (no uncertainty), and the rest of the field rank sizes are adjusted to reflect this added information. Conditioned on the match in Fig. 5, the estimated conditional play potential from the volumetric approach is adjusted to lie in the range from 84 to 145.106 t.o.e, with probability 0.9 and with an expected value of 116.106 t.o.e. Conditioned on the match in Fig. 6, the remaining play potential predicted by the LDP model is adjusted to be in the range from 110 to 1 8 0 . 1 0 6 t.o.e, with probability 0.9 and with an expected value of 144.106 t.o.e. The uncertainties in the size of fields remaining to be discovered are reduced by attaching probability 1.0 to the discoveries in Fig. 4. The largest projected remaining field is to be about 40.106 t.o.e, and the mean size of the remaining fields is 18.106 t.o.e. Table 6 shows a comparison of the play potentials estimated from the two methods. Although the final result of the play potentials estimated from these two methods are a function of the difference in numbers of remaining fields, the mean values of the estimated parent distributions are less than 1.106 t.o.e, apart.
Discussion Play potential, field size/frequency distribution and the number of economic accumulations have been estimated and cross-checked by simultaneously apply-
Table 6 Summary of the estimates (expected values)
Rem. fields Mean rem. size Mean par. size Rem. resources Total resources
Vol. [I]
Vol. [II]
L D P [I]
L D P [II]
7 20.6 37.9 144.0 645.1
7 16.6 36.3 116.2 617.3
8 17.5 35.6 139.9 641.0
8 18.0 35.8 143.9 645.0
Total discovered reserves: 501.1 9106 t.o.e Total number of discoveries: 10 Vol. [I] and Vol. [II]: volumetric estimates before and after the integrated approach. L D P [I] and L D P [II]: L D P model estimates before and after the integrated approach.
ing both the volumetric approach and the lognormal discovery process model. The field size distribution and play potential of this pre-rift play could have been overestimated if the volumetric method was applied alone because of uncalibrated bias in the geological variables. In contrast, without additional geological constraints and the use of other type of information, the LDP model, in this case, could not have generated a reasonable estimate of the expected number of economic petroleum accumulations and could therefore have produced a severely underestimated play potential. The sensitivity of the discovery modelling approach to the licensing process and differences in the reserve figures of discoveries given by different sources were studied by using the three reserve figures and the hypothetical order of dis-
11 4
R. Sinding-Larsen and Z. Chen
covery given in Table 1. No significant departure in the estimation results was found. This application shows that cross-validation is a valuable tool for controlling the quality of resource assessment. By cross-validation, differences in the initial assessment results will force the assessor to check the applicability of the methodology in use and cast doubts on the geological model. Finally if the differences are reduced to an acceptable level, the credibility of the assessment result will be improved by the degree of the agreement obtained between the assessment results from the two methods used individually. The quality control of resource assessment can be established by setting standards for the comparison of assessment results, for example, by the methodology followed in this paper and ensuring that these standard are met in routine assessment. Subsequent to the presentation of this paper the Saga Petroleum Company made a discovery in well 6406/2-1. Total recoverable resources are estimated to lie in the range from 50 to 100.106 tons of oil equivalents and would fit perfectly the largest undiscovered field predicted in Fig. 5 or Fig. 6.
References Blystad, P., F~erseth, R.B., Larsen, B.T., Skosgeid, J. and TCrudbakken, B., 1995. Structural elements of the Norwegian continental shelf, Part II. The Norwegian sea region, NPD Bull., 8, 45 pp. Brekke, H. and Kalheim, J.E., 1996. The Norwegian Petroleum Directorate's assessment of the undiscovered resources of the Norwegian Continental S h e l f - background and methods. In: A.G. Dor6 and R. Sinding-Larsen (Editor), Quantification and Prediction of Hydrocarbon Resources. Norwegian Petroleum Society (NPF), Special Publication 6, Elsevier, Amsterdam, pp. 91103 (this volume). Chen, Z., 1993. Quantification of petroleum resources through sampiing from a parent population and as a function of basin yield. Doctorate dissertation, Norwegian Institute of Technology, 1993, 223 pp. Chen, Z. and Sinding-Larsen, R., 1994. Discovery process mod-
R. SINDING-LARSEN Z. CHEN
elling, a sensitivity study. Nonrenewable Resources, 3(4): 295303. Ehrenberg, S.N., 1990. Relationship between diagenesis and reservoir quality in sandstones of the Garn Formation, Haltenbanken, Mid-Norwegian continental shelf. AAPG Bull., 74(10): 15381558. Fagerland, N., 1990. Mid-Norway shelf-hydrocarbon habitat in relation to tectonic elements. Nor. Geol. Tidsskr., 70: 65-79. Jackson, J.S. and Hastings, D.S., 1987. The role of salt movement in the tectonic history of Haltenbanken and Tr~enabanken and its relationship to structural style. In: Spencer et al. (Editors), Habitat of Hydrocarbons on the Norwegian Continental Shelf. Norwegian Petroleum Society (NPF), Graham and Trotman, London, pp. 241-258. Koch, J.O. and Heum, O.R., 1995. Exploration trends of the Halten Terrace. In: S. Hanslien (Editor), Petroleum Exploration and Exploitation in Norway. Norwegian Petroleum Society (NPF) Special Publication 4, Elsevier, Amsterdam Kalheim, J.E., 1989. A play description scenario offshore Norway (Scenario 1.3). Workshop material prepared for CCOP, The Coordinating Committee for Offshore Prospecting in South East Asia, Bangkok. Unpublished manuscript, 32 pp. Kaufman, G.M., Balcer, Y. and Kruyt, D., 1975. A probabilistic model of oil and gas discovery. In: J.D. Haun (Editor), Methods of Estimating Undiscovered Volume of Oil and Gas Resources. AAPG Stud. Geol., 1: 113-142. Kaufman, G.M., Crovelli, R.A., Chow, S. Grace, J.D., SindingLarsen, R., Sollie, B.H. and Wang, P.C.C., 1988. Oil and gas resource modeling and forecasting. In: C.F. Chung et al. (Editors), Quantitative analysis of Mineral and Energy Resources. Reidel, Dordrecht, pp. 695-700. Lee, P.J. and Wang, P.C.C., 1985. Prediction of oil or gas pool sizes when discovery record is available. Math. Geol., 17(2): 95-113. Lee, P.J. and Wang, P.C.C., 1986. Evaluation of petroleum resources from pool size distributions. In D.D. Rice (Editor), Oil and Gas Assessment - - Methods and Applications. AAPG Stud. Geol., 21: 33-42. Lee, P.J., Snowdon, L.R. and Wang, P.C.C., 1990. Petroleum resource evaluation, short course notes. Canadian Society of Petroleum Geology, 1990 Convention, Calgary, 108 pp. NPD (Norwegian Petroleum Directorate), 1992. NPD Annual Report, 195 pp. NPD (Norwegian Petroleum Directorate) 1993. Petroleum resources Norwegian Continental Shelf. External report, Norwegian Petroleum Directorate, 40 pp. Taylor, G.C., Procter, R.M. and Menely, R.A., 1991. Petroleum Resource Appraisal System. Geol. Surv. Can., Open-File Rep., 2374, 65 pp.
Department of Geology and Mineral Resources Engineering, The Norwegian Institute of Technology, N-7034, Trondheim, Norway Department of Geology and Mineral Resources Engineering, The Norwegian Institute of Technology, N-7034, Trondheim, Norway
11b
The Russian method for prediction of hydrocarbon resources of continental shelves, with examples from the Barents Sea K.O. Sandvik and E.V. Zakharov
The Russian method of carrying out offshore hydrocarbon resource assessments is briefly described. This method is applied in three examples based on analogies between productive units in northwest Russia and prospective units in the Norwegian Barents Sea. Although there has been little public data available to the authors in the latter area, it is believed that use of Russian methods and data might bring some new ideas about what prospect types might be productive in the Norwegian sector of the Barents Sea.
Introduction Very little co-operative work has been carried out between people doing hydrocarbon resource assessment in Russia and in the West. This largely reflects the fact that Russian/Soviet authorities have been reluctant to provide background data. Furthermore, the situation has been compounded by the lack of a common assessment language. This paper addresses both these questions. The examples presented could have been more accurate and detailed if all existing data, both Norwegian and Russian had been available. Hopefully the future will show a more open and fruitful collaboration between geoscientists of all nationalities with interests in this matter. In Russia, several organizations are involved in the development and use of hydrocarbon assessment methods. This has resulted in a voluminous domestic literature on the subject. It is not the intention of this paper to compare the Russian methods to each other, since this is a subject in itself and deserves more attention than can be given here. The first petroleum resource assessment in the USSR was made by Gubkin in 1937. Later, petroleum resource assessment was established as an integral part of the 5-year planning system starting with the assessment made in 1958. The methods used were published by Bujalov et al. (1962). Several modifications have been made and an updated version was published in 1990 by Bujalov et al. The methods described in the present paper were first published by Zakharov (1971). They have since been applied in various regions of CIS (the Barents, Kara, Caspian,
and Black Seas), other CIS countries, the former GDR, and Cuba. The basis of the Russian assessment method is the use of the specific concentration of ISR (initial oil in place) of hydrocarbons in 1000 tonnes/km 2. ISR includes both the produced hydrocarbon reserves, remaining reserves, and the potential hydrocarbon resources. Table 1 shows a comparison between the Russian classification system and other world-wide systems. The Russian nomenclature for resource classification is as follows: A: reserves produced or under production B: reserves that yielded commercial flows from wells at different depth levels. C l: reserves characterised by flow of commercial quantities of oil and gas from some wells, combined with positive geological and geophysical results in untested wells. C2: reserves in untested zones adjacent to reserves of higher categories, or in untested beds within the producing sector of a field. C3: prospective resources in mapped traps prepared for exploration drilling, or in developed fields in beds untested by drilling but proved to be productive in other fields. D l: prognostic resources in sedimentary deposits on major trends with proved commercial potential. D2: prognostic resources in sedimentary deposits within major regional structures without proved commercial potential. The three groups C3, D~, and D2 constitute the undiscovered resources and differ from each other in
Quantification and Prediction of Petroleum Resources edited by A.G. Dor6 and R. Sinding-Larsen. NPF Special Publication 6, pp. 115-122, Elsevier, Amsterdam. 9 Norwegian Petroleum Society (NPF), 1996.
116
K. O. Sandvik and E. V.. Zakharov
Table 1 Hydrocarbon reserve and resource classifications (slightly modified from Oil and Gas, 7 Oct. 1992)
Russia
France, North Germany, African Netherland,, nations
U.S., Canada, Saudi Arabia Drilled developed
Measured
proved
Prove n
Explored
Demonstrated Undeveloped
B
reserves
Proved Proved
Identified
Indicated
C1
Probable Preliminarily C2
evaluated
Prospective C3
Inferred
Probable
Possible
Non-proven reso u rces
D1
Hypothetical
D2
Speculative
Prognostic
the reliability of the information on the oil and gas bearing potential of the geological structures in the area under consideration. Four assessment levels
The assessments are made in accordance with a division into four levels: regional, sub regional, zonal, or local. At the regional level, the whole part or a major part of the sedimentary section of a basin is addressed. At a zonal level, a particular part of the sedimentary section is dealt with, i.e. Jurassic or Upper Jurassic. The four levels are described in Table 2. The validity of the prediction is not only dependent on the degree of investigation, but also on the dimensions and complexity of the geological conditions. The greater the volume of sedimentary rocks or the more complex the structure, the less the possibility of using advanced evaluation methods. For each of the four levels, there is a corresponding level of geological and geophysical input, along with information on the oil and gas bearing ability of the evaluated area. The basis for the latter input is information from analogous oil and gas bearing reference areas. Whether this evaluation gives valid values is of course dependent upon how close the similarities are between the evaluated area and the reference area. If both the evaluated and the reference units are of the same level and in the same oil and gas bearing province ("inner" analogy), the quality of the assessment will be much better than
with a reference area in another oil and gas bearing province, even if the latter is located in a similar tectonic setting or has the same age ("outer" analogy). Zakharov and Kondakov (1978) presented a new approach for systematising resource evaluations in which the prediction of oil and gas potential was linked to the actual exploration stage. This systernatisation required the following conditions to be fulfilled: (1) All available data from the area should be included in the evaluation. (2) All actual trap-types should be placed in a hierarchy dependent on their structural setting and size. (3) The prognosis assessments should be in accordance with subdivisions D1 and D2. When comparing areas from different structural settings, it is important to be aware of the transition zone between oceanic and continental crust in order to determine how far offshore one can apply onshore analogies. In principle, the geological criteria for present marine areas and adjoining onshore areas should be the same, but in practice many factors (e.g. thickness and reservoir and seal properties) may change. For instance, in the North Sea, accumulations of oil were discovered in Jurassic, Cretaceous, and Palaeogene reservoir-rocks which have no analogy in the adjoining onshore areas. When making "inner" and "outer" analogies, for offshore oil and gas resources, we should compare reference and evaluation units of similar geometry and size. The boundaries of these units should be
The Russian method for prediction of hydrocarbon resources of continental shelves
delineated similarly, i.e. based on the same principles and type of geological-geophysical data. Absolute geological analogy between two areas does not exist in nature. To take this into account, correction factors are included which consider the main calculation parameters in the evaluated areas in comparison with the reference ones. The most important parameters for calculating the prognosed resources are the extent of the prospective areas, net reservoir thicknesses, and porosity of the reservoirs. The practical value for explorationists increases dramatically when going from regional to zonal level, i.e. from dealing with a whole basin to dealing with separate reservoir successions.
Calculation of D~ and D2 resources using specific densities of ISR of hydrocarbons In Russia, the use of specific densities of ISR of hydrocarbons is the standard method for the prediction of offshore resources. Prediction for Russian offshore oil and gas provinces and basins has so far mainly been carried out at a regional level, as defined in Table 2. This places the resource assessments, with a few exceptions, within category D2. The ISR method is characterised by its simplicity and acceptable results. It is based on a transfer of the average specific ISR from the reference areas to the area under evaluation. Correction coefficients are used to compensate for estimated differences between the two areas. Of the two major correction parameters,
117
the first is linked to the thickness of the stratigraphic/ lithologic units or the net reservoir thickness. The second is connected to the porosity of the reservoir units. These evaluations are normally based on information from well logs. More information about correction coefficients has been published by Zakharov (1985). Different specific concentrations of recoverable ISR of hydrocarbons have been proposed by Weeks (1979). These are as follows: - Transition between onshore and offshore part of a basin: 35,000 tonnes oil equivalent per square kilometre (t/km2). Shelf, intercontinental and inter-shelf basins" 17,000 t / k n l 2 . - Deep-water continental slope basins: 13,000 t/ km 2. - Deep-water continental rise basins: 6000 ffkm 2. In 1986, Zakharov showed that Weeks' numbers were too high for the combined offshore/onshore basins and too low for intercontinental basins. Zakharov's numbers, resulting from studies of more than 100 basins around the world, are presented in Figs. 1 and 2 together with a division between oil and gas. In this scheme, the greatest ISRs occur in Mesozoic and Palaeozoic deposits in platform basins, and in Cenozoic sediments in intermontane depressions. The smallest values occur in intermontane basins filled by Palaeozoic deposits. When this new concept was used, the results showed a systematic tendency to somewhat downgrade the most prospective areas while upgrading the least prospective areas. -
118
K. O. Sandvik and E. V. Zakharov
Fig. 3. Sketch of how different trap types are handled in the resource prediction method. 1 = Areas with proven reservoirs; 2 = anticline; 3 = regional fault sealing the productive unit; 4 = borders of eroded trap facies; 5 -- boundary of eroded trap facies; 6 = syncline axes; I -- dome/anticlinal traps; H -- fault-sealed traps; III = stratigraphic traps - - lithology related; IV -- stratigraphic traps - - unconformity related.
Fig. 2. The ratio of ISR of oil to ISR of gas in the uppermost 7 km of basins in different settings around the world. Compiled by E.V. Zakharov, 1987.
Volumetric method to calculate D~ resources on zonal level For predicting D1 resources on the zonal level, Zakharov (1971) developed a modification of the volumetric method. When a part of a basin is investigated it is possible to estimate how much gas will be found as op-
posed to oil. When transferring this knowledge to a new basin, one should look for similarities such as presence of source rocks, reservoirs, and seals. In addition, the structural settings of the reservoirs should be compared; a Russian system is available for this comparison. In some cases, reservoirs should be compared in terms of stratigraphic conditions along secondary migration pathways. A sketch of how different trap types are handled is shown in Fig. 3. Traps may be simplified into three categories: anticlinal, fault, and stratigraphic. Stratigraphic traps are controlled by either impermeable rocks of the same age (lithologically related) or by erosion and later sedimentation of impermeable rocks (unconformity related). Anticlinal traps form both clastic and carbonate reservoirs. In fault-sealed traps, the vertical displacement has to be less than the thickness of the cap rock or greater than/equal to the thickness of the reservoir. Resource assessment made on evaluation units at the zonal level for category D1, is based on a mapping
The Russian method for prediction of hydrocarbon resources of continental shelves
of all prospects, both confirmed and unconfirmed reservoirs. For anticlinal reservoirs the predicted total resources of a new area, Q2, are calculated using the following formula:
Ql Q2 = g2" E ~ vl
(1)
where V1 is the total maximum reservoir volume taken from the previously explored reference area, V2 is the volume of the reservoirs in the new area with the same geological age as in the reference area, and Q1 is the ISR of the oil and gas detected in the reference area. The most difficult parameter to determine in faultsealed traps is the size of the area under consideration. In practice, however, statistical analyses in several areas (Volga-Ural, Pre-Caucasus, Azerbaijan, and Middle Asia), shows that the average width of this type of accumulation is relatively small (in the range of 1.5-2.5 km). For the fault-sealed types, the following modifications to Eq. (1) are recommended:
Q2- ~Q1
L2.K1-K2
(2)
L1 where L1 is the fault displacement in the reference area, L2 is the corresponding length of the evaluated area, K1 is the ratio of the oil/gas column in the new area compared to the same thickness in the reference area (heval/href), K2 is the ratio of the porosity of the reservoirs in the new area and the porosity in the reference area (4~eval/4~ref). The resources in stratigraphic traps may be estimated using the following formula:
Q2 -~ A 9d . h .q -K2
(3)
where A is total area of the supposed prospective zone in km 2, d is ratio of the sum of productive trap areas found in the reference area to the total area of the reference area, h is expected net reservoir thickness in the evaluated area (in metres), q is average specific initial oil and gas reserves per km 2 and metre of the saturated reservoir thickness as measured in the reference area. The main sources of error in predicting hydrocarbon resources are: (1) insufficient level of study of both the reference and evaluation units; (2) errors in the determination of input parameters It is necessary to emphasize that the coefficient of productive area [d in Eq. (3)] can only be used in the case of proven or supposed non-uniform areal distribution of reservoir rocks. Correction factors for changes in reservoir rock thickness (K2) should be taken into consideration only in the case of a close
119
lithological relationship between deposits in reference and evaluated areas. Within each particular region, or separate stratigraphic level of the region, the character of the correction coefficients may differ. Most informative are those parameters which have the closest correlation to already proven reserves. Correlations carried out at VNIIGAS (Zakharov and Yudin, 1988) have shown the following: - T h e variation of the most informative parameters is related to the tectonic style and geological evolution of the area. - We cannot exclude that in some regions, some important parameters are not taken into account (because of insufficient study at the time of estimation).
Evaluation of C3 and D~ resources on local level In Russia, evaluation of resources of categories C3 and predicted resources D1 in fields prepared for drilling or local structures are made using the volumetric method of Zhdanov et al. (1966).
Generation and accumulation potential When making prognoses at zonal and local levels, the accumulation potential is the main parameter used as input to the assessment of the area under evaluation. When dealing with regional units, such as basins, we must take into account both generation and accumulation potentials in the form of generalised indices. The generation potential of sedimentary deposits is found to be directly related to the sedimentation rate of these deposits (in meters/million years). The accumulation potential of the same deposit is characterised by the ratio of hydrocarbon ISR to the average summary thickness of the individual reservoir beds. This can be determined in areas with proven oil and gas accumulations. It is given in million tonnes HC/metre. These generalised parameters, multiplied by each other, provide an estimate of what proportion of the prognosed resources is confined to 1 million years of deposition for a particular stratigraphic subdivision (in million tonnes HC/million years). For less investigated subdivisions of sedimentary sections, the generation and accumulation potential can be determined using the ratio between the average ISR in the reference units and the total reservoir thickness in evaluation unit. This thickness is multiplied by a factor which describes the most likely proportion of reservoirs in the basin under examination. The latter factor changes according to the dominant facies of the deposits. Gomelkova et al.
120
(1978) recommend these factors to be as follows: - marine clastics 25-30%, - marine carbonates 5-15% (for reefal varieties 3540%), - mixed marine clastics and carbonates 20%, evaporitic deposits 10%, continental-marine deposits 1-3%. In cases where various lithofacies are interbedded, it is recommended that weighted values representing the different facies be used. After this procedure, one can estimate specific values of each of the complexes (in%) and with this result distribute the predicted resources in the whole sedimentary section. -
-
Examples from the Norwegian Barents Sea The reported methods have been applied to resource prognoses for both Russian and Norwegian areas of the Barents Shelf. In the selection of the analogies, the following conditions were taken into account: tectonic setting, geological evolution, types of traps, lithology/ sedimentary facies, and the phase condition of the HC accumulation. In the described examples, the analogies have been applied to separate lithological units and not to the complete sedimentary section. With more detailed geologic input from the assessed Norwegian areas a less fragmentary approach could have been applied. Examples are described for three units. The areas are defined on Fig. 4, which is taken from Zakharov et al. (1993).
Example 1: Loppa High (A) Fault-sealed, dolomitized limestones of Lower Carboniferous age in the northern part of the Loppa High evaluated with reference to Serpukhovian limestones of the Sorokin Bar in the Timan-Pechora Basin. In the reference area, the Serpukhovian limestone of the Sorokin Bar, the ISR of HC is 3.5 tonnes/km 2. The input to Eq. (2) is L1 -- 55 km, L2 = 53 kin, K1 = 4.0/5.1, and K2 -- 8%/15%. D1 resources of this part of the Loppa High are thereby calculated to be 1.4 million tonnes (10.4 million barrels) of oil. (B) Anticlinal traps with Middle CarboniferousLower Permian carbonates and sandstones of the Loppa High, evaluated with reference to the Kolvin Bar in the Timan-Pechora Basin. The published data from the Loppa High contain little information on the quality and distribution of the reservoir rocks. Based on maps and profiles presented by Gabrielsen et al. (1990), the volume of the possible
K.O. Sandvik and E.V. Zakharov
oil-filled pores in the Middle Carboniferous-Lower Permian carbonates and sandstones on the Loppa High may be in the order of 6 km 3. (V2 = A . d . h . q 5 = 3892 k n l 2 90 . 5 90.022 k m . 0.14 = 5.994 km 3. Factor d refers to the proportion of the total area that is prospective. In this case, d is the same as in the reference area.) For the reference area, Kolvin Bar, the prognosed resources are calculated to be 31.2 million tonnes of oil and the volume of the possible oil filled pores, V1, is estimated to be 6.320 km 3. The anticlinal D2 resources on the Loppa High are then estimated to be: 5.994-31.2/6.320 = 29.6 million tonnes of oil (219 million barrels). (C) Total resources at Loppa High. Summing the above results the total estimated resources on the Loppa High are 31 million tonnes. With a recovery factor of 0.33, this equates to ca. 10 million tonnes of recoverable oil on the Loppa High.
Example 2: Mercurius High Triassic stratigraphic traps on the Mercurius High, evaluated with reference to the northern Kildin Nose. The reference area, i.e. Northern Kildin Nose, has an area of 1875 km 2. Only 25% of the traps are filled. The average effective HC saturated thickness is 8.4 m, the porosity is 19%, and the average concentration of ISR is 65,000 tonnesNm 2. The total area of the Mercurius High is 6900 km 2. The extension of the Mercurius High is taken from Zakharov et al. (1993). If only 25% of this area is prospective, as in the reference area, the effective area will be 1750 km 2. The effective average thickness of the reservoir is estimated to be 4.3 m and the porosity to be 25%. The prognostic resources of Mercurius High are then estimated to be: 1750-4, 3/8, 4-65,000.25/19 = 77 million tonnes oil equivalent (t.o.e.). In the reference area, the ratio between oil and gas + condensate is estimated to be 0.22. Using this figure on the Mercurius High gives total oil resources of 17 million tonnes. With a recovery factor of 0.33 the recoverable resources will be in the order of 5.5 million tonnes. The total resources of gas + condensate are 60 billion m 3 of gas and 2 million tonnes condensate. With a recovery factor of 0.85 for gas and 0.65 for condensate, the recoverable gas and condensate will be in the order of 51 billion m 3 and 1 million tonnes respectively.
Example 3: West Malingiskaya Saddle (A) West Malinginskaya Saddle evaluated with reference to the Ludlov Saddle.
The Russian method for prediction of hydrocarbon resources of continental shelves
121
Fig. 4. Example areas. 3 -- West Maliginskaya Saddle; 8 = Loppa High; 9 = Mercurius High. From Zakharov et al. (1993).
One discovery, Ludlovskaya, has been made on the Ludlov Saddle. In this discovery, the thickness of the Jurassic is 800 m, out of which 90 m is saturated with hydrocarbons. From Johansen et al. (1993), the thickness of the Jurassic on the West Malinginskaya Saddle is estimated to be 180 m. Applying the same saturation ratio as at the Ludlov Saddle, the saturated thickness will be 20.25 m. The prospective part of the Ludlov Saddle has an area of 38,965 km 2, and at the West Malinginskaya Saddle, 9062 km 2. D1 resources of the Ludlov Saddle are calculated to be 750 million t.o.e. The resources of West Malinginskaya Saddle will then be: D 2 = 7 5 0 . 2 0 . 2 5 / 9 0 . 9062/38,956 = 39.68 million t.o.e. (B) West Malinginskaya Saddle evaluated with reference to the Eastern Fedinsky Nose. Two discoveries have been made on the Eastern Fedinsky Nose, Shtokmanovskaya and Ledovoye.
Here the average concentration of ISR is 90,000 t.o.e./km 2. The prospective part of the structure represents 25% of the total area and the saturated thickness is 90 m. The prospective part of West Malinginskaya Saddle is: 9062 km 2. 0.25 = 2265 klTl2 Alternatively, using the same saturated thickness as above, i.e. 20.25 m, the resources on the West Malinginsakya saddle can be calculated as: D 2 = 2 2 6 5 . 2 0 . 2 5 / 9 0 . 9 0 , 0 0 0 = 45.9 million t.o.e. (C) Resources of West Malinginskaya Saddle. From (A) and (B) the resources of the West Malinginskaya Saddle are estimated to be greater than 39.7 million t.o.e., or 39.7 billion m 3 gas equivalent since the Russian discoveries have shown that the fluids are gas and condensate. The condensate content is 50 g/m 3 gas. By using a recovery factor of 0.85 for gas and 0.65 for condensate, the resources of the West
K.O. Sandvik and E.V. Zakharov
1 22
Malinginskaya Saddle are estimated to be in the order of 37 billion m 3 of free gas and 1.5 million tonnes of condensate.
Conclusions Exploration success in the Norwegian Barents Sea to date has been disappointing compared to the discoveries made in the northwestern part of Russia. This concerns both Late Palaeozoic and Mesozoic hydrocarbon accumulations. Only rudimentary geological information was available for the three Norwegian examples presented in this paper: Loppa High, Mercurius High, and West Malinginskaya Saddle. In spite of this, the resource estimates correlate well with Norwegian exploration results. The total recoverable resources of investigated sections of the Loppa High are estimated to be 10 million t.o.e., on the Mercurius High 51 billion m 3 gas, and 1 million tonnes condensate, and on the West Malinginskaya Saddle 37 billion m 3 gas and 1.5 tonnes of condensate. More detailed geological information from these areas would probably improve the accuracy of these estimates. It is an obvious task for geoscientists experienced in both Western and Russian assessment methods and with access to detailed exploration results from northwest Russia, to co-operate in improving the reliability of estimates of the hydrocarbon potential of the Norwegian Barents Sea.
References Bujalov, N.I., Vasilyev, V.G., Erofeeva, N.S. et al., 1962. Methodology of Evaluation of Oil and Gas Predicted Reserves. Gostoptechizdat, Moscow, 82 pp. (in Russian). Bujalov, N.I., Vinnikovsky, S.A., Zakharov, E.V., Kontorovich, A.E.
K.O. SANDVIK E.V. ZAKHAROV
et al., 1990. Methodological Basis for the Prediction of Oil and Gas Potential. Nedra, Moscow, 248 pp. (in Russian). Gabrielsen, R.H., Fa~rseth, R.B., Jensen, L.N. Kalheim, J.E. and Riis, F., 1990. Structural Elements of the Norwegian Continental Shelf, Part 1. The Barents Sea Region. Norwegian Petroleum Directorate, Bull., 6, 33 pp. Gomelkova, N.R, Modelevsky, M.S. and Polster, L.A., 1978. Some characteristics of sedimentary thickness of oil and gas bearing basins. In: Modern Problems of Oil and Gas Geology. MGU, Moscow, pp. 165-172 (in Russian). Johansen, S.E., Ostisty, B.K., Birkeland, 0., Federovsky, Y.E, Martirosjan, V.N., Bruun Christensen, O., Cheredeev, S.I., Ignatenko, E.A. and Margulis, L.S., 1993. Hydrocarbon potential in the Barents Sea: play distribution and potential. In: T.O. Vorren, E. Bergsager, O.A. Dahl-Stamnes, E. Holter, B. Johansen, E. Lie and T.B. Lund (Editors), Arctic Geology and Petroleum Potential. Norwegian Petroleum Society (NPF), Special Publication No. 2. Elsvier, Amsterdam, pp. 273-320. Weeks, L.G., 1979. Geology of Continental Margins, Vol. 3. Mir, Moscow, pp. 313-327 (in Russian). Zakharov, E.V., 1971. Methods for prediction of reserves of oil and gas of subgroup DI . Geol. Oil Gas, 4: 48-51. Zakharov, E.V., 1985. The account of analogy when predicting oil and gas resources. J. Azerbaijan's Oil Ind., 9: 12-14. Zakharov, E.V., 1986. The Specifics of Continental Shelfs Oil an Gas Bearing Ability Prediction. Survey Information Series. Geology and Survey of Marine Oil and Gas Fields. M., VNIIEGAZPROM, 196, 1, 44 pp. (in Russian). Zakharov, E.V. and Kondakov, A.V., 1978. Systematization of prediction for oil and gas bearing units. Oil Gas Geol. Geophys., 9: 17-22 (in Russian). Zakharov, E.V. and Yudin, S.G., 1988. Geological appearance of predicted hydrocarbon resources in sedimentary deposits. J. Azerbaijan's Oil Ind., 4: 5-9. Zakharov, E.V., Kulibakina, I.B. and Bogoslovskaya, G.N. 1993. The Jurassic complex is an object of oil and gas prospecting in the Barents Sea. In: T.O. Vorren, E. Bergsager, O.A. Dahl-Stamnes, E. Holter, B. Johansen, E. Lie and T.B. Lund (Editors), Arctic Geology and Petroleum Potential. Norwegian Petroleum Society (NPF), Special Publication No. 2. Elsevier, Amsterdam, pp. 257260. Zhdanov, M.A., Grishin, F.A. and Gordinsky, E.V., 1966. Geological Reservoir Evaluation of Oil and Gas Fields. Nedra, Moscow, pp. 158-159.
IKU Petroleum Research, N-7034 Trondheim, Norway VNlIGAS International, p. Razvilka, Leninsky raion, Moskovskaya oblast, 142717, Russia
123
Offshore Brazil: analysis of a successful strategy for reserve and production growth S. Jack Morbey
Major oil companies cannot continue to tolerate the high cost of exploration failure in a world of low future oil prices. There is a trend in the industry to a rationalization of worldwide exploration and production portfolios. Companies are looking to reduce their cost of finding by lessening the risk of failure at all levels of the exploration process, and stimulating resource replacement and growth through the efficient exploration of basins and plays. The coastal basins of Brazil offer an opportunity to assess the effectiveness of turbidite exploration along a passive continental margin, where non-marine, shelf, and deepwater exploration and production have taken place. The predominant role and strategy of Petrobras, the Brazilian national oil company, in the development of exploration offshore Brazil is worth looking at closely. When oil companies are planning new country-new basin ventures in passive margin deeper water environments worldwide, a lot can be learned from comparisons with similar but mature exploration environments with successful proven petroleum systems. Petrobras are showing the way forward in efficient, lower-risk, passive margin exploration and exploitation at more acceptable costs of E&E An exploration strategy based on play focus, play environment, and deepwater technology development seems to be working in the Campos Basin, offshore Brazil. Driven by a desire for Brazilian energy self-sufficiency before the year 2000, Petrobras' mission is to achieve significant reserve and production growth through an exploration focus on the deep water Campos Basin. Exploration and production success in the Campos Basin is enabling Petrobras to manage their exploration portfolio better, and to minimise the cost of failure in other passive margin basins. A thematic statistical approach to the analysis of future play potential has given rise to a clearer assessment of offshore exploration risk, particularly exploration on the shelf versus in deeper water. This has helped Petrobras to develop a more focused exploration strategy, and has enabled them to allocate their E&P annual budgets more cost efficiently in their drive for oil self-sufficiency. New windows of opportunity are now opening for exploration success in Brazilian passive margin coastal basins hitherto considered past their best. The recent ability of Petrobras to explore outside the Campos Basin in new deeper water frontier passive margin basins, and achieve early discovery success, shows the value of experience and strategic focus. Worldwide, the exploration for new, often subtle, clastic turbidite plays is a potentially high cost enterprise. Passive margin exploration is moving swiftly into deeper waters of the South Atlantic off West Africa, as well as the North Atlantic West of the Shetlands. These areas offer a high-risk, high-reward challenge, and drilling environments in need of innovative new E&P technology and creative initial exploration strategies. Brazil, through its national oil company Petrobras, leads the way in this effort.
Country focus The vulnerability of the Brazilian economy to fluctuations in world oil prices, in addition to government-subsidised "below market" oil prices and recent hyperinflation, have spurred a drive for selfsufficiency and the expansion of energy resources. Today, Brazil is the top non-OPEC replacer of its production worldwide. This follows significant annual oil and gas reserve additions and production increases since 1980 (Fig. 1), which are associated with considerable reductions in exploratory well drilling, a marked increase in discovery rates, and the ongoing control of exploration and production costs. Petrobras currently pursue a program of exploration and production ventures focused mainly on
the offshore and the challenging deeper water areas of the Campos Basin. This has been achieved through a highly focused exploration and production strategy. High rates of exploratory drilling success in 1992 of 35% offshore and 25% onshore reflect the impact of this strategy. The accelerated growth in domestic proven oil and gas reserves in the last 15 years is the result of a significant change in offshore exploration strategy towards exploration in turbidite plays. Resource growth has been fueled by ongoing deeper water exploration and the development of new deep water production technology in the Campos Basin. At the present time, offshore reserves amount to 87.5% of the total Brazil reserves, and 77% of these offshore reserves are to be found in water
Quantification and Prediction of Petroleum Resources edited by A.G. Dor6 and R. Sinding-Larsen. NPF Special Publication 6, pp. 123-133, Elsevier, Amsterdam. 9 Norwegian Petroleum Society (NPF), 1996.
S.J. Morbey
124
Fig. 2. Brazil - - coastal basins.
Offshore Brazil: analysis of a successful strategy for reserve and production growth
125
Exploration efficiency
Fig. 3. Brazil m passive margin sequences.
deeper than 400 m in the Campos Basin (Fig. 2). The Campos Basin currently accounts for 65% or 470,000 b/d of Brazil's total oil production, and 38% of Brazil's total gas production, largely from clastic turbidite reservoirs of Tertiary age. Rift, pre-rift, and transitional post-rift reservoirs of the Brazilian coastal basins (Fig. 3) were the mainstay of exploration and production from 1938 to 1980 onshore and in the shallow water offshore areas of the continental shelf. However, internal Brazilian oil supply over this period proved inadequate given a rise in downstream demand. A 10 year decline in oil and gas resource growth from 1970 to 1980 (Fig. 1) saw reserve replacement fail to keep up with the growing production needs of Brazil. This failure was a function of the overall high costs of finding and development, the nature of the plays and their reservoir quality, and the discovery of mainly small fields of insufficient size (<50 MMBOE) to impact the supply/demand ratio.
In pursuit of a politically driven goal of oil selfsufficiency by the year 2000, Petrobras have become a world leader in deep water technology, exploration, and production. Technology development aligned to E&P cost control, exploration expansion, and focus on key basin and play environments is driving the Brazilian oil sector with Petrobras dominating the upstream and downstream energy market. Their recent success in the upstream domain has been their ability to apply and refine exploration strategies from their mature basins to rift and post-rift turbidite sequences and reservoirs in adjacent poorly explored basins. Petrobras have been under considerable pressure from the Brazilian government to efficiently find large reserves at commercially acceptable costs of exploration and production. They have reduced their costs of finding and exploitation as they have become more efficient in their exploration process. The cost of exploration per barrel discovered is down from US$ 3.00 in 1991 to US$ 2.00 in 1992. The production cost for the deepwater "supergiant" Marlim Field located at a water depth of 600-790 m in the Campos Basin has been put at US$ 9.8 per barrel (Freire, 1990). Since 1980, Petrobras have reduced significantly the number of exploration wells completed both onshore and offshore, whilst achieving high discovery rates and large reserve additions from post-rift Tertiary deep water submarine fan turbidite reservoirs in the Campos Basin (Fig. 4). Over this period, they have demonstrated a high degree of awareness of the requirements for consistent ongoing exploration success in passive margin rift sequences (see Bruhn et al., 1988; Bruhn, 1990; Figueiredo and Martins, 1990; Mello et al., 1992). Over 95% of current Brazilian oil and gas production comes from the Atlantic passive margin coastal basins. Brazil still has outstanding potential for giant hydrocarbon discoveries (Davison, 1991). Petrobras have developed a highly successful and focused exploration strategy, supported in recent years by the application of play-specific new technology such as 3D seismic imaging and amplitude analysis. Oil and gas reserves increased in Brazil by 19.6% and 10.4% respectively during 1992 with 593 million barrels of oil (including condensate) and 455 BCF of natural gas added. This reserve growth, however, excludes the very deep water oil accumulations discovered at water depths greater than 1000 m.
Thematic exploration analysis Thematic exploration analysis looks at the way in which countries, through their state or national
1
2
6
S
.
J
.
Morbey
Fig. 4. Campos Basin m Proven Reserves by Play 1981-1992.
oil companies (NOC's) and foreign multi-national oil companies, have historically undertaken and managed their exploration, achieved commercial success, and significantly increased their oil and gas resource base. A thematic approach to the assessment of exploration risk enables a clearer understanding of where and how future potential economic success can be achieved for individual petroleum plays (e.g. Holtz, 1993). This type of detailed statistical analysis examines the petroleum system in a basin, the impact of exploration over time, the changing nature of where exploration has taken place (e.g. offshore versus onshore, deep water versus shallow water), and the overall cost of exploration and production success per barrel. A company strategy can then be developed that addresses the type of exploration exposure and technology program needed to reduce future exploration risk in specific play environments. Priority can then be given to investment opportunities, where the risk of exploration failure is manageable, and the commercial success potential high. Petrobras' success at resource growth and E&P cost reduction in the coastal basins of Brazil is based on a focus to key basins and play environments, and the application of new technologies. They are using their experience and monopoly status most effectively in their management of exploration risk, particularly in deep water. Their current success has been achieved through better portfolio management, based to some degree on internal thematic statistical analyses of oil and gas play
success and failure per basin and tectonic setting (see Bruhn et al., 1988; Bruhn, 1990; Figueiredo and Martins, 1990; Mello et al., 1992).
Campos Basin - - exploration and production success
The Campos Basin is an excellent role model for passive margin exploration strategy analysis (see Edwards and Santogrossi, 1990). The exploration strategy of Petrobras in the Campos Basin has changed emphasis totally over the last twenty years from basal post-rift post-salt Albian marine carbonate targets on the continental shelf to post-rift/drift Cretaceous and Tertiary turbidite plays in greater than 400 rn water depth (Figueiredo and Martins, 1990; Guardado et al., 1990). Of the 22 exploratory wells drilled in 1992 in the offshore Campos Basin, 19 were in water depths greater than 400 m, and only 1 well addressed an exploration target on the shelf. A comparison with the Kwanza and Congo passive margin basins, offshore West Africa shows only 2 of the 35 exploratory wells drilled in 1992 in deepwater beyond the 200 m shelf break with no exploration beyond 400 m water depth. The majority of the wells in these West Africa basins were drilled on the continental shelf in less than 100 m water depth. Since 1974, the continued recognition of new play types in the Campos Basin associated with post-rift turbidites has stimulated the trend to deeper water,
Offshore Brazil: analysis of a successful strategy for reserve and production growth larger reserve, exploration and production for both oil and gas close to the Rio-Sao Paulo industrial market. Pre-salt syn-rift and pre-rift shelf exploration has proven over time to add little resource value with small reserves and a high finding cost. The discovery of the Namorado Field in 1975 at a water depth of 166 m with recoverable oil reserves of 250 million barrels in Cenomanian/Turonian marine turbidites (Bococcoli et al., 1980) marked a watershed shift to turbidite exploration. This discovery still represents the largest oil find to date on the Brazilian continental shelf, and marked the point at which Petrobras began to seriously look at the merits of marine post-rift Cretaceous and Tertiary turbidite exploration for resource growth from increasing water depths. The shift in play focus in the Campos Basin to deepwater and Tertiary turbidites has had a significant impact. These post-rift/drift Tertiary reservoirs represent an exploration effort focused since the early 1980's on subtly developed combination traps with depositional, stratigraphic, and/or structural seals. The pre-1980 historically "successful" Albian post-salt carbonate play, pre-salt lacustrine limestone coquina play, and pre-rift volcanic play are now almost redundant for oil exploration in the Campos Basin. Furthermore, Szatmari et al. (1985) have highlighted a critical hydrocarbon success factor to be the relationship between key tectonic fault trends and fault types and hydrocarbon migration and entrapment in the coastal basins of Brazil.
127
By 1989, Petrobras had discovered 32.0 BBOEIP (billion barrels of oil equivalent in place) in the Campos Basin (Figueiredo and Martins, 1990). The majority (65%+) of the proven Brazil reserves of 4.5 BBOE have been discovered in Campos Basin "giant" Tertiary oil fields of Eocene to Miocene age. An additional 4-6 BBO have also been "discovered" recently as potentially exploitable reserves in the Campos Basin in water deeper than 400 m. The associated and non-associated gas potential of these deeper water discoveries may exceed 5-10 TCF recoverable. Campos Basin - - deep water
Exploration and production activity in water depths beyond the continental shelf break started in 1980 and 1983, respectively. Since 1984, three deep water "giant" fields have been discovered in Tertiary turbidites in the Campos Basin at water depths of between 300 and 2000 m: the Marlim (8.2 BBOIP - - Oligocene), Albacora (4.1 BBOIP m Miocene to Albian), and Barracuda (3.3 BBOIP m Eocene) fields (Fig. 5). These three fields contain the majority of the current country oil reserve base with production currently viable down to a water depth of 1000 m. Petrobras are currently developing the technology to exploit oil reserves at very deep water depths of 1000 to 2000 m by 1996, and forecast 60% of new reserve additions and 60% of total production to be from deep and very deep water greater than 400 m by
Fig. 5. CamposBasin -- deepwater"giant" fields. Oil in place per play.
1
2
8
S
.
the year 2000 (Franca et al., 1992). Petrobras started oil production in April 1994 from a well at 1027 m water depth, a world record depth for offshore oil extraction. The discoveries of the "giant" Albacora and Marlim Fields in 1984 and 1985, and Barracuda in 1991, have proven to be the turning point for Petrobras in their drive for self-sufficiency (Souza et al., 1989; Peres, 1993). These discoveries have critically impacted Petrobras' exploration strategy, and are of vital significance to the future of the Brazilian economy. Petrobras have recently announced the discovery of four new giant offshore oil fields in the Campos Basin, all probably in turbidite reservoirs, with estimated total reserves of 1.0 billion barrels. These new fields increase Brazil's proven and likely oil reserves from 8.8 BBO to 9.8 BBO. By the year 2000, production from these four fields will enable total production to reach 1.5 million b/d or 90% of Brazil's projected demand.
Recent trends Since 1987, only two Brazilian basins (Campos and Santos) have demonstrated an offshore discovery success rate greater than 20% based on wildcat wells (Fig. 6). A comparison of offshore well discovery with water depth over this period 1987 to
J
.
Morbey
1992 clearly demonstrates the success of Petrobras' exploration strategy in water deeper than 400 m, and their continuing lack of success and inefficiency in shelf and slope exploration at water depths less than 400 m. This environment of drilling factor is important. It demonstrates that the risk of failure and the cost of exploration is currently very high on the continental shelf relative to deep water (Fig. 7). This is also the case offshore West Africa, particularly the passive margin shelf offshore Gabon and Congo, but not the Angola shelf with presumably superior source rocks in the lacustrine rift sequence (see Fig. 12). Along the Brazilian coastal passive margin, continental shelf exploration has focused on higher-risk syn-rift pre-salt and transitional post-rift top post-salt sequences, with their inherent seismic imaging, reservoir quality and definition problems, and restricted facies extent at depths of 500 to 3000 m below sea level. Deeper continental slope and abyssal plain exploration appears to offer lower-risk higher-reward post-rift/drift exploration for subtle turbidite traps of significant potential extent and high reservoir quality at depths of 2500 to 4000 m below sea level (Fig. 8) (see Bruhn, 1990; Abreu and Savini, 1994). Oil discoveries on the shelf rarely hold reserves greater than 100 MMBO, whilst the deep water offers significant potential for discoveries of greater than
Fig. 6. Brazil -- offshorecommercialsuccess per basin 1987-1992.
Offshore Brazil: analysis of a successful strategy for reserve and production growth
129
Fig. 8. Brazil n coastal sedimentary basin play strategy (after Petrobras).
500 MMBO. Consequently, if the costs of finding and exploitation can be managed in deepwater through efficient portfolio management, realistic initial work programs, and technology usage, the economic risks of exploration in deepwater areas of passive margin basins should produce an acceptable return on capital employed. The "giant" deepwater discoveries in the Campos Basin are about to come onstream. The Campos Basin
has experienced an impressive deeper water finding rate with 82% of the offshore discoveries over the six years 1987-1992 in water deeper than 400 m. The very deep water (> 1000 m water depth) has experienced a 66% plus discovery rate. This compares to only a 16% discovery rate for the continental shelf (Fig. 9). An analysis of exploration success and cumulative reserve additions of between 20 and 500 MMBOE shows that, since 1980, the Campos Basin
130
S.J.
Morbey
areas of the Campos Basin. The best potential for significant near term oil discoveries and resource growth in the Brazilian coastal basins lies in the Campos Basin, where there is still considerable opportunity, but at water depths of 1000 m to 2500 m. The Campos and Santos Basins have achieved collectively over 75% of Brazil's recent coastal basin success. The other coastal basins have had few discoveries, comparatively little hydrocarbon success, and minor reserve growth (Fig. 11). As Petrobras begin to explore beyond the continental shelf in other Brazilian coastal basins, the historical experience Fig. 9. Campos Basin m commercial success vs water depth 1 9 8 7 1992.
has dominated oil and gas resource growth for all Brazilian coastal basins (Fig. 10). The Campos Basin represents a significant role model for passive margin exploration success analysis (Pereira and Macedo, 1990).
Future trends Petrobras are currently achieving considerable deep water success with the ongoing discovery of large new oil and gas reserves. This is largely a consequence of their strategy to focus exploration and production ventures on the challenging deeper water
Fig. 11. Brazil -- offshore commercial discoveries by well per basin 1987-1992.
Fig. 10. Brazil coastal basins J cumulativeproven reserves 1980-1992.
Offshore Brazil: analysis of a successful strategy for reserve and production growth
gained from the Campos Basin over 30 years of passive margin exploration success and failure is proving invaluable. The application of a thematic approach to exploration is now beginning to pay off in exploration cost efficiency in adjacent Brazilian Atlantic coastal passive margin basins, where sustained commercially viable resource growth has to date proven elusive. Recent exploration in the deep water frontier of the Sergipe-Alagoas Basin has led to the discovery in 1987 of tested oil reserves at a water depth of 1100 m (well 1-SES-92), and the first deep water oil accumulation outside the Campos Basin. A recent well 1-SES-106 completed in January 1993 at a water depth of 779 m has shown further encouragement for future commercial success. The well flowed 1540 BOPD of 32 API oil. These recent, as yet, "non-commercial" deep water discoveries rank as a significant strategic success for Petrobras in a new deep water frontier basin with little available well control beyond the shelf break. The main reservoir objectives are Upper Cretaceous to Oligo-Miocene clastic turbidites as in the Campos Basin (see Van der Ven et al., 1989; Aquino and Lana, 1990). The key to continuing passive margin success and Brazilian growth to self-sufficiency in hydrocarbons lies with a continued reduction in the cost of finding by minimizing exploration failure, and the development of cost efficient technology to explore for and produce from very deep water. The offshore deep water areas of Brazilian basins have a considerable potential remaining reserve, presently estimated at about 40 billion barrels of oil in place (Franke, 1992). In addition, future deep gas exploration close to a ready onshore industrial market appears to be a major exploration and production growth area for the late 1990's and beyond (see Bruhn, 1990). In line with this new strategic direction, Petrobras have allocated short term budget funds of US$ 300 MM for natural gas and light oil/condensate exploration of deep shelf and slope reservoirs below a depth of 4000 m subsea in the Santos Basin. I envision a major expansion in natural gas exploration and production over the next 10 years in both the Santos and Campos Basins, geared to the deep continental shelf presalt and transitional post-rift turbidite sequences of Cretaceous age, and the deeper water clastic turbidites of Tertiary age.
131
passive margin (see Clifford, 1986). Consequently, in comparison with Brazil's rush over the last 15 years to reserve growth through successful deeper water Tertiary exploration and exploitation, deep water exploration is still in its infancy offshore West Africa (Akinosho, 1993b; Perrodon, 1993). Some may argue that West African governments have historically created insufficient financial incentives to attract multinational oil companies to explore in these costly but attractive deeper water areas. Exploration offshore West Africa has yet to really address deep water post-rift Tertiary submarine fan plays with a cohesive strategy based on a realistic awareness of the cost of exploration and production success in a present-day deep water environment. The costs of exploration failure and the production technology required to exploit a new reserve find have to be thought out carefully prior to signing an E&P contract with a host government, and committing to a particular exploration program. Relatively few new wildcat wells have been drilled in deep water offshore West Africa, and no significant deep water discoveries have been made. Coastal passive margin basins like the Lower Congo and Kwanza basins off Angola have seen few offshore exploratory wells drilled beyond a water depth of 200 m with the majority of wells at water depths less than 100 m in recent years (Fig. 12). Most of the West African passive margin fields discovered in the salt basins from Cameroon to Angola are in the rift/pre-salt lacustrine or post-salt transitional marine post-rift sequences of Cretaceous age with median reserve sizes between 40 and 74 MMBOE (Clifford, 1986). The Atlantic margin basins of West Africa have produced 8.1 BBO from Cretaceous lacustrine rift sequences (Perrodon, 1993). A thematic analysis of play success by Teisserenc and Villemin (1990) for Gabon has shown that 60% of the total recoverable reserves of 1.8 BBO were being produced from Upper Cretaceous post-rift/drift turbidites (Batanga and Anguille). The pre-salt Aptian fluvial sandstones of the transitional rift sequence (Gamba) accounted for 36% of the reserves, and less than 4% of production came from the Tertiary. They commented that if future economic conditions permit deep water oil production, a comparable Tertiary turbidite reserve potential to the Campos Basin could develop off Gabon.
Success and failure --- a comparison with West Africa
Summary
Unlike Brazil, no one country, political ideology, or NOC (national oil company) in West Africa has influenced exploration strategy within and between the various African coastal basins of the South Atlantic
Major oil companies cannot continue to tolerate the high cost of exploration failure in a world of low future oil prices. There is a trend in the industry to a rationalization of worldwide exploration and pro-
S.J. Morbey
1 32
Fig. 12. Angola m offshore exploratory well success 1987-1992.
duction portfolios. Companies are looking at ways of lessening their risk of failure at all levels of the exploration process, and stimulating resource replacement and growth through the efficient exploration of basins and plays. The Brazilian exploration experience, particularly in the Campos Basin, is worth looking at closely when oil companies are planning new country-new basin ventures in passive margin basin settings worldwide. Petrobras are showing the way forward in successful deeper water, lower-risk, passive margin exploration and exploitation at increasingly acceptable costs of E&E An exploration strategy based on play focus, play environment, and deep water technology development seems to be working for Petrobras offshore Brazil in their drive for self-sufficiency. The ability of the company to explore in a new frontier basin and minimise the risk of failure through the application of an efficient thematic exploration strategy is worthy of note. New windows of opportunity are opening for exploration success in Brazilian passive margin basins hitherto considered passed their best. Exploration for clastic turbidite plays is a potentially high-cost enterprise. Exploration is now moving swiftly to target these plays in deeper waters of the South Atlantic off Angola and Nigeria, West Africa, as well as in the North Atlantic West of the Shetlands. These areas and other passive margin basins
worldwide offer a high-risk, high-reward challenge, and drilling environments in need of innovative new E&P technology. They also require thoughtful initial exploration programs and strategies. Brazil, through its national oil company Petrobras, leads the way in this effort.
Acknowledgements I would like to thank Amoco Exploration and Production for permission to publish this paper. The views expressed in this paper represent a personal opinion of the exploration process in the South Atlantic.
References Abreu, V.S. and Savini, R.R., 1994. Major paleoceanographic events of the Brazilian continental margin: relationships with the giant oil fields of the Oligo-Miocene of the Campos Basin, Brazil. 26th Annual OTC (OTC 7411), Offshore Technology Conference, Houston, pp. 335-343. Akanni, F., 1993. Nigeria's focus now offshore. Offshore/Oilman, 53(7): 10. Akinosho, T., 1993a. West Africa's deepwater basins may prove very productive. Offshore/Oilman, 53(3): 27-29. Akinosho, T., 1993b. Operators still optimistic despite NNPC share funding reduction. Offshore/Oilman, 53(9): 57. Anonymous, 1994. Brazil. Oil Gas J., 92(35): 44-47. Anonymous, 1994. Brazil's development plan centers on deepwater oil. Oil Gas J., 92(1): 18-20.
Offshore Brazil: analysis of a successful strategy for reserve and production growth Aquino, G.S. de and Lana, M. da C., 1990. Sergipe-Alagoas Basin: current exploration status. B. Geoci. Petrobras, 4(1): 75-84. Bococcoli, G., Morales, R.G. and Campos, O.A.J., 1980. The Namorado oil field, a major discovery in the Campos basin, Brazil. AAPG Mem., 30: 329-338. Boeuf, M.A.G., Cliff, W.J. and Hombroek, J.A.R., 1992. Discovery and development of the Rabi-Kounga field: A giant oil field in a rift basin onshore Gabon. Proc. 13th World Petroleum Congress, Buenos Aires, 1991. J. Wiley and Sons, New York, Vol. 1, Pt., 4, pp. 33-46. Bruhn, C.H.L., 1990. Deep hydrocarbon reservoirs: the next Brazilian exploration frontier? B. Geoci. Petrobras, 4(4): 349-386. Bruhn, C.H.L., Cainelli, C. and Matos, R.M.D. de, 1988. Habitat of petroleum and exploration frontiers in the Brazilian rifts. B. Geoci. Petrobras, 2(2/4): 217-253. Capen, E.C., 1993. A consistent probabilistic approach to reserve estimates. SPE Hydrocarbon Econ. and Evaluation Symp. (Dallas 1993) Proc., pp. 117-122. Chang, H.K., Kowsmann, R.O. and Figueiredo, M.E de, 1988. New concepts on the development of east Brazilian marginal basins. Episodes, 11(3): 194-201. Clifford, A.C., 1986. African Oil - - Past, Present and Future. In: M.T. Ha|bouty (Editor), Future Petroleum Provinces of the World. AAPG Mem., 40: 339-372. Davison, I., 1991. Brazil's many sedimentary basins offer attractive exploration targets. Oil Gas J., Aug. 5, pp. 52-55. DeSorcy, G.J., Warne, G.A., Ashton, B.R., Campbell, G.R., Collyer, D.R., Drury, J., Lang, R.V., Robertson, W.D., Robinson, J.G. and Tutt, D.W., 1993. Definitions and Guidelines for Classification of Oil and Gas Reserves. J. Can. Pet. Technol., 32(5): 10-21. Doust, H. and Omatsola, E., 1990. Niger Delta. In: Divergent/ Passive Margin Basins. AAPG Mem., 48: 201-238. Edwards, J.D. and Santogrossi, EA., 1990. Summary and Conclusions. In: Divergent/Passive Margin Basins. AAPG Mere., 48: 201-238. Figueiredo, A.M.E de and Martins, C.C., 1990. The Campos Basin: Twenty years of activity and deep-water exploration success. B. Geoci. Petrobras, 4(1): 105-123. Franca, L.C., Giozza, W.E, Araujo, M.S., Molle, L., Jr. and Assayag, M.I., 1992. Part 2: Outlook on exploitation concepts. Proc. 13th World Petroleum Congress, Buenos Aires, 1991, J. Wiley and Sons, New York, Vol. 2, E&P, pp. 307-314. Franke, M.R., 1992. Discovered and potential petroleum resources, deep offshore Brazil. Proc. 13th World Petroleum Congress, Buenos Aires, 1991, J. Wiley and Sons, New York, Vol. 1, Pt. 2, pp. 71-73. Freire, W., 1990. An outlook on deep water exploration and production in Brazil. In: F.L.L.B. Carneiro et al., International Symposium on Offshore Engineering (7th: 1989: Rio de Janeiro, Brazil) Offshore Engineering, Pentech Press Ltd., London, pp. 1-22. Guardado, L.R., Gamboa, L.A.E and Lucchesi, C.E, 1990. Petroleum Geology of the Campos Basin, Brazil; A model for a producing Atlantic type Basin. In: Divergent/Passive Margin Basins. AAPG Mem., 48: 3-79. Holtz, M.H., 1993. Estimating oil reserve variability by combining geologic and engineering parameters. SPE Hydrocarbon Econ. and Evaluation Symp. (Dallas 1993) Proc., pp. 85-95.
S.J. MORBEY
"133
Knott, D. and Petzet, G.A., 1993. Namibia wildcat drilling, licensing round planned. Oil Gas J., 91(36): 93-95. Macedo, R.A.V. de, 1992. Brazil near future offshore projects: development trends and market opportunities. 10th Offshore Northern Seas Int. Conf. Proc., 1, F6, 20 pp. McHaffie, E.R., Jarvis, M.G. and Barber, S.A., 1993. Governments, contractors seen headed for era of cooperation. Oil Gas J., 91 (17): 54-57. Mell.o, M.R., Mohriak, W.U., Koutsoukos, A.H.M. and Figueira, J.C.A., 1992. Brazilian and West African oil: generation, migration, accumulation and correlation. Proc. 13th World Petroleum Congress, Buenos Aires, 1991, J. Wiley and Sons, Vol. 3, Pt. 1, pp. 153-164. Mohriak, W.U., Mello, M.R., Karner, G.D., Dewey, J.F. and Maxwell, J.R., 1989. Structural and stratigraphic evolution of the Campos Basin, Offshore Brazil. In: A.J. Tankard and H.R. Balkwill (Editors), Extension Tectonics and Stratigraphy of the North Atlantic Margins. AAPG Mem., 46: 577-598. Moraes, M.A.S., 1989. Diagenetic evolution of Cretaceous-Tertiary turbidite reservoirs, Campos Basin, Brazil. AAPG Bull., 73(5): 598-612. Palhares, C.A.C.A., Jr., Rangel, H.D., Wolff, B. and Figueiredo, A.M.F. de, 1991. Lagoa Parda Field - - Brazil. Espirito Santo Basin, Southeastern Brazil. AAPG Treatise of Petroleum Geology. Atlas of Oil and Gas Fields. Stratigraphic Traps, II, pp. 349-360. Pereira, M.J. and Macedo, J.M., 1990. Santos Basin: the outlook for a new petroleum province on the southeastern Brazilian continental shelf. B. Geoci. Petrobras, 4(1): 3-11. Peres, W.E., 1993. Shelf-fed turbidite system model and its application to the Oligocene deposits of the Campos Basin, Brazil. AAPG Bull., 77(1): 81-101. Peres, W.E. and Ramos, A.L., 1986. O uso da sismica na delimitacao de acumulacoes de hidrocarbonetos na bacia de Campos - - Brasil. Bol. Tec. ARPEL, 15(2): 101-113. Perrodon, A., 1993. Overview of African petroleum systems. Oil Gas J., 91(28): 115-118. Riva, J.E, Jr., 1992. Exploration Opportunities in Latin America. PennWell Publishing Co., Tulsa, OK 74101, USA, 266 pp. Santa Cruz, C.E. de, Barrocas, S.L.S. and Appi, C.J., 1987. Depositional model of Oligocene/Eomiocene turbidite reservoirs in the Albacora oil field, Campos Basin, Brazil. B. Geoci. Petrobras, 1(2): 215-223. Santos, C.F. and Braga, J.A.E., 1990. Reconcavo Basin: current status. B. Geoci. Petrobras, 4(1): 35-43. Souza, J.M., Scarton, J.C., Candido, A., Souz Cruz, C.E. and Cora, C.A.G., 1989. The Marlim and Albacora Fields: geophysical, geological, and reservoir aspects. 21st Annual Offshore Technology Conference (OTC), Houston, pp. 109-118. Szatmari, E, Milani, E., Lana, M., Conceicao, J. and Lobo, A., 1985. How South Atlantic rifting affects Brazilian oil reserves distribution. Oil Gas J., Jan. 14, pp. 107-113. Teisserenc, E and Villemin, J., 1990. Sedimentary Basin of Gabon Geology and oil systems. In: Divergent/Passive Margin Basins. AAPG Mem., 48:117-199. Van der Ven, EH., Cainelli, C. and Fernandes, G.J.E, 1989. Geology and exploration in the Sergipe-Alagoas Basin. B. Geoci. Petrobras, 3(4): 307-319.
StrategicExploration, WEBG, Exploration and Production, Houston, USA
This Page Intentionally Left Blank
135
Risk analysis: from prospect to exploration portfolio and back Gordon M. Kaufman
Improper accounting for covariation of uncertain quantities that characterize prospect, play, basin and corporate exploration portfolio uncertainties can lead to seriously distorted appraisals of exploration and investment risk. The roles played by covariability up the ladder of aggregation from prospect to exploration portfolio are examined. The distinction between systematic risk induced by the dependence of all exploration opportunities on price variation (all boats rising and falling on the tide of prices) and non-systematic or diversifiable risk (geologic and engineering) is highlighted in the context of deciding how to allocate exploration effort among competing exploration opportunities. Valuation of projects that can be flexibly managed over time cannot be correctly valued using probabilistic net present value methods that employ a fixed discount rate. Modern asset value methods derived from the theory of stock options allows correct accounting for flexible timing of exploration and development decisions. An example shows how to employ these methods.
Introduction In a 1962 visit to the offices of General Petroleum in Los Angeles, the corporate exploration manager proudly informed a colleague and me that, starting with the company's next annual exploration budget allocation meeting, a geologist must assign a "dry hole" probability to each prospect that he promotes. This was the company's first venture into the world of quantitative risk assessment. Few companies went so far at the time. We have come a long way since this time. The sophistication of probabilistic, statistical and economic aids to oil and gas exploration has developed in tandem with advances in geophysical, geochemical and geological techniques that yield descriptively richer, more precise and more accurate data. It is now routine for oil and gas firms to employ information systems that integrate large geological, geochemical and geophysical data bases, mapping protocols and economic decision-making paradigms to assist management in deciding how much to spend on exploration, when to spend it and where it should be spent. These systems are shaped by intellectual frameworks constructed from a blend of geological, geophysical, geochemical and engineering theory and practice, probabilistic and statistical analysis of earth science and economic data and micro-economic theory of profit maximizing behaviour of firms in the presence of large project risks. The papers presented illustrate the broad range of systems that can be constructed from a common core of ideas and they give
us a view of techniques for exploration and development uncertainty and risk analysis at the cutting edge of current knowledge. My comments hinge on four basic themes: The first theme is the importance of accounting correctly for covariation of uncertain quantities at all levels of aggregation~prospect, play, basin, and exploration portfolio. Omission of dependencies among uncertain quantities can lead to seriously distorted estimates. Covariability plays different roles as analysis steps up the ladder of aggregation. Statistical analogy can be employed to account for covariation of uncertain quantities in much the same way that we employ geological analogy to provide a benchmark for understanding geological features of frontier areas. The second theme is closely related to the first. Decisions about allocation of exploration effort among basins, among plays and among prospects within plays should account for covariability of returns to exploration effort introduced by price and cost uncertainties and, in particular, should distinguish between systematic and unsystematic risk. Meanvariance portfolio analysis and its generalizations can be employed to this end. The third theme is that when exploratory risk analysis is based on personal assessments of uncertainties by one or more technical experts, post mortems aimed at measuring the ex p o s t quality of such assessments are much talked about, but unfortunately too few are done. In addition, direct elicitation of judgements
Quantification and Prediction of Petroleum Resources edited by A.G. Dor6 and R. Sinding-Larsen. NPF Special Publication 6, pp. 135-152, Elsevier, Amsterdam. 9 Norwegian Petroleum Society (NPF), 1996.
G.M. Kaufman
136
about dependencies among uncertain quantities is almost always difficult and often infeasible. Ignoring even modest correlations can distort both the accuracy and precision of assessments of field size. The final theme is the importance of correctly valuing project flexibility. The most frequently used method of project valuation is to base it on properties of the probability distribution of project net present value computed at a pre-assigned discount rate. This method is appropriate when, once the project is under way, management cannot alter its future course. If, however, management can expand, contract, or abandon the project as the future unfolds then, unfortunately, the method does not correctly account for project value. Possession of a lease to explore a tract and then to develop it if oil or gas is found in commercial quantities is just such an option. In fact, it is an option (to develop or not depending on the outcome of exploratory drilling) within an option to explore or not before lease expiration. Modem finance theory offers option valuation methods that correctly account for flexibility in the timing of exploration and development. An illustrative example is given later, in the section on options for exploration and development. A simple taxonomy of types of risk faced by oil and gas explorationists can be given. It sets the stage for our discussion of uncertainty and risk. The classification appearing in Table 1 is self-explanatory. The distinction between systematic risk and nonsystematic risk is fundamental: non-systematic risks are those that can be reduced by geographic and geological diversification of exploration and development activities, whereas systematic risks are those risks that cannot be so reduced. Oil and gas price risks are systematic. Economic returns to all oil and gas projects, like boats on a rising and falling tide, move up and down together with rising and falling oil and gas prices. I show later how the aggregate risk
Table 2 Risk reduction - Increase resolving power of detection technology. - Do "smarter" intensive and extensive data analysis, and interpretation. - Increase calibration, accuracy and precision of expert judgment. - Diversify intelligently. - Employ up-to-date decision aids that fit your style.
of an exploration prospect portfolio can be split into systematic and non-systematic parts. A limited but powerful battery of tools with which to reduce risk are available (Table 2). The state of the art is well illustrated by the presentations given at the conference from which this volume is derived. They range widely. Grant et al. develop the concept of play uncertainty maps and play risk maps. Dahl and Meisingset show how state-of-the-art menu-driven basin modelling software can yield quick basinal assessments and projections of accumulation histories, and Krokstad and Sylta show us a method for assessing uncertainties in source rock yields and trapped hydrocarbons. In their discussion of a model-based approach to evaluation of exploration opportunities, Duff and Hall suggest a distinctive approach to play definition and modelling that emphasizes closure style. They correctly emphasize the importance of post mortem evaluation of model performance as a way of correcting biases and adapting to new information and tag probability intervals (risk tranches) with specific descriptions of the type of data that supports assignment of a geologic event to an interval. Morbey's examination of historical play efficiency throughout the South Atlantic rift system is, in effect, a post mortem of play successes and failures at an aggregated basin level. Ex post analysis of the quality of subjective probabilistic assessments made by explorationists is an
Table 1 Types of risk u risk = uncertainty + undesirable consequences
Political
Fiscal
Non-fiscal
" ... not maintaining a contract for exploration, development and production perceived as equitable by both an oil company and a host government in the face of changing political and economic conditions." (Kobrin, 1979. J. of Int. Bus. Stud., 10(1)).
"Ex post changes in fiscal terms of a contract." (Blitzer et al., 1989, Energ. J., VI, Spec. Tax Issue).
Country risk unrelated to that country's government actions v i s a v i s a contract. (Sangsnit, 1993).
Exploration
Development
Costs and prices
Potential economic losses stemming from uncertainties about the existence of hydrocarbons in a prospect field or basin, and about magnitudes of oil and of gas in place conditional on their presence.
Potential economic losses arising from uncertainties about technical features of field, pool, reservoir development.
Potential economic losses associated with uncertainties about exploration and development costs and uncertainty about future oil and gas prices.
Risk analysis: from prospect to exploration portfolio and back Table 3 Desirable attributes in expert judgment and risk reduction Well calibrated: announced probabilities are in good correspondence with relative frequencies of occurrence. Discriminating: announced probabilities are extreme and well calibrated. Accurate: probability of coverage of a state of the world is large. Precise: dispersion of a probabilistic assessment is small. "The three most influential (determinants of exploration venture) values are target size, discovery probability and finding cost" (Rose, 1987).
increasingly attractive way to improve the quality of risk assessment. Many companies are riding this bandwagon. Some desirable properties of subjective probability assessments are shown in Table 3. Without a vigorous effort to encode and analyze the historical performance, expert judgement is not likely to improve. In a 1987 paper, Rose tells us how to improve the precision and accuracy of probability judgements. And in testimony to the Texas Water Commission, Rose says: ...
most technical people have almost no idea as to their
their degree of uncertaintymthey cannot differentiate between 98 percent confidence and 30 percent confidence! Moreover, the prevailing pattern is one of overconfidencekwhen asked to make estimates at, say, 90% confidence, they characteristically set predictive ranges that actually reflect about 35-40% accuracy. As Capen says, "people tend to be a lot prouder of their answers [i.e., predictions] than they should be." This bias is nearly universal (scientists and engineers are not exempt!) and expresses itself specifically in forecasts that are exceeded by subsequent events (or not met at all). That is, in their quantitative predictions, experts usually set their predictive ranges far too narrow. In qualitative forecasts, this bias is expressed by a strong tendency to rely on only one or two hypothesesmrather than on many--in carrying out a scientific investigation. Put simply, most scientists and engineers are overconfident-they think they know more than they do! So they frequently find themselves surprised by Nature's outcomes . . . . over the past 10 years I have tested more than 100 technical audiences, totalling well over 5000 professional scientists and engineers. The results are always the same--they are significantly overconfident, actually estimating at about 40% confidence while believing they are estimating at 80% confidence . . . . We have found that, with training and practice, scientists and engineers can improve significantly, but even after considerable effort, they have a hard time consistently setting ranges that really do correspond to demonstrable uncertainty."
Now let us turn to covariability. Covariability - - when you can neglect it and when you cannot
Covariability of uncertain physical and economic variables underlying oil and gas exploration risk assessment is the rule rather than the exception. Some variables covary at all levels of aggregation~
137
prospect, pool, play and basin levelsmand may covary spatially as well. This fact leads to a small paradox; while sophisticated multivariate statistical techniques specifically designed to parse vectorvalued observations of geological, geochemical and geophysical phenomena into probabilistically independent components are now widely employed, probabilistic dependencies at other levels in the chain of analysis for decision making are not often treated with precision and are sometimes omitted when they should not be omitted. Many of the studies presented in this volume alert us to the importance of probabilistic dependencies among geological, geochemical and engineering variables. Several presenters call attention to the importance of proper modelling of probabilistic dependencies: Grant et al. highlight the distinction between prospect specific and play risk, and Snow et al. model risks generated by a portfolio of prospects. The introduction of play level uncertainties can introduce probabilistic dependencies that do not vanish even after drilling has confirmed the play's existence. This is the case for discovery process models based on the idea that discovery is akin to sampling a finite population of deposits without replacement and proportional to size. Damsleth's model (this volume) for the total number of discoveries that can be made in an area incorporates prospect dependence in the form of expert judgement about pair-wise dependence of prospects within a cluster of prospects that share the same marginal probabilities and pair-wise (joint) probabilities of success. Sinding-Larsen and Chen's (this volume) integration of discovery process models and volumetric accumulation models automatically incorporates dependencies among sizes of fields remaining to be discovered. It is well understood that geologic play risk is sensitive to dependencies among geological events such as timing, presence or absence of migration paths, existence of reservoir rock and seal. Hermanrud et al. (this volume) highlight the sensitivity of North Sea economic risks to variations in levels of, and dependencies among such geologic events and to variations in infrastructure, field size and water depth. It is less well understood that even mild correlations among the primary physical variables that determine field size~area of closure, average feet of pay and yield per acre foot, for example~can induce large differences in properties of a field size distribution relative to a field size distribution that does not have these correlations. The Lloydminster play is an excellent example. The Lloydminster play in Canada's Western Sedimentary Basin is unusual in two ways. First, features of all deposits in this well explored play have been
138
G.M. Kaufman
I
Table 5
I
Covariance structure for the Lloydminster play
log Acres log Pay log Quality
w
<>
log Acres
log Pay
log Quality
2.459 0.133 0.044
0.371 0.011
0.019
1.00 0.139 0.182
0.134
Correlation matrix
0
log Acres log Pay log Quality
x ua
OIP -- acres x net pay x quality.
C3
m
0
1.00
1.00
-2 Table 6
-3 -4
Lloydminster play: properties of oil-in-place (millions of barrels)
0
1
I
10
20
30
LOGOIP Fig. 1. Lloydminster play, Western Canada Basin. Plot of logarithm of oil-in-place (bbls) against a normal probability scale (vertical axis) measured in standard deviations.
Mean Standard deviation Mode Median Inter-quartile range 2 2 O'with/O'withou t =
measured down to single-well deposits using a uniform protocol. Second, there are 2509 deposits. Figure 1 is the empirical cumulative distribution function (ecdf) of the logarithm of oil-in-place plotted against a Normal probability scale. The horizontal scale is in logarithmic units and the vertical scale is constructed so that if fractiles of the logarithm of oil-in-place correspond to a Normal distribution, then the graph of the ecdf will appear as a straight line. The assumption that oil in place is approximately Lognormal is clearly reasonable (see Kaufman, 1993, for further discussion of this play). Lloydminster play statistics appear in Table 4. The large number of deposits in this play enable us to pin down with precision the covariance structure of deposit area, net pay and yield per acre-foot. Table 5 displays covariance and correlation matrices for the logarithms of area, pay and yield. None of the pair-wise correlations are large. Since oil-in-place in a deposit is the product of
Without covariances
With covariances
14.59 (x 1.21) 58.83 (x 1.47) 0.203 (x 0.69) 3.51 (1.13, 10.97)
17.68 86.32 0.140 3.51 (1.05, 11.97)
1.13.
these three variables, if we adopt the working assumption that all three variables are jointly Lognormal, then we can easily compute properties of the distribution of oil-in-place as a function of area, pay and yield. Table 6 offers a comparison of properties of the distribution of oil and place assuming independence of log area, log pay and log yield with the distribution that arises if we account for empirically derived correlations (those of Table 5). Even though pair-wise correlations are small, ignoring them substantially distorts estimates of the mean, mode and standard deviation of oil-in-place (Table 6). Specifically, the mean deposit size is underestimated by about 21%, the standard deviation by about 47% and the mode is overestimated by about 30%. A 21% underestimate of mean deposit size results in an underestimate of about 3.109 barrels of oil in place! Both distributions are so positively skewed that, on a scale of oil-in-place, they are virtually indistinguishable. (Fig. 2).
Table 4 Statistics for the Lloydminster play ~.2
log log log log log
Area (acres) Pay (feet) Acre-feet Quality (bbls/acre-feet) OIP
6.206 1.557 7.850 7.229 15.072
2.459 0.371 3.091 0.019 3.221
Area Pay Acre-feet Quality OIP
~
~ ( e a2
1836 5.71 12.034 1.392 17.584.000
6003 3.83 55.145 192.8 86.32.106
_
1)1/2
Mean M = exp{/x+ 102}" standard deviation = M(e a2-1) 1/2./~ and 6 2 are unbiased estimates of # and 0"2. The estimate of 37,/= exp{/2+ 89 of M is biased. For simplicity, we eschew bias correction here.
Risk analysis: from prospect to exploration portfolio and back Table 7
0.4
I
I
I
Exploratory risk-return tradeoffs via mean-variance portfolio analysis
0.3
a.
139
0.2
Inputs
Outputs
- Prospect alternatives available - Prospect uncertainties Prospect expected NPV Prospect dispersion (variance of NPV) prospect covariabilities (covariability of NPVs) - Exploration budget constraints - Activity constraints
Allocation of exploratory effort that: (1) Minimizes portfolio dispersion (variance) Subject to: (2) Achieving target expectation of ROR
0.1
0,0
0
I
I
5
10
i
15
20
Fig. 2. Probability density plot of Lloydminster oil-in-place (horizontal axis) showing positive skewing of distribution. Variance = 3.221.
When a deposit size distribution is computed from assessments of individual components of deposit size, component correlation structure can be ignored only if the assumption that components are independent is absolutely convincing. When the assumption of independence is not warranted, import the covariance structure of a good geological analogy if you can, rather than ignore covariation. Exploration, d e v e l o p m e n t and r i s k - r e t u r n tradeoffs
Further up the ladder of aggregation, the aggregate economic risk of a corporation's exploration and development portfolio may be strongly influenced by covariabilities of geological, engineering and cost risks and is always influenced by price risk. The fashion in which an exploration budget is allocated among basins, among plays within a basin and among prospects within a play determines an exploration portfolio's relative exposure to systematic and to non-systematic risks. As stated earlier, the distinction between these two types of risk is fundamental: Non-systematic risk = Diversifiable risk Systematic r i s k - Non-diversifiable risk By appropriate spreading of the investment budget within its oil and gas exploration and development opportunity set, a corporation can reduce the component of dispersion (variability) of investment rate of return (ROR) that depends on geological and engineering uncertainties. Farm-outs, bottom hole contributions and syndication of lease bids are other
familiar devices for sharing risks. However, cost and price risks cannot be diversified away in this fashion. Price risk can be reduced by hedging in oil price futures markets or by diversifying away from oil and gas markets (e.g., buy airline company stocks to introduce negative correlation with oil and gas exploration and development ROR), but our interest here centres on those risks associated with exploration and development management. It is possible to measure the relative contributions of systematic and non-systematic risks to overall risk by appropriate application of financial portfolio theory. The principal aim of this theory is to identify allocations of a fixed investment budget that, subject to a budget constraint and to activity constraints, minimizes dispersion or variability of ROR while achieving a target expected ROR. Table 7 outlines inputs and outputs of such an analysis restricted to an exploration prospect opportunity set. Associated with each target expected ROR is a minimum variance portfolio. The graph describing how the standard deviation of such portfolios varies as a function of target expected ROR and budget size is a valuable display of available risk-return tradeoffs and we shall discuss an example. Before doing so, however, we note an important property of portfolio variance. By use of a formula well known to statisticians, the variance of ROR for any exploration and development portfolio can be split into a sum of two pieces, one systematic and the other non-systematic (Table 8). Total portfolio variance is seen to be the sum of a diversifiable risk component and a non-diversifiable risk component. Each of these two components, shown in
Table 8 Decomposition of variance of ROR Var (ROR)
= Ep Var (RORIP) + Varp[E(RORIP)] Y non-systematic
Y systematic
Non-systematic = E.V. of variance of ROR as a function of price P Systematic
= Variance of E.V. of ROR as a function of price P
1
4
0
G
.
Table 8, can be interpreted as a function of uncertain future prices. The diversifiable component is the expectation with respect to uncertain future prices of the variance of ROR conditioned on future prices. The nondiversifiable component is the variance with respect to future prices of the expectation of ROR conditioned on future prices. Price variability introduces positive correlation of the RORs of individual prospects and the contribution to overall variability due to these correlations is generally large. The formula in Table 8 enables us to see why even the most extreme diversification of geological and engineering risks leaves price risk intact. Consider a portfolio of N prospects with a fraction 1/N of the exploration budget (scaled to equal one) allocated to each. Suppose that the risk characteristics of each prospect are identical and in addition that, given known future prices, the outcomes of drilling these prospects are uncorrelated. Denote the variance of ROR of a generic prospect given future prices P by V (RORI P) and let E (RORI P) be the expected ROR of a generic prospect given P. In turn let V equal the expectation with respect to P of V (RORI P). Because drilling outcomes are assumed to be uncorrelated, the unsystematic component of ROR variance is V/N and the systematic component is the variance of a single E(RORIP) with respect to P. As N gets larger and larger, V/N approaches zero, while the systematic component, the variance of E(RORI P), remains unchanged. To illustrate these ideas, mean-variance tradeoffs for an exploration opportunity set available to a U.S. independent are shown in Figs. 3 and 4. The set consists of sixty U.S. gas prospects. Since this is an il-
70
Z
9
[< >
I
60
-
50
-
40
-
30
-
U~
a <
I
I
I
9
2010
-
9
9
.
80
9
70
9
60
9
50
9
40
9
30
9
20
~
Z
< [-,
9
9
rael O
O0
i
01
I
I
02
03
!
0.4
05
ROR Fig. 3. Mean-variance analysis for gas prospects with fixed price of $6.15/MCE Horizontal axis = target rate of return. Vertical axis = standard deviation of value of exploration portfolio. Dots denote percentage of exploration budget allocated to the prospect set.
M
Kaufman
. I
I
I
9
100
Z
9
9
too
.
90
9
80
9
70
9
60
9
<
> u~
"
i
9
9
9
9
50
9
9
9
Z
0
1
0,O
0.1
I 0,3
[
02
9
50
40
9
9
< [..
,
30
.
I 0,4
20
0,5
ROll Fig. 4. Mean-variance analysis with 70% price volatility. See text and caption to Fig. 3.
lustrative example, we have assumed that drilling outcomes measured in recoverable gas equivalent MCF (thousand cubic feet) are uncorrelated and in addition that conditional on success, production profiles are fixed and known. Working interest in a prospect is constrained to be less than or equal to 100% (the allocation cannot oversubscribe desirable prospects), but fractions of working interest are allowable. Price uncertainty is denominated in terms of a single uncertain price that scales a future price vector. At the time that these prospects were in contention, the relevant unregulated price was about $6.15/MCF because of market distortions introduced by the Natural Gas Policy Act of 1978. For each choice of exploration budget level there is an "efficient frontier" consisting of the set of pairs of target expected ROR and minimum variance of ROR given this target. Associated with each such pair is an allocation of the budget to prospects. Figure 3 displays efficient frontiers conditioned on a fixed price of $6.15/MCF. Figure 4 displays efficient frontiers assuming price uncertainty and price volatility of 70% (standard deviation of percent change in price from one period to the next). Price volatility introduces a risky shift of efficient frontiers. Systematic risk as a percent of total risk varies with target ROR and budget size, as shown in Fig. 5. Unregulated wellhead prices up to $10/MCF appeared at the time of this example; Anadarko Basin (Fletcher Field) deep gas wellhead price is a case in point. At prices as high as $6.15/MCF it is possible to achieve target RORs of 20-50% by investing only a fraction of allocated budget. Thus with price volatility of 0.707 and a 50 million dollar budget, here is
Risk analysis: from prospect to exploration portfolio and back 0.35
0.31
I
-
i
i"
i
Z 9 ll, 9
~,
;
9
027
I
9
9
"
9
"
9
.
I
9
o
_,
, ,
9 9
9
9
,
,
9
.
.
*
.
"...~00
*
9
-
"
. ;
9
023
9
9
9 9
-
9
9
I
.
9
9 *
.
9
9
9
0.19
0.15
-
0.0
*
90
Ot'l
.
ov
.
/u
.
60
.
Ju
,
40
,
I 0.2
t 0.3
-
t 0.4
20
0.5
ROR
Fig. 5. Plot showing how systematic risk/total risk proportion (vertical axis) varies with target rate of return and budget size. See text and caption to Fig. 3.
how the fraction of budget invested varies with target ROR. Target ROR 20% 30% 40%
Budget fraction 0.190 0.343 0.440
If target ROR is held fixed, as the amount invested in this fixed prospect opportunity set increases, systematic risk increases as a proportion of total risk. As the target ROR increases with the budget fixed, this proportion decreases. An intuitive explanation for the decrease in systematic risk divided by total risk for a fixed budget as target ROR increases, rests on three features of the structure of the portfolio model adopted for this example. First, price uncertainty is represented by a single price that linearly scales the expected NPV of each prospect. Second, the large value of current price relative to the variance of price even in the presence of substantial volatility of 0.707 leads to: Var (Price) = 0.5 [Current Price] 2
Some managers object to the choice of minimum variance as a criterion because it treats upside and downside variations in ROR symmetrically. They prefer to minimize the expectation of downside risk subject to achieving a benchmark expected ROR. A decomposition of both the expectation of downside risk of a portfolio and the variance of downside risk into systematic and non-systematic risk components is possible. How this can be done is a story for another day!
Looking forward" options, exploration and development
30
9
I O. 1
_
] 41
Modern asset valuation methods are designed to address the problem of valuing projects that can be flexibly managed over time. These methods are an outgrowth of a compelling theory of stock option valuation developed in the 1970s by three MIT professors, Black, Merton and Scholes (Black and Scholes, 1973; Merton, 1973). Their discovery caused an explosion of stock and commodity market investment activity and stimulated research designed to extend the key ideas of their theory of stock option valuation to managerial project valuation. Since exploration and development projects are conceptually analogous to stock options, this work should be of particular interest to exploration managers. It has long been recognized that possessing the right to explore a tract for a specific period of time is an operating option that is informally equivalent to a call option on a stock. The analogy is sketched in Table 9. All explorationists appreciate the option value of pre-emption--get into a highly prospective frontier area before the competition in order to maximize exploration flexibility and to pick off the most promising fields. If a discovery is made, another option arises" develop now, postpone development or not develop at all. This development option is embedded within the option to explore at some point in time prior to expiration of the lease. Thus, ex ante, the exploration manager faces a compound option. Only recently, however, has a conceptually sound and
and
Expectation (Price Squared) = 1.5. [Current Price] 2 The non-systematic risk component of portfolio variance is weighted by 1.5 and the systematic component of risk by 0.5. Third, if the budget is held fixed and target ROR is increased, effort is increasingly allocated to riskier prospects. As this happens, nonsystematic risk increases more rapidly (roughly with weight 1.5) than systematic risk (with weight 0.5).
Table 9 Exploration and development options: analogy with the stock market
Exploration
Stock market
Option to time exploration expenditures
=
Relinquishment date
=
Exercise a stock call option: pay exercise price and receive stock Call expiration data
Development Option to pay development costs and install productive capacity to receive developed reserves
A second call option
142
G.M. Kaufman
easily explainable method for valuing such options been in place. Just what is this method? Trigeoris and Mason (1987) point out that extensions of stock market option theory to embrace valuation of project flexibility are a special, market-adjusted version of decision tree analysis that explicitly prices out the value of operating flexibility. Why is this important? Because, according to all published accounts, the method overcomes inadequacies of approaches to project management that rest on choice of a single, risk-adjusted discounted rate of return to compute the probability distribution of project net present value (NPV): - Traditional NPV methods do not properly capture the value of a manager's ability to modify a project as uncertainty is resolved. - As uncertainty unfolds and risk increases or decreases, so does the "correct" risk-adjusted discount rate, i.e., no single discount rate may be appropriate. - The ability to adapt the operation of a project to future contingencies introduces asymmetries in future project value resulting in an overall increase in value, relative to static NPV analysis. According to Pickles and Smith (1993), "Discounted Cash Flow (DCF) analysis typically ignores the added value brought to the project through management's ability to make operating decisions during the life of the project and to adjust the investment to existing market conditions as they change over time".
Options or contingent claims methods of valuing project flexibility have advantages: - There is no need to forecast the mean path of future prices. - The appropriate discount rate is the risk free rate, i.e., there is no need to pick a risk-adjusted discount rate as risk adjustment is intrinsic to the method. These methods do depend on an estimate of price volatility (price variance) and the assumption that the relevant oil and gas market is in equilibrium. Paddock et al. (1988) document the practical importance of option valuation: "The governmentuses valuations to establish pre-sale reservation prices and to study the effect of policy changes on revenues it expects to receive from lease sales. Because the bidding process involves billions of dollars, it is important to obtain accurate valuations [of tracts posted for bidding]. [As of 1988] Government valuations have tended to underestimate industry bids. Using the same geological and cost data as the government, our option valuations are closer to industry bids." The analysis of stock option value done by Merton (1973) and by Black and Scholes (1973) requires an understanding of advanced mathematical concepts (Ito's stochastic calculus and parabolic partial differential equations with moving boundaries). Paddock et al. (1988)and Jacoby and Laughton (1992) show how to value an exploration option using the originators' mathematical approach to the problem. Bjerksund and Ekem (1990) compare traditional NPV methods and option valuation and show how distinct types
EXAMPLE: B = 100M BBLS COSTS I = $500M TO DEVELOP
DECISION WITHOUT
. 3 ~
$7.32" $232M
PRICE CLAIRVOYANCE $4.68" -$32M
CurrentPrice/bbl
E.V. D = $47.2M
--~
o THE CLAIRVOYANT CAN FLIP THE TREE!
SHE GAINS $22.4M
DECISION WITH PRICE CLAIRVOYANCE
D
$232M
$7.32 - ~ 6 CurrentPrice/bbl $
~ i
$4.68 D ~ - ~
0 $-32M0
E.V. "D IF $7.32 AND D IF $4.68" = $69.6M
Fig. 6. Decision tree analysis showing development option with and without oil price "clairvoyance". See text for discussion. D = develop, /9 = do not develop.
Risk analysis: from prospect to exploration portfolio and back Time to expiry (yr) Length of time period (yr) Volatility (annualized) Real risk-free interest rate Payout rate Price at time zero, per bbl Exercise price, per bbl Percent of costs expensed Development delay (yr)
1 43
After-tax risk-free interest rate, p
T dt a
= 6.00 = 0.2400 = 22%
Upward price change in one timer period
r 8
-- 4% = 10%
Downward price change in one time period
=r,(1-tax) = e x p { c r , ~d-t}
S X
= $6.00 = $3.00 expense = 30.00% delay = 6.00
= 1-
e x p { - c r , ~/-d--/}
=
1.90%
=
11.38%
=
10.22%
Change in expected price in one time period = exp{(p - 8) 9 - 1 = -1.92% I f probability of upward change -- p, then p 9 increased price 4- (1 - p) 9 decreased price -- expected price, therefore p = 0.3840
Present value of price Present value of exercise price (after tax) Calculated option value
PV(S)
= $3.293
and probability of downward change
PV(X) OV
= $2.528 = $0.765
Discount per period at after-tax risk-free rate
(1 - p)
=
= exp(-p,
dt) - 1
0.6160
= -0.46%
Possible prices and option values ($/bbl) Initial
(-
After period 1
[
After period 2
After period 3
A ~ e r period 4 ---+ (25 periods, total)
I 5.0676 X
2.5397 4.5498 x
2.0219 4.0850
4.0850
X
X
1.5570
1.5570
3.6676
3.6676
X
X
1.1396
1.1396
3.2929
3.2929 x 0.7649
X
0.7649 2.9564 O 0.4582
3.2929 X
0.7649 2.9564 O 0.4582
2.6544 O 0.2704
2.6544 O 0.2704 2.3832 O 0.1563 2.1397 O 0.0878
Fig. 7. Introduction of tax and delay with a 25-period model (from Pickles and Smith, 1993).
of operating flexibility affect value. Their theoretical analysis is made concrete with an excellent discussion of how operating flexibility changes the value of a North Sea oil field. Pickles and Smith (1993) adapt a simplified approach to option valuation developed by Cox et al. (1979) to oil and gas development options. This latter approach is based on an easily understood binomial model of price variation that, as the time span between periods of price change approaches zero, converges to the same continuous time distribution of price changes employed by Merton et al. In addition, the limiting valuation formula is identical to the continuous time formula. I believe that Pickles and Smith's approach is by far the easiest to understand
and their explanation is difficult to improve upon. They provide several examples, one of which is a prototypical valuation of a North Sea oil field. A simplified example of a development option will help set the stage for a discussion of real compound options. You own a 100 million barrel field that will cost $500 million to develop. Once developed, you will immediately sell the field. However, between now and sale time, the market price per barrel in the ground will rise to $7.32 with probability 0.3 and fall to $4.68 with probability 0.7. Thus the expected value of developing the field now is $47.2 million. This calculation is shown in the top decision tree of Fig. 6. A clairvoyant claims that she has perfect foresight. She can tell you with certainty whether
G.M. Kaufman
144
Initial
T=I
T--2
T=3
T=4
Initial
T = 1
5.0677 2.5397 0.0000 2.5397 DEVELOP
explore 0.1943 0.1781 0.1943
4.0849 1.5569 0.0000 1.5569 DEVELOP
2.9563 0.4283 0.3696 0.4283 DEVELOP
2.9564 0.4284 0.4108 0.4284 DEVELOP
2.6541 0.1261 0.0000 0.1261 DEVELOP
2.6542 0.1262 0.1933 0.1933 HOLD 2.3830 -0.1450 0.0482 0.0482 HOLD p = 0.384 ( 1 - p) = 0.616 d - 0.9954 D = 2.528
0.1128 0.1007 0.1128
0.1128 0 0.1128
explore Exercise value Hold value Option value Exercise?
0.0397 0.0532 0.0532 HOLD
explore 0.0396 0.0431 0.0431 HOLD
0 0.0165 0.0165 HOLD 3.2927 0.7647 0.0000 0.7647 DEVELOP
3.2928 0.7648 0.6982 0.7648 DEVELOP
Price 3.2929 Exercise value 0.6983 Hold value Option value 0.6983 Exercise?
explore
3.6675 1.1395 1.0640 1.1395 DEVELOP
3.6676 1.1396 1.0641 1.1396 DEVELOP
T= 3 0.2851 0 0.2851
4.5499 2.0219 1.9254 2.0219 DEVELOP 4.0850 1.5570 1.4715 1.5570 DEVELOP
T -- 2
2.1394 -0.3886 0.0000 0.0000 HOLD
Fig. 8. Development option values given discovery of 100 million barrels (see text). For parameters, see Table 10.
the price/barrel will be $7.32 or $4.68, but has not yet done so. What is the value of being able to postpone choice until the clairvoyant has spoken? (She will speak in time for you to decide whether or not to develop at each possible price just cited.) The clairvoyant flips the decision tree for you! Choice is postponed until price is revealed. This tree is shown at the bottom of Fig. 6. The prior expected value of the option to postpone choice until price is revealed is a gain of $22.4 million over and above the value of the best (static with respect to price) choice in the top tree. Notice the asymmetry introduced at choice nodes: if price turns out to be $4.68, the value is zero (in contrast to - $ 3 2 million if choice is to be
0 0 0 HOLD 0 0 0 HOLD
p -- 0.348 (1 - p) = 0.616 d = 0.9954 D -- 2.528
0 0 0 HOLD
Fig. 9. Values of exploration rights with probability of discovery of 0.2 and exploration costs of S0.1/barrel.
made without clairvoyance). Any operating option, no matter how complicated, possesses these features. Additional work on valuation of exploration and development as a compound option remains to be done. Paddock et al. (1988) assume that if exploration is successful, then the decision to undertake development is made at the same time as the decision to explore. This ignores the value of development delay. Pickles and Smith (1993) assume that the decision to explore is taken immediately, but that the option exists to delay development. This ignores the value of exploration delay. Valuation of the compound option via Pickles and Smith's representation of the development.option can be done by interfacing a decision tree representing the complete choice set and layered spread sheets, one for each binomial lattice of development option value as a function of field size. The appendix presents key steps in such an analysis. A student of mine, Brant Liddle, calculated the compound option value of a four-period option. Exploration is allowed to take place up to and including the third period and development can be delayed up to and including the fourth period. The parameters chosen are based on Pickles and Smith's (1993) figure 5, a 25-period development option with tax and delay. (Shown here as Fig. 7; see also Table 10.) For simplicity, it is assumed that if a discovery is made, 100 million developable barrels will be dis-
145
Risk analysis: from prospect to exploration portfolio and back Initial
T= 1
T=2
Initial
T=3
T= 1
T=2
0.2351 0 0.2351 explore 0.0943 0.0786 0.0943 explore
0.1443 0.1284 0.1443 explore
Exercise value Hold value Option value Exercise?
Exercise value Hold value Option value Exercise?
0 0.0240 0.0240 HOLD
0 0.0323 0.0323 HOLD
0 0.0049 0.0049 HOLD
0 0.0161 0.0161 HOLD
0 0 0 HOLD
0 0 0 HOLD p (1
-
d D
= p) = = =
0.348 0.616 0.9954 2.528
0 0 0 HOLD
0 0.0119 0.0119 HOLD
0 0 0 HOLD
0 0.0092 0.0092 HOLD
0.0128 0 0.0128 explore
0.0128 0.0390 0.0390 HOLD
0.0628 0 0.0628 explore
0.0628 0.0699 0.0699 HOLD
T=3 0.1851 0 0.1851 explore
0 0 0 HOLD
Fig. 10. Values of exploration rights with probability of discovery of 0.2 and exploration costs of $0.15/barrel.
Table 10 Assumptions for Pickles and Smith's (1993) 25-period model with tax and delay Have to develop by the fourth period and have to explore by the end of the third period. Upward price change in one time period -- 11.38% - Downward price change in one time period = 10.22%
-
-
- Probability of upward change p -- 0.3840 - Probability of downward change (1 - p) = 0.6160 Discount per period at after-tax risk-free rate d = 0.9954 Present value of after tax cost of development D = $2.528 - Present value of price $3.293 = Current Price - Probability of discovery q = 0.2 - If discovery, field size is 100 M barrels
p = 0.348 (1 - p) = 0.616 d = 0.9954 D = 2.528
0 0 0 HOLD
Fig. 11. Values of exploration rights with probability of discovery of 0.2 and exploration costs of $0.2/barrel.
covered. Figure 8 presents development option values given discovery of 100 million barrels. Figures 9, 10, and 11 present binomial lattice values of exploration rights with development timing flexibility built in, each at different choice of exploration cost per barrel. Table 11 and Fig. 12 summarizes how values at t = 0 change with exploration cost and presents corollary
-
Table 11
-
Exploration options, illustrating the gains in value from flexibility for a range of exploration costs Exp. cost/bbl a
Value of exploration rights
Flexibility gains
0.10 0.15 0.20
0.0532 0.0323 0.0161
0.0135 0.0323 0.0161
Available acts Initial
Time = 1
Explore Explore Explore Explore Explore
Develop
Time = 2
Time = 3
Time = 4
Develop
FLEXIBILITY GAINS Develop Develop
Explore Explore Explore Explore
Develop Develop Develop Explore Explore Explore Explore Explore
Develop EXPL COST/BBL Develop
a With low exploration cost/bbl, the inflexible option "E now, D one period later" has positive E.V; with high exploration cost/bbl the inflexible option has negative E.V.
G.M. Kaufman
146
Value of Flexibility 0.04 0.03
0.02
0.01 1
0
0.1
probability of discovery = .25
0.2
i
I
,
I
0.3 0.4 Exploration Costs
0.5
0.6
Fig. 12. Plot of exploration costs in S/barrel against value of flexibility in development timing in $/bbl. See text and Table 11.
flexibility gains. Even though the value of exploring and developing a tract is negative at the current time, the value of the option to explore may be positive. At an exploration cost of $0.2 per barrel in our example, the expected value of exploring at t = 0 is negative, but the (compound option) value of exploration rights is positive. The reader is encouraged to examine the schematic outline of analysis given in the appendix. A decision tree for a three-period exploration and development option is constructed (a four-period tree is too large to display conveniently), and evaluated by backward induction. The value at t = 0 of "Wait and See" (HOLD) is compared with the value of "Explore Now" (EXERCISE) unconditionally as regards an optimal development strategy. Straightforward extension of this scheme by partitioning more finely the binomial lattice and by accounting for uncertainty about discovery size prior to exploratory drilling will yield a practical tool for valuation of gains from the ability to delay independently exploration and development. Conclusions
Our focus on the role of covariability of geological, engineering and economic variables in oil and gas
risk assessment led us from prospect risks to field size distributions, up the ladder of aggregation to prospect portfolio risk via mean-variance portfolio analysis and back down to valuation of operating flexibility at the prospect level. The risk management messages that emerged are as follows:
Expert judgement and risk reduction: Requires consistent monitoring effort; Evaluate and calibrate because the payoff is large.
Covariability of geologic variables: If present, neglect at your peril! Import statistical analogies.
Covariability of project returns and efficient allocation of exploration effort: Use mean-variance portfolio analysis or its generalizations; Separate systematic and non-systematic risk and measure both.
Value project flexibility correctly: Avoid under-estimation of project value; Eliminate need to forecast mean path of futureprices.
147
Risk analysis: from prospect to exploration portfolio and back
Appendix A. Valuation schematic for exploration and development as a compound option Scheme 1. Acts available t=0 al
t--1
t=2
E
D
-
E E
-
D -
a4
-
-
-
a5
-
E
D
a6
-
E
-
a7
-
-
-
a2
a3
9 Development lags exploration by one period. 9 Development must begin at t = 2 or earlier.
Scheme 2. Components of compound option Costs, prices, interest rate: C Po r p
Acts: E /~ D /)
-= = =
explore do not explore yet develop do not develop yet
= -=
exploration cost current price risk-free interest rate 1/(1 + r)
l__~p
u Po } - Next period prices
dpo Values: D(B, Po) = value of developing B barrels if price is Po {)(B, Po) = pd(B, lzPo) + (1 - p)O(b, dPo)
State: B -- bbls discovered
v(B, Po) = max[0, D(B, Po)] f)(B, Po) = pv(B, uPo) + (1 - p)v(B, dPo)
Scheme 3. Where does p = Prob (Price up) come from ? [Pickles and Smith (1993)] r
= risk-free rate = payout rate on developed reserves Po = current price
Equilibrium :=~ (1 + r - 6)/~ = p(uPo) + (1 - p)(dPo) Their example: (volatility = 22%) r=4%,
~=
13%,
u=
1.22,
d=0.78
=~0.91=p(1.22)+(1-p)(0.78)
=~p=0.2955~0.3
148
G.M. Kaufman
Scheme 4
Y p/~o-~~
pD(B, uP )
~ q
pfU2 o_
D/
pD(B, dP ) O X r',/ o 1-p N... ud, Dr.n,,~" ( du)Po
dPo~ D/
P2D(B, u2P ) O
0 p2D(B, udP ) O
0 92D(B, d2P )
@
_p~___d2p~
B=O
O
D~~ 0
p2D(B, u2dP ) O
p
u2p~
Po B
'-p (ddu) o Po
B=O
--
0 p2D(B, u dP ) O
D
p2D(B, d 2P ) O
d2Po-~
Po
=0
0
149
Risk analysis: from prospect to exploration portfolio and back
Scheme 5
v(B, Po ) - max {O, D(B, Po) }
pD(B, uP )
D
0
~-u2P :p2v(B, u2P ) O
O
pD(B, dPo 1 - p ~ q
Q/
/
dP~N ~
o
(ddu)Po: p2v(B, u dP )
'P~._d2Po: p2v(B, d2PO )
B=O
E
f
u2Po:p2v(B, u2Bo)
1-p
(Ud)D : p2v(B, u dP)
PO
7"
G
E .. '
E
"
B
=
0
--'du"o
"~0
Q ~
'-P\ 0
B=0
d2Po: p2v(B,
d~p~)
150
G.M. Kaufman
Scheme 6 m
v (B, Po ) = pv(B, uP) + (1- p)v (B, dPo)
p f
uP: max{pD(B, UPo)' p2 V(B, UPo))
1
dP :max{pD(B dP ), p2V(B, dP )} O
o/
O
O
B=0
E
P
O
B'P2V'(B, uP ) O
@ B=0
B" 92 ~r(B, dPo )
1-p
d~o~_~~ 1_
~--0 0
151
Risk analysis: from prospect to exploration portfolio and back
Scheme 8
Scheme 7
/~
E [ p max {pD(B, uP), ~2V'(~, uP)} + (1- p) max {D(B, dP), p2 V (I~, dP) }]- C -C
Explore now: E.V. of "Explore at t = 0, develop at t = 1" @ Value of option to develop at t -- 1 or postpone decision to develop or not to t - 2 Wait and see: E.V. of "Explore at t -- 1, develop at t = 2" @ Value of option to develop or not at t = 2
po
\ F_u [p max {qp2V (~, u P ) - pC, 0} + (1 - p){max qpZv-(B, dP) - p C, 0}]
Scheme 9 Explore now: Explore at t = 0, develop at t = 1
- C + q p D ( B , Po)
Value of option to postpone development decision (D or D) by one period
+
+
Pmax{0, qp[pfJ(B, uPo) - D(B, uPo)]} ] (1 - P)max{0, qp[p~(B, dPo) - D(B, dPo)]}
J
Wait and see: Value of option to wait and see price at t -- 1 before exploration decision
Explore at t = 1, develop at t -- 2
p [ - C + q p D ( B , Po)]
+
Pmax{C - q p D ( B , uPo), qp[pf)(B, uPo) - D(B, uPo)]} 1 + (1 - P ) m a x { C - q p D ( B , dPo), qp[pf)(B, dPo) - D(B, dPo)]}
Acknowledgements I wish to thank James Smith and L. James Valverde A., Jr. for valuable discussions, Brant Liddle for computing options valuation examples and John Maglio for programming assistance.
References Bjerksund, E and Ekern, S., 1990. Managing investment opportunities under price uncertainty: from last chance to wait and see strategies. Financial Manage., 19(3): 65-83. Black, F. and Scholes, M., 1973. The pricing of options and corporate liabilities. J. Polit. Econ., 81: 637-657. Cox, J.C., Ross, S.A. and Rubenstein, M., 1979. Option pricing: a simplified approach. J. Financial Econ., 7: 229-264. Dahl, B. and Meisingset, I., 1996. Prospect resource assessment using an integrated system of basin simulation and geological mapping software: examples from the North Sea. In: A.G. Dor6 and R. Sinding-Larsen (Editor), Quantification and Prediction of Hydrocarbon Resources. Norwegian Petroleum Society (NPF), Special Publication 6, Elsevier, Amsterdam, pp. 237-251 (this volume). Damsleth, E., 1996. A method for the statistical assessment of total undiscovered resources in an area. In: A.G. Dor6 and R.
Sinding-Larsen (Editor), Quantification and Prediction of Hydrocarbon Resources. Norwegian Petroleum Society (NPF), Special Publication 6, Elsevier, Amsterdam, pp. 83-90 (this volume). Duff, B.A. and Hall, D., 1996. A model-based approach to evaluation of exploration opportunities. In: A.G. Dor6 R. Sinding-Larsen (Editor), Quantification and Prediction of Hydrocarbon Resources. Norwegian Petroleum Society (NPF), Special Publication 6, Elsevier, Amsterdam, pp. 183-198 (this volume). Grant, S., Milton, N. and Thompson, M., 1996. Play fairway analysis and risk mapping--an example using the Middle Jurassic Brent Group in the northern North Sea. In: A.G. Dor6 and R. Sinding-Larsen (Editor), Quantification and Prediction of Hydrocarbon Resources. Norwegian Petroleum Society (NPF), Special Publication 6, Elsevier, Amsterdam, pp. 167-181 (this volume). Hermanrud, C., Abrahamsen, K., Vollset, J., Nordahl, S. and Jourdan, C., 1996. Evaluation of undrilled prospectsisensitivity to economic and geological factors. In: A.G. Dor6 and R. SindingLarsen (Editor), Quantification and Prediction of Hydrocarbon Resources. Norwegian Petroleum Society (NPF), Special Publication 6, Elsevier, Amsterdam, pp. 325-337 (this volume). Jacoby, H.D. and Laughton, D.G., 1992. Project evaluation: a practical asset pricing method. Energ. J., 13(2): 19-47. Kaufman, G.M., 1993. Statistical issues in the assessment of undiscovered oil and gas resources. Energ. J., 14(1): 183-215. Krokstad, W. and Sylta, 0., 1996. Risk assessment using volumetrics from secondary migration modeling: assessing uncertainties in source rock yields and trapped hydrocarbons. In: A.G. DorE
1 52
G.M. Kaufman
and R. Sinding-Larsen (Editor), Quantification and Prediction of Hydrocarbon Resources. Norwegian Petroleum Society (NPF), Special Publication 6, Elsevier, Amsterdam, pp. 219-235 (this volume). Merton, R., 1973. Theory of rational option pricing. Bell J. Econ. Manage. Sci., 4: 141-183. Morbey, S.J., 1996. Offshore Brazil: analysis of a successful strategy for reserve and production growth. In: A.G. Dor6 and R. Sinding-Larsen (Editor), Quantification and Prediction of Hydrocarbon Resources. Norwegian Petroleum Society (NPF), Special Publication 6, Elsevier, Amsterdam, pp. 123-133 (this volume). Paddock, J.L., Siegel, D.R. and Smith, J.L., 1988. Option valuation of claims on real assets: the case of offshore petroleum leases. Q. J. Econ., Vol. CIII Aug. 1988, Issue 3, pp. 479-508. Pickles, E. and Smith, J.L., 1993. Petroleum property valuation: a binomial lattice implementation of option pricing theory. Energ. J., 14(2): 1-26. Rose, P.R., 1987. Dealing with risk and uncertainty in exploration:
G.M. KAUFMAN
how can we improve? Am. Assoc. Pet. Geol. Bull., 71(1): 1-16. Rose, ER., 1992. Testimony to Texas Water Commission, June 30 (personal communication). Sinding-Larsen, R. and Chen, Z., 1996. Cross-validation of resource estimates from discovery process modelling and volumetric accumulation modelling: example from the Lower and Middle Jurassic play of the Halten Terrace, offshore Norway. In: A.G. Dor6 and R. Sinding-Larsen (Editor), Quantification and Prediction of Hydrocarbon Resources. Norwegian Petroleum Society (NPF), Special Publication 6, Elsevier, Amsterdam, pp. 105-114 (this volume). Snow, J.H., Dor6, A.G. and Dorn-Lopez, D.W., 1996. Risk analysis and full-cycle probabilistic modelling of prospects: a prototype system developed for the Norwegian shelf. In: A.G. Dor6 and R. Sinding-Larsen (Editor), Quantification and Prediction of Hydrocarbon Resources. Norwegian Petroleum Society (NPF), Special Publication 6, Elsevier, Amsterdam, pp. 153-165 (this volume). Trigeoris, L. and Mason, S.E, 1987. Valuing managerial flexibility. Midland Corp. Finance J., 5(1): 14-21.
SloanSchool of Management, MIT, Room E53-375, Cambridge, MA 02142, USA
153
Risk analysis and full-cycle probabilistic modelling of prospects: a prototype system developed for the Norwegian shelf J.H. Snow, A.G. Dore and D.W. Dorn-Lopez
A probabilistic modelling methodology for reserves, engineering and economic risk analysis was developed for the evaluation of exploratory prospects on the Norwegian shelf. The methodology utilizes a Monte Carlo-type parameter-sampling technique to derive risk versus reserves and risk versus net present value probability distributions; these provide standard decision criteria for a prospect. The system includes an analysis technique that integrates, where possible, historical success ratio with conventional risk factors. Reserves are estimated by sampling distributions for the usual reservoir parameters. After a reserve value is calculated, a broad suite of engineering and economic parameters is sampled assuming that the reserves are developed and produced. A resulting net cash flow is determined based on estimated costs and revenue, and net present value is calculated. This process is repeated for each iteration and reserves and net present value distributions are generated. After the Monte Carlo simulation is completed, the reserves and net present value distributions are modified to include failure cases (dry hole), and total risk versus reserves and total risk versus net present value probability distributions are generated. The mean of the resulting net present value distribution is a measure of the risked net present value (i.e., decision tree solution) of the prospect.
Introduction Analyzing risk and estimating value for exploratory prospects are significant challenges to the oil and gas industry. Following the Norwegian 13th Licensing Round, Conoco Norway set out to study the issue of exploration risk analysis and to make improvements in its approach. This paper presents a probabilistic solution that was developed by a multi-disciplinary team of geoscientists, engineers, economists and computer scientists, from both inside and outside Conoco Norway. This approach is a full cycle analysis which includes evaluation of reserves, engineering and economics, and an estimation of the associated risk of these elements for individual exploratory prospects. The method is called the "Integrated Prospect Evaluation" system (IPE). It is referred to as integrated because it includes a comprehensive evaluation of the reserves, drilling, development, production and economics, for a range of possible outcomes for an exploratory prospect. These outcomes include drilling dry holes, finding reserves and abandoning them, or finding reserves and developing and producing them. The integrated nature of the system encourages co-operation between geoscientists,
engineers and economists in the early assessment of the value of a prospect (see Corrigan, 1993, for a discussion of the desirability of such interaction). The method is probabilistic. Most of the input parameters that are used for the model can be defined as frequency distributions, rather than as single deterministic values (e.g. Cronquist, 1991). The advantage of the probabilistic approach is that the parameter statistics can generally be derived or estimated from historical or analog data, and the uncertainty in these parameters can be quantified. The results from this approach are distributions of total probability versus reserves, and total probability versus net present value (NPV). Inverse cumulative curves are used to represent these distributions, shown schematically in Fig. 1.
Methodology The IPE methodology incorporates decision tree logic, as described in Newendorp (1975). The decision tree, shown in Fig. 2, includes four possible outcomes: the right branch is the zero value if no well is drilled. The left branch of the tree includes the NPV of the dry hole commitment, and two discovery
Quantification and Prediction of Petroleum Resources edited by A.G. Dor6 and R. Sinding-Larsen. NPF Special Publication 6, pp. 153-165, Elsevier, Amsterdam. 9 Norwegian Petroleum Society (NPF), 1996.
154
J.H. Snow, A.G. Dor~ and D.W. Dorn-Lopez NPV'S OF
CHANCE OF DRY HOLE ( ZERO RESERVES )
100% C U
~ UNECONOMIC
PROBABILITY OF TESTABLE HC
M.
DISCOVERIES
100% -- D c
NPV OF
U
J
M.
DRY HOLE ECONOMIC
UNECONOMIC
P
BREAKEVEN
R
O
B
A
B I
I
~
BREAKEVEN , RESERVE
~
NPV'S OF ECONOMIC
L
I
S
T Y
o%
0%
0
-
~,
RESERVESIZE
0
~_~-!-
NET PRESENTVALUE
Fig. 1. Final reserve and Net Present Value inverse cumulative distributions from IPE analysis. The mean or average value of the NPV distribution is risked NPV or EMV, and the yes solution to the drill? decision tree branch.
yes ~
MONTE CARLO SIMULATION
no
DISCOVERY? Phc 1 -Phc
yes
In~ 1
I MINIMUM RESERVES? I Pdev 1 -Pdev
IIII
NPV OF TOTAL APPRAISAL, TESTINGAND DEVELOPMENT
III
17
NPV OF APPRAISAL & TESTING
II NPV I OF DRY HOLE I
I
COMMITMENT
NPV OF DOING NOTHING (ZEROVALUE?)
Fig. 2. Integrated Prospect Evaluation decision tree. Outlined branches indicate portion of the analysis done by Monte Carlo analysis.
outcomes: NPV if a discovery is developed, and NPV if a discovery is tested, found to be to small, and abandoned. These last two branches are solved by a Monte Carlo simulation (Newendorp, 1975; White and Gehman, 1979). A deterministic approach to solving this type of decision tree would be to generate representative NPVs for the dry hole and discovery categories, assign probabilities at the discovery branch, and then calculate risked NPV. This approach is not particularly repeatable because the methodology is poorly defined. Any two people could derive significantly different NPVs and probabilities on the same prospect and, consequently, they would also derive significantly different risked NPV measures for the value of the prospect.
The final net present value distribution, the curve on the right in Fig. 1, includes outcomes from the three "drill" branches of this decision tree. The mean, or expected value of this distribution, is the solution or risked net present value of the left branch of the tree. In contrast to the deterministic approach, this value is obtained by statistically averaging a broad range of outcomes of the model according to their probability of occurrence. The reserve and NPV distributions are obtained from the product of two different probability estimation processes (Fig. 3). The first of these is estimation of the chance of finding any testable hydrocarbon before a well is drilled. The second is a Monte Carlo simulation which estimates reserve size and value,
155
Risk analysis and full-cycle probabilistic modelling of prospects
Fig. 6. Basic method for combining historical success rate and comparison coefficients to determine chance of testable hydrocarbons.
currences of finding zero reserves. In this cartoon, 33% of the time there is a dry hole or zero reserves and 67% of the time, reserves of some none-zero size are found. The bottom graph, an inverse cumulative distribution display of the middle frequency distribution, shows the total probability for finding a particular reserve size or greater, for the prospect. The same approach is also applied to the NPV distribution except that occurrences of the net present value of the dry hole cost would be added to the NPV distribution given discovery.
IPE modules
Fig. 4. Basic concept of the IPE system. Top: standard frequency distribution for a factor such as reserves or NPV. Center: frequency distribution including dry-hole or non-discovery cases. Bottom: derived inverse cumulative distribution of center frequency distribution.
given that testable hydrocarbons have been found. This is most easily understood as a sum of two frequency distributions (see Fig. 4). The top graph is a frequency distribution of the occurrence of different possible reserve sizes given that a discovery has been made. This is the result obtained from the Monte Carlo simulation. The middle graph has added the estimated occurrence of drilling a dry hole, or in other words, oc-
Fig. 5 shows that the input to the IPE methodology is organized into modules. These modules generally represent the major parameter categories in the model. Also indicated are the key people that would be responsible for defining the parameters required for that portion of the process.
Chance of hydrocarbons module The first step in the methodology is estimation of the chance, or probability, of finding any testable hydrocarbons. This process is summarized in Fig. 6. In this calculation the probability of hydrocarbons is the product of the historical success rate and four coefficients. These coefficients represent numerical comparisons of the prospect's reservoir, seal, trap, and source characteristics. The comparison is made
156
J.H. Snow, A.G. Dor~ and D.W. Dorn-Lopez
against the prospects/play from which the historical success rate was derived.
general, reliability will improve with the maturity of the of the play, basin or analog under consideration.
Historical success rate
Comparison coefficients
One method for risking reserves is to estimate a probability of finding a specific reserve size. Studying this approach, we concluded that it was unreasonable for someone to try to estimate the probability associated with a particular reserve size, and that it was not repeatable. Rather, it was more reasonable for someone to estimate the chance of simply finding some testable hydrocarbons, without having to define how much. Testable hydrocarbons does not mean economic hydrocarbons. It is any amount that could be tested or flowed to surface; essentially, reserves greater than zero. A key data point for this analysis is the historical success rate for finding testable hydrocarbons. This is defined for the play or trend containing the prospect. The historical success rate is a critical measure of the prospect's chance of having any oil or gas (White and Gehman, 1979), and the prospect being evaluated is compared, component by component, to the play from which the success rate was derived. From this comparison, the prospect's chance of finding testable hydrocarbons is calculated. Historical success rate will change as a play matures. In plays where essentially all prospects in the play are available for license, as prospects are tested and experience is gained the success rate should improve. Once a play matures, the pool of available prospects diminishes, more of the poorer quality prospects are tested and the success rate will eventually begin to decline. In plays of this type the current success rate, or an extrapolation from the latest trend of the historical success rate, should be used for evaluating undrilled prospects. In areas where only a limited number of selected licenses are offered, for example the Norwegian type of licensing rounds, the current historical success rate may be influenced more by the quality of prospects that were awarded in the latest round and may not be representative of the play over a longer period of time. In these types of plays, an obvious increasing or decreasing trend with time may not emerge, and an average historical success rate would be the more appropriate measure. In the absence of historical data for the play, success rate for the basin or the nearest analog can be used as an indication of the prospect's chance of hydrocarbons. Comparison would then be made between the prospect and the analog. The reliability of the historical success rate for a play, basin or analog should be addressed for each prospect evaluated. In
The comparison coefficients range from "worse than" to "better than", measured against those prospects from the historical play. As shown in Fig. 7, the coefficient values are determined from the qualitative answers to two questions. The first question is: how does the prospect's characteristic, such as reservoir quality, compare with that of prospects in the play? The second question is: how good is your knowledge to make this comparison? The answers have a matrix of weighting profiles associated with them, two examples of which are shown in Fig. 8. If the answer to the questions were, "worse than", with "intermediate knowledge", a coefficient value of 0.8 would be returned (see Fig. 8). This single coefficient value is obtained by taking the products of the "worse than" with "intermediate knowledge" weighting profile values, which range between zero and one (dashed line in Fig. 8), and coefficient values, which are 1.5, 1.25, 1.0, 0.75 and 0.5. These products are taken at five discrete points then summed. In the second example (solid line in Fig. 8), the answer was "much better" with "direct knowledge". The coefficient here would be nearly 1.5.
COEFFICIENT VALUES ARE DETERMINED FROM T W O QUESTIONS:
1. HOW DOES THIS FACTOR (RESERVOIR, TRAP, SEAL OR SOURCE) COMPARE TO THAT OF PROSPECTS FROM THE HISTORICAL PLAY? better >1, same =1, worse <1
2. HOW GOOD IS THE INFORMATION YOU ARE USING TO MAKE THIS COMPARISON? direct data - high certainty intermediate data - moderate certainty indirect d a t a - low certainty
Fig. 7. Determination of comparison coefficient values.
1.0
.~
"WORSE" "INTERMEDIATE DATA" AVERAGE VALUE 0.80
"MUCH BETTER" "DIRECT DATA" AVERAGE VALUE .~.. 1.47
j
-
iq o-,d
1.5 MUCH BETTER
! 1.25 BETTER
SAME AS
0.-~5
WORSE
(.5 MUCH WORSE
COEFFICIENT VALUE
Fig. 8. Example weighting profiles and derived average values.
lb/
Risk analysis and full-cycle probabilistic modelling of prospects As these two examples demonstrate, when there is better knowledge about a specific coefficient, that coefficient will have a greater impact of in modifying the historical success rate, As knowledge decreases, the influence of the coefficient will decrease. These questions are answered for each characteristic: reservoir, source, seal and trap. A value between 0.5 and 1.5 is obtained for each coefficient. The product of the four coefficients is then used to modify or adjust the historical success rate, to obtain the expected chance for the prospect. Given the mathematics of this approach (Fig. 6), there is the possibility that the product of the four "better than" coefficients and a high historical success rate could generate a nonsensical value above one (greater than 100% chance of discovery). To avoid this, we apply the simple solution of never allowing the product to exceed one. Our experience, however, has shown that there is a more practical limiting factor. When the historical success rate for finding testable hydrocarbons is high, it is very difficult for more than one or two of an undrilled prospect's characteristics to be "better than" those of the play. A good prospect's characteristics are more often judged to be the "same as" those of the high success rate play, but rarely better. In practice, the chance of hydrocarbons estimated using this approach rarely needs to be limited. Clearly, subjectivity remains with this method, and the accuracy of the technique is uncalibrated at present. However, more critically, is that the method gives results that are consistent and repeatable and allows for future calibration as results are obtained. Our experience in using this approach shows that different people can generally agree on a historical success estimate for a specific prospect. And while they may argue over whether a coefficient is, for example, the same as or worse, they would both agree that the coefficient is not better. Consequently, the relative risk between prospects can be measured more consistently with this approach and re-calibration using statistical results should be possible in the future.
Reserves module The next module in the IPE process is reserve estimation (Fig. 5). This is the first step in the Monte Carlo simulation. The parameters and equations for reserves calculation are the standard ones shown in Fig. 9. These include: oil and gas gross rock volumes, net to gross ratio, porosity, water saturation, formation volume factor, gas expansion factor and recovery factors for oil and gas. Associated gas and condensate are calculated from the oil and gas reserves using the gas oil ratio and condensate yield.
[R oil = G R V o i l * N E T / G R O S S
* Q * (1-Sw) * I/FVF * RF] +
i
I R ass g a s = R oil * I R gas = G R V g a s * N E T / G R O S S
GOR I i
* Eli * ( 1 - S w ) * G E F * RF I
-I.
I R ass c o n d = R g a s * C O N D E N S A T E
YIELD I
I RESERVES I Fig. 9. Standard equations used in calculation of reserves for hydrocarbon phases.
Parameter distribution types In the model, each of these parameters is defined as a frequency distribution. Two examples are shown in Fig. 10. The frequency distribution defines how that parameter is sampled during the Monte Carlo simulation. A mean, standard deviation, and specified distribution types are generally used for most parameters. These distribution types could be normal, log normal, triangular or even constant. These statistical measures are generally obtainable from historical data for most of the reservoir parameters. For example, a mean value and standard deviation for water saturation can easily be calculated from a frequency distribution of nearby or analogous wells. The distribution type could be observed from a histogram plot of those data. Where historical statistics are lacking, an estimated range and predicted "mid value" can be specified with a triangular distribution, based on the interpreter's estimate of the minimum possible, the maximum possible and a mid or average value.
Gross rock volume distribution The most critical and influential parameter impacting the reserve estimate is gross rock volume. While the estimate of most parameters varies by less than one order of magnitude, gross rock volume can vary by up to 10 orders of magnitude or more. Consequently, when the gross rock volume variance is large, which is generally the case with exploratory prospects, the shape of the gross rock volume frequency distribution largely determines the shape of the resulting reserves distribution. The importance of this parameter is demonstrated by the two reserve distribution examples shown in Fig. 11. The only difference in these two distributions is how the frequency distribution of the gross rock volume parameter was defined. All other reservoir parameters were the same in both models. Both the distribution shape and consequent mean values are significantly affected.
158
J.H. Snow, A.G. Dord and D.W. Dorn-Lopez
Fig. 12. Model used to risk-weight hydrocarbon fill of gross rock volume in IPE system. Left: simple model for volume versus depth, derived from mapping. Right: frequency distribution assigned to weight hydrocarbon fill cases.
159
Risk analysis and full-cycle probabilistic modelling of prospects Because of this high degree of influence, gross rock volume deserves special attention. In defining the gross rock volume frequency distribution, one is in fact, defining the occurrence or frequency of how the prospect's available container is filled with hydrocarbons, and where the hydrocarbon-water contact is. The schematic example in Fig. 12, shows a cross section of a trap, which has an available hydrocarbon column and volume. The maximum volume is defined by the maximum column supported by the seal or the spill point or leak point of the structure. If hydrocarbons are there, somewhere within this volume we expect to find an oil-water, gas-water or perhaps a gas-oil contact. The frequency distribution on the fight side defines how the oil-water contact and associated volume is selected for each iteration. In this case, a distribution has been defined that indicates the structure is equally likely to be filled between zero and about half way. Beyond that, it has continually decreasing likelihood for filling to the maximum available column. When defining frequency distributions for the trap fill, the interpreter is encouraged to take into consideration such factors as placement of fault compartments and sealing faults. However, in addition to trap fill, there is uncertainty in the geometry of the closure from mapping, depth conversion, data quality, etc., which affects the estimation of gross rock volume. In the current model, once a hydrocarbon fill has been selected for that iteration, the associated gross rock volume is assumed to be 100% correct for that trap fill. We recognize that gross rock volume uncertainty is an additional parameter. It is a limitation to the current model that this is not simulated specifically, and the interpreter must also take this uncertainty into account when defining the trap fill frequency distribution.
Multiple hydrocarbon phases For multi-phase reservoirs an additional parameter, called percent gas column, is defined. The percent gas column parameter, also input as a frequency distribution, determines how the total gross rock volume is divided into oil and gas volumes. For each total gross rock volume chosen, there is an associated total hydrocarbon column. The sampled percent gas column parameter then determines how much of that total column is gas. This in turn defines the gas volume and the remaining volume is then oil. Defining the weighting on the gross rock volume and percent gas distributions is potentially the most critical part of the reserve estimation process. How well this weighting is understood and defined for the prospect determines, to a large extent, how good
I Total Oil = R oil Res I + R oil Res 2 + .... I -i-
I Total Ass Gas = R ass gas Res 1 + R ass Gas Res 2 + ... I §
[ Total Gas = R gas Res 1 + R gas Res 2 + ..... I +
I Total Ass Cond = R ass cond Res 1+ R ass cond Res 2+.. ] I T O T A L P R O S P E C T RESERVES I
Fig. 13. Total reserves calculation for all hydrocarbon phases.
the reserve estimate is. These need to be carefully assessed for each prospect.
Total reserves A full set of reservoir parameters for each reservoir in the prospect is defined. As shown in Fig. 13, oil, gas and associated products from each reservoir are then summed and the total reserves are calculated for each iteration.
Production and expenditure forecasting module The third module in IPE deals with exploration and development scheduling, production forecasting and costing. The main criteria associated with developing a discovery are summarized in Fig. 14. These are: the exploration phase which includes seismic acquisition, exploratory and appraisal drilling, the construction and development drilling phase, and finally, the oil and gas production phase.
Exploration expenditure The key elements of appraising, developing and producing a particular reserve are how much it will cost, how fast it will produce, and how these elements will be scheduled over time. Consequently, the parameters which comprise the engineering module mostly define how much or how long, and are generally dependent on reserve size for a given prospect. In the real world, the number of wells and duration of exploratory or appraisal drilling of a discovery will vary, depending on the size of reserves found and the complexity of the prospect. Four parameters to define this phase are well cost, testing cost per well, the number of wells, and the total duration of the appraisal drilling period. The last two parameters are defined by look-up tables. An example of a look-up table is shown in Fig.
J.H. Snow, A.G. Dor# and D.W. Dorn-Lopez
160
Fig. 14. Full-cycle project scheduling as simulated in the IPE system.
Duration of Appraisal Drilling (Years) Reserves (MMSCM)
Triangular Distributions Low Mid High (years) (years) (years)
0.00
2.80
3.00
5.00
16.00
3.00
4.00
7.00
50.00
4.00
5.00
8.00
80,00
5.00
6.00
10,00
Fig. 15. Example look-up table for determination of the duration of appraisal drilling.
15. In this example, for a reserve size greater than 16 million standard cubic metres (MMSCM), a triangular frequency distribution for duration of appraisal drilling ranging between 3 and 7 years would be sampled for that iteration. For a 50 MMSCM reserve size, a different distribution, between 4 and 8 years, would be sampled. The default look-up tables are based on an experience database for the Norwegian shelf. In this area, as elsewhere, larger discoveries tend to take longer to appraise and develop. Using this technique provides appraisal costs and durations appropriate for the reserve size calculated in that iteration. For the seismic acquisition phase, specific years that seismic data would be acquired are defined. These times are not distributions and are not reservedependent. However, the cost would be defined as a distribution, one for each year designated.
Production forecasting The next step in the process is estimating the oil and gas production profiles. A schematic production profile is shown in Fig. 16. The start year for production is determined from the sum of: the duration of exploratory drilling, lag time, and the duration of construction. Lag time is simply a distribution of additional time that can be added to reflect a possible delay between when the prospect is delineated and when construction starts. For each iteration, oil reserves are produced first. Gas production then starts when the oil production rate has started to decline. Minimum quantities of both oil and gas are defined, such that if a small amount of, say, oil were calculated for that iteration, only the gas would be produced. Conversely, only the oil would be produced if the gas reserve was smaller than the minimum. The key parameters for defining the production rates are shown in the schematic production profile (Fig. 16). Plateau rate and the amount of reserves produced on plateau are both calculated as a percentage of reserves. These parameters both use look-up tables to define reserve size versus a triangular distribution of percentages. This type of relationship is demonstrated in Fig. 17, which plots plateau rate as a percentage of reserves from producing Norwegian fields. Superimposed are the graphical representations of the triangular distributions which are defined in the look-up tables. In this example, the modelled distribution is conservative with respect to the historical data for low reserve size cases.
161
Risk analysis and full-cycle probabilistic modelling of prospects PRODUCTION
R A T E (per yr)
P L A T E A U R A T E (per yr) =% R E S E R V E S
16.00
S T A R T OF D E C L I N E A T % TOTAL RESERVES PRODUCED
14.00
12.00
L_ (9
>.
10.00
O
8.00
03
E
%
D E C L I N E R A T E (per yr) = % PLATEAU RATE
6.00 START PRODUCTION
4.00
PLATEAU BUILD UP = % P L A T E A U RATE/# Yrs to plateau
2.00
0.00
(kl
"~"
~0
O0
0
Od
~
~0
O0
0
C~
~
~0
O0
0
C~I
~
QO
O0
0
O~ v-
O~ v-
O~ v-
O~ ~-
0 OJ
0 OJ
0 OJ
0 O4
0 C~
0 OJ
0 OJ
0 O~
0 OJ
0 C~
0 OJ
0 C~
0 C~
0 ~
0 CW
0 OJ
YEARS Fig. 16. Typical production profile used in the IPE system. Production levels and durations are derived from reserve size.
30 - - - IPE Default High or)
20
or)
*" rr
. . . . . IPE Default Low 9 Field Data
or) rr O
IPE Default Mid
25
-p % I
15
~"~
~
m
,.__I
-1_
limB
~ Q.
10
-',
'
9
9
,. . . . . . . i I
0
I i
9
9
[]
....................................................................................
100
200
300
400
500
600
Oil Reserves (106 Sm 3 ) Fig. 17. Plateau rate per year versus oil reserves (field data from County Natwest Woodmac, 1992).
The other two parameters in the production profile are the plateau build-up and decline rate. Plateau build-up is defined from the plateau rate and the number of years estimated to reach plateau. The decline rate is a normal distribution of percentage which defines the plateau rate reduction per year. If both the oil and gas reserves are less than a predetermined amount, no development will occur for that iteration. Only the exploratory costs, both seismic and drilling, will then be included. This cut-off is usually set as a reasonable minimum reserve for the area, based on reserve sizes that have been developed. This is a reserve size criterion, not economic value, and differentiates between the "de-
velop" or "walk away" branches of the decision tree (Fig. 2).
Construction and development expenditure Reserve-dependent look-up tables are also used for the duration of construction and the duration of development drilling. Together, these define the total duration for the development phase of the project. The total costs associated with the construction and development phase are derived from a correlation of total capital expenditure with estimated oil equivalent production rate per year. This relationship is derived from data for 36 fields
162
J.H. Snow, A. G. Dord and D. W. Dorn-Lopez
that have been developed or will be developed on the Norwegian shelf (e.g., County Natwest Woodmac, 1992). In practice, this general correlation is not used for all exploratory projects, recognizing that a prospect's location, particularly with respect to infrastructure and water depth, can significantly impact the costs. Instead, a customized relationship must be derived for each prospect. This provides cost estimates, specific to the prospect, for the full range of costs associated with the prospect reserves distribution. Of course, comparison of the customized curve to the general curve is done to confirm that cost estimates are realistic. For each iteration, a value for the total capital expenditure is calculated. This calculated value is used as the mean of a normal distribution, and a percentage of that expenditure is defined as a standard deviation. This normal distribution is then sampled for that iteration. Once a total capital expenditure value is determined, operating costs and abandonment costs are then estimated as percentages of that value. This relationship is based on statistical analysis of historical data. Once the total capital expenditure and the duration of development have been obtained for an iteration, the cost expenditure profile is generated. This then defines how expenditures are scheduled over time. Many of the engineering parameters are based on look-up tables which correlate that parameter to reserve size. As stated earlier, all of these parameters have been defined by a set of default values, based on historical data from the Norwegian shelf or common engineering rules-of-thumb. These defaults are
UNECONOMIC DISCOVERIES
1 /0 0 % - -C~ U
M. R
O
'~
B A B I
The fourth module in IPE, and last part of the Monte Carlo simulation, is the economic and tax analysis (Fig. 5). From the engineering analysis, the exploration costs, development expenditures, operating expenditures and production, have all been determined and scheduled over time. Key parameters for the economic analysis are working interest, inflation rate, discount rate, tariffs and product prices. For oil and gas prices, a broad triangular distribution is used to represent the price forecast. A distribution is specified for each year, for as many years as are needed in the model: typically 30 to 60 years. The distribution would typically be based on the current corporate forecast for oil and gas prices. Four tariffs are included: oil transportation and processing and gas transportation and processing.
100% C "
M. p
BREAKEVEN
RESERVE
L T Y
~
~
NPV'S OF U DISCOVERIES NEcONOM cI
I
ECONOMIC
REAKEVEN
R
O
B A B I
ES
L T Y
I
0%
Economics module
U
~
P
suitable "as is" for some prospects. However, each of these look-up tables should be examined, and where appropriate, customized to the prospect being evaluated. Such factors as trap geometry and complexity, expected drive mechanism, reservoir homogeneity, etc., should be taken into account. For some prospects, there may be parameters for which the relationship with reserve size is not clear or even in correct. In these cases, this type of model may not be the most appropriate method for estimating their worth. However, experience in evaluation of prospects on the Norwegian shelf, is that the reservedependent parameters give average results that are very similar to what would be derived from deterministic analysis.
I
0
RESERVESIZE
~
0%
- ~
0 NETPRESENTVALUE
~+
Fig. 18. Reserve and NPV inverse cumulative distributions derived from Monte Carlo analysis. These represent the unrisked or given discovery case.
Risk analysis and full-cycle probabilistic modelling of prospects Any distribution type can be used to represent these different tariffs depending on the data available for estimating these parameters. From the engineering results and economic assumptions, the before-tax cash flows, tax payments and the key result, after-tax cash flow, are calculated. From the cash flows, net present value and internal rate of return are determined. This then completes the full-cycle Monte Carlo analysis.
Given discovery and before discovery results The entire process, reserves through to net present value, is repeated for each iteration. The distributions shown in Fig. 18 summarize these iteration data. These are referred to as the "given discovery" reserves and NPV distributions for the prospect. They are the "unrisked" probability distributions, unburdened by the risk of finding hydrocarbons. The frequency of drilling dry holes, which was estimated from the probability of testable hydrocarbons module, is used to scale these distributions to derive the "before discovery" final results, or the total probability versus reserves and NPV curves (Fig. 1). From these final distributions, it is possible to determine the complete range of outcomes and associated probabilities from the model. These include the probabilities of discovering any particular reserve sizes, of making certain amounts of money, the probability of losing money, and how much money can be lost if everything goes wrong. A minimum economic or break-even reserve size can be graphically determined.
Break-even reserves
163
ities. Fixing some parameters and varying others shows very clearly which parameters have the greatest impact on the results. In the example shown in Fig. 21, we have progressively fixed parameters, starting with the reserves, then engineering and finally economics. The resulting standard deviations are numbered 1 through 5. Reserve uncertainty has the greatest impact for this prospect. However, uncertainty in engineering cost and product prices also cause a substantial variation in the net present value of the project. In case 2 (Fig. 21), the fixed reserve case, the
CUMULATIVE FREQUENCY
I
I
I
-1'0
-5
0
i
5
I
1~0 1'5
....
NPV (UNITS ARBITRARY)
SELECTED RANGE OF NPV CENTERED AROUND ZERO
1
Break-even reserves can be determined in a more elegant or rigorous way during the simulation by capturing reserve values that generate NPVs around the zero value. This is shown schematically by the two graphs in Fig. 19. This produces a reserve distribution, the mean of which represents the average economic break-even reserve size. As shown in the actual reserve distribution example in Fig. 20, there is a large range of reserves that can give a small or zero NPV. This is typical for most models. This also shows that, while the reserves may be larger than what is considered the economic breakeven, the project can still make no money or lose money in some cases.
Sensitivity analysis Another aspect of the full-cycle Monte Carlo model is the ability to analyze parameter sensitiv-
Fig. 19. Statistical method for deriving break-even reserves in the IPE system. The upper plot is an inverse cumulative distribution for Net Present Value. The lower plot is a frequency distribution of all reserves cases (sampled by the system) that resulted in close to zero NPV. The expected value (mean) of the lower distribution is a measure of the break-even size.
J.H. Snow, A.G. Dor~ and D.W. Dorn-Lopez
164
In a traditional deterministic evaluation of a prospect, all of the input assumptions may have validity and result from meticulous ground-work. The result of this detailed work, however, is a single set of figures which statistically has a very small chance of being realized. The basic value of the IPE approach is that it includes a broader range of outcomes, both good and bad, than can be evaluated using deterministic methods. Input is statistically founded on real, historical data and uncertainty can be quantified. Consequently, the approach provides a more complete, realistic and consistent evaluation of the possible outcomes for a prospect.
Fig. 20. Example break-even reserve histogram.
value of the project can still vary by plus or minus 500 million Norwegian kroner. In other words, the value ranges between 250 and 1250 million kroner, a substantial range.
Conclusions The IPE methodology integrates traditional decision tree analysis with Monte Carlo simulation and applies these techniques to the entire life cycle of a prospect. The integrated nature of the model invites criticism that the system is a "black box" or "magic bullet" for solving problems or making decisions. Care has been taken, therefore, to make the system interactive and to make all implicit assumptions visible to the user.
1.0
I
Acknowledgements The authors wish to thank Conoco for permission to publish this paper. Also gratefully acknowledged are the contributions made by Thorbj0n Pedersen, Randy Theilig, Kurt Thomas, Bob Dixon, Anders Gjesdal and Kjetil Ausland, to the various IPE modules. We thank Stuart Anderson (ICTS) for his major part in programming the system. Andrew Conway, Michael Frost, Tina Langtry, Sayers Kyle, Jim McColgin and Matt Strickland (Conoco), and B.A. Duff and D. Hall (Fina) are acknowledged for critical reading of the manuscript. Most of all, however, we are grateful to the geoscientists, engineers and economists of Conoco Norway Inc. for their extensive beta testing of the IPE system during the Norwegian 14th Round and other exercises.
I
I
I
1. FULL IPE 2. FIXED RESERVES 3. FIXED RESERVES + ENGINEERING (NOT CAPEX) 4. FIXED RESERVES + ENGINEERING + CAPEX 5. TYPICAL CASE
0.9 0.8 0.7 0.6 ..Q ..O O
0.5
\
0.4 0.3 0.2 0.1 0.0 I
-500
.
.
.
.
.
0
Mean: 638.6 Standard deviation' 831.7
.
.
.
500
1000
1500
2000
2500
3000
After-tax NPV Given Discovery (106 NOK)
Fig. 21. Comparison of IPE runs with different levels of fixed parameters. The horizontal bars show the standard deviation associated with each case.
165
Risk analysis and full-cycle probabilistic modelling of prospects
References Corrigan, A.F., 1993. Estimation of recoverable reserves: the geologist's job. In: J.R. Parker (Editor), Petroleum Geology of Northwest Europe, Proceedings of the 4th Conference. Geological Society, London, pp. 1473-1481. County Natwest Woodmac, 1992. Denmark, Ireland, Norway. Northwest Europe Service, Reference section Volume 2.
J.H. SNOW A.G. DORI~ D.W. DORN-LOPEZ
Cronquist, C., 1991. Reserves and probabilities m synergism or anachronism? J. Pet. Technol., 43: 1258-1264. Newendorp, ED., 1975. Decision Analysis for Petroleum Exploration. Petroleum Publishing Company, Tulsa, 668 pp. White, D.A. and Gehman, H.M., 1979. Methods of estimating oil and gas resources. Am. Assoc. Pet. Geol. Bull., 63(12): 21832192.
Conoco Norway Inc., P.O. Box 488, N-4001 Stavanger, Norway Statoil UK Ltd, Swan Gardens, 10 Piccadilly, London WI V OHL, UK Conoco Norway Inc., P.O. Box 488, N-4001 Stavanger, Norway
This Page Intentionally Left Blank
167
Play fairway analysis and risk mapping: an example using the Middle Jurassic Brent Group in the northern North Sea Shona Grant, Nick Milton and Mark Thompson
A methodology for play fairway analysis and risk assessment is illustrated using the Middle Jurassic Brent Group play in the northern North Sea as an example of the technique. This is the most successful play in the northern North Sea with over 50% of the discovered reserves. Prospect risk can be subdivided into play risk and prospect-specific risk. The play risk comprises the regional risk elements, i.e. those elements of risk which can be estimated and mapped regionally without detailed mapping of the prospect structure. Prospect specific risk reflects local risk elements within the fairway. In order to assess relative risk across the play fairway, each play can be subdivided into several risk elements or factors. These include: - the presence and effectiveness of a reservoir; - the presence of a source rock and the effectiveness of the carrier system; - the presence of an effective vertical seal. Each risk element is then subdivided into areas of common risk. This is done by subdividing each element into areas of low, moderate or high risk and assigning a corresponding green, yellow or red colour to each to produce a common risk segment (CRS) map. In the case of reservoir, source or seal presence, the delineation of common risk areas is based on our regional understanding of the basin stratigraphy. Further data are then integrated to assess the risk that the predicted stratigraphy provides an effective reservoir, carrier system and/or seal. This data include core porosity and permeability, pressure and leak-off data, well test results, geochemical data, well log analysis and thermal/basin modelling work. Other regional risk elements (such as timing of trap formation and biodegradation) are often mapped but are not important regional risk elements for the play discussed here. Individual common risk segment maps can be combined to provide play fairway summary maps and composite common risk segment (CCRS) maps for each play. These provide a powerful pictorial representation of relative risk within the play fairway. Uncertainty maps are also produced for each play. These illustrate our confidence in the geologic model. They are controlled by the density, quality and reliability of well and seismic data. They are used in conjunction with the risk maps, taking care not to confuse risk with uncertainty.
Introduction B P has developed a system of regional play analysis, which was used extensively in the evaluation of acreage during the 14th Licensing round on the Norwegian continental shelf. All potential plays in the northern North Sea were analysed in terms of relative risk. In this way it was possible to high grade plays within the blocks on offer, and to use drilling success ratios as a constraint for risking individual prospects. The methodology is described below using the Middle Jurassic Brent Group as an example. The North Viking Graben of the Northern North Sea (60~176 is a major petroleum province (Fig. 1) with several working play systems at many different stratigraphic levels. The reserves discovered in the major plays for the UK and Norwegian sec-
tors (UKCS and NOCS) combined are illustrated in Fig. 2a. Also shown in Fig. 2b is the number of target tests by play. The Brent Group play is the most successful to date with over 50% of the discovered hydrocarbon reserves. This corresponds to some 15.6 billion barrels of oil in 82 separate pools. The Brent Group comprises Aalenian to Early Bathonian age deltaic sediments (Mitchener et al., 1992). Oil is mostly sourced from Middle to Late Jurassic marine mudstones and trapped in tilted fault blocks (Bowen, 1992). A large numbers of prospects and leads remain undrilled in the Norwegian sector (NOCS). The play is analysed in more detail below as an example of the use of common risk segment maps (defined below), play maps, and drilling success ratios for risk estimation and play fairway analysis. The maps demonstrate the underlying geolog-
Quantification and Prediction of Petroleum Resources edited by A.G. Dor6 and R. Sinding-Larsen. NPF Special Publication 6, pp. 167-181, Elsevier, Amsterdam. 9 Norwegian Petroleum Society (NPF), 1996.
168
S. Grant, N. Milton and M. Thompson
Fig. 1. Location map showing existing discoveries in the North Viking Graben.
ical understanding of the play system under review. In the text, the North Viking Graben stratigraphy is described in terms of the established lithostratigraphy (Deegan and Scull, 1977; Vollset and Dor6, 1984) for the Northem North Sea and B P's genetic sequence stratigraphy (presented in Mitchener et al., 1992, and Partington et al., 1993). The relationship between the two is shown in Fig. 3. For example, the
Middle Jurassic Brent Group reservoir interval corresponds to the J20-J32 genetic sequence stratigraphic units.
Terminology A play is defined here as a grouping of prospects with one or more common factors. Throughout this
169
Play fairway analysis and risk mapping
Fig. 2. Discovered reserves (a) and number of wells drilled (b) subdivided by sequence for the North Viking Graben.
study, plays have been defined by their reservoir stratigraphy, thus all prospects with a Brent Group reservoir form the Brent play. A play fairway, in this case, is therefore the area defined by the maximum possible extent of reservoir rocks within the stratigraphic interval of the play. The limits of the Brent play are defined by the maximum possible limits'of the Brent Group reservoir rocks. Prospect-specific risk consists of those elements of risk which can only be determined locally (e.g. trap risk, specific fault seal risk, local reservoir erosion). This is equivalent to the prospect success factor of White (1992). Play risk consists of regional risk elements, which can be determined by regional mapping, and are not just specific to a single prospect. Our play risk is equivalent to the play chance of White (1992). Thus the probability of discovering petroleum in a prospect often includes elements of play risk as well as those risks solely associated with the prospect, i.e."
Overall Prospect Chance = Play risk x Prospect specific risk It is possible to draw a map for each of the regional risk elements showing areas of common relative risk, for example, areas where there is a relatively high risk (low chance) of finding an effective reservoir due to depth of burial. These areas are termed risk segments, and the map of a regional relative risk element is termed a common risk segment (CRS) map. All of the regional risk elements for a specific play can be combined into one map which illustrates the overall variation in play risk across the fairway. Such maps are termed composite common risk segment (CCRS) maps or play fairway summary maps.
Play fairway definition The extent of a play fairway is here defined as the maximum possible extent of reservoir rocks. This is based on sequence stratigraphic analysis of well
170
S. Grant, N. Milton and M. Thompson
Fig. 3. Pre-Cretaceouslithostratigraphyand geneticsequencestratigraphyin the NorthVikingGraben. and seismic data to predict the distribution of systems tracts. The maximum basinward extent of reservoir is based on the predicted distribution of the low stand
deposits. An example of the resulting depositional environment map for the Late Bajocian part of the Brent Group (which includes the Upper Ness Fm.) is
171
Play fairway analysis and risk mapping shown in Fig. 4. Similar maps have been constructed for all sequences within the Brent Group, in order to define reservoir limits (Milton and Ewen, in prep.).
Risk segment mapping Play risk can be subdivided into various regional risk elements. The regional risk elements which have been mapped and evaluated for the North Viking Graben are: - reservoir presence; - reservoir effectiveness; - top seal; - hydrocarbon charge. Other potential regional risk elements, such as timing of trap formation, or biodegradation of trapped oil, are uniformly favourable (low risk) for the Brent play. A map is produced for each element showing areas of common relative risk. The assessment of risk is qualitative in terms of low (green), moderate (yellow) and high (red) risk. "Low risk", for a particular regional risk element, implies that this element will be favourable, e.g. there is evidence from regional work that reservoir will be present. "High risk" implies evidence that the element is unfavourable, e.g. that there is evidence that reservoir will be absent. "Moderate risk" may reflect conflicting or absent evidence. Maps are presented below for the Brent Group play to illustrate the technique.
Reservoir presence risk Reservoir presence risk maps have been derived from semi-regional geological and geophysical analyses, and show the risk of reservoir facies being present. Generally these maps show low risk (green) segments where reservoir facies is proven by drilling, or can reasonably be predicted by all likely geological models. High risk (red) segments contain some evidence for the absence of the reservoir interval, while moderate risk (yellow) segments are often zones of little knowledge. The geological model for the deposition of the Brent Group is based on sequence stratigraphic analysis of biostratigraphic, wireline and core data (Milton and Ewen, in prep.). An example of the depositional environment map for the Late Bajocian is shown in Fig. 4. The resulting reservoir presence common risk segments map is presented as Fig. 5. The low risk area (green) represents the area of the proven conventional Brent play. The yellow area to the north represents the unproven extension of the conventional play. It includes a postulated low stand system north of conventional limit of the Brent Delta (Milton and Ewen, in prep).
Reservoir effectiveness risk Reservoir effectiveness maps illustrate the risk that the reservoir sandstone has effective porosity and permeability. The maps have been derived from a regional study of reservoir quality, involving an extensive well database from the UK and Norwegian sectors of the North Sea. The reservoir quality database was analysed statistically to investigate and identify the key controls on reservoir quality. From this the following factors were found to be important for prediction: - depth (maximum depth of burial if area is uplifted); - overpressure; - facies; - grain size; - pore fluid type (petroleum versus water); - thermal/diagenetic history. Porosity and permeability versus depth cross-plots were generated for each of the major reservoir intervals from the Triassic to Tertiary. Examples of cross-plots for the Brent Group are shown in Figs. 6 and 7. In Fig. 6 the data are subdivided into three overpressure categories: < 1500 psi, 1500 to 3500 psi and >3500 psi. For any given depth of burial the highest porosities and permeabilities are found in the highly overpressured wells (corresponding approximately to an increase of two porosity units for every 1000 psi of overpressure at any given depth of burial). Multivariate analysis using both depth and overpressure as variables produces the tighter regression in Fig. 7. The equations that result from this analysis use depth and overpressure as variables to predict poroperm. The reservoir effectiveness map is then based on the Top Brent Group depth map (taking into account Tertiary uplift where necessary) and the predicted overpressure distribution in the Brent Group. The actual depth cut-offs assumed for the Brent Group are given in Table 1. These were calibrated against well test results. Because gas can be produced from lower permeability sandstones than oil, a deeper depth cut-off applies for the moderate to high risk segment. The common risk segment map for effective reservoir is given in Fig. 8. It illustrates a complex distribution of common risk segments. Throughout the Brent Province the maximum depth limit for low-risk effective reservoir (green) is around 3700 m (Fig. 6). Even allowing for the effects of overpressure the axis of the North Viking Graben is mostly high risk (red). In deeper prospects diagenesis is responsible for seriously reducing porosity and permeability. If petroleum enters the trap before or during cementation it may retard porosity and permeability reduction. This results in steep porosity and permeability-depth
Fig. 4.Depositional environment map for the Late Bajocian (includes Upper Ness Fm.).
Fig. 5. Common risk segment map for Brent Gp. reservoir presence.
173
Play fairway analysis and risk mapping Table 1 Estimated depth cut-offs for effective reservoir in the Brent Group Gas
Risk segment
Oil depth range (m)
K range (mD)
depth range (m)
K range (mD)
Low Moderate High
<3750 3750-4500 >4500
K > 50 10 < K < 50 K < 10
<3650 3650-4900 >4900
K > 30 1 < K < 30 K < 1
The depths correspond to present-day depth, maximum depth of burial (if uplifted), or overpressure corrected depth (see Fig. 7).
35
o
i
!
b
l
30
.
.
.
.
.
.
.
.
.
.
\o
i
35
i
I + 0-1500psi .
.
I O 1500-3500psi i-I ~ >3500psi '
.
o
i
30-
\
oC, 25-
~
+ 0-1500psi o 1500-3500psi A >3500psi
X\_ ......
+
i
20
O
o+ ~
O
o. 20-
....
....\
\
\
15-
10
t 2000
1
3000
4000
5000
Depth (mTVDss)
10 1000
........
++
'
o
_
\, 1500 2000 2500 3000 Overpressure corrected DEPTH (mTVDss)
3500
Fig. 6. Average porosity-depth data, differentiated by reservoir overpressure, for the Brent Group.
Fig. 7. Multivariate analysis, using depth and overpressure as variables, of average porosity data for the Brent Gp.
trends within many North Sea fields (e.g. Emery et al., 1993; Gluyas et al., 1993). This observation has been used to lower the reservoir effectiveness risk (from yellow to green or from red to yellow) on large structural closures in Fig. 8 where thermal modelling indicates an early petroleum charge to the trap and the potential for poroperm preservation (e.g. the Brent Group prospect in Block 34/11).
The Brent Group charge map in Fig. 9 is a combination of source presence and charge effectiveness risks. The main source rocks are the Statfjord Fm. coals, the Brent Group coals, the Heather Fm. oilprone source rocks, and the Draupne Fm. oil-prone source rocks. Maturation studies have been carried out by 1D and 2D basin modelling. Potential carrier systems are identified from regional geological work and from pressure data. The Brent Group is commonly charged via downward migration from overlying oil-prone source rocks, and internally by gas generated from coals within the Ness Formation of the Brent Group. The Brent Group is locally stratigraphically separated from the Late Jurassic Draupne Formation source by thick Heather Formation. Here downward migration from the Draupne Formation into the Brent Group is restricted to high relief half grabens. As a result one of the main sources for oil in Brent Group traps is the Lower Heather Formation mudstone which lies stratigraphically closer to the reservoir. Gas sourced from the intra-formational coals
Charge risk Charge risk maps have been derived from semiregional geochemical studies and fault seal studies. These illustrate the regional variation of risk for an effective charge mechanism. The hydrocarbon charge model is based on an extensive database, including source rock analyses, 1D thermal modelling, 2D modelling of two regional lines using TEMISPACK (Duppenbecker and Dodd, 1993), geochemical analysis of hydrocarbon samples, RFT and well test data, and geochemical data from a regional sea bed coring programme.
174
S. Grant, N. Milton and M. Thompson
Play fairway analysis and risk mapping can easily migrate within the Brent Group carrier system. Long distance lateral migration along spill chains linking footwall traps is possible in the west of the North Viking Graben due to the linear nature of the structures and the absence of major crossfaults. It is likely to be more difficult and hence higher risk on the east flank of the North Viking Graben due to compartmentalisation by sealing faults. Charge effectiveness is likely to be high risk in the shallow peripheral areas, where source rocks are locally absent or immature. There are examples of migration shadows in the north, where charge was not able to move down dip to fill traps; examples include areas west of the Don and Thistle Fields where dry wells are located on valid targets. Collectively these comprise the high (red) and moderate (yellow) risk areas on the CRS map for the Brent Group (Fig. 9).
Seal risk Seal risk maps have been derived from the semiregional geological and geophysical analyses in the same way as reservoir presence maps, and show the risk of an effective top seal facies being present. Top seal to Brent Group prospects is provided in the first instance by Callovian to Oxfordian Heather Formation mudstones within genetic sequences J30, J40 and J50 (Fig. 3). Where these are eroded over fault block crests, top seal may be provided by Kimmeridgian to Ryazanian Draupne Formation mudstones, or marine claystones of the Lower Cretaceous Cromer Knoll Group. Depositional environment maps for the Late Jurassic were combined with pressure data to assess the risk of effective top seal. An example of the depositional environment map for the Early Callovian interval is shown in Fig. 10. The map is based on the predicted distribution of low stand deposits. Most of the area is mud-prone resulting in low risk for effective top seal but there are two areas where sandy facies are predicted to be present above the Brent Group reservoir. The larger of the two sand-prone areas is the prograding shelf system on the east flank of the North Viking Graben (Krossfjord Fm.), the other is a smaller shelf system centred around the Emerald Field in UK Block 2/10. The east flank progradational system continued from the Early Callovian (J33) through to Early Kimmeridgian (Fig. 3) resulting in a relatively thick sequence of sandstone and mudstones. As a result Brent Group prospects may not have an effective top seal in this area. This is illustrated in the common risk segment map in Fig. 11. The moderate to high risk area around the Snorre
175 fault block in Fig. 11 reflects the local development of the J70 (Munin Formation) sandstones which do not provide an effective seal to the underlying Brent Group sandstones.
Play fairway summary Common risk segment maps can be combined to produce a composite common risk segment (CCRS) map. This illustrates the regional risk of an effective reservoir being present, below an effective top seal, and with access to an effective source. It provides a regional summary and relative measure of play risk, which allows ranking of acreage before analysing a local dataset. The map is constructed by assuming that the composite risk at any given point reflects the highest risk component of an individual CRS map. Thus the presence of only one high risk (red) component is sufficient to cause the corresponding area in the CCRS map to be red. For an area to be assigned a low risk (green) all the corresponding component risk maps must also be green in that area. All remaining areas are assigned a moderate risk (yellow). A play fairway map is complementary to a CCRS map and provides additional geological data for the play, including the reason for the high risk (red) assignation in the CCRS map, the location of discovered pools, and dry target tests. A play fairway map illustrates high graded and prospective areas of the play (in green and yellow) and provides a geological sense check for the CCRS map (which only illustrates relative risk). Play fairway maps are particularly useful in relatively mature exploration areas where considerable well data are available for incorporation into the overall play fairway analysis. In frontier exploration where there are very few data points, play fairway maps may be less useful and may not provide any more data than can be illustrated on a CCRS map. The play fairway map and the composite common risk segment map for the Brent Group play are shown in Figs. 12 and 13. The large area of the conventional low risk Brent play (green) explains the highly successful nature of this play. There is no play risk in such areas, i.e. play risk can be assigned a value of one (100%). The yellow areas on the play fairway summary map, north of the proven conventional play, illustrates an area where there is play risk, ranging from 0.5 to 0.95. Here the presence of reservoir sandstones is the key risk. The red areas are relatively high risk with a play risk of less than 0.5. This does not necessarily mean that wells should not be drilled in the high risk areas, rather that these areas are downgraded relative to the lower risk parts of the fairway.
Fig. 10. Depositional environment map for the Callovian (Heather Frn.).
Fig. 1 1. Common risk segment map for effective top seal to the Brent Gp
Fig. 12. Composite common risk segment map illustrating relative play risk for the Brent Gp. play.
Fig. 13. Play fairway summary map for the Brent Gp. showing the location of discovered pools, dry target tests and an explanation of the critical risk elements within the fairway.
178
S. Grant, N. Milton and M. Thompson
It is often the case that the green (low risk) areas of a play fairway have seen considerable exploration, and that all traps with low prospect-specific risks have been drilled. Further exploration may involve pursuing low risk traps in areas of moderate or high play risk (yellow or red areas), or pursuing high risk trapping geometries (stratigraphic or hanging wall traps) where play risk is low (green
areas).
Drilling statistics The play fairways in the North Viking Graben have been sampled by a large number of wells in both the UKCS and NOCS. Within each play fairway the drilling success ratios, well failures, and pool size distributions have been analysed statistically. These data are used to calibrate prospect specific risks and prospect sizes within the play fairway. The drilling success ratio (DSR) is defined as: DSR =
No. of Technical Successes Total No. of Target Tests
where a target test is a well drilled to test a specific play and a technical success produced hydrocarbons on test. The results of the well failure analysis for the Brent Group are shown on Fig. 14. About half the failures were due to invalid trapping (no fault seal, or drilled outside closure). This is especially true of hanging wall traps. A number of target tests are interpreted to be located in migration shadows, mostly in synclines. Some of the larger migration shadows are shown on
the regional play fairway summary (Fig. 13). Other failed wells are located on the crests of tilted fault blocks where the Brent Group reservoir interval has been eroded. Fig. 15 illustrates the pool size distribution for the Brent Group play. The vast majority of reserves discovered to date have been found in footwall traps (over 15 billion barrels) with a drilling success ratio of 1 in 2 (140 tests). Hanging wall tests total 34 with a drilling success ratio of 1 in 3, and contain less than 2% of the reserves discovered to date.
Uncertainty mapping The play fairway is defined from the geological model for the maximum possible extent of reservoir. Many areas of the North Viking Graben, however, have a sparse database, with no or few wells and poor seismic quality to constrain the geological model. In these areas interpretations must be extrapolated from areas with greater data density, and there is a variable degree of confidence in the final interpretation. To delineate these areas interpretation confidence maps have been produced. For geological studies, these maps directly reflect data density and the proximity of control points. For seismic maps they reflect the density and quality of the data, the existence of well ties, and uncertainties in the horizon pick. These maps are divided into areas of high, moderate and low interpretation confidence. An example for the Brent play is shown in Fig. 16. Interpretation confidence maps and CRS maps are complementary. A true estimate of risk across the
Fig. 14. Well results pie chart where the Brent Group was the primarytarget. This illustrates the breakdown of critical play elements that brought about well failure.
179
Play fairway analysis and risk mapping
Fig. 15. Discovered pool size distribution for the Brent Gp. play.
play fairway requires perfect knowledge. The confidence in the play risk is a function of interpretation confidence. Low confidence should not however be equated with high risk. The absence of direct evidence may produce low confidence, however, this is a function of ignorance and not true geological or exploration risk. In general, low confidence areas should be equated with moderate, not high, risk.
Summary The methodology for play fairway analysis has been described above, and is summarised in Fig. 17. It involves several key steps: (1) The play fairway is defined by a geological model for the maximum possible extent of reservoir. Confidence in this geologic model varies across the fairway, depending on the amount of seismic and well data. This is illustrated by interpretation confidence maps. (2) The perceived risk across the fairway is subdivided into three main risk elements: (a) the presence and effectiveness of a reservoir; (b) the presence of a source rock and the effectiveness of the carrier system;
(c) the presence of an effective top seal. These are illustrated as separate common risk segments (CRS) maps where each segment is assigned a low, moderate or high relative risk. (3) Individual CRS maps are combined to produce composite common risk segment and play fairway summary maps. These illustrate the overall variation in relative play risk for the fairway. They are used to high grade parts of the fairway and provide a technical foundation for prospect analysis. (4) Well failure analysis is used to understand the critical play elements. For example trap definition and trap integrity are a key concern for the Brent Group play in the North Viking Graben. This is hardly surprising for a mature exploration province where the regional geological model for the play is well constrained. In a frontier area, with little well data, reservoir or source presence, for example, may be the key risk. (5) Prospect specific risks can be calibrated using drilling success ratios for the play.
Acknowledgements B P Norge are acknowledged for granting permission to publish. The large number of colleagues in
S. Grant, N. Milton and M. Thompson
180
Fig. 16. Interpretation confidence map for the geological model of the Brent Gp. play.
B P are thanked for their contributions to the regional work during the 14th Round Application. Grateful
thanks go to Pia Walmsn~ess, Ellen Lindland and Kjell Falnes for drafting the figures.
181
Play fairway analysis and risk mapping
Fig. 17. A summary of the methodology for play fairway analysis.
References Bowen, J.M., 1992. Exploration of the Brent Province. In: A.C. Morton, R.S. Haszeldine, M.R. Giles and S. Brown (Editors), Geology of the Brent Group. Geol. Soc. Spec. Publ., 61: 314. Deegan, C.E. and Scull, B.J., 1977. A proposed standard lithostratigraphic nomenclature for the central and northern North Sea. Institute of Geological Sciences, Rep. 77/25. Norwegian Petroleum Directorate Bull., 1. Duppenbecker, S. and Dodd, T., 1993. Petroleum charge model for Brent accumulations - - application of integrated basin modelling. E.A.EG. 5th Conference. Extended Abstr. F028. Emery, D., Smalley, EC. and Oxtoby, N.H., 1993. Synchronous oil migration and cementation in sandstone reservoirs demonstrated by quantitative description of diagenesis. Philos. Trans. R. Soc., A 344:115-125. Gluyas, J., Robinson, A.G., Emery, D., Grant, S.M. and Oxtoby, N.H., 1993. The link between petroleum emplacement and sandstone cementation. In: J.R. Parker (Editor), Petroleum Geology of
S. GRANT N. MILTON M. THOMPSON
Northwest Europe, Proceedings of the 4th Conference. Geological Society London, pp. 1395-1402. Milton, N. and Ewen, D. (in prep). A regional sequence stratigraphy for the Brent Group. Mitchener, B.C., Lawrence, D.A., Partington, M.A., Bowman, M.B.J. and Gluyas, J., 1992. Brent Group: sequence stratigraphy and regional implications. In: A.C. Morton, R.S. Haszeldine, M.R. Giles and S. Brown (Editors), Geology of the Brent Group. Geol. Soc. Spec. Publ., 61: 45-80. Partington, M.A., Mitchener, B.C., Milton, N.J. and Fraser, A.J., 1993. Genetic sequence stratigraphy for the North Sea Late Jurassic and Early Cretaceous: distribution and prediction of Kimmeridgian-late Ryazanian reservoirs in the North Sea and adjacent areas. In: J.R. Parker (Editor), Petroleum Geology of Northwest Europe, Proceedings of the 4th Conference. Geological Society London, pp. 347-370. Vollset, J. and Dor6, A.G., 1984. A revised Triassic and Jurassic lithostratigraphic nomenclature for the Norwegian North Sea. Norwegian Petroleum Directorate Bull., 3.
BP Norge UA, P.O. Box 197, Forusbeen 35, 4033 Forus, Norway BP Norge UA, P.O. Box 197, Forusbeen 35, 4033 Forus, Norway BP Norge UA, P.O. Box 197, Forusbeen 35, 4033 Forus, Norway
This Page Intentionally Left Blank
183
A model-based approach to evaluation of exploration opportunities B.A. Duff and D. Hall
Effective exploration comprises a cycle determined by the synthesis of all available technical data into a unifying basin model, the testing of model predictions through the acquisition of further exploration information, and a "post-mortem" evaluation phase during which the model and its associated process-response systems are refined to accord with the new information. Evaluation of exploration opportunities, and resource assessment, is approached through a process-based understanding of exploration plays. Risked reserves estimation for plays and prospects is facilitated by the recognition of process and chance domains within the play, which are based on the specific predictions of the basin model. This model-based exploration cycle provides the most efficient and cost-effective framework for reducing risk to prospectivity prior to drilling wells or taking up acreage. Accordingly, it guarantees that these business decisions are optimized.
Introduction More than ever before, explorationists are using sophisticated analytical and interpretation techniques to minimize risk to prospectivity and maximize reward of their exploration portfolios in a cost-effective way. However, whilst this aim of exploration and the associated tools are well known, the strategy under which these tools are cost-effectively selected, applied and interpreted to add value to exploration assets has been far less apparent (Fig. 1a). The logical strategy for achieving this is the application of conceptual models of the processes impacting on hydrocarbon prospectivity. Such models of hydrocarbon trap-forming processes within individual basins have often been used to facilitate and enhance the interpretation phase of the exploration cycle. At the other extreme of scale, "global" causal models of basin processes, calibrated by world-wide data-sets (Nederlof, 1982; Nijhuis and Baak, 1990), have been used to estimate prospect reserves within particular basins. A model-based understanding of relevant basin processes and their responses however should be pivotal to all phases of cost-effective exploration. Furthermore, through meaningful play analysis this should be directed at the intra-basin scale, and the resulting model predictions should underpin prospect evaluation and resource assessment (Fig. l b). In this paper we propose a practical, model-based method for improving the consistency with which
exploration risks and the uncertainties of volumetric variables are estimated for input to probabilistic reserves evaluation. This is founded on an awareness of the value of exploration information and how this is unified predictively within the conceptual basin model. It overcomes the subjectivity of individual intuition and of the "Delphic" approach.
Common shortcomings in exploration The following failures to fully utilize exploration knowledge of a basin are very familiar: (1) Separate, sequential application of selected analytical techniques at the expense of other, possibly more appropriate, methods. In extreme cases, single analytical and interpretive tools are sometimes applied as if they alone are an exploration panacea rather than useful adjuncts to a broader data-set. (2) The lack of a conceptual framework to guide what data to acquire, process or interpret, when to use particular techniques, and what coverage, sample density and parameters to use. (3) Inadequate post-mortem analysis of the results of successful or unsuccessful drilling. In our view, the most obvious solution to these problems has frequently been overlooked. For this we need to consider the properties of the conceptual model and how these may be of practical value in reserves estimation and exploration risk reduction. First, however, we consider the nature of exploration risk and uncertainty, and the value of information.
Quantification and Prediction of Petroleum Resources edited by A.G. Dor6 and R. Sinding-Larsen. NPF Special Publication 6, pp. 183-198, Elsevier, Amsterdam. 9 Norwegian Petroleum Society (NPF), 1996.
184
B.A. Duff and D. Hall
Exploration in a Conceptual Vacuum: Costly, directionless
What next 19.
Za
vlimii:i:::....I? TM
1 '"'~ ~
I"
I :'e.~timatiIonI9 ......
The Exploration Cycle: Cost-effective, focused exploration Update model
........
"Tools"
o ...... o.
1~ ..
9 .
Appropriate tool?
........... SIS ] ~
Organization of in, . . . . ti~ ~
I --
~
~
Eci,vl
/"
I Probabilistic ..... S............ timation ON ~ I~~. I ..........
Fig. 1. (a) Directionless exploration without a strategy for using results of data acquisition to achieve the exploration objective is costly and inefficient. (b) The basin model is pivotal at all stages of efficient, cost-effective exploration: it "drives" the "exploration cycle".
Risk and uncertainty in exploration It is now generally accepted that the volumetric variables and their product, recoverable reserves, must be described as statistical distributions (e.g. White and Gehman, 1979; Nederlof, 1982; Nijhuis and Baak, 1990): net Recoverable reserves = Bulk rock volume x gross x porosity x hydrocarbon saturation x
1
x recovery factor (1) volume factor Conversely because uncertainty is always present to a greater or lesser degree it is meaningless to d e scribe a prospect s reserves by a single deterministic value Monte Carlo simulation is the multivariate statis tical method typically used to honour this impreci sion surrounding the volumetric variables and their product This builds up an accurate estimate of the probability distribution for recoverable reserves by
repeated random sampling of the individual distributions considered by the explorationist to best describe the uncertainty associated with each of the volumetric variables. In practice, the exploration risk of failing to prove reserves is often distinguished from the uncertainty expressed by the variance or spread of values associated with the reserves distribution for a given prospect (Fig. 2). The most useful way to represent exploration risk and uncertainty is the expectation curve (e.g. Nederlof, 1982; Nijhuis and Baak, 1990). This portrays the cumulative probability (P1 in Fig. 2) that a certain reserves value (X1, Fig. 2), or a value greater than this, will be realized. The intercept of the curve indicates the chance of success, and the slope is a measure of the uncertainty, or conversely the precision, associated with reserves estimation. Probabilistic modelling of reserves has certainly helped improve resource estimates. However, too often it seems to be considered as an end in itself rather than a useful statistical tool for analyzing the uncertainty of knowledge of the process-response systems impacting on prospectivity.
Effective exploitation of exploration knowledge The value of information We define effective exploration as the evaluation of a portfolio of exploration opportunities in a manner which minimizes risk to prospectivity, and maximizes potential reward in the most cost-effective and efficient way. Exploration opportunities may comprise both new ventures acreage acquisition and options to drill a wildcat exploration well on a prospect, or appraisal drilling of an existing discovery. The evaluation strategy must have sufficient generality to address each of these categories as they are all likely to be competing for the same, finite budget. In other words, effective exploration is a process of risk reduction (Fig. 3). Exploration typically commences during a new ventures stage with only sparse data available, and the associated uncertainty is expressed probabilistically as a rather broad, weakly-sloping expectation curve. The effect of ongoing exploration is to progressively decrease the uncertainty associated with each of the volumetric variables, which together determine the expectation curve for reserves (Fig. 3). In this way, the steepness and therefore precision of the expectation curve estimated for reserves is successively increased by the incremental addition of exploration knowledge. The acquisition of exploration information does not proceed indefinitely however. Eventually a stage of rapidly diminishing returns in terms of further
185
A model-based approach to evaluation of exploration opportunities 1.0-
o~
0.8
~
RISK OF DRY HOLE (I-POD) = 0.6
0.6
==
/ ~
P RO BABILITY OFADN~~;:EE1sY~IPO2)E;0"4
0.4
ES DISTRIBUTION
Pl 0.2-
0.0
Xl
Reserves
Fig. 2. The principle of the expectation curve and its application in hydrocarbon reserves estimation. P1 is the cumulative probability that reserves of X1 or greater are realized. Expectation curves can also be used to depict distributions associated with volumetric variables such as porosity (modified after Nijhuis and Baak, 1990).
1.0 .~
IER--~ i
tlL ~
-~ ........ g ,m
Prosnec~t I z Field appraised mature'for ~ ~ (development plan) drilling ~ ~[~,~ ~
~1 II
Actual reserves value
.Q 0,5 .Q
2e~ ._>
E 0
509 100 Reserves (mmbo)
Fig. 3. Schematic expectation curves portraying the value added in information, or conversely risk reduction, as a lead identified early in the exploration cycle evolves toward the curve associated with the Irreducible Exploration Risk (IER), which defines a prospect mature for drilling. The further decrease in risk following a discovery through the appraisal and development phases is also shown.
risk reduction per unit of exploration information is reached. The reserves expectation curve during this stage tends toward an intercept on the cumulative probability axis which we here term the irreducible exploration risk (IER; Fig. 3). The IER represents notionally the lowest possible estimated risk to prospectivity achievable without drilling a well. The value of exploration information is therefore clear. There is value in the acquisition, processing and interpretation of geological, geophysical and geochemical data prior to drilling only as long as this contributes significantly to a reduction in the estimated risk to prospectivity toward the IER. Here, data should be distinguished from knowledge. Additional
data only imply improved knowledge if they can be used to refine the basin model. From the inception of exploration in a basin there are many possible pathways toward the IER, depending on the specific exploration programme of information acquisition pursued. When exploration is conducted arbitrarily (Fig. l a), these will be grossly inefficient in terms of identifying the best sequence of activities comprising the programme, and judging when the optimum amount of data has been acquired. As a result, they will be unnecessarily costly, and invariably characterized by all or some of the shortcomings indicated previously. Our main thesis is that a model-based foundation for probabilistic reserves estimation maximizes the chance that the most efficient and cost-effective exploration programme is identified and followed. It may not be possible to quantify the IER as a unique criterion. However, by application of the basin model and judicious incorporation of its predictions, it should still be possible to judge the point of diminishing incremental value added in the description of risk attached to an exploration opportunity. Gauging this point can help indicate when exploration activities other than drilling should be discontinued to contain costs. We recommend that the term prospect be restricted to features that have, in this way, been matured for drilling, and that all other less mature features be referred to as leads. This view of exploration therefore has the merits of highlighting the logical relationships between risk, uncertainty and reward, and the acquisition costs of exploration information. We now turn to the conceptual model and its vital and very practical role in
186
controlling the acquisition of this information, and its organization for optimum reserves estimation.
The conceptual model The two fundamental properties of a conceptual model are also its great strengths: its predictive power and its susceptibility to continual testing, refining or even rejecting. Indeed the criterion of falsifiability is a fundamental pre-requisite for any hypothesis which aspires to being truly scientific (Popper, 1959). As more information is acquired, the number of working models is reduced as those which are denied by this new information are discarded. Eventually, one model remains which can be refined by further data acquisition to a highly-detailed description of particular natural processes, with very specific predictions. This model is sometimes referred to as the processresponse model. Basin analysis can usefully be regarded as the application of this methodology in the geological context of a sedimentary basin. But in this context we must recognize that typically many processes have, during the life of the basin, contributed to an even greater number of responses (e.g. Krumbein and Sloss, 1963). The term basin model is used in this paper to refer to the totality of individual processresponse systems that have been active in the basin. We are concerned in particular with those which together govern hydrocarbon prospectivity (Fig. 10). These comprise models of depositional, diagenetic, structural, halokinetic, and hydrocarbon expulsion and migration processes, and their specific, measurable responses, the knowledge of which should constitute a prime aim of an exploration programme. In the exploration context, the following properties of the basin model follow from the nature of conceptual models in general. (1) It is predictive: concrete responses are predicted for all of the various process-response systems of which it is comprised. (2) As a result, it is always testable and potentially falsifiable. The basin model is therefore the only logical context for planning acquisition, processing and interpretation of geological, geophysical and geochemical data. (3) The unified basin model is more powerful predictively than the sum of its individual constituent data elements. To take a simple example, porosity preservation may be linked to hydrocarbon charge retention and because of this the reservoir model cannot be fully understood without also understanding the predictions of the charge model. (4) Application of the basin model optimally explains spatial and geological temporal variance
B.A. Duff and D. Hall
throughout a multivariate data-set, dimensions often ignored to the cost of effective prospect evaluation. (5) Application of the basin model approach is an iterative process, comprising phases of data acquisition, processing and interpretation, followed by model modification and refinement. From these properties, it follows that the basin model provides the soundest basis for risking overall prospectivity and estimating parameter distributions for prospects within a play.
The evaluation of exploration opportunities
The exploration cycle The post-mortem review and the basin analysis phase are inextricably linked through "feed-back" to the basin model, and this makes exploration, when it is pursued effectively, cyclic (Figs. l b and 4). If this exploration cycle is the factory processing and adding value to an exploration portfolio, then the basin model is its engine, ensuring that all available exploration data are optimally exploited in order to maximize the value added. In short, the basin model is pivotal in cost-effective exploration (Fig. lb). We now suggest specific ways in which this general approach to exploration can be applied to the challenges of effective play analysis, and the associated evaluation of prospects and assessment of resource potential.
Model-based play analysis The definition of play There appears to be no clear consensus as to what constitutes an exploration play, and this has unfortunately made it a blunt tool in articulating what explorationists believe concerning the prospective potential of a basin. Consistent with our view of the pivotal role of the basin model, we define the exploration play as follows: one or more closures of similar structural, depositional or hydrodynamic style, which result from a specific set of tectonic, depositional/ diagenetic or halokinetic processes within a sedimentary basin, and which, with suitable reservoir and sealing lithologies, and hydrocarbon charge, may form prospective hydrocarbon traps. This definition identifies the major geological processes and their unique responses within the basin as determining in turn the uniqueness of plays. In particular, it emphasises closure as the key distinguishing process-related element of a play. The term closure (e.g. Fig. 6) refers only to the morphology of the sealing surfaces enveloping the single gross rock volume. Closures of structural or
A model-based approach to evaluation of exploration opportunities
187
Fig. 5. Some examples of play families, and their constituent plays (P1, P2 . . . . . Pn), illustrating the definition of play as resulting from a unique closure-forming process. Play families are determined by first-order processes such as halokinesis (a), extensional tectonism (b), depositional and diagenetic processes (c and d), compressional tectonism (e) and basin compaction (f). These generate one or more particular closure styles which represent the individual plays.
B.A. Duff and D. Hall
188 .r
3D SEISMIC
p... ~
2D SEISMIC
PLAY 1 (P1)
PLAY 2 (P2)
SALT ANTICLINES
EXTENSIONAL FAULT BLOCKS
PLAY 3 (P3)
INVERSION
i
P2 0.8 - 0.9
Fig. 6. A schematic example of plays (P1, P2 and P3) distinguished by the closure-forming process and associated style. In the case of closure, process domains (a) correspond to the entire extent of the play. The corresponding closure chance domains (b) result from the application of the risk tranches method.
depositional/diagenetic origin within a given play only become traps when the essential ingredients for prospectivity, reservoir, seal and hydrocarbon charge, are added to the closure. The responses of specific first-order tectonic, depositional, diagenetic and halokinetic basin processes are "play families" These comprise one or more plays which, whilst possessing quite different styles of closure, have all originated from the same, first-order processes (e.g. Fig. 5). Play-based procedures for estimating the volumetric parameter distributions and the risks of absence of reservoir, seal and hydrocarbon charge are now considered.
within the spatial extent of any play. However, this is typically not a single, uniform variation. Rather, there are specific domains within which a process generates rather similar responses, and between which these responses show much greater variance. This arises because natural processes within the upper crust are themselves discretized into domains. Accordingly we define reservoir, seal and charge process domains as follows: Reservoir process domains are specific areas within a play within which the same or similar processes of deposition and diagenesis operated, and between which these processes varied significantly. Seal process domains are specific areas within a play within which the same or similar processes of deposition, burial, diagenesis, deformation, and hydrocarbon hydraulics operated, and between which these processes varied significantly. Charge process domains are specific areas within a play within which the same or similar processes of hydrocarbon expulsion, migration and entrapment occurred, and between which these processes varied significantly. Examples of these are given in Figs. 7, 8 and 9. Processes such as deformation and diagenesis typically overprint and may modify the primary reservoir and seal domains inherited from the depositional environment. For example, diagenesis may enhance or
a)
9
/
o~4e2x.~j e-/t/~TIc "T~'~=" ~
( "t')
oW /
I
/
SANDS
b)
Play process domains We wish to estimate the volumetric parameter distributions for porosity, net-to-gross (n/g), gross rock volume (grv) and hydrocarbon saturation (Sh), and to assign estimates to the chances of reservoir, seal, charge and closure. We begin by recognizing that natural processes such as deposition, diagenesis, and deformation interact in the depositional environment or with the older basin fill sequence to produce observable responses. Furthermore, these responses can vary significantly
0.6 - 0.7
0.6
ill
0.8
Fig. 7. A schematic example of reservoir process domains (a) and corresponding chance domains (b).
A model-based approach to evaluation of exploration opportunities
Fig. 8. A schematic example of seal process domains (a) and corresponding chance domains (b).
degrade reservoir, and deformation may compromise sealing integrity. However, basin analysis experience suggests that certain depositional processes and their response lithofacies "pre-condition" subsequent processes in the basin's history. It is therefore not unusual within a play to observe a high degree of spatial correlation (or anti-correlation) between the responses of a number of processes. A common example is the restriction of specific types of diagenesis in both carbonate and siliclastic settings to specific lithofacies depositional belts; the textural and mineralogical maturities of these belts effectively control the type of cementation and framework alteration that occurs. This impacts on both sealing and reservoir quality. Accordingly, it is common for reservoir and seal domains to follow the primary depositional lithofacies belts within any play, and this tendency should be fully exploited in defining a play's reservoir and seal domains, and simplifying play analysis. The reservoir, seal and charge process domains of a play are derived from the spatially-discretized variation in lithologies, structural style and hydrocarbon type predicted by the depositional, diagenetic, struc-
189
tural, and expulsion and migration process-response systems comprising the basin model (Fig. 10). The reservoir and charge domains provide a process-based way of best-estimating the volumetric parameter distributions (e.g. P05, P50, P95 percentiles) for porosity, n/g, and Sh for input to probabilistic reserves modelling. Furthermore, these domains also facilitate estimation of the hydrocarbon recovery factor, oil shrinkage and gas expansion factor, because these also generally relate to both reservoir and hydrocarbon phase properties. The definition of the spatial extent of process domains is only as good as the associated processresponse model, and therefore the amount of information on which this is based. For example, a geologist's reservoir lithofacies domains may, in detail, be in error in areas of the play for which data are few. In the next section, a procedure which consistently addresses the irregularity in data distribution within and between process domains is described.
Model-based estimation of chance of reservoir, seal, charge and closure: chance domains The key risks to hydrocarbon prospectivity within a play are absence of reservoir, absence of one or more seals to fully seal this reservoir throughout the
190
B.A. Duff and D. Hall PROCESS (EXAMPLES)
PROCESS RESPONSE
CHANCE DOMAINS
----1
DISTINCT RESERVOIR FAIRWAYS
DEPOSITIONAL I~ DIAGENETIC
RESERVOIR
LITHOFACIES BELTS POROSITY DOMAINS
[, SECONDARY DISSOLUTION I j r '
A LU -r 0
,
..J
U.I
o=i
Z u) ,<
Z < E p.
,,
DISTINCT STRUCTURAL STYLES
TECTONIC
HALOKINETIC
EG. IMBRICATED CLOSURES, TURTLE-BACK AN~CLINE
E
OISTINCT AREAS WITH VARYING CHARGE EFFECTIVENESS & POST-MIGRATION MODIFICATION OF HYDROCARBON . . .Q U. A L I T Y
a
SEAL
n
m HYDRODYNAMIC
FLUSHING
*1
HEAT FLOW
I'~,
1t
tu
_J m
CHARGE
___1
I
CHARGE (EXPULSION + MIGRATION)
CLOSURE Fig. 10. Relationship between play definition, the model process-response systems and the chance domains which they determine within a play.
extent of the closure, and absence of hydrocarbon charge of the sealed closure. A further risk associated with individual prospects in the play is absence of closure. It is generally more convenient to use the probabilistic complements of these risks: the chance of reservoir (Pr), the chance of the reservoir being sealed (Ps), the chance of hydrocarbon charge (Pch), and the chance of closure (Pcl). The composite risk to prospectivity for any closure within the play, the probability of discovery (POD), is then given as: POD = Pr x Ps x Pch x Pcl
1. Stratigraphic trap I
I
x \ \ \ \ ~ ~ ' ~ ~ ~ ~..~..
DEFINITON OF PINCH OUT = CLOSURE CHANCE FACTOR (Not seal chance factor)
I
I
"
~ ! INTEGRITY= SEAL CHANCE FACTOR ]
2. Fault trap
(2)
For both stratigraphic and structural closures the closure probability (does a closure exist?) should not be confused with the chance of sealing (is this closure completely sealed?). Fig. 11 illustrates some of the differences between the two chance factors. If at the inception of exploration in a new basin no information exists to describe that basin, there would, of course, be no basis for formulating an initial conceptual model, and therefore in turn no way to define the process domains of a play. Indeed, from our definition of play it follows that the play itself would not be recognized. This may be visualized as a "flip of a coin" type probability of eventual success of one or more initially unrecognized plays, with an equal 0.5 chance that prospectivity could be affirmed or denied by the acquisition of information and eventual drilling. This is strictly notional, however, because in practice there would generally be some information al-
INTEGRITY=SEAL CHANCE FACTOR
SEAL CHANCE' ~ FACTOR ~ ....... I
I
~ ~
v I
t-I
]
.
c...,..o~=osu.~ I]--t/ GEOMETRY = CLOSURE CHANCE FACTOR
]
CONTINUITYOF TOP SEAL = SEAL CHANCE FACTOR re chance factor)
I~ \ )
CERTAINTY OF FAULT PRESENCE = CLOSURE CHANCE FACTOR
Fig. 11. Comparison of closure and seal chance factors (schematic).
ready available, such as gravity, magnetic and seismic surveys of the basin, or previous drilling results, which would permit an initial basin model to be formulated. Even if such data were lacking, known basin analogues are frequently available. Consequently, realistic estimates of the chance of success of a play
I
A model-based approach to evaluation of exploration opportunities
191
Fig. 12. The risk tranches method for standardizing the consistent estimation of probability of presence of closure, reservoir, seal and charge.
at the beginning of an exploration cycle are usually lower than 0.5. In order to assign risk consistently, we introduce the concept of risk tranches. These are logically and symmetrically arranged around the notional "no knowledge" case of 0.5 risk as follows (Fig. 12): 0.4-0.6: Little or no available data on which to base a model; play may with equal likelihood eventually be vindicated or denied. 0.6-0.8: Sufficient data on which to base a model which predicts that play may possibly be affirmed by subsequent data acquisition, including drilling. 0.2-0.4: Sufficient data on which to base a model which predicts that play may possibly be denied by subsequent data acquisition, including drilling. 0.8-1.0: Further data which strengthen the model predictions to: "play will probably be affirmed by drilling". 0.0-0.2: Further data which strengthen the model predictions to: "play will probably be denied by drilling". Specific technical criteria which increase in severity and number are invoked in order for the assignment of risk to reservoir, seal, closure and charge to evolve with the acquisition of knowledge through the
risk tranches from "do not know" (0.5), through "possibly" (0.6 to 0.8), to "probably" (0.8 to 1.0). Conversely, the direction of denial passes successively through "possibly not" (0.2 to 0.4) and "probably not" (0 to 0.2) with increasing knowledge. These conditions (Fig. 13; Table 1) test the predicted responses of the basin model, and impose the discipline of a set of specific technical tests, which can be made more or less severe depending on corporate policy. The appropriate risk tranche is that for which all tests or conditions (e.g. Table 1) are passed (a "logical sieves" approach; Fig. 14). For example, a top-sealing unit which passes all tests in the tranche 1.0-0.8 (Table 1) would qualify for a probability estimate (Ps) in this range. However, if sealing is proven in only one offset well, and seismic quality is inadequate to demonstrate lateral continuity of this sealing unit over the closure, we must drop down through the tranches until all (weaker) conditions are met; in this case, within the 0.8-0.6 tranche (Fig. 14). An identical approach applies to estimating the chance of presence of reservoir, charge and closure (Table 1). This scheme therefore minimizes the subjectivity and maximizes the consistency of risk assignment, because the logical symmetry and tranches of risk are intuitive to the human mind in terms of a commonly-
B.A. D u f f and D. Hall
192 Table 1 Guidelines for the estimation of geological chance scores using the risk tranches method Closure Score: 1.0-0.8
0.8-0.6
0.6-0.4
0.4-0.0
Presence of minimum structural or stratigraphic closure is clearly indicated by seismic coverage. Available well and seismic data allow accurate depth conversion. Closure should be identified from the top reservoir pick, which should be clearly registered on seismic. Stratigraphic closures should be further defined by a reliable base reservoir pick, and wedge-out geometry should be clearly resolved on seismic. Presence of minimum structural or stratigraphic closure is probable based on seismic coverage and depth conversion. Closure should be identified from the top or near-top reservoir pick. For stratigraphic traps wedge-out geometry should be clearly apparent on at least some seismic lines. Based on seismic coverage and depth conversion there is a near equal chance of minimum structural or stratigraphic closure being present or absent. This may be because the mapped seismic horizon is significantly above the target as a result of limited seismic quality. You may wish to consider further seismic acquisition, seismic reprocessing and a review of the depth conversion. Closure is inadequately defined by seismic data. Additional seismic acquisition and/or seismic reprocessing is required.
Seal Score: 1.0-0.8
0.8-0.6
0.6-0.4 0.4-0.2 0.2-0.0
(Note: Top seal and base seal if relevant.) Presence of seal is clearly calibrated by wells and seismic. The integrity of seal is confirmed by seismic facies analysis; there is no evidence of seal lithofacies deterioration between wells and prospect. Predicted reservoir pressure is not sufficient to break seal (consider capillary entry pressure of seal lithology). There is no evidence of structural breaching such as faults, jointing or fracture cleavage. Presence of seal is proven in at least one well, and its presence over the prospect or lead is confirmed by seismic. It may not be possible to predict seal from seismic facies analysis. Available reservoir pressure data are insufficient to deny seal integrity. At worst there is only a small risk of structural breaching. Presence of seal is neither confirmed nor denied by well or seismic data. In rank wildcat areas the chance of seal presence will often be the same as risk of seal absence. Wells and seismic data indicate possible absence of a seal. Reservoir pressure data suggest some risk of seal failure. Structural breaching of the seal is also possible. Well, seismic, or reservoir pressure data indicate a high risk of seal failure. You should seriously consider whether to recommend this prospect or lead.
Reservoir Score: 1.0-0.8
0.8-0.6
0.6-0.4 0.4-0.2
(Note: Presence defined by porosity > cut-off.) Presence of reservoir is clearly indicated by wells and seismic. The reliability of reservoir presence is confirmed by seismic facies analysis (ie. there is no evidence of reservoir deterioration between wells and prospect). Reservoir presence may also be supported by seismic attributes. Both wells and seismic data yield a consistent depositional and diagenetic model. Presence of reservoir is proven in at least one well, and its presence over the prospect or lead is confirmed by seismic (facies and/or attributes). It may not be possible to predict from seismic facies analysis. However, a positive indication should come from the depositional and diagenetic model. Presence of reservoir is neither confirmed nor denied by well or seismic data and the associated depositional and diagenetic model. In rank wildcat areas the chance of reservoir presence will often be the same as risk of reservoir absence. Wells and aseismic data indicate possible absence of a reservoir. Seismic facies analysis and the depositional and diagenetic model indicate the possibility of reservoir absence.
Charge (the chance that a particular hydrocarbon phase is present). Scoring: The score range used to estimate chance of charge is determined by the most pessimistic of the charge parameters (ie. source rock, expulsion, migration and timing). For example, if source rock, expulsion and migration qualify for the range 0.8-0.6, but timing only qualifies for the range 0.6-0.4, then the overall chance of charge must be scored in the range 0.6-0.4. Score: 1.0-0.8
Source rock: presence of source rock is clearly implied by wells and seismic. Source rock quality (predicted or directly measured) should be of primary grade (potential yield > 7 kg/ton).
Expulsion: hydrocarbon expulsion in the source rock kitchen is clearly indicated (eg. borehole shows, hydrocarbon seeps, and possibly seismic direct indicators). The source rock kitchen is clearly defined and of sufficient volume to source the prospect or lead, and fields and discoveries within the same drainage area, and untested closures downdip along the same migration pathway. Migration: a viable migration pathway is clearly supported by the distribution of surrounding hydrocarbon shows, and possibly seismic DHIs. The geometry and effectiveness of the migration pathway should be clearly apparent on seismic section. Timing: prospect/lead closure should clearly pre-date the main phases of hydrocarbon expulsion. 0.8-0.6
Source rock: presence of source rock is probable based on well and seismic data and basin model. Source rock quality (predicted or directly measured) should be of primary grade (potential yield > 7 kg/ton). Slightly leaner source rocks may be considered if it can be demonstrated that the migration pathway is highly efficient. Expulsion: hydrocarbon expulsion in the source rock kitchen is probable based, for example, on borehole shows, hydrocarbon seeps, and possibly seismic direct indicators (DHIs). The source rock kitchen is probably of sufficient volume to source the prospect or lead, and also the fields and discoveries within the same drainage area, plus untested closures downdip along the same migration pathway. Migration: a viable migration pathway is probable as implied by the distribution of surrounding hydrocarbon shows, and possibly seismic DHIs. A probable migration pathway should be apparent on seismic section. Timing: it should be at least probable that the prospect or lead closure pre-dates the main phases of hydrocarbon expulsion.
A model-based approach to evaluation of exploration opportunities
193
Table 1 (continued) 0.6-0.4
Source rock: source rock may or may not be present based on well and seismic data and basin model. There may be no data to support or deny the presence of primary grade source rock.
Expulsion: hydrocarbon expulsion in the source rock kitchen is supported by maturation modelling, although this may be uncalibrated. The basin model and seismic interpretation should give some indication of kitchen volumes. The source rock kitchen may or may not be of sufficient volume to source the prospect or lead, and also the fields and discoveries within the same drainage area, plus untested closures downdip along the same migration pathway. Migration: a viable migration pathway may or may not exist. Timing: the prospect or lead closure may or may not pre-date the main phases of hydrocarbon expulsion. 0.4-0.2
Sourcerock: well and seismic data and the basin model indicate that primary grade source rocks may be absent. Expulsion: maturation modelling indicates the possibility that kitchen source rock volume is insufficient to source the prospect or lead. Migration: the distribution (or absence) of hydrocarbon shows and possible seismic indicators, or results of seismic structural mapping indicate the possibility that the prospect or lead does not lie on a viable migration pathway. Timing: seismic interpretation and basin modelling indicate the possibility that the prospect or lead closure post-dates the main phases of hydrocarbon expulsion.
0.2-0.0
Sourcerock: well and seismic data and basin model indicate that primary grade source rocks are probably absent. Expulsion: maturation modelling indicates the probability that kitchen source rock volume is insufficient to source the prospect or lead.
Migration: the distribution (or absence) of hydrocarbon shows and possible seismic indicators, or results of seismic structural mapping indicate the probability that the prospect or lead does not lie on a viable migration pathway. Timing: seismic interpretation and basin modelling indicate the probability that the prospect or lead closure post-dates the main phases of hydrocarbons expulsion.
DEPTH ~ CONVERSION
CLOSURE ~ PRESENCEOF MINIMUM CLOSURE
(MAPPED HORIZON
Seismic Calibration
SEISMIC COVERAGE )
Reservoir Pressure No Structural Breaching
SEISMIC) RESOLUTION
Confirmed > 1 Well
Limited Seismic Calibration RESERVOIR ~.~ PRESSURE J ~ // l
SEAL --~ ~ CAPILLARY }~"
~ENTRYPRESSUREJ
,
Reservoir Pressure
, J'~ CALIBRATIONBY~ ~_.~LLS J
SEAL INTEGRITYOFSEAL OVERcLoSuREMINIMUM I - - ( LITHOFACIESVARIATION~'~ I I I FROMWELLSTOPROSPECTI " t"" ~ ~ "~"~'~ ' ~ (Seismic calibration) J
CRO SS-FAULT) SEALS
(
No Structural Breaching Confirmed by I Well
STRUCTURAL BREACHING )
All ,,,
Factors Are Ambiguous
SEISMIC ~ . ~ AI-rRIBUTES ~
RESERVOIR
Jf I~~.
CALIBRATIONBY wELLs J
CHANCETHATA RESERVOIRLITHOLOGY ISPRESENT J // LITHOFACIESVARIATION~'~ I I FROMWELLSTOPROSPECTI ~.. (Seismic calibration) J
~ DEPOSITIONAL/ DIAGENETIC MODEL
One or More Factors Suggest Failure
Ic
MIGRATION
~
~ f ~. .(. ~ [ SOURCEROCK 9 9 ~[ PRESENCEAND ' ' I ~k~ QUALITY I-4 CHARGE CHANCETHATA I~ ~'~ ( TIM PARTICULARHYDROCARBON EXPULSION -'~ JPre-migration structuration; J ] PHASEIS PRESENT I |Direct indications (eg.shows);I spillage j , I | maturation modelling; | ~,~ k adequate charge volumes J Distribution of shows; arrier bed effectiveness; structural complexity
C )
I / I~) I / I
~
)
p- EXAMPLESOFINPUTCONSIDERATIONS
Fig. 13. Specific criteria which guide determination of the risk tranches score (Fig. 12; Table 1).
Fig. 14. The "logical sieves" analogy for using the risk tranches method to estimate the chance of closure, reservoir, seal or charge. This simplified example applies to chance of seal.
shared perception of degree of risk (or conversely no risk), and because it is possible to define qualification for these discrete tranches in terms of specific technical tests. The precise attribution of specific tests between the different tranches will undoubtedly vary with dif-
194
B.A. Duff and D. Hall
ferent corporate perspectives, and the example shown in Table 1 is just one possible application which seeks to be as comprehensive as possible in systematically applying all possible constraints on prospectivity. However, the explorationist following the general methodology, irrespective of how in detail he defines the cut-offs between different tranches, can be sure of consistently risking all of the prospects in his play. Furthermore, because this risk assignment is rooted in the specific predictions of the basin model, all available data are optimally exploited. The final step in the assignment of chance probabilities to the actual prospects of a play is to apply the risk tranche scores to the reservoir, seal and charge process domains to create reservoir, seal and charge chance domain maps for the play (e.g. Figs. 7-10). A closure chance domain map is also made by applying the risk tranche score for closure to all prospects within the play (e.g. Fig. 6). The chance domains are therefore sub-areas of the process domains (and the play area in the case of closure chance domains) within which the range of probabilities of presence of reservoir, seal, charge or closure is the same. Although chance scores for each domain may be expressed by a range of values, for individual prospect evaluation it is necessary to assign discrete chance estimates for reservoir (Pr), seal (Ps), charge (Pch) and closure (Pcl) in order to determine a discrete probability of discovery (POD). This is achieved by considering prospect scale aspects of the data density and domain-
related processes. The range of values for each domain provide a spatially consistent guide for the specific values assigned in this way to individual prospects. Consistent estimates of the POD for any prospect within a given play are then given by the product of individual chance probabilities [Eq. (2)]. Accordingly, the composite chance domains of a play are those sub-areas of the play characterized by a constant POD value. The organization of prospect risking through the chance domains and the derivative POD ensures a consistency of risking of all prospects within any play. This in turn provides a logical basis, founded on understanding of geological processes, for prospect ranking across one or more plays. With increasing data acquisition and the associated progression through the risk tranches, there is an increase in precision of the corresponding expectation curves specific to each type of process domain. The maps of these domains and the composite play map are also updated correspondingly. This provides the dynamic basis for the exploration cycle, founded on the basin model. In moving through the risk tranches as the basin model is progressively refined, the aim is to minimize the composite risk to prospectivity before taking the decision to drill. The effect of drilling a prospect in a particular composite chance domain of a play is either to affirm (POD = 1.0; risk of 0) or deny (POD = 0; risk of 1) the prospectivity of that domain. Within this and other domains, however, the volumetric variables
L., Rangeof porosities ,.J I"actually presentwithinDI" I
1.0 9
,"~ ~/" . D2 (polymodal~ Q ~ ~distribution) \
~
O1 (unimodal) "Target" distributions resulting from genuine geological variance associated with distinct processes in separate process domains D1, D2 & D3.
~
I~.
Pre-drilling distributions estimated for
-~9 -Q
Post-drilling distributions for D1, D2& D3(D3is a completely tight reservoir domain).
n" P(D]) ( ;-
Estimated pre-drilling chances of reservoir in domains
..Q
D1,D2& D3.
D1,D2& D3.
.g --, P(D; I'~ o P(D3) ~
3 (non-reservoirdomain) 0
10
20 Porosity
30
(%)
Fig. 15. Illustrating the distinction between the irreducible variance of genuine geological origin characteristic of specific process domains, and the best-estimates of these distributions predicted by the depositional and diagenetic process-response models, based on a finite data-set before (light solid) and after (dashed) calibration of these domains by drilling. This schematic example portrays three reservoir process domains of a play (D1, D2, D3). However the same distinction applies also to seal and charge domains. Sub-areas within D1, D2 and D3 of equal risk (1-Pr) are the reservoir chance domains, determined by data density.
A model-based approach to evaluation of exploration opportunities
still have a finite, irreducible variance of genuine geological origin (Fig. 15). The composite chance for reservoir, seal, closure and charge for other prospects in that same domain, should therefore be set to 1 or 0, respectively, depending on whether drilling has yielded a discovery or not. In the case of undrilled prospects within domains calibrated by a discovery, the only remaining uncertainty relates to the size of reserves, and the possibility that the exact spatial extent of the domain has been incorrectly estimated from the basin model. The impact of the drilling results on other undrilled composite chance domains comes through the postmortem drilling review, when the basin model is updated for these results (Fig. 4), and the spatial extent of the composite chance domains is, as a result, also updated. Residual risk, however, will continue to characterize these domains until they, too, are eventually drilled. The following hypothetical examples (Figs. 6-9) illustrate the relationship between process domains and chance domains. In most cases a scoring range has been assigned to each domain, rather than one specific value. This allows chance domain scores for individual closures to be "fine-tuned" by prospect scale considerations.
Closure chance domains example Three play or closure types are present, each covering a discrete spatial domain (Fig. 6). Part of the region is covered by 3D seismic data and part by 2D data. For play type 2 the certainty with which closures can be identified corresponds directly to the density of data coverage. Although the general trend of these closures is revealed by the 2D data-set, structural complexity is not. Consequently there is a large risk that a well located on the basis of 2D seismic will not test a valid closure. For play type 1, the nearest mapped horizon to top reservoir is an overlying drape horizon. Although the acquisition of 3D seismic data reduces the risk of spatial aliasing of mapped closures, it does not solve the problem of resolution at the level of top reservoir (base top seal). This inability to produce accurate top reservoir maps represents an irreducible geological risk. Play type 3 closures occur within a structurally complex zone of inversion and risk would also be reduced by the acquisition of 3D seismic data. Nonetheless, the domain score is relatively high because it is known by analogy with similar basins that this play has a good chance of containing viable closures. A further consideration that has not been addressed in this example are the uncertainties in the interpretation of structural geometry created by imprecision in the depth conversion model.
195
Reservoir chance domains example In this example (Fig. 7) the reservoir process domains are associated with a simple shallow-marine siliclastic model. The probability of reservoir presence is determined primarily by the spatial arrangement of the various lithofacies belts and their shifting location through time. As shown in Fig. 7, the certainty of encountering reservoir is highest within the inner shelf and littoral facies belts. However, even within the inner shelf area there may be sand-poor areas depending on spatial variations in sediment supply. Consequently, the highest scoring chance domain corresponds to the area where presence of reservoir is calibrated by well control. Acquisition of additional data may permit the definition of more detailed process domains (e.g. areas of sand-rich and sand-poor lithofacies). The difficulty in predicting the presence of reservoir in the outer shelf zone, characterised by the deposition of grain-flow sands, is reflected in lower risk tranche scores. However, there is always at least a 0.5 chance of encountering reservoir. Seal chance domains example The seal domains shown in Fig. 8 are determined primarily by the extent of the sealing lithofacies, modified by the geometric relationship of top seal and top reservoir, and structural style or play type. Probability of seal is highest for play type 1 closures owing to the presence of a conformable, continuous sealing unit. The broad scoring band for play types 2 and 3 reflects the variable influence of faulting on seal integrity. Consequently the score of individual closures within each domain will depend on additional prospect-scale considerations such as amount of fault throw, and intensity of faulting. The score for play type 3 closures is influenced by the presence of a gas chimney, which possibly indicates that the seal of some of the closures has been breached. Other considerations not included in this example include the influence of hydrocarbon hydraulics and associated capillary leakage. When risking stratigraphic traps, the integrity of top and base seal should be considered separately because these often correspond to different lithological units. Charge chance domains example The charge chance domains (Fig. 9) are determined by the combined consideration of source rock presence and quality, the degree to which hydrocarbon expulsion has occurred, the effectiveness of the migration route, and the timing of closure formation relative to migration. The chance domains also relate specifically to one type of predicted hydrocarbon, in
196
this case oil, as different hydrocarbon phases may have contrasting migration efficiencies. The highest scoring chance domains surround the areas of maximum hydrocarbon expulsion, as it is predicted that the lateral migration impedance is relatively low. Clearly, the arrangement of domains would be more complicated if lateral impedance was high and migration depended on specific fracture systems. The slightly lower score coinciding with the main part of the expulsion kitchen relates to uncertainty in the relative timing of expulsion and trap creation by inversion. The lowest scoring chance domains are areas predicted to be in the migration shadow zone, or associated with carrier bed discontinuities.
Application to the assessment of prospective potential Prospect evaluation of drilling opportunities The composite chance map describes the spatial arrangement of composite chance domains D 1, D2, D3, D4, . . . , Dn for a play. The inventory of prospects mapped for the play comprises prospects present in one or more of these chance domains. The estimation of distributions for the volumetric variables porosity, net-gross reservoir, and hydrocarbon saturation, like estimates of the chance of reservoir, seal, and charge, are based on the specific predictions of the basin model for the appropriate process domains. Distributions for gross rock volume of closure on the other hand are estimated individually for each prospect from considerations of the precision of the actual volumetric technique used, and the style of closure. For each of these volumetric variables, the distribution to be estimated is the underlying target distribution associated with the irreducible variance of pure geological origin (Fig. 15). These distributions can be used to estimate percentiles (e.g. P05, P50, P95) for input to Monte Carlo reserves simulation. The type of distribution used (e.g. normal, lognormal, Poisson, etc.) is that most consistent with the corresponding process-response system. For example, available data may suggest that lognormal porosity and normal n/g distributions may be assigned to prospects within a reservoir process domain inferred from a shallow-marine siliclastic depositional model to represent a shoreface sand lithofacies belt. Typically, distributions for hydrocarbon saturation (Sh) can be estimated from consideration of both the charge process domains in order to prognose the particular hydrocarbon phase predicted by the charge model, and the reservoir process domains, to the extent that Sh is typically controlled by capillary
B.A. Duff and D. Hall
effects. Capillary effects are in turn dependent on reservoir poroperms. The organization of prospects into the process domains (Figs. 7-9) therefore ensures that the variance of each of the volumetric distributions input to Monte Carlo analysis is founded on the predictions of the basin model, and its constituent depositional, diagenetic and charge process-response systems. As in the corresponding assignment of chance probabilities, it therefore fully exploits the available information, and minimizes the variance compared to any other method of prospect evaluation. Risking the undiscounted reserves distribution of a prospect from Monte Carlo simulation by the POD yields a geologically-risked mean reserves estimate (risked "reward"). The prospect inventory can then be consistently ranked in terms of this estimate. This in turn can be used as the basis of a rational drilling strategy which prioritizes the lowest composite risk (1-POD), highest reward prospects. Predictive application of the basin model through the logically-consistent framework of the process and chance domains of any play, guided by the risk tranches method, therefore optimizes estimation of both composite geological risk to prospectivity, as well as the distributions of volumetric variables required as input to Monte Carlo reserves simulation. Furthermore, the composite chance (POD) domains for a play yield a clear insight into the overall variation of risk across all prospects in the play (the risk spectrum of the play), and therefore a risk-based prospect ranking, in advance of probabilistic reserves modelling.
Resource assessment Agencies such as national geological surveys and energy departments, as well as oil companies, frequently need to estimate the total hydrocarbon prospective potential of plays in one or more sedimentary basins. Many methods have evolved, mostly probabilistic, which attempt to allow for the great variability in information available to characterize basins and their plays. One of the main difficulties for many of these resource assessment methods is how to best address the problems of interpolation and extrapolation of risk and volumetric parameter distributions when, as commonly occurs, available data are sparse and inhomogeneously clustered in parts of the basin. The methods advanced in this paper permit hydrocarbon resource assessments which maximize use of the available exploration data. In particular, the problems of extrapolation and interpolation into areas of little or no data are optimally addressed because estimates of the volumetric parameter distributions and risks are based on the actual spatial variations in these properties predicted by the basin model.
A model-based approach to evaluation of exploration opportunities
Play Process Domains Reservoir~] Seal o
] 97
~
~~
charge I/
Chance domains
~POD~~D~,~
----~
POD1 _ ~
2
Prospects,~, Closure
Total Basin/Country Resource Potential = PRP
r
Fig. 16. Resource assessment for one or more basins using the methods proposed in this paper. reserves for each prospect; N = the number of prospects in each chance domain.
Our proposed method for resource assessment (Fig. 16) comprises the following steps: (1) Based on all available data, identify plays within the basins, following the play definition proposed in this paper. (2) Within each play of every basin, use all available data and the risk tranches method to assign process and composite chance domains with discrete probability estimates. (3) For each composite chance domain within all plays, assign estimates of the volumetric parameter distributions using the associated reservoir and charge process domains, and estimate the composite chance (POD), based on the risk tranches method. (4) From these distributions and consideration of the closure-generating process defining the play, estimate deterministically, or if there are enough data, probabilistically (Monte Carlo method), mean reserves for each composite chance domain in all plays. (5) From considerations of the particular closuregenerating process defining the play, estimate the number of prospects most likely to be present within each composite chance domain for all plays. (6) For all composite chance domains for all plays compute the total geologically-risked mean reserves estimate from the product of number of prospects, mean reserves and the POD. (7) Sum this risk-weighted value across all composite chance domains in order to estimate a total risked reserves estimate characterizing each play in every basin. Sum these estimates in turn across all plays to determine an overall risked hydrocarbon reserves estimate for the basin or country.
Conclusions The austere economic climate affecting the oil industry has heightened awareness of the prime aim
Total Play Resource = Potential, POD * NPRP, * MR
POD --
probability of discovery;
MR = m e a n
of exploration: cost-effective minimization of risk to prospectivity, and maximization of reward in terms of discovered reserves. At the same time, major advances in exploration techniques such as 3D seismic help to realize this aim. It appears, however, that there is little awareness of how the results of these techniques can most cost-effectively be organized to minimize risk prior to committing funds to drilling or taking up acreage. This minimization of risk is best achieved by the synthesis of all available technical results into a unifying basin model comprising the key depositional, diagenetic, tectonic, halokinetic and charge process-response systems which impact on hydrocarbon prospectivity. The testing of model predictions through the acquisition of further exploration information, and a post-mortem evaluation phase during which the model and its associated process-response systems are refined to accord with the new information, give rise to the exploration cycle. This view of exploration allows the notions of exploration risk and reward to be linked to the value of exploration information, and highlights the irreducible exploration risk as a guideline for cost control by the curtailment of exploration other than drilling. Prior to reaching this point, the testing of model predictions within the context of the exploration cycle provides the most logical and cost-effective way of programming the type and amount of exploration data to be acquired. The model-based approach to evaluation of acreage and exploration drilling opportunities starts with a process-based understanding of exploration plays, in which closure style is taken to be the principal defining characteristic. Process domains are sub-areas of a play within which the same or similar processes operated, determining the volumetric parameters porosity, net-to-gross, and hydrocarbon
B.A. Duff and D. Hall
1 98
saturation, as well as the risk of absence of reservoir, seal, charge and closure. Between these domains the processes varied significantly. They can therefore be used both for the estimation of the volumetric parameter distributions for input to probabilistic modelling of reserves, as well as for risking individual prospects in any play. The risk tranches method maximizes consistency and minimizes the subjectivity of estimating and assigning probabilities to the presence of reservoir, seal, charge and closure, and therefore to the composite geological risk to prospectivity. Application of this method to the reservoir, seal and charge process domains of a play, and to the entire closure-defined play area, yield sub-areas within which the composite geological risk to presence of reservoir, seal, charge and closure is constant. These play chance domains permit an early ranking based on composite risk of prospects within a play. Through the basin model, composite chance domains take full account of all available information and its spatial variability. They therefore provide a way of optimally estimating geologically-risked reserves for prospects and discoveries, using standard probabilistic Monte Carlo methods. This in turn allows a rational exploration or appraisal drilling strategy to be formulated. The methods also provide a process-based way of assessing play and basin resource potential, which fully exploits the available scant and clustered information. Organizing all exploration data through a predictive model of prospectivity in the context of the exploration cycle guarantees that the business decision
B.A. DUFF D. HALL
concerning acreage acquisition or drilling is optimized, because the precision of prospect evaluation is greater than if this evaluation were based directly on the results of individual exploration techniques, as often appears to be the case.
Acknowledgements We wish to thank the management of PetroFina s.a. and Fina Exploration Norway for permission to publish this paper, and for their encouragement to develop the ideas expressed. In particular, we thank Messrs. G.A. McLanachan and M. Green for their constructive comments. We stress, however, that the views expressed are those of the authors, and do not necessarily reflect those of PetroFina s.a. Finally, we thank R. Miller, whose helpful comments as referee undoubtedly improved the text.
References Krumbein, W.C. and Sloss, L.L., 1963. Stratigraphy and sedimentation. In: J. Gilluly and A.O. Woodford (Editors), W.H. Freeman and Company, San Francisco, Calif. Nederlof, M.H., 1982. Estimation of undiscovered hydrocarbons: methods and pitfalls. Paper presented at the 6th exploration seminar of the Egyptian General Petroleum Corporation, Cairo. Nijhuis, H.J. and Baak, A.B., 1990. A calibrated prospect appraisal system. Proceedings Indonesian Petroleum Association, 19th Annual Convention, October, pp. 69-83. Popper, K.R., 1959. The Logic of Scientific Discovery. Unwin Hyman press, London. White, D.A. and Gehman, H.M., 1979. Methods of estimating oil and gas resources. AAPG Bull., 63(12): 2183-2192.
Fina Exploration Norway, Skogstostraen 37, P.O. Box 4055 Tasta, N-4004 Stavanger, Norway Present address: Fina Exploration & Production, Rue de l'Industrie, 52, B-1040 Brussels, Belgium Fina Exploration Norway, Skogstostraen 37, P.O. Box 4055 Tasta, N-4004 Stavanger, Norway Present address: Fina Exploration & Production, Rue de l'Industrie, 52, B-1040 Brussels, Belgium
199
Risk and probability in resource assessment as
functions of parameter uncertainty in basin analysis exploration models S. Cao, A.E. Abbott and I. Lerche
Basin modelling development can be separated into three stages. The first stage is model development during which developing mathematical/computer models is the main theme. As basin modelling techniques are widely used in hydrocarbon exploration, the uncertainty and sensitivity associated with basin modelling become major issues (the second stage). For a given model, one must examine how the model results are influenced by the changes (uncertainty) in the model assumptions, parameters of the model, and by errors in, and finite sampling of, the input data used as control information. The third stage is that of risk analysis in basin modelling. Because of the uncertainties associated with basin modelling, the model results have to be assigned a risk factor. In principle, risk analysis in basin modelling should take a Monte Carlo approach, i.e. simulating probability distributions of model results by considering all possible values of assumptions, parameters, data effects, and all possible outcomes of the uncertainties. Unfortunately, the Monte Carlo approach is not appropriate in practice, because the many uncertainties and variables that are associated with basin modelling mean that it takes too much computing time to complete even a few Monte Carlo simulations. In this paper, we present a cumulative probabilistic procedure for risk analysis in basin modelling that is numerically quick, and which provides a risk assessment without loss of accuracy. The procedure does not require massive Monte Carlo computer runs and so is of importance in providing risk assessments in a timely manner. Applications of the cumulative probabilistic method are given for a real case history to illustrate how one can assess the scientific risk associated with variations and uncertainties, and also to show how one can discriminate sensitive from insensitive controls on the risk factors.
Introduction Geological processes related to petroleum generation, migration and accumulation are complicated and no model can simulate these processes exactly. A comprehensive examination of how the results of the model are influenced by changes in the assumptions and parameters of the model, and by errors in the data, is required, thereby making the model a more effective tool in basin analysis. Integrated models consist of three parts: a geohistory model, a thermal history model and a hydrocarbon generation model. (a) Geohistory model. Through a simulation of the fluid flow movement in sediments caused by the compaction of the sediments, the geohistory model reconstructs the burial history, basement subsidence, vertical and horizontal fluid flow, and the changes of porosity, permeability, pressure and fluid flow rate with both time and depth across a basin. Also, the evolution of cementation, dissolution and fracturing, caused by abnormally high pore pressure, are
simulated in terms of the change in formation permeability. The input data required to run a geohistory model are the depth and age of each formation base at a suite of locations in the basin, the lithology and palaeowater depth of each formation. (b) Thermal history model. Based on the burial history created in a geohistory model, the thermal history model reconstructs the thermal history by: (1) comparing predicted thermal indicator values (such as vitrinite reflectance) to measurements down a borehole; and (2) adjusting the palaeoheat flux to minimize discordances. The outputs are: (i) a heat flow change with time; (ii) a temperature change with time and depth; (iii) vitrinite reflectance changes with time and depth. The input data required to run the thermal history model are the temperature at the sediment surface, bottom hole temperature (or thermal gradient/present-day heat flow), and some thermal indicator measurements with depth. (c) Hydrocarbon generation model. The hydrocarbon generation model is based on the kinetics of kerogen degradation, and the general scheme of
Quantification and Prediction of Petroleum Resources edited by A.G. Dor~ and R. Sinding-Larsen. NPF Special Publication 6, pp. 199-218, Elsevier, Amsterdam. 9 Norwegian Petroleum Society (NPF), 1996.
200 evolution uses mathematical models to simulate hydrocarbon generation. Based on the burial history and temperature history produced in the model, the generation model gives the absolute amount of hydrocarbons generated (per gram kerogen) with time and depth. The input data required to run the generation model are the content of different kerogen types of each formation. Results which need to be examined in the analysis include: formation thickness, formation porosity, formation permeability, formation pressure, heat flow with time, temperature with time and depth, vitrinite reflectance with time and depth, "oil window" in terms of time and depth, and amount of hydrocarbons generated. Three groups of variables which influence the results are: input data, equation parameters and intrinsic assumptions. (1) Input data. Very commonly there are errors in the measurements of the geological and geochemical data and, in addition, the data are non-uniformly sampled with depth at different locations as well as being of different total number at each location. The variables which need to be examined in the input data are: depth and age of each formation base, lithology and palaeowater depth of each formation, unconformity time and eroded thickness, temperature at the sediment surface, bottom hole temperature, heat flow at present-day, thermal gradient, vitrinite reflectance, kerogen type and content of each formation. (2) Equation parameters. Most of the parameters/ constants in the equations used in the models are based on empirical data (for example, a depositional porosity of shale, 0.62, is often used as a default in geohistory models). The following parameters need to be examined for the sensitivity of output results to variations in the input parameters: depositional porosity, permeability and frame pressure, viscosity of the fluid in the sediments, parameters in void ratio/frame pressure functions and in the void ratio/ permeability functions for each lithology (see later), critical temperature and doubling temperature in the vitrinite reflectance equation, thermal conductivity of each lithology, activation energies and frequency factors in the generation model, etc. (3) Intrinsic assumptions. Most models contain intrinsic assumptions such as the 1-D nature versus a 2-D problem, such as the replacement of complete isostatic movement of basement by partial flexural compensation, etc. Examination of these assumptions would make for a very long paper, and on this ground are not included, although we recognize the pressing need for such a development (Cao and Lerche, 1989). The results of models have to be evaluated for four factors: (i) the sensitivity of the model results
S. Cao, A.E. Abbott and I. Lerche
to changes in input assumptions, parameters or data, i.e. how much model results vary with input changes; (ii) the resolution of the model results i.e. whether the mean output value of a quantity is well-resolved (low variance) or is poorly resolved (high variance); (iii) the precision of output results (e.g. there is little point in quoting output temperature values with depth to decimal level values if the input controls are uncertain by much larger factors); (iv) the uniqueness of the model results (e.g. it can happen that two very different models of palaeoheat flux variations with time are consistent with observed thermal indicator data at the present day. In that case the model results are not unique.) Thus, on the path to achieving an evaluation of hydrocarbon accumulation site determinations, the four major concerns on model results are: (i) resolution; (ii) sensitivity; (iii) uniqueness; (iv) precision. These four factors are controlled to a greater or lesser extent by the three overriding inputs of: (i) intrinsic assumptions; (ii) parameter values; (iii) quality, quantity and frequency distribution of data. A detailed analysis of the evaluation of basins taking into account the three input factors and the four output concerns has been given elsewhere (Lerche, 1989, 1990) for dynamical, thermal and hydrocarbon assessments. It cannot be stated too strongly that every case history investigated should be evaluated in relation to the points made above. However, the appropriateness of any quantitative basin analysis result in relation to the evolution of a basin can be no better than the quality and quantity of the modelled behaviours. At any given stage of knowledge of a sedimentary basin, there are always the questions of determining what can be garnered without equivocation, what can be achieved with a degree of equivocation that is acceptable, what cannot be evaluated without further data, and what measures of "acceptable equivocation" can be provided. The purpose of this paper is to provide a risk analysis assessment for quantitative basin analysis studies when uncertainty is allowed for. The development of quantitative basin analysis can be separated into three stages (Fig. 1): the first stage of model development concentrates on the construction and use of mathematical and numerical models of basinal evolution; the second stage is uncertainty and sensitivity in which the main aim is to examine the stability of model results related to the uncertainties and ranges of variations in assumptions, parameters and input data; the third stage, and the one we are dominantly concerned with here, is risk analysis, in which model results are assigned risk factors dependent on the uncertainties and sensitivities of the models used, as in stage two above.
Risk and probability in resource assessment as functions of parameter uncertainty
Basin Modeling
Third Stage Risk
Analysis Second Stage
Uncertainty & Sensitivity First Stage Model Development Fig. 1. Developmentof basin modelling.
The purpose, then, is to provide quantitative probability measures to outcomes of model calculations in order to have objective, reproducible estimates of the likelihood of uncertainty. From such measures strategic exploration decisions can be made, tied also to the probability of finding economic hydrocarbon accumulations (Lerche, 1992). According to Yukler and Kokesh (1984), mathematical models are used to simulate complex processes with one or more variables and thus they are essential in the assessment of hydrocarbon resources. Mathematical models are applied either as statistical models or as deterministic models to reconstruct and predict geological processes assuming that these processes are deterministic. The statistical models are mainly used in estimation of hydrocarbon resources because they cannot directly analyze the dynamic processes of hydrocarbon generation, migration and accumulation. A Monte Carlo simulation is usually employed in the statistical models to construct various probability curves, with the most commonly used assessment methods being geological analogy (Weeks, 1952; Conybeare, 1965; Bally, 1975; Pitcher, 1976; Warren, 1979), delphi (Miller et al., 1975), areal and volumetric yields (Stoian, 1965; Walstrom et al., 1967; Smith, 1968; Jones, 1975; Newendorp, 1975; Roadifer, 1975), geochemical yields (Conybeare, 1965; McDowell, 1975), and field distributions (Atwater, 1956; Roy et al., 1975). The need for such measures is illustrated by the following specific example. The Chukchi Sea basin, Alaska, was (prior to 1989) a true frontier basin in that only seismic information was available from five regional lines; no wells were then available. Nevertheless, an estimate of likely hydrocarbon proneness of the basin was required in order to decide on future drilling and leasing commitments. The "background" geology of the basin, in relation to palaeogeography and continental plate motion,
201
can be estimated so that a rough idea of basin setting, palaeotemperature (warm/cold), likely tectonic events influencing the basinal development, and likely lithologies are to hand. With this background information, it is then possible to identify, from the seismic sections, tectonic sequences (compression/extension/ rotated blocks, etc.) and to suggest, from sequence stratigraphy, the possible chronostratigraphy of sedimentary units as a function of sea level variations. From the structural and stratigraphic identifications, and from estimates of sedimentary load at each location, some inference can then be made of basement motion (isostatic/flexural/visco-elastic) and of fault development with time. Seismic velocities, tied to a Wyllie-type (1963) model relating velocity and porosity, can often be used to assess a rough porosity with depth behaviour, as can gravity (i.e. density) modelling across the basin. At this minimal data stage, the determination of palaeoheat flux, or even present-day heat flux, is a problem for there are only a few indications to guide and control. Some form of model is required. At least four possibilities are available: (i) arbitrarily assign a basement heat flux with time according to idiosyncratic preference m this procedure is recommended only as a last resort; (ii) tie the palaeoheat flux model at each location to a basement motion (McKenzie-type rift; Royden intrusive model, etc.); (iii) tie the motion of the section as a whole, with sediment loading, to the depth to detachment under a model of regional load compensation assume this detachment is always at a fixed brittleductile temperature (450-550~ is conventional); (iv) use gravity/aeromagnetics and/or seismic to figure an estimate of the depth today to the brittle/ductile transition. Estimate the temperature at the transition from analogy. Construct a present-day temperature gradient estimate and then allow for an uncertainty range (e.g. 2.5 4-0.5~ m); assume this range of gradients covers all possibilities. For preference it is appropriate to try to do the last three of the above four, and so run several cases. In this way some idea of the range of output results can be achieved even with minimal data. Such considerations have been given in detail elsewhere (Cao et al., 1993). For illustration purposes we concentrate here on the palaeoheat flux. As noted in Cao et al. (1993), there is uncertainty in the palaeoheat flux because there were no wells at all in the Chukchi Sea Basin prior to 1989. Three possible extreme heat flux patterns were proposed based on geological and geothermal data, with the geothermal data coming from the nearest wells ~ which are several hundred kilometres from the centre of the basin. Fig. 2 shows the three palaeoheat flux patterns used (labelled Increase, Constant and Decrease) based on
S. Cao, A.E. Abbott and L Lerche
202
7:
5.0
-5.0
4.0--
_ 4.0
3.0-
- 3.0
_2.0
-~ 2 . 0 -
-1.0
1.0- "v
0.0
I
I
200.0
250.0
150.0
!
100.0
I
50.0
-0.0 0
Geological Time (MaBP) Fig. 2. Three possible heat flow patterns in the Chukchi Sea basin, Alaska.
Table 1 Palaeotemperature (in degrees centigrade) at the bottom of the Pebble Shale in the Chukchi Sea Basin, Alaska Depth (m)/time (Ma BP)
4800/75 5000/55 5900/23
Heat flow pattern I
II
III
179 190 229
224 223 244
268 255 259
I = decreasing heat flow with time (lower in the past)" II = constant heat flow with time; III = increasing heat flow with time (higher in the past).
both individual well studies (shown by dashed lines) and on geothermal models of rifting: Increase (Decrease) implying higher (lower) heat flux in the past. (The extreme values were obtained by varying uplift amounts and uplift timing, together with parameters describing equations of state, for the wells. The point is that, no matter how determined, the three extreme estimates of heat flux are used here to illustrate application of the probability procedure.) The three palaeoheat flux patterns give three considerably different temperature histories for sediments in the basin, as shown in Table 1 for the Pebble Shale Unit (considered to be a likely source rock for hydrocarbons based on extrapolation from the National Petroleum Reserve, Alaska where the Pebble Shale is known from drilling to be a good source rock). The differences in palaeotemperature are sufficient to suggest significant uncertainties in the amount and timing of thermal maturity, and also in the "oil window" depth and time in relation to the development of likely reservoir structures. For example, Fig. 3 shows a burial history at a location in the Chukchi Sea basin with superimposed vitrinite iso-reflectance lines of 0.6% for each of the
three palaeoheat flux cases. The reflectance value of 0.6 is often considered as indicative of entry into the oil window for sediment maturity (Hunt, 1979; Tissot and Welte, 1984). Inspection of Fig. 3 shows that there is about a 50 Ma spread in time in the Ro = 0.6 onset estimates, and a spread of more than 3 km in depth for onset of the hydrocarbon oil window. The large ranges of uncertainty are direct consequences of the palaeoheat flux uncertainty. There is then a serious concern to provide some probabilistic measure of correctness of the above behaviours, in order to lessen the risk of estimating hydrocarbon charge to reservoirs. Cumulative probability and error assessment In principle Monte Carlo procedures should be applied to the study of risk assessment and most probable outcomes of basin analysis models (Smith and Buchee, 1985). The probability distribution of each model output result should be evaluated by considering all possible values of each assumption, each parameter, and each datum point, within determined ranges. Then, an intrinsic probability distribution of finding each such input control, assigned within its given range (e.g. uniform, Gaussian, triangular, log normal, etc.), should be computed. Using a multiMonte Carlo approach to sample values for each and every input control a large number of times, a basin modelling calculation can then be run through to produce a set of outputs. The probability distributions of the respective outputs can then be recorded and used in risk assessment (Figs. 4, 5). While the Monte Carlo procedure is completely general and can be applied to any basin analysis model, once the range and underlying probability distributions for each parameter (and for the measurement errors in data) have
203
Risk and probability in resource assessment as functions of parameter uncertainty
Geological 0
2OO
250
Time
,
(MaBP) lOO
I5 0
I
~
5o
0
,.\,\\
1,000 J 2,000
E v
t'" ~9 t'~ 0
n
3,000 4,000 5,000
~
6,000 L_ E3 fn
7,000
Iso Ro=0.6 l in
I ] I i i I
8,000
9,000
Upper B r o o k i a n
Upper
Lower Brookian
IilllllIlllllll
Ellesmerian
Lower Eilesmerian
Fig. 3. Burial history at a location of the Chukchi Sea basin with superimposed vitrinite iso-reflectance lines of 0.6% for each of the three palaeoheat flux cases.
Monte carlo Approach:
Simulating probability distributions of model results by considering all possible values of parameters and all possible outcomes of the uncertainties. r
OUTPUT Porosity Permeability Pressure Temperature Ro Oil Window HC Generation HC Migration
INPUT DATA )
|Depth, Age, lilhology.... [ ITemperature, Heat Flow, I
~
/
LR~
Kinetics....
J
~-{'Basin Modelin~'~~ PARAMETERS "1 -
~._
~ ,
- L A'B....
'~rASSUMPTIONS
Appendix is done deliberately so as not to interrupt the flow of the main thrust of the paper, which is the use of quantitative measures to handle the uncertainty in basin modelling computations. In the body of the paper focus is given to actual applications in order to illustrate, by direct example, the ability of the cumulative probability method to generate measures of probable behaviour for particular attributes of interest.
J
~
[ 1) Basement ........... l L 2) Temperature ...... fl
Fig. 4. Monte Carlo approach for risk analysis in basin modelling.
been provided, the problem is that such a numerical investigation can be extremely computer intensive, perhaps one of the reasons that such methods are not routinely applied to the results derived from basin analysis models. It would, therefore, be of help if alternative procedures could be used, which would provide the same probability outcome, but which would not demand such massive computer time investment as do the Monte Carlo methods. This paper demonstrates how to use cumulative probability techniques to achieve such a goal. The Appendix gives not only the general mathematics showing how to develop a cumulative probability methodology for handling the uncertainty, but also how to use the procedure practically. This relegation of the quantitative mathematical procedures to an
An application of cumulative probability techniques to the Navarin Basin COST No. 1 well A detailed quantitative geological investigation of the dynamical, thermal and hydrocarbon histories of the COST No. 1 well has been given elsewhere (Cao and Lerche, 1989) and does not need to be repeated here. The location of the COST No. 1 well is shown in Fig. 6a, and the corresponding geological column is shown in Fig. 6b. For the purposes of illustrating how the cumulative probability and risk assessment procedure operates, we use a onedimensional fluid-flow/compaction code to investigate dynamic and thermal maturity evolution with time in the well. A one-dimensional fluid-flow/compaction code allows deposition of each lithologic unit with time, ascribes a relation between permeability and porosity (void ratio) to each lithology, as well as a frame pressure/void ratio equation of state for each lithology, and then allows the sediments to compact according to fluid loss rates controlled by Darcy's law
S. Cao, A.E. Abbott and L Lerche
204
p(x) 1.0
Uniform
I
(a) Pi ~x
0.0
a
I
] I
x.i
b
(xi-a) Pi - (b-a)
~x
p(x)
f(x) i
1.0
Triangular
(b)
For x i <- x p'
(xi-a)2 Pi- (b-a)(xp-a)
For xi >= xp'
l)i- 1-
[
__ a
xP
b
~x
I
0.0
(b'xi)2 (b-a) (b-Xl))
~x a
Xp x i
!)
Fig. 5. Four common probability distributions and their corresponding cumulative probability distributions: (a) uniform; (b) triangular.
in a compacting system. Sediment overburden is allowed for as is the ability of fluid to escape from formations with time, depending on both the permeability and fluid-pressure drive. Empirically (Dutta, 1989) it would appear that each lithology has a permeability/void ratio equation of state of the form: B k -
k,
--
e,
where k , ( e , ) is the depositional permeability (void ratio) of the lithology; k ( e ) are the values after burial, and B is a parameter in the rough range of about 3 + 3 for most lithologies. (Porosity, 4), is related to void ratio, e, by 4) = e/(1 + e)). Equally, the empirical relation between framepressure (effective stress), pf, and void ratio, e, for each lithology is of the general form: pf-
pf,
-e,
where pf, is a scaling value, and A is a constant, also in the rough range 3 4- 3. Each lithology (sand, shale, sandy-shale, carbonate, salt, etc.) has different values and ranges for each of the parameters k,, e,, pf,, A, B. The evolution of excess fluid pressure, porosity, and other dynamical quantities of interest is then influenced by choices made for the parameter values. The difficulty, as always, is that we do not know the values of these parameters with precision; so any dynamical predictions of, say, the build-up of excess fluid pressure with time will entail an uncertainty. The
task of the cumulative probability technique is to address that uncertainty in some objective, reproducible, quantitative manner. To illustrate the method, consider the two lithologies shale and sandy-shale which constitute the lithologic zones of the Navarin Basin COST No. 1 well from mudline to near base Oligocene (at 4000 m). Start with the empirical ranges thought to represent approximately the values of parameters considered geologically acceptable based on whatever criteria are considered controlling factors on the ranges. For instance, observations and extrapolations of presentday porosity with depth trends from many wells suggest that it is very unlikely that the surface void ratio for a sandy-shale lithology lies outside the broad range 0.16 _< e, _< 2.03 with a prevalent value of 1.07. In Table 2 we present the minimum, maximum and most likely values of 9 parameters which influence both the dynamical (8 parameters) and thermal (8 dynamical plus the rate of change of heat flux with time) evolutions. The thermal parameter, fl, is determined from the heat flow as: Q(t) -
Qo
exp(flt)
where Qo is present-day inferred heat flux, t is geological time (with t - 0 being the present day) and fl (units of Ma -1) measures whether palaeoheat flux was higher (/3 > 0), constant (fl - 0) or lower (/3 < 0) in the past. The range of fl can be assessed either from a model behaviour (e.g. a rifting model) or from uncertainty in fl when attempting to perform a thermal inversion with control against present-day observed thermal indicators with depth.
205
Risk and probability in resource assessment as functions of parameter uncertainty
p(x)
f(x)
~
ttistogranl
0'9991t
Normal
0.84
o.soo+
(c)
0.159i___
-
X
I
/ ~x
0.001_1
i
~.t- I2(n x) / 2n
[ ! i
"
~..x
2
2
o - ~_(x-~t)/ N
fx) ~t:(x
0.999 t
Log Normal
(ci)
Xp ~t
p
)/3 I11 ;ix
02=1.t2/2 -(x . x
0.841F . ~ 0.500-{0.159[------ ,._o.I I ~1"-~
Xmin
+x+x iiiin
xp(~t) 2
[_..x,,,cxi)(.~t )
~x 0.001_1 Xmax
Inlll
+x (X . +x......))/6 II:aX
I)
111III
x = ~1(I+o2/~t2)"in 1/2
xl/zexi)(-~t2) - ~t(1+O2/~L2)"3/2 xs/2exp(~t) - xs/zexp[(ln(l+oZ/lt2)) I/2]
~log (x) XI/2
Fig. 5 (continued). (c) Normal; (d) log-normal.
Table 2 Uncertainties in the nine parameters used in the model, presented as minimum, maximum, and most likely values Parameter value
Minimum value
Maximum value
Most likely value
Heat flow/3 (Ma -l) Shale A Shale B Shale k* (md) Sandy shale A Sandy shale B Sandy shale k* (mD) Sandy shale p* (atm) Sandy shale e*
-0.015 0.60 1.10 70.0 1.20 0.80 10.0 0.21 0.16
-0.450 5.00 5.10 220.0 10.00 4.70 530.0 3.00 2.03
0.015 1.77 2.48 160.00 4.97 2.50 300.0 1.70 1.07
With the range and most likely value of each parameter provided, the burial and thermal maturity histories of the COST No. 1 well can be run in the manner outlined in the previous section. Two simple pictures can be drawn: first one can plot present-day predicted behaviour with depth for a quantity of interest and superpose on the plot both any observed or inferred values of the quantity as well as the cumulative probability likelihood values. We do so here for porosity, fluid pressure, and vitrinite reflectance. Second, one can plot the burial history with time (which varies for each computer run depending on the parameter values chosen for each run) and superpose the development with time of each quantity of interest (porosity, fluid pressure, and vitrinite reflectance) together with the cumulative probability. In addition to these two simple pictures, one can also plot the present-day cumulative probability of a particular iso-
value being reached at a given depth and, with time, the corresponding cumulative probability. From the perspective of when fluids were most likely in motion in the sub-surface and when thermal maturity was sufficient to generate hydrocarbons, the cumulative probability plots with time are extremely useful.
Porosity and probability Following the prescription given above, and using the ranges of parameter values of Table 2, Fig. 7 plots the predicted probability of porosity with depth at the present day. In addition the nominal measured porosity values are represented by filled circles. No attempt has been made to allow for uncertainty on the measured porosity values, although the scatter in the data raises a cautionary flag that some error of measurement (or inference) is likely present, depending on how the porosity was measured. Nevertheless, in order to focus on the essence of the cumulative probability procedure, the error in measurement is not discussed here (see, however, Lerche, 1993). Four curves are drawn on Fig. 7, representing cumulative probability iso-values of 30, 50, 70 and 90%, respectively. For instance, at a present-day depth of 3000 m there is a 70% chance that the porosity will be less than about 17% but only a 30% chance the porosity will be less than 8%. A differential probability can be obtained by subtraction: thus the probability that the porosity lies between 28% and 8% at 3000 m depth is 60% (90% cumulative probability of less
S. Cao, A.E. Abbott and L Lerche
206
Fig. 6. (a) Location of COST No. 1 well in Navarin basin, Bering Sea, Alaska.
than 28%, 30% cumulative probability less than 8%). Clearly, as the ranges allowed for different parameters are chosen to vary in different ways so, too, will the cumulative probability values. An alternate way to view the uncertainty in outputs is to plot cumulative probability with depth for fixed values of the porosity. Thus, in Fig. 8 are given the likelihoods of obtaining a fixed porosity. For example, at 3000 m the cumulative probability that the porosity is greater than 10% is 45%, while the cumulative probability that the porosity is less than 20% is 80%. Again, a differential statement can be constructed by subtraction: the probability for the porosity to lie between 10 and 20% at 3000 m depth is about 35% (80% cumulative probability of less than 20% porosity; 45%
cumulative probability of less than 10% porosity). In one extreme (Fig. 7) one plots the variation of porosity with depth for constant cumulative probability values, while in the other extreme (Fig. 8) one plots the variation of cumulative probability with depth for constant porosity values; both plots allow information to be assessed quickly as to the degree of uncertainty of information at the present day in different formats. Of equal or greater importance than assessing the values of present-day porosity outputs is the development of porosity with time, which influences migration of fluids, development of excess pressure, and thermal gradients. Again two different types of plots show different aspects of the uncertainty. For example Fig. 9 shows the burial history versus poros-
Risk and probability in resource assessment as functions of parameter uncertainty
,
FEET ,
i
DEPTH BELOW SEA LEVEL
LITHOLOGIC ZONE
SEISMIC PERIOD EPOCH SEQUENCE
207
Sea Bottom
PlioPleistocene
,,
METERS
! -
1,000
First Sample --
Pliocene
I
2,000
A-1 -
3,000
-
4,000
A-2
-
1,000
Miocene
II < [.-,
C- 1
- 5,000
C-2
- 6,000 2,000
7,000
i D-1
III
Oiigocene D-2
r
8,0O0
1
9,000
- 10,000
- 3,000
- 11,000
IV
i- 12,000 Eocene 1 - 13,000
Low Cretaceous
G&H
9 < [-
-
4,000
- 14,000 15,000
LOW
Cretaceous
- 16,000 - 5,000
Total Depth --"
i 17,000
Fig. 6 (continued). (b) Corresponding geological column for the well (modified from Turner et al., 1984).
ity plots at different constant cumulative probability values. Thus on Fig. 9a, for example, there is only a 30% chance that the porosity values are less than or equal to the values shown on the figure; while if the cumulative probability is set to 70%, as in Fig. 9b, then there is a 70% chance that the porosity values are less than the values given; while at the extreme case of a 90% value for the cumulative probability the porosity is 90% likely to be less than the values recorded on the burial history curve of Fig. 9c. For instance at 20 m.y. B P it is 90% certain that the porosity at 2000 m depth is less than about 4050%, it is 70% certain that the porosity is less than 20-30%, and it is only 30% certain that the porosity
is less than 10-20%. Thus an evaluation of likely porosity evolution with time for each formation, or with burial history depth, can be obtained. By flipping the argument around we can ask: for a fixed porosity value what is the likely cumulative probability evolution with burial history? Such a situation is sketched in Fig. 10 for porosities of 10%, 20% and 30%, respectively. Note that at 20 m.y. BP at a depth of 2000 m, the probability is 30% that the porosity is less than 10%, the probability is 40% that the porosity is less than 20%, while the probability is 90% that the porosity is less than 30%. Thus it is unlikely (3 chances out of 10) that a porosity less than 10% occurred in the
208
S. Cao, A.E. Abbott and L Lerche
Fig. 9. Burial history versus porosity at different cumulative probability values. (a) 30% cumulative probability; (b) 70% cumulative probability; (c) 90% cumulative probability.
Fluid pressure and probability Fig. 8. Porosity versus depth with different cumulative probability curves for Navarin COST No. 1 well.
formation which was at 2000 m depth at 20 m.y. BE but it is also unlikely (1 chance in 10) that the porosity was greater than 30% at that time and depth. In short an estimate can be made rather quickly of the expected evolution of porosity and the likely range of variation of porosity based on the range of variation of the eight input parameters of Table 2.
Because the fluid-flow/compaction code allows fluids to escape according to Darcy's law, and because the difference between overburden weight and frame pressure is supported by a fluid pressure in excess of hydrostatic, it becomes a relatively simple matter to calculate the influence of varying ranges of parameters on the likely total fluid pressure (excess fluid pressure plus hydrostatic pressure) with depth at the present day. For the parameter ranges given in Table 2, Fig. 11 presents observed fluid-pressure measurements with depth (again without ascribing
Risk and probability in resource assessment as functions of parameter uncertainty
209
Fig. 10. Burial history versus cumulative probability for porosity. (a) 10% porosity; (b) 20% porosity; (c) 30% porosity.
any error or uncertainty to the measurements), together with cumulative probability curves. One can observe, for instance, that there is only a 30% predicted chance of a fluid pressure less than 150 kg c m - 2 at 3000 m, a 70% chance of less than 300 kg c m - 2 and a 90% chance that the fluid pressure is less than 600 kg cm -2. Curves of present-day predicted behaviour can be projected in a different manner, as in Fig. 12, where cumulative probability of fluid pressure with depth is plotted for selected values. For instance at 2000 m depth, Fig. 12 indicates less than 10% chance of obtaining a fluid pressure less than 50 kg cm -2, about
Fig. 12. Predicted cumulative probability with depth at fixed fluidpressure values at present day for Navarin COST No. 1 well.
40% chance of a fluid pressure less than 100 kg cm -2, a 65% chance of fluid pressure less than 150 kg c m - 2 and nearly 80% chance of a fluid pressure less than 200 kg cm -2. Viewed from the perspective of Fig. 12, this way of presenting results is important for drilling operations because a guide can be given as to the likely pressures expected to be encountered and the chance of encounter. The build-up of overpressure with time is also of significance in that such build-up indicates the
210
development of sealing capability of the system for likely hydrocarbon retention, the preservation of porosity, and thermal gradient increase due to lower thermal conductivity than would occur under nonoverpressured sediment compaction ~ of interest in developing earlier genesis of hydrocarbons than might otherwise have been thought to occur. Fig. 13 shows the burial history with cumulative probability curves superposed. Thus, in Fig. 13a, for a fixed value of 50 kg cm -2 fluid pressure, the cumulative probability curves indicate that the chance of a regime deeper than 1000 m having less than
S. Cao, A.E. Abbott and L Lerche
50 kg c m -2 fluid pressure is less than 40% throughout the whole burial history. Fig. 13b, drawn for a fluid pressure of 100 kg c m -2, indicates that the probability is less than 40% throughout the whole burial history that the fluid pressure is less than 100 kg c m - 2 at depths in excess of 2000 m (60% chance the pressure is greater than 100 kg c m - 2 ) . Fig. 13c (drawn for 200 kg c m - 2 fluid pressure) indicates that it is 80% likely that the fluid pressure is less than 200 kg c m - 2 during the whole burial history at all depths shallower than about 3000 m. The other projection of information is given in Fig. 14 which shows fluid-pressure development with burial history for different, fixed, cumulative probability values. Thus, Fig. 14a shows that there is only a 30% chance that fluid-pressure development will attain values less than those, at any time and depth, on Fig. 14a; Fig. 14b shows that there is a 70% chance of obtaining values less than those drawn; while Fig. 14c provides a 90% chance that fluid-pressure values will be less than those given. Thus one can assess the likelihood of the amount and timing of fluid-pressure development in terms of probabilistic ranges determined by the intrinsic assigned uncertainties in the input parameters of Table 2.
Thermal maturity and probability
Fig. 13. Burial history with cumulative probability curves for different fluid-pressure values. (a) 50 kg cm-2; (b) 100 kg cm-2; (c) 200 kg cm -2.
Vitrinite reflectance provides an indirect measure of thermal maturity of hydrocarbon proneness in a basin. The reflectance is dependent not only on parameters influencing the dynamical behaviour of the system, but also on both the present-day heat flux and the palaeoheat flux. Indeed measurements of vitrinite reflectance with depth have often been used in an inverse sense to determine, or at least bracket, the palaeoheat flux (Lerche, 1989). In Table 2 we provide a range of palaeoheat flux values considered likely to bracket the extremes of the true palaeoheat flux. The predicted evolution of vitrinite reflectance depends on both the palaeoheat flux and on parameters controlling palaeotemperature due to dynamical conditions (thermal conductivity is tied to porosity; formation thicknesses and depths are tied to permeability, etc.). In Fig. 15 we provide present-day measured values of vitrinite reflectance with depth (once again, the data are taken "as is, where is" and no attempt has been made to include error or uncertainty measures of the data per se); superposed on Fig. 15 are the corresponding cumulative probability curves based on the ranges of the 9 parameters of Table 2. For instance, if one takes a value of reflectance of 0.6% as marking the onset of the oil window, then
211
Risk and probability in resource assessment as functions of parameter uncertainty Vitrinite 0 0
0
0.5 !
(%)
Reflectance
1.0 I
1.5 I
2.0 I
2.5 I
3.0 I
/z,\,.
1,000-
2,000-
'iX,..
{'h 3 000-
4,000-
Cumulative
Probability
(< =)
30 50 70 90
5,000-9
Input
Data
Fig. 15. Vitrinite reflectance versus depth with cumulative probability values for Navarin COST No. 1 well.
0
0
~ / ,'//'
Cumulative
Probability
~0 2o 3o 40
so
I
I
I
!
I
6o I
7o I
(<=)
80 9o ~00 I
!
I
/i/II
1 ,000-
E
2 j 0 0 0 --
z::
.l..a (D
n
Fig. 14. Burial history with fluid pressure for different cumulative probability values. (a) 30%; (b) 70%; (c) 90%.
3,000-
4,000-
Vitrinite i
Fig. 15 indicates there is 90% chance that this value of reflectance is currently at a depth greater than 1200 m, 70% chance of the value being at a depth greater than about 2500 m, 50% chance of being at a depth greater than 3500 m, and only 30% chance of occurring at a depth in excess of 3800 m. Equally, if a reflectance value of 1.2% is taken as marking the onset to the gas window, then Fig. 15 indicates that there is 90% chance of the value occurring at a depth greater than 2700 m, and 70% chance of being at a depth greater than 3700 m. The burial history goes no deeper, today, than 3800 m so it is difficult to establish any further probability statements for greater depths. But the point is made. For instance: it is 70%
5,000-
reflectance I
(%)
0 4 0 6 O 8 1 0 1 2 1 4 1 6
Fig. 16. Predicted cumulative probability of vitrinite reflectance with depth for Navarin COST No. 1 well.
likely that onset of the oil window (as measured by a reflectance of 0.6%) occurs at a depth today greater than 2500 m but shallower than 3900 m; it is 70% likely that the gas window today occurs only at depths in excess of 3800 m. Again another projection throws light on the information extractable. Fig. 16 shows plots of cumulative probability with depth for
212
various values of vitrinite reflectance. From Fig. 16, we see, for instance, that at 2000 m depth there is 55% chance of a reflectance less than 0.4, 80% chance of a reflectance less than 0.6, 90% chance of a reflectance less than 0.8; while at 3000 m depth, there is only 40% chance of a reflectance less than 0.4, 70% chance of a reflectance less than 0.6, and 85% chance of a reflectance less than 0.8. Thus the onset of the oil window can be ranked either by the choice of a fixed vitrinite reflectance value (and then an assessment provided of the probable depth range at which this value occurs), or one can choose a fixed depth and then assess the probability that the vitrinite reflectance range at that depth will encompass the oil window onset today. Apart from present-day estimates of depths to oil and gas windows based on vitrinite reflectance, there are also the questions of the timing of the onset of maturity throughout the burial history of the sediments. Accordingly, in Fig. 17 the cumulative probability behaviours are plotted as a function of burial history for different fixed values of vitrinite reflectance. For example, Fig. 17a (drawn for a vitrinite reflectance of 0.6%) shows that most of the sediments are likely immature (reflectance <0.6%) throughout most of the burial history. There is a 90% chance that reflectance will be less than 0.6% for all formations except below a line commencing at about 2500 m at 20 m.y. BP and rising to 1500 m today. Similarly in Fig. 17b, drawn for a reflectance of 1.2%, it can be observed that almost all formations have a 90% chance that, throughout the whole burial history, they never reached 1.2% reflectance, with the exception of the deepest formation, currently at 3800 m, which had a 20% chance of exceeding a reflectance of 1.2% in the last 5 million years. Fig. 17c shows that there is almost a 100% chance that no formation ever reached a vitrinite reflectance of 1.6% at any time in the burial history. The plots of vitrinite reflectance with burial history for fixed cumulative probability values are recorded in Fig. 18. Fig. 18a, drawn for a probability of 30%, indicates the depth-time range values; there is only a 30% chance that the reflectance values will be smaller than those shown. Fig. 18b provides the 70% chance of obtaining a reflectance value smaller than shown, while Fig. 18c provides the 90% chance burial historyreflectance behaviour. Again one sees the shift in pattern of behaviour as the probability is increased. For example, reading off from Fig. 18c there is 90% likelihood that the vitrinite reflectance at 2000 m at 10 m.y. BP is greater than 0.5 % but less than 1%. The points to note about the Navarin Basin COST No. 1 well example are that the uncertainties in dy-
S. Cao, A.E. Abbott and L Lerche
Fig. 17. Burial history with cumulative probability for different vitrinite reflectance values. (a) Ro = 0.6%; (b) Ro = 1.2%; (c) Ro -- 1.6%.
namical parameters, and the degree of uncertainty of the palaeoheat flux,can both be used to assess the likely dynamical and thermal maturation histories with a quantitative degree of error assignment directly tied to the uncertainty of each and every input parameter. The risk of, say, reaching a particular overpressure or of not reaching a particular vitrinite reflectance value can then be assessed both for the present-day (where some control data are available to constrain the degrees of uncertainty) and in the past. The control of the probabilistic values in this example is facilitated to some extent by the use of the measured data as though those data: (a) contained no
Risk and probability in resource assessment as functions of parameter uncertainty
213
Perhaps the salient point from the present example is an understanding of how probability procedures can be used to assess risk for hydrocarbon proneness in the face of uncertainty of parameter values, or in the face of uncertainty in geological factors caused by a lack of good quality, high quantity, well-sampled data. It would seem that the example presented here illustrates these points.
Conclusion
Fig. 18. Burial history with vitrinite reflectance for different cumulative probability values. (a) 30%; (b) 70%; (c) 90%.
error; (b) were of sufficient number to control the system uncertainties; and (c) were sampled sufficiently densely with depth to provide accurate control of the system evolution throughout the total burial history. When such a situation does not obtain, then errors and uncertainties due to sparse, poorly sampled, and uncertain measurements can also provide a degree of uncertainty on the likely values to be expected for outputs of basin models, as has been detailed elsewhere (Cao and Lerche, 1990) where it was shown that only with further well information can an appreciation be obtained of the likelihood of the correctness of any particular choice for thermal maturity.
After many years of research and development, basin modelling has become a useful tool in today's petroleum exploration, especially in those areas where not many wells are available. However, when data availability decreases, the uncertainty involved will increase and so will the risk. Therefore we have to know the uncertainty and associated risk when applying basin modelling techniques to these areas with high uncertainty and high risk. It is necessary to perform risk analysis in basin modelling in order to know the resolution, sensitivity, uniqueness and precision of modelling results. To perform risk analysis in basin modelling,the Monte Carlo approach is not appropriate in practice because of the complexity of the system we are modelling (i.e. processes related to the tectonic and strata development, sediment deposition and diagenesis, petroleum generation, migration and accumulation in a sedimentary basin). Too much computer time is taken to perform risk analysis for any basin analysis model including all possible uncertainties of intrinsic assumptions, parameter ranges, and data vagaries. The probability method presented in this paper for risk analysis in basin modelling is numerically quick without loss of accuracy in risk assessment. The procedure does not require massive Monte Carlo computer runs and can be performed on a personal computer in a timely manner, which is of importance in risky and competitive petroleum exploration.
Appendix Numerical application of cumulative probability procedures to basin analysis problems is based on three considerations: first is knowledge (or surmize) of the underlying, intrinsic, probability distributions for each parameter being used in the computation; second is knowledge (or surmize) of how multiple parameters combine to provide a probability distribution for each computational run, and how to extract the cumulative probability behaviour of each required output from knowledge of the variations in each input variable parameter; third is the pragmatic concern
214
of how to operate numerically to obtain the desired statistical information on attributes of interest. In this Appendix we consider each of these three factors in turn. While some of the procedures are standard statistical manipulations which can be found in just about any reference work (e.g. Feller, 1960; Bharucha-Reid, 1968; Lumley, 1970), and while such statistical methods have been later applied in reserve estimation assessments (Warren, 1981, 1982, 1983a, b, 1988; Smith and Buchee, 1985; Withers, 1992), nevertheless the statistical development is spelled out here in detail, for several reasons. First, familiarity and standard use of techniques in one discipline do not imply that they are either well known or standard in other disciplines; in our opinion the application of cumulative probability techniques in quantitative geological models, such as the basin analysis results presented here, marks one of the first times probabilistic methods have been applied in this discipline. Second, the logic of the method, as applied numerically to problems of geology, is of interest in its own right. Third, a consistent nomenclature is required to set the stage of development not only within a familiar framework but also to illuminate the appropriate hypotheses and assumptions underlying applications of the method. For these, and other, reasons we give here the detailed development.
S. Cao, A.E. Abbott and L Lerche
normalized to unity so that the frequency distribution then provides an approximate empirical assessment of the probability of occurrence of a parameter in every interval range, based on the data to hand. Often, interest centres not so much on the frequency distribution (e.g. what is the probability that the porosity is between 1% and 2%?) but rather on the cumulative frequency distribution (e.g. what is the probability that the porosity is greater (less) than 1%7). In one case (greater than) the cumulative frequency distribution histogram is usually the residual fractional area still lying beyond a parameter value. In the other case (less than) the measure is the fractional area contained up to the parameter value. This is a mutually exclusive set (i.e. the probability of a porosity greater than 1% plus the probability of a porosity less than 1% must sum to unity). Thus i f p ( x ) d x measures the frequency distribution (normalized) of occurrence of x in the range x to x + dx, then: P(y > x)-
measures the cumulative frequency distribution, i.e. the chance of exceeding a particular value y, while: P ( y < x) =
Mathematical considerations
f0yp ( x ) dx
fo
-
In order to make full use of the techniques of probability measures, knowledge would be required of the intrinsic probability of obtaining an event. This knowledge is, by and large, based on data to do with scientific conditions. But the data available to work with are very limited at the early exploration assessment stage, usually imprecise or derived by analogy, or dependent on conditions about which little knowledge (as opposed to surmise) is available. For these reasons, approximations and assumptions are introduced in attempts to obtain relatively robust estimates of data-related quantities from which some form of assessment can be made.
-= l -
p(x)dx
p(x)dx
P ( y > x)
(2)
measures the chance of not exceeding a particular value y. In a large number of circumstances in exploration assessments, it is often difficult to provide the frequency distribution even in a rough form. For that reason moments of the underlying distribution are often used as approximations. The mean value, E1 (x), of x for a frequency distribution p ( x ) dx is: E1 (x) -
x p ( x ) dx
(3a)
oo
while the mean square value Ez(x) is: Ez(x) -
x 2 p ( x ) dx
(3b)
oo
Single parameter distributions By and large, parameters or variables that are not too well known (and we shall be more precise later) are usually treated as randomly varying in some manner around a mean value. The random component is customarily represented by a frequency distribution histogram which provides the relative number of times a parameter has been observed in a given interval range compared to all interval ranges. Customarily, the area under the histogram is
(1)
p ( x ) dx
The variance, a2, around the mean is given by" 0 -2 - -
E2(x)- E1 ( X ) 2
>__ 0
(3c)
where cr is the standard deviation. In most of the situations we shall deal with, only multiple powers of distributions will be needed, defined by: Ej (x) -
Z_ oo
x j p ( x ) dx
(4)
Risk and probability in resource assessment as functions of parameter uncertainty
We will also deal with the median value, X 1~2, of a frequency distribution defined as that value of x such that: P(y < x)--
P(y > x)--
1
(5)
and with the mode Xm (for a unimodal distribution) defined as that value of x at which p ( x ) has its maximum value. Three basic types of frequency distributions play fairly dominant roles in exploration strategy assessments. They are binomial, normal or Gaussian, and log normal distributions. The binomial distribution is particularly useful when considering the probability that wildcat wells will be dry (or contain oil) given that the intrinsic probability, q, of a given event (dry or oil-bearing) is estimated based on other knowledge. Thus: if q is the intrinsic probability that a well will be dry (and so 1 - q is the intrinsic probability of the well containing oil) the probability after n wildcats that k of them will be dry is given by: p~ (k) =
n! k!(n - k)!
qk (1 - q)~-k
(7a)
while the variance is: (7b)
For instance, an unbiased coin has an intrinsic probability q - 1/2 of coming up tails on any given trial. Thus the probability of obtaining precisely k tails out of n flips of the coin is:
n, with a mean, expected, number E l ( k ) - n / 2 , i.e. half the trials should be tails. For a biased coin in which the intrinsic probability of a tail is q 0.4, then El(k) - - 0 . 4 n , while pn(k) = ( n ! / k ! ( n k)!)(0.4)k(0.6) "-k. Thus on three trials (n -- 3), the probability of 2 tails is about 0.27, while for the unbiased coin the value would be 0.375. The normal distribution is particularly useful when there are errors in measurements or when several random parameters have to be combined together. For a mean value of (x) and a variance of o.z, the probability of finding a value x in the range x to x + d x is: p(x)dx-
erf (z) -- Jr -1/2
(27rcr2)-l/2exp[
-
( x - ( x ) ) 2] 20.2 dx
(9)
when x can range in ( - e o , ec). The cumulative probability of finding a value less than x, subject to
21/2o.
I ze x p ( - x 2) dx
(11)
Note, for further use, that at the two values xo (x) + ~, P(x I (x), or) takes on the approximate values 0.84 (at (x) + o.) and 0.16 (at (x) - o . ) . The mean, median, and mode values are identical at (x) for the normal distribution and occur at P - 1/2. The log-normal distribution occurs physically in many situations ranging from the areal size distribution of sunspots to lease sale bid distributions. The normal distribution cannot be appropriate when there is a constraint on a variable, e.g. area cannot be negative, bid values must be positive. Under such conditions, empirical evidence suggests that an approximate measure of cumulative frequency distribution is provided by a log normal behaviour with: 1[ -- ~
1 -+-erf
(ln(x/xl/2))] 21/2/z
(12)
with the mean value of x, E1 (x), given through: Xl/2 exp ( - ~ )
(13a)
the mode value by" Xm - Xl/2 e x p ( - # )
and the variance in x, E z ( x ) by: O-Z
(8)
-2
1 + erf
where:
El(x)-
O-2 __ nq(1 - q)
k!(n - k)
P ( x I (x), o-) - ~
P(XlXl/2,/z)
El (k) -- qn
Pn (k) -
the constraint that the mean value be (x) and the variance be cr 2, is"
(6)
The mean value of k is"
215
E l ( x ) Z [ e x p ( # 2 ) _ 1]
(13b) El(X) 2 ~ o"2, given
(13c)
where X l/2 is the median value. At x -- x~ = xl/zexp(#) we again have P ( x a [ Xl/2, ~ ) -- 0.84 while on x - Xm [= Xl/Zexp(-#)], we have I x l/Z,/x) - - 0 . 1 6 . In this case note that P = 1/2 on Xl/2, but P -~ 0.68 on x = E l ( x ) > A sketch of the cumulative log-normal probability distribution is given in Fig. 5. Empirically, it is often difficult, if not impossible at the beginning of an exploration project, to obtain enough information to determine the precise shape of the frequency distribution of a particular parameter or variable. Indeed, quite often it is considered a fairly good achievement to be able to estimate a likely minimum, Xmin, a likely maximum, Xmax, and a likely most probable value, Xp, for a parameter. A rough idea of relevant mean and variance can then be obtained from Simpson's triangular rule, as sketched in Fig. 5, although other forms of underlying distribution (uniform, Gaussian, log-normal, etc.)can
P(xm
Xl/2.
21 6
S. Cao, A.E. Abbott and L Lerche
also be used to generate a cumulative probability behaviour, as also shown in Fig. 5. For the triangular distribution one has the estimates"
necessarily normally distributed) add to give a sum SN(-- Xl + x2 :t:x3 + . . . + Xu) which is approximately normally distributed as N becomes large, with mean value:
1 E1 (x) ,~ -- g(Xmin + Xp -Jr- Xmax)
E1 (SN) ~ E1 (Xl) -at- E1 (x2) --[-...-a t- E1 (Xu.)
(14a)
(17a)
0.2 __~ 21E1 (X) 2 -- ~1[XminXmax -~- Xp(Xmin -'~ Xmax)]
(14b) (14c)
E 2 ( x ) = E1 (x) 2 -qt- 0-2
If it is further assumed that the variable is log normally distributed, it is possible to work Eqs. (13) in reverse to obtain estimates of tx, x, X l/2 and Xm. Thus from Eq. (13c) we obtain: #--
In 1 +
El(X)2
(15a)
and then, from Eqs. (13a) and (13b), we can estimate:
and with variance" N o-(SN) 2 -- ~ 0-(Xi) 2 i=1
(17b)
Products of parameters. It is well known that multiple independent random variables X, Y, Z . . . . from log-normal distributions combine in generic product form X a Y b Z ~ . . , to give a distribution for the product which is also precisely log-normally distributed with mean value: EI(Xaybzc
. . . ) - EI(Xa)EI(yb)EI(ZC)
...
(18a) Xl/2-El(X)
=-- E1 (x)
exp ( - ~ )
[
1 -t- E1 (x) 2
with second moment:
]
E2(xayazc
< E1 (x)
Xm - Xl/2 exp(-/x) < xl/2
xo -- xl/2 exp
[(( In
1 + El(x) 2
(15c)
))121
(15d)
Multiple parameter distributions In assessments of exploration economic objectives many parameters occur, either alone or in combination with other parameters, and each of the parameters has its own uncertainty. We need to have available practical procedures for estimating the combined effects of uncertainty of parameters on an exploration project. Two sorts of fundamental parameter combinations seem to be prevalent: sums of parameters and products or ratios of parameters. Consider each in turn.
Sums of parameters. It is well known that two or more independent random variables A and B, both with normal distributions, combine to give a sum (A + B) which is also precisely normally distributed with mean value" E l ( A :t: B) -- E1 (A) 4- E1 (B)
(16a)
and variance" cr(A -+- B) 2 -- 0-(A) 2 + o ' ( B ) 2
...
(18b) and scale factor/z given through:
and, from Eq. (13b), we have:
while
. . . ) - E2(Xa)E2(Yb)E2(ZC)
(15b)
(16b)
Empirically it appears that N independent random variables from any frequency distributions (not
[ E2(xaybzc...) /z 2 = In
El(Xaybz c
1
)2
(18c)
Empirically it appears that N independent random variables from any frequency distributions (not necessarily log-normally distributed) tend to combine to produce a product PN (= -- X~ X~X~ . . . X d) which is approximately log-normally distributed with: El(PN)-
EI(X~)EI(X~)...
(19a)
and scale parameter # given through: lZ2~_ln[
E2(PN) ] EI(PN) 2
(19b)
Numerical procedures Empirically, the determination of ].L2, which controls the slope of the cumulative log probability curve at each coordinate, proceeds as follows. Consider, first, that a vector set of parameters, p, are to be used in a model. The output of a specific positive quantity, R, from the model is then available at spatial coordinates x, at time t, and is dependent on the specific values, p, used for the parameters, i.e. R - R(p; x, t). For brevity, throughout this development, the dependence of R on x and t is not written out explicitly, although one must keep in mind that an output quantity is being calculated at each x and t of the model computation. Then write: p - E1 (p) + 6p
(20)
Risk and probability in resource assessment as functions of parameter uncertainty
with E1 (6p) --= 0, where E1 (p) is the average value of the vector p. With given minimum Pmin, most likely Plikely, and maximum Pmax, values for p, the Simpson triangular rule yields: 1
E1 (40) ~" ~ (Pmin -+- Plikely -+ Pmax)
(21)
Taylor series expansion of R yields:
R(p) = R[EI (p) + 6p] "-" R[EI(p)]+ Z 6 p i (OR(P)) + i
Opi
(02R(p)) 3piOPj
1
,
-+- - . .
(22)
-2 Z ( ~ p i 3 P j i * where ( ) . means evaluate the bracketed quantity on p = El (p). It follows that:
EI[R(R)] = R[E1 (p)] 41 ( o2R )
E1 (6p]) + - . .
~
3Pi
(23)
.
It also follows that: E 2 [ R ( p ) ] ~ E1 ( R ( p ) 2) "" E1 {R[E1 (10)]} 2 -/-
ZEl((~p2)(O~pi)2+... i
Then:
-
~2 ~ In
(24)
*
( ER[R(R)])
{
EI[R(R)] 2
In
Opi
i
.§
}
(25)
Now:
,26) ,
Pmax,i -- Pmin,i
(27)
With:
sin20i
_
_
Plikely,i -- Pmin,i Pmax,i -- Pmin,i
(28)
it follows from the Simpson triangular rule applied to 6p, that: E1 (8p2) _ ~8 (Pmax, i _ Pmin,i)2(sin40i + COS20i)
(29) Hence: /z2 ~ In { l + g 1 Z ( s i n 4 0 i
-Ji-COS20i) X
i [ R ( P m a x ) - R(Pmin)] 2 x {R[EI (p)] -'1- R(Pmax) -Jr- R(Pmin)}-e }
(30)
where: 1
E l ( p ) ~ 5(Pmax +Plikely + P m i n )
Thus the two quantities, El(R) and /Z2, which control the cumulative log probability distribution of an output quantity, R, are evaluated at each vector spatial coordinate, x and for each time, t, in terms of computer runs with the parameter vector, p, set to its maximum, minimum and average values, respectively. Accordingly, several strategies are clear for determining parameters which sensitively control an output. For N parameters one can: (a) run all at once; (b) run each independently; (c) run groups of parameters. Cumulative probability outputs can then be evaluated for each case, so that sensitive dependences to a particular parameter and its range of uncertainty, or to a group of parameters, can be evaluated quickly. In terms of minimizing the computer run time for basin analysis models, the mathematical procedure developed above has the following pragmatic operational format: (1) For the impact of each of N input parameters on the behaviour of output results, the basin analysis model is run N times for the N uncertainty ranges of the N parameters. This procedure produces a set of output values of interest. (2) The cumulative probability for each output value is then determined from the N runs of the model. (The cumulative probability can be assessed using any or all of the intrinsic differential distributions exhibited in Fig. 5.) (3) The probability uncertainty range of particular output behaviours can then be assessed, and a risk factor assigned to the likelihood of say, porosity being in the reservoir window, or of excess pressure being above a given value, etc.
Acknowledgments
while: E1 ((~p2) __ E2(Pi) - E1 (pi)2
217
(31)
The work reported here was supported by the Industrial Associates of the Basin Analysis Group at USC. Saga Petroleum is singled out for its special help.
References Atwater, G.I., 1956. Future of Louisiana offshore oil province. AAPG Bull., 40: 2624-2634. Bally, A.W., 1975. A geodynamic scenario for hydrocarbon occurrences. Proc. 9th World Petroleum Congress, Tokyo, Vol. 2, pp. 33-44. Bharucha-Reid, A.T., 1968. Probabilistic Methods in Applied Mathematics. Academic Press, New York, 291 pp. Cao, S. and Lerche, I., 1989. Geohistory, thermal history and hydrocarbon generation history of Navarin Basin COST No. 1 well, Bering Sea, Alaska. J. Pet. Geol., 12: 325-352. Cao, S. and Lerche, I., 1990. Basin modelling: applications of sensitivity analysis. J. Pet. Sci. Eng., 21: 523-542. Cao, S., Tang, J., Liu, J. and Lerche, I., 1993. The Chukchi Sea planning area, Alaska: structural development, basin analysis and hydrocarbon potential. Energ. Explor. Exploit., 11: 235-283.
21 8
S. Cao, A.E. Abbott and L Lerche
Conybeare, C.E.B., 1965. Hydrocarbon generation potential and hydrocarbon yield capacity of sedimentary basin. Bull. Can. Pet. Geol., 13: 509-528. Dutta, N.C., 1989. Fluid flow in low permeable porous media. In: B. Doligez (Editor), Migration of Hydrocarbons in Sedimentary Basins. Editions Technip, Paris, pp. 567-596. Feller, W., 1960. An Introduction to Probability Theory and its Applications, Vol. 1. John Wiley and Sons, New York, 461 pp. Hunt, J.M., 1979. Petroleum Geochemistry and Geology. W.H. Freeman and Company, San Francisco, Calif., 617 pp. Jones, R.W., 1975. A quantitative geologic approach to prediction of petroleum resources. AAPG, Stud. Geol., 1: 186-195. Lerche, I., 1989. Basin Analysis: Quantitative Methods, Vol. 1. Academic Press, San Diego, Calif., 562 pp. Lerche, I., 1990. Basin Analysis: Quantitative Methods, Vol. 2. Academic Press, San Diego, Calif., 570 pp. Lerche, I., 1992. Oil Exploration: Basin Analysis and Economics. Academic Press, New York, 178 pp. Lerche, I., 1993. A probabilistic procedure to assess the uncertainty of fractal dimensions from measurements. Pure Appl. Geophys., 140:503-517. Lumley, J.L., 1970. Stochastic Tools in Turbulence. Academic Press, New York, 194 pp. Magoon, L.B. and Bird, K.J., 1986. Organic carbon content, hydrocarbon content, visual kerogen and vitrinite reflectance data within NPRA (National Petroleum Reserve in Alaska) - - contour maps to evaluate petroleum source rock richness, type, and thermal maturity. In: G. Gryc (Editor), Geology of the National Petroleum Reserve in Alaska. U.S. Geol. Surv. Prof. Pap. 139. McDowell, A.N., 1975. What are the problems in estimating the oil potential of a basin? Oil Gas J., June 9, pp. 85-90. Miller, B.M. et al., 1975. Geological estimation of undiscovered recoverable oil gas resources in the United States. U.S. Geol. Surv. Circ., 725: 78. Newendorp, ED., 1975. Decision Analysis for Petroleum Exploration. Petroleum Publ. Co., Tulsa, Okla., 345 pp. Pitcher, M.G., 1976. U.S. discovery rate tied to technology. Oil Gas J., March 22, pp. 34-35. Roadifer, R., 1975. A probability approach to estimate volumes of undiscovered oil and gas. In: J.C. Davis (Editor), Probability Methods in Oil Exploration. American Association of Petroleum Geologists Research Symposium, Stanford University, p. 18.
S. CAO A.E. ABBOTT I. LERCHE
Roy, K.J., Procter, R.M. and McCrossan, R.C., 1975. Hydrocarbon assessment using subjective probability. In: J.C. Davis (Editor), Probability Methods in Oil Exploration. American Association of Petroleum Geologists Research Symposium, Stanford University, pp. 56-60. Smith, M.B., 1968. Estimating resources by using computer simulation methods. Oil Gas J., March 11, pp. 81-84. Smith, EJ. and Buckee, J.W., 1985. Calculating in-place and recoverable hydrocarbons: a comparison of alternative methods. SPE 13776. Stoian, E., 1965. Fundamentals and applications of the Monte Carlo method. J. Can. Pet. Technol., 4:120-129. Tissot, B.E and Welte, D.H., 1984. Petroleum Formation and Occurrence, 2nd ed. Springer-Verlag, New York, 699 pp. Turner, R.E, McCarthy, C.M., Seffy, D.H., Lynch, M.B., Martin, G.C., Sherwood, K.W., Flett, T.O. and Adams, A.J., 1984. Geological and Operational Summary, Navarin Basin COST No. 1 Well, Bering Sea, Alaska. OCS Report MMS 84-0031. US Department of the Interior Mineral Management Service, Anchorage. Walstrom, J.E., Mueller, T.D. and McFarlane, R.C., 1967. Evaluating uncertainty in engineering calculation. J. Pet. Technol., 19: 15951603. Warren, J.E., 1979. Basin evaluation. Society of Petroleum Engineers Economics and Evaluation Symposium, Dallas, February, 1979. Warren, J.E., 1981. The Development Decision: Frontier Areas. SPE 9558. Warren, J.E., 1983a. A Strategic Exploration Model. SPE 11442. Warren, J.E., 1983b. The Development Decision: Value of Information. SPE 11312. Warren, J.E., 1988. The Evaluation of Oil and Gas Plays. SPE 17554. Weeks, L.G., 1952. Factors of sedimentary basin development that control oil occurrence. AAPG Bull., 36: 2071-2124. Withers, R.J., 1992. The value of reservoir geophysics. Geophysics, The Leading Edge of Exploration, SEG Publication, Tulsa, OK, pp. 35-39. Wyllie, M.J.R., 1963. Fundamentals of Well Log Interpretation. Academic Press, New York, 293 pp. Yukler, M.A. and Kokesh, F., 1984. An overview of models used in petroleum resource estimation and organic geochemistry. In: J. Brooks and D. Welte (Editors), Advances in Petroleum Geochemistry, Vol. 1. Academic Press, New York, pp. 69-73.
Department of Geological Sciences, University of South Carolina, Columbia, SC 29208, USA Department of Geological Sciences, University of South Carolina, Columbia, SC 29208, USA Department of Geological Sciences, University of South Carolina, Columbia, SC 29208, USA
219
Risk assessment using volumetrics from secondary migration modelling: assessing uncertainties in source rock yields and trapped hydrocarbons Wenche Krokstad and Qyvind Sylta
Modelling of secondary hydrocarbon migration is useful in risk assessment because it enables the quantities of oil and gas in undrilled prospects to be calculated. Monte-Carlo simulation techniques can be used together with secondary migration modelling to assess the oil and gas in-place resource distributions. Probability maps show the relationship between trapped oil and trapped gas in prospects. Many simulation runs will not match the known oil and gas fields, but calibration to discoveries can be made by choosing appropriate weighting functions. Weighting as a function of distance between prospect and known hydrocarbon accumulations, migration pathway distances and positions relative to source rock kitchens can yield different distributions of trapped oil and gas. The type and quantity of organic matter in the source rocks vary in a sedimentary basin. Hydrocarbon yield uncertainty maps can be constructed from quantitative three-dimensional source rock models. These maps can be used to make higher frequency source rock hydrocarbon yield maps, which reflect a spatial variation that cannot be appropriately determined from well data because of insufficient well coverage in the kitchen areas. A yield anomaly is here characterized by its height, length, width and orientation. The values for each of these variables are chosen from statistical distributions based on the yield uncertainty maps. Using Monte-Carlo techniques and secondary migration modelling, each yield map determines the amount and type of hydrocarbons within the undrilled traps. When a sufficient number of computer simulations have been run, a source rock yield uncertainty distribution of predicted amounts of trapped oil and gas can be assembled. These distributions are characteristic of each trap and drainage area, as illustrated by modelling results from Norwegian block 35/4. The procedures presented here can therefore provide important constraints in the risking of undiscovered resources.
Introduction Probabilistic approaches are widely used to assess exploration uncertainties and risks (Megill, 1992; Rose, 1992). A thorough review of the techniques and their application to petroleum exploration is found in Steinmetz (1992). For many years oil companies have used Monte-Carlo simulation in the risking of new prospects (Capen, 1992). The advantages of the MonteCarlo simulation techniques include ease of use, i.e. by means of existing software programs, and the speed of use. It only takes a few minutes to run a Monte-Carlo simulation program once all the input data have been created. Basin modelling has been used as one of several techniques to constrain input variables in risking software, e.g. assess the probability of mature source rocks, probability of oil and gas expulsion etc. So far, basin modelling results have not been used to their full potential in the risking process. Because modelling is quantitative and uses computers, it should be possible to transform results from modelling into better, i.e. less subjective, risking of undrilled prospects.
In his preface to the conference proceedings from the NPF conference "Basin Modelling: Advances and Applications" (Dor6 et al., 1993), Dor6 stated that "Irwin et al.'s work on probabilistic modelling of hydrocarbon charge in the Egersund basin signals a direction many of us believe basin modelling must eventually take." The work which is reported here further advances this integration of basin modelling techniques with Monte-Carlo simulation techniques to provide a better risk assessment. Irwin et al. (1993) focused mainly on thermal modelling of the oil and gas generation history of important source rocks in a fairly well defined catchment area. They applied a simpler approach to the modelling of hydrocarbon migration. In our approach, a more comprehensive migration modelling is carried out for an area substantially larger than the catchment area of a single prospect. A more extensive calibration of the model to existing oil and gas fields has therefore been achieved. This calibration is subsequently used in the weighting of Monte-Carlo simulation results to arrive at a better risk assessment of the individual prospects.
Quantification and Prediction of Petroleum Resources edited by A.G. Dor6 and R. Sinding-Larsen. NPF Special Publication 6, pp. 219-235, Elsevier, Amsterdam. 9 Norwegian Petroleum Society (NPF), 1996.
220
Schroeder and Sylta (1993) showed how a secondary migration model could be used to construct a risk matrix for generation and migration. This risk matrix could then be combined with other risk factors, e.g. reservoir rock porosity, top seal leakage etc., to compute the final prospect evaluation. A deficiency of the risk matrix approach is that the evaluation of risks associated with different system uncertainties, like source rock quality, reservoir thicknesses and expulsion efficiencies, is not at all simple when only pluses and minuses are used (see table 3 in Schroeder and Sylta, 1993). Frequently, the uncertainties from some risk factors will be much more important than others. Hence, there is a need of applying the migration model in a more quantitative manner in the risk assessment procedure. In this paper we use the results from a migration modelling study performed at IKU in 1992. The study was based on work reported in Sylta (1993), and utilized the SEMI secondary migration program to model secondary migration in the study area shown in Fig. 1. The results we present here are only included to show the principles involved as the intention of this paper is not to discuss the case study in any detail. We therefore only document data required to understand the methodology and the objectives of the methods employed.
Methodology Migration modelling and Monte-Carlo simulation The Viking Group of the North Viking Graben includes the very rich Draupne source rocks (Fig. 1). The Draupne Formation is separated from the Brent carrier beds by the leaner Heather Formation. Heather Formation thickens towards the basin centre, and may therefore cause expulsion from the Draupne to be less efficient in the deeper parts of the Viking and Sogn Grabens (Fig. 2). The Lower Jurassic Dunlin Group is considered to be a less efficient source for oil and gas. Secondary migration was therefore modelled using ray-tracing (Sylta, 1991, 1993) within a single carrier bed and immediately below a laterally extensive cap rock (the Viking Group). The structure map, the regionally extensive "Top Middle Jurassic", which represents the top of the carrier and reservoir rocks, was decompacted to provide paleomigration directions (Fig. 2). The grid resolution used was 750 by 750 m, which gives a satisfactory resolution for regional studies. The migration model quantifies oil losses during migration, cracking of oil to gas within traps, leakage of trapped gas through the cap rock, and fill and spill of trapped oil and gas (Sylta, 1991, 1993).
w. Krokstad and O. Sylta
A regional best fit to seven calibration fields (Visund North and South, Agat, 35/8-1, 35/8-2, 35/9-1 and the Veslefrikk Field; Fig. 2) was obtained using a Monte-Carlo simulation approach and varying five important input parameters (see later). The best calibration resulted in an average deviation between measured and modelled trapped oil and gas of 6 (106 Rm 3 oil and 109 Sm 3 gas). The secondary migration simulator used in this work is deterministic, whereas the results we intend to construct are probabilistic (e.g. statistical distributions). A simple way to obtain statistical results from a deterministic model is to put the deterministic model inside a Monte-Carlo simulation procedure. This requires that one is able to adjust the variables of the deterministic model for each randomly selected input parameter value specified by the Monte-Carlo simulation procedure. The approach used in this paper is shown in Fig. 3. It starts with an established hydrocarbon generation model for the area, represented as a series of digital oil and gas generation maps of various source rock units, in this case the Viking and Dunlin Groups (Fig. 1). Hydrocarbon discoveries have been made in the Brent Group (Fig. 1) and in Upper Jurassic sands to the east in the study area, e.g. the Troll Field. The discoveries have usually been made in the uppermost sandy units of the Jurassic, and occasionally in the Lower Cretaceous (the Agat Field). The Jurassic and Lower Cretaceous reservoirs can be considered to be one single migration system. For each simulation run, the main input parameters are selected randomly and independently. One may wish to adjust a large number of input parameters in such a procedure, but in this study we chose to vary only five input variables. These input variables were Upper Jurassic oil and gas expulsion factors in a thickness-dependent expulsion model, Lower Jurassic oil and gas expulsion efficiencies (in percents of generated oil and gas) and vertical gas leakage from traps. These five variables were considered to be the least constrained at the time of the study. For each migration simulation run, results for all traps (calibration fields and prospects) are stored. The results include trapped oil, solution gas and free gas. After storing the results, the procedure is repeated as shown below and in Fig. 3.: - select new input values randomly; - perform a migration simulation run; - store results. This iterative procedure is repeated until a sufficient number of simulations has been run. The probability distributions of trapped oil, solution gas and free gas can thereafter be established from the stored simulation results.
Risk assessment using volumetrics from secondary migration modelling
Weightingprocedure An important aspect of our procedure is the weighting of simulation results according to how well they match the discoveries in the study area (Fig. 3). The choice of weighting procedure and the correct weights is crucial to the reliability of the results from the method described here. The weighting can be based on trap sizes, distances between prospect and oil and gas fields, migration distances between prospect and fields, and the overall error (misfit) between modelled and observed trapped oil and gas in the calibration fields. Among these, the
221
most important weighting factor is the misfit to the calibration fields. In this study the misfit is defined as" 1 X---" [{N mod gobs) 2 misfit 2 = 2 N Z.. WiLk Voi . / + il
( mod _ "gi
gi ,]
2]
(1)
where V~ is observed oil in place of field i ( 10 6 Rm 3), Vom~ is modelled oil in place of field i (10 6 Rm3), Vgt~ is observed gas in place of field i (10 9 Sm3), Vgm~ is modelled gas in place of field i (10 9 Sm3), W/is weight of field i, and N is number of calibration fields.
222
W. Krokstad and 0. Sylta
Fig. 2. Decompacted top Middle Jurassic carrier bed at 94 Ma with calibration fields/discoveries and prospect 35/4 used in the study. A = line showing distance between prospect and calibration fields. B = "migration distance" between prospect and calibration fields, a-a' = distance between Veslefrikk and prospect catchment areas.
Risk assessment using volumetrics from secondary migration modelling
Hydrocarbon generation modelling input t o ~ igration ~ L
s Random:
sion -l-expul eakage
1 -expulsion anomalies
migration
modelling
trapped oil and gas compute misfit
determine weight
compile probability distribution
Fig. 3. Flow-chart for Monte-Carlo simulation of secondary migration.
Here the V-terms reflect the inability of the simulator to compute correctly the volumes of oil and gas in the calibration fields. The W-term contains the (manually) determined weight associated with each calibration field. If W is set to 1 for all fields, no calibration field weighting is performed and the misfit is considered to represent a "global misfit". This global misfit describes how accurate a simulation run (with one set of values for the input parameters) matches the calibration fields, i.e. the average error in the modelled trapped oil and gas volumes. Eq. (1) combines oil and gas terms. We use scaling factors of 1.106 Rm 3 for the oil and 1-109 Sm 3 for the gas in this study. The scaling of the gas and oil terms in Eq. (1) determines the most important phase in the matching of the calibration fields. The gas term in Eq. (1) can represent free gas or total gas. We have used total gas because the free gas volume will be zero for all simulations that cause traps to be undersaturated. The misfit function can then become insensitive to changes in the input parameters. Another approach could be to expand the equation with a free gas as well as a solution gas term. We have not yet enough experience with the methodology to decide which approach that produces the most reliable answers.
223
The V-terms in Eq. (1) are computed directly from the simulation results, whereas the values of the Wterms have to be determined by the geologist. A number of different geological constraints can be included in this latter weighting factor. Thus, the relative importance of each calibration field can be quantified in the W-term. A simple weighting approach is that of weighting by distance (see straight lines in Fig. 2). The weights resulting from using distance weighting in combination with capacity weighting are listed in Table 1 (cap/dist). Note in particular that Visund N has achieved a weight of 21.9%, even though it is located on the opposite side of the Viking Graben and may receive oil and gas from areas that do not source the 35/4 trap. We therefore conclude that weighting by distance, although being simple, does not yield sufficiently good weighting of the calibration fields. However, using the migration distances between the prospect and the calibration fields results in a more realistic weighting. The migration distance between two fields is here defined as the shortest distance between the two traps along migration paths. This principle is illustrated in Fig. 2, where the curved lines between trap 35/4 and the calibration fields represent migration distances for trap 35/4. The Veslefrikk Field (Fig. 2) is not adjacent to the 35/4 drainage area. Therefore we suggest the use of a straight line between the two drainage areas to define the complete migration distance (see a-a' in Fig. 2). Weights computed from migration distances (Fig. 2) are listed in Table 1. The weights are defined as:
Wi-
Oi
O-[-1
(2)
i=1
where Wi is weight for calibration field i, Di is migration distance for calibration field i, and N is number of calibration fields. When trap capacities are included, Eq. (2) can be expanded into:
[
W/- - CAP/
CAP/ -~i
Di
]_l
(3)
i=1
where CAP/ is the capacity (hydrocarbon porevolume) of calibration field i, and the other parameters are as defined in Eq. (2). The incorporation of migration distances in the weighting of the Monte-Carlo simulations causes a reduced importance of the Visund Field, whereas the importance of matching the 35/8 discoveries (Table 1) is enhanced. The Veslefrikk Field is of relatively little importance (5%), whichever method is used. This may not seem representative of the location of the
W. Krokstad and 0. Sylta
224
Table 1 Hydrocarbon resource data and weights (%) applied to the calibration fields Field
Agat 35/8-1 35/8-1 35/9-1 Visund N Visund S Veslefrikk
Capacity (106 Rm 3)
Trapped oil ( 106 Rm 3)
gas (10 9 Sm 3)
0. 0. 0. 12.5 30. 10. 90.
61. 19. 10. 15. 43. 20. 3.
Weights (%)
245. 77. 40. 60. 173. 81. 90.
cap/dist a
Cap/D
25.2 22.0 7.7 8.8 21.9 8.3 6.1
27.7 31.0 10.9 6.4 13.3 5.6 5.1
b
Dc 10.3 36.7 24.8 9.7 7.0 6.3 5.2
a cap/dist -- weights from capacity/distance in Eq. (3) with trap distance instead of migration distance. b cap/D -- weights from trap capacity/migration distance, Eq. (3). c D = weights from the use of migration distance in Eq. (2).
150-
m o
=" m ,,.B
lOO-
E m o x.__
e~
E
:3 Z
50-
0
I
i
I
i
!
I
1
I
I
10
20
30
40 Misfit
50
60
70
80
9O
Fig. 4. Number of simulations versus global misfit. A total of 1700 simulation runs are included in the histogram.
field in the basin (see Fig. 2), because the structural location of Veslefrikk is in some respects quite similar to the 35/4 prospect. They are both very close to the deeper parts of the Viking Graben. As such, it could be correct to remove the a-a' part of the migration distance (Fig. 2) from the weighting of the Veslefrikk Field. However, trap 35/4 also receives oil and gas from the Sogn Graben to the northeast, and allowing Veslefrikk to be too important may therefore not be a very good approach. The fact that the 35/8 discoveries are given high weights is reasonable, since 35/4 spills into these structures. In conclusion, we decided to use weights incorporating trap capacities and migration distances in the construction of the 35/4 probability distributions. The numbers used (cap/D in Table 1) are thus obtained by relatively objective means, i.e. independent analysis by different geologists should
yield very similar numbers when this approach is used. Fig. 4 shows the global misfit (no weighting) distribution for 1700 simulations. The results cluster around some of the misfit values. More than 150 simulations yield global misfit values around 27, whereas only very few simulations result in global misfit values less than 9. The relative importance of each of the "misfit equals 27" simulations will be 100 times less than the low misfit simulations since the square root of the misfit is used in the compilation of the probability distributions. Because of the much larger number of simulations causing higher misfits, more emphasis will be placed on the high misfit simulations if they show some coherency in their results, i.e. if they all give similar estimates of oil and gas in the prospect. In order to prevent any
Risk assessment using volumetrics from secondary migration modelling
225
such effects, the simulation results are weighted with the inverse of the number of simulations. The group of simulations which all have a misfit value of 27 represents a broad range of input parameter values, and the likelihood for producing very similar output values from these simulations is therefore not high. The construction of the probability distribution now becomes a two-pass operation. First the data in Fig. 4 are tabulated and misfit range weights are computed. In our procedure we have divided the total range of global misfit into 20 groups. Each group is given a misfit range weight. This weight is then used in the final compilation of the probability distributions.
HEIGHT WIDTH
Quantifying source rock yield uncertainties Viking Graben source rocks have generated appreciable volumes of oil and gas. These generated hydrocarbons have been expelled into the highly permeable Jurassic carrier and reservoir rocks. Several procedures have been proposed to model this expulsion of oil and gas from the Upper Jurassic Viking Group and Middle and Lower Jurassic shales. In previous migration modelling work several models have been used, including simple expulsion efficiencies (Sylta, 1993) and saturation dependent expulsion (Skjerv0y and Sylta, 1993). The thickness of the Upper Jurassic rocks increases from less than 100 m along the margins of the North Viking Graben to in excess of 1 km in the deep grabens. Expulsion from the Draupne Formation into the B rent Group carrier bed may therefore be less efficient in the basin centre than along the flanks of the basin, where the Draupne Formation is separated from the Brent by only a thin Heather unit. In this study a thickness dependent expulsion model was therefore chosen to account for lateral changes in expulsion. Other expulsion models proposed and used by the industry include diffusion in a kerogen network (Stainforth and Reinders, 1990) and two-phase forced Darcy fluid flow (Ungerer et al., 1987). Hermanrud (1993) and Lerche (1993) discuss alternative expulsion models, without concluding definitely as to which model is correct. The numerous publications incorporating different expulsion models suggest that many of the expulsion models proposed can indeed be adjusted to provide a reasonable match to calibration data. An improvement of the input parameters can be achieved by calibration of a secondary migration model to known fields. The regional trends of the generated and expelled oil and gas may then become better constrained. Lateral changes in the geological system are, however, the rule rather than the
LENGTH ANOMALY DIRECTION
Fig. 5. Source rock yield anomaly geometry. The anomaly is defined by centre height, length, width, location [h, a, b, x, y in Eq. (4)] and anomaly direction.
exception. These changes have so far not been well accounted for in prospect risk assessments. Input variables that account for lateral changes in source rock quality, expulsion efficiencies, reservoir permeabilities etc., provide for a quantification of the resulting risks for a prospect. In order to achieve a quantification of the lateral source rock variability, a "source-rock yield anomaly" (SYA) can be used as a building block. Such an anomaly will consist of a basic geometric shape and associated parameters. A simple and very flexible geometry is shown in Fig. 5 and Eq. (4): f ( x , y) = (
h /1 + cos
IYr ( ( X -
X0) 2 +
(Y-Y~ b2
_~
(y-
y0) 2
Ik
and: (x - xo) 2
a2
+
< 1
(4)
where x is axis of anomaly at an angle to the westeast direction (Fig. 5), y is axis normal to x, a is elliptical length of anomaly in the x-direction, b is elliptical length of anomaly in the y-direction, x0 is x, location of centre of the anomaly, Y0 is y, location of centre of the anomaly, and h is height of the anomaly at the centre. One of the advantages of Eq. (4) is that the derivative is zero at the boundary, as is the function value. It is therefore possible to stack a series
W. Krokstad and O. Sylta
226
of partly overlapping anomalies into a composite anomaly map (Sylta, 1991). This latter map can then be combined with oil and gas expulsion maps to make a basin-wide hydrocarbon expulsion anomaly scenario. The anomaly scenario can be used as input to the secondary migration modelling, and the resulting changes to the trapped oil and gas can thus be deduced (Fig. 3). Assigning values to the SYA can be achieved by means of a Monte-Carlo simulation approach. Input values are the location, the direction and the size of the anomalies. In the North Viking Graben area, the uncertainty of the source rock distribution, e.g. thicknesses, concentration of total organic carbon (TOC) and kerogen type, are considered to be large in the basinal parts, where only a few wells have been drilled. The uncertainty in the vicinity of exploration wells is significantly less, but not zero. A relative source rock anomaly location probability map for the study area is shown in Fig. 6. This map was constructed from geological reasoning. High values imply that the probability of finding a source rock anomaly is higher. A value of 1 gives twice the probability of an anomaly value of 0.5. The conceptual source rock model used in this study suggests a roughly concentric pattern, and is also characterised by an increase in the total organic carbon (TOC) content and hydrogen index towards the basin depocenters (Huc, 1988; Thomas et al., 1984). Therefore, the source rock anomalies are more likely to parallel the basin axis of the Viking Graben. The direction angles of anomalies are intrinsically uncertain, and we select the angles within -+-20 degrees from the anomaly direction map (Fig. 7 and Table 2). Each anomaly scenario modelled in this study is computed from the variable ranges listed in Table 2. There is no reason to define any value within the ranges listed in Table 2 as more likely than others, and therefore a rectangular probability distribution is used for all parameters, except for those that receive values according to the maps, as discussed above. Parts of an anomaly scenario map may then look like Fig. 8a, whereas another realisation is shown in Fig. 8b.
Table 2 Range of input variables defining source-rock anomalies (SYA) Minimum # anomalies Location: map Length Width Height Direction a
Maximum
1
10
20 10 0.25 -20
40 20 4 20
a Added to anomaly direction map (Fig. 7)
Unit
km km fraction degrees
Fig. 9a shows an oil expulsion map, without source rock anomalies, which was used to construct the oil flow-rate distribution shown in Fig. 10. The oil anomaly map (Fig. 8a), which was constructed by Monte-Carlo simulations of the parameters in Table 2, was multiplied with the oil expulsion map (Fig. 9a) to produce the oil expulsion anomaly map (Fig. 9b). Secondary migration with Fig. 9b as input, caused changes to the oil flow-rates (Fig. 10) shown in the lower right of Fig. 10. Each of the anomaly maps is multiplied by oil and/or gas expulsion maps in all the time steps of a simulation (Fig. 3).
Results The number of simulations required to establish correct probability distributions determines whether or not the method proposed here is feasible to use in exploration. If too many simulations are required, it will not be possible to run enough simulations within the short time frames of most exploration projects. Fig. 11 shows compiled probability distributions of oil charge for the same input parameters, but with the number of simulations varying from 160 to 1600. A minimum of 1000 simulations were in this case required to achieve a reasonable correct distribution. The distributions are too uneven when less than 1000 simulations are used, whereas they become smooth and less subject to change when more simulations have been included. The number of simulations versus the global misfit (to the right in Fig. 11) suggests that the sampling is too uneven when less than 1000 simulations are used. A total of 1500 Monte-Carlo simulation runs should in this case be sufficient to guarantee high reliability in the risking process. Fig. 11 (left) shows a standard type of probability distribution which can be constructed for trapped oil as well as trapped gas, i.e. for a single variable. "Probability maps" for two parameters can also be constructed. Fig. 12a shows that if the 35/4 prospect contains 3 5 0 . 1 0 6 Rm 3 of oil, it will contain at least 30.109 S m 3 of free gas, and most likely the trap will be filled to spill point, i.e. the gas cap will contain 100.109 Sm 3 of gas. Fig. 12a also clearly indicates that there is a 27% probability of finding 5 5 0 . 1 0 6 rn3 of oil with 20.109 S m 3 gas in the gas cap. The modelling results suggest that there is no risk for a dry trap, provided that the model concepts used are valid for the study area. The probability maps in Figs. 12b-d show the uncertainties related to lateral changes in the source rock yields. Three scenarios have been tested, as follows: (1) oil expulsion anomalies (Fig. 12b); (2) gas expulsion anomalies (Fig. 12c); (3) oil and gas expulsion anomalies (Fig. 12d).
Risk assessment using volumetrics from secondary migration modelling
227
Fig. 6. Map of source rock yield anomaly location distribution. Values show relative probability of having the centre of an anomaly located in that particular position. See Fig. 2 for explanation of lines.
228
W. Krokstad and 0. Sylta
Fig. 7. Source rock yield anomaly direction map. Values show most likely angle of deviation of each anomaly. North is 90 degrees, and lower values show directions to the northeast (see arrows). See Fig. 2 for explanation of lines.
Risk assessment using volumetrics from secondary migration modelling
229
Fig. 9. Oil expulsion before 94 Ma without (a) and with (b) multiplication by source rock yield anomaly in Fig. 8a. The area is the same as in Fig. 8.
Oil anomalies may be due to increases in type II kerogens or a result of more efficient expulsion of oil from the source rocks due to good vertical communication with the carrier rock system. Gas anomalies may be caused by local increases in gasprone source rocks, by more cracking of oil generated within existing source rocks due to less efficient expulsion of the heavier components. Oil and gas anomalies typically represent a change (increase or decrease) of the source rock richness. Figs. 12b-d depict the resulting oil, gas and oilgas anomaly distributions of trapped hydrocarbons in prospect 35/4. In these simulations the minimum global misfit scenario (most likely value in Fig. 12a) was used as the base case, i.e. the five input variables to migration were kept constant. Thus, we were able to assess how much more uncertainty resulted from anticipated lateral source rock variations.
The composite probability maps (Fig. 12) enable a direct comparison of results from the three anomaly types. The gas anomalies follow a straight line that results when a trap is modelled to be filled to spill point in all low-misfit simulations. In contrast, the oil anomalies cause prospect 35/4 to be not filled to spill point if less than 400.106 Rm 3 of oil is trapped. The oil-gas anomalies, which represent uncertainties in source rock richness, produce a very narrow risk distribution. The 35/4 prospect is therefore not very sensitive to local variations in source rock richness.
Discussion and conclusions We have shown that a Monte-Carlo simulation approach to secondary migration modelling is feasible and can produce probability distributions that can
230
W. Krokstad and ~). Sylta
Fig. 10. Oil flow rate without use of source rock anomalies and with (insert lower right) source rock anomalies. Colour values are lOgl0 (m 3 km -2 m.y.-1), except for trapped oil (dark green ) and trapped free gas (dark brown). Note the increased oil flow-rates in areas with source rock yield anomalies (Fig. 8a). Insert contains the same area as in Figs. 8 and 9.
Risk assessment using volumetrics from secondary migration modelling
231
Fig. 11. Probability distributions for oil charge to prospect 35/4 (left) and number of simulations versus global misfit (right) for 160, 385, 1000 and 1600 simulations. The 160 simulation runs are included in the 385 simulation run plot etc.
Fig. 12. Trapped oil and free gas probability distributions for prospect 35/4. Figure also shows trapped oil and free gas probability map for 35/4 (probabilities in%). Twenty bars in each probability distribution. The four scenarios plotted are total 35/4 probabilities varying 5 input variables (a), oil expulsion anomalies (b), gas expulsion anomalies (c), and oil-gas expulsion anomalies (d). See text for explanation of anomalies.
Risk assessment using volumetrics from secondary migration modelling
be used directly in the exploration risk assessment. The method elucidates the interdependence of oil and gas trapped within a prospect, thereby increasing the consistency in exploration decisions and ultimately reducing the uncertainty involved in selecting the right trap as a drilling target. The probability distributions produced in this study tend to contain several maxima. Some of the distributions almost seem to be constructed by adding at least two independent distributions together (e.g. Fig. 12a). The classical normal and log-normal distribution shapes (Capen, 1992) do not seem to fit very well in this case, because the distributions decrease very rapidly and exponentially from the maxima. In the usual risking process, a number of independent distributions are convolved into a resulting probability distribution which takes the form of a normal or log-normal distribution (generally log-normal). In the secondary migration Monte-Carlo simulation approach, the probability distributions of the input variables are no longer independent of each other. The weighting of the simulation results by the misfit value of each simulation run may cause a strong linking of the distributions, and therefore the "rule of thumb" of a resulting log-normal distribution for the aggregated prospect risk can no longer be assumed to be correct. The work presented here therefore suggests that more complex distributions may result when secondary migration modelling is included in the risking process. Some very important variables in the risking process have not been varied in this study. Possibly the most important of these is the size of the "container" in which the hydrocarbons are trapped. Thus, the hydrocarbon pore volume of the 35/4 prospect was maintained at 680.10 6 m 3 in all the simulations. This may partly explain why the resulting probability distributions show a "reverse log-normal" shape rather than the more frequent log-normal shape. It is evident that attempts have to be made to incorporate variations in the trap sizes in future applications of the described methodology. One approach may be to use the "SYAs" to construct lateral changes in the porosity and net/gross grids which are used by the secondary migration software. If such a procedure is used, it should be calibrated to existing trap size probability distributions in known areas. A second approach is to apply a probabilistic description of the (palaeo) depth maps. This approach would, however, require more research before it can be applied to real cases. The migration model used in this study used 20 min to run a single simulation on a Sparc-II workstation. The number of grid nodes in each map totalled 40,000 grid nodes. Hence 1500 simulations required
233
20 days to complete. Four Monte-Carlo sequences were invoked in this case study (Fig. 12), and if only one workstation had been available, the simulations would have run for almost three months. This is obviously not feasible in most exploration projects. In order to be suitable for hydrocarbon exploration, the method should take no more than a week to run, and the simulations must therefore be run almost 10 times faster than in our study. The new generation of hardware (1994) workstations will be at least five times faster than the Sparc-II, and the method which is described in this paper can therefore be applied if one can have full access to two such workstations for at least one week. One may anticipate that more high-resolution maps will be constructed from larger seismic databases at the same time as computers become faster. The use of 3D data in seismic interpretations, sometimes covering several North Sea blocks, often results in accurate depth-converted maps at a grid spacing in the range of 100 by 100 m. If migration modelling is to be performed with such grids, the CPU requirements may increase more than the improvements in computer hardware, and parallel computing techniques may have to be used. Two approaches may be suggested to reduce the CPU requirements. First, the incorporation of inversion techniques in the Monte-Carlo flow-chart (the Random box in Fig. 3) may help to increase the number of low misfit simulations. This could reduce the number of simulations required to construct the probability distributions. Secondly, a statistical analysis may be applied to a limited number of simulations. Such analysis may be used to construct the complete response function of the migration model to changes in the input variables (E. Loomis, pers. commun.). The response function may then be used in the final risk assessment, and thus reduce the required time by at least an order of magnitude. Including the description of lateral changes in the expelled oil and gas from source rocks results in important new input data for prospect risking with respect to hydrocarbon generation and migration. In particular, the study of the probability distribution resulting from the SYA analysis will give new insight into which prospects are more sensitive to lateral geological uncertainties. The anomaly sizes used in the SYA may, however, influence the resulting probability distributions. It is not realistic to use smaller anomalies than the ones listed in Table 2, because very small anomalies will only influence one catchment area. Anomalies that are located entirely within the Visund catchment area will change the modelled oil and gas trapped in Visund without changing the trapped hydrocarbons in 35/4. If too many such anomalies
W. Krokstad and 0. Sylta
234
150-
100-
Large Anomalies
500
i i
i
E
m m
0
0
300-
E
250-
z
0
IIIillllll[lllllll,tlll,,,,llllllll,l,,t,,,,,, ..........,........... illltl,,,,,,,, !
I
I
I
I
I
!
I
5
10
15
20
25
30
35
40
200150-
100-
50- I il !ilitl l,t,l,r"~ !, II,,I,l,,,..,
0
o
5
10
........ l . . . . . . . . .
15
2O
,-
25
, ...... 50
I!, I
i
35
40
Misfit Fig. 13. The effect of source rock yield anomaly sizes. Large sizes (above) are twice as large (width and length) as the normal size anomalies (below). Axis values represent number of simulation runs (y) versus global misfit (x, see also text). A total of 1500 runs are included in each histogram.
are included in the analysis, the results will include significant noise and the geological interpretation will be less confident. A comparison of the SYA method using input from Table 2 and with larger anomalies is shown in Fig. 13. The figure shows that the most significant result of increasing the anomaly sizes is to increase the number of large misfit simulations, and to reduce the number of low-misfit simulations. Results from the larger anomalies therefore were considered less reliable in the case described.
The approach described in this paper requires information that is specified at an early stage in the ranking process, e.g. weights for calibration fields, the source rock model, prospect mapping. An advantage of using secondary migration modelling together with Monte-Carlo simulation techniques in a multiprospect area is that one series of simulations will enable the construction of probability distributions for many prospects. These prospects can therefore be ranked in a fairly objective manner, i.e. without too much independent interpretation of the different traps.
Risk assessment using volumetrics from secondary migration modelling
Acknowledgements The structural time maps used in this work were provided by Nopec a.s., and are greatly appreciated. We thank IKU and Nopec a.s. for permission to publish this paper.
References Capen, E., 1992. Dealing with exploration uncertainties. In: Richard Steinmetz (Editor), The Business of Petroleum Exploration, Treatise of Petroleum Geology, Handbook of Petroleum Geology. AAPG, Tulsa, Okla., pp. 29-61. Dor6, A.G., Augustson, J.H., Hermanrud, C., Stewart, D.J. and Sylta, 0. (Editors), 1993. Basin Modelling: Advances and Applications. NPF, Special Publication 3, Elsevier, Amsterdam. Herbin, J.P., Geyssant, J.R., M61i6res, E, Muller, C. and Penn, I.E., 1991. H6t6rog6n6itEe quantitative et qualitative de la mati6re organique dans les argiles du Kimmeridgian du val de Pickering (Yorkshire, UK). Cadre s6dimentologique et stratigraphique. In: Revue de l'Institut Frangais du P6trole, 46(6): 675-712. Hermanrud, C., 1993. Basin modelling techniques m an overview. In: A.G. Dor6, J.H. Augustson, C. Hermanrud, D.J. Stewart and 0. Sylta (Editors), Basin Modelling: Advances and Applications. NPF, Special Publication 3, Elsevier, Amsterdam, pp. 1-34. Huc, A.Y., 1988. Aspects of depositional processes of organic matter in sedimentary basins. In: L. Mattavelli and L. Novelli (Editors), Organic Geochemistry in Petroleum Exploration, Part I. Advances in Organic Geochemistry, pp. 263-272. Irwin, H., Hermanrud, C., Carlsen, E.M., Vollset, J. and Nordvall, I., 1993. Basin modelling of hydrocarbon charge in the Egersund Basin, Norwegian North Sea: pre- and post-drilling assessments. In: A.G. Dor6, J.H. Augustson, C. Hermanrud, D.J. Stewart and 0. Sylta (Editores), Basin Modelling: Advances and Applications. NPF, Special Publication 3, Elsevier, Amsterdam, pp. 539-548. Lerche, I., 1993. Theoretical aspects of problems in basin modelling. In: A.G. Dor6, J.H. Augustson, C. Hermanrud, D.J. Stewart and 0. Sylta (Editors), Basin Modelling: Advances and Applications. NPF, Special Publication 3, Elsevier, Amsterdam, pp. 35-65. Megill, R.E., 1992. Estimating Prospect sizes. In: Richard Steinmetz (Editor), The Business of Petroleum Exploration, Treatise of Petroleum Geology. Handbook of Petroleum Geology, AAPG, Tulsa, Okla., pp. 63-70.
W. KROKsTAD O. SYLTA
IKU Petroleum Research, N-7034 Trondheim, Norway IKU Petroleum Research, N-7034 Trondheim, Norway
235
Rose, P R., 1992. Risk behaviour in petroleum exploration. In: Richard Steinmetz (Editor), The Business of Petroleum Exploration, Treatise of Petroleum Geology. Handbook of Petroleum Geology, AAPG, Tulsa, Okla., pp. 95-104. Schroeder, EW. and Sylta, O., 1993. Modelling the hydrocarbon system of the North Viking Graben: a case study. In: A.G. Dor6, J.H. Augustson, C. Hermanrud, D.J. Stewart and 0. Sylta (Editors), Basin Modelling: Advances and Applications. NPF, Special Publication 3, Elsevier, Amsterdam, pp. 469-484. Skjerv0y, A. and Sylta, 0., 1993. Modelling of expulsion and secondary migration along the southwestern margin of the Horda Platform. In: A.G. Dor6, J.H. Augustson, C. Hermanrud, D.J. Stewart and 0. Sylta (Editors), Basin Modelling: Advances and Applications. NPF, Special Publication 3, Elsevier, Amsterdam, pp. 499-537. Stainforth, J.G. and Reinders, J.E.A., 1990. Primary migration of hydrocarbons by diffusion through organic matter networks, and its effect on oil and gas generation. In: Advances in Organic Geochemistry, 1989. Org. Geochem., pp. 16-74. Steinmetz, R. (Editor), 1992. The Business of Petroleum Exploration, Treatise of Petroleum Geology. Handbook of Petroleum Geology, AAPG, Tulsa, Okla. Sylta, 0., 1991. Modelling of secondary migration and entrapment of multicomponent hydrocarbon mixtures using equation of state and ray-tracing modelling techniques. In: W.A. England and A.J. Fleet (Editors), Petroleum Migration. Geol. Soc. London, Spec. Publ., 59:111-122. Sylta, 0 , 1993. New techniques and their applications in the analysis of secondary migration. In: A.G. Dor6, J.H. Augustson, C. Hermanrud, D.J. Stewart and 0. Sylta (Editors), Basin Modelling: Advances and Applications. NPF, Special Publication 3, Elsevier, Amsterdam, pp. 385-398. Thomas, B.M., Mr P., Whitaker, M.E and Shaw, N.D., 1984. Organic facies and hydrocarbon distributions in the Norwegian North Sea. In: B.M. Thomas et al. (Editors), Petroleum Geochemistry in Exploration of the Norwegian Shelf. Proceedings of a Norwegian Petroleum Society (NPF) Conference, 22-24 October 1984, pp. 3-26. Ungerer, P., Doligez, B., Ch6net, P.Y., Burrus, J., Bessis, E, Lafargue, E, Giroir, G., Heum, O. and Eggen, S., 1987. A 2-D model of basin scale petroleum migration by two-phase fluid flow. Application to some case studies. In: B. Doligez (Editor), Migration of Hydrocarbons in Sedimentary Basins. Technip, Paris, pp. 415456.
This Page Intentionally Left Blank
237
Prospect resource assessment using an integrated system of basin simulation and geological mapping software: examples from the North Sea Birger Dahl and Ivar Meisingset
A software system for faster turnaround of basin modelling studies has been developed. It combines basin simulation software (Norsk Hydro's 1D and 2D basin simulation programs) and a mapping software system (the IRAP Petroleum Mapping System). The system reduces manual operations dramatically and allows pseudo-3D basin modelling studies to be completed much quicker and with less expense than previous systems. The system is described here to demonstrate the potential benefits in hydrocarbon quantification and prediction that can result from such integrated models. Assessment of generated and expelled oil and gas volumes through time is carried out using general-purpose IRAP grid arithmetic functions. Structural reconstruction can be carried out either from direct mapping of basin simulation palaeodepths, or by backstripping in IRAE IRAP Monte Carlo simulation can be used to incorporate parameter uncertainties. The system can be used for quick assessment of basins, or for detailed predictions of accumulation histories in particular structures. An example of a quick basin study is shown from the Egersund Basin. Examples of the standard output plots available from the system are shown for the Oseberg Field area.
Introduction Basin modelling is a commonly used tool for assessment of source rock hydrocarbon generation, expulsion, migration and entrapment in prospect appraisal studies ( e.g. Yukler and Welte, 1980; Dor6 et al., 1993). Such studies are multidisciplinary and laborious and may normally require a longer time than available in typical Norwegian licence round projects. Many of these labour intensive and manual operations are suitable for computer handling. This paper presents a system that collects and organises 1D and 2D basin simulation data and pseudo-3D basin modelling data (grids). The main function of the system is to transfer simulation data digitally to general purpose mapping systems such as IRAP, statistical analysis systems such as SAS, or spreadsheets or any other program with the ability to read tabulated data. The link to IRAP has been found especially useful. IRAP is extremely well suited for pseudo-3D model building and it can perform all of the necessary volume calculations with or without Monte Carlo simulation. This paper focuses on the link module itself (BasXYZ), sometimes referred to as the Simulation Database, and on the Basin Assist module, which is
an IRAP system written in the IRAP programming language (IPL). Other basin modelling systems are often based on densely spaced 1D simulations (e.g. Skjerv~y and Sylta, 1993) and maps can be made by direct computer contouring of simulation parameters from these points. The system presented here is different in that it can use a combination of 1D and 2D simulations without the requirement for dense or regular spacing. The maps are made indirectly using interpolation techniques similar to those used in depth conversion and reservoir model building. This implies that existing software, e.g. 1D and 2D simulation programs, depth conversion software and reservoir building programs (IRAP) etc. can be used in combination. The only requirement is the link that transfers the data digitally (BasXYZ). It has been suggested previously (Dahl et al., 1988; Yukler and Dahl, 1993) that integration of basin simulation programs with other geosoftware will provide advantages and will be a future activity both in basin modelling and in general computerassisted exploration and field development. To the knowledge of the authors, a comprehensive system of the type described here has not yet been presented in the literature.
Quantification and Prediction of Petroleum Resources edited by A.G. Dor6 and R. Sinding-Larsen. NPF Special Publication 6, pp. 237-251, Elsevier, Amsterdam. 9 Norwegian Petroleum Society (NPF), 1996.
238
B. Dahl and L Meisingset
mentologists, structural geologists, geochemists and seismic interpreters. At this stage the basin modeller is optimising 1D real wells and should ensure that the 1D model predictions are consistent with the observations made by the geoscientists. Any problem that is encountered will be fed back to the rest of the basin modelling team who will adjust the conceptual model. This is a very creative, but critical phase of the study and major errors will cause wrong output data and consequently wrong predictions and prospect appraisals. The number of manual operations in this phase is not particularly large: primarily transfer of reasonably small tables to the 1D simulation program. (2) In the simulation phase the conceptual model is applied to a much larger number of wells, pseudowells and/or 2D grid nodes. Pseudowells and 2D grid nodes are picked from seismic interpretation, which in the conceptual phase was tied to the primary control wells. Picking and transferring of all these data to the simulation program could be a rather large operation. Previously it was carried out manually, and prevented many organisations from engaging in larger basin modelling studies. The simulation phase comprises simulation runs with optimisation for thicknesses and checks on organic maturities and temperatures from wells close to simulated grid points. The thickness optimisation in some systems is taken care of by the simulation software itself. This phase is the easiest of all the steps in a basin modelling project and is primarily labour intensive with respect to data transfer, most of which can be carried out by technical assistants. (3) The synthesis and prediction phase starts with posting of the simulation data on paper, forming the
Motivation
The development of this computer system was motivated by the enormous amounts of data that must be handled in basin modelling studies. This data handling can be very labour intensive if it must be done by hand. Those parts of a basin modelling study that involve much data handling and would be the most labour intensive if done manually, can be identified by considering the way these studies are organised.
Organisation of basin modelling studies Integrated basin modelling studies can be organised according to the flow chart (Fig. 1) suggested by Dahl and Yukler (1991). The work is subdivided into four phases with significantly different ratios between intellectual and manual work. (1) The conceptual phase is the most critical stage of a basin modelling study and requires large efforts from geoscientists. The objective of this phase is to acquire the necessary basin understanding to optimise the subdivision of the sedimentary column into appropriate geological events. These include depositional, nondepositional or erosional episodes. With respect to the erosional episodes, the significance, timing and magnitudes have to be determined (e.g. Dahl and Augustson, 1992). Depths, thicknesses and other parameters must be determined for each event in each of the wells used to define the conceptual model. The events should be tied to the seismic interpretation. Further seismic interpretation is often required at this time. The entire process requires interaction between complementary geoscientists, such as stratigraphers, sedi-
I Single well studies, i I CONCEPTUAL PHASE
Regional studies
>~ Model input concept
II SIMULATION PHASE
III SYNTHESIS AND PREDICTIVE PHASE IV VERIFICATION PHASE
1-D or 2-D simulation ~[ Regional geochemical studies
Synthesis
<~
Predictions ,
Verifications
Fig. 1. Organisationchart for integratedbasin modelling studies.
Structural studies I
Prospect resource assessment using an integrated system of basin simulation and geological mapping software
basis for cross-sections or maps. Most 1D and 2D programs contain graphics which can plot properties on graphs or cross-sections as a function of depth or time. In 3D studies, maps for various properties are required either for structural reconstruction, hydrodynamic assessment or generation and expulsion calculations. In the latter activities mathematical operations on a series of maps are required (Dahl and Yukler, 1991). This is a very tedious and time-consuming task to perform manually. For each operation, intermediate maps are produced. Consequently, this is an appropriate job for a computer. When the maps are finished the interpretation, prediction and prospect appraisal can commence. This activity requires the skills of the geoscientists in the basin modelling team. (4) The verification phase also requires geoscientific effort when the model predictions are compared with field observations from drilled traps. Large-scale manual operations hardly exist in this last phase. The above description points to two phases that potentially require much manual intervention: phase 2, in which the data are input to simulation, and phase 3, in which the pseudo-3D model is built (in map form) and then used in hydrocarbon volume calculations and predictions. Fig. 2 shows project plans for a regional pseudo-3D basin modelling study. The upper part of the diagram shows a plan for a "manual" study, where simulation data are hand posted and hand contoured prior to gridding and the computer volume calculation. The lower part shows a "computerised" study as it would be carried out by the system presented, with digital transfer of simulation input data and direct gridding in IRAP to form the appropriate pseudo-3D maps. The study incorporates 10-20 1D simulations in real wells in the conceptual
239
phase, plus about 50 1D simulations of pseudo-wells in the simulation phase, and covers an area of about 10,000 km 2 (close to two North Sea quadrants). There are three source units yielding two hydrocarbon phases (gas and oil) at five time steps. The basin modelling team includes 1 seismic interpreter, 1 geologist, 1 geophysicist/computer mapping specialist, and 1 geochemist/basin modelling specialist. The "manual" study also requires a technical assistant. Comparison between the two approaches shows that the turn-around time would have been doubled if the "manual" approach had been used.
System description Description of the basin simulation software Two modelling software packages have been used in conjunction with this integrated system: an improved deterministic dynamic basin model based on an earlier concept by A. Yukler (Yukler and Welte, 1980) and an "in-house" rewritten software based on the same concept. Both programs can be operated in either a 1-dimensional or 2-dimensional mode and determine the geologic history, palaeopressure, palaeotemperature and generation of hydrocarbons in a sedimentary basin (Yukler and Welte, 1980; Dahl and Yukler, 1991). The simulation software describes the physicochemical processes occurring at and after sediment deposition in mathematical terms. The geological processes taking place as a function of time are divided into events, characterised by positive (deposition), zero (non deposition) or negative (erosion) thicknesses of homogenous lithologies that occurred during a specific time interval. The sediments deposited
Fig. 2. Resource assessment of basin modelling studies not using and using the integrated basin simulation and IRAP mapping system.
B. Dahl and L Meisingset
240
Basin Assist
Seismic/log Work station
1 , 2 a n d 3D
simulation
Simulation database
9Volumes 9Migration 9Structural
"BasXYZ"
General database
P R
IRAP
-->.
reconstruction 9Monte Carlo simulation
-~
o A S P E C
---~
E V
L U A T
T I 0 N
Map database Fig. 3. Flow chart describing the integrated basin simulation-mapping system.
are chronostratigraphic units described by lithological composition, present day porosity, palaeobathymetry and palaeotemperatures at the sediment-water interface. The model uses heat flow histories rather than constant geothermal gradients. The heat flow history is determined by iteration until the calculated vitrinite reflectance trend matches the measured trend from a well. This optimisation is then checked against sterane (20S) and triterpane (22S) isomerization data. Temperature histories are used in a kinetic subroutine modified from Tissot and Espitali6 (1975), which calculates the transformation of kerogen into maximum 4 component groups: C1, C2-C5, C6-C14, C15+ plus cracking of oils to gas components.
Description of the IRAP Petroleum Mapping System (PMS) IRAP PMS is a conventional petroleum mapping system, with a general purpose command language that can be used for programming. The command language, called IPL (for IRAP Programming Language), also has commands that generate Motif user interface panels. Arithmetic and logical operations on grids are available, including Monte Carlo simulation. IRAP is much used for reservoir model building, and is therefore well suited for pseudo-3D basin modelling. Technically these are very similar processes. Consequently IRAP can be used in basin modelling interpretations, e.g. in the multiplication of maps for volumetric assessments, for structural reconstruction by backstripping of maps and as a general purpose plotting program in 3D studies.
Description of BasXYZ and Basin Assist Basin Assist consists of a series of modules, that form the basis for a complete pseudo-3D basin modelling system. Fig. 3 shows how Basin Assist is or-
ganised. The simulation database module (BasXYZ) sits between basin simulation and mapping. Basin Assist modules in IRAP help the user perform several important tasks, especially calculation of generated and expelled hydrocarbon volumes, interpretation of secondary migration, structural reconstruction and Monte Carlo simulation. The input to the process may come from seismic or log correlation workstations. The output is normally used in prospect evaluations. Project databases support data transfers between the individual programs, while general databases supply additional information. The system is loosely integrated. This means that data are available to all applications through copying. Data updated in one application do not automatically change elsewhere. The Basin Assist IRAP interface layout is shown in Fig. 4. This is a simple version, with only one simulator connected. A pulldown menu in IRAP is used to access all necessary programs, including HYDROBAS 1D simulation, the simulation database (BasXYZ) and the IRAP-internal pseudo-3D modules. The simulation database (BasXYZ) stores all input and output data from 1D, 2D and 3D simulations. BasXYZ is written in C-+-+, and may be used separately from the rest of the system, which is written in IPL (IRAP Programming Language). Basin Assist is a software system for management of basin modelling data in addition to numerical compilations and calculation of the basin modelling output data (e.g. generated oil volume assessments). It consists of several modules" (1) The IRAP data management module organises the grids that make up the pseudo-3D model, the drainage area polygon files, and control point data and other file types. This module is based on a structured file naming convention. Each project goes into a separate directory. Within the directory, all event files must be prefixed "EV_", all surfaces "SU_", and all drainage area files "DR_". The poros-
Prospect resource assessment using an integrated system of basin simulation and geological mapping software
241
Fig. 4. Basin Assist module interface layout.
ity map file of Upper Draupne Fm may thus be called "EV_135-128_REF0_POROSITY.GRI". Here "135128" represents the time interval of the event used in the conceptual input model to the basin simulation program. "REF0" indicates that the file is referring to 0 mill. years ago (Present), "POROSITY" is self evident and ".GRI" means grid, a standard IRAP annotation. This naming convention may result in long filenames, but it is easy to use and becomes very powerful when management of a thousand such files are necessary. It is even more powerful in Basin Assist, where index lists of each of the 4 elements in the file name assist the user in making complicated searches for data in a very "visual" and user-friendly manner. (2) The plotting module is an extension of the "IRAP data" management module. A file can be selected and plotted either as hard copy or on the screen by pressing a single button. Before plotting the first time in a given project, the user must set up a number of parameter menus. These allow the system to be taught that ".GRI" means grid, that "POROSITY" must be plotted using a special colour scheme, etc. There is a plot style available for each kind of "IRAP data" file. Each file will be plotted according to its own style when the user presses the plot button. The IRAP data management module with the plotting module provides the user with a tool for browsing through data. There may easily be hundreds or possibly thousands of maps (grids) in a single pseudo 3D study and it is not practical to include hard copies of every one of them in study reports. This browsing tool, along with equally powerful browsing tools in the simulation database, makes Basin Assist a supplement to written reports and a good long-term storage database system. (3) The grid interpolation module allows a thick layer to be subdivided into thinner layers, assuming
either onlap, truncation or conformity. The module analyses the model, detects missing depth and/or isopach grids and allows the user to create the missing grids in various ways. This module is particularly helpful when input and output data from simulations with varying levels of event detail must be merged. (4) The hydrocarbon volume calculation module uses the completed pseudo-3D model to calculate generated and expelled hydrocarbon volumes by fractions and time steps. A recent study in the North Viking Graben used 4-component group simulation kinetics, C1, C2-C4, C6-C14 and C15+ modified after Burnham and Dahl (1993), 7 time steps and 4 source rock horizons with 4 associated component potential maps. This study consequently generated 112 grids for distributions of various type of expelled hydrocarbon volumes. A part of the selected study area was subdivided into 30 drainage areas yielding a total of 3360 volume numbers when applied to the 112 grids. (5) The computation utilities module includes four useful, general-purpose mapping utilities, i.e. well tie, approximation gridding, cross-section display and grid cross plot. The approximation gridding utility is the most important of these. It allows the user to grid basin simulation data using a variety of different approaches, all of which will end up with a grid that ties all control points to perfection. Approximation gridding is different from regular gridding in that an existing grid must be used to define a trend surface on which the shape of the final grid is based. The most powerful form of approximation gridding has 3 steps. First the parameter to be mapped is crossplotted against an existing grid (e.g. porosity against depth) and a "look-up" curve (IRAP term for a best fit line) is drawn through the point cloud. Secondly the look-up curve is used to make an intermediate version of the map. Thirdly a
242 well-tie utility is called upon to adjust the map until it ties all control points perfectly. At the end it is possible to use the grid cross plot utility to verify that the grids produce the same kind of dependency that the look-up curve originally was based upon.
Description of the integrated system The complete integrated system is shown as a flow chart in Fig. 3. The three most important modules are the basin simulation software, the simulation database "BasXYZ" and "Basin Assist". The philosophy is that digitised data are transported between the modules and that the modules can acquire data from other peripheral sources: seismic workstation or electric log interpretation programs for input data, map database (stored IRAP grids) or a general geological/ geochemical database. Thus the input data (digitised on, for example a seismic workstation) are transferred to the simulation program, and the output from this stage is transferred further and "manipulated" by other interpretational programs. In principle all numbers necessary in modelling studies should be transferred between the modules all the way through to the interpretation and prediction stage. The punching of large tables into the computer is therefore not necessary with this system. The Basin Assist module plays the most central role in the system and the other modules can be accessed from Basin Assist. This is illustrated in Fig. 4, which shows the Basin Assist user interface layout
Practical application of the integrated system The most important aspects of pseudo-3D basin modelling studies are of course the evaluation of the geohistory of a structure and its associated drainage area, to assess structure formation and its dynamic evolution in relation to the drainage area's hydrocarbon generation, expulsion and migration. To achieve these goals mapping of simulation output data and their synthesis with other quantitative geological information is essential:
Mapping of basin simulation output data in pseudo-3D studies Maps can be made directly from the simulation output data. This is, however, dependent on the gridding of the 1D pseudowells and/or the positions and densities of 2D profiles relative to the subsurface topography. In cases with few pseudowells and large spacing between them, maps are made mostly using the "lookup curve" version of approximation gridding (see above) of erosion isopachs, palaeobathymetries, decompaction factors, porosities and vitrinite reflectances.
B. Dahl and L Meisingset
In a recent in-house study, the three first of the map (grid) sets (erosion isopach, palaeobathymetry and decompaction factors) were used in structural reconstruction at 7 selected reference ages. This involved stretching the grids vertically to compensate for decompaction and adding palaeowater depth and erosion (if any). Layers younger than the current reference age were stripped off. All older layers were reconstructed. Palaeodepth and isopach grids were output. Lookup curves between vitrinite reflectance and kerogen transformations have been used. These reduced the numbers of hydrocarbon generation maps that needed to be produced, for example through conventional hand contouring and subsequent digitising and gridding. From the simulation data in this particular limited area, a good relationship was found between the vitrinite reflectance data and the hydrocarbon generation data. Ro was modelled for each of the 4 source events at each of the 7 reference ages. Transformation ratio grids could therefore be created via lookup curves from vitrinite reflectance grids (Fig. 5). The fourth set of grids, porosity, was used together with transformation ratios, isopachs and hydrocarbon potential ($2) grids to compute expelled hydrocarbons. Approximately 500 grids were involved in these calculations.
Assessment of generated and expelled hydrocarbons As indicated above assessment of generated and released hydrocarbon volumes in 3D studies are carried out outside the simulation module by multiplying maps of source rock properties and basin simulation outputs. Using a standard volume assessment formula (e.g. Dahl and Yukler, 1991) these calculations are fairly simple and straightforward and give the integral volume at a given geological time. For detailed prospect appraisal and filling/leakage histories incremental volumes for time slices are appropriate. In addition, when 4-component kinetics plus cracking are used, the operation becomes more complex and requires a detailed book-keeping system. Because the 4-component kinetics (modified after Burnham and Dahl, 1993) used in the simulation software calculate large amounts of heavy components relative to lighter material a simple expulsion efficiency module has been applied. It is evident from geochemical observations that heavy and more polar components are preferentially retained in source rocks during expulsion (Tissot and Welte, 1978; Waples, 1985). Straight-forward application of the data from the basin simulation would then give erroneous results. Of several proposed mechanisms for
Prospect resource assessment using an integrated system of basin simulation and geological mapping software
243
1.0 0.9
ICracking,,~,
0.8-
, 0.7 O
0.6
Heavy oil
,=9 0.5 E 0.4 r
C
~- o.3 0.2
/
/f
~Wet gas
016
018
--.-__
0.1 0.0
0.0
0'.2
014
110
112
1'.4
Vitrinite reflectance (Ro)
1'.6
1'.8
2'.0
2'.2
2'.4
Fig. 5. "Look-up" curve between HC-generation and vitrinite reflectance.
1.0 0.8o ,,,.,. o,,.
o tO Ig u. ,..,,
(3
L_
0.6 0.4 0.2
0
0.1
012
013
0.4
015
0.6
0'.7
018
o.9
1.0
Transformation ratio (TR)
Fig. 6. Fractional oil flow as a function of kerogen transformation for various oil-water viscosities. expulsion (Tissot and Welte, 1978), pressure-driven flow controlled by relative permeabilities and viscosity contrasts combined with saturation dependency was selected as an appropriate mechanism in the expulsion calculations. The pressure-driven flow was based on Braun and Burnham (1992) and Palciauskas (1991). The saturation dependency was derived from observations of Rock-Eval, S 1 values and extraction yields from source rocks. These observations were indications of the amounts of retained oil and indicated the magnitudes of the retention threshold as a function of source richness and state of thermal maturity. In the calculations these threshold values are controlled by the degree of compaction/shale porosity. The volumes of oil and gas exceeding the threshold values can flow out of the source rock controlled by the water/petroleum viscosity contrast. Since the simulation program does not calculate hydrocarbon-
induced overpressure, it is assumed that the necessary pressure drive exists. A lookup curve for fractional flow (expulsion efficiency) versus kerogen transformation for various water-hydrocarbon viscosity contrasts was modified from Palciauskas (1991) (Fig. 6) and used in the calculations. The material retained in the rock matrix is cracked to lighter material using the stoichiometric coefficients proposed by Braun and Burnham (1992). In the Basin Assist pulldown menu system the viscosity contrasts for the various generated hydrocarbon groups, cracking (Palciauskas, 1991) stoichiometric parameters, densities etc., can be set to override the default values. The Basin Assist subroutine is written in IPL (IRAP Programming Language) and one run took 7 hours in a study on the scale described above (4 source rocks broken down into 4-component potential maps and 7 time slices).
B. Dahl and L Meisingset
244
In cases where the uncertainties in the source rock properties are so large that average values are more convenient to use than mappable distributions, Monte Carlo simulation with input probability distributions can be applied from "Basin Assist". In this respect Basin Assist can combine deterministic and probabilistic approaches.
Field studies Oseberg area: fast standard source volumetrics and reporting As shown above, the system was constructed to provide fast turnaround in quantitative basin modelling studies. In this respect regional and subregional
studies have been carried out in which map databases of distributions of generated and expelled hydrocarbons have been established. These map databases can be used for general prospect assessment of any trap in the particular region. The Basin Assist module can quickly and easily acquire generated and expelled volumes of hydrocarbons available for secondary migration to any potential hydrocarbon trap in a study area by defining the trap's drainage area. A standard layout system, consisting of a series of diagrams, has been made for prospect appraisal documentation. An example of such a series of diagrams is shown in Figs. 7 to 12. It is taken from the map database of a large regional North Viking Graben study. The area around the Oseberg Field, an oil and gas accumulation in a Middle Jurassic reservoir in the northern
Fig. 7. Drainage pattern in the Oseberg area based on IRAP orthocontours.
Prospect resource assessment using an integrated system of basin simulation and geological mapping software
Norwegian North Sea (Dahl and Yukler, 1991), is used as an example. The first diagram shows the drainage area determined by use of orthocontouring (orthocontours are simply contours that can be drawn perpendicular to, for example, depth contours). This map gives a qualitative indication of secondary migration of hydrocarbons along the carrier bed and of areas where migration is focused. Fig. 7 shows also the drainage area, with depth contours on the top of the source rock unit and Fig. 8 shows the average distribution of oil-prone organic material within that rock unit for the same drainage area. There is a similar map for gas-prone material (not shown). The next two diagrams (Figs. 9 and 10) show the source thickness and the kerogen transformation of oil prone kerogen within the drainage area.
245
A similar kerogen conversion map is also produced for the gas- prone kerogen (not shown). The last map in the series (Fig 11) is the result of the volumetric calculation: the distribution of expelled oil per unit area. Similar maps are constructed for generated gas. The last diagram in the series (Fig. 12) is a plot of volume oil generated versus time. (Gas versus time is not shown). These curves are valuable in prospect appraisal as they indicate in a qualitative manner which geological periods had strong generation of oil relative to gas, and how oil and gas compete for the trap volume in cases where generated volumes are in excess of the trap volume. These curves can also be assessed in relation to strong basin movements and periods of likely seal rupture and trap leakage (Dahl and Yukler, 1991) in order to predict accumulation
Fig. 8. Distribution (wt%)of oil prone kerogens in the Viking Group, Huldra-Oseberg drainage area.
B. Dahl and L Meisingset
246
Fig. 9. Total isopach (m) Viking Group in the Huldra-Oseberg drainage area.
histories and likely dominant hydrocarbon phase in the trap.
Egersund Basin: fast prospect appraisal with parameter uncertainties The Egersund Basin is a small subbasin located east of the Central Graben in the Norwegian sector in the North Sea. Upper Jurassic source rocks are present in the basin, but are probably marginally mature, and only minor amounts of oil have been found (Ritter et al., 1987; Hermanrud et al., 1990; Irwin et al., 1993). In-house work had identified structures and associated drainage areas (A, B and C) with fair amounts of potential reserves in blocks that were open in the 13th Norwegian concession round
(Fig. 13). With the knowledge of the Egersund Basin being marginally mature the task was to determine if the Statoil 9/2-1 find had indications of other source rocks than the Upper Jurassic shales. If no imprint of other active sources were apparent, the objectives were to quantify the volumes of hydrocarbons that had been available for migration towards the identified structures. If the volumes were too small to be commercial, the area would be abandoned and further exploration would cease. An oil-source correlation study (in-house report) suggested the 9/2-1 oil to be a low maturity Upper Jurassic derived oil with no sign of other possible source rocks. The basin modelling study that followed using the basin simulation-IRAP integrated system was
Prospect resource assessment using an integrated system of basin simulation and geological mapping software
247
Fig. 10. Average transformation of oil prone kerogen in the Viking Group, Present.
carried out by 2 persons over 4 days. Available for the study were a few digitised subsurface maps, field data including the well completion log and the geochemical data from Well 9/2-1. Two approaches were applied. The first and simplest used average values for the various appropriate source rock properties, and volumes were calculated for different "cases" (Table 1). The other approach used parameter uncertainties in Monte Carlo simulation to produce volume probability distributions. To assess the general heat-flow history and identify possible significant erosional events, Well 9/2-1 was simulated and optimised against vitrinite reflectance and sterane isomerization. The heat-flow history from this well was transferred to a pseudo-well in the central parts of the basin, which was simulated.
Table 1 Cases used for the calculation of generated and expelled petroleum in addition to the assessment of migration efficiencies in the Egersund Basin Source rock case number Oil prone TOC (%)a
I 2.24
II 5.44
Heat flow case number Tertiary heat flow (HFU)
I 0.90
II 0.95
III 1.00
a The source rock has a hydrogen index of 700 kg HC/tonne org. C.
Because of uncertainties shown by the work on Well 9/2-1, three slightly different heat-flow histories were applied (Fig. 14). It can be seen from this diagram and Fig. 13 that the top of the source rocks in the deepest part of the basin has not reached peak oil
B. Dahl and I. Meisingset
248
Fig. 11. Distribution of expelled oil (m3/m3) in the Huldra-Oseberg drainage area, Present.
generation and that small changes in the heat flow can be critical for the volumetric assessments. The area was modelled using the different heatflow histories shown in Fig. 14. By calculating the generated volumes at present for the drainage areas of three structures using the source rock cases in Table 1, together with a source-rock isopach map, the results in Table 2 were obtained. It should be noted that the in-place volume from the 9/2-1 structure was used actively for calibration and comparison with calculated volumes. The table shows the in-place potential volumes for the various structures against the total volume generated and expelled in the respective drainage areas. The accumulation efficiencies listed represent volumes found over volume expelled from the 9/2-1 structure. The calculated accumulation effi-
ciencies are between 26% and 7%. The accumulation efficiencies for the different heat-flow and TOC cases are used to calculate the potential volumes accumulated in the undrilled prospects. Table 2 demonstrates clearly that the accumulation efficiencies required to fill the structures to their spill points are much higher than the range of efficiencies determined for the 9/2-1 structure. In fact, some cases require more than 100% accumulation efficiencies. It is evident that it is difficult to fill the appraised structures completely. The second approach involving parameter uncertainties (Fig. 15) and Monte Carlo simulation (500 simulations), revealed expected (P50) generated volumes that were in the same range (Table 3) as those from the first approach. Even the more optimistic re-
249
Prospect resource assessment usingan integrated system of basin simulation and geological mapping software 7000
0.0 L_
6000 5000
=m
-
1.0-
03
E
--
E
4000
Cumulative
-
E _= 3000-
o
_o>_
0
A
2.0-
L_
-
O
G)
1ooo
Rate 0
o
E
2000 -
-' 20
I
-100
1
-80
-60
-40
-20
0
suits (P80) indicate that too small petroleum volumes were generated to fill any structures to their spillpoints. These results supported the conclusions from the first approach. Based on this study, work on the area was abandoned and manpower could be directed towards acreage regarded as more prospective.
Conclusions Pseudo-3D basin modelling can be carried out quickly, and at lower cost, when data transfer between different programs can be done digitally. Norsk Hydro
O.90HFU_
o
=,..
4.0-
0.95HFU 1.00HFU
_
Age (mill. years)
Fig. 12. Volume of expelled oil versus time report in the HuldraOseberg drainage area.
/
3.0-
r L_
3 '='3
5.0
I
I
I
0.0 Oil g e n e r a t i o n
rate
0'3
I
0.4
(gHC/gTOC)
Fig. 14. Conversion of oil prone kerogen simulated in an Egersund Basin pseudo-well using three different heat flow histories, Present.
has developed such a software system, which uses the calculation abilities of IRAP to perform volume calculations of generated and expelled oil, with or without Monte Carlo simulations. This system consists of three main parts: an internal 1D and 2D simulation program, a simulation database module (BasXYZ) and an IRAP based mod-
Fig. 13. Egersund Basin; depth (m) to top Vestland Group with drainage areas for prospects A, B, C and the 9/2-1 structure.
B. Dahl and L Meisingset
250
Table 2 Volumes for prospects A, B, C and proven oil content in the 9/2-1 structure HF cases
TOC cases
A
B
C
92
140
24
9/2-1 12
Proven oil volume (mill. m 3) Prospect volume (mill. m 3) Expelled oil (mill. m 3) Migration efficiency (%) Potential volume accumulated (mill. m 3)
I I I
I I I
35
60
1
10
16
3
Expelled oil (mill. m 3) Migration efficiency (%) Potential volume accumulated (mill. m 3)
I I I
II II II
115
195
5
10
18
0.5
Expelled oil (mill. m 3) Migration efficiency (%) Potential volume accumulated (mill. m 3)
II II II
I I I
50
90
2
9
17
0.4
Expelled oil (mill. m 3) Migration efficiency (%) Potential volume accumulated (mill. m 3)
II II II
II II II
155
265
11
19
III III III
I I I
80
130
12
20
Expelled oil (mill. m 3) Migration efficiency (%) Potential volume accumulated (mill. m 3)
45 26 135 9 65 19
9
175 7
0.6 5
80 15
0.8
Volumetric results from the deterministic approach using the heatflow values from Fig. 15 and the TOC cases in Table 1.
Source oil prone TOC (%)
1.8
2.5
Depth conversion (m)
5.4
-200
0
Tertiary heat flow (HFU)
+200
0.90
0.95
1.00
Fig. 15. Source rock variables for Monte Carlo simulation.
Table 3 Volumetric results for the various prospects in the Egersund Basin using the Monte Carlo approach Probability case
A
B
C
9/2-1
92
140
24
12
Proven oil volume (mill. m 3) Prospect volume (mill. m 3) Expelled oil (mill. m 3) P20 migration efficiency (%) Potential volume accumulated (mill. m 3)
P20 P20 P20
43
71
1.4
11
19
0.4
Expelled oil (mill. m 3) P50 migration efficiency (%) Potential volume accumulated (mill. m 3)
P50 P50 P50
65
109
11
19
Expelled oil (mill. m 3) P80 migration efficiency (%) Potential volume accumulated (mill. m 3)
P80 P80 P80
90
159
10
17
ule (Basin Assist). We have experienced very significant reductions in project and tumaround time by using this system compared with previous manual studies.
3 0.5 5
47 26 69 17 12 107 11
0.6
Acknowledgements The authors wish to thank Norsk Hydro A.S. for permission to publish this paper. We are also indebted
Prospect resource assessment using an integrated system of basin simulation and geological mapping software
to A.G. Dor6 and O. Sylta for advice that improved this publication. References Braun, R.L. and Burnham, A.K., 1992. PMOD: a flexible model of oil and gas generation, cracking, and expulsion. Org. Geochem., 19(1-3): 161-172. Burnham, A.K. and Dahl, B., 1993. Compositional Modelling of Kerogen Maturation. In: K. Oygard et al. (Editors), Organic Geochemistry. Poster session from the 16th Meeting on Organic Geochemistry, Stavanger, pp. 241-246. Dahl, B. and Augustson, J.H., 1993. The influence of Tertiary and Quaternary sedimentation and erosion on the hydrocarbon generation in Norwegian offshore basins. In: A.G. Dor6, J.H. Augustson, C. Hermanrud, D.J. Stewart and 0. Sylta (Editors), Basin Modelling: Advances and Applications. Norwegian Petroleum Society (NPF), Special Publication 3, Elsevier, Amsterdam, pp. 419431. Dahl, B. and Yukler, M.A., 1991. The role of petroleum geochemistry in basin modelling of the Oseberg area, North Sea. In: R.K. Merrill (Editor), AAPG Treatise of Petroleum Geology Handbook. Source and Migration Processes and Evaluation Techniques. American Association of Petroleum Geologists, Tulsa, pp. 65-85. Dahl, B., Kvalheim, O.M., Christie, A.A. and Yukler, M.A., 1988. Multivariate data analysis in quantitative petroleum geology: applications to basin analysis. AAPG Research Conference, Petroleum Potential of Sedimentary Basins, Techniques, Methods and Approaches, Leesburg, Virginia, April 25-29, Abstract. Dor6, A.G., Augustson, J.H., Hermanrud, C., Stewart, D.J. and Sylta, 0. (Editors), 1993. Basin Modelling: Advances and Applications. Norwegian Petroleum Society (NPF), Special Publication 3, Elsevier, Amsterdam. Hermanrud, C., Eggen, S., Jacobsen, T., Carlsen, E.M. and Pallesen, S., 1990. On the accuracy of modelling hydrocarbon generation and migration: the Egersund Basin oil find, Norway. Org. Geochem., 16(1-3): 389-399. Irwing, H., Hermanrud, C., Carlsen, E.M., Vollset, J. and Nordvall, I., 1993. Basin modelling of hydrocarbon charge in the Egersund
B. DAHL I. MEISINGSET
25 ]
Basin, Norwegian North Sea: pre- and post-drilling assessments. In: A.G. Dor6, J.H. Augustson, C. Hermanrud, D.J. Stewart and 0. Sylta (Editors), Basin Modelling: Advances and Applications, Norwegian Petroleum Society (NPF), Special Publication 3, Elsevier, Amsterdam, pp. 539-548. Palciauskas, V.V., 1991. Primary Migration of Petroleum. In: R.K. Merrill (Editor), AAPG Treatise of Petroleum Geology Handbook. Source and Migration Processes and Evaluation Techniques. American Association of Petroleum Geologists, Tulsa, pp. 65-85. Ritter, U., Leith, L.T., Griffiths, C.M. and Schou, L., 1987. Hydrocarbon generation and thermal evolution in parts of the Egersund Basin, Northern North Sea. Can. Soc. Pet. Geol., Mem., 12 pp. 75-85. SkjervOy, A. and Sylta, 0., 1993. Modelling of expulsion and secondary migration along the southwestern margin of the Horda Platform. In: A.G. Dor6, J.H. Augustson, C. Hermanrud, D.J. Stewart and 121. Sylta (Editors), Basin Modelling: Advances and Applications. Norwegian Petroleum Society (NPF), Special Publication 3, Elsevier, Amsterdam, pp. 499-537 Tissot, B.P. and Espitali6, J., 1975. L'evolution thermique de la mati~re organique des sediments: Application d'une simulation math6matique. Rev. Inst. Fr. Pet., 30: 743-777. Tissot, B.P. and Welte, D.H., 1978. Petroleum Formation and Occurrence. Part IV, Springer-Verlag, Berlin. Waples, D.W., 1985. Geochemistry in Petroleum Exploration. D. Reidel Publishing Co., Boston. Yukler, M.A. and Welte, D.H., 1980. A three-dimensional deterministic dynamic model to determine geologic history and hydrocarbon generation and accumulation in a sedimentary basin. In: Fossil Fuels: Hydrocarbons C2, 26th International Geological Congress. Editions Technics, Paris, pp. 271-285. Yukler, M.A. and Dahl, B., 1993. Future potential of basin modelling techniques. In: A.G. Dor6, J.H. Augustson, C. Hermanrud, D.J. Stewart and 0. Sylta (Editors), Basin Modelling: Advances and Applications. Norwegian Petroleum Society (NPF), Special Publication 3, Elsevier, Amsterdam, pp. 67-70. Yukler, M.A. and Welte, D.H., 1980. A three-dimensional deterministic, dynamic model to determine geologic history and hydrocarbon generation, migration and accumulation in a sedimentary basin. In: Fossil Fuels. Technip, Paris, pp 267-285.
Norsk Hydro Research Centre, Bergen, Norway Present address: Dept. of Geology, University of Bergen, Norway Norsk Hydro Exploration and Production, Oslo, Norway
This Page Intentionally Left Blank
253
Enhanced oil recovery---the international perspective J.J. George Stosur
This paper presents information on the status of worldwide Enhanced Oil Recovery (EOR) activities with focus on applicable technologies, resulting production rates, number of projects by EOR methods employed and regions of the world, economic performance and technology trends. EOR is making a significant and growing contribution to the world's oil supply, particularly in the USA and Canada. Despite the lower world oil prices since the mid-1980s and the decline in new project starts, EOR production rates have grown steadily. Production rates from gas flooding have sharply increased, chemical flooding has almost disappeared, and thermal recovery has stabilized.
Introduction
Worldwide oil production by enhanced oil recovery (EOR) is now two million barrels per day, or 3.5% of the world's total oil production. About onehalf of the world's total EOR production is in North America, where oil fields are mature and prospects for discovery of new large oil fields are small and declining. Currently available oil recovery technology allows about one-third of the discovered oil to be produced; the remaining two-thirds cannot be recovered economically. This creates a very large and tempting target for improved oil recovery methods. This paper reviews the state-of-the-art of EOR, focusing on the United States, where there were by far most field experiments and commercial projects. Since other countries and regions of the world will eventually face exactly the same problem and will want to maximize oil recovery from their reservoirs before those reservoirs are abandoned, the experience from the United States should be of more than passing interest. E O R --- s t a t e of t h e art
The life cycle of mature oil fields usually exhibits three, more or less distinguishable stages. At the end of the primary stage, when the reservoir's energy has been reduced, the injection of water or gas to increase oil recovery has been practiced for several decades. That practice became known as the secondary recovery stage. The third crop of oil involved the introduction of more sophisticated and expensive processes which, at least at the beginning, were applied at the end of secondary methods and therefore were referred to as
Fig. 1. U.S. EOR production by primary, secondary and tertiary methods.
tertiary. To reduce ambiguities in connection with the last two reservoir life cycles, the term enhanced oil recovery (EOR) is often used to indicate any process aimed at increased oil recovery. Relative amounts of oil produced from these three reservoir life cycles in the United States are shown in Fig. 1. Proven EOR processes are classified into three categories: gas miscible, chemical, and thermal. The following is a brief description of how each process works to recover incremental oil.
Gas-miscible EOR Gas-miscible EOR includes any process in which a gas, such as carbon dioxide, nitrogen, flue gas, or enriched natural gas, is injected into a reservoir at sufficiently high pressure to achieve miscibility and to
Quantification and Prediction of Petroleum Resources edited by A.G. Dor6 and R. Sinding-Larsen. NPF Special Publication 6, pp. 253-260, Elsevier, Amsterdam. 9 Norwegian Petroleum Soc!ety (NPF), 1996.
J.J. George Stosur
254 mobilize and displace water flood residual oil. These processes have been shown to be highly effective in both sandstone and carbonate reservoirs, especially if injected at a pressure high enough to cause the injected gas and components of the oil to completely mix and stay mixed at the miscibility pressure. The injected gas generally flows into the previously water-swept zones where it displaces the mobile water and mixes with and swells the oil left in the pore space. With repeated contact of injected gas and oil, the gas extracts the more volatile portions of the crude oil to form an enriched injected gashydrocarbon mixture. This mixture then displaces most of the oil it contacts, leaving behind a very small quantity of tar-like residue. Because the injected gas has a low viscosity relative to the residual oil and water, it tends to finger through the more permeable parts of the reservoir and often moves more quickly through the top of the reservoir, "overriding" the oil. To minimize these effects, water is often injected with the gas in alternating "slugs" (particularly in carbon dioxide floods), to increase the portion of the previously water-swept zone contacted by the injectant. Other materials (e.g., surfactant foams) are under development to enable larger portions of the reservoir to be contacted. The combination of swelling, mixing, and sweeping can effectively contact, mobilize, and recover a significant portion of the immobile oil remaining in the reservoir. As gas injection continues, water, oil, and the gas injectant are recovered at the producing wells. In larger projects, the recovered injectant gas is separated, repressurized, and reinjected.
Chemical EOR Chemical EOR involves the injection of chemicals into a reservoir to reduce the interracial tension between residual oil and water and/or improve the mobility ratio (the contrast in viscosities and relative permeabilities) between the injected, displacing fluids and the residual crude oil. The major chemical processes are surfactant flooding, alkaline injection, and polymer flooding. Surfactant flooding involves the injection of a chemical that is both oil and water soluble to reduce the interfacial tension between these fluids at reservoir conditions. The chemical "slug" may include surfactant, water, hydrocarbons, alcohols, polymers, and inorganic salts. Polymer slugs, injected after (and sometimes before) the surfactant, provide mobility control and help maintain the integrity of the chemical slug. Due to the relatively high cost of surfactants, the slug is typically small relative to the volume of mobility control agents injected before and after. Sur-
factant floods must be designed specifically for each reservoir. The most effective combination of chemicals is highly dependent on the temperature, salinity, rock properties, and crude oil composition of the reservoir. Thus, the process is relatively complicated, expensive, and therefore risky compared to other EOR processes. However, pilot tests have yielded recoveries of up to 22% of the original oil in place. The high cost of chemicals combined with limited field applications have made these projects economically risky, especially in periods of fluctuating oil prices. The development of low cost, widely applicable chemicals is necessary for this technique to achieve its full potential. One subset of chemical EOR processes is permeability profile modification with chemical agents. Profile modification reduces the volume of injected water that flows through high permeability, previously swept layers of the reservoir. The treatments reduce rock permeability somewhat in all layers, with the greatest effect in the highest permeability layers. Reducing the permeability contrast between layers allows more injected water to enter and recover oil from low permeability layers. This increase in oil recovery effectively extends the economically productive life of water flood projects. Some of this bypassed oil can also be recovered by redesigning water floods to place injector and producer perforations in bypassed zones, particularly where these zones are separated by tight streaks or shale breaks forming non-communicating layers.
Thermal EOR Thermal EOR processes include steam drive, steam soak, and in-situ combustion. The hydrocarbon displacement properties and the economics of steam soak and steam drive processes have been extensively demonstrated worldwide, particularly in California. Steam EOR projects account for about two-thirds of all EOR production in the United States and are recognized as commercially proven technologies in heavy oil (less than 20 ~ API gravity) reservoirs at depths less than about 1000 m. Thermal recovery processes involve the introduction of heat to reduce the viscosity of residual crude oil, to partially "crack" the heavy oil into lighter constituents, or to produce a pressure gradient to help drive the oil through the reservoir. Steam flooding involves the injection of steam to heat the reservoir oil, thus reducing its viscosity and moving it towards a production well, where it is pumped to the surface. When steam flows through the reservoir, it condenses as the latent heat is transferred to the rocks and reservoir fluids above and below. The steam vapor also tends to rise to the top of the reservoir, while the
255
E n h a n c e d oil recovery ~ the international perspective
condensed water tends to under-run the steam zone. This "gravity segregation" of the vapor and liquid phase is due to their different densities. In-situ combustion, also know as fire flooding, is normally applied to reservoirs containing lowgravity oil, but has been tested over perhaps the widest spectrum of conditions of any EOR process. Heat is generated within the reservoir by injecting air and burning part of the crude oil, creating heat which reduces the oil viscosity and partially vaporizes the oil-in-place. The lighter residual oil is driven forward by a combination of steam, hot water, and gas drive. The relatively small portion of the oil that remains after these displacement mechanisms have acted provides the fuel for the in-situ combustion process. In some applications, the efficiency of the total in-situ combustion operation can be improved by alternating water and air injection. The injected water tends to improve the utilization of heat by transferring heat from the rock behind the combustion zone to the rock immediately ahead of the combustion zone.
EOR production trends - - q u o vadis The EOR contribution to worldwide oil production is no longer insignificant at almost 2 million barrels per day. Fig. 2 provides a comparison of EOR production rates in major oil producing nations, or regions of the world, based on published and privately assembled data. It shows that the United States is the largest EOR producing country, with a 39% share of the world's EOR production, followed by Indonesia, Venezuela, CIS, Canada, China, Europe and the rest of the world combined. The cases of Indonesia and Venezuela are somewhat special. Indonesia's Duri heavy oil field, almost single-handedly in that country, will produce at a rate
of about 330,000 barrels of oil per day at its peak. It is already the largest single EOR project in the world. Venezuela has mounted a very ambitious effort to produce that country's heavy oil resource and registered a large increase in oil production by EOR in the past two years. Oil production from EOR activities in the CIS has stabilized or declined slightly since 1990 (Mamedov et al., 1992). Fig. 3 provides a breakdown by major EOR technologies in the United States and Canada. The popularity of EOR technologies in the United States and Canada differs significantly. Thermal methods are by far the most common in the United States where as much as 461,000 barrels per day (60%) is attributed to thermal recovery. In Canada, however, the most often used method is miscible gas flooding, which contributes 140,000 barrels per day (83%) to their EOR production. Not included in the Canadian figures are about 280,000 barrels per day of bitumen produced from the Alberta tar sands. The popularity of thermal recovery in the United States (mostly steam flooding) is attributed to unusually favorable heavy oil reservoirs in Kern County, California. Likewise, the frequent occurrence of carbonate reefs in Alberta, Canada, and better availability of enriched natural gas is conducive to miscible gas drive projects. These differences illustrate that commercial applications of different EOR processes evolve at various rates and depend strongly on local conditions, such as suitability of the reservoir to a particular technology (steam flooding in California), availability of injectants (natural gas in Alberta, carbon dioxide in the Permian Basin, Texas), local market conditions, and economic incentives. One unique aspect of oil recovery in the United States is that this is the only country in the world
Fig. 2. Worldwide EOR production (1000 bbls/day).
J.J. George Stosur
256
Fig. 4. U.S. EOR production by process type. that produces its marginal wells on a significant basis (Interstate Oil and Gas Compact Commission 1994). So-called stripper wells, which by definition produce less than 10 barrels of oil per day, represent the ultimate in conservation. Table 1 shows statistical information on U.S. stripper wells for 1992. There are nearly one-half million such oil producing wells in the United States. Collectively, they contribute about 15% to the total oil production rate, but the
Table 1 Stripper wells in the United States (data for 1994) Number of stripper wells Total yearlyproduction (bbls) Average daily production per well (bbls/day) Number of stripper wells abandoned (1994) Fraction of U.S. produced oil by stripper wells (%)
452,248 355,961,000 2.16 16,914 15
average production rate per stripper well is just over two barrels of oil per day. Fig. 4 shows several important trends with respect to the popularity of various technologies in the United States. Thermal methods (nearly all steam flooding) continue to contribute the largest fraction of EOR production, but the growth rates have stopped since 1986. Actually, the thermal fraction of total EOR production has declined from 73% in 1986 to 60% in 1992. Gas miscible and immiscible projects (utilizing carbon dioxide, natural gas or nitrogen) have rapidly increased in the same period of time, while chemical methods have severely decreased. Gas injection methods are clearly best able to compete in the current low oil price regime, while chemical methods are too expensive and unpredictable at this time. These conclusions should not be extrapolated indiscriminately since too much depends on
257
E n h a n c e d oil recovery m the international perspective
individual reservoir situation, availability of injectants, specific local conditions, and incentives offered by governments. E O R a c t i v i t i e s a n d oil price
Fig. 5 illustrates a surprisingly strong correlation between new EOR project starts in the United States and the price of oil for the period 1960-1992. Although the project starts curve appear volatile when compared to the average oil price curve, the cause and effect relationships are unmistakable. Both curves reflect a slight increase in the 1960s and tremendous increase in 1980, followed by sharp decline in 1982. There also appears to be a slight delayed reaction by project starts to some of the major changes in oil price (Pautz et al., 1992). EOR project starts have continued to drop since 1986 in spite of stabilizing oil prices. This may indicate that resources and technology are being applied by expanding ongoing promising projects rather than by starting new projects. Also, the better EOR projects demonstrated a positive cash flow, while new projects would not be profitable at expected oil prices. For whatever reason, companies are not now investing in new EOR projects at the rate they were through 1986 ~ the year crude oil prices dropped to $12 per barrel. The data in Fig. 5 is for new EOR projects only and does not relate to the ongoing success of EOR projects which were started during periods of strong oil prices and continue to produce successfully in a lower price regime. The trend points to an increasing amount of EOR production coming from a smaller number of projects. This is a "natural selection" process whereby the most effective and
efficient processes continue to produce and flourish while those less efficient and effective are reduced or retained for additional technology development. While it is difficult to estimate costs of various EOR methods since they are so dependent on specific reservoir conditions and many other factors, a recent study done for the U.S. Department of Energy (Interstate Oil and Gas Commission, 1993), estimated relative cost elements and their respective labor and material components (Table 2). Overall, despite the deteriorated health of the petroleum industry, which has experienced steady decline of oil production in the United States since the mid-1980s, EOR activities have remained competitive with conventional methods of oil production and have managed to record steady production rate increases throughout that time. These observations seem to confirm the underlying strength of the technology, given that the trend has continued and even accelerated since 1990. The growth of EOR producTable 2 Labor and materials as percentage of total cost in EOR projects % Labor Drilling wells Work-overs Equipping wells Pipe installation Plant installation
18 15 10 10 16
55 63 50 50 46
27 22 40 40 38
33 38 0 100
58 57 100 0
9 5 0 0
Expenses Field operations Plant operations Production treating Overhead
m
$30
..Q
$20
140
100
I
80
ILl
~: $15
OIL PRICE
13.
"
.--I
5
~160 120 er
= $25 ----
$10-
$5/
$0 -Y'
1960
J
,
I
\ 1970
% Other
Investments
$35
.r
% Materials
1980
"l
60
I
| PROJECT | STARTS
40
V",,,,,
20
1990
YEAR Fig. 5. EOR project starts in the United States compared to crude oil prices.
uJ
oee' o
IdA
J.J. George Stosur
258
tion in the face of declining oil prices and declining EOR projects is an example of how the troubled petroleum industry adapts to the rapidly changing economic conditions and finds new routes to success.
costs significantly less to drill a well than it did ten years ago, including a horizontal well. A number of other small technical improvements should also offset the lower oil prices. Another reason is that a steady supply of domestically produced oil is becoming a strategically important issue in the United States. The outlook for various EOR technologies differs significantly. Steam injection has been, and will remain, the most commercially successful EOR technology because heavy oils are so receptive to enhanced production with steam and because of the huge resources of heavy oil throughout the world. The decrease of the viscosity of heavy oils and bitumen with temperature is so p r o n o u n c e d - and the viscosity of heavy oils must be reduced before it can be produced - - that there is no emerging competition to steam flooding. Canada alone holds over 40% of the world's tar sand resources and has made significant progress in the development of these resources which are important for Canada. The development of the Alberta tar sands with surface mining and, lately, with in-situ recovery methods using steam, is a stable and growing source of oil supply in Canada. Synthetic oil from upgraded bitumen, combined with heavy oil production, now comprise about one-third of Canada's total crude oil production. Venezuelan extra heavy oil resources in the Orinoco Belt are also comparable in size to those of the Athabasca in Canada. Indonesia, a newcomer in EOR, already boasts the largest single steam-based EOR project in the world, with oil production still increasing. Miscible and immiscible gas injection techniques enjoy the largest rate of growth, particularly in the
L o o k i n g f o r w a r d to the future
Whether EOR production growth will continue to increase or stabilize at current levels will depend, in largest measure, on the uncertainty associated with the oil price. Assuming that the oil price will stabilize at the level of around $20 per barrel for a few years, EOR production is estimated to reach a plateau near the current production levels. In this context it is interesting to compare the predictions of a major study on EOR (National Petroleum Council, 1984) completed in 1984 for the U.S. Department of Energy by the National Petroleum Council (NPC). The study estimated EOR production levels at three price scenarios through the year 2010 (Fig. 6). Actual production since the 1980s is marked in solid circles and is seen to be close to the $20 per barrel price curve (Moritis, 1994). Back in 1984, the prediction was that there would not be much future growth in EOR production at $20 per barrel and that the production would peak around 1992. In fact, the agreement between the actual EOR production that transpired and the forecast for the $20 per barrel is reasonably good, which gives confidence that the remaining price scenarios may also be reasonably accurate. EOR production declines are not expected. This is, in part, because some operating cost improvements already have been made, and more are expected. It now
1,600 1,400
9 9 9 $40/bbl
9 9 "~ 1,200
9
~
~
a
~
~
_~~ " 5 > ~" _]~'__ 9
.~
1,oo0
8
800
~~...~
9
9 9
~ ~ _ _ . . $30/bbl .~
600 O
!--
-
PRODUCTION
"~,,,,~ $20/bbl
DATA
200 1980
I
1985
I
1990
I
I
1995 2000 YEAR
I
2005
I
2010
Fig. 6. National PetroleumCouncil Study on EOR for three price scenarios vs. actual EOR production data.
259
E n h a n c e d oil recovery ~ the international perspective
United States where several large pipelines were completed to carry CO2 to oil fields in the Permian Basin in Texas. In Canada, hydrocarbon gas flooding is widely used because of accessibility to, and availability of, natural gas. Nitrogen flooding is limited to relatively deep reservoirs where high pressure makes miscibility possible. Viscous fingering (frontal instabilities caused by uneven viscosity of the injected and produced fluids) is probably the most important factor limiting miscible flood performance. Therefore, current efforts focus on gaining a better understanding of factors affecting CO2 mobility and reduction of viscous fingering effects. Carbon dioxide flooding has achieved a remarkable growth rate in the United States since 1986: a 390% increase in production, compared to only 26% for all EOR methods. It is expected that CO2 flooding in the United States and hydrocarbon gas flooding in Canada will continue to grow faster than other EOR processes, particularly as important strides are now being made to reduce gas mobility and thus increase oil sweep efficiency. Surfactant flooding has been the subject of intensive R&D for about 30 years and there have been many field tests of the process. The most recent field tests were, for the most part, considered technically successful, yet the process has not become commercial (Gall et al., 1993). A recent review of surfactant technology (as opposed to low tension polymer flooding or alkaline/surfactant/polymer processes, which were commercial in some instances), lists the following reasons why the process is not yet commercial: - high costs, even excluding chemicals; high risk and uncertainty compared to alternative projects; negative perception of the technology by oil company managers; level of expertise required to design and operate a successful project; limited range of experience under difficult conditions, e.g., carbonates, high temperatures, fractured reservoirs, offshore, at-large well spacings, with moderately heavy crudes; - high sensitivity of process to reservoir characteristics, such as permeability variations; - l a c k of standard, large-scale source of proven, effective surfactants and polymers at competitive prices; lack of standard, accurate, fast, robust commercial reservoir simulators of the process that could be used routinely to design and optimize large field projects; high uncertainty in crude oil prices, environmental regulations and fears, downsizing by many oil companies. -
-
-
-
-
-
h
i
g
h
Some of the promising and recently field-tested developments include a variety of fluid diversion techniques via permeability modification profile control, such as gel or foam treatments to improve reservoir sweep. These techniques do not always work, but sometimes spectacular results can be obtained. This class of well treatments to improve reservoir conformance requires further research and field testing to improve its predictability, and may show significant progress in the long-term. Improved production in low productivity or mature reservoirs using horizontal wells is an attractive alternate to traditional EOR methods. The use of horizontal wells has gained great popularity as drilling methods became more reliable and less costly. Thousands of horizontal wells have been drilled worldwide. The promise of the horizontal well in a conventional oil reservoir is to give many times the productivity of a vertical well, with only a small increase in cost over the vertical well. In general, the reservoir production methods are similar or the same for the vertical or horizontal well, except that the horizontal well contacts and drains a much larger section of the reservoir. The use of EOR methods and horizontal wells, both alone and in combination, appears exciting. There is no doubt that over the next few years we will see many attempts at combining horizontal wells with EOR applications.
Conclusions (1) More than one-half of oil production in the United States and Canada now comes from secondary and tertiary (EOR) projects, attesting to the maturity of North American oil fields and the growing reliance on improved recovery efficiency from the already discovered oil fields. (2) EOR continues to make a significant and growing contribution to the world's oil supply, particularly in the United States and Canada, which jointly contribute approximately one-half of the world's EOR production. (3) Since the mid 1980s, significant readjustments were made to accommodate to the lower world oil prices; the number of new project starts has dramatically declined, as did the number of active EOR projects, but EOR production rates kept growing steadily. Production rates from gas flooding have sharply increased, chemical flooding activity has almost disappeared, and thermal recovery has stabilized. (4) About 10% of total oil production in the United States now comes from EOR; 13% in Canada. Counting heavy oil and bitumen production from the Alberta tar sand deposits, the EOR share would be close to 50% of Canada's total oil production.
J.J. George Stosur
260
(5) At crude oil prices of about $20 per barrel, EOR production rates in the United States are estimated to level off. The views in this paper are those of the author and do not necessarily represent those of the U.S. Department of Energy.
References Gall, B.L., Llave, EM. and Tham, K.K., 1993. National Institute for Petroleum and Energy Research/Department of Energy (NIPER/ DOE) Chemical EOR Workshop. Final Report NIPER-698. October 1993, pp. 1-92. Interstate Oil and Gas Compact Commission, 1993. An evaluation
J.J. GEORGE STOSUR
of known remaining oil resources in the State of Oklahoma. Interstate Oil and Gas Compact Commission, III-18. Interstate Oil and Gas Compact Commission, 1994. National Stripper Well Survey, Interstate Oil & Gas Commission, Oklahoma City, Oklahoma, pp. 2-3. Mamedov, Y.G. et al., 1992. Applications of Improved Oil Recovery Technologies in a Wider Europe. 4th EC Symposium, Berlin. Moritis, G., 1994. EOR dips in U.S. but remains a significant factor. Oil Gas J., September 26, pp. 51-79. National Petroleum Council, 1984. Enhanced Oil Recovery Potential in the United States. Washington, D.C., p. 68. Pautz, J.E et al., 1992. Enhanced oil recovery projects data base. National Institute for Petroleum and Energy Research Rep. 583, pp. 14-17.
U.S.Department of Energy, Fossil Energy-33, Germantown, Washington, D.C. 20586, USA
261
Nessie: a process analysis of a generic North Sea field life cycle P. McGaughrin, K. Ashton, N. Fuller, R.G. Heywood, R.W. Holt, B.L. King and S. White
A benchmarking study of 99 fields, conducted in 1991, reviewed the length of time North Sea operators took to move their project from acreage acquisition to commercial production. As part of Phillips' Quality Improvement Programme, a multidisciplinary Quality Action Team was set up to produce a generic plan to compress its core business cycle. Such a plan for improvement required an understanding of what the key geological, geophysical, engineering and commercial activities were and how they interrelated. An activity network was therefore set up on a project planning computer programme, using a generic, medium-sized UK Central North Sea oil field "Nessie" to analyse how things were typically done at Phillips. A critical path through the Nessie network gave the "AS IS" cycle time and resources required. Using problem solving techniques the areas of potential time and cost savings were identified, solutions generated and a new "TO BE" network cycle, where the business process was optimised, was produced. Potential savings of over 30% on cycle time, 9% on manpower resources, 22% on development costs and an enhancement of over 500% on the Net Present Value of the project were achievable, by applying the key solutions. These include: the need for agreed objectives and strategies for every asset, setting up a multidisciplinary strategic planning group, a focus on key strategies, partner agreement on objectives and strategies, risk analysis for optimum decision making, continuous engineering, construction cost reduction and benchmarking. The Nessie network, being generic, is suitable for use as the basis for analysis and improvement of a wide variety of North Sea projects. The key solutions have been or are currently being implemented at Phillips.
Introduction It usually takes longer than expected from the time an oil company acquires offshore acreage to the time oil or gas is flowing from a field on that acreage. It had often been felt, within Phillips, that it took too long and that there was considerable scope for improvement. In 1991 Phillips UK addressed this problem as part of an overall Total Quality initiative aiming for continuous improvement in all aspects of its business. A problem statement was produced by the management team which stated that "it appears to take too long for PPCo UK to convert existing and potential acreage into profitable producing operations." A team was set up to look at the problem. Since the problem covered an area of the business where a number of disciplines interacted, the team was multidisciplinary in make up, consisting of geoscientists from both the Exploration and Development Groups, a reservoir engineer, a project engineer and representatives from the Engineering and Construction and the Commercial Departments. The initial task the team set itself was to establish whether the problem statement was true: did it take
Phillips longer than the best of its competitors to get fields in the UK North Sea on stream?
Benchmarking the performance To answer the question, the team commissioned a benchmarking exercise from public sources of information, e.g. the DTI Brown Book. Ninety-nine currently producing UK North Sea fields were analysed and segregated into oil and gas, by reserves size and by the type of production facility. The Phillips' Maureen oil field which has reserves of 212 million stock tank barrels (MMSTB) producing from a single fixed platform was part of a group of seven, 100 to 250 MMSTB oil fields, including Miller, Hutton and Clyde. The 4 trillion cubic feet (TCF), multiple platform Hewett gas field was compared with four similar fields over 3 TCF including Leman, Indefatigable and Frigg. The other Phillips producing fields, Audrey, Moira and Della, were likewise bracketed with their peer groups. Fig. 1 summarises the results for the study. Maureen took 2.5 years from block award to discovery, 5 years from discovery to a development decision and
Quantification and Prediction of Petroleum Resources edited by A.G. Dor6 and R. Sinding-Larsen. NPF Special Publication 6, pp. 261-271, Elsevier, Amsterdam. 9 Norwegian Petroleum Society (NPF), 1996.
262
P. McGaughrin, K. Ashton, N. Fuller, R.G. Heywood, R.W. Holt, B.L. King and S. White
Fig. 1. Field cycle times on the UK continental shelf, showing "best in class" and Phillips examples.
5.5 years from then to first oil, a total of 13 years for the complete cycle. One of the seven fields made its discovery within 2 years of award. Another appraised its discovery within 3 years and one developed it in 1 year. The best overall time for the complete cycle, however, was Conoco's Hutton field with a time of 11 years. This was 16% faster than the Maureen cycle. The Hewett field was quickest in its class in the appraisal phase taking just 1 year and quickest in class for the total cycle with 5 years. The Moira field was 86% slower then the best in class overall, taking 18 years through the exploration phase alone. Audrey was 28% behind the best in class in its group and Della 13% behind, despite individual best in class performances for Audrey development and Della appraisal and development. The lengths of time in the exploration phase spoiled the performance. The obvious conclusion from this comparison is that overall Phillips had taken too long, compared with the best of its competitors, to get projects through the cycle and that an improvement of at least 29% on average was feasible judging by the best historical performances. The data base, as should be expected, is not ideal and there are a number of anomalies. However, the general conclusion is thought to be sound and the 29% gives a reasonable target for improvement.
So the team could now agree on an objective: "to establish a genetic plan to convert existing and potential acreage into producing operations as quickly and profitably as possible." This was set against a 1991 UK Division target to do things 30% cheaper and 30% faster and the team agreed to aim at the 30% level. Identification of causes
To achieve the improvement, a standard problem solving route was taken, i.e. identifying causes of the problem, identifying solutions to eliminate the causes, then coming up with a plan to implement the solutions. Through the process of brainstorming a number of causes of the cycle delay were identified. The main ones are listed below: Poor data management. Too much time was being spent searching for and retrieving data. Lack of early focus on development. Exploration had occasionally been done in the past in an area or block at the expense of appraising a potential development. Misalignment of objectives. Different groups within the company had different ideas on what to do with a prospect or discovery and more importantly the overall objective was occasionally not agreed within the coventurer group.
Nessie: a process analysis of a generic North Sea field life cycle
No strategy for every asset. There was no clearly defined strategy for many blocks and prospects, and where there was one it was often not internally agreed. Incomplete risk analysis procedures. Risk analysis was used for project evaluation, but not fully within other parts of the company, as part of the decision making process. Poor post-project analysis. There was little documentation on what was done well and what was done poorly on wells or developments so that benefits could be passed on and lessons learned from one project to another. Complex contract bidding procedures. The use of simple "call off" and "single source" contracts, contract strategies and standard terms and conditions could be more widely applied. Project handover delay. Time was being wasted with rework and relearning as a project was handed over from one department to another. Overmanagement. There were too many layers of reporting. These were the reasons why the cycle took so long. How could they be overcome? What were the solutions?
Process analysis: Nessie It was still not clear to the team what all of the major steps were within each of the stages of the project life cycle. The geologist knew his part,
EXPLORATION
"~ ' I i I I I I I I I I I
COMMERCIAL
Ir
MILESTONES
263
and had a rough idea what the reservoir engineer did to turn the geologist's hydrocarbons in place (HIP) into a reserves figure and what the commercial representative did to turn that into an economic value. However, the overall flow and interrelationships of the key activities was not fully clear to him or to the team. It was felt that the team could come up with a better way of doing things if they understood what the key activities were at that time in the Company and how they fitted together. The approach the team took was to set themselves the task of acquiring, discovering, appraising and developing a generic oil field on their own and asking themselves what was involved in the process. The Nessie prospect was invented for the purpose, based loosely on a Phillips' block in the Outer Moray Firth. Specific premises were assigned to Nessie to make it realistic and to allow quantification of the time and cost elements. Nessie had Lower Cretaceous and Jurassic target reservoirs, the deepest at 11,000 feet, HIP of 250 MMSTB Oil and 250 BCF gas, reserves of 150 MMSTB oil and 200 BCF gas and would require two exploration and two appraisal wells to take it to a development decision. On the development side it would require a fixed drilling and quarters platform with ten production and six water injection wells and would have initial production rates of 65,000 barrels per day (BPD) oil and 20 million cubic feet per day (MMCFD) gas. A 30 mile gas pipeline and a 20 mile oil pipeline would be needed for export.
f I DETAILED II i REVIEW i ~ STRATEGY & .I.. -: ACREAGE I EVALUATIONi f :DEFINE AREAS; I I ACQUIRE i I I OF INTERESTI PURCHA,~'E I I I SPEC.DATA
I I I I I I I I
COMMERCIAL PREMISES
I ANALYSE i i ! ! ,-.I PROSPECTS, I...1 CORPORATE PARTNERS L ! SHORTLIST i l APPROVAL ~ APPROVAL ! BLOCKS, I i t
"1 t
I 8, PRopos,= I, PROCESSII PROCESS ,
IWRK PROGRAMi ECONOMICS FOR PROSPECTS (GENERIC i ASSUMPTIONS)
i DEn BLOCKS el ON OFFER ~1 I I[ ANNOUNCED I
DEV. GEOSCIENCE DRILLING
RESERVOIR ENGINEERING
PROSPECTS
DEVELOPMENT ENGINEERING
DEVELOPMENT "ASSUMPTIONS "
PROFILES FOR
EX
t
3
Fig. 2. Extract from the Nessie "AS IS" exercise, showing the Acreage Evaluation phase.
P McGaughrin, K. Ashton, N. Fuller, R.G. Heywood, R.W. Holt, B.L. King and S. White
264
With Nessie thus described, what activities would have been required typically in 1991, to progress the cycle? What is the "AS IS" situation. Each of the team
came up with their discipline's activities and the team on a whole attempted to work out the sequence and interdependence of the activities.
I (DISCOVERY) I I ESTABUSH H.,.P. ' ~ AND ASSESS } COMMERCIALffY
EXPLORATION
it
I I I I ~ ECONOMICIS (GENERIC ASSUMPTS)
COMMERCIAL
. . . .
I
MILESTONES
i
EXPORT ROUTE SCREENING
PRELIMINARY SALES DISCUSSIONS
| p -. m ~-i
I I HANDOVERI q TO I DEVELPMNTr ~ I GEOSCIENCE II I_GROU..P_ I I
iI I
I DISCOVERY ~
I
t
I I I I
PRELIMINARY EXPORT DISCUSSIONS
j
SEISMIC
PARTNERS| I
I I I
IC ....
DEVELOPMENT GEOSCIENCE
=
W~H p,
I I I
.
t I I
I PROCESS, IGEOLOGICAL
tq . . . . . . .
t.
.
.
.
.
.
.
.
.
_l. . ,
SEISMIC
c~ L"IEVALUATION L |. ||v| r'nrnr | I' I" HiP I 9 SEISMIC ,, i "'" , . . . l i
I
[ .
DRILLING RESERVOIR ENGINEERING
PROFILES FROM DISCOVERY
WELL
EX6
1
EX7
I
"P'
I
AP 2
Fig. 3. Extract from the Nessie "AS IS" exercise, showing the Early Appraisal phase.
APPRAISAL
ACQUISITION / EXPLORATION EX1 LICENSE R O U N D STRATEGY/PRELIM. EVALUATION
EX2
EX3
EX4
DEn BLOCK ANN.
DETAILED BLOCK EVALUATION
DEn BLOCK AWARDS
EX5
EX6
DRILL FIRST EXPLORATION WELL
DRILL DIS. WELL
EX7
AP1
POST GEN/AGREE DISCOVERY APPRAISAL ANALVSm STRATEGY
AP2
AP3
APPROVE DRILL 1st WELL
AP4
EVALUATION STRATEGY
AP5
Znd APPRAISAL i UPDATE WELL / EVAL.
L
9
COM
L.oJ-
I I
I I I I I
- i l l DEV GEO DRLG
I
r>
!
'
I I
I J
I I I I ,i,
I I
I I
I I
I I ,
FI
I
I
I I
I ,
,
Ii
I
r
p
-
I
o, '~%1,
I-I
I I
I I
"
"
;I L . , _ ~ l L - o . - - /"l
RES ENG
9 4 J
Fig. 4. Complete Nessie "AS IS" network flowchart. The grey dots show the extracts in Figs. 2 and 3. The dashed line shows the critical path.
265
Nessie: a process analysis of a generic North Sea field life cycle
Nessie "AS IS"
the total Nessie cycle; the dots or nodes represent the activity boxes shown in Figs. 2 and 3. Related activities are grouped together into "sub processes" which allow the cycle flow to be more easily followed. EX1 to EX7 take the prospect through the Exploration/ Acquisition phase from initial licence strategy generation to post-discovery analysis. AP1 to AP6 chart the appraisal phase and D1 to D3 cover development. The dots show those areas described in the excerpts (Figs. 2 and 3), and the dashed line is the critical path.
Fig. 2 shows a simple example of a group of related "AS IS" activities. After the UK Department of Energy (DTI) announces which blocks are on offer, the Exploration Department refines the strategy and identifies the areas of interest. This takes 8 weeks. They then spend 12 weeks conducting detailed acreage evaluation by identifying and quantifying prospects. The HIPs are given to the reservoir engineer to generate profiles which are given to the Commercial Department for economics analyses. The Exploration Department will then complete the prospect analysis and submit a shortlist of bid blocks to management and partners for approval. Fig. 3 shows an example from the early appraisal phase of the cycle. After the discovery well the process is similar to the pre-acquisition stage above. If the project is probably commercial, it would be handed over to the Development Geoscience Group who would then agree an appraisal strategy and in the case of Nessie, shoot a 3D seismic programme prior to any further drilling. There are just two extracts. The whole activity flow chart was about 20 feet long and consisted of over 140 activities. The project planning software "Artemis" which is used internally for field project management was used, on the suggestion of the project engineer, to network these activities. Fig. 4 shows schematically the Artemis product for APPRAISAL
Another technique introduced from the project management area was the concept of cost, time and resource analysis (CTRs). Each activity was examined and quantified in terms of what it costs, how long it takes and what manpower resource is required to complete it. In that way, the activities could better be understood by the team and scope for improvement quantified. Each discipline completed CTRs for their activities. An example is shown in Fig. 5. This exploration activity is referenced EX045 and is done within the Exploration Department and can be seen as one of the activity boxes in Fig. 2. The questions that are asked in the CTRs are: What INPUT is needed to allow the task to be done and who is the SUPPLIER of that information? What does the PROCESS or task consist of?
DEVELOPMENT
AP5
AP6
UPDATE
DRILL 2nd APP.
EVAL.
Cost, time and resource analysis
D
D 2
1
PREPARE DEVELOPMENT
SELECT DEVELOPMENT
PROJECT EXECUTION
PLAN AND BUDGET
OPTION
WELL
D 3
EXPL.
r I~
I
I
l
|u
L._.-J
-J
---_-
i
9.. ! l.= J. . IL= . _
I
I
I
I I I |
A
I 9
A
! I
~e-.
19
I I I
;-
I__
!
~ :/o/
LA~,~
I I I
!_LJ-
I 9
_
,, ~
~
I ' ~,=
/
!
!_
I
I
, I I I I I I
I I I i !
I I I I n I
:-
I_1 ;';
._--
_ :
:F
',_ ya= _J l_
DEV GEO DRLG
I
~1 ~1
RES ENG
I
DEV ENG
I I I I
E&C
P McGaughrin, K. Ashton, N. Fuller, R.G. Heywood, R.W. Holt, B.L. King and S. White
266 ACTIVITY
(ref no. ) EX045
9
CYCLE TIME
( weeks )
912
TITLE"
DETAILED ACREAGE EVALUATION
INPUTS
Selected blocks, spec seismic data, regional studies ( G&G )
PROCESS
Block mapping, regional block geolical evaluation,prospect identification, regular/continual partner review.
OUTPUT
List of blocks with prospects considered, significant for detailed evaluation.
SUPPLIERS
G & G ,Exploration services,CoVenturers,? External Consultants, Contractors.
CUSTOMERS
G & G Commercial,Reservoir Engineering (CM030) (RE000) COST ( s
RESOURCES
(inc ext. manpower) MAN WEEKS (time written)
POTENTIAL FOR SAVINGS ( Total savings using final list of solutions )
s 528,000 110 RESOURCES
CYCLE
Internal Study Costs Manpower & ss
TIME
0
Low
( Fixed )
VALUE
Specific value to the project
High
Consequence of failure
Poor Acreage
COMMENTS
Part of Detailed Evaluation SubProcess ML1 - > ML2
Fig. 5. Example of the Cost, Time and Resource Analysis (CTR) for the Nessie "ASIS" evaluation. This example is for exploration activity. What is the OUTPUT or product of that task and who is it given to, i.e. the CUSTOMER? What RESOURCE does the activity need in terms of costs and manpower? The VALUE of that activity and POTENTIAL FOR SAVINGS are also reviewed. The cycle time and resource costs were put into Artemis to allow the process to be measured. Solutions
At that point the "AS IS" network had been completed, showing the current situation, each individual activity had been analysed by the CTR evaluation and
the principal causes of delay had been identified. The team was then able to focus on how best to overcome the obstacles to compressing the cycle. Through brainstorming dozens of solutions were identified. These were prioritised and grouped. The main ones are listed below as recommendations. Establish and agree a clearly defined objective, strategy and plan for every asset. Teams need to be set up to define objectives, strategies and action plans for every block and prospect. The management team would approve the plans and align them with divisional objectives, strategies and budget. The asset team would then gain partner concurrence and incorporate the plans in an Asset Portfolio.
Nessie: a process analysis of a generic North Sea field life cycle Set up a multidiscipline strategic planning group. This would facilitate the coordination of Exploration, Development and Exploitation Department planning and the alignment of these plans with divisional and corporate planning. Optimise the use of funds by focusing funds on key strategies. Funds are rarely available for all assets at any given time. By defining strategy priorities based on expected values, funds can be focused on maximum value added projects. Implement standard risk analysis procedures. This will allow management to look at all projects and prospects on a comparative basis and make better decisions at the key cycle steps. Establish and maintain a continuous sequence of engineering. Select a development concept at the earliest stage and establish a continuous engineering sequence from them to the completion of detailed design to provide the maximum degree of design/cost definition at critical project phases and decision points. Reduce construction costs. This would be done through better construction strategies, innovative fabrication methods and by maximising the use of new fabrication provinces. Set up benchmarking initiative. This would involve an identification and measurement of Phillips key business processes, an identification and measurement of top performing companies in these processes and an identification of practices that enable those companies to perform so well.
EXPLORATION
"I II I. I I I I I I I I I II
COMMERCIAL
MILESTONES
267 Implement a database management system covering all departments requirements. Significant time savings can be made by having an integrated, relational database of wells, prospects, reserves, economics, budget and licence information, having it stored just the once, updated only by an appointed representative and with access on line for all teams. A total of 23 solutions were recommended and for each, a detailed implementation plan was developed. These listed the problem and the solution, documented the plan to implement the solution, covered the cost, timing and responsibility for implementation, quantified the savings and addressed the benefits and concerns.
Nessie "TO BE" The team then looked again at the activity network and asked itself what that would look like if all of the solutions were incorporated. What is the optimum way of moving Nessie through the project cycle? The network was then redone as it "SHOULD BE" or as it was "TO BE". Fig. 6 shows an extract from the "TO BE" cycle modified from the "AS IS" cycle extract shown in Fig. 2. The major difference is, now that the team has been set up early, the fight people are involved; rather than the Commercial and Exploration Departments being responsible for development premises, these
rI IDENTI~-: r ;E~AI-LE-D-'I f-q AREASOF t'" -'1 ACREAGEt= I ==zINTERESTDEFI &Ill NE IIEVALUATIONI ..I ,_ . S,.TEO,. . . . . j I I I I
I I zi- ...... ANALYSE !II CORPORATE;...j PARTNERS] ]'7, PROSPECTS,~ APPROVALi: APPROVAL li SHORTLIST ! : PROCESS Ii L PROCESS: Ii . . . ..,oc . . . s
ECONOMICS FOR PROSPECTS (GENERIC ASSUMPTIONS
I
DEn BLOCKS II ON OFFER ~1 :ANNOUNCED :
DEV. GEOSCIENCE DRILLING RESERVOIR ENGINEERING
PROFILESJ
FOR PROSPECTS
EX 3 Fig. 6. Extractfrom the Nessie "TO BE" cycle for the AcreageEvaluationphase (cf. Fig. 2).
P. McGaughrin, K. Ashton, N. Fuller, R.G. Heywood, R.W. Holt, B.L. King and S. White
268
EXPLORATION
COMMERCIAL
I EVAL.H.I.P.I
--~ I ---I I I I I I I I
/
I
I.
1
H
! .r- ;E~i''; MILESTONES
(,.CLUOE TEAM
r
,t
PRODUCEL--J ECONOMICSI |I
H
-~
~L) PREUM' I .~ UPDATED SALES/ L rXPORT / COMMERCIAL ISCUSSlONS / PREMISES
II
COMMERCIALly'~ PREMISESI
DISCOVERYLa WELL -" ~
Is..Bs_MLc GO~
; ECO.O.,C; I. . . .
1
1
I
II .rgrrAff ql I A~I~SAL :1
DEVELOPMENT GEOSCIENCE
--
AGREE ;.SEISMIC ' 13DPROGRAMI
NEGOTIATE SALES/ EXPORT H OFA
~'G~E"] I APPRAISAL, r
:"~I~S~L" I I t'l STRATEGY~ I; AGREED , I
I' AGREE 3 C'l APPRAISALI'~ I ; WELLNo,1 , I
I
I
_L'_----" . . . . . . . I
I; I . APPROVALS .... .I I~I LI STRATEGYU_ I STRATEGY: J I. I (INCLUDINGiJ (EXTENDED,~" -I1( MGT/CV's) I PERMISSION) I i iEST1NGr',l I CONSIDER1, I APP.WELL11 IDETRACL?tI. IPHASEI)DEV; ...... . . . . . . . . . . . . . . . . . . EAN~ "; ~_N~TT "ri. . . . . . . ON SEISMIC;......i PROCESS& ~1 WEATHER-- i INTERPRET'1 I WINDOW I SEISMIC I I
_
l
•
; 'I
I I
I i IL . . . . .
,, ! ~'~.irL-.
APPRAISAL'1 - I WELLAND ".
DRILLING
RESERVOIR ENGINEERING DEVELOPMENT ENGINEERING
~1
_•
1,1,1Zre IJJ
rua ~ - ~ ~
~.~1 ~:~a. I IN~ Ima.
DERIVE PROFILES
..]DERIVE / DEVELPMNTE~, PREMISES|
LSUSPENDJ
IN
~<142 ~
EARLY
AP 0 / 3
PRODUCTION
OPTION
AP 1
AP 4
Fig. 7. Extract from the Nessie "TO BE" cycle for the Early Appraisal phase (cf. Fig. 3).
are generated by the project engineer on the team. Fig. 7 shows the "TO BE" network portion which was covered in its "AS IS" condition in Fig. 3. This time, however, the team is already in place and because of pre planning, there is (1) a strategy agreed with management and partners, (2) a contract option on a 3D seismic boat in the event of a discovery, and (3) agreement in principle to evaluate the need for a long-term production test and/or the feasibility of early production. The project team can therefore go straight into action (1) to asses the economic viability, come up with an appraisal strategy and progress the appraisal drilling, at the same time as (2) finalising the 3D programme and giving the boat the go-ahead to shoot the programme and (3) reviewing the feasibility, strategy and premises for a long-term test/phased development. Fig. 8 shows an extract from the project execution phase of the development cycle. After the Annex B (Government consent) and AFE (Development Plan and Budget) have been approved and detailed design is complete, all the contracts for the platform, template and pipeline can be bid, evaluated and awarded, followed by fabrication and installation. The complete "TO BE" network is shown schematically in Fig. 9. The exploration and development
activities have been optimised but the structure of the process is largely similar to the "AS IS" of Fig. 4. Where the structure changes is in the appraisal phase, where a number of activities are done in parallel rather than in series. The 3D seismic is shot soon after the discovery is made and if the reservoir definition is reasonably good, a follow-up well could be drilled on a near trace processed or partial 3D data set. The above sequence is somewhat out of date since 3D is now commonly shot prior to drilling the exploration well. This would significantly reduce the chance of drilling it in the wrong place and would compress the cycle even more. In parallel with these activities, the team would embark upon a preliminary evaluation of a long-term production test/phased development option. Again, this would only work in certain circumstances, e.g. where reservoir continuity is a problem. Risk analysis would be used to assess the various uncertainties and to decide on the optimum route. This is only one of a number of scenarios that could have been chosen to suit a number of different circumstances. The principles are the same, however. Planning, teamwork and risk management can significantly accelerate a project life cycle. If early production is feasible, it can be implemented
269
Nessie: a process analysis of a generic North Sea field life cycle
COMMERCIAL
ANNEX B APPROVED& PROJECTGO!
MILESTONES
I
DETAIL i DESIGN COMPLETE
DEVELOPMENT GEOSCIENCE
B.E.A. DRILLING RIG CONTRACT
DRILLING RESERVOIR ENGINEERING
DEVELOPMENT ENGINEERING
APPROVAL OF AlE
ANNEX B
SUBMISSION
I F~176176
q
"~'~-11
ENGINEERING AND
..J
CONSTRUCTION
II /
.,ji
B.E.A. I TEMPLATE FABRICATION J
B.E.A. I
I FABRICATE I i i ... 1 IP'='.PROCURE
1 I.
'. ENGIINEERINGI I !PHASE 3 1750 r ~ -" Pi FABRICATIONI
~_c.M~JLI
q
CONTRACTORI
SULK STEELS
I TEMPLATE I
I
ENG /DESIGN I I ENG.PHASE4 I I (75-1oo%) I
iL-lh
I;[ .~T,O. I_ . I FABRICATIONI t. . . . . 1
. . . . . . .
I PROCURE 111 . . . . 36-/ "~ LONG LEAD ;-'[ . . . . . I. rrEMS B.E.A. [ ONSHOREENG.] INSTALLATION - PLATFORM I INSTALLATIONI CONTRACTOR AUTHORITY APPROVALS
ND 3 Fig. 8. Extract from the Nessie "TO BE" cycle showing the project execution phase of the development cycle.
through a sub-sea tie in or, in the case of Nessie, through a floating production system and be on stream prior to selection of the main development concept. Production from the completed discovery would continue while the main development is being planned, approved and executed and would be demobilised as late as immediately prior to the main productions start up.
Schedules The relative timing of the "AS IS" and "TO BE" cycles is shown in Fig. 10. Appraisal drilling is complete in the "AS IS" cycle 8 years after the block has been awarded. This is reduced to 4 years in the "TO BE" cycle. The full "AS IS" cycle takes 14.5 years; 10 years in the "TO BE" case. (In the early production scenario the field can be on stream in 4 years.) The main time savings are in the Exploration and Appraisal phases and are due largely to having the optimum strategy, having it agreed by management and partners and having the team in place to implement it.
Time and cost savings The cycle compression from 14.5 to 10 years represents a time saving of about 32% (Fig. 11). There
is a corresponding reduction of 9% in manpower resource and a cost reduction of about 22% from s MM to about s MM.
E c o n o m i c impact Fig. 12 shows the cumulative cash flow of the Nessie project for the three cases. The net present value (NPV) of the "AS IS" case is around s MM, measured from block award. The "TO BE" case has an NPV of s with most of the increase due to production being 4 years earlier. The "TO BE" phased development case has a value of s MM, again primarily by producing earlier. The revenue from early production is used to offset the major costs of full scale development. This shows the sizeable opportunity to increase a project's value through time compression.
Results/conclusions The Cycle Compression team reported its findings in April 1992. Since then most of the key recommendations have been or are being implemented within Phillips. A Strategic Planning Group has been set up; an Asset Register has been established with
270
P. McGaughrin, K. Ashton, N. Fuller, R.G. Heywood, R.W. Holt, B.L. King and S. White
Fig. 9. Complete Nessie "TO BE" network flowchart. The dashed line is the critical path. Grey indicates early production scenario.
DEn AWARD BLOCK DRILL 1st WELL
J
I
41~lb
AS I
DRILL DISC. WELL
i
DRILL 1st APPRAISAL DRILL 2nd APPRAISAL
I
I
15 I
41~lb
AFD APPROVAL
m
PROJECT GO FIRST PRODUCTION
I
DRILL 1st WELL DRILL DISC. WELL DRILL 1st APPRAISAL
m I
TO
m
AFD APPROVAL
nn
PROJECT GO
m
FIRST PRODUCTION
END
mmmm mmmm mmmm
t m
||
m
S T A R T E. PROD'N
lie Iill
DRILL 2nd APPRAISAL
z r 0 m
I
E. PROD'N
YEARS
13
!!
Fig. 10. Relative timing of the Nessie "AS IS" and "TO BE" cycles.
goals, strategies, work programmes, schedules, HIR reserves, economics, risks, budgets and licence information for all assets; a Database Management System developed on Oracle for wells by Phillips Norway is being extended to cover all key database items, including those in the Asset Portfolio; Exploration Risk Analysis is now standard practice; and a Direct
Gas Marketing venture is in place between Phillips and Southern Electric. In addition the recommendations have been incorporated in current development projects' strategies with significant cost savings to date. It is too early yet to identify proven time savings. Finally, what is the value of Nessie as a tool.
271
Nessie: a process analysis of a generic North Sea field life cycle
Fig. 11. Timing and cost savings between "AS IS" and "TO BE" cycles.
As a process analysis of a company's core business, provides a basis for: (1) Management and teams to understand their business better and to improve it; (2) More detailed departmental or functional key business process analysis; Nessie is still a fairly high level network of activities but it does provide a framework to analyse at specific areas of the cycle or individual disciplines in greater detail; (3) Measuring performance in any of the activities, sub-processes or phases of the cycle to see how well or poorly it is being done and to measure improvement; and
(4) Benchmarking, to compare the companies performance with that of the "best in class" and to improve the performance by embracing best practices.
Acknowledgements The authors would liked to thank Phillips Petroleum Company United Kingdom Limited for permission to publish this paper and A.J. Keamey for helping to keep the project focussed. Special thanks are also due to Geoff Oglethorpe for his planning input and to the Phillips drawing office for many hours of high-quality drafting.
272
P. McGaughrin, K. Ashton, N. Fuller, R.G. Heywood, R.W. Holt, B.L. King and S. White s 200 "0
..--'" P H A S E D
c :3 0
100 -
TO
SSS
9
BE AS
!
O
/
m
r s sNN
_x.
x
. . . . . .
l
t_
I .....
IS
.~.,~'
0
,,.J I_L
.....
100 -
Z
-200
I 1992
I 1995
,,9
I 2000
\-.C.xji I 2005
i ./"
I 2010
I 2015
YEAR Fig. 12. Nessie economics, expressed as cumulative cash flow versus time, comparing "AS IS" and "TO BE" cycles.
R MCGAuGHRIN K. ASHTON N. FULLER R.G. HEYWOOD R.W. HOLT B.L. KING S. WHITE
Phillips Petroleum Company UK Ltd., 35, Guildford Road, Woking, Surrey GU22 7QT, UK Phillips Petroleum Company UK Ltd., 35, Guildford Road, Woking, Surrey GU22 7QT, UK Phillips Petroleum Company UK Ltd., 35, Guildford Road, Woking, Surrey GU22 7QT, UK Phillips Petroleum Company UK Ltd., 35, Guildford Road, Woking, Surrey GU22 7QT, UK Phillips Petroleum Company UK Ltd., 35, Guildford Road, Woking, Surrey GU22 7QT, UK Phillips Petroleum Company UK Ltd., 351 Guildford Road, Woking, Surrey GU22 7QT, UK Phillips Petroleum Company UK Ltd., 35, Guildford Road, Woking, Surrey GU22 7QT, UK
273
Changing perceptions of a gas field during its life cycle: a Frigg field c a s e study Eivind Torheim
Well 25/1-1 was drilled in 1971 and discovered what is presently known as the Frigg field. This was the first giant gas field that was discovered in the northern North Sea, and it is located on the boundary between UK and Norway. The field was declared commercial in 1972 and came on stream in 1977. In total 48 development wells were drilled from two platforms located on opposite sides of the border line. The gas has been transported to the UK gas market via the St. Fergus land terminal. The production has constituted a significant part of the British gas consumption, providing as much as 45 million Sm 3 of gas per day during plateau production. After a start-up phase of two years the production was kept at plateau for seven years, before a relatively long decline period started. Following the discovery, extensive work was performed through all phases of the life cycle of the field, covering a wide range of disciplines within petroleum technology. Ultimately the work always had the same main target: to assess quantities and distribution of hydrocarbons and optimise production profiles and economy. A major part of the uncertainties on Frigg has been, and still is, related to the aquifer in the area. Accordingly much of the monitoring effort has been connected to this. The main objective of this paper is to summarise the impact of reservoir monitoring results on perceptions and reservoir management of the Frigg field, where evolution in technology during the lifetime of Frigg plays an important role.
Introduction The Frigg field is a giant hydrocarbon accumulation, discovered mid-1971 by well 25/1-1 drilled by the semi-submersible rig Neptune P 81. The field is usually regarded as a dry gas field. However, in a thermodynamic sense it is a retrograde gas condensate initially overlying an oil disc in the reservoir. The condensate yield was about 4.3 g/Sm 3 at initial conditions. The field is located approximately 190 km westnorthwest of Haugesund, Norway, 180 km east of the Shetland Islands, and 390 km northeast of Aberdeen, Scotland, on the border between Norway and United Kingdom (Fig. 1). The water depth in the area is about 100 m. On the Norwegian side the main part of the field lies within block 25/1 with a small extension into block 30/10 to the north. On the United Kingdom side the field is mostly located within the blocks 10/1, 10/6 and 9/10. It also touches block 9/5. The field's location on the national borderline has complicated the allocation of interests, and international negotiations and agreements between Norway and United Kingdom have been necessary. The field is now unitized with Total Oil Marine plc. and Elf Enterprise Caledonian as owners
on the British side and Statoil, Norsk Hydro, Total Norge A/S, and Elf Petroleum Norge a.s. as owners on the Norwegian side. Elf Petroleum Norge a.s. is operator. The field is considered to be a submarine fan with mass transportation from the Shetland platform located to the west-southwest of the depositional area. This explains the "bird's foot" configuration as displayed by the structural mapping. The Frigg Formation is of Early Eocene age, with the upper part constituting the hydrocarbon-bearing interval. Below is the Balder Formation followed by the Paleocene formations Sele, Heimdal/Lista and Maureen before reaching the Cretaceous. It is probable that all the formations between upper Frigg and Cretaceous contribute, to various extents, to the aquifer impact on the Frigg field (Fig. 2). The geological setting is given in detail by Brewster and Jeangeot (1987). The field covered an area of slightly over 100 klTl2 at the initial gas oil contact. The maximum gas column was 160 m, overlying an oil disk with an average thickness of 8.6 m. The reservoir was full to spill point with the initial gas oil contact at - 1 9 4 8 m MSL. Middle Eocene marine shales create the seal. The Upper Frigg Formation consists of a rather unconsolidated sand interbedded with shale layers
Quantification and Prediction of Petroleum Resources edited by A.G. Dor6 and R. Sinding-Larsen. NPF Special Publication 6, pp. 273-289, Elsevier, Amsterdam. 9 Norwegian Petroleum Society (NPF), 1996.
274
E. Torheim
Fig. 1. Frigg area location.
and calcareous stringers. The petrophysical properties are generally very good with porosities ranging from 27% to 32% and permeabilities from 1 to 4 darcies. The initial static gas reservoir pressure was 197.9 bar at the chosen reference level of -1900 m MSL and the initial aquifer pressure (Sele/Lista Formations) was found to be 223.4 bar at -2191 m MSL. The present estimate of original gas in place is 238.109 Sm 3, of which close to 184.109 Sm 3 are considered recoverable. The field was declared commercial in 1972 and the development plan was finalized in 1974 following the signing of the first sales contract with the British Gas Corporation for gas from the British part of the field, and the agreement of the Norwegian government for the sale of the Norwegian gas. The concept of both British and Norwegian involvement was incorporated in the development plan. The quarters platform (QP) was the first installation in the field in 1975, located on the British side. Two 24-slot drilling and produc-
tion platforms (CDP1 and DP2), one on each side of the border line, were installed in 1975 and 1976. For treatment of the gas two processing platforms (TCP2 and TP1) were sited in 1976 and 1977, again one on each side of the national boundary. The treated gas is transported to St. Fergus in Scotland through two 32" sea-lines. The pipelines were installed between 1974 and 1976 and hooked-up during 1977 and 1978. The St. Fergus terminal and the sea-lines, including a manifold and compression platform between Frigg and St. Fergus, are operated by Total. Commercial deliveries of Frigg gas started in September 1977, and the field has now been in production for 17 years and has delivered about 180.10 9 Sm 3 of gas to the British market (Fig. 3). Throughout the lifetime of the field new dynamic information has become available more or less continuously, modifying the understanding and the image of the field to various extents. This paper is predominantly about the development of the comprehension of the aquifer
Changing perceptions of a gas field during its life cycle: a Frigg field case study
275
Fig. 2. General stratigraphic section.
activity on the field and the consequences for field management.
First period - - discovery (up to 1971) In 1965 a seismic survey was performed, covering the Frigg area with a grid of 15 km by 22 km. The interpretation of these data revealed an interesting structure at a depth beginning at the top of the Upper Cretaceous chalk horizon. New seismic was shot in 1968 and this information was used in 1969 for selection of blocks in the second round of licence application. This was when the production licence 024, coveting Block 25/1, was awarded. New seismic was acquired and the 25/1 area was covered by a 5 km x 5 km grid. This set of data led to the definition of the Frigg structure at several levels, particularly at
the top of the basal Tertiary sands. In 1970 licence P 118, including Block 10/1 on the British side covering the western part of the structure, was awarded. From March to July 1971, 6 years after the first indications of a possible hydrocarbon accumulation in the area were obtained, the discovery well 25/1-1 was drilled. The well encountered a 128.8 m thick gas column in sands of Eocene age, later defined as the Frigg Formation, with 6.5 m of oil bearing sands below. From the oil-water contact to TD, 96 m deeper, all sand intervals were water-bearing. Below the Frigg Formation the more laminated and tuffaceous Balder Formation was penetrated. It was considered likely that the tuff could represent a barrier between the Eocene intervals and the Paleocene formations, and thus isolate a local Eocene aquifer in lower Frigg and upper Balder from a regional aquifer in the underlying formations.
Fig. 3. Production history. (DCQ = Daily Contract Quantity.)
277
Changing perceptions of a gas field during its life cycle: a Frigg field case study Table 1 Summary of simulation results Period
Study
OGIP ( 109 Sm 3)
Recovery (%)
Reserves ( 109 Sm 3)
1
Elf, 1972, without water drive Elf, 1972, with moderate water drive
270 270
78.5 85.6
212 231
2
Elf, 1973 Franlab/Norsk Hydro, 1973 GCA/BGC, 1973 DGMN and official reserves, 1977
228 287 229 265
75 70 82 81
171 200 188 215
3
Elf, including satellites, 1979 Elf, perm. window, Sgr = 19%, 1984 Elf, perm. window, Sgr = 29%, 1984
265 265 265
86 82 76
227 217 202
4
Elf, Elf, Elf, Elf, Elf,
early 1985 late 1985 1987, base case, Sgr = 0.29 1987, Sgr = 0.24 1987, max OGIR Sgr = 0.29
265 256 223 223 246
55.8 67.6 75.3 79.4 77.6
148 173 168 177 191
5
Elf, Elf, Elf, Elf, Elf,
March 1989 November 1989 1990 1991 1993, no artificial aquifers
235 235 235 238 238
75.7 76.6 76.0 76.9 77.3
178 180 181 183 184
Some very early studies of accumulation and recoverable reserves were performed based on well 25/1-1 and seismic results. The studies arrived at a figure of 270-109 Sm 3 of original gas in place (OGIP). Several uncertainties for the reserves were naturally recognised at the time, among which the possible impact of an underlying aquifer was very important. Two different cases were studied, one with no water drive and a second with a limited homogeneous water drive. The ultimate recovery was limited by well head pressure in both cases and a residual gas saturation (Sgr) of 20% was assumed behind the water front in the case of water drive (see Table 1, period 1).
Second period - - appraisal and development (1971-1979)
Appraisal drilling During the first couple of years following the discovery in 1971, five appraisal wells were drilled and additional seismic was acquired to study the crest of the field. Well 25/1-2 confirmed the presence of hydrocarbons in the northern extension of the field, and the maximum eastern extension was verified by well 25/1-3 (Fig. 4). On the British side well 10/1-1 was drilled with the intent to appraise the western area of the field. Well 10/1-2 was the first appraisal well not encountering any hydrocarbons, but it delineated the western extension of the Frigg structure. Well 30/10-1, drilled by Esso, assessed the maximum northern extension of the structure.
Coring The reservoir was cored in most of the wells. However, due to the unconsolidated nature of the sand combined with the limitations of conventional coring techniques, it was difficult to recover representative samples. Still, due to concerns about the possibility of water encroachment into the hydrocarbon bearing sands, studies were performed on the available material at lab conditions in order to evaluate residual saturations behind water. The results showed a residual saturation of 25% for the oil and 19% for the gas when swept by water.
Studies During this first appraisal phase several evaluations were made, more or less in parallel, to assess the accumulation and reserves of the Frigg field. In 1973 Elf reviewed its previous studies, incorporating the new data. The main change was a reduced OGIP from 270.109 Sm 3 to 228.109 Sm 3. The most probable ultimate recovery was considered at the time to be 75%. Franlab completed a three-dimensional simulation study in 1973. The work was based on a structural map made by Norsk Hydro, giving an OGIP of 287.10 9 Sm 3. The regional aquifer was separated from the Eocene aquifer by a layer with vertical permeability of 2 mD. This resulted in a projected recovery of 70% and a pressure drop of 108 bar in the gas after 15 years of production. Although this was then considered to be a strong water drive, the
278
E. Torheim 441 9O 6655000
|
l , l
|
,,,
#5ooo,
, , ,
, , , 4 , 5 o , o 0 0 , , ,
m
ssooo,
, ,
, , ,
, ,
,s
i
6650000-
6645000-
o
1
6640000-
..
(~
I
6635000-
9 CDPI.weUs (abandoned)
~ / /
9 DP2-weUs 9 Observation.wells m
, 6630000 4400O0
I
'
I
'
I
'
I
'
I
'
445000
I
'
I
'
I
'
I
'
I
'
450000
I
'
I
'
I
'
I
'
I
'
455000
I
'
I
'
I
'
I
|
460
Fig. 4. Field outline and well locations.
model predicted the production to remain water-free. Another assessment of reserves was made by Gaffney, Cline and Associates (GCA) on request from the British Gas Corporation. Their report confirmed the ability of the Frigg reservoir to fulfil the contractual requirements between the operators and the British Gas Corporation. The proven OGIP was calculated to be as low as 227.109 Sm 3, but with a high recovery of 82%. It was concluded in the report that the exact recovery was dependent on the strength of the water drive, and that the most probable scenario was a limited water influx. The results from the
studies performed in 1973 after six wells were drilled on the structure, are summarized in Table 1, period 2.
Further evaluation The unitization of the field between the two countries was a controversial matter. In 1973 it was mutually agreed by the companies and the governments to appoint DeGolyer and MacNaughton (DGMN) as an international expert to evaluate the volume of gas in place and the division of the ownership. The available seismic information was considered insufficient and
Changing perceptions of a gas field during its life cycle: a Frigg field case study
a new detailed seismic survey was performed during the summer of 1973. After evaluating the new seismic data, DGMN still deferred their results. Three more appraisal wells were required in areas where the seismic horizons had been questioned by various partners. The wells 25/1-5, 10/1-3 and 10/1-4 were then drilled. In addition, well 30/10-5 to the north provided information from the Frigg reservoir although its main objective was to explore Jurassic intervals (Fig. 4). DGMN published their results and conclusions in 1976, indicating that the OGIP of Frigg was 265.109 S m 3 (Table 1, period 2), of which 60.82% was located on the Norwegian side. The split has remained unchanged to present in spite of being re-definable every four years. In addition to the discovery and appraisal wells located on the Frigg field itself, five wells were drilled to appraise the deposits on the extensions of the Frigg structure and to investigate the relations between possible satellites and the main Frigg structure. Wells 25/2-1 and 25/2-2 were drilled on East Frigg Alpha and Beta, respectively, 25/1-4 explored the North East Frigg accumulation, and the wells 30/10-2 and 30/10-3 drilled by Esso proved the existence and independence of the Odin field (Fig. 1).
Field interpretation The general picture of the field before production start-up in 1977 (Fig. 5) can be briefly described as follows: - The reservoir was homogeneous. Shale and limestone units were considered to be local in extent.
279
- The "local" aquifer in the Eocene sands, immediately below the hydrocarbons, was separated from the underlying regional Paleocene aquifer to an unknown degree by the tuffaceous Balder Formation. - T h e limiting factor of the recoverable reserves would probably be the pressure depletion with a minimum well head pressure of 65 bar. However, interpretation of the field still suffered from the following problems: - T h e average areal coverage per well was still more than 10 km 2. - V e r y poor recovery from conventional core barrels. - Q u a l i t y of seismic data not ideal for lithological description. - P e r m e a b i l i t y barriers impossible to locate under static conditions.
Development drilling The development wells on the field were drilled from 1976 to 1979. Due to the very good flow properties of the rock/gas system of the Frigg reservoir, all production wells were located on the apex within an area of 5 km 2 (Fig. 4). As the possibility of an active aquifer had been recognised, only one third of the gas bearing sand column was penetrated by the 47 production wells. The 48th well, 25/1-A22, was drilled through the Eocene and completed in the Paleocene formations to the northeast of the well cluster (Fig. 5). The aim of this well was to monitor pressure and water contact evolution in the regional aquifer, again as result of the identified potential of an
Fig. 5. Simple geological model.
280
E. Torheim
P/Z (bar) 250 I 200 150 |January 1979 J 100-
~ ~
~ ~ ~
[ January 1983 J
50[ OGIP'26S BSm3 I 0
0
I
I
I
1
I
l
1
I
20
40
60
80
100
120
140
160
Gp (BSm3)
180
Fig. 6. P/Z-aquifer support.
active aquifer. Unfortunately the gas-liquid contact coincided with a thick shale, preventing observations of any liquid rise for several years. Due to the concentrated pattern and limited penetration of the development wells no significant new geological information was obtained to change the existing image of the field. Some shaly layers and calcareous stringers were encountered in the wells, but as no correlation was possible nor predicted by the depositional model, it was concluded that they were of limited extent.
Third period - - first years of production (1977-1984) The need for a new simulation model to evaluate various aspects of the gas satellites in relation to the Frigg field became apparent. In 1979 new simulation studies were performed incorporating Odin, North East Frigg, and East Frigg. The opportunity was then taken to include the most recent geological work on the extension of the Eocene and Paleocene aquifers together with some new interpretations of the vertical permeability of the Tuff barrier. Important assumptions included in the new simulation model are listed below: -Original gas in place 265.10 9 Sm 3 as before. - Homogeneous reservoir. Impact of the aquifer increased as the permeability across the tuffaceous layer in the Balder Formation increased. Trapped gas saturation in water-swept zones still 19%, as suggested by lab measurements on available core material. -
-
In the model, final abandonment was still a consequence of the field reaching the well head pressure limitation. However, water encroachment in the wells occurred simultaneously. It became evident that the Frigg depletion was affected by water drive (Fig. 6), and by the autumn of 1980 an important support from the deeper regional Paleocene aquifer was apparent from the pressure divergence between the model and the reality. But water coning below the platforms was not yet a concern as only limited contact movements were observed in the monitoring well 25/1-A22. The increasing difference in forecasted and actual pressure behaviour lead to a review of the permeability and homogeneity of the tuff barrier. Laboratory measurements showed that the tuffaceous event in the Balder Formation was nearly impermeable, and log data suggested that the tuff was missing in the southwestern areas of the field. These observations resulted in a new approach in modelling the tuff barrier. A permeability window in the southwestern area of the field was introduced in the simulation model. Concurrently, further laboratory studies had been performed, showing the possibility of a residual gas saturation of 29% rather than 19% in water-swept zones. New simulations were performed incorporating the new hypothesis on the aquifer barrier, and residual gas saturations of both 19% and 29% were tried. The main result was that a tilted water front due to the focussed water entry was obtained, now with a significant water rise below the platforms. A potential for unrecovered reserves with the existing well pattern and facilities was identified. The new simulations
Changing perceptions of a gas field during its life cycle: a Frigg field case study
revealed that water entry and not pressure depletion would determine the ultimate recovery. The various simulation studies performed since production start-up are summarised in Table 1, period 3. As a consequence of the new model results and the continued lack of water rise in well 25/1-A22, it was decided to deepen well 10/l-A12 to validate the geological model and confirm the simulation results (De Leebeck, 1987).
Fourth period - - reappraisal phase (1984-1987) New wells Well 10/1-A 12 was a centrally located production well in the southern cluster of wells. It was deepened (and renamed 10/1-A25) during autumn 1984. The simulation model had predicted a water rise between 38 m and 42 m while the well found as much as 55 m. Other important observations and immediate conclusions were: - N o pressure differential was encountered across the tuff, confirming the existence of a permeability window, and also calling into question the concept of the tuff as a permeability barrier (Fig. 7). - Some pressure shifts were observed across shaly intervals, demonstrating their ability to act as barriers to some extent. -Residual gas saturation measured in situ behind the water front confirmed the more pessimistic lab results of 29%. - The possible effects of water encroachment were amplified as the time to water breakthrough at the platforms appeared to be shorter than forecasted. The consequences of the observations in 10/1-A25 were significant. An extensive reappraisal program was initiated, predominantly aimed at defining the water front, investigating sweep efficiency and residual gas saturation, improving the geological understanding, and re-evaluating the OGIE The program comprised drilling of three reappraisal wells (10/1-5, 25/1-7 and 25/1-8) (Fig. 4), deepening of three production wells (25/1-A14, 10/1-A26 and 25/1-A3A), and shooting of new 3D seismic. The drilling of well 10/1-5 to a bottom hole location on the southwestern comer of the northern well cluster commenced immediately after completing 10/1-A25. The well confirmed a significant water rise below the platforms, although, not as much as in 10/1-A25. The tendency of shale intervals to act as water flow barriers rather than the tuffaceous member of the Balder Formation was also verified by the well. The new wells 25/1-7 and 25/1-8 investi-
281
gated the water sweep respectively in the northern and northeastern part of the structure. The wells were temporarily abandoned for future water front monitoring by re-entry from a semi-submersible rig on a yearly basis.
Coring Extensive coring and logging of all reappraisal wells provided a significant source of information for improvements in the petrophysical and geological description of the field. Coring technology had developed since the first appraisal drilling and the use of fibre glass core barrels improved the recovery and quality of the samples significantly. Special core analysis performed during 1986 proved a relationship between residual gas saturation and porosity as indicated from a few earlier experiments on less representative core material. The average value of about 29% Sgr at the average field porosity was confirmed (Fig. 8). Well logs obtained in water swept zones during the reappraisal phase supported this value. The large amount of core material made it possible for a more detailed stratigraphical division of the reservoir by use of palynology and seismic stratigraphy. Extrapolations to uncored wells were done by log correlations. RFT (repeat formation tester) measurements obtained under dynamic conditions proved extremely useful for identifying shaly intervals with sufficient extent and permeability properties to act as barriers (Fig. 9).
New seismic In 1985 a complete 3D seismic grid was acquired on the Frigg structure. This information was used in the detailed geological modelling of the reservoir, and in a new evaluation of the OGIR Both deterministic and probabilistic calculations were made, giving quite similar answers of 223.109 Sm 3 and 225.109 Sm 3, respectively. For the probabilistic case a standard deviation range of 204 to 246.109 Sm 3 was given.
Simulations During the reappraisal phase the estimates of recoverable reserves varied with the availability of new information and interpretations. A preliminary model was made early 1985 to incorporate the high liquid level observed in 10/1-A25. This gave a very modest recovery and a short remaining field life. Results from deepening and drilling of five wells in the reappraisal program were incorporated in a new simulation update by the end of 1985.
282
E. Torheim
9 ~ '
o
9. 9 ~
=
.
'.
'=
,
-.I--
.
~ ...~.
:~1~.
9
~
.
.
i
~
;
o.-..-~__~--)
.<
,
. "..'
, _../',p=O.SDar
9
9
9
.
9
.
..
:'1229
9
.
.
9
9" 9".. ' - " .
Ap=2.6bar
ii__.--i~.
:
......
!
.
.
.
.
.
.
.
,
~
',,
.
.
.
. . . . = 1.2bar
.
.
.
.
.
.
.
.
.
.
.
.
=:
9.':7".'.'-7." 9
9
.
:-'''"
~
Ill
~
El"" w
. ~
iTuJ
w
a..J ~.
'~
~ -,-
-=-..----~. 9. . .
-
'."
.
..
9 ~
"r"
" . ' . ' j
.
.
. . . -". .
.
%
9t
"f
"l
"'~ ,.,
i
.,
9 /
--.
~,
LL
.,..,.
I~,
l
9I
,,~,-
I%'~.."0
\
,
.
%~
~
.
.
.
.
.
~. . . . . . . . . . . . .
.
\
.-~
Z .J
~<:r. ..:
9
. . . , 9 . . . . .
. .
.
.
.
'. .
.
.
.
. .
. 9 oo
.
. . . . .
,
GO
'
9
.'. - - - . . . .
'"'"
I
~
<
~
.
.
.
.
.
.
.
.
.
.
.
\
A
,,
~Lu
:',. ":.-.__" "9' ' "
~
(,~
.
o
A p = 0 .\ 8 b a. r., \ _
~
~
. . . .
.
z ,r
"."
9 .
, , ~
.
"
.
.'.'o i.
\
.
_
,it
.': ,
, I 2250
'
I 2300
i
~ 2350
'
I 2400
'
: 2450
Pressure
:
~ 2500 (HP
'
!
2550
- Gauge)
Fig. 7. RFT-10/1-A25.
-'
I 2600
,
(PSIA)
I 2650
i
I 2700
i
~ 2750
'
; 2800
:
!
2850
9
283
Changing perceptions of a gas field during its life cycle: a Frigg field case study
40
Sgr (@ Swi=lO%)
35
xx
30 25 o
201510 20
25
I
I
I
30
35
40
POROSITY (%) Fig. 8. Porosity vs. residual gas saturation - - laboratory measurements.
10/1-A26 mMSL
10/1-A25
10/1-5
25/1A14
25/1-8
25/1-7
1900
2000
\~X ~ ~ ~ ' ~ ~ ~ ~
X
\\=~~~~
~-
"Odin" Shale X
N__ "lntra Frigg" Shale
2100
2200
\N \ 2300 2400Fig. 9. RFT profiles from appraisal wells on Frigg.
A reduced OGIP figure of 256.109 Sm 3 based on reprocessing of the 1973 seismic was incorporated. Nevertheless, this model predicted a significantly better final recovery. All substantial information from the reappraisal operations were available in 1987, comprising a comprehensive new geological description and new OGIP figures based on the 3D seismic (Maritvold, 1989). On this foundation a completely new simulation model was built on which all later modifications have been based. The behaviour of this model approached what is presently regarded as realistic. Sensitivity cases were run to investigate the effect of reduced Sgr and increased OGIP.
An overview of the simulation history during the reappraisal phase can be seen in Table 1, period 4. Fifth period - - decline phase (1987 and later)
Remote observation wells After the reappraisal phase considerable uncertainty was still connected to the movement and impact of the water front. The strong aquifer support and the function of the permeability barriers had been recognised, and a detailed geological model had been established. Nevertheless, the efficiency of a barrier could only be properly detected by RFT measure-
284
ments at dynamic conditions after the barrier had been reached by the water front, or by time lapse liquid level measurements. Important criteria such as sweep efficiency and water front location with time would therefore still be difficult to predict, and also difficult to monitor in remote areas away from the platforms. The wells 25/1-7 and 8 in the northern areas of the structure became key wells in this respect as they were re-entered on a yearly basis for monitoring purposes. TDT logs were run five times from the drilling in 1985 until final abandonment in 1990 and 1991 respectively. RFT measurements obtained in open hole when the wells were drilled were supplemented with cased hole RFTs just before final abandonment. One very important observation in these wells was that no free gas was trapped beneath the extensive shale barriers existing in this area. Well 25/1-10 was drilled in 1988 on the northern arm with the main purpose to explore a deeper Jurassic prospect. Information obtained in the Frigg Formation demonstrated that the reservoir was completely swept in this area. The last drilling operations on Frigg were performed in 1989 and 1990 when the wells 25/1-A17A and A4A were deviated from the northern drilling and production platform, DP2. The purpose of A17A was to drain a structurally high area to the north of the well cluster where simulations had demonstrated a high potential for inefficient sweeping. A4A was deviated to the south to produce the remaining gas on the structural top below CDP1 after abandonment of this platform. In addition to being producers, both wells act as observation wells for pressure and also liquid level after water entry.
Production wells Uncertainty existed as to the behaviour of the production wells with respect to production potential and recoverable reserves after being reached by the water front. The main concern when designing the completion of the production wells was the danger of sand production. Therefore most of the wells were equipped with sand screens located in 8.5" open holes. Monitoring of sand production by surface detection equipment and top sediment level measurements showed that this was a minor problem. Since all production wells initially only penetrated about one third of the gas pay, the opportunity was taken to test the performance of a well with screens across the gas liquid contact when well 25/1-A14 was deepened in 1985 as a part of the reappraisal program. This showed for the first time that significant gas production from wells that had been reached by
E. Torheim
the water front could be expected, and that an expensive recompletion program could be avoided. Later this was confirmed when production wells produced gas at reasonable rates until only a few metres of screen were left above the liquid level. The good performance is due to a very small pressure drop from the formation into the well during gas flow, a situation facilitated by a careful and successful completion program. For production potential predictions it became important to describe the well performance as a function of liquid level position. The CDP1 wells located on the southern structural high were the first to be flooded by water. This area is very homogeneous in terms of petrophysical properties. A simple relationship of remaining screen height versus water free gas production potential could be established (Fig. 10). This was applicable for all the CDP1 wells since a similar completion design was used in all wells. Step rate tests at known liquid levels were continuously performed to confirm the relationship. When the water reached the DP2 wells it soon became evident that the same correlation did not apply. The northern apex, where the DP2 wells are located, is more heterogeneous and well by well correlations had to be made. By using a simplified description of water being lifted inside the well bore as a function of the pressure drop caused by the gas flow into the well, individual correlations of water free gas production potential versus remaining screen height were established (Fig. 11). This method has given good potential predictions, essentially due to the favourable porosity and permeability characteristics and the significant gravity difference between the gas and the water. Relationships for some wells have been modified as test data have become available after water entry. Monitoring of the water front position in wells utilized several techniques. In wells where the static liquid level in the formation is concealed by a solid liner or other types of blank pipe, pulsed neutron tools such as the thermal decay time (TDT) log have been used. The liquid level evolution is normally easy to track by time lapse methods, providing lithology variations such as shaly intervals do not disturb the picture. Since the liquid level inside the wells coincides with the water front in the formation when it is located within the completed intervals, a simple liquid level tool was developed. This tool is run on slick line and consists of one metre high cylindrical chambers assembled in a string in which a small hole at the top of each chamber allows for fluid entry. Important cost savings compared to more sophisticated methods were achieved by this technique. In spite of the good behaviour of the production wells after being reached by the formation water, it
285
Changing perceptions of a gas field during its life cycle: a Frigg field case study
Z~ CDP1 Screened Wells
z
zX
lz3~
Ax
Z~z3 /k
N
._o -20
>
:
zx
~-
E
..J
i~zx
Z~
L~
L
i,=
Q.
Zx
Z
A o
Zl
~ b
A Z~ ~XZ~ZX
,~
~
-30
-40
~
-50 0
0.4
0.8
1.2
1.6
2.0
=
0.05
2.4
x A H
2.8
Water Production Limit (10"6 Sm3/d)
Fig. 10. C D P 1 w e l l s - - m a x i m u m w a t e r free gas rate vs. remaining screen h e i g h t .
Depth
Lith.
-1830 . . . . .
.
R o w profile (%)
Screens
Water free flowrate (10 Sma/d) 1.o 2.0
lOO o
,j
-1840 . . . .
I
I -1850 . . . .
J
I
==_m=_
-1860 --~
i
=
-I I I
-
!
I I I
-1870 --
i I
I
/
j
/
i
/
Well: 25/1-A21
k= 1 6 0 0 m D (h= 90m) D= 2.7 S= 2.7 (Based on test 12/84)
.
I
3.0
is
\
20-
I 25-
30-
I
40Fig. 1 1 . 2 5 / 1 - A 2 1 - - m a x i m u m w a t e r free gas rate vs. remaining screen height.
286
was obviously beneficial to improve the flow potential, provided that it could be done by simple and reasonably cheap methods. As the field entered further into its decline phase, with CDP1 permanently abandoned and more wells on DP2 severely affected by water, it was decided to test the effect of setting bridge plugs inside the completion intervals. This had been considered earlier, but the idea was abandoned for two main reasons. Firstly, a plug which could pass through a production tubing with a minimum ID of 3.8" and at the same time be expandable to the screen ID of 6" in a gas well, was not readily available. Secondly, due to the completion design with an open annulus outside the screen, the expected effect of a plug was limited. If the open hole size had remained unchanged since drilling, only about 65% of the cross-sectional area of the well bore would be blocked by the plug and the liner together. In recent years plugs with the capabilities described above have become available, while the actual candidates for plugging have shifted from the CDP1 wells to the more northerly DP2 wells where the occurrence of shaly intervals is more frequent. The possibility that a shale could swell and plug the annulus was recognised. Therefore, in 1993 plugs were set in three wells in front of shaly intervals above the gas water contact. Significant increases in water free gas production potential were obtained in all cases, confirming the sealing of the shales. This experience represented another important aid in optimising the production potential of the field.
Seismic monitoring A well-known problem in typical gas fields that suffer water encroachment is caused by the clustering of the production wells. As all traditional reservoir monitoring is done via the wells, the areal knowledge is a function of well locations. A better areal coverage would make for improved liquid-level monitoring. In the case of Frigg large parts of the field were not covered, in spite of remote monitoring wells. This situation deteriorated with the final abandonment of 25/1-7 and 8. The idea of utilising seismic technology as a tool for water front monitoring was tried as early as in 1983 (Maritvold, 1989). The main objective then was to evaluate possible migration of gas from the satellites towards Frigg. A gas expansion on the North East Frigg field was detected using time lapse techniques on high resolution seismic acquired in 1979, 1981 and 1983. Pressure measurements have later confirmed this observation. Following the encouraging results from North East Frigg, further research was initiated aimed at broad
E. Torheim
monitoring of the water front movement. The 1973 seismic (shot before production start-up) was reprocessed, used as reference data and compared with the detailed seismic acquired in 1985. By studying the relative amplitude difference between a seismic event close to the top of the reservoir (the Frigg seismic marker, FSM) and the fiat spot for the two sets of data, it was possible to detect contact movements. However, most of the results could only be used qualitatively. In 1991 three new seismic lines were shot coveting old lines from 1973 (Fig. 12), again with the main objective to quantify water rise along the lines. Compared with the work performed in the late eighties a few premises had changed, giving good hope for quantitative results. The water rise had increased making it easier to detect the difference in the seismic response, computer support had improved, and interpretation technology had advanced, especially regarding use of amplitude versus offset (AVO) attributes. Another important innovation was the improved capability to transform well log data and to incorporate it into the seismic evaluation. Together this lead to a positioning of the liquid level along the seismic lines. The results compared very well with the simulation model along the east-west line (Fig. 13). The northsouth line demonstrated a lack of gas in the simulation model to the east of the platform area (Fig. 14). This was incorporated and matched, and contributed to a continuation of the nominated production level. With the seismic monitoring technology a new tool for liquid level monitoring on gas fields has been developed. The advantages of the method are obvious: good areal coverage and very low cost compared to drilling and re-entry of remote observation wells. On the other hand, there is still some uncertainty in the results. The accuracy is a function of the seismic resolution, and since human interpretation is involved, the results will always be affected by subjectivity to some extent.
Simulations The probabilistic OGIP of 225.10 9 S m 3 from 1987 was modified in 1988/89 to include new geological information from well 25/1-10, a new interpretation giving a reduced difference between the Frigg seismic marker and the top of the pay, and improved mapping of the flank areas involving a correction of the edge effects. This resulted in an OGIP of 235.109 Sm 3 with a standard deviation of 14.109 Sm 3. This information was incorporated when updating the simulation model in 1989. In addition, the concept of water free gas production in wells reached by the water front was used (a function of remaining screen height above the liquid level).
Changing perceptions of a gas field during its life cycle: a Frigg field case study
287
Fig. 12. Seismic lines acquired 1991.
To respect the individual behaviour of the DP2 wells and increase the accuracy of the water encroachment description, refined gridding and an improved detailed geological description were incorporated in the well area in the 1991 simulation. Until now, each model well had represented several real wells, but after the modifications all remaining active wells were modelled individually. The field reached its minimum reservoir pressure in 1990 (Fig. 3) due to the reduced production level. This made it easier to balance the effect of gas volume and aquifer impact in the pressure matching. After some time with the field in compression, difficulties in matching the aquifer pressure behaviour occurred. The model gave a stronger aquifer pressure support than actually experienced. To investigate the
possible consequences and to avoid an unnecessary decline in the gas sale nominations, it was decided to extend the existing simulation model to avoid the use of artificial aquifers. The model was kept unchanged in the Frigg area except that the artificial aquifers representing the regional Paleocene aquifer volume were replaced by a new coarse grid surrounding the original grid. Regional geological information was incorporated together with matching of pressure from exploration wells. The Heimdal field (Fig. 1) was now a part of the model which improved the control of the pressure effect from this gas pool and its depletion during Heimdal production. The seismic liquid level results were used and an excellent match of the water front was obtained, respecting all well measurements. The aquifer pressure evolution was now better de-
288
E. Torheim
Fig. 14. Comparison of GLC from seismic and simulations. North-south cross-section.
scribed and generally a good pressure match was achieved. In terms of reserves the changes compared to the previous model were moderate. However, an improved tool to evaluate the remaining field life was obtained. Period 5 in Table 1 shows the evolution of the simulation models since 1987. The history of recovery factor estimates is plotted in Fig. 15.
Conclusions Data acquisition and interpretation activity on the Frigg field has taken place continuously for nearly 30 years since the first seismic was shot in 1965. Some
uncertainty in the recoverable reserves will remain until the field ceases to produce. Nonetheless, continuous improvements in knowledge have been crucial for reservoir management of the field, in particular to optimise production. The amount of reservoir work performed must be seen in connection with the contractual obligations towards the gas buyer, as gas sales and rates have to be nominated a considerable time in advance. Many of the problems and challenges in the reservoir management of the Frigg field have been linked to the aquifer activity. Based on this experience, some conclusions may be drawn:
289
Changing perceptions of a gas field during its life cycle: a l~rigg field case study
90
% VARIATION
RECOVERY FACTOR (%)
10
80
-
f
70
-
60
-
I PRESENT VALUE ]
-10
-20
-30 I
50 1970
1972
I 1974
f 1976
I 1978
I 1980
I 1982
I 1984
I 1986
I 1988
I 1990
I 1992
Fig. 15. Evolution of estimated recovery factors.
(1) For reservoir engineering purposes an appraisal program should be as extensive as possible. Sparse information leads to an over-simplification of nature. The economical constraints on a field appraisal should be considered in relation to the possible consequences of such over-simplifications. (2) Experience from Frigg has shown that extensive coring can be very beneficial, both for detailed geological description and petrophysical purposes. Residual gas saturation in swept zones is very important in the case of a water drive. (3) In a typical gas field development a sensible pattern of "deep" observation wells is very useful. (4) Care should be taken when designing the completion program for production wells, both in terms of equipment and clean-up. Wells with low skin have been very beneficial on Frigg when producing after water breakthrough. (5) Good high-resolution 3D seismic should be acquired at an early stage, primarily for better geological description. A second reason for acquiring the 3D is for potential seismic monitoring of liquid level rise. (6) RFT measurements at dynamic conditions have been very valuable on the Frigg field. Permeability
E. TORHEIM
barriers are not often recognised before being reached by the encroaching liquid (Fig. 8). (7) The ultimate goal of reservoir monitoring is to optimise reserves and production. In addition to the obvious benefit of improved cash flow, another significant economical incentive would be to accelerate the recovery of the reserves giving a reduction of operational expenditures.
Acknowledgements The author wishes to thank the management of Elf Petroleum Norge a.s. and the Frigg Unit Partners, Norsk Hydro A/S, Statoil, Total Norge A/S, Total Oil Marine p.l.c, and Elf Enterprise Caledonia for permission to publish this paper.
References Brewster, J. and Jeangeot, G., 1987. The production geology of the Frigg field. North Sea Oil and Gas Reservoirs, pp. 75-89. De Leebeck, A., 1987. Frigg field reservoir: characteristics and performance. North Sea Oil and Gas Reservoirs, pp. 89-100. Maritvold, R., 1989. Frigg field reservoir management. North Sea Oil and Gas Reservoirs, II, pp. 155-163.
Elf Petroleum Norge a.s., P.O. Box 168, N-4001 Stavanger, Norway
This Page Intentionally Left Blank
291
Choosing between rocks, hard places and a lot more: the e c o n o m i c interface Helge Hove Haldorsen
The most central aspects of the oil and gas business are exploration (is there oil or gas?), production (market? infrastructure? how fast can we produce and how much can we recover?), licence and tax regulations (acceptable and stable? incentive systems?), safety, health and environmental concerns, and the future oil price (will the oil business remain an attractive business-segment to invest in, will we obtain a satisfactory return on the capital employed?). Exploration is in many ways retrospective; a real understanding of how geological formations and fluids within basins were formed and modified is a prerequisite for success. Production and oil price forecasting are intensely quantitative and prospective (forward-looking). The main application of economic analysis in the oil business is to arrive at economic indicators for proposed future activities (given a set of assumptions and estimates) and to determine the effect a certain investment will have on the company's financial position. As a future oil business venture is never deterministic (in terms of geological and/or financial outcomes) and since several courses of action are possible given the information at hand, project economics may be computed for a range of assumptions (input data) and for several alternative scenarios, where each scenario and/or assumption may additionally be assigned a probability. In this manner, the expected economic outcome may be computed along with downside and upside estimates. Also, project economics play a decisive role when a number of alternatives are competing for implementation; based on technicaleconomic optimization studies, we select the optimum development concept, the optimum number of wells, the optimum fluid handling capacities, etc. with due consideration of subsurface and other relevant uncertainties. The economic optimum is chosen based on comparative analyses. This paper will demonstrate that the 'economic interface' is essential for most oil business decisions: for deciding whether or not to explore or develop, for choosing between exploration prospects ('rocks') in a licensing round, for choosing between basins or countries ('places') in which one wants to explore and produce, and for choosing between development concepts, recovery mechanisms, plateau rates . . . . (and a lot more), in field development planning. Due to the largely unpredictable nature of the subsurface and the future oil price, the concepts of 'risk' (= possibility of a financial loss or an unachieved objective) and unfortunately, to a lesser extent 'grisk' (= possibility of a financial gain in excess of the objective) are quite central in the oil business. Illustrations of 'risk' and 'grisk' will be presented.
The oil price forecast: the king of the company! The only relevant risk? Although the future oil price is the cornerstone of the oil industry in terms of the industry's sustainable attractiveness as a high-return business segment, predicting this price correctly has proven difficult, if not impossible (Austvik, 1991). Figs. 1 and 2 show historical values, and Figs. 3 and 4 show the inherent optimism in historical forecasts. The one thing we seem to learn from oil price history is that it is difficult to learn anything useful from oil price history! Every oil company, finance ministry and a large number of institutions every year come out with new oil price forecasts. They never quite agree. The list of factors which could cause a rise or a fall or a steady
oil price is, however, usually the same (Fig. 5). In the end, a base-case oil price scenario is arrived at, and in oil companies, this 'official' oil price scenario (along with 'official' inflation, exchange, interest, and discount rates) is used for evaluating exploration and investment projects competing for funds from a (usually) limited purse. Since oil is sold in dollars, the dollar - your currency - exchange rate (into the future) is also a very important forecast to be made. Oil companies traditionally work in partnerships to spread risk and to secure that the operator constantly has a number of constructive devil's advocates to keep him 'honest'. Since each company has its own 'official' oil price forecast, we can get the paradoxical situation that a project can look like a 'big loser' to one company and like a 'company maker'
Quantification and Prediction of Petroleum Resources edited by A.G. Dor6 and R. Sinding-Larsen. NPF Special Publication 6, pp. 291-312, Elsevier, Amsterdam. 9 Norwegian Petroleum Society (NPF), 1996.
H.H. Haldorsen
292
PREDICTING THE FUTURE OIL PRICE IS DIFFICULT!
$/BBL I
5O -
4O
30 10
....
-40
9
i?__i,
-
-
30- -
,o--
-
10 - I 1930 1940 1950 1960 1970 1980 1990 2000 2010 2020
Year Fig. 1. Crude oil prices (1987 $).
3't( 36
OK, WHO PREDICTED THIS? Brent B l e n d ) [ Oct.
33
R)
i'
30
!
27 24 21 18 15
:eb.
12
I 1986
I Year
1991
Fig. 2. Crude oil prices detail, 1985-1991.
to another. Companies also have a required rate of return (often a country-dependent hurdle rate) which must be met before project approval. Assuming that all other estimates and forecasts (CAPEX, OPEX, schedule, production vs time) are correct (of course, they seldom are, as we shall discuss later), the required rate of return, in effect, would translate into a required future oil price. A rather interesting requirement! Companies with too optimistic 'official' forecasts will continuously disappoint their shareholders with a less than 'required' return on investments (always blaming the oil price in quarterly and annual reports) while companies with cautious or too pessimistic oil price forecasts will find fewer acceptable projects, but those that pass the hurdle will deliver (unless something else went wrong). A company's tax position, portfolio size and upstream-downstream diversification may alter this conclusion; it only serves to point out an old truth taught in business-schools: the only relevant risk in the oil industry is the fu-
Fig. 5. Factors influencing crude oil prices to year 2000 (source: de Ruiter, 1993).
ture oil price risk. As we shall discuss later, many upstream disciplines (explorationists, reservoir engineers, productionists) will strongly disagree with this (over)simplification. Fig. 6 shows, however, that a strong historical relationship between oil price and return on capital employed exists. Today, very few companies maintain the 'hockey stick' approach to oil price forecasting (Fig. 7a). The industry has become very cautious; too cautious, some will say, particularly so far as the long-term oil price is concerned. Some scientists note that the world's hydrocarbons
Choosing between rocks, hard places and a lot more: the economic interface
35 -7
30~ *c ~
I
-35
i~-,,
III II
~
I
Oil Price 1 Rateof Return
-30 -25 .D
\
I
#
-20 =
'~ . 15
- 15
O
"= a.
-10 o
lo
-5
1978' ~80' 19'82' 19'~' 1186' 1~88' 19'90 Year
Fig. 6. Rate of return for the international oil and gas industry and oil prices during 1977-1990, USA excluded (source: Energy Information Administration, USA).
293
In summary, a company's official oil price forecast (and exchange rate, discount rate, etc.) is king. As we proceed to discuss the nuts and bolts of the business, we shall see that the 'economic interface' always takes us from 'Jurassic exploration parks', wells and mud, platforms and pipelines, and horrible partial differential equations in time and space (reservoir engineering) to the bottom line: will we make money? What are the financial risks? We go from one largely imaginary reality (rocks at 9000 ft) to another (the oil price 10 years from now). This is why we all love the oil business: 'invisible risks' in time and space, which we need multidisciplinary organizations and advanced technology to quantify before we can make a decision, which in turn triggers the application of unparalleled engineering technology for safely constructing, installing, and operating a North Sea Platform, ready to produce oil at a price, we hope, at least as high as the company's official forecast.
From resources to reserves" the global challenge
20
~15-
~, 1o-
~. 5 -
s;
b
0
I
1000
I
2000
Time
I
3000
4000
Fig. 7. (a) "Hockey-stick" forecasts strictly forbidden. (b) Oil production in a historical perspective.
may run out in 50 years (Fig. 7b) as the world's population doubles and fission/fusion still cannot power cars. Some upstream projects actually last for all those 50 years!
Consider that, right now, all of us reading this volume form a new oil company, Jurassic Park Oil and Gas (JPO&G). We have money, technology, competence and 1000 years of cumulative international E&P experience. We have one major problem, however; we have no Prospective Resources (undrilled acreage which we think contains oil), no Technical Resources (discovered, or proven oil in the ground, but not yet proven commercial), no Developed Reserves (platform installed, production ongoing which taps reserves), and no Undeveloped Commercial Reserves (discovered and commercial oil which we plan to phase in later). So, how do we get started? We can (1) buy another company which has all of the above, (2) buy into producing properties held by other companies, (3) buy into acreage with Technical Resources, or (4) acquire exploration acreage (and hence, Prospective Resources) through licensing rounds or farm-ins. Let us assume that our only interest is option (4): we want to explore our way to production! The world has, however, hundreds of sedimentary basins, production is ongoing in a large number of countries, each with a different commercial and fiscal framework. We can go for high-risk frontier areas ('elephant-hunting'), we can go for licences 'on a trend' in producing basins with proven prospectivity, we can go onshore and/or offshore, we can go for operatorships and/or only partner roles, we can target gas rather than oil etc. Hence, 'choosing between places' (countries or basins) and 'choosing between rocks' (licenses or prospects) is a m o m e n -
294
H.H. Haldorsen
Region A
I I El
0
:~0
60 (No. of Untested Basins)
~,0
10-
-b-
JQ .Q r n9.
5-
v i=.
2
Q=
b
0
I
5m
.0
4-
O r --9
3-
o
2-
Z
I-
~> r
0
--6--
8
m t k
--if@ US$20 per barrel (including exploration and development costs)
Fig. 8. (a) Untested basins per region or country. (b) Value of discovery; profit on a US $20 barrel for various countries. (c) Value to producer; NPV 10% per barrel produced assuming an oil price of US $20/bbl for various countries.
tous task for JPO&G, even though our 1000 years of experience should help us some part of the way. For simplicity, let us assume that after a very comprehensive study of all relevant factors we picked the 10 most interesting basins (from a prospectivity point of view) in the 10 most interesting countries from a commercial setup/fiscal risk-reward balance point of view. The 'economic interface' in this exercise could be to take three imaginary fields (size 1: average field size; 2: small; 3: large), put them at 'typical' reservoir depths in the 10 basins, construct very coarse production profiles/development schemes (based on analogue field data) and finally, estimate economic indicators for the conceived venture with the prevailing terms and conditions in the various countries. 'Ranking of places' could be done as shown in Fig. 8. Let us further assume that we are capital(and manpower-) constrained and not opportunityconstrained in these 10 basins and that we have thoroughly analyzed all regional and field-specific
data available, to a point where we feel that we really understand the key to exploration success in these basins. The fact that we are capital-constrained will have a bearing on: (1) what percentage of a licence we will want, and (2) offshore vs onshore preference. If we end up with an average (portfolio) discovery probability of 10% and we assume that only 50% of all discovered technical resources turn out to be commercial, only an average of one well in twenty will yield a commercial discovery. If our total exploration budget is 50.10 6 USD, we should ask ourselves: what do we get for 50.10 6 USD? One obvious answer is, of course: probably a lot more onshore wells than offshore wells (although again dependent on participating interest) due to the different cost levels. There are other important differences which should be remembered when choosing between onshore and offshore participation; (1) scope, in terms of engineering requirements and capital expenditure, tends to be one or two orders of magnitude greater for offshore projects; (2) offshore projects tend to have a much longer development schedule before they come onstream; (3) onshore is more 'pay as you go', 'invest as you investigate', or even better, 'invest as you earn'; (4) offshore well productivities need to be substantially greater than onshore to cover the greater capital expenditure and operating cost, respectively. These differences should be kept in mind as we proceed. Typical cash flow profiles for onshore and offshore projects are shown in Fig. 9. In order to choose among competing prospects (which are all acceptable from a 'strategic' point of view), a consistent methodology for prospect evaluation should be followed. Although all companies' approaches differ somewhat, the basic idea for interfacing the subsurface with economics is usually the same. A brief overview is presented next.
Prospect evaluation" from Jurassic exploration parks to economic indicators Oil companies are fundamentally concerned with replacing and, hopefully, increasing economically producible reserves. The value of their stock is strongly tied to the 'health' of their reserve base (or reserves-to-production ratio). Licensing rounds and farm-ins are the two most common ways of acquiring prospective resources (which in due time, we hope, will turn into commercial reserves). Assume that a farm-in possibility (in a basin where we have previously studied the regional setting, trends and field data) is available. The offer is to join '2 for 1' (take over 20% of the licence and pay for 40% of the total future committed exploration expenses), i.e. a typical
295
Choosing between rocks, hard places and a lot more: the economic interface 15009- - - -
Typical onshore
project
-- -- Typical offshore project
1000I I
500-
$ $
I
$ m~ ~
~
~ mm mm m ~ .m m .......-
0 $ $ $
-500 -
-1000
#
s 9
I
I
I
0
5
10
I
I
I
I
15
20
25
27
Year Fig. 9. Typical before tax cash flow profiles for offshore and onshore projects.
Fig. 10. The big cash flow picture; from initial G & G to abandonment.
'carry' arrangement. The licence obligation is 1500 km of seismic and 1 exploration well. A 'typical' Production Sharing Agreement (PSA) with detailed cost-recovery principles, profit oil split, etc. regulates the licence. When executing the farm-in agreement, JPO&G is required to pay 20% of cumulative costs to date and to take over the operatorship. JPO&G explorationists evaluate all available data and arrive at structure maps, geological models, volumes, volume factors, fluid type (gas vs oil) and combine all relevant elements in the geological risk chain (trap, reservoir, source, seal, migration/preservation) into an overall discovery probability, Po, for the
prospect. Let us assume, for simplicity, that the licence at hand only has one prospect. Next, estimates of seismic (acquisition and interpretation) costs, drilling (including coring, testing, logging . . . . ) costs, appraisal costs (new 3-D seismic and appraisal wells), and costs of all technical studies until the development decision has been arrived at are made. Assuming a discovery, estimates of oil rate vs time, CAPEX, OPEX, etc. are made for a conceived field development. Cash flow vs time from the initial G&G evaluation to field abandonment 25 years from now is shown in Fig. 10. A common project-economic indicator (Boye, 1985; Bradley et
H.H. Haldorsen
296
Fig. 11. Input and assumptions from many disciplines are required before the net present value (NPV) can be computed.
al., 1988) the Net Present Value (NPV), is shown in Fig. 11, where some of the necessary input data from the various disciplines are indicated. In general, the risked NPV (probability of discovery x NPV) of the development (including all costs after the exploration well) should be equal to or larger than the net present value of costs (up to and including the exploration well) with the company's official oil price and inflation/discount rate/etc, utilized in the calculation. The PSA terms (which vary significantly from licence to licence, country to country) must, of course, be correctly implemented in the economic model. Many other economic indicators may be computed along with relevant sensitivities in order to check the robustness of the project (to further aid decision makers). Note, however, that computing economic indicators means collapsing '25 years and hundreds of cash-flow elements' into one magic number! Most companies utilize decision trees in the economic evaluation of exploration (and many other) projects in order to 'organize' options, outcomes and outcome probabilities in a systematic manner. Each branch-end of the tree will have a NPV and a composite probability associated with it. The project's weighted average NPV is the sum of the products: (NPV)Branchi • (Probability)Branch i over all branches in the tree (Fig. 12).
In exploration, it is essential that the assessment of geological risks are harmonized against the rest of JPO&G's exploration portfolio (to ensure consistency and to avoid bias). Senior, experienced explorationists note, however, that regardless of how much data you have and how fancy the computers you use to process them, in the final analysis its always a 'question of faith'. JPO&G has a standing panel of senior explorationists which sees every exploration proposal to ensure such harmonization. This is because the discovery probability, Po, is perhaps the single most important factor to assess correctly, based on the available data. With several obligation wells drilling several prospects, dependent probabilities (for a given hit-fail sequence) must be estimated and used for computing the risked NPV of that specific outcome. So, we see that the economic interface is essential for choosing between rocks. We also see, however, that many disciplines contribute with forecasts and estimates necessary for the overall economic assessment. The validity of the economic indicators are only as good as the input data. A realistic view of the subsurface is the key to whether or not a correct exploration conclusion has been arrived at. Unfortunately, we can only find out by drilling the well.
Choosing between rocks, hard places and a lot more: the economic interface
High Reserves (P=.2) Discovery (P=.I)
Yes Explore ?
Expected Reserves (P=.4)
Low Reserves (P=.2) Not Commercial (P=.2)
Dry (P=.9)
297
i
Probi
NPVi
Probi x NPVi
1
0.02
A
(.02) A
2
0.04
B
(.04) B
3
0.02
c
(0.2) c
4 5
0.02 0.90
D E
(.02) D (.90) E
6
1.00
0
0
No 5 = ~ (Probi) x NPVi)=G i=i
9Weighted AVG. NPV ('Yes-Branch') 9NPV ('No-Branch') =0 9IF G>O, Explore! 9A-E; Shows the spread in N P V - O u t c o m e for this venture Fig. 12. Decision tree for an exploration project.
Pre-development costs Let us assume that we actually made a promising discovery with our obligation well. Prior to drilling, the operator would have proposed (via an AFE) a data-gathering package (in case of discovery): MWD, VSP, logging, coting, RFT, testing, sampling, etc. It is essential to obtain sufficient and reliable data, but 'data for the sake of data' should be avoided and the value of the information obtained should always justify its acquisition (Fig. 13). Fig. 14 shows a typical pre-development cost-split for some North Sea fields. Note that drilling and well-related data gathering activities represent 72% of the total. Based on the data from the first well, all maps are updated, formation/fluid characteristics are re-evaluated and hydrocarbon volumes re-estimated (Fig. 15). A coarse field development plan is constructed (including the remaining pre-development costs, Fig. 10) and economic indicators are again computed (Fig. 11). Note that at this stage, several alternative development concepts seem viable. More subsurface information is clearly needed, and (concept) engineering studies should focus on options, not details. At this point, we also note one of the great 'reliefs' of the upstream oil business: the wonderful concept of sunk costs. We just love to forget all our spending to date, because sunk costs do not affect 'yet-to-spend economics' (except for tax effects). Past expenditure cannot be 'unspent', so we ignore it! Shareholders and those who compute reserves replacement costs, however, have a longer 'fiscal' memory. We must continuously try to learn from past experience in order to avoid unnecessary costs in future projects.
We now establish an integrated project team, shoot and interpret 3-D seismic data, drill and test two appraisal wells, and set an ambitious target: finalize a field development plan within one year! The 3-D survey and the two appraisal wells were justified based on a classical 'value of information' analysis. Deciding when to stop appraising (and hence, when to 'freeze' the design basis) is difficult; freezing the design basis too early can lead to a suboptimum development plan (expensive retrofits later) and freezing it too late makes total project costs excessive (Fig. 16). Of course, over- and underdimensioning are, or should be, equally embarrassing. Below, we discuss key issues relevant to designing and optimizing fields before developments today. The main focus is on describing the 'geological plumbingsystem' sufficiently well to avoid 'building the wrong factory' and to arrive at optimum choices for those things which the operator is responsible for choosing in a field development project. Optimizing field development plans Before ordering optimally designed onshore modules or offshore platforms and before predrilling/ drilling wells in optimum positions, the operator of a field needs to have demonstrated clearly and beyond doubt for himself, for partners, and for the authorities that the chosen recovery mechanism, platform facility position(s), plateau rate, processing capacities (topside facilities), number of wells, types of wells, positions of wells, perforated intervals, tubing diameters, offtake strategy (reservoir management), and many other choices, are optimum choices, also taking
H.H. Haldorsen
298
(~ Eo No Gather Additional Information?
!
Good News ~ fG
Yes I I ] Bad New s ~ (l-f G) /
Right fGR
Wrong (1 "fGR) Right fBR
(~EGR
(•EGw ~ EBR
Wrong ,, ( ~ EBW (1-fBR)
9CMax = The maximum amount that the 'information' may be worth 9CMax= fG~GR E G R , (I-fGR) EGW~+ (I-fG) ~BREBR+ (I-fBR)
EB~-EO (f-probabilities,E-netpresent values)
9It would be reasonable to spend up to CMAX for obtaining new information (3D-seismic, one more well, .... ) but no more. If CMAX is negative, the information gathered would have no positive value. 9The cost of gathering the information is not included in the estimated outcomes. If it is included, the information gathering activity should be initiated only if CMAX >0.
Fig. 13. "Valueof information"--decision tree (source:Lohrenz, 1988).
Fig. 14. Typical split of pre-development costs (North Sea).
into consideration subsurface and other future uncertainties. Such a demonstration is only possible if it is documented that all suboptimum choices have been considered and rejected and that relevant 'what ifs' have been checked during the search for the optimum.
The project team and its challenges Integrated multidisciplinary field development teams must be empowered to organize and steer the search, to quickly reduce the number of scenarios,
iterations and simulation runs, and be challenged to create an innovative and 'value creation' spirit within all disciplines involved. The reservoir simulation model is the industry's major tool for forecasting the future production performance; this is key technical input data for arriving at optimum economic choices (by investigating a large number of alternatives) and for finding the effects of every discipline's major uncertainty by running a large number of geological and other 'what-if' sensitivities (Fig. 17).
Choosing between rocks, hard places and a lot more: the economic interface Table 1
9 Seismic Uncertainties Mapping Uncertainties Fluid Contact Uncertainties
Optimizing a project by maximizing its Net Present Value (from Behrenbruch, 1991)
4,
The overall strategy is usually to maximize risked NPV, optimizing the project as follows:
Most likely GRV + distribution of GRV (GRV-gross rock volume)
(GRV)
- Minimise number of wells - Maximise reserves - O p t i m i s e production profile (maximum early production, short production life)
(N/G)
(~)
(I-Sw)
(Bo) (BG)
100 ~ [ ~
m~'
(Mean)I ] l%
001 u T ~ -
STOOIP (OGIP)
STOOIP (OGIP)
- Incorporate flexibility, simplicity and contingency
Fig. 15. Estimation of initial hydrocarbons in place.
- Build on previous experience (competitive edge) - Proper regard for value of information (i.e. optimum field appraisal; 3-D seismic . . . . ) - Proper regard for external factors (oil price . . . . )
I
costs
"~/
~/'~,~
um time
Total
project
Reservoir knowledge:
10o% I
{static+ dynamic info.}->"
I
duration
~'Optimum
area, freeze consept design
- Minimise capital expenditure (CAPEX) - Minimise operating expenditure (OPEX) - Short construction schedule - Regard for safety and environment - M i n i m i s e time between discovery and development and predevelopment cost - Minimise risk
(L)(M)(H)
Total project
299
basis,
finalize development plan, and cover remaining subsurface uncertainties with contingency plans and/or 'cheap' flexibility
Fig. 16. Optimization of time for "freezing" concept design basis.
As noted by Saleri (1993), field-wide cumulative production forecast accuracies (in an absolute sense) tend to range from 10% to 40%. Individual well performance forecasts are much less successful than field-wide predictions. This may sound alarming, but remember that the majority of (reservoir) engineering and economic decisions require choosing based on comparative analyses. If we are wrong about the subsurface, at least we are perhaps consistently wrong. A major problem is, however, that the various recovery mechanisms exhibit counterintuitive and changing sensitivity to changes in the reservoir description. Hence, we may not be consistently wrong as we compare economic indicators for several competing recovery mechanisms.
Several of the above aspects oppose (or compete with) each other and that is where a particular company's values come into play. It should also be noted that outside influences can play a major role, for example oil price variations and, to a lesser degree, changes in inflation and exchange rates. In summary, constraints may be technical, geographical, environmental, legal, fiscal, political or related to marketing and safety.
The goal is to arrive at the optimum combination of 'man-controllable settings' (Fig. 18, Table 1) when up against an uncertain subsurface (or several parallel imaginary subsurface realities) and to compute the final project-economic outcome range (Fig. 19). Although conflicts must arise between maximizing project-economic indicators and maximizing the ultimate recovery factor (with one million wells the recovery factor would be 96%, but the developer would lose billions of dollars), this important and often political issue is largely left out in the ensuing discussions. Operators seek to arrive at a good balance between prudent resource management and return on investments while adhering to all local rules and regulations. Safety and environmental concerns can never be compromised!
Optimum choices Recovery mechanism As a minimum, pure depletion, water injection and gas injection with high-angle, horizontal, multi-bore, or 'designer' wells should be studied (Fig. 20). In addition, relevant IOR/EOR possibilities, such as WAG (miscible and immiscible), polymer (or gel) injection, hydraulic fracturing, etc., should be investigated and documented separately. Combinations, such as
H.H. Haldorsen
300
Fig. 18. Searching for the optimum field development plan is like trying to solve a 4-D Rubik-puzzle.
a period of pure depletion followed by water injection, water injection followed by WAG, or updip gas injection concurrent with downdip water injection (producers in the middle) should also be studied for merit. Constraints, such as lack of infrastructure for (associated) gas export, may eliminate some options, and create or necessitate others.
In principle, each onshore and offshore field could be gridded up into candidate (x, y) positions for the platform (processing facility) and through an iterative process, the overall project NPV could be computed for all candidate (x, y) positions. The maximum point of this NPV (x, y) surface would then in principle be the optimum (x, y) position for the platform (facility).
Platform (facilities) position
Plateau rate
The areal distribution of hydrocarbons in-place should be investigated. For offshore fields, this information often dictates the necessary number of platforms and their optimal positions. Drilling reach, seabed site surveys, indications of shallow gas, positions of fluid contacts and the chosen concept (GBS vs a DCF with subsea wells, etc.) also play a major role when determining the optimum location of processing/drilling facilities.
Several plateau rates for a field should be tried (Fig. 21a). The number of wells should be varied for each choice of plateau rate to see how the NPV (Fig. 21b) and the recovery factor (Fig. 21c) are affected. Even if the recovery factor often increases as the number of wells is increased, a case of diminishing return is often evidenced (Fig. 21c). This is due to the fact that the last 'new' wells do not produce enough oil to pay for the added expense of including
Choosing between rocks, hard places and a lot more: the economic interface
301
P. HC ___~.VV
500
~oo l
GEOLOGICALMODEL (BASECASEAND OTHER POSSIBILITIES)
I
I
I
E
~J300, ~'~
CONCEPT,PROCESSINGCAPACITIES NO. OF WELLS, PLATEAU RATE......
i"i""
~mmmmmm~~ PERFORMANCE ECONOMI FORECASTI CS NG
~00 -
TIME
18
~ 1O0 RF (@ 20 YEARS)
~1 O0
-
~:200 1. Cum. Probability
~ 99.
NPV
Fig. 19. A project's "what-if machine"; from geology to economy (and back again) many times.
NATURAL DRIVES
DRAINAGE
/1
--/I
INFLUX
l
INFLUX
/
-
so.u ,o. GAS DRIVE
/
CONVENTIONAL OIL RECOVERY
ENGINEERED DRIVES
"FLOOD HORIZONTAL WELLS
MINING & EXTRACTION I
I
" LEACHING
III
o,s REINJECTION
ENHANCED & IMPROVED OIL RECOVERY (EOR & IOR)
/
m k
HYDRAULIC & THERMAL FRACTURING
-HYDROCARBON - CO NI'~ROGEN - FLUE GAS - FOAM, WAG - CYCLING - SUPER-CRITICAL
~
L-I
m
- SURFACTANT
m l " POLYMER I - CAUSTIC
9 I - ALCOHOL 9 I - ARTIFICIAL 9I CONINGBARRIERS 9 I " "MINI-EOR":SURF. 1 TREATM.OF INJ.
T.E...L /
[-- 9 -
9
STEAM IN-SITU
I
9 COMBUSTION 9 - ELECTRIC (MW) 9 ~ 1 9 9
IMPROVED/
~ ENHANCED OIL
k I" IMPROVED COMPL.TECH I N m II- (HORIZ. WELLS) WORKOVN OPTIMIZATION~'WELLPOSN 9 I OFFTAKE,GAS LIFT)
m i-INRLLWELLS I- CO-MINGLEDAND
MULTI-PURPOSEWELLS I- ELECTRICDH PUMPS I- DUMP-FLOODING
RECOVERY
n
9
9 9 9 9
Fig. 20. Recovery mechanisms.
them (perhaps the whole development concept must be changed from a ship to a GBS as the number of wells goes from 18 to 45). A plot of NPV against the number of wells and/or the plateau rate will often exhibit a maximum value (Fig. 21b). A plot of the recovery factor (at a fixed time, for a fixed number of wells) against plateau rate will demonstrate if rate sensitivity is a problem (Fig. 21 d). If this is the case, the number of wells may be increased, which in turn will increase costs, and the search continues. As discussed, a case with a lower ultimate recovery
factor may still be the most profitable (have highest return of NPV) if the oil rate is high during early times. Generally, 'late oil' has (unfairly some say) less value than one should think due to the discounting process. In some cases, however, early high oil rates will require increased topside handling capacities (Table 2) and larger tubing sizes or the number of wells may have to be increased to achieve it. These additions will again erode the NPV. At this time, the need for fully and continuously integrating economics in the optimization process should be evident.
302
H.H. Haldorsen
Ibl
f
.!v Nwells QoiI,P
l
t
QoiI,P
lal
RF I Nwel I @fixedtimeI QOiI,P
Time
Icl
Nwells
,
~"
Insensitive
Sensitive
"1~1
Fig. 21. Optimization methodology.
Gas/Water Handling Capacity Level
oO| G
T
9
= Oil Rate loo~176176176 ~l-- ~%~' v O
Rate
i
.,
oo.
"~
Time
~,
v
Fig. 22. Effect of gas or water handling limit (1-8) on plateau length (1-8).
Table 2 Capacity of facility limits must be decided for a large number of critical platform elements - Capacity for shipping pumps - Capacity for water injection pumps - C a p a c i t y for gas compressors - C a p a c i t y for gas injection pumps - Capacity for separation vessels - Capacity for oillwater separation - Well capacity = f (qT, WOR . . . . ) - Etc . . . .
For all fields developed to date in the North Sea, one can compute the percentage of STOOIP (or reserves) that is produced per year (on plateau). The chosen plateau rate should be checked against a good analogue. Also, history has shown that on average in one specific province, only 65% of the (pre-production) estimated plateau rate was achieved. In nearly all cases, water injection rates (injectivities) were overestimated.
Processing capacities (topside facilities) For a given optimum oil plateau rate, the next task is to determine the optimum capacity choice for all potential bottlenecks in the topside system (Table 2). This is done by running the reservoir simulation model with a fixed oil plateau level and an increasing gas and/or water handling capacity roof (Fig. 22) to see how much extra oil is generated or how much oil is accelerated, etc. Also, here, one comes to a point where additional handling capacity no longer generates sufficient value to pay for the increase in weight/cost/OPEX. In some 'size windows', a small increase in capacity may add a disproportionately large step-wise increase in weight/cost (Fig. 23). A plot of NPV against handling capacity (Fig. 24) will yield the optimum choice. Again, a full economic analysis is necessary for each sensitivity run. For many platforms currently in operation, one has underestimated the need for gas and/or water handling capacity has been underestimated. The platforms have
Choosing between rocks, hard places and a lot more: the economic interface
30:3
Must Change Concept Due to Weight
I 1 to2 [ Trains I , J
T 10 000
20 000
30 000
Water Handling Capacity, (Sm3/D)
Fig. 23. When "one more Sm3/D '' capacity gets really expensive.
I q~
Plateau Fixed
100% Regularity I
I
V |
Wells
| | | | | | | | | | | | | | | | | | | | | | | | |
NPV (Of revenues and costs)
T
NPV
I
10 I
20 I
30 I
p
Water Handling Capacity (xl03Sm 3) --"-I
Fig. 24. Locating the optimum amount of water handling capacity.
become facility-constrained after a surprisingly short time and expensive retrofits have been necessary, if at all possible. Some platforms have demonstrated an impressive handling capacity increase by cheap de-bottlenecking. Flexibility (and contingency plans) to compensate for production profile uncertainty is discussed later. This large number of (too early) facility-constrained platforms in the North Sea led L.E Dake (pers. commun., 1989) to note that in oil fields, 'oil is a by-product'! Here, it pays to focus on the oil field revenue 'enemies', i.e. gas and/or water.
Regularity (= no. of days the platform is in full operation divided by 365) Fig. 11 shows that a facility's operational regularity influences the income stream. Designing for 100%
{Regularity
}---~-
Low oil price
NPV of profits
High oil pricel
~
Optimum curve
Fig. 25. Regularity is expensive; the optimum value is a function of oil price expectation.
regularity is very expensive (you basically need a double set of everything). It turns out that the optimum regularity choice is a function of future oil price expectations. Consider Fig. 25, where NPV of revenues (at a high and a low oil price) and costs are plotted vs regularity. We see that the maximum distance
304
H.H. Haldorsen
Ship Semi
~
GBS
~l-"
TLP DCF
Weight & Cost
I 5
10
20
30
I
I
I
,,
,
No. of Wells, Nw
Fig. 26. When "one more well" gets really expensive.
between revenues and costs (points A and B) occur at two different regularities. Hence, if you think the future oil price will be low, you do not buy 'Cadillacs', and you lower the regularity requirement. Conversely, if you believe in a 'hockey stick' oil price forecast (Fig. 7a), you can afford anything and largely disregard the price of your choices!
Number of wells Offshore, the number of wells is a very critical design variable since it translates directly into the necessary number of 'slots' on a platform, the necessary number of subsea wells or the necessary number of risers in a turret. From a purely reservoir engineering point of view, the choice of the number of wells would
be based on plateau rate, plateau length, recovery at end-plateau, project lifetime, and ultimate recovery considerations. In the final analysis, however, a plot of NPV against the number of wells (Fig. 21b), where the concept may change as the number of wells is changed (Fig. 26), will determine the technical-economic optimum. Nearly all fields in the North Sea have made good use of all their 'slots'. Some fields now have twice as many wells as slots since wells have been drilled twice, in new oil pockets when the water cut became too high in the old positions. Undetected (pre-production) faults which compartmentalize the reservoir much more than anticipated (Ruijtenberg et al., 1990), discontinuous pay (between wells), overestimated (unachievable) oil plateau rates and water injection rates have led operators to wish that slots beyond those installed were available.
Types of wells Several types of wells should be considered for application: - 'Slim hole' or coiled tubing for exploration. - Vertical wells. - Horizontal/high angle/extended reach/'designer'/ multi-bore wells (Table 3, Figs. 27-29). -Combination wells: first horizontal oil producer (from a thin oil zone) and when gas/water rates become too high, conversion to gas producer (Fig. 29). Example: Gamma North, Oseberg Fields, Norway. - C o m i n g l e d completion: simultaneous or phased perforation of vertically separated hydraulic units.
Fig. 27. Horizontal wells; typical applications (source: Kydland, 1992).
Choosing between rocks, hard places and a lot more: the economic interface
305
Fig. 28. The Troll oil development is made possible by horizontal wells (courtesy: Norsk Hydro a.s.).
-
-
-
Artificial lift: gas lift vs pumps or contingency for same (important for well design). Hydraulic fracturing of producers/injectors or surfactant 'clean-up' of injectors in the oil-zone to increase krw towards 1.0. Ratio between injectors and producers: a very important design variable. In a waterflood operation designed for pressure maintenance, one should every day be able to inject: qwi - -
qo Bo + qwp
Table 3 Why horizontal wells? (from Kydland, 1992) -
Increased recoveries Higher production rates/accelerated production Fewer production/injection wells Increased reach from platforms (avoid expensive sub-sea wells) Developments of otherwise uneconomic/marginal fields Several reservoirs/fault-blocks from the same well More information about reservoir properties
(1)
where qwi is water injection rate (Sm3/d), qo is oil production rate (Sm3/d), Bo is oil formation volume
factor (m3/Sm3), and qwp is water production rate (Sm3/d).
306
H.H. Haldorsen
..... Gas:::::::i:i:: ....... . . . . . . . . . . . . . . . .iiiiiii!iiiiii:: .... j
~iiiiiii!iii!iii(~i
~'1i1111111111111111111111111111111111111111111~
Fig. 29. Combination well; first, horizontal oil production well. Second, gas production well.
Eq. (1) also influences the choice to be made among pattern floods, contour floods, periphery floods and updip/downdip combined floods as discussed earlier.
Positions of wells In a flooding operation and even for optimum pure depletion scenarios, the operator's ability to really 'outsmart nature' is a strong function of the selected well positions. The structure map, the number and sealing capacity of faults, the internal geological architecture and many other input variables are, however, quite uncertain prior to commencement of production. Hence, 'trial and error' of alternative production and/or injection well positions, with a view to maximizing the performance of wells positioned in an assumed correct reservoir representation, is much like fishing in an aquarium; the hook can be placed fight in front of the fish (because we can see both the fish and the hook). An alternative approach would be to: (1) Have three different geophysicists interpret the (same) 3-D seismic data (in isolation). (2) Review the three alternative interpretations (alternatives, not just sensitivities) to locate the most and least robust potential well positions/areas. (3) Arrive at optimum (x,y) positions (within the robust areas) for wells through 'cheap' simulation runs where hundreds of different positions are considered and simulated by front-tracking (Screide et al., 1992) or streamline models where the vertical flow behaviour is captured (by first doing cross-sectional simulation runs) in a pseudo-relative permeability curve which is put into the 2-D well position optimization models. Some companies find it so unreasonable to optimize well positions in an assumed perfectly known
reality that they multiply the optimized rates and recoveries by, for example, 0.85 to compensate for 'having gone fishing in an aquarium'. In many fields redrilling and sidetracking, or turning planned producers into injectors, have been necessary as the structure map has changed with each new well (making the old 'optimum' position impossible to live with). Clearly, new, fast, cheap and easy-to-use tools for finding optimum well positions are needed. When 3-D, 3-phase numerical simulators (with topside facility constraints, wellbore hydraulics and 1,000,000 grid blocks) get very cheap to run for a 25-year simulation period, the optimum well positions may be found by running hundreds of position possibilities overnight (Fig. 18). Until this is possible, cheaper tools for relative ranking of alternative positions should be developed. Also, the ability to see 'moving fronts' on the screen while running the simulator is now possible. From visual inspection, the operator may then terminate runs with wells in obviously 'unsmart' positions. As we drill up the field and obtain dynamic (including 4-D seismic) data, optimum new well positions in a history-matched model do not usually coincide with pre-production (x,y) picks! 'Plan-Do-Check-Adjust' is an exercise that never ends.
Tubing diameter and gas lift optimization As pointed out by Carroll and Horne (1992), traditional analysis of production systems treats individual 'nodes' one at a time, calculating feasible, but not necessarily optimal, solutions. The approach suggested by these authors consists of two phases: the development of a model that determines the economic benefit of a well, and the optimization of a well's economic benefit with multivariate optimization techniques. Several objectives (NPV, cumulative recovery, total investment per barrel produced, etc.) may be used to optimize, for example, the initial tubing diameter and what tubing changes should be made (if any) through time as the state of the reservoir and the well-stream change. For field developments relying on gas lift, Edwards et al. (1990) and many other authors show techniques for arriving at the optimum amount of allocated gas lift gas per well as well conditions (PI, water cut, GOR) change through time.
Perforated intervals Again, 'trial and error' runs on detailed (finely gridded) cross-sectional models (with a variable width to mirror changing fluid velocities in the 'flow tube' under study) is the most common procedure. Operational
Choosing between rocks, hard places and a lot more: the economic interface
307
Fig. 30. Offtake strategy optimization; many oil rate per well combinations yield the desired plateau rate. The choice (initially) and over time is ours!
issues such as workovers (i.e. regularity issues) must be considered. We want a low OPEX and a high regularity apart from sweep and other reservoir considerations (perforation strategy will also influence handling capacities). Vertical shot density variations (per zone) may be used to control (optimize) the production/ injection profile. Also, opening/closing of different zones through time is a possibility. If subsea wells are chosen, gravel packing (if sand production is a potential problem) may be a good investment since later interventions in subsea wells are very expensive.
Optimization of Offtake Strategy Offtake strategy can be defined as the science of determining (optimizing) which areas or formations in fields/wells should be produced on a given day and their associated rates. Offtake strategy affects both plateau length and recovery. The main issues are: - R a t e or capacity allocation per well or area or field prior to reaching the installed (total) liquid/ water/gas handling capacity limit (Fig. 30). - R a t e or capacity allocation per well or area or field after reaching the installed (total) liquid/water/ gas handling capacity limits.
While on plateau At the onset, all wells have low GORs or water cuts and excess capacity exists in the gas/water handling system. Therefore, offtake strategy has not yet had to address the allocation of a limited resource (i.e., gas/water handling capacity). The major objective of offtake strategy during the plateau period is to use the excess gas/water handling capacity to maximize ultimate recovery (and economic indicators) while maintaining the plateau (oil) offtake rate.
Post plateau Increasing GORs or water cuts will eventually eliminate the excess gas/water handling capacity. After the decline point, subsequent increases in GORs/ water cuts will force corresponding decreases in oil rate. The objective of offtake strategy in the postplateau period is to allocate the available gas/water handling capacity in such a way as to maximize (for example) the present worth of the remaining reserves. This is done by performing gas/water allocation studies; relationships between present worth and amount of gas/water allocation are established for different types (e.g., coning vs cusping) of high GOR/ water-cut wells (or areas, formations, fields). Usually, reservoir simulation models are set up for each well type (or area or field) to evaluate the production performance as a function of increasing amounts of allocated handling capacity. In this manner, relationships between allocated gas/water offtake, instantaneous rate, ultimate recovery and present worth is found for each type. The optimum allocation-split is finally found by linear programming or calculus-based solution techniques. In other words, we seek to operate the wells through time (or areas of a field or different formations in a field or several fields using the same facilities) in such a manner that we maximize an objective function whilst observing the correct constraints on all levels. It is therefore essential to have facility models with all constraints 'on top of' the simulation model (Miertschin and Weiser, 1989) for all relevant constraint levels (well, platform, transport line, etc.). The offtake strategy optimizer is looking for offtake strategies which maximize one or several of the following: the daily production rate, ultimate recovery, NPV, etc., by trying several Well Ranking Schemes (WRS). The WRS' allocate production on each time step based on user-specified criteria (e.g.,
308
maximum production from high water cut/GOR wells or the opposite, minimum production from high water
cut/GOR wells, etc.). This decision after each time step can be made by a 'brain on the simulator'; decisions are made based on hard-coded logic in terms of what to do in a given circumstance. A 'smart' offtake strategy also shortens the economic life. The economic limit is reached when the current cash expenditures associated with a production rate start to exceed current cash receipts. This is a great time for someone in the neighborhood to inquire after leasing any spare processing capacity you may have!
The 'what-if' machine At this point it is clear that a seemingly endless amount of model runs are necessary to check ideas, to locate the various optimum field development choices and to expose the uncertainty inherent in the future production forecasts. Consider a hypothetical field where: - five different gas sales rate scenarios, five different total liquid handling capacities, - five different horizontal well numbers, two different well lengths, two different vertical well positions, are to be simulated in order to arrive at optimum choices for each. In addition, fifteen different geological and other input variables are considered uncertain (i.e., have specified low, base case, and high input values). If all combinations should be investigated, 15,500 runs would be necessary! For this reason, it is common practice to run operational sensitivities (to locate optimum choices) with a base case or reference reservoir description. When the optimum choices have been made, the uncertainty level, due to subsurface uncertainties, is found by running geological sensitivities one parameter at a time or based on a strategy of experimentation (Damsleth et al., 1991). In this manner, the range of outcomes for the production profile or the recovery factor can be determined (Ovreberg et al., 1992). Economic simulations will finally reveal the total project-economic uncertainties (Haldorsen, 1991). Should the optimum 'man-controllable' choices not be robust with a changing reservoir description, this could warrant design changes to build in flexibility for the future. If there is 60% chance of rain tomorrow, would you bring your umbrella? If there is 60% chance that the chosen water handling capacity would be insufficient after 2 years, would you install a little more to be prepared? Or would you save platform space for installing a new module later, when or if it is needed. -
-
-
H.H. Haldorsen
Flexibility (contingency) to compensate for outcome uncertainty As shown earlier, a large number of (preproduction) unverifiable assumptions about the reservoir, from faults to relative permeabilities are inherent in the forecasts. 3-D seismic data, long-term production testing and other activities are of course recommended (if 'value of information analysis' (Gerhardt and Haldorsen, 1991) is positive) to reduce as many of these uncertainties as possible. However, since actual and predicted outcomes always tend to differ, operators desire 'cheap' flexibility and options for the future so that they can modify/alter plans if necessary during the project. Some examples of data-gathering activities and uncertainty-flexibility relationships are given below.
Relative permeability Many operational assumptions, design criteria and 'results' are driven by this parameter (Fig. 31): water/ gas handling capacities, injectivity, and the need for, and timing of artificial lift. In the past, overdimensioning of water/gas handling capacities seems to have been the most common means of providing flexibility for this type of uncertainty. The price of this flexibility is, however, vast offshore and the value of knowing for sure if k~w (Sor) is 0.65 or 0.1 in a field ready for waterflooding could be hundreds of millions of dollars! Perhaps pressure coring to obtain as representative data
Fig. 31. Relative permeability impact on production profile (source: Hinderaker, 1991).
Choosing between rocks, hard places and a lot more: the economic interface
as possible would in some cases be a good investment. History matching of long-term tests which triggered two- and/or three-phase flow to the well will also aid in reinforcing your input assumptions.
Aquifer support If the aquifer support is very strong, planned water injection wells may be converted to producers. If the aquifer support is weak, water injectors may be utilized as planned.
Reservoir continuity If the reservoir is more discontinuous than assumed, extra available platform slots and infill drilling options should be planned for. Early pulse/ interference testing/RFT surveying should be initiated in order to reduce the uncertainty regarding pressure communication in the reservoir (e.g., perpendicular to channel sands, across faults, etc.). Long-term production/injection testing before taking major decisions is now commonly performed to eliminate 'disaster-scenarios' and to optimize the development.
Injectivities If water injectivities are less than expected, the formation may possibly be fractured during (cold) water injection in order to achieve the planned water injection rate. Alternatively, a surfactant treatment of injectors in the oil zone may boost injectivities in addition to being a 'mini-EOR' application. Many other examples of uncertainty-flexibility relationships could be mentioned. Insight about experiences in analogue fields and 'post mortem' reviews are essential in order to prepare well for avoiding nature's traps and unnecessary mistakes in the next field development. And by then, Improved Reservoir Characterization (IRC) may again have made it easier 'to choose between rocks and hard places' within a field (Haldorsen and Damsleth, 1993).
Can the industry adapt? The oil industry's ability to adapt is an amazing thing to observe. In 1986, when the oil price nosedived and remained relatively low, the oil industry reacted with some panic; nothing seemed economic or worthwhile anymore should these oil prices prevail. In 1993, the oil price is still at about the same level, if not lower, but development after development passes the companies' 'hurdle rate', is approved and initiated. Cost and schedule cutting, re-engineering of work processes, TQM, new ways of working
309
with contractors (partnering, outsourcing), increased use of 'off the shelf' components, a critical review of functional requirements and standards (but not compromising safety and environmental concerns), smarter solutions, use of existing infrastructure, unmanned platforms, 3-D and 4-D (time-lapse) seismic, horizontal wells, two-phase pumps and many other technological advancements continue to make more and more possible at lower oil prices.
Organizational design in the oil business: the economic interface In the past most oil companies were organized according to function or discipline: the exploration department, the reservoir engineering department, etc. Today, many companies are moving away from only this dimension of organizational design (claiming that 'science is king' in this setup when 'making money' should have been) into a composite design consisting of Business Units (BUs) where the name itself is supposed to guarantee a financial (asset management) focus. BUs are often defined based on assets, divided up according to geography. Science is not forgotten in the new BU-setup: the many project teams within BUs are staffed with brilliant and experienced personnel from the company's discipline groups (reservoir engineering, production geology, etc.). The advantage of BUs is supposed to be: increased ownership for everyone involved, healthy competition, focus on money and asset management in one dimension (BUs) and focus on competence/technology/science/R&D in another (the discipline groups). Although clearly 'in vogue', there is no empirical evidence that demonstrates that a BU-setup is better than the old-fashioned functional setup. Empirical evidence shows, however, that changing from one to the other is a time-consuming and often expensive process as external consultants 'help out' for a price perhaps equivalent to participating in a well which might yield a commercial discovery! Perhaps one should attempt to make a 'value of reorganization analysis' prior to embarking on it. Many companies note that the process of going from one design to another is useful in itself: old truths are questioned, alliances and loyalties must shift, and new and exciting goals 'empower' the new organization.
Other oil business issues with an economic interface This paper has been limited to discussing the most 'classical' upstream oil business issues where
310
H.H. Haldorsen
the economic interface plays a decisive role. Other important issues have been omitted, such as: Unitization (redetermination) of fields, Environmental issues, 'physical' (operational) risks, Infrastructure and transportation agreements, downstream projects, and gas sales contracts, Choosing between offshore loading and a pipeline to shore, - T h e nature of tax/royalty regimes or PSAs in various countries, Setting a price on (discovered) oil in the ground, Project financing and insurance; evaluation of 'political risk', - The 'correct' discount rate to use for a company, Formal definitions of all project-economic indicators and a discussion of their strengths and weaknesses when used for decision making, Portfolio balancing, strategic E&P planning, - Theories for prioritizing among competing E&P projects, Corporate finance issues (tax effects, financial ratios); balance sheet management, instruments for managing petroleum (market) risks (contract/spot-pricing, forward/ futures trading, netback arrangements, crack -
-
-
-
-
-
-
-
-
-
S
h
o
r
t
-
t
e
r
m
spreads, option pricing, derivatives . . . . (Rennemo, 1991), - SEC reporting of reserves, Bidding and contract strategies, tariff issues, Joint Bidding Agreements, - Abandonment Agreements, - Capacity Allocation Agreements, Joint Operating Agreements (JOAs), Unit Agreements, Pre-unit or Planning Agreements, - Instruments for managing petroleum risk at government level (Oien, 1991), - And many, many other important issues with fiscal implications. It just shows that 'oil and money' are so interrelated that they, in effect, are inseparable; in this business, the economic interface shows up everywhere! As demonstrated in this paper, the economic interface is essential for choosing between 'rocks and hard places' for arriving at optimum field development plans, and for quantifying the associated uncertainty levels by passing on relevant error bars from 'geology to economy'. Exploration ventures, field developments and in fact the whole E&P business are based on a large number of estimates and assumptions which 'stack -
-
-
-
-
Fig. 32. Why the oil industry is a "risky business".
Choosing between rocks, hard places and a lot more: the economic interface
311
NPV (10%)
109NOKS(1987)
3.0
Invest~rnentCosts 2100 ~-20-10.1.0
O"~\~\Ge ,.~,.co~er~b~e rt~, _,,,~eS teS~'-
t
10 20 30 40 50'~'~7"0~~
% change relativeto 100 basecase
-2.0 -3.0 Fig. 33. Sensitivity analysis; one parameter at a time ("spider diagram").
Grisk I
543-
-3-
-41%
P(NPV>0)
99%
Cum. Probability
Fig. 34. NPV vs probability of NPV being larger than the value on the Y-axis (Grisk = good risk).
up' (Fig. 32). Even if only optimum decisions apparently have been made, a range of 'parallel imaginary realities' should be considered (either one at a time, Fig. 33, or by allowing simultaneous variations (by doing economic simulations, Fig. 34), or by looking at all NPVs at the end of each branch in the alternative outcome or decision tree, Fig. 12) to expose the possible financial outcome range. In the oil business, all players agree that there can be no harvesting of 'grisks' without first taking some (calculated) 'risks'.
List of symbols Bo
--
CAPEX C DCF Dffl
= = =
Formation Volume Factor, RB/STB (res m 3/stock tank m 3) Capital Expenses, $ Value of Information, $ Deep Concrete Floater Direct Hydrocarbon Indicator
E E&P EOR EPC
= = =
f
= = = -
g G G&G GRV GBS GOR GRISK HCPV H IOR IRe JPO&G
-
-
-
= = = -
Net Present Value, $ Exploration and Production Enhanced Oil Recovery Engineering, Procurement and Construction Contract Probability, % or fraction Gas Weighted Average NPV, Geology and Geophysics Gross Rock Volume Gravity Based Structure (platform) Gas-Oil Ratio Upside Potential at a Certain Probability Level Hydrocarbon Pore Volume, m 3 High Value Improved Oil Recovery Improved Reservoir Characterization Jurassic Park Oil and Gas Company
3 K
Krw L M MWD NPV N N/G OGIP OPEX PDO PI PSA P QOlL,P
qo qwi qwp r
RF RFT RISK
STOOIP SW TQM VSP WAG WOR WRS x
y
1 = = ---= --= = = -= = = = = --= --= = =
2
H
.
Permeability, m 2 Relative Permeability to Water Low Value M e d i u m Value Measurement While Drilling Net Present Value, $ Number of Years/or Wells Net-to-Gross Ratio, fraction Original Gas in Place, Sm 3 Operating Expenses, $ Plan for Development and Operation Productivity Index Production Sharing Agreement Probability, % or fraction Oil Plateau Rate, Sm3/D Oil Production Rate, Sm3/D Water Injection Rate, Sm3/D Produced Water Rate, S m 3/D Discount Rate, % or fraction Recovery Factor, fraction Repeat Formation Tester Downside Potential at a Certain Probability Level Stock Tank Oil Originally In Place, Sm 3 Water Saturation, % Total Quality Management Vertical Seismic Profiling Water-Alternating-Gas Drive Water-Oil Ratio, fraction Well Ranking Scheme Position, x-coordinate, m Position, y-coordinate, m Porosity, % or fraction
Acknowledgements Thanks are expressed to Norsk Hydro for permission to publish this paper, to Sidsel Gaustad, Kit Smoot, and Duoc Vikan for typing and revising the manuscript, and to Nina Anthun making figures and tables. A special thanks to my colleagues Hans Traaholt and Aage Frohde for valuable input and suggestions.
References Austvik, O.G., 1991. Limits to oil pricing; a scenario planning approach. The 12th Bergen Conference on Oil and Economics
H
.
Haldorsen
m Business Risk in the Oil industry, April 22-23. Behrenbruch, R, 1991. Offshore oilfield development planning: project feasibility and key considerations. SPE Paper No. 22957 presented at the SPE Asia-Pacific Conference held in Perth, Western Australia, 4-7.11. Bradey, R., Myers, and Stewart, C., 1988. Principles of Corporate Finance. McGraw-Hill, New York. Boye, K., 1985. Finansielle Emner. Bedrifts~konomenes Forlag, Oslo. Carroll III, J.A. and Home, R.N., 1992. Multivariate optimization of production systems. JPT, 44(7): 782-831. Damsleth, E., Hage, A. and Volden, R., 1991. Maximum information at minimum cost: a North Sea field development study using experimental design. SPE Paper No. 23139 presented at Offshore Europe, Aberdeen. De Ruiter, H., 1993. From a presentation made at the Howard Weil Energy Conference, New Orleans, March 22. Dyrhaug, L.T., 1986. Kan ckt integrasjon stabilisere oljemarkedet, Norge Industri No. 4. Edwards, R., Marshall, D.L., and Wade, K.C., 1990. A gas-lift optimization and allocation model for manifolded subsea wells. SPE Paper No. 20979 presented at EUROPEC 90, The Hague. Energy Information Administration, USA, 1991. Data for Fig. 6. Gerhardt, J.H., and Haldorsen, H.H., 1991. On the value of information. SPE Paper No. 19291. Haldorsen, H.H. and Damsleth, E., 1993. Challenges in reservoir characterization. AAPG Bull., 77(4): 541-551. Haldorsen, H.H., 1991. Reservoir characterization. 6th European Symposium on Improved Oil Recovery, Proc., Vol. 3, Stavanger, pp. 59-68. Hinderaker, L., 1991. Verdien av Initielle Data og Betydning ved Planlegging av Feltutbygginger; Et Tilbakeblikk p~ Felt i NordsjCen (internal presentation). Kydland, T., 1993. Horizontal wells; applications and economic advantages. Paper presented at the conference How to Develop Cost-Reducing Strategies Within the E&P-Industry and Offshore Developments, Oslo Plaza, Oslo. Lohrenz, J., 1988. New values of our information. JPT. Manne, A.S. and Schrattenholzer, L. 1987. International Energy Workshop. Overview of Poll Responses. Energy Modelling Forum, Stanford University, July. Miertschin, J.W. and Weiser, A., 1989. A flexible approach to predictive well management via user-defined strategies. SPE Paper No. 19848 presented at the 64th SPE Annual Technical Conference and Exhibition held in San Antonio, Tex. Rennemo, S., 1991. Managing Petroleum Risks at the Corporate Level. PETRAD Publication No. 4, Selected Lectures. Ruijtenberg, P.A., Buchanan, R. and Marke, R, 1990. Threedimensional data improve reservoir mapping. JPT, pp. 22-25, 59-61. Saleri, N.G., 1993. Reservoir performance forecasting: acceleration by parallel planning. JPT, pp. 652-657. Screide, I., Flach, T.A., and Rian, D.T., 1992. Automatic optimization of well locations using a front tracking reservoir simulator. SPE Paper No. 24278 presented at the European Petroleum Computer Conference held in Stavanger, pp. 171-182. Ovreberg, O., Damsleth, E. and Haldorsen, H.H., 1992. Putting error bars on reservoir engineering forecasts. JPT, pp. 732-738. Oien, A., 1991. Managing Petroleum Risk at Government Level. PETRAD Publication No. 4, Selected Lectures, pp. 93-97.
H.H. HALDORSEN Norsk Hydro a.s., P.O. Box 200, N-1321 Stabekk, Norway
313
The usefulness of resource analysis in national economic
planning --examples from the Norwegian Shelf Jan-Erik Kalheim and Harald Brekke
Petroleum resources have an important influence on the economic life in Norway. To the official authorities it is therefore essential to have good and detailed knowledge of these resources. Questions like "what are the total resources", "how much of the resources have been produced" and "how much is expected to remain" should be answered. A classification system and an updated resource account is therefore required. Even though there is currently a well defined classification system and a resource account which is annually updated, estimates of discovered resources are associated with considerable uncertainty for some fields and discoveries. The uncertainties are not only connected to the mapping of field sizes but also to the possible potential for improved recovery, particularly in the light of new or sophisticated recovery methods. During the last two years the Norwegian Petroleum Directorate (NPD) has paid considerable attention to the potential for improved oil recovery. This is because of the time-critical aspects in the producing fields. The assessment of undiscovered resources also includes large uncertainties. However, regardless of the uncertainties such assessments are important for long term national planning. Decisions concerning opening of new areas for future exploration will have a considerable impact on the future level of the petroleum activities. To national authorities it is therefore important to have an understanding of what economic potential the undiscovered resources may represent. To resource analysts it will be important to know how the results from the geological assessments are used in the economic calculations. Different resource assessments of a single play model are compared, and a comparison between different prospect evaluations of one single prospect are also presented. This approach is used to focus on the most critical uncertainty factors.
Introduction The intention of this paper is to give a brief overview of how resource analysis and the results from such analysis are used in Norway. Topics covered include the impact of petroleum resources on the national long-range planning and a general classification of the Norwegian petroleum resources. Also discussed are uncertainties related to the field evaluations, improved oil recovery and undiscovered resources, including prospect evaluations, play assessments and economic evaluations. The petroleum resources on the continental shelf have contributed greatly to the economic life and growth in Norway from the beginning of the 1970s. At the time of writing (December, 1993) the estimated total expected recoverable petroleum resources are on the order of 10 billions tonnes of oil equivalent (Fig. 1). About 5.9 billion t.o.e. (tonnes of oil equivalent) have been discovered, with an oil/gas proportion of about 40/60. The discovered recoverable resources are found in more than 120 fields and discoveries. Forty-one fields, constituting of about 3.85 billion t.o.e, have been produced, are in production or have been decided to be developed, and about 80 fields
and discoveries are under planning or evaluation. The recoverable reserves in the producing and developing fields may be increased by about 0.44 billion tons of oil by implementing improved oil recovery methods. The undiscovered resources are estimated to be about 3.67 billion t.o.e., with an uncertainty range from 2.1 to 6.0 billion t.o.e.
National economic prognoses Petroleum activity plays a considerable role in the Norwegian economy. For example, in 1990, about 14 percent of the Gross National Product (GNP) was related to the petroleum sector. The investments in this sector rose to almost 50 billion NOK (Norwegian Kroner) in 1992, and were more than three times as high as in the land-based industry (Fig. 2). These figures also give an indication of how important the offshore activity is for the Norwegian land-based industry dealing with delivery of goods and services. Most of the Norwegian Government's net income from the petroleum sector is derived from direct taxes, production fees and area fees. The development of these income figures has mainly followed the oil price. In addition, in 1985 the Government divided
Quantification and Prediction of Petroleum Resources edited by A.G. Dor6 and R. Sinding-Larsen. NPF Special Publication 6, pp. 313-324, Elsevier, Amsterdam. 9 Norwegian Petroleum Society (NPF), 1996.
314
J.-E. Kalheim and H. Brekke
Fig. 2. Comparison between investments in the petroleum sector and in other Norwegian industry. (Source: Norwegian Petroleum Directorate, 1992.)
The usefulness of resource analysis in national economic planning h examples from the Norwegian Shelf
315
Fig. 3. A comparison between the net cash-flow, presented in Fig. 3a, derived by the Norwegian petroleum activity and the historical development of the oil price, presented in Fig. 3b. (Sources: Ministry of Industry and Energy and The Government's Long-Range Programme 1994-97.)
the Norwegian state oil company (Statoil) owner share into a State owner share (SDI) and a Statoil owner share. Because of major investments the State owner share presented a negative cash-flow in the period 1985 to 1988, but today this share gives a positive cash-flow. In 1991 the State net cashflow was about 40 billion NOK, in which about 5 billion. NOK were derived from the SDI and the remaining part from taxes and fees (Fig. 3). Since 1991, the petroleum sector has also paid a fee related to emission of carbon dioxide. In the years 1991 and 1992 this fee was in the order of 2 billion NOK annually. The petroleum sector contributes heavily to the welfare of Norwegian society. Major industrial environments with a large number of employees exist as a result of the petroleum resources. During the last 5 years more than 60,000 employees have been at work in petroleum-related industries. Such industries are connected to exploration and production, delivery of goods and services and refining of petroleum products. The international orientation of this industry, and the great demands for quality and technology,
have also given important impulses to other industrial activities in Norway. Fig. 4 shows an overview of the contribution that the petroleum sector played in the Norwegian economy, in the years 1985 and 1993, and the prognosis for the years 2001 and 2010. In 1993, about one third of the total national export value from Norway was related to sales of oil and gas. The investments in the petroleum sector in 1993 are believed to be more than one third of the total investments in Norway, while the net income, measured in % of the GNP, is estimated to be only 3%. By the shift of the Century, the investments are expected to be reduced by half, while the net income is predicted to be doubled. About 20 years from now, petroleum exports and investments are predicted to play a less important role in the Norwegian economy, but the net income from this sector is still predicted to be about 5% of the GNP. These figures indicate that the petroleum sector in Norway has a long-range perspective, and is expected to have a considerable impact on the Norwegian economy in the future. We should, however, keep in mind that this sector is very capital demanding.
316
J.-E. Kalheim and H. Brekke
Fig. 5. The prognosis for the net cash-flow derived from petroleum activity until year 2020. (Source: The Government's Long-Range Programme 1994-97.)
Therefore, the petroleum industry is highly vulnerable to oil prices and market conditions. When the Ministry of Finance develops prognoses for the next 20 to 30 years, the assessment of petroleum resources plays a very central role. Fig. 5 shows the prognosis for net cash-flow derived from petroleum activity until the year 2020. The State net cash-flow is assumed to increase considerably, particularly since the investments related to the State Direct Interest will decrease in the future. The oil price assumption behind this prognosis was 130 NOK/BBL, while the oil price in November 1993 was about 105 NOK/BBL. All these previously mentioned prognoses are based on estimated production profiles, which are worked out on basis of estimates of the total petroleum resources.
Petroleum resources
Production profiles for the Norwegian Continental Shelf are presented in Fig. 6. The prognosis for future oil and gas production is based on existing production plans for fields in production and under development, as well as on expected production plans for discoveries under evaluation. In addition, a production profile based on expectations of future discoveries, and the most likely potential for improved oil recovery is added at the top of this figure. In the future Norway will also become a considerable gas-producer. Fig. 6b shows the expected future distribution of oil and gas production. This prognosis includes both contracts already agreed upon and contracts expected in the
The usefulness of resource analysis in national economic planning m examples from the Norwegian Shelf
2001 a) 160
120
80
40
[i:..:.":
0-7" 1971
".':-" .::.':ii
1980
1990
2000
2010
2020
IOR Fields under evaluation Fields under development Fields in production Historical production
200
160 -
317
future. About 20 years from now, gas production is expected to be more than 50% of the total petroleum production in Norway. On the Norwegian Continental Shelf (Fig. 7), a total of 40 oil and gas fields will be producing in the near future. Thirty-eight of these fields are located in the North Sea, consisting of 28 fields in production and 10 fields which are decided to be developed. One gas field in the North Sea is already produced. The remaining two fields are both located in the Norwegian Sea: these are the Draugen Field which came on-stream in October 1993 and the Heidrun Field which will start production in 1995. In the Barents Sea, there are no fields either in production or under development. In addition, by December 1993 about 80 fields and discoveries are planned for development or under evaluation. Fifty-five of these fields and discoveries are located in the North Sea, of which 11 are fields planned to be developed and 44 are discoveries under evaluation. In the Norwegian Sea two fields (SmCrbukk SCr and Midgard) are under development planning, and 11 discoveries are under evaluation. However, some of these discoveries in the SmCrbukk-Midgard area are awaiting a future gas transport solution for the Haltenbanken area. In addition, a very aggressive plan has been made for a future development of the Norne discovery in the Nordland II area, north of Haltenbanken. So far 12 gas discoveries are registered in the Barents Sea, but for the time being none are regarded to be of commercial interest. The Snohvit gas discovery is of considerable size, about 95 million t.o.e., but due to its long distance from the gas market even this discovery is regarded as marginal.
Uncertainties in field evaluations 120 -
O/L
0
'~
1971
1980
1990
2000
2010
2020
2030
Fig. 6. Prognosed production profile for oil and gas production, respectively subdivided by using the NPD classification system (Fig. 6a) and by prognosed oil and gas production (Fig. 6b).
After more than 25 years of petroleum activity in Norway we have learned that the uncertainties in resource estimates even for fields in production are considerable. Fig. 8 shows how the reserve estimates have changed for some fields in Norwegian waters. The examples are taken from the Ekofisk, Statfjord, Frigg and Valhall Fields. The Ekofisk Field started production in 1971, while the Frigg Field started in 1977, the Statfjord Field in 1979 and the Valhall Field in 1982. The most remarkable change is found in the reserve estimates of the Ekofisk Field. This is mainly due to an increase in the recovery factor, which increased from an original figure of about 15% to about 30% in 1992. Estimates of the recoverable reserves in the Ekofisk Field have increased from about 250 million t.o.e, in 1976 to about 480 million t.o.e, in 1992. This field originally had the largest oil-in-place
318
J.-E. Kalheim and H. Brekke -30"
"
-200
~
~.~,;~"
"
/
M,YE./
/
,~:/-.;.."~-RIDGE /
;ii!i;;! i;i:~-., /
'"-----/._ .....o"
ii':/
0*
10"
20*
30*
~:':':':'~':
/"~--../il]]]]]]i;~}., .........~ ~
/
70*
-10"
/
/ ~
/
.#
/
i
40*
i ...... '
50*
'
'
", ""-~-~8~
~:!:i:~ ,,::i:]:i~;:~:!:;:i:!!!i';!;!;i~ii!;;;!i!!:::i;/
i::i~iiiii::!iiii::i::i::i!iiiii::iii!if::ili! ::i:ii ::ii1 BARENTS 70*
NORWEGIAN
Ii~ Opened for exploration 60*
New areas opened for expl. I~[~ Opened for exploration on seperate conditions SEA
60*
Not opened for exploration 20~
30
Fig. 7. Map of the main exploration provinces on the Norwegian Continental Shelf.
Fig. 8. Historical reserve estimates for some Norwegian oil and gas fields. The estimates are taken from the figures published in the Norwegian Petroleum Directorate annual reports (1975 to 1993).
The usefulness of resource analysis in national economic planning m examples from the Norwegian Shelf
319
Fig. 9. Standard deviation in some field and discovery estimates. The reserve and resource estimates were carried out by different oil companies and institutions for five fields and discoveries in the Norwegian Sea. The standard deviations are plotted as percent of the average value.
reserves in the North Sea. The changes in reserve estimates for the other three fields are partly due to an increased recovery factor and partly due to alterations in the estimated in-place volumes. This is mainly a result of new seismic and geological mapping, and/or improved reservoir models. These annual trends in reserve estimates may lead us to conclude that all field estimates are going to increase as a function of time. However, we should keep in mind that the present estimate of the Statfjord Field has still not reached the initial resource estimate of 1975. Another example of uncertainties is taken from some field evaluations carried out by different oil companies and institutions at about the same period. In Fig. 9 the average reserve estimate for each field is set at the zero-line, while the standard deviation is presented in percentages, relative to the average estimates. These variations may also be related to different resource classification systems, with respect to terms such as "proven", "expected", and "probable" resources.
Improved oil recovery The uncertainty in the reserve estimates is also a function of possible potential for improved recovery. This relates partly to new and sophisticated recovery methods and partly to conventional methods not implemented in the original production plans. Fig. 10 shows an overview of the technical maturity of improved oil recovery (IOR) methods for
the Norwegian Continental Shelf. The two columns to the right represent methods which have been implemented for entire fields or as pilot tests. The three columns to the left give an indication of ongoing studies. The length of the bars indicates the activity level for each method. The total potential for improved oil recovery by December 1993 is estimated to be about 440 million t.o.e. However, the uncertainty in this estimate is considerable, and the range in possible values is from 250 to 800 million t.o.e. During the last two years NPD has paid considerable attention to this factor, not least because improved oil recovery from producing fields has an important time critical aspect for the next 5 to 15 years. The official definition of improved oil recovery is presented in a published report by the NPD (NPD, 1993b): Improved Oil Recovery refers to actual measures resulting in an increased oil recovery factor from a reservoir as compared with the expected value at a certain reference time. This definition could also be expressed in a more broad and simple way: Improved recovery of petroleum is regarded as any increase, without any consideration of methodology, which is not related to an increase in the estimate of in-place volumes. This includes conventional as well as advanced recovery methods. NPD has also carried out calculations on the profitability of improved oil recovery projects (Fig. 11). The results indicate that projects accounting for about 65% of the presently identified improved oil recovery potential have an acceptable profitability. The aim of
320
J.-E. Kalheim and H. Brekke
Fig. 11. Profitability of increased oil recovery projects. Income and cash flow is estimated from the profitable part (290 million t.o.e) of the total potential for improved oil recovery (440 million t.o.e).
NPD's strategy is for improved oil recovery to be incorporated in updated field production plans. Undiscovered resources
The total undiscovered resources on the Norwegian continental shelf are estimated to be on the order of
2.1 to 6.0 billion t.o.e, with an expected value of about 3.7 billion t.o.e. The expected oil and gas proportion is about 35/65. Those estimates are based on a play analysis of the entire shelf (see Brekke and Kalheim, 1996, and NPD, 1993a). More than 70% of the total expected resources in the Norwegian North Sea are already discovered.
321
The usefulness of resource analysis in national economic planning - - examples from the Norwegian Shelf
.
The undiscovered resources are estimated to be in the magnitude of 1.1 to 2.4 billion t.o.e, with an expected value of about 1.7 million t.o.e. The oil and gas proportion is expected to be about 45/55. In the North Sea the future prospect sizes are assumed to be rather small on average, since this area has been well explored for more than 25 years. In the Norwegian Sea the undiscovered resources are estimated to be on the order of 0.2 to 2.2 billion t.o.e, with an expected value of about 0.85 million t.o.e. The considerable spread in the estimate is due to the presence of large unexplored areas, particularly in the western part of the Norwegian Sea, and reflects the uncertainty in play models. If some untested play models in areas such as the Vcring Basin should prove to be very productive, this region may have a considerable upside potential. The undiscovered resources in the Barents Sea are estimated to be in the order of 0.4 to 2.6 billion t.o.e. with an expected value of about 1.15 million t.o.e. The oil/gas-proportion is estimated to be about 15/85. Estimates of undiscovered petroleum resources are of great importance to the Government's national planning of future petroleum activity. At the beginning of 1994, the Ministry of Industry and Energy (1994) presented a white paper to the Norwegian Parliament (Stortinget) concerning petroleum activity on the Norwegian Continental Shelf. In the paper the Ministry proposed opening of new areas in the Norwegian Sea for exploration licensing. These included the greater part of the remaining unopened areas in this region, including the Mere- and VCring Basins, and parts of the near-coast areas. However, some of the coastal areas are vulnerable with respect to fishery activities and the ecology. Because of the long distance to the gas market, a future development in the Barents Sea most probably depends on an oil discovery. During the last months of 1993, a group consisting of representatives from the industry and the authorities have been discussing the framework conditions for the future exploration in this region. This working group would not have been possible, unless the participating members had assessed the area and come to the conclusion that this region still might hold considerable undiscovered resources. The results from these discussions were published in the white paper (Ministry of Industry and Energy, 1994), and some changes in the framework conditions for the Barents Sea were proposed. (1) Historically, the Norwegian authorities have clearly expressed that individual licence applications are preferred. Now, the Ministry will encourage group applications for new licences in this area. (2) The application documents should be considerably simplified. (3) The work com-
mitment should in the first phase only consist of a seismic work program. (4) The area fee should be reduced. (5) The participating share should be increased for each company. (6) The licence acreage should increase to about 1 to 2 quadrants (12 to 24 blocks).
Uncertainty in assessment of undiscovered resources Prospect size and the probability of discovery are two important factors, and form the necessary input to the economic evaluation of a given prospect. This evaluation constitutes the basis for deciding whether this prospect should be drilled or not. However, it is important to consider the large uncertainties in such an evaluation. Fig. 12 shows two examples of prospect evaluations for two prospects (A and B) carried out by different oil companies for some concession rounds in the 1980s. These figures plot the expected resource estimates on the horizontal axis and the probability of the discovery on the vertical axis. The spread in evaluations is remarkable. We should expect that high resource estimates correspond to low probability of
PROSPECT A 1 0,9 ~o 0,8 ._o 0,7 "o 0,6 o
~
o,5
e,
0,4
.~ o ~ I1.
0,3 0,2
!
0,1 0
0
20
40
60
80
Estimated resources
100
120
(mill t.o.e.)
PROSPECT B 1
~, 0,9 ~o 0,8
m_
~ 0,7 "o
9
9
9
9
0,6
0,5 0,4 j~ 0,3 o ~- 0,2 0,1 0
9
0
mm
100
9
200 Estimated resources
9
9
mm
300
400
500
(mill t.o.e.)
Fig. 12. Examples of prospect evaluations for two prospects (A and B) carried out by different oil companies in concession round applications in the mid-1980s.
322
J.-E. Kalheim and H. Brekke
Fig. 13. An example of assessments of total resources of the Middle and Lower Jurassic Sandstone Play on the Halten Terrace at four stages in the play history (including the present), made by a number of different groups of assessors. Not all groups completed the work on all four stages.
discovery values, and vice versa. However, there is no such correlation. These prospects were later drilled, and proved to be discoveries. After one wildcat, remapping and a couple of appraisal wells prospect 'A' seems to be between 5 and 10 million t.o.e. The other structure is believed to be between 50 and 100 million t.o.e. The uncertainties are still considerable, even after several wells and 3D-seismic mapping. If we compare some of the input parameters in the different prospect evaluations, the largest spread is related to the volumetric factors, such as gross rock volume, net/gross-ratio and trap fill. In this comparison it is also of interest to note that some companies, about 10 years ago, did not use spread in the volumetric factor in their calculations. We are, however, aware that oil companies differ in their approaches to prospect evaluation. Due to established company practice, some of the companies only use the highest possible volumetric factor in the prospect evaluation. These evaluations form the basis for the economic profitability calculations of a prospect. The key questions are: (a) is this prospect drillable, or worth a concession application from the companies? (b) is it profitable for the Norwegian State to allocate a new production license, which contains this prospect? Looking at the examples in Fig. 12, it is easy to understand why the oil companies differ in their priorities when applying in the Norwegian concession rounds. In the play analysis for the entire Continental Shelf, NPD used the computer program FASPUM developed by the United States Geological Survey (USGS) (Crovelli and Balay, 1988). This computer program and results are presented in more detail in Brekke and Kalheim (1996).
The computer program has also been used in other exercises. NPD is participating in programs funded by the organization Norwegian Aid for Developing Countries (NORAD). As a part of a development program in South East Asia, NPD contributed the technique of play assessment. In order to introduce the play assessment concept to these countries, some examples from plays in Norwegian waters were presented. This situation gave a unique opportunity for Asian geologists, who were not familiar to Norwegian petroleum geology, to make an objective test of the play analysis method. As an exercise, different groups were asked to assess the Middle Jurassic Sandstone Play on the Halten Terrace. Three different scenarios were given related to different stages of the exploration period. The task was to assess the remaining undiscovered resources at each stage. The results are shown in Fig. 13, compared to the real accumulated resource curve for this play. In each scenario, the assessments were based on exactly the same amount of geological information. The figure shows that there is a significant spread in the results, resulting from the way that the different assessment groups interpreted the geological information given in each scenario. Fig. 14 shows a summary of the variation in some of the input parameters for the above exercise. The spread is annotated by standard deviation in percent, where the zero-line is set at the average value for each parameter. The standard deviation is plotted for the minimum, P50 and maximum input values for each parameter in the three given scenarios. The figure shows clearly that the most significant factors are: "rock volume size" and "number of prospects". The number of prospects in a play model is strongly related to how the play is delineated.
323
The usefulness of resource analysis in national economic planning m examples from the Norwegian Shelf
Fig. 14. Variation on some input parameters in the play analysis exercise for the Middle and Lower Jurassic Sandstone Play on the Halten Terrace. The standard deviation is noted in percent, and is plotted for the minimum, P50 and maximum input values for each parameter in the three described scenarios (~ee also Fig. 13).
Probability of discovery also plays an important role. This example shows that the uncertainty in such assessments is considerable with standard deviations from 5 to almost 40%. It is therefore important to pay sufficient attention to these uncertainties when economic prognoses are developed.
Economic
evaluations
2oo
I I I I I
~80 160
14o =; d ._
The economic calculations have several purposes: to estimate the economic potential for all plays in a geological province, and to evaluate and compare different areas. From the Government's point of view, such calculations will have implications for which areas should be announced in the forthcoming concession rounds, and on the exploration strategy in each area. Important economic input parameters in such calculations are exploration costs (seismic and drilling), development costs, production costs (including tariffs) and expected income (oil and gas price) from potential commercial fields/discoveries. The geological input parameters needed for such calculations are estimated prospect size distribution, number of prospects and average probability of discovery. Following the resource analysis published by NPD (1993a), the economists in NPD initiated an internal project to evaluate the profitability of the Norwegian Shelf. Fig. 15 shows an example of their work from the Barents Sea. Based on a set of economic and
I
E
oil price
I
120 100 "'....
/
80
'
60 40
~,~,~"
i
/
I I I I
unit cost
~.~"
~.~.~,'costt~,e~ eX~l.
lo
2o
j
20 0 -5(
4o
ao
2o
io
o
~o
4o
so
Variation in %
Fig. 15. Sensitivity analysis on the economic cut-off value for an oil prospect in the Barents Sea. See the text for further explanation.
geological assumptions, it was estimated that for an oil prospect to be marginally profitable in the Barents Sea, it had to contain reserves in the order of 50 million t.o.e. This calculation was based on the following series of assumptions: (1) the prospect size was based on a log normal distribution, where the total number of drillable prospects was set to n = 2600; (2) the average probability of oil discovery,
324
J.-E. Kalheim and H. Brekke
P -- 0.03; (3) the exploration cost per wildcat was set by historical experience to be 205 million NOK; (4) the unit cost for developing and producing a Barents Sea oil field was set to 105 NOK/BBL; and (5) the oil price was set to 130 NOK/BBL, based on the prognosis in the Norwegian Government's LongRange Programme (at the end of 1993, the rate of exchange was 1 USD = 7.50 NOK). Fig. 15 shows the sensitivity to these economic parameters, and how changes in the parameters influence the result. All the curves in this figure cross at the 0% variation value on the x-axis, and at the 50 million t.o.e, value on the y-axis. By making a change in one of the parameters, but fixing all the others, we are able to read the resulting change in economic cut-off value for an oil prospect in the Barents Sea. For example, by changing some assumptions (e.g. a 30% unit cost reduction, or 100% increase in chance of discovery) the economic cut-off value may easily be reduced to about 30 million t.o.e. When we also know the uncertainty in other geological input parameters, such as prospect rock volumes and number of prospects, it is quite obvious that such estimates should be handled with considerable caution. Examples presented by several oil companies have given economic cut-off values for oil prospects in the Barents Sea ranging from about 25 million t.o.e, to about 100 million t.o.e.
Summary In this paper, we have tried to present an impression of how resource analyses influence the future Norwegian economy by showing some examples from the government's long-range programme. We have also attempted to demonstrate how uncertainties
J.-E. KALHEIM H. BREKKE
occur in all kinds of resource estimates, from field evaluations to evaluation of undiscovered resources on a play model basis. Assessments of undiscovered resources have in some contexts been regarded as exercises for wizards or alchemists. However, we believe it is possible to produce a sensible and scientifically-based prognosis for undiscovered resources, while remaining aware of the inherent uncertainties. These projections are vital to the government agencies, oil companies, rigowners, seismic contractors, purveyor industries and others who must produce plans for future investment in the oil and gas sector.
References Brekke, H. and Kalheim, J.E., 1994. The NPD assessment of the undiscovered resources of the Norwegian Continental Shelf. Background and methods. In: A.G. Dor6 and R. Sinding-Larsen (Editor), Quantification and Prediction of Hydrocarbon Resources. Norwegian Petroleum Society (NPF), Special Publication 6, Elsevier, Amsterdam, pp. 91-103 (this volume). Crovelli, R. and Balay, H., 1988. FASPUM metroc version: Analytic petroleum resource appraisal microcomputer programs for play analysis using a reservoir-engineering model. U.S. Geol. Surv. Open-File Rep. 87-414, 14 pp. Norwegian Petroleum Directorate 1975 to 1993. Annual reports. Norwegian Petroleum Directorate 1993a. Petroleum resourcesNorwegian Continental Shelf. External report, 40 pp. Norwegian Petroleum Directorate 1993b. Increased Oil Recovery Norwegian Continental Shelf. External Report, 35 pp. Norwegian Petroleum Directorate 1993c. Description of the Norwegian Petroleum Directorate's Resource Classification System. NPD, Contrib., 37, 13 pp. The Royal Ministry of Finance, 1992. Stortingsmelding nr. 4 (199293). Langtidsprogrammet 1994-1997, 385 pp. (The Norwegian Government Long-Range Programme 1994-1997; in Norwegian only.) The Royal Ministry of Industry and Energy 1994. Stortingsmelding nr. 26 (1993-1994). Utfordringer og perspektiver for petroleumsvirksomheten p~ kontinentalsokkelen. (White Paper No. 26 (1993-1994); in Norwegian only.)
Exploration Department, Norwegian Petroleum Directorate, P.O. Box 600, 4001 Stavanger, Norway Exploration Department, Norwegian Petroleum Directorate, P.O. Box 600, 4001 Stavanger, Norway
325
Evaluation of undrilled prospects - - sensitivity to
economic and geological factors C. Hermanrud, K. Abrahamsen, J. Vollset, S. Nordahl and C. Jourdan
Economic prospect evaluation at an early stage involves personnel with different skills, such as geoscientists, reservoir engineers, construction engineers and economists. Data are transferred between these groups of people who often have only a vague understanding of the accuracy of the data they receive. This lack of communication naturally limits the correctness of the results. To improve this communication, the complete process of prospect evaluation (including both geological and economical aspects) has been followed here in order to show the different data sets that are transferred and to comment upon their accuracy. Although this paper is based entirely on Statoil's methodology, it is nevertheless believed to be of general relevance. In Statoil's methodology, prospect volumes calculated by geoscientists are given as likelihood distributions. Post-drilling examination of such volume distributions show that historically they have been too optimistic. However, historical prospect risking has correctly identified the most important risk factors and has been able to separate low-risk from high-risk prospects in a satisfactory manner. The number of appraisal wells that are needed before the development of a field can be decided upon is often crucial to the economic evaluations. This number, however, is usually underestimated during the early stages of exploration, probably because data limitations mask reservoir heterogeneities. Reservoir performance is of utmost importance to early economic calculations as it influences both the drilling costs and the production of hydrocarbons vs. time. Of course, reservoir productivity is highly uncertain when judged prior to drilling the first well. Historical data show that reserve estimates of producing fields tend to be upgraded as reservoir depletion proceeds, although several fields have had their reserve estimates downgraded shortly after production start-up. The operational and investment costs are not generally predictable to a greater accuracy than 40% at early stages of exploration. This, and the fluctuations in oil price, adds to the uncertainty of early economic calculations. Sensitivity analysis of calculated net present values for exploration prospects shows that the investigated prospects are sensitive to many different factors, and that a separate sensitivity analysis for each prospect is needed.
Introduction Offshore Norwegian acreage is allocated for petroleum exploration by the Norwegian government in concession rounds which are announced approximately every second year. In response to this, oil companies rank the announced blocks and apply for ownership of the blocks of their choice within a given deadline. The government then composes partner groups for each new licence, which are largely based on the technical quality of the applications and on each individual company's previous experience on the Norwegian Continental Shelf. The major exploration geology task for Norwegian-based oil companies is thus prospect ranking and application for acreage in concession rounds. The two last concession rounds (the 13th and 14th licencing rounds) have both included a large number of blocks covering areas spread all over the Norwegian Continental Shelf, with a significant number of mappable
prospects (at least 50) in each round. Some of these prospects can quickly be written off as being too small to be of economic interest, while others will need a more in-depth study before assessments of their economic potential can be made. The economic evaluation of a prospect includes prospect mapping and risking, as well as calculation of income and expenditure in the case of development. Such economic evaluations are highly uncertain as they include both geological, economic and political risk. The geological risking includes play and prospect assessments of recoverable resources and fluid type. Economic risks include exploration, appraisal, investment, drilling and operational costs, and also income (oil/gas price, total production, production profile vs. time). Note that both cost and income to a large extent depend on geological factors as reservoir performance and heterogeneities influence both appraisal and production drilling costs (number of wells) and
Quantification and Prediction of Petroleum Resources edited by A.G. Dor6 and R. Sinding-Larsen. NPF Special Publication 6, pp. 325-337, Elsevier, Amsterdam. 9 Norwegian Petroleum Society (NPF), 1996.
326
C. Hermanrud, 1~. Abrahamsen, J. Vollset, S. Nordahl and C. Jourdan
well performance (and thereby hydrocarbon production). The political risks cover the economic consequences of changing rules (e.g. nationalization, tax regulations, changes in write-off legislation), and also cost increases or income losses due to civil unrest (e.g. causing increased security and repair costs or production disruption). The purpose of this paper is to discuss the various geological and economic risk factors in hydrocarbon exploration and their interrelationship. Knowledge of the relative importance of these risk factors is especially important not only for the planning of concession round work but also because such knowledge increases awareness of their importance. The discussions and examples used are based on Statoil's (1993) system for technical and economic prospect evaluation. The following discussions are restricted to cases with no political risk and to exploration in well established geological plays. Only oil prospects will be discussed. For a more general discussion of exploration economics, see Newendorp (1975), Megill (1985), Lerche (1992) and Steinmetz (1992).
Economic evaluation of prospects ~ general procedures Economic evaluation of a new hydrocarbon prospect is a stepwise process. First the prospect is identified by the exploration geologist/geophysicist and hydrocarbon volumes and prospect risk are calculated. The hydrocarbon volumes are given as a cumulative distribution. The reservoir engineers then forecast the reservoir performance and lifetime and the well demand to reach this production performance. These results are in turn used as a basis for considering production facilities and transport options for the products. The cost estimates for these development scenarios are combined with estimates of operational expenditures and exploration costs, taxes, discount rate and sales income to give a calculated net present value (NPV) of the prospect. The NPV is given both as an unrisked and as a risked (expected) value (ENPV). The ENPV includes the probability of a dry exploration well according to the equation: ENPV = p . N P V - (1 - p ) . exploration costs where p is the prospect chance. Here we follow White's (1993) definition of prospect chance as the probability of making a discovery of producible hydrocarbons in a well drilled on the prospect. The prospect chance is the complement of the prospect risk ~ a low-risk prospect has a high prospect chance. The value of the prospect chance p, which
gives an ENPV of zero, is termed the break-even probability (BEP). To find the ENPV that most correctly shows the potential of the prospect, different production mechanisms should often be evaluated. Such mechanisms include variations in secondary recovery processes, extended reach wells, varying number of wells and different transport solutions. Each of these solutions results in separate production and cost profiles. Searching for the best development plan for the prospect generally improves the calculated NPV. While this improvement is desirable for a given prospect, optimization of some (but not all) prospects in a concession round may lead to an unwarranted bias in favour of the optimized prospects. This warning does not, of course, apply to optimization of prospects which have been drilled and are being considered for development. The discussion of risk factors in the different stages of economic prospect evaluation is best performed by following a typical evaluation process. The following discussion will therefore begin with exploration risk at the prospect evaluation stage. This will be followed by a forecast of reservoir productivity and will conclude with the risk attached to the cost estimates of investments (capex) and operations (opex) and the effect of uncertainty in the oil price.
Volume and prospect risking The explorationist who has identified a prospect is asked to give a low (10%), modal and high (90%) estimate of the parameters which determine the hydrocarbon pore volume: prospect area, hydrocarbon column, reservoir thickness, net/gross ratio, porosity and hydrocarbon saturation (Table 1). These factors are combined with the shrinkage/expansion factor and the recovery factor to give the resource estimates for the prospect. It is required (by Statoil) that the explorationist provides written documentation for the choice of each parameter distribution. Such documentation can be based solely on statistical evidence (e.g. porosity distributions in the same formation of neighbouring wells at the same depth or depth-normalized), but in most cases it is influenced by the geological judgement of the explorationist. Errors may arise because of unforseen geological circumstances (e.g. a change in facies), inaccurate knowledge of the processes in operation (e.g. trapping capacity of a stratigraphic trap) or inaccurate calculations of fundamental variables such as depth (e.g. because of inaccurate depth conversion). One should note that the volume estimates illustrated in Table 1 have probably been performed by an explorationist who believes in his own seismic inter-
E v a l u a t i o n o f undrilled p r o s p e c t s ~ sensitivity to e c o n o m i c a n d g e o l o g i c a l f a c t o r s Table 1 A typical reserve estimation form Nation: Licence: Block: Structure: Chronostratigraphy: Lithostratigraphy:
Basin:
Depth to prospect (m): Water depth (m): Hydrocarbontype, main phase: GOR (volume/volume): Depth to spillpoint (m):
3010 85 OIL 26 3260
Low
Probable High
Estimation of reserves Reservoir thickness (m) Hydrocarbon column (m) Area of prospect (km 2) Rock volume, block (x 109 m 3) Rock volume, total (x 109 m 3) Net/gross Porosity Hydrocarbon saturation
80 200 2.300 0.160 0.160 0.800 0.140 0.720
90 200 2.600 0.240 0.240 0.820 0.150 0.735
90 250 7.300 0.350 0.350 0.840 0.160 0.750
Hydrocarbon pore volume (106 m 3) Within the block Main phase Total Main phase
17.9 17.9
22.6 22.6
27.6 27.6
Shrinkage/expansion factor
0.880
0.890
0.900
Hydrocarbons in place (oil: 106 Sm 3 ~ gas: 109 Sm 3) Within the block Main phase 16.0 20.1 Ass. phase 0.4 0.5 Total Main phase 16.0 20.1 Ass. phase 0.4 0.5 Recovery factor
Main phase
0.25
Reserves (oil: 106 Sm 3 m gas: 109 Sm 3) Within the block Main phase 4.7 Ass. phase 0.0 Total Main phase 4.7 Ass. phase 0.0
24.5 0.6 24.6 0.6
0.30
0.35
6.0 0.0 6.0 0.0
7.4 0.0 7.4 0.0
Probabilities (min. estimate or more): 1. Closure 2. Reservoir 3. Porosity 4. Source/migration 5. Timing 6. Trap 7. Recovery
0.9 0.8 1.0 0.5 1.0 0.6 1.0
Probability of discovery (prospect chance)
0.22
pretation. Other geologists may interpret the prospect differently, the outcome of which is significantly different views concerning risking and volumetric uncertainty. There is therefore an unseen spread of risk and parameter values depending on the range of interpretations possible which is very difficult to account for in exploration models. Seven independent chance factors are multiplied to give the prospect chance in Statoil's (1993) evaluation system. These factors describe the chance for: (seismic) closure, the existence of a reservoir rock, the existence of sufficient porosity in the reservoir rock, source/maturation/migration of hydrocarbons, timing (e.g. trap formation before hydrocarbon migration), trapping/leakage and recovery. As with the
327
volumetric parameters, the level of accuracy varies widely. Irwin et al. (1993) demonstrated that a cumulative distribution of migrating hydrocarbons could be given with considerable accuracy in a case where well control could be used to constrain the results from basin modelling of hydrocarbon charge. This modelling was based on an accuracy analysis of the input parameters, some of which could be given a likelihood distribution based on statistical verification. However, the risking of other factors (such as long distance migration in heavily faulted terrains) will be largely subjective. The subjectivity of several of the chance factors and volumetric parameters can never be completely removed, although efforts should be made to base assessments as far as possible on statistical evidence. The subjectivity involved makes post-drilling inspection of risk and volumetric assessments all the more important. However, such inspections can only substantiate whether the estimates were unbiased on average; they cannot be used to verify the risk or volumetric assessments of a new prospect within well-specified error limits. A proprietary investigation of the accuracy of risking and volumetric estimates performed in the mid and late 1980s gave the following main conclusions: (1) Volumes in undrilled prospects were generally overestimated (by 49% on average), as also noted by Rose (1987) but contrary to the finding of Uman et al. (1979). (2) Prospect risking was somewhat optimistic. (3) The relative risking of prospects was satisfactory in that the discovery rate was significantly higher in low risk than in high risk prospects. (4) The relative importance of the various risk factors was apparently correctly identified (Fig. 1). (5) Hydrocarbon phase prediction was almost perfect in those cases where the explorationist had a strong view as to whether the prospects would contain oil or gas. The identification of the "post drilling risks" of Fig. 1 was performed by monitoring the frequency of failure attributed to each of the chance factors (e.g., a factor of 0.97 for parameter 1 indicates that 3% of the drilled prospects failed because they were drilled off closure). This same procedure was followed by White (1993). Note that identification of the risk factor which caused the unsuccessful prospects to fail is uncertain and that the post-drilling results of Fig. 1 are not entirely accurate. The overoptimism revealed by this investigation was to a large degree due to the tremendous exploration success offshore Norway in the late 1970s and early 1980s, with discoveries of giant fields such as Troll, Gullfaks, Snorre and Oseberg. The sobering ex-
328
C. Hermanrud, K. Abrahamsen, J. Vollset, S. Nordahl and C. Jourdan
Fig. 1. Pre- and post drilling assessments of risk parameters for prospects on the Norwegian continental shelf. 1 =closure; 2 =reservoir rock existence; 3 =porosity; 4 =source, maturation and migration; 5 =timing; 6 =trapping/leakage; 7 =recovery. See text for further explanations.
perience following this period has led to a revision of risking and volumetric assessments which have been substantially improved and continue to be so.
Risking of reservoir complexity and performance Once a discovery has been made, additional drilling is required to determine the extent and complexity of the reservoir. Such information is needed to plan the development and production of the field. The number of wells that are likely to be needed before completion of a plan for development and operation (PDO) is estimated as a part of the concession round work. This estimate is based on pre-drilling perceptions of reservoir heterogeneities. The early forecasts of appraisal well requirements are based on highly inaccurate knowledge of the reservoir (which at this stage has yet to be proven). The seismic data base generally consists of 2D data only, and the structural/stratigraphic picture is invariably oversimplified. This may easily lead to the notion that the reservoir itself is simple, e.g. because faults which are not identifiable on seismic are often erroneously believed to be non-existent. Consequently, an underestimation of the number of appraisal wells should be expected during early stages of exploration. Internal changes in reporting routines, and the fact that relatively few of the late oil discoveries on the Norwegian Continental Shelf have been developed,
preclude a statistical analysis of estimates vs. the actual need for appraisal wells. However, a comparison of relatively recent pre-discovery estimates for appraisal needs vs. the number of actual appraisal wells drilled on producing fields supports the contention that the need for appraisal wells is often severely underestimated in the early stages of exploration, although decreasing field size can partly explain the same observation. The small size of present exploration targets necessitates keeping the number of appraisal wells low in order to ensure economic viability and exploration and appraisal wells may be converted for use as producers. The distinction between exploration and production wells may therefore be clouded in future. The future development of small fields will involve different development strategies than those adapted up until now. One alternative may be to develop a small field based on subsea templates connected to a host platform. The prospect may initially only contain one exploration well and one appraisal well, which can later be converted to a water injector, and be covered by one 3D seismic survey. Further segments of the reservoir can then be explored in a stepwise manner using sidetrack wells from producers and injectors once production from the first well has proven successful. Further success may warrant the installation of another template. The advantage of such a development scheme is that initial costs and thus economic risk is kept at a low level, with possibilities for expansion of the development after
Evaluation of undrilled prospects ~ sensitivity to economic and geological factors
329
Fig. 2. Annual resource/reserve estimates for the Norwegian Petroleum Directorate for producing oil fields through time, normalized to the time of production start-up.
removal of major risks attached to reservoir existence and performance. Extended well testing with small production ships may further reduce risk prior to development. Given a fixed hydrocarbon pore volume, early estimates of reservoir performance and recovery factor are also uncertain. Fig. 2 shows how the annual resource estimates of the Norwegian Petroleum Directorate have varied through time for Norwegian oil fields. The resource estimates of Figs. 2-4 are based on estimates of the most likely resources rather than reserves which are often reported in similar US studies. Most of the fluctuations immediately after production start-up are due to significant misconceptions concerning the reservoir geology which became evident during production and affected both oil in place and the recovery factor; the pre-production fluctuations mainly record changes in estimated oil in place. Even though the reserve estimates for several fields vary by several hundred percent, this is well within the accuracy ranges for such estimates documented by Rose (1987). Fig. 2 shows that in most cases the reserve estimates were upgraded between the first well and production start-up. This upgrading may historically have compensated for volumetric overestimation prior to drilling of first well, which in most cases led to a reduction in estimated resources relative to the predrilling expectancy. It is also clear from Fig. 2 that several fields had their reserve estimates reduced by
about 50% during that first years of production. Later adjustments have generally upgraded reserves, a feature that has also been documented in other parts of the world (Attanasi and Root, 1994). All of the fields in Fig. 2 which had their reserves significantly reduced after a short period of production belong to the Chalk play of the Central Graben area. These fields produce hydrocarbons from fractures in otherwise porous but very low permeability limestones. The changes illustrated in the figure resulted from poorer-than-expected performance of this reservoir. The later reserve upgrades in some of these fields result from pressure maintenance through water injection, which was not included in the original development plans. Fig. 3 shows similar information to that in Fig. 2 for the UK sector of the North Sea. The data were compiled by Corrigan (1993) from the "Brown Book" of the UK Department of Energy. This figure resembles Fig. 2 in that several of the fields had their reserve estimates reduced shortly after production start-up, with later reserve estimates close to constant or slowly rising. The fields which experienced a significant drop in estimated total reserves did so because of failure to recognize structural ~omplexities, compartmentalisation, and strong facies-related permeability contrasts. Most of these fields (Dunlin, Thistle, NW Hutton and Hutton) produce oil from the Middle Jurassic Brent Group. It is interesting to note that Brent fields in the Norwegian
330
C. Hermanrud, K. Abrahamsen, J. Vollset, S. Nordahl and C. Jourdan
Fig. 3. Annual resource/reserve estimates by the UK Department of Energy for producing fields through time. normalized to the time of production start-up.
sector of the North Sea do not show similar declines in reserve estimates. Here the reservoir quality is generally better as they are not always as structurally complex as their British counterparts and Norwegian workers have benefitted from the UK experience. However, as field sizes become smaller the effects of internal complexity will almost certainly be seen again. Note also that Corrigan (1993) did not attempt to provide representative coverage of the UK fields, but rather wanted to show some examples where the changes in reserves had been significant. A common feature of fields in the Norwegian and UK sectors is the general tendency to upgrade reserve estimates at medium to late stages in a producing field's life. This feature emerges clearly from Fig. 4, which contains the same fields as Fig. 2 but has been normalised to a time three years after production started. These improvements result from better reservoir descriptions, which led to more efficient depletion of the reservoirs, and improved well technology (such as horizontal drilling). The improved recovery thus comes at the cost (for example) of additional drilling and 3D seismic and does not necessarily yield a proportional increase in net income. Uncertainty in reservoir performance also influences the production profile vs. time. Overestimation of reservoir performance may imply that more
wells must be drilled to reach planned plateau production, which is then delayed or reduced. Such overestimations reduce the NPV both because of the postponed income and because of the excess drilling costs. Upgrades of reserves and/or reservoir performance after production start-up come too late to influence platform designs and, as a rule, lead to a less favourable development strategy than that which would have been chosen if the reserves had been accurately known in advance. The expected extra costs/ benefits arising from down- or upgrading reserves after production start cannot therefore be simply and accurately quantified on the basis of historical production data.
Risking of costs and sales income The economic calculations which form the basis for NPV assessments require input in the form of Table 2. All cost in this and subsequent tables are given in Norwegian kroner (NOK) unless otherwise specified (1 US dollar equals about 7 NOK). The different costs and incomes are given on a yearly basis, and a summation of these is combined with tax and internal interest (= discount) rates to produce unrisked NPV assessments. The numbers in Table 2 form the basis for a sensitivity analysis study and will be referred to later.
Evaluation of undrilled prospects m sensitivity to economic and geological factors
331
Fig. 4. Annual resource/reserve estimates by the Norwegian Petroleum Directorate for producing fields through time, normalized to three years after production start. Table 2 Base case input values for sensitivity analysis of prospect A Installations Producers/injectors Capacity GOR Tariffs: Year
Prospect risk 9%
Transport of oil Transport of gas Expl. costs
Appr. costs
Investments
M93NOK
M93NOK
Development M93NOK
Drilling M93NOK
Operations M93NOK
CO2 M93NOK
Sum
90
150
1.060
835
1057
1993 1994 1995
90
25 85
260 575
OPEX
Oil-production M Sm 3
Gas-production G Sm 3
875
9.9199
0.000
30 70 115 115
25 25
1.795 1.825
2001 2002 2003 2004 2005
115 112 100 100 100
25 25 140 161 182
1.799 1.224 0.895 0.656 0.483
2006 2007 2008 2009 2010
100 100
151 151
0.353 0.169
1.057 527 450
875 365 298
1996 1997 1998 1999 2000
0% discount rate 8% discount rate 10% discount rate
4O
90 83 82
150 128 123
6O 600 400
1.060 761 703
835 539 486
3:32
C. Hermanrud, K. Abrahamsen, J. Vollset, S. Nordahl and C. Jourdan
The different costs listed in Table 2 are all associated with some inaccuracy. The appraisal costs are most sensitive to the number of appraisal wells, but they are also sensitive to the costs of acquiring and processing 3D seismic data during appraisal and to the costs of drilling a well. Drilling costs are generally accurate to within 30%. The CO2 costs shown in Table 2 are environmentally motivated extra costs imposed by the Norwegian government. Their present value is just above 10 cents/1 (0.8 NOK/Sm 3) of burned gas ~ independent of whether the hydrocarbons are used for energy production onshore or offshore or are flared (the latter being generally disallowed in Norway). The cost estimates for operational expenditure (opex) and investments are not given within predetermined error limits. Later cost estimates (screening studies after a discovery has been made) are designed to give cost estimates within 40% of the true value. This requirement is refined to 20% at the PDO stage, and is further refined before the investments are actually made. The calculated incomes are uncertain both because of volume estimates and because of fluctuations in the oil price (Fig. 5).
Historical data concerning estimates of oil prices show that the uncertainty is generally underestimated and that there has been a tendency to extrapolate shortterm trends rather than to accept their oscillatory nature (Fig. 6). Table 3 shows the results of a poll among six Norwegian experts in 1981 as to what the oil price would be in 1985 and 1990; note that figures of around 100 USD/bbl are not uncommon. The opinions of these experts were quite accepted at the time, even though the oil price had just risen to a historical high (DeGoyler and MacNaughton, 1984; Yeager, 1985). Fluctuations in oil price make, to some extent, a comparison of cost estimates then with later actual costs incurred of limited value, as at that time almost any expenditure was warranted if it was believed to result in even a modest increase in oil production. Cost estimates for operating fields in licences which were allocated after the collapse of the oil price are too few to permit these data to be used as a basis for controlling present-day cost estimates. The reduction of operational expenditures on the Statfjord Field from 1991 onwards by more than 20% relative to the 1988 budgets confirms that significant cost reductions from previous estimates can sometimes be achieved.
Fig. 5. Fluctuations of the oil price through time, measured in 1993 Norwegian (NOK) and US (USD) currency. Slight differences in trend between the NOK values and the USD values are due to exchange rate fluctuations. The numbers for 1993 exclude November and December.
Evaluation o f undrilled prospects m sensitivity to economic and geological factors
333
Fig. 6. Expert estimates of the 1993 oil price, extrapolated from the forecasts for 1990 made by ten financial experts (Dagens Industri, 1981). The price forecasts have here been adjusted to account for the rate of inflation from 1981 to 1993, which has been less that forecasted by the experts in 1981. All prices have been recalculated to 1993 USD/bbl. Table 3 Anticipated 1985 and 1990 oil prices as suggested by ten international experts in 1981 (Dagens Industri, 198 l) Price forecasts in 1981 USD 1981
1985
Price forecasts in 1993 USD 1990
1993
intl. (%)
10
15
10
15
Constantine Fliakos, Vice President, Meril Lynch Pierce, Fenner and Smith Inc., New York
40
60
70
110
190
95
Gregory Shuttlesworth, Senior Energy Expert, J W Levy Consultant Corp., New York
40
65
80
120
180
104
Ted White, Director, Petroleum Economics Ltd., London
45-50
65-75
80-90
100-120
160-175
102
Odd Halvorsen Byr~sjef (head of department) at the Price and Market Office, DOE, Oslo
40
65
75
110
165
Oystein Noreng, Responsible for oil economy studies, Bedriftsokonimisk institutt, Oslo
40-50
75-95
90-110
110-210
165-310
139
Matthias Rapp, Director Scandinavian Trading, Stockholm
32-40
55-70
65-80
110-140
165-200
113
Herman Attinger, Director, Energy Centre, SRI International, London
40
60
80
115
180
101
William Randol, Vice President and Senior Petroleum analyst, Saleman Bros., New York
40
60
70
85
130
68
Michael Kelly, Senior Associate, Jensen Associates, Boston
40
70
80
140
200
124
Abdulaziz M. A1-Dukheil, President, the Saudi Centre for Finance and Investment, prev. Vice-Minister of Finance, Saudi Arabia
40
70
80
140
200
124
Average oil price USD/bbl
40
67
80
125
190
109
93
The forecasts for 1981, 1985 and 1990 are taken directly from Dagens Industri, 1981. Experts quoted with more than one number gave ranges rather than exact forecasts. The column to the far right was obtained by extrapolating the predicted trend from the period 1985-1990 to 1993, by using averages for the experts who gave ranges instead of single numbers in 1981, and by recalculating their forecast to what they would have been if the inflation from 1981 to 1993 had been known. The numbers in the far right column have been transformed to 1993 USD, the other numbers are those given in 1981.
C. Hermanrud, K. Abrahamsen, J. Vollset, S. Nordahl and C. Jourdan
334 Table 4
Results from the sensitivity analysis of prospect A at 8 and 10% discount rate
Base case, prospect chance = 8% prospect chance = 15% prospect chance = 4.5% Exploration costs reduced from 90 to 65 MNOK One added appraisal well, production postponed one year 20% reduction of investment and drilling costs 20% reduction of operational expenditures Tail production stopped in year 2004 Tail production extended in three years Annual production reduced by 20% Drilling costs increased by 400 MNOK in year 2000 Plateau production extended by three years Drilling costs reduced by 40% OPEX increased by 90% Investment increased by 38%
NPV @ 8%
NPV @ 10%
ENPV @ 8%
ENPV @ 10%
BEP @ 8%
290 290 290 290
200 200 200 210
10 40 -10 10
0 20 -10 10
6 6 13 4
240 370 330 250 290 100 210 680 350 130 130
160 280 240 170 200 30 130 530 270 60 40
10 10 10 10 10 -10 0 40 20 0 -10
0 10 0 0 0 -10 0 30 10 -10 -10
7 5 5 7 6 16 8 3 5 12 12
Sensitivity analysis of NPV assessments The numbers in Table 2 refer to the economic evaluation of prospect A. Different cost parameters were changed one at a time to investigate their relative importance on NPV, ENPV and BEE Table 4 displays the results of the sensitivity analysis. The values 8 and 10% refer to the discount rate which was used in the calculations. A flat oil price with time of 17 USD/bbl was used throughout unless otherwise specified. It appears from Table 4 that prospect A, which is rather small, is quite robust. A doubling of the plateau production period (from three to six years) has approximately the same effect on the ENPV as a doubling of the probability of discovery. A 20% reduction in the annual oil production has approximately the same effect as reducing the prospect's probability of discovery by 50%; it also has a more severe influence on the NPV than a 20% reduction of operational or investment costs. Table 5 shows the input parameters for a larger prospect (B) and Table 6 the results of the sensitivity analysis. The prospect risk was varied from 5 to 30% to investigate the sensitivity of this parameter. In addition to a modest reduction in CO2 costs and a reduction of opex by 20%, two combinations of parameters were set up to mimic possible changes in exploration and production strategy due to negative appraisal or production results. The combination with an NPV at 8% discount rate of 482 MNOK describes a situation where the initial appraisal drilling showed that the reservoir was more complex than previously believed. This complexity resulted in the need for three additional appraisal wells, as well as causing the production period to be postponed by two years. This combination of parameters reduced the NPV at
an 8% discount rate by MNOK 79 and increased the break-even probability from 8 to 10%. The parameter combination which gives an NPV at 8% discount rate of 182 KNOK was designed to accommodate serious production problems. An excessive extra drilling cost of 500 million NOK was added in each of the years 2004 and 2005, and the field reached plateau production two years behind schedule. This scenario reduced the NPV at 8% discount rate by 2/3 and increased the break-even probability from 8 to 25%. The introduction of more pessimistic numbers in the calculations caused the break-even probability (BEP) of a prospect to increase, thus indicating that a higher probability of discovery is needed to make the NPV positive. The importance of a given risk factor can thus be measured by its influence on the BEE To evaluate the relative importance of several risk factors, they were varied to find how large a change was needed to change the BEP of prospect A from 6 to 12%. The results of this investigation (Table 7) demonstrate clearly how prospect A is more sensitive to expenditure in the exploration than in the production stage, and also how (late) operational expenditure is less influential than earlier (investment) costs. The parameter variations applied to the sensitivity analysis are quite modest. Table 8 contains data for two real prospects (prospects C and D) where different estimates of volumes and prospect risk caused quite significant changes in NPVs. The time span between the first and last estimate in both cases is short (two years or less), and none of the revisions results from new well information. Note that the NPV is significantly different between these estimates and break-even probabilities vary from 7 to 66%. Models a and b for prospect D result from two
Evaluation of undrilled prospects ~ sensitivity to economic and geological factors
335
Table 5 Best case input values for sensitivity analysis of prospect B Installations Producers/injectors Capacity
Prospect risk 9%
GOR
Tariffs:
Transport of oil Transport of gas
Year
Expl. costs
Appr. costs
Investments
OPEX
Oil-production
Gas-production
M Sm 3
G Sm 3
Development
Drilling
Operations CO2
M92NOK
M92NOK
M92NOK
M92NOK
M92NOK
M92NOK
Sum
150
295
7.256
2.533
13.731
882
47.157
5.714
1992 1993 1994 1995
40 110
158 370 730 730 730
49 49 49
1.253 3.293 5.265
0.150 0.395 0.632
2006 2007 2O08 2009 2010
730 730 730 730 730
49 49 49 49 49
5.475 5.475 4.934 3.771 3.071
0.657 0.657 0.592 0.453 0.394
2011 2012 2013 2014 2015
730 730 730 730 730
49 49 49 49 49
2.690 2.272 2.069 1.750 1.462
0.329 0.297 0.248 0.210 0.175
2016 2017 2018 2019 2020
730 730 730 730 730
49 49 49 49 49
1.221 1.020 0.852 0.711 0.573
0.147 0.122 0.102 0.085 0.069
13.731 3.453 2.547
882 213 155
30
1996 1997 1998 1999 2000
95 90 20 60 1.074
2001 2002 2003 2004 2005
0% 8% 10%
1.873 2.289 1.508 512
150 122 116
295 203 185
7.256 3.428 2.870
depth conversions, where one was believed to be three times more likely than the other. The development of prospect D is envisaged with one exploration and one appraisal well. The cost of the appraisal well was removed from the calculations for a sensitivity test. The removal of this well caused the EPNV at 8% discount rate to double (from 10 to 20 MNOK) and the B EP to decrease from 16 to 10%.
Discussion and conclusions Economic calculations during prospect evaluation are hampered by large uncertainties. Both income and
63 363 1.075 1.095
2.533 1.064 866
expenditure may be uncertain by 50% or more, even when a discovery is made which is in reasonable accordance with expectations. The explorationist produces a likelihood distribution of hydrocarbon pore volumes and gives a corresponding prospect risk. This distribution, however, cannot be directly used for economic calculations. The different volumes in the distribution would require different development solutions, and a full analysis of the total economic consequences would require a large number of development scenarios to be made. This is impractical within the time constraint of a concession round. The use of historical
C. Hermanrud, K. Abrahamsen, J. Vollset, S. Nordahl and C. Jourdan
336
Table 6 Results from the sensitivity analysis of prospect B at 8 and 10% discount rate NPV @ 8%
Base case, prospect chance -- 8% prospect chance -- 5% prospect chance -- 15% prospect chance -- 30% Operational expenditures reduced by 20% CO2 costs reduced from 49 to 40 MNOK Three additional appraisal wells, production start postponed by two years Drilling costs increased by 500 MNOK in 2004 and 2005, production start unchanged, but plateau production start postponed by two years Operational expenditures reduced by 40% No CO2 costs
BEP @ 8%
0 -10 23 74 9 1
-16 -19 -7 11 -9 -15
8 8 8 8 6 8
482
160
-5
-19
10
-18 18 3
-31 -2 -14
25 5 7
-ll0
182 935 614
Equivalent sensitivity to variations of different factors for prospect A
Volume reduction Reduced plateau production Exploration wells: 2 wells + 3D Appraisal wells: 5 Investments increase 38% OPEX increase 90%
ENPV @ 10%
206 206 206 206 346 213
Table 7
Oil price
ENPV @ 8%
561 561 561 561 745 571
curves of development costs vs. reserve volumes (Snow et al., 1996) or forecasts of costs and incomes given as probability distributions (Otis and Schneidermann, 1994) may lead to alternative ways of performing an economic analysis. Finding the best way of representing volume distributions from prospect evaluation for economic calculations clearly remains a challenge.
Breakeven probability
NPV @ 10%
6 ~ 12% 17 --+ 14.50 USD/bbl 16%
1 year 230 MNOK 425 MNOK 720 MNOK 1738 MNOK
An increase in the prospect chance from 6 to 12% will compensate
for the various negative changes in parameter values displayed in the table.
490 246
It is difficult, if not impossible, to adequately characterize the total uncertainty of an early economic estimate. There has been an historical reluctance to collect data for verification of previous economic estimates, and rapid changes in technology, oil price, calculation procedures and reserve estimation techniques related to the period from prospect evaluation to field production restrict the usefulness of such information. Nevertheless, it appears clear that an analysis of past successes in economic and technical predictions aids future prospect evaluation. The ENPV of most fields (but not all, judging from previous experience) is sensitive to prospect risk. This fact is generally acknowledged by oil companies who put considerable effort into determining this parameter. Less awareness seems to exist about the quite marked influence on NPV of a suggested appraisal programme. The pre-drilling appraisal programme is highly uncertain and is usually over-optimistic during the early stages of exploration. The accuracy of early estimates of capex and opex (which are in the range of multi-million NOK)
Table 8 Different economical evaluations of prospects C and D through time Volumes M Sm 3
Prospect chance
28.7 12.1 9.2
23 9 9
7.6 7.6 3.6 3.1 5.3 5.3 5.3 8.5
13 13 20 6 19 19 19 19
BEP
NPV @ 8%
NPV @ 10%
ENPV @ 8%
ENPV @ 10%
2 3 6
780 400 290
500 290 200
170 20 10
100 10 0
16 15 22 66 16 10 12 7
130 160 70 10 110 190 140 260
50 60 20 -30 50 130 90 190
-5 0 0 -20 0 20 10 30
-15 -20 -10 -20 -10 10 0 20
Case C: Version 1 Version 2 Version 3
Case D: Version Version Version Version Version Version Version Version
1 l b (new development plan) 2 3a (complementary to 3b 3b
3b, sensitivity (minus appraisal well) 3b (new development plan) 4
Evaluation o f undrilled prospects ~ sensitivity to economic and geological factors
are usually no better than within 40%. Nevertheless, this uncertainty is not large enough to render early economic evaluations worthless. General experience with economic calculations for a large number of prospects has demonstrated that different prospects vary widely in their sensitivity to various factors. This is exemplified by the different sensitivities related to an extra appraisal well in prospects A and D. Both prospects are small and of marginal economical value. Yet another appraisal well in case A only increases the break-even probability from 6 to 7%, while the removal of an appraisal well in case D decreases the B EP from 16 to 10%. It has proved very difficult to anticipate which factors a given prospect will be sensitive to; this suggests that a separate sensitivity analysis should be performed for each prospect.
Acknowledgements We thank E. Siring for compiling and systemizing the historic data on reserve estimates, S. Heiberg for providing useful information on the early history of the Chalk fields and A.T. Buller for improving the English language of the paper. Lars Reistad and Terje Olsen are thanked for preparing the figures and tables, and Statoil is thanked for the permission to publish this material.
References Attanasi, E.D. and Root, D.H., 1994. The enigma of oil and gas field growth. Am. Assoc. Pet. Geol. Bull., 78(3): 321-332. Corrigan, A.E, 1993. Estimation of recoverable reserves: the geo-
C. HERMANRUD K. ABRAHAMSEN J. VOLLSET S. NORDAHL C. JOURDAN
Statoil, Statoil, Statoil, Statoil, Statoil,
Postuttak, 7004 Trondheim, Norway P.O. Box 300, 4001 Stavanger, Norway P.O. Box 300, 4001 Stavanger, Norway P.O. Box 300, 4001 Stavanger, Norway P.O. Box 300, 4001 Stavanger, Norway
337
logist's job. In: J.R. Parker (Editor), Petroleum Geology of Northwest Europe. Proceedings of the 4th Conference, Geological Society of London, pp. 1437-1481. Dagens Industri, 1981. Tio internationella experter om oljepriset 1985-90. In: the Swedish financial newspaper 'Dagens Industri', April 30, p. 17. DeGoyler, E.L. and MacNaughton, L.W., 1984. Twentieth century petroleum statistics, 1984. Dallas, Texas, p. 126. Irwin, H., Hermanrud, C., Carlsen, E.M., Nordvall, I. and Vollset, J., 1993. Computation of hydrocarbon charge in the Egersund Basin: pre- and post-drilling assessments. In: A.G. Dor6 et al. (Editors), Basin Modelling Advances and Applications. Elsevier, Amsterdam. Lerche, I., 1992. Exploration Economics. Pergamon Press, Oxford. Megill, R.E. (Editor), 1985. Economics and the explorer. Am. Assoc. Pet. Geol., Stud. Geol., Tulsa, Oklahoma, 19. Newendorp, ED., 1975. Decision Analysis for Petroleum Exploration. The Petroleum Publishing Company, Tulsa, Okla., 668 R Otis, R.M. and Schneidermann, N., 1994. A process for validation of exploration prospects. Am. Assoc. Pet. Geol., 1994 Annu. Conv. Off. Progr., 3: 228. Rose, ER., 1987. Dealing with risk and uncertainty in exploration: How can we improve? Am. Assoc. Pet. Geol. Bull., 71(1): 1-16. Snow, J.H., Dore, A.G. and Dorn-Lopez, D.W., 1996. Risk analysis and full-cycle probabilistic modelling of prospects: A prototype system developed for the Norwegian shelf. In: A.G. Dor6 and R. Sinding-Larsen (Editor), Quantification and Prediction of Hydrocarbon Resources. Norwegian Petroleum Society (NPF), Special Publication 6, Elsevier, Amsterdam, pp. 153-165 (this volume). Steinmetz, R., 1992. The Business of Petroleum Exploration. In: E.A. Beaumont and N.H. Foster (Editors), Treatise of Petroleum Geology. Handbook of Petroleum Geology, The American Association of Petroleum Geologists, Tulsa, Okla., 382 p. Uman, M.E, James, W.J. and Tomilson, H.R., 1979. Oil and gas in offshore tracts: Estimates before and after drilling. Science, 205(3): 489-491. White, D.A., 1993. Geologic risking guide for prospects and plays. Am. Assoc. Pet. Geol. Bull., 77(12): 2048-2061. Yeager, J.G., 1985. Significant trends in the downstream sector m the pricing of crude oil. In: R.E. Megill (Editor), Economics and the Explorer. Am. Assoc. Pet. Geol., Stud. Geol., 19: 63-70.
This Page Intentionally Left Blank
339
The World Bank's financial support to the petroleum sector in developing countries Eleodoro Mayorga-Alba and Scott Smith
This paper presents the World Bank Group's role in the petroleum sector of developing countries. It addresses separately the role of the International Bank for Reconstruction and Development (IBRD), the International Finance Corporation (IFC) and the Multilateral Investment Guarantee Agency (MIGA) in the upstream, downstream, and natural gas subsectors. Using specific examples, it describes the World Bank's role in promotion exercises, infrastructure projects, policy reform, mobilization of the private sector, and provision of political risk insurance. Considering that bank lending in the hydrocarbon sector meets only about 1% of the industry's capital requirements, the paper argues that the World Bank Group is best suited to use its unique resources to catalyze private sector investment and to provide an environment conducive to market-driven development.
Introduction The World Bank is a particular financial institution. Its operations encompass some characteristics of a commercial bank, some of a private enterprise, and some which are unique to itself. In general, the World Bank acts like a commercial bank in that it provides capital for investment projects. It acts like a firm in the way that it designs, implements, supervises and evaluates the projects it finances. But it provides far more scrutiny in the application of its Capital than any normal bank, and has fewer capital constraints than many firms. The World Bank behaves also as a consulting firm in the sense that it often provides policy advice to its clients. Indeed, structural and sectorial adjustment operations have become an increasingly prominent part of the Bank's activity, taking the place of the once typical infrastructure project. Despite these similarities with other organizations, what really makes the World Bank unique as a financial institution is its lending portfolio which is directed entirely at the developing world. As such, it has funded roads, bridges, irrigation systems, hydroelectric dams, telecommunication systems, policy reform, and education and health projects in Africa, Asia, Latin America, and now in Eastern Europe. The annual lending program of The World Bank has averaged US$22 billion over the last three years. Energy lending represents nearly 15% of the World Bank's portfolio, of which the power sector absorbs approximately two thirds. Hydrocarbons account for
the remaining third, or about US$1 billion. Compared with the capital requirements of the world's oil and gas industry, estimated at more than US$100 billion per year, the Bank's participation, at 1% of capital requirements, is not significant. Furthermore, the Bank's involvement in the problem sector has always been different from its involvement in other sectors of the economy. This is due mainly to the peculiarities of petroleum financing: the high risks and capital costs involved, the strategic importance of oil supplies to governments ~ both as a vital resource and as a tax base ~ and the unpredictable shifts in global prices. These and other volatile factors combine to make the economics of petroleum projects very different from most. As a result, the Bank's involvement has sometimes seemed hesitant. Through experience, however, it has found an important niche that it alone can fill by virtue of its constant dialogue with governments in the developing world and its importance factors into account, the Bank's role in the petroleum sector can be characterized by the emphasis on institutional and policy reforms to create an environment that attracts private capital. This paper will look in greater detail at the Bank's role in the petroleum sector of the developing world, differentiating between upstream, downstream and natural gas subsectors, and examining not only the role of the IBRD, but also of IFC and MIGA. 1 I The World Bank was created along with the International Monetary Fund as a result of the Bretton Woods Conference in 1944. At
Quantification and Prediction of Petroleum Resources edited by A.G. Dor6 and R. Sinding-Larsen. NPF Special Publication 6, pp. 339-344, Elsevier, Amsterdam. 9 Norwegian Petroleum Society (NPF), 1996.
340
The International Bank for Reconstruction and Development (IBRD) The Oil and Gas Division Before broaching the role that the Bank plays in the petroleum sector, it would be appropriate to explain the way it operates in general. Six regional Vice Presidencies direct the operational work of the Bank. Within each Vice Presidency, country departments are responsible for one or a group of countries, and, within each department, divisions are in charge of operations in specific sectors. The Bank's energy lending is done through the Industry and Energy or the Infrastructure Divisions. The Bank calls these divisions Sector Operation Divisions (SODs). Responsibility for lending to the oil and gas sectors lies with the SODs that, in turn, call on the Bank's Oil and Gas Division (IENOG) when carrying out complex operational activities. IENOG is a division of the Industry and Energy Department, subdivided thematically into groups for petroleum, gas, and economics and restructuring. IENOG concentrates the Bank expertise in the sector and acts as the in-house consulting arm, providing technical advice to the SODs on a variety of oil and gas projects around the world.
The upstream subsector Prior to 1973, oil was cheap, seemed plentiful, and was easily produced by the developing world. The Bank consequently directed most of its energy sector lending to the power subsector. But between 1970 and 1977, the price of oil rose over six-fold, creating serious balance of payments problems for oil importing developing countries (OIDCs), leading the Bank to question how it might assist these countries in mitigating such adverse effects. One of the answers came in the form of Petroleum Exploration Production Promotions (PEPPs). These promotions were born from the realization that many countries had prospective areas which were not being explored, primarily from a lack of geological knowledge and an inadequate institutional framework for investment. The PEPPs had, in essence, three goals" -compile and improve the available geological and geophysical data to support the exploration of potential petroleum reserves; that time it included the International Bank for Reconstruction and Development (IBRD) and the International Centre for Settlement of Investment Disputes (ICSID). The International Finance Corporation (IFC) was founded in 1956, and the Multilateral Investment Guarantee Agency (MIGA), the most recent of the Bank Group institution, was created in 1988.
E. Mayorga-Alba and S. Smith
-assist developing countries to establish the legal framework and policies to attract private investment in hydrocarbon exploration and production; -procure and/or develop the skills needed by the governments to negotiate agreements and manage and supervise their hydrocarbon subsectors. The program was started in 1980 and has been carried out in nearly 50 countries. Just over US$300 million was spent to finance PEPPs (compared to over $6 billion for downstream lending in the petroleum sector). Over half of the PEPPs took place in Africa, with mixed results. This can be partly explained by the fact that the PEPPs were designed for small countries with little or no production history, where the emphasis was on the verification and promotion of the resource base. The program was substantially reduced in the late 1980s, when the price of oil collapsed. The Bank's entry in the oil business, as a possible major source for financing Exploration and Production (E&P) Projects, spawned a controversy. The International Oil Companies (IOCs) opposed the Bank's involvement in the sector on the grounds that it was unnecessary and would perhaps have a negative effect on normal market behavior. They argued that where the geological conditions were promising and where the political risks were manageable, IOCs would undertake the required exploration efforts. The Bank, conscious of these fears, has been careful never to displace private risk capital from the exploration activities and has focused its involvement on the development phase of upstream projects. The IFC has adopted this approach after financing some exploration efforts. One of the main successes of the PEPPs was in fostering the dialogue between the IOCs and governments (Oduolowu, 1992). Fears of Bank interference were replaced by the realization that the Bank's presence ensured fairer and more efficient E&P contract terms. Companies also found the Bank instrumental in helping them form joint ventures with national oil companies (NOCs). On the government's side, there was an appreciation that the Bank increased the level of competition for their acreage and provided an opportunity for exposing their countries to IOCs. Both the successes and the failures provided important lessons about the necessary and sufficient policy reforms. Above all, where they were most successful, the PEPPs were tied to a comprehensive national energy strategy which provided assistance for expanding the infrastructure and encouraged the creation of an investment-friendly policy environment. Recognizing that sedimentary basins often straddle several countries, the Bank went beyond the PEPP operations by promoting sub-regional geological stud-
The Worm Bank's financial support to the petroleum sector in developing countries
ies. One was carried out in the sub-Andean basin of Latin America with Canadian funding. In Africa, the Red Sea/Gulf of Aden project was designed to enlist the cooperation of each of the countries on this subregion (Egypt, Sudan, Ethiopia, Djibouti, Somalia, the Republic of Yemen and Saudi Arabia) in promoting the entire sedimentary basin instead of a single country's share. Though the project suffered from a pull-out of one of the cofinanciers, using this comprehensive, multinational approach, it did succeed in attracting at least twelve active exploration contracts where before there were none.
The downstream subsector Most of the World Bank's petroleum portfolio goes to the downstream sector. Downstream lending is either in the form of project loans to build infrastructure or policy loans to reduce the weight of the public sector and to encourage efficiency gains through private sector competition in procurement, refining, and distribution activities. The need for sector reform becomes clear when one looks at the preponderant importance of petroleum products to the balance of payments of the OIDCs, or to the fiscal revenues in any other developing country. Since oil imports must be paid in foreign exchange, and since foreign exchange is often dangerously scarce, oil imports represent a significant macroeconomic burden. Inappropriate policies hinder oil producing countries as well. Often, in these countries, petroleum product prices are set on a cost-plus basis, or even subsidized below cost, leaving no capital for reinvestment and vastly distorting resource allocation. Africa is probably the continent which most dramatically illustrates these problems. In most of subSaharan Africa (SSA) the supply of petroleum products is handled by state-owned monopolies, which were created for political as much as economic reasons at the time of the countries' independence. Heavy government influence in markets combined with such factors as the lack of foreign exchange for purchasing petroleum products and the use of suboptimal practices in procurement, refining and distribution, have, over time, provoked an unnecessarily large increase in the cost of supplies. In addition, landlocked countries suffer from the monopolistic exploitation of transport corridors by coastal countries. According to a recent Bank-commissioned study (Cuneo e Associati/World Bank, 1993), these factors together push the cost of petroleum products to about $400 per ton ($235 for procurement, $57 for refining and $110 for distribution). The study suggests that this price could feasibly be reduced by almost $50
341
per ton, saving SSA the equivalent of $1.4 billion per year. Of this, nearly 65% could be realized through changes in operating procedures, institutional set-up, and refinery rationalization, without any investment; the remainder would require investments in infrastructure rehabilitation. The required policy changes are all in the direction of opening markets, encouraging private sector participation, making pricing policies more marketsensitive, and shifting the government's role from a participant in the sector to a regulator of the sector. The World Bank has made such policy reform an integral part of its lending policy for most SSA countries. Of the total lending portfolio, approximately 15% of commitments goes specifically to sector adjustment, through Structural Adjustment Loans. These loans, in all sectors, are designed to support sectoral programs of policy and institutional change, including restructuring of capacity, mobilization of resources, and efficiency in resource allocation. Policy-based loans in the energy sector have followed the rising trend; it has been estimated that approximately one third of future energy lending may be of this type. The most direct way to involve the private sector is, however, to privatize State Owned Enterprises (SOEs). The Bank is starting to play an important role in privatization endeavors in the petroleum sector. In essence, it is providing the following types of support: (1) analysis of the sector (pinpointing which enterprises to privatize and how); (2) drawing up of the necessary regulatory frameworks; (3) providing specific loans to mitigate short-term adverse macroeconomic effects; and (4) providing credibility to the sales process by being involved in marketing trips and trade shows. One of the most comprehensive sector privatization efforts has been taking place in Argentina, where the Bank has assisted the process with both financial and technical support. The national oil company, YPF, held a major stake in every facet of Argentina's hydrocarbon activity, from exploration, development and processing through transportation and distribution. The first step in privatizing was to convert YPF into a public, limited liability corporation and to allow all oil producers to freely sell their production. This was done in 1991. By early 1993, half of YPF assets, over $200 million worth of refining facilities and pipelines, shipping fleet and port facilities, storage center, retail outlets, in-house drilling and exploration services, and laboratories had been sold. In addition, Gas del Estado was split up and privatized. The next step will be the sale of up to 50% of the YPF shares still held by the government. In parallel, the sector has been deregulated and liberalized. Private operators can explore areas of acknowledged poten-
342 tial and develop discoveries on a concession basis, or build and operate gas lines. Restrictions on prices and exports have been completely eliminated. In short, the World Bank has assisted in creating efficient operations, increasing investments, work opportunities (despite initial layoffs), and fiscal revenues, all within a liberalized macroeconomy. The Bank has been especially vigorous in promoting downstream projects in Eastern Europe and Central Asia. In this region, there are 22 projects worth a total of $4,495 million (a staggering 57% of the total sector lending) "in the pipeline". Currently, hydrocarbon projects are planned in Azerbaijan, Bulgaria, Kazakhstan, Romania, Russia, the Slovak Republic, Turkey, Turkmenistan, the Ukraine, and Uzbekistan. Most of these projects, for both gas and oil, attempt to rehabilitate existing infrastructure (for both production and distribution), effect regulatory and pricing reforms, and provide technical assistance to improve practices and efficiency. There is a great hop that the energy sector of these countries, if rescued soon enough, can provide a more than adequate base for developing these economies.
The natural gas subsector The objectives of the Bank's lending in this subsector are related to planning and regulation, institutional framework, pricing, rationalization of gas utilization, and expansion of private sector investments. While the Bank's natural gas projects and adjustment loans have had success ratios significantly higher than the overall average for Bank operations, there has been only mixed success in certain important regions (Khelil et al., 1991; Mor et al., 1993). In particular, difficulties remain in diverting natural gas to high-value consumption (power and industry), in encouraging rational producer pricing (gas is mostly underpriced and hence discourages foreign investment), and in taxation and regulatory policies (taxation and pricing are often not treated independently, resulting effectively in huge taxes on windfall profits, which also discourages foreign investment). Finally, consumer prices remain inefficiently low, though this has been somewhat mitigated recently by a decline in world gas prices. The development of natural gas often entails comprehensive programs involving significant investment in infrastructure coupled with sector adjustment and policy reform. A recent study by the Bank in Indonesia deemed the reserves and market size sufficient to recommend that the country embark upon such a comprehensive program. The estimated total investment required for field development and transmission and distribution networks in Indonesia is about $6
E. Mayorga-Alba and S. Smith
billion. Essentially, adequate pricing and institutional changes are necessary to provide incentives to operators to develop the proven and potential reserves. This type of study is typical of the Bank's advisory role. Similar support is being done in other countries with large gas reserves in Africa (i.e. Nigeria, Cameroon) and in Latin America (i.e. Colombia, Bolivia, Trinidad and Tobago).
Conservation and environment issues Along with promoting energy development, the Bank has been active since its inception in encouraging energy efficiency. Beginning in the 1980s, however, as environmental consciousness rose to the fore, the Bank began to promote new kinds of projects with physical components and policy initiatives that specifically targeted energy efficiency and conservation and economically justified fuel-switching. This emphasis notwithstanding, energy efficiency in the developing world has remained low. It has been estimated that with the existing capital stock 20 to 25% of energy consumed could already be saved through greater efficiency, much of it on the supply side. However, with additional investments in energy efficient capital equipment, these savings could rise to between 30 and 60% of consumption. Even though these results indicate that there is a substantial opening for efficiency improvement, a review of the Bank's policy efforts in this regard show them to be generally sound; if energy conservation is not all that it should be, the Bank is not entirely to blame. Nonetheless, it has been recognized within the Bank that there is scope for better integrating energy conservation issues into the Bank's country policy dialogue. In particular: (1) energy prices should reflect real costs which include environmental concerns; (2) markets should be competitive, allowing suppliers to compete, encouraging efficiency; (3) supply-side restructuring should be encouraged; (4) a more rapid transfer of competitive energy-efficient technology should be effected.
The International Finance Corporation and the Multilateral Investment Guarantee Agency Recognizing the private sector's value as an engine of development, the World Bank Group includes two organizations devoted to enlisting private sector support in development projects. The International Finance Corporation (IFC) was founded in 1956. Its mandate was (and remains) to assist in the economic and social growth of its developing member countries through the development
The World Bank's financial support to the petroleum sector in developing countries
of the private sector. With an equity capital of $1.8 billion, it is the largest multilateral source of direct financial assistance to the private sector in developing countries. Its financing is directed solely at the private sector. Unlike virtually all comparable organizations (including the IBRD), its articles specifically forbid it from accepting government guarantees, thus subjecting it to the discipline of the market. Due to the unique risks of petroleum projects in developing countries, their high capital costs, and the high degree of technical sophistication required, it is no surprise that the IFC has played a fairly substantial role in the petroleum sector. Over the past ten years, IFC has participated in financing oil and gas projects with cumulative total project costs of about $3 billion. Of this, IFCs participation totalled $500 million, reflecting its catalytic role in attracting capital, most often through the syndication of loans. IFC has been particularly valuable in facilitating non-recourse financing for projects. In non-recourse financing, the private sector seeks to minimize the amount of capital it places at risk, and shift the burden of risk disproportionately to local partners or multilateral agencies. In essence, IFC agrees to shoulder the portion of risk which the company deems excessive. IFC is able to do this because of its unique mandate and because it evaluated country risk differently from commercial banks. Its sister agency, the Multilateral Investment Guarantee Agency (MIGA), was founded in 1988 and is the most recent member of the World Bank Group. Though it too is devoted exclusively to the private sector, it differs from IFC in that it offers no financing and instead provides insurance to private companies investing in the developing world. The concept of political risk insurance is not new; more than twenty countries created bilateral investment insurance programs to help their national investors before the creation of MIGA. However, these national programs were not overly successful due to particular eligibility criteria and sometimes restrictive mandates. MIGA was created to provide uniform protection, regardless of the investor's nationality. It was expected to provide coverage of political risk more broadly and effectively than any of the national agencies. Some 115 countries have signed its charter, and in the last three fiscal years, the cumulative number of MIGA contracts has risen from four to thirty-six. MIGA provides long-term coverage against a number or risks, including currency transfer, war, breach of contract, and civil disturbance. Most of its guarantees are for fifteen years, but they have been for as little as three and as many as thirty. A facet of MIGA that is particularly useful to the high risks of petroleum projects is its flexibility under
343
its Convention to write other forms of political risk coverage on a case by case basis to suit an investor's need. When one considers other available financing mechanisms, be they from private banks, IFC, or local lenders, this flexibility allowed by MIGA can be the vital factor that allows a project financing package to be assembled.
Conclusions The unique set of risks presented by investment in developing countries and investment in petroleum projects calls for unique instruments for projects which combine both. Economic development is little more than the transfer of human energy to technological energy, so the Bank's involvement in the energy sector, and particularly in petroleum, is at the heart of its development agenda. The World Bank increasingly recognizes that among the most important of its challenges is the establishment of a proper investment environment to allow for economic growth by unleashed private sector market-driven forces. The World Bank, IFC and MIGA each focus on different areas, each working synergistically to channel investment towards the developing world. The Bank, while continuing to finance projects, increasingly insists on sector reform. IFC assists foreign direct investment and also participates in the policy dialogue. MIGA does not directly fund projects, but guarantees projects funded by the private sector. The Bank has had an impact in opening doors to investment and development in the petroleum sector of the developing world that would otherwise surely have remained shut. In the future, we hope to see an expansion in the use of natural gas, greater exploration and exploitation of endogenous petroleum resources, and greater efficiency in the downstream petroleum sector. We would like to help achieve these goals with the cooperation of national governments and, wherever possible, with the capital and expertise of the private sector. The Bank has at its disposal both capital and a large, varied, and appropriate knowledge base. Where it has worked best it has worked as a catalyst, using its resources as a lever to set larger forces into motion, or a spark to ignite a more powerful engine: the engine of development.
References Cuneo e Associati/World Bank, 1993. Regional Study on the Rationalization of Petroleum Products Supply and Distribution in Sub-Saharan Africa. Executive Summary, 52 pp. Khelil, C., Gutierrez, L. and Joyce, T., 1991. World Bank Strategy for the Natural Gas Sector in LAC. World Bank, Latin America and the Caribbean Technical Department, Regional Studies, Report No. 1, 85 pp.
E. Mayorga-Alba and S. Smith
344 Mor, A., Bond, J. and Mayorga-Alba, E., 1993. Natural Gas in Sub-Saharan Africa: An Overview. World Bank, Africa Technical Department, AFTPS Division Note No. 1, 37 pp. Oduolowu, A., 1992. An Evaluation of World Bank Funded
E. MAYORGA-ALBA S. SMITH
Petroleum Exploration Promotion Programs, 1980-1990. World Bank Industry and Energy Department Working Paper, Energy Series Paper No. 59, 30 pp.
The World Bank, 1818 H Street N~, Washington, DC 20433, USA The World Bank, 1818 H Street NW, Washington, DC 20433, USA
345
References index AAPG Explorer, 40 Abraham, K.S., 19 Abrahamsen, K., 151 Abreu, V.S., 128, 132 Aburish, S.K., 17, 18, 19 Ackleson, S.G., 48, 55 Adams, A.J., 207, 218 Akanni, E, 132 Akinosho, T., 131,132 All6gre, C.J., 47, 55 Amos, J.E, 48, 55 Appi, C.J., 133 Aquino, G.S. de, 131,133 Araujo, M.S., 128, 133 Arps, J.K., 95, 103 Ashton, B.R., 19, 61, 62, 133 Assayag, M.I., 128, 133 Attanasi, E.D., 20, 329, 337 Atwater, G.I., 201, 217 Augustson, J.H., 219, 235, 237, 238, 251 Austvik, O.G., 291,312 Baak, A.B., 183-185, 198 Balay, H., 322, 324 Balay, R.H., 98, 103 Balcer, Y., 109, 114 Bally, A.W., 201,217 Barber, S.A., 133 Barrocas, S.L.S., 133 Barry, R.A., 19 Barton, C.C., 20, 44, 50, 55 Bee, A.C., 19 Behrenbruch, P., 299, 312 Bekker, H., 63, 70 Berli, K., 63, 64, 70 Bessis, E, 225, 235 Bharucha-Reid, A.T., 214, 217 Bird, K.J., 218 Birkeland, ~., 121,122 Bjerksund, P., 142, 151 Black, F., 141,142, 151 Blatt, H., 48, 55 Blitzer, et al., 136 Blystad, P., 106, 114 Bococcoli, G., 127, 133 Boeuf, M.A.G., 133 Bogoslovskaya, G.N., 120, 121,122 Bond, J., 342, 344 Bowen, J.M., 167, 181 Bowman, M.B.J., 167, 168, 181 Boye, K., 295, 312 BP Review of World Gas, 40 BP Statistical Review of World Energy, 3, 19, 40 Bradey, R., 295, 312 Braga, J.A.E., 133 Braun, R.L., 243, 251
Brekke, H., 106, 114, 320, 322, 324 Brewster, J., 273, 289 British Petroleum Company, 19, 43, 51, 55 Brooks, J.M., 48, 55 Browne, E.J.P., 19 Bruhn, C.H.L., 125, 126, 128, 133 Bruun Christensen, 0., 121,122 Buchanan, R., 304, 312 Buckee, J.W., 202, 214, 218 Bujalov, N.I., 115, 122 Burnham, A.K., 241-243, 251 Burrus, J., 225, 235 Cainelli, C., 125, 126, 131,133 Caldwell, R.H., 19, 63, 64, 70, 71, 76 Campbell, C.J., 3, 7, 19, 43, 51, 55 Campbell, G.R., 19, 61, 62, 133 Campos, O.A.J., 127, 133 Candido, A., 128, 133 Cao, S., 200, 201,203, 213, 217 Capen, E.C., 133, 219, 233, 235 Carlsen, E.M., 219, 235, 246, 251,327, 337 Carmalt, S.W., 4, 19 Carroll III, J.A., 306, 312 Central Intelligence Agency, 33, 35, 40 Chang, H.K., 133 Charpentier, R.R., 20, 44, 50, 55 Chen, Z., 109, 110, 114, 152 Ch6net, P.Y., 225, 235 Cheredeev, S.I., 121,122 Chow, S., 109, 114 Christie, A.A., 237, 251 Clarke, R.H., 48, 55 Cleveland, C.J., 19 Cleverly, R.W., 48, 55 Cliff, W.J., 133 Clifford, A.C., 131,133 Collyer, D.R., 19, 61, 62, 133 Conceicao, J., 127, 133 Conybeare, C.E.B., 201,218 Cora, C.A.G., 128, 133 Corrigan, A.E, 153, 165, 329, 330, 337 County Natwest Woodmac, 162, 165 Cox, J.C., 143, 151 Cronquist, Ch., 77, 81, 153, 165 Crovelli, R.A., 95, 98, 103, 109, 114, 322, 324 Cuneo e Associati/World Bank, 341,344 Dagens Industri, 333,337 Dahl, B., 151,237-239, 241,242, 245, 251 Dake, L.P., 303 Damsleth, E., 151,308, 309, 312 Davies, P., 19 Davison, I., 125, 133 De Leebeck, A., 281,289 De Ruiter, H., 292, 312
346 Deegan, C.E., 168, 181 Degens, E.T., 47, 55 DeGoyler, E.L., 332, 337 Dekker, H., 77, 81 Demaison, G., 4, 6, 19, 20 DeSorcy, G.J., 2, 19, 61, 62, 63, 70, 77, 81,133 Dewey, J.F., 133 Dodd, T., 173, 181 Doligez, B., 225, 235 Dolton, G.L., 91,103 Dor6, A.G., 137, 152, 168, 181, 219, 235, 237, 251,336, 337 Dorn-Lopez, D.W., 137, 152, 336, 337 Doust, H., 133 Drury, J., 19, 61, 62, 133 Duckworth, R., 48, 55 Duff, B.A., 151 Duppenbecker, S., 173, 181 Dutta, N.C., 204, 218 Dyrhaug, L.T., 292, 312 Edwards, J.D., 126, 133 Edwards, R., 306, 312 Eggen, S., 225, 235, 246, 251 Ehrenberg, S.N., 106, 114 Ekern, S., 142, 151 Eleventh World Petroleum Congress, 58, 62 Emery, D., 51, 55, 173, 181 Emrich, L.J., 88, 90 Energy Economics, 19 Energy Exploration and Exploitation, 40 Energy Information Administration, 6, 19 Energy Information Administration, USA, 293, 312 Energy Statistics Source Book, 40 Enron Corporation, 25, 30, 40 Erofeeva, N.S., 115, 122 Espitali6, J., 240, 251 Ewen, D., 171,181 Exploration for Coal, Gas and Oil in Onshore and Offshore China, 40 Fagerland, N., 106, 114 F~erseth, R.B., 106, 114, 120, 122 Federovsky, Y.E, 121,122 Feller, W., 214, 218 Fernandes, G.J.F., 131,133 Figueira, J.C.A., 125, 126, 133 Figueiredo, A.M.E de, 125-127, 133 Flach, T.A., 306, 312 Flett, T.O., 207, 218 Folinsbee, R., 5, 19 Franca, L.C., 128, 133 Franke, M.R., 133 Fraser, A.J., 168, 181 Freire, W., 125, 133 Fuller, J.G.C., 19 Gabrielsen, R.H., 120, 122 Gall, B.L., 259, 260 Gamboa, L.A.P., 126, 133 Gehman, H.M., 154, 156, 165, 184, 198 Gerhardt, J.H., 308, 312 Geyssant, J.R., 235 Giozza, W.E, 128, 133 Giroir, G., 225, 235 Gluyas, J., 167, 168, 173, 181 Gomelkova, N.P., 119, 122 Gordinsky, E.V., 119, 122 Gorsuch, D.P., 58, 62 Grace, J.D., 2, 19, 63, 64, 70, 71, 76, 109, 114 Grant, S.M., 137, 151, 173, 181
References index
Greer, C., 18, 19 Griffiths, C.M., 246, 251 Grishin, EA., 119, 122 Guardado, L.R., 126, 133 Guinasso, N.L., Jr., 48, 55 Gutierrez, L., 342, 344 Hage, A., 308, 312 Halbouty, M.T., 19 Haldorsen, H.H., 308, 309, 312 Hall, D., 151 Harbaugh, J.W., 47, 48, 55, 83, 90 Hastings, D.S., 106, 114 Haun, J.D., 3, 19 Hawken, P., 7, 18, 19 Heather, D.I., 19, 63, 64, 70, 71, 76 Herbin, J.P., 235 Hermanrud, C., 151,219, 225, 235, 237, 246, 251,327, 337 Heum, O.R., 106, 107, 114, 225, 235 Hinderaker, L., 308, 312 Hoagland, J., 18, 19 Holtz, M.H., 126, 133 Hombroek, J.A.R., 133 Home, R.N., 306, 312 Hubbert, M.K., 12, 19, 44, 55 Huc, A.Y., 226, 235 Huizinga, B.J., 19 Hunt, J.M., 47, 55, 202, 218 Ignatenko, E.A., 121,122 Internal studies and documentation, 76 International Energy Agency, 13, 18, 19 Interstate Oil and Gas Compact Commission, 256, 257, 260 Ion, D.C., 63, 70, 77, 81 Irwin, H., 219, 235, 327, 337 Irwing, H., 246, 251 Ismail, I.A.H., 19 Ittekot, V., 47, 55 Ivanhoe, L.E, 6, 19, 20 Jackson, J.S., 106, 114 Jacobsen, T., 246, 251 Jacoby, H.D., 142, 151 James, W.J., 327, 337 Jarvis, M.G., 133 Javoy, M., 47, 55 Jeangeot, G., 273, 289 Jensen, L.N., 120, 122 Johansen, S.E., 121,122 Jones, R.W., 201,218 Jourdan, C., 151 Joyce, T., 342, 344 Kalheim, J.E., 106, 107, 114, 120, 122, 320, 322, 324 Karner, G.D., 133 Kaufman, G.M., 95, 103, 109, 114, 138, 151 Keegan, W., 40 Keith, D.R., 58, 62 Khelil, C., 342, 344 Klemme, H.D., 20, 40, 45, 55 Knott, D., 133 Knudsen, K.R., 63, 64, 70 Kobrin, 136 Koch, J.O., 106, 107, 114 Kokesh, E, 201,218 Kondakov, A.V., 116, 122 Kontorovich, A.E., 115, 122 Koutsoukos, A.H.M., 125, 126, 133 Kowsmann, R.O., 133
347
References index
Krokstad, W., 151 Krumbein, W.C., 186, 198 Kruyt, D., 109, 114 Kulibakina, I.B., 120, 121,122 Kvalheim, O.M., 237, 251 Kvenvolden, K.A., 47, 48, 52, 55 Kydland, T., 304, 305, 312 Lafargue, E, 225, 235 Laherrere, J.H., 4, 6, 20 Lana, M., 127, 131,133 Lang, R.V., 19, 61, 62, 133 Larsen, B.T., 106, 114 Laughton, D.G., 142, 151 Lawrence, D.A., 167, 168, 181 Leckie, G.G., 6, 20 Lee, P.J., 107, 109, 114 Leith, L.T., 246, 251 Lerche, I., 27, 40, 200, 201, 201,203, 205, 210, 213, 217, 218, 225, 235, 326, 337 Liu, J., 201,217 Llave, EM., 259, 260 Lobo, A., 127, 133 Lohrenz, J., 298, 312 Loomis, E., 233 Los Angeles Times, 6, 20 Lucchesi, C.E, 126, 133 Lumley, J.L., 214, 218 Lynch, M.B., 207, 218 MacDonald, I.R., 48, 55 Macedo, J.M., 130, 133 Macedo, R.A.V. de, 133 Macgregor, D.S., 46, 48, 49, 55 MacNaughton, L.W., 332, 337 Magoon, L.B., 218 Mamedov, Y.G., 255, 260 Manne, A.S., 292, 312 Margulis, L.S., 121,122 Maritvold, R., 283, 286, 289 Marke, P., 304, 312 Marshall, D.L., 306, 312 Martin, A.J., 20 Martin, G.C., 207, 218 Martfnez, A.R., 63, 70, 77, 81 Martins, C.C., 125-127, 133 Martirosjan, V.N., 121,122 Mason, S.P., 142, 152 Masters, C.D., 6, 13, 20 Matos, R.M.D. de, 125, 126, 133 Maxwell, J.R., 133 Mayorga-Alba, E., 342, 344 McCarthy, C.M., 207, 218 McCrossan, R.C., 201,218 McDowell, A.N., 201,218 McFarlane, R.C., 201,218 McHaffie, E.R., 133 McKelvey, V.E., 63, 70, 71, 76 Megill, R.E., 219, 235, 326, 337 Meisingset, I., 151 M61i6res, F., 235 Mello, M.R., 125, 126, 133 Menely, R.A., 111,114 Merton, R., 141, 142, 152 Middleton, G., 48, 55 Miertschin, J.W., 307, 312 Milani, E., 127, 133 Miller, B.M., 201,218 Miller, R.G., 20, 43-50, 55
Milton, N.J., 137, 151, 168, 171,181 Minerals Management Service, 40 Mitchener, B.C., 167, 168, 181 Modelevsky, M.S., 119, 122 Mohriak, W.U., 125, 126, 133 Molle, L. Jr., 128, 133 Mr P., 226, 235 Monaghan, P.H., 47, 48, 50, 56 Mor, A., 342, 344 Moraes, M.A.S., 133 Morales, R.G., 127, 133 Morbey, S.J., 152 Moritis, G., 258, 260 Mueller, T.D., 201,218 Muller, C., 235 Murray, R.C., 48, 55 Myers, 312 National Petroleum Council, 258, 260 Nederlof, M.H., 83, 90, 183, 184, 198 Nehring, R., 20 Newendorp, P.D., 153, 154, 165, 201,218, 326, 337 Nijhuis, H.J., 183-185, 198 Nordahl, S., 151 Nordvall, I., 219, 235, 246, 251,327, 337 NPD (Norwegian Petroleum Directorate), 78, 79, 81, 81, 91, 103, 106, 107, 114, 314, 318-320, 323, 324 Odell, ER., 43, 48, 55 Oduolowu, A., 340, 344 Oien, A., 310, 312 Oil and Gas Journal, 5-7, 20, 40 Omatsola, E., 133 Ormaasen, E., 63, 64, 70 Osanik, A., 47, 48, 50, 56 Ostisty, B.K., 121,122 Otis, R.M., 336, 337 Ovreberg, O., 308, 312 Oxtoby, N.H., 51, 55, 173, 181 Paddock, J.L., 142, 144, 152 Palciauskas, V.V., 243, 251 Palhares, C.A.C.A. Jr., 133 Pallesen, S., 246, 251 Partington, M.A., 167, 168, 181 Pautz, J.E, 260 Pautz, J.E et al., 257 Penn, I.E., 235 Pereira, M.J., 130, 133 Peres, W.E., 128, 133 Perrodon, A., 4, 6, 20, 131,133 Petroconsultants S.A., 20 Petzet, G.A., 133 Phipps, S.C., 10, 20 Pickles, E., 142-145, 147, 152 Piedmonte, M.R., 88, 90 Pineau, E, 47, 55 Pitcher, M.G., 201,218 Polster, L.A., 119, 122 Popper, K.R., 186, 198 Porter, E.D., 40 Potential Gas Agency Report, 40 Price, L.C., 47, 48, 50, 56 Procter, R.M., 111,114, 201,218 Ramos, A.L., 133 Rangel, H.D., 133 Reinders, J.E.A., 225, 235 Rennemo, S., 310, 312
348 Rian, D.T., 306, 312 Richardson, F.H., 40 Riis, E, 120, 122 Ritter, U., 246, 251 Riva, J.P. Jr., 133 Roadifer, R.E., 4, 20, 201,218 Roberts, T.G., 95, 103 Robertson, W.D., 19, 61, 62, 133 Robinson, A.G., 173, 181 Robinson, J.G., 19, 61, 62, 133 Rogers, M.A., 47, 48, 50, 56 Root, D.H., 20, 329, 337 Rose, P.R., 137, 152, 219, 235, 327, 329, 337 Ross, S.A., 143, 151 Roy, K.J., 201,218 Rubenstein, M., 143, 151 Ruijtenberg, P.A., 304, 312 Saleri, N.G., 299, 312 Sangsnit, 136 Santa Cruz, C.E. de, 133 Santogrossi, P.A., 126, 133 Santos, C.F., 133 Sassen, R., 48, 55 Savini, R.R., 128, 132 Scarton, J.C., 128, 133 Schneidermann, N., 336, 337 Scholes, M., 141, 142, 151 Schou, L., 246, 251 Schrattenholzer, L., 292, 312 Schroeder, EW., 220, 235 Schuler, G.H.M., 17, 20 Scull, B.J., 168, 181 Securities and Exchange Commission, 77, 81 Seffy, D.H., 207, 218 Shaw, N.D., 226, 235 Sherwood, K.W., 207, 218 Siegel, D.R., 142, 144, 152 Sinding-Larsen, R., 109, 110, 114, 152 Skjerv~y, A., 221,225, 235, 237, 251 Skosgeid, J., 106, 114 Sloss, L.L., 186, 198 Smalley, P.C., 51, 55, 173, 181 Smith, J.L., 142, 142-145, 147, 152 Smith, M.B., 201,218 Smith, P.J., 202, 214, 218 Smith, S., 63, 70, 77, 81 Snow, J.H., 137, 152, 336, 337 Snowdon, L.R., 107, 114 Society of Petroleum Engineers, 58, 60, 62 Sollie, B.H., 109, 114 Solomon, C., 20 SCreide, I., 306, 312 Souz Cruz, C.E., 128, 133 Souza, J.M., 128, 133 SPEE, 63, 70 St. John, B., 4, 19 Stainforth, J.G., 225, 235 Steinmetz, R., 219, 235, 326, 337 Stewart, C., 295, 312 Stewart, D.J., 219, 235, 237, 251 Stoian, E., 201, 218 Sylta, 0., 151,219-221,225, 226, 235, 237, 251 Szatmari, P., 127, 133
References index
Takin, M., 20 Tang, J., 201,217 Taylor, G.C., 111,114 Teisserenc, P., 131,133 Tham, K.K., 259, 260 The Economist, 16, 19 The Royal Ministry of Finance, 315, 316, 324 The Royal Ministry of Industry and Energy, 78, 81, 81, 315, 321, 324
The Royal Ministry of Oil and Energy, 81 Thomas, B.M., 226, 235 Thompson, M., 137, 151 Time Magazine, 41 Tissot, B.P., 47, 55, 202, 218, 240, 242, 243,251 Tomatite, T., 20 Tomilson, H.R., 327, 337 TCrudbakken, B., 106, 114 Townes, H.L., 3, 20 Trigeoris, L., 142, 152 Turner, R.E, 207, 218 Tutt, D.W., 19, 61, 62, 133 U.K. Dept. of Energy Brown Book, 41 Ulmishek, G.F., 13, 20, 40, 44, 45, 50, 55 Uman, M.F., 327, 337 Ungerer, P., 225, 235 USGS-MMS, 41 Van der Ven, P.H., 131,133 Vasilyev, V.G., 115, 122 Villemin, J., 131,133 Vinnikovsky, S.A., 115, 122 Volden, R., 308, 312 Vollset, J., 151, 168, 181,219, 235, 246, 251,327, 337 Wade, K.C., 306, 312 Walstrom, J.E., 201,218 Wang, EC.C., 107, 109, 114 Waples, D.W., 242, 251 Warne, G.A., 19, 61, 62, 133 Warren, J.E., 201,214, 218 Weeks, L.G., 117, 122, 201,218 Weiser, A., 307, 312 Welte, D.H., 47, 55, 202, 218, 237, 239, 242, 243, 251 Whitaker, M.F., 226, 235 White, D.A., 96, 103, 154, 156, 165, 169, 184, 198, 326, 327, 337 Wilson, D.C., 58, 62 Wilson, R.D., 47, 48, 50, 56 Withers, R.J., 214, 218 Wolff, B., 133 Woods, T.J., 41 World Oil, 11, 20 World Petroleum Trends, 41 Wyllie, M.J.R., 201, 218 Yeager, J.G., 332, 337 Yergin, D., 20 Yudin, S.G., 119, 122 Yukler, M.A., 201,218, 237-239, 242, 245,251 Zakharov, E.V., 115-121,122 Zhdanov, M.A., 119, 122
349
Subject index Africa, 341 aggregation, 135 Albacora Field, 128 Alberta tar sands, 258 Anadarko Basin, 140 anticlinal traps, 118 appraisal drilling, 160, 184, 198, 268, 269, 277, 281,334 appraisal wells, 322 Argentina, 341 artificial lift, 305 Asia, 23, 25, 30, 119, 322, 339, 342 asset management, 309 Audrey gas field, 261 Balder Formation, 273, 279 Barents Sea, 79, 93, 97, 115, 120, 122, 321,323, 324 Barracuda Field, 128 basement heat flux, 201 basin analysis, 186 basin modelling, 136, 167, 173, 186, 193, 199, 201-203, 213, 219, 237-240, 242, 244, 246, 249, 327 benchmarking, 261, 271 Brazil, 123 break-even probability, 326, 334 break-even reserves, 163 Brent Group, 167 Business Units, 309 Campos Basin, 123, 125-132 CAPEX, 292, 295, 299, 311 capital expenditure, 162 cash flow, 269 Central Graben, 246, 329 Chalk, 329 chance factors, 84, 327 charge risk, 173 charge, 194 chemical EOR, 254 China, 6, 8, 13, 27-30, 39, 255 Chukchi Sea basin, Alaska, 201 CIS, 22-25, 28-30, 39, 54, 115, 255 closure, 194, 327 CO2 costs, 332 Coal Bed Methane, 26 Colombia, 5 Common Risk Segment (CRS), 167, 169, 171,172, 174-179 comparison coefficients, 156 concession rounds, 325 conditional probability, 90 conservation, 342 construction and development expenditure, 161 conventional oil, 52 cost, time and resource analysis, 265 covariance, 108 covariation, 84, 135 cross-validation, 105
cumulative probabilistic method, 199 cumulative production, 1 cycle compression, 269 cycle time, 261,262, 266 deceleration, 54 decision trees, 142, 153, 296-298 deep water technology, 125, 132 definitions of resources, 58 Della gas field, 261 Delphic approach, 183, 201 dependency, 84 deterministic approaches, 244 deterministic calculations, 281 developed reserves, 293 developing countries, 352 development cost, 73, 136, 141,261,323, 336 development drilling, 279 development option, 143, 145 discount rate, 135, 136, 142, 163, 291,293, 296, 310, 326, 334-336 discoverability parameter, 109 discovered recoverable resources, 313 discovered resources, 79 discoveries, 79 discovery modelling, 109 discovery pattern, 3, 4 discovery probability, 296 distribution pattern, 3 Draugen Field, 317 Draupne Formation, 220, 241 drillable prospects, 100 drilling success ratio, 178 Dunlin Field, 329 Dunlin Groups, 220 Eastern Europe and Central Asia, 342 Eastern Fedinsky Nose, 121 economic analysis, 162, 336 economic evaluations, 323, 326 economic input parameters, 323 Egersund Basin, 237, 246 Ekofisk Field, 317 Enhanced Oil Recovery (EOR), 253-259, 299, 311 Canada, 255 m chemical, 254 gas miscible, 253 Indonesia, 255 production trends, 255 United States, 255 n Venezuela, 255 environment, 342 Eocene, 279 exchange rate, 73, 291,293, 299, 332 existing infrastructure, use of, 309 expectation curve, 184, 185, 194 expected resources, 319
Subject index
350 expert judgement, 137 exploration cost, 39, 131, 144-147, 163, 323, 324, 326, 334 exploration cycle, 197 exploration expenditure, 159 exploration options, 145 exploratory drilling success, 123 expulsion efficiencies, 220 expulsion models, 225 expulsion, 242 FASPUM, 91, 98-101,322 fault-sealed traps, 118 field development plans, 297 field size distribution, 111, 113 field size, 111 fields, names of, 81 financial portfolio theory, 139 finding cost, 24, 30, 37, 127, 137 fluid pressure, 208 flux rate, 43 frequency distribution, 157 Frigg Field, 261,273, 275, 277, 279-281,283, 285-289, 317 Frigg Formation, 273 frontier areas, 293 gas injection, 254, 256, 258, 299, 300, 302 gas price, 32, 136, 163, 323, 342 gas (see natural gas) gas-miscible EOR, 253 geohistory model, 199 geological mapping software, 237 geostatistical techniques, 72 Germany's production, 10 gross rock volume, 157 Gullfaks Field, 327 half-life, 45 Halten Terrace, 106, 322 Haltenbanken, 317 HC saturation, 100 heat flux, 201 heat-flow histories, 24 Heather Formation, 220 heavy oil, 45, 254 Heidrun Field, 317 Heimdal field, 287 Hewett gas field, 261 historical success rate, 156 horizontal wells, 258, 259, 304, 305, 308, 309 Hutton Field, 262, 329 hydrate (clathrate), 51 hydraulic fracturing, 305 hydrocarbon charge, 219, 327 hydrocarbon generation, 237, 242 hydrocarbon migration (see also migration), 94 Hydrocarbon Pore Volume (HCPV), 79, 80, 233, 311,326, 327, 329, 335 hydrocarbon saturation, 196 hydrocarbon volume calculation, 241 hydrocarbon yield, 219 hydrocarbons in-place, 300 Improved Oil Recovery (IOR), 299, 319 integration, VI International Bank for Reconstruction and Development (IBRD), 340 International Finance Corporation, 342 International Oil Companies (IOCs), 340 interpretation confidence maps, 178
IRAP Petroleum Mapping System (PMS), 240 Iraq, 6 Irreducible Exploration Risk (IER), 185 Japan, 25, 26, 30-32, 38, 39, 53 kerogen transformations, 242 kerogen, 245 kinetics, 242 Kuwait, 6 Latin America, 3, 5, 23, 24, 339, 341,342 life cycle of oil and gas fields, 164, 253, 261, 263, 265, 267-269, 271,273, 275, 277, 279, 281,283, 285, 287, 289 likelihood distribution, 335 Lloydminster play, 137 log-normal, 84, 111,205, 215, 216, 233 log-normal distribution, 84, 110, 233, 323 Loppa High, 120, 122 Ludlov Saddle, 121 Ludlovskaya, 121 Marlim Field, 128 maturation, 327 Maureen oil field, 261 mean, standard deviation, 86 mean-variance analysis, 140 Mercurius High, 120, 122 Mexico, 6 mid-East, 22-25, 28-31, 38, 40, 23 Middle Jurassic Brent Group, 329 Middle Jurassic, 322 Midgard Field, 317 migration, 48, 220, 237, 242, 327 Moira gas field, 261 Monte Carlo simulation, 61, 63, 64, 68, 72, 79, 87, 89, 153-155, 157, 162-164, 184, 196, 197, 199, 201-203, 213, 219, 237, 240, 244, 247-250 Mere Basin, 321 multi-national oil companies, 126, 131, 126 Multilateral Investment Guarantee Agency, 342, 343 National Oil Companies (NOCs), 123, 126, 131,132, 340, 341 natural gas, 1, 21, 25, 31, 125, 131,253, 255, 256, 259, 339, 342, 343 Navarin Basin, 203 Net Present Value (NPV), 64, 65, 135, 136, 139, 141,142, 153-155, 162-164, 261, 269, 294, 296, 299-304, 306, 307, 311, 312, 325, 326, 330, 334, 336 non-systematic or diversifiable risk, 135 normal (Gaussian) distribution, 84 Norne Field, 317 North America, 21, 23-25, 27, 38, 39, 253 North Sea, 7, 51, 54, 58, 78, 79, 85, 93, 96, 97, 99, 106, 116, 137, 143, 167, 168, 171, 173, 233, 237, 239, 245, 246, 261, 263, 265, 267, 269, 271, 273, 293, 297, 298, 302-304, 317, 319-321,329, 330 North Viking Graben, 167, 241 northern Kildin Nose, 120 Norway, 1,313 Norwegian Aid for Developing Countries (NORAD), 322 Norwegian Continental Shelf, 317, 325, 328 Norwegian economy, 313, 315 Norwegian government, 325 Norwegian licensing procedure, 106 Norwegian North Sea, 245, 320 Norwegian Parliament (Stortinget), 321 Norwegian Petroleum Directorate, 77, 329
Subject index Norwegian Sea, 79, 93, 97, 321 Norwegian Shelf, 160, 323 Norwegian State Oil Company (Statoil), 315 NW Hutton Field, 329 offtake strategy, 307 oil disk, 273 oil half-life, 46 oil price forecasts, 291 oil price risk, 292 oil prices, 1, 7, 13, 17, 18, 63, 77, 123, 131,136, 253, 254, 257-260, 291-293, 309, 316, 332, 333, 336 oil reserves, 45 oil resources remaining, 1 m reserves, 1 total discovered, 1 m ultimate, 1 undiscovered, 1 oil window, 210 OPEC, 5, 6, 13, 32, 37, 38, 58, 123 operating flexibility, 142 OPEX, 292, 295, 299, 302, 307, 312, 331,332, 334-336 optimization, 326 Original Gas In Place (OGIP), 277 Oseberg Field, 237, 244, 327 palaeoheat flux, 201 Paleocene, 279 passive margin basins, 123 passive margin, 128 permeability barrier, 281 Petrobras, 123, 125-132 Plan for Development and Operation (PDO), 80, 328 plateau rate, 73, 161, 291,297, 300-302, 304, 307 play, 94, 168, 186 play analysis, 81, 91, 94, 101, 102, 106, 111, 167, 183, 186, 189, 320, 322, 323 play chance, 101 play fairway, 167, 169, 171,173, 175, 177-179, 181 play model, 322 play potential, 113 play process domains, 188 play risk, 137, 167 play summary maps, 96 political risk, 326 porosity, 99, 101, 102, 107, 108, 117, 119, 120, 157, 167, 171, 173, 184-186, 188, 189, 192, 196, 197, 199-201, 203-210, 214, 217, 220, 233, 240-243, 281,283, 284, 312, 326-328 Potential Gas Agency, 26 pre-development costs, 297, 298 price volatility, 142 probabilistic approaches, 244 probabilistic calculations, 281 probabilistic methods, 27, 63, 64, 199, 214 probabilistic modelling, 153 probability distribution, 46, 59, 63, 64, 83, 84, 87, 90, 91, 94, 98100, 136, 142, 153, 163, 184, 199, 202, 204, 213, 215, 217, 220, 224-226, 229, 231-234, 244, 247, 336 probability maps, 226 probability of discovery, 87, 321 probability, 205 probable resources, 319 processing capacities, 302 product prices, 162 production forecasting, 160 production profiles, 316 production well, 74, 254, 279, 281,284, 286, 289, 306, 328 Production Sharing Agreement (PSA), 295
351 prospects, 63 prospect appraisal, 244 prospect chance, 101,326 prospect evaluation, 294, 321 prospect mapping, 325 prospect risk, 327, 334, 336 prospect-specific risk, 167 prospective resources, 83, 293 proved and probable reserves, 2 proved reserves, 22, 59 proved resources, 319 pseudowells, 238, 242 PVT data, 100 Rate Of Return (ROR), 32, 139-142, 163, 292, 293 recoverable reserves, 184 relative permeability, 308 remote observation wells, 283 reported reserves, 6 reserve estimates, 329 reserves, 71 deterministic, 59 probabilistic, 59 reservoir porosity, 99 reservoir rock, 94 reservoir, 194, 327 resources assessment, 196 resources classification, 76, 78, 115, 319 resources in place, 79 resources, 45, 71 RFT (Repeat Formation Tester), 173, 281,283, 284, 289, 297, 309, 312 risk analysis, 91, 101-103, 135, 137, 139, 141, 143, 145, 147, 149, 151, 153, 155, 157, 159, 161, 163, 165, 199, 200, 203, 213, 261,263, 267, 268 risk factors, 326 risk reduction, 184 risk tranches, 191 risk-return tradeoffs, 139 risk of reservoir effectiveness, 171 risk of reservoir presence, 171 risked NPV, 296, 299 risking, 325, 326, 328 Rock-Eval, 243 Russia, 115 Santos Basin, 130, 131 Saudi Arabia, 6, 13, 17, 18, 29, 333, 341 seal risk, 175 seal, 194 secondary hydrocarbon migration, 219 secondary recovery, 253 Securities and Exchange Commission (SEC), 58 seepage, 43-46, 48-50 segment, 67 3D seismic, 72, 125, 195, 197, 265, 268, 281, 283, 289, 295, 297, 309, 322, 328, 330, 332 data, 308 programme, 265 seismic monitoring, 286 sensitivity analysis, 72, 334, 337 sensitivity, 27, 72, 86, 106, 110, 113, 137, 163, 164, 199, 200, 213, 259, 283, 296, 298, 299, 301, 302, 306, 308, 311, 323-325, 327, 329-331,333-337 Sergipe-Alagoas Basin, 131 Seven Sisters, 32 shrinkage factor, 49 small fields, 328 SmCrbukk S0r Field, 317
Subject index
352 Snchvit Field, 317 Snorre Field, 327 Society of Petroleum Engineers (SPE), 58 Sogn Graben, 224 source rock distribution, 226 source rock yield, 225 source rocks, 47 standard deviation, 322 Statfjord Field, 7, 317 statistical models, 201 steam injection, 258 stratigraphic traps, 118 stripper wells, 256 submarine fan, 273 sunk costs, 297 surfactant flooding, 254 surfactant, 254, 305 suspect reserves, 5 swing producers, 1 systematic risk, 135
153, 159, 164, 165, 167, 178, 183-185, 194, 195, 199-207, 209-213, 215-217, 219, 220, 225, 226, 229, 233, 237, 244, 246-248, 258, 259, 268, 273, 277, 283, 284, 286, 288, 291, 298, 300, 303, 308-310, 313, 317, 319, 321-327, 330, 332, 335-337 unconventional resources, 26 undeveloped commercial reserves, 293 undiscovered resources, 78, 83, 313, 320, 321,324 undrilled prospects, 327 United States Geological Survey (USGS), 27, 30, 71, 77, 91, 98, 106, 322 United States, 17 Valhall Field, 317 Veslefrikk Field, 223 Viking Group, 220 viscous fingering, 259 Visund Field, 223 vitrinite reflectance, 210, 242 volumetric calculations, 58 volumetric modelling, 107 Vcring Basin, 321
tariff, 73, 163, 310, 323, 331,335 technical resources, 293 testable hydrocarbons, 154-157, 163 thematic exploration analysis, 125 thermal EOR, 254 thermal history model, 199 thermal maturity, 210 Thistle Field, 329 trap fill, 99, 159 trap formation, 327 trap sizes, 233 trap types, 100 traps, 118 Troll Field, 327 turbidite plays, 132 turbidite reservoirs, 125
Water Alternating Gas (WAG), 74 water encroachment, 277 water flood projects, 74, 254 water flood, 305 water injection, 263,299, 300, 302, 304, 305, 309, 329 weighting, 221 well-head price, 30 West Malinginskaya Saddle, 121,122 Western Europe, 23, 24, 29, 30, 38 world consumption, 25 world midpoint of depletion, 13 world production, 16, 22, 23 World Bank, 351 World Petroleum Congress (WPC), 58, 63
uncertainty, 2, 26, 27, 29, 32, 37, 46, 55, 57-59, 61, 65, 67, 68, 7177, 79, 83-86, 98, 102, 112, 113, 135-137, 139-142, 146;
yet-to-find oil, 7, 45 yet-to-produce oil, 7
