Cover:Example_Design 23/09/2009 10:38 Page 1 An assessment of the evidence for a near-term peak in global oil production Global Oil Depletion

An assessment of the evidence for a near-term peak in global oil production UK Energy Research Centre September 2009 September Centre Research UK Energy UK ENERGY RESEARCH CENTRE

58 Prince’s Gate Exhibition Road London SW7 2PG tel: +44 (0)20 7594 1574 email: [email protected] www.ukerc.ac.uk

UK ENERGY RESEARCH CENTRE Cover:Example_Design 23/09/2009 10:38 Page 2

An assessment of the evidence for a near-term peak in global oil production

A report produced by the Technology and Policy Assessment function of the UK Energy Research Centre

Steve Sorrell Jamie Speirs Roger Bentley Adam Brandt Richard Miller

August 2009

ISBN number 1-903144-0-35 Prelims_Intro:UKERC 23/09/2009 10:31 Page i Page 10:31 23/09/2009 Prelims_Intro:UKERC n n n n n n n Technology (ICEPT). The contributors were: Univ The assessment w n n n n n n are containedinseven in-depth The Synthesis Report presentsthemainfindingsofthisassessment.Moredetailedresults 1 http://www.ukerc.ac.uk/support/tiki-index.php?page=TPA%20Overview sector. Theassessmentaddressesthefollowingquestion: which is comprised of independent experts from government, academia and the private energy sectorstakeholders andupontherecommendationofTPA AdvisoryGroup, global oildepletion.Thesubjectofthisassessmentwas chosenafterconsultationwith This reportsummarisesthemainconclusionsfromTPA’s assessmentofevidencefor explaining results in a way that is useful to policymakers. provide authoritative reportsthatsethighstandardsforrigourandtransparency, while field throughcomprehensive assessmentsofthecurrentstateknowledge.Itaimsto Assessment (TPA) function.TheTPA was setuptoaddresskey controversies intheenergy This reporthasbeenproducedbytheUKEnergyResearch Centre’s Technology andPolicy Preface the UKERCwebsite Technical Report5 Simon Wheeler, Independent Consultant ( Godfrey Boyle, Director, EERU, TheOpenUniversity ( Roger Bentley, Department of Cybernetics, University of Reading ( Richard Miller, IndependentConsultant( Adam Brandt University of California, Berkeley ( Erica Thompson,Grantham Institute,ImperialCollege( Technical Report7 Technical Report6 Technical Report4 Technical Report3 Technical Report2 Technical Report1: ersity of Sussex and Jamie Speirs of the Imperial College Centre for Energy Policy and What evidence isthere to support the proposition the that global supply of ‘ co nventional oil’ will beconstrain as led by the Steve Sorrell of the Sussex Energy Group (SEG) at the 1 : : Methods of estimating ultimately recov : Comparison of global supply forecasts : Methods of forecasting future oil supply : Decline rates and depletion r : Nature and importance of reserve growth : Definition and interpretation of reserv Data sources and issues Technical Reports ed by phys ed by Technical Reports4 Technical Report 7 Technical Report6 which area ates ical depletion Technical Report7 Technical Reports2 e estimates er v ) and able resources ailable todownloadfrom before 2030? 7) Technical Report 7 ) ) and 3) ) i Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Prelims_Intro:UKERC 23/09/2009 10:31 Page ii Page 10:31 23/09/2009 Prelims_Intro:UKERC

ii Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Prelims_Intro:UKERC 23/09/2009 10:31 Page iii Page 10:31 23/09/2009 Prelims_Intro:UKERC analysis. The comparison of forecasts in Fabiana Gordon(ImperialCollegeStatisticalAdvisoryService)assistedwiththestatistical Jean Laherrèreprovided usefulcommentsonanearlydraft of publication ofdatafromtheirPEPSdatabase. House) andDavid Strahan. We arealsovery grateful toIHSEnergyforallowingthe Skrebowski (Peak Oil Consulting), Michael Smith (Energyfiles), Paul Stevens (Chatham Expert Group, namelyKen Chew(IHSEnergy),JohnMitchell(ChathamHouse),Chris Energy Programme). Useful commentsandadvicehave alsobeenreceived fromour Kaufmann (University ofPennsylvania) andLuciavan Geuns(ClingendaelInternational We arevery grateful forthecommentsreceived fromourtwopeerreviewers,Robert Acknowledgements energy. UKERCisfunded bytheUKResearch Councils. and use while developing and maintaining the means to enable cohesive research in undertakes world-class research addressing the whole-systems aspects of energy supply and source of authoritative information and leadership on sustainable energy systems. It The UKEnergyResearch Centre’s missionistobetheUK'spre-eminentcentreofresearch About UKERC none are responsible for the content of this report. The above individualsrepresentarange ofviewsonthe risksofglobaloildepletionand editing. (UKERC TPA) for assistance with drafting and Phil Heptonstall (UKERC TPA) for copy Director) and Rob Gross (Head of UKERC TPA function). Thanks also to Philip Greenacre We are grateful for the guidance and patience provided by Jim Skea (UKERC Research and JeanLaherrère. Hilmar Rempel (BGR),JamesSmith(Shell),Ken Chew(IHSEnergy),RichardHardman 7 Garry Brennand (OPEC) and David Freedman (Shell E&P). The authors of Supply), Colin Campbell (ASPO), John Staub (US Energy Information Administration), Consulting), LeifMagneMeling(StatoilHydro),LaurentMaurel(Total Exploration and Schindler andWerner Zittel(LudwigBölkow Systemtechnik), ChrisSkrebowski(Peak Oil without the model descriptions and data provided by Michael Smith (Energyfiles), Jörg would alsolike tothankFatih Birol(IEA),NimatAbuAl-Soof andcolleagues(OPEC), Technical Report 7 would not have been possible Technical Report5 Technical Report while iii Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Prelims_Intro:UKERC 23/09/2009 10:31 Page iv Page 10:31 23/09/2009 Prelims_Intro:UKERC

iv Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Prelims_Intro:UKERC 23/09/2009 10:31 Page v Page 10:31 23/09/2009 Prelims_Intro:UKERC depletion, supplemented by data analysis and guided by an Expert Group. A total of sev The report is based upon a thorough review of the current state of knowledge on oil This report addresses the following question: Overview sufficient to meet global demand well into the 21 required. Incontrast, othercommentatorsarguethatliquidfuelsproductionwillbe dislocation, with alternative sources being unable to ‘fill the gap’ on the timescale depletion oftheresource.Many believe thatthiscouldleadtosubstantialeconomic subsequent terminal decline in the production of conventional oil as a result of the physical oil. Butagrowingnumberofcommentatorsareforecastingnear-term peakand economies andatpresentthevast majorityofthesefuelsareobtainedfrom‘conventional’ Abundant supplies of cheap liquid fuels form the foundation of modern industrial Executive Summary to signaloildepletioninasufficientlytimelyfashion. addition, there are uncertainties over the extent to which the market may be relied upon rate atwhichdemandreductionandalternative sourcesofsupplymay berequired.In the rate atwhichproductionmay beexpectedtodeclinefollowing thepeakandhence conventional oilproductionhasbeenafocusofdebate,whatappearsequallyimportant is appropriate investments aretobemade.Whilethetimingofafuturepeak(orplateau) in economic, environmental andsecurityimplicationswhichneedtobeanticipatedifthe economy. Inaddition,thetransition away from conventional oilwillhave important decline intheproductionofconventional oilcouldhave amajorimpactontheglobal Without sufficientinvestment indemandreductionandsubstitutesourcesofenergy, a cheap oiliscomingtoanend. of many upstreaminvestment projects.Thereisagrowingconsensusthattheageof Agency (IEA)iswarning ofanear-term ‘supplycrunch’owing tothecancellationanddelay has broughtoilpricesdownfromtheirrecordhighofJuly2008,theInternationalEnergy analysts has increased concerns about oil security. While the global economic recession rising prices, declining production in key regions and ominous warnings from market analysts remain sceptical. But beginning in 2003, a combination of strong demand growth, depletion, several oilcompanieshave beenpubliclydismissive andthemajorityofenergy influence onenergyandclimatepolicy. Mostgovernments exhibitlittleconcernaboutoil Despite muchpopularattention,thegrowingdebateon‘peakoil’hashadrelatively little whether therelevant organisationswillhave theincentive andabilitytoinvest. how depletion can be mitigated by investment and new technology. A concern for both is depletion willhave adominant influence onfutureoilsupply, whilethelatteremphasise of ‘non-conventional’ resources such as oil sands. The first group claims that physical exploration anddiscovery, theenhancedrecovery ofconventional oilandthedevelopment What evidence isthere to support the proposition the that global supply of ‘c o nventional oil’ will beconstrain ed by phys ed by st century, as rising oil prices stimulate ical depletion before 2030? en v Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Prelims_Intro:UKERC 23/09/2009 10:31 Page vi Page 10:31 23/09/2009 Prelims_Intro:UKERC

vi Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production • The mechanisms leading to a ‘peaking’ of conventional oil production are well 1. The main conclusions of the report are as follows: Key conclusions priorities for future research. feasibility of different approaches to mitigating such shortages, although both are The report does not investigate the potential consequences of supply shortages or the peak oildebate. Nevertheless, this report focuses primarily on the latter since they are the focus of the ground’ and‘belowrisksareinterdependentdifficulttoseparate. even ifeconomic andpoliticalconditionsprove morefavourable. Inpractice, these‘above whether physical depletionis security, even in the absence of ‘below-ground’ constraints. What is disputed, however, is resource nationalism)andseveral ofthesemay poseasignificantchallengetoenergy influenced byamuchwiderrange ofeconomic,politicalandgeopoliticalfactors(e.g. investment can overcome these physical constraints is contested. Global oil supply is also mediated byeconomic,technicalandpoliticalfactors,theextenttowhichincreased the distributionofresourcesbetweendifferentsizes offield.Whiletheseareinvariably conventional oilcanbeproduced,includingtheproductionprofileofindividualfields and The report also focuses on the broadly ‘physical’ factors that may restrict the rate at which is beyond thescopeofthisreport. fashion. Whiletheeconomicpotentialofnon-conventional fuelsisofcriticalimportance,it supply if‘non-conventional’ sourcesareunabletosubstituteinasufficientlytimely A peakinconventional oilproductionwillonlybeassociatedwithapeakinliquidfuels supply of liquid fuels in the period to 2030 and its resource base is comparatively depleted. natural gasandbiomass. Conventional oilisanticipatedtoprovide thebulkofglobal natural gas liquids (NGLs) but to exclude liquid fuels derived from oil sands, oil shale, coal, The reportfocuseson‘conventional oil’, definedheretoincludecrudeoil,condensateand assesses the risk of a near-term peak in oil production. uncertainty associated with key issues, compares contemporary forecasts of oil supply and the size ofoilresourcesandforforecastingfuturesupply, highlightsthedegreeof oil’ debate,identifiesthestrengthsandweaknessesofdifferentmethodsforestimating website. Thissynthesisreportclarifiestheconceptsanddefinitionsrelevant tothe‘peak supporting reportshave beenproducedandareavailable todownloadfromtheUKERC o ‘below-ground’ factors and little is to be gained by emphasising one set of v Oil supply is determined by a complex and interdependent mix of ‘above-ground’ and and global lev understood andprovide identifiable constraints onitsfuturesupplyatboththe regional ver theother. Nevertheless, fundamentalfeatures oftheconventional oilresource el. also likely toconstrain globalproductioninthenear-term, ariables Prelims_Intro:UKERC 23/09/2009 10:31 Page vii Page 10:31 23/09/2009 Prelims_Intro:UKERC • .There is potential for improving consensus on important and long-standing 3. • • Despitelargeuncertaintiesintheavailable data,sufficientinformationisavailable to 2. • therefore of critical importance for future supply. these fields, their future production profile and the potential for reserve growth are or so and few new giant fields are expected to be found. The remaining reserves at their peak of production, most of the rest will begin to decline within the next decade cumulativ for two thirdsof 100 fieldsaccountforhalfofproductionandup to500fieldsaccount approximately 25 fields account for one quarter of the global production of crude oil, increasingly well understood. Although there are around 70,000 oil fields in the world, The distributionofconventionalbetween different sizes oil resources of fieldis controv understanding of the nature and limitations of the available data. producing countries.Hence,thedebateonoildepletionwouldbenefitfromimproved global level which could potentially offset any overestimation of those reserves by key regional total. Doing so may lead to an underestimation of reserves at the regional and simply addtheestimatesof‘proved’ reserves fromdifferentoilfieldstoobtaina towards standardisation in reserve reporting. For example, it is statistically incorrect to Data uncertainties are compounded by errors in interpretation and the slow progress members, that hold the majority of the world's reserves. proportional totheirimportance–beinglowestforthosecountries,includingkey OPEC forecasts mustrelyuponassumptionswhoselevel ofconfidenceisinversely necessarily reliableforallregions.Intheabsenceofauditedreserve estimates,supply sources are better in this regard, but are also expensive, confidential and not limitations are insufficiently appreciated. The databases a Publicly available datasourcesarepoorlysuitedtostudyingoildepletionandtheir allow thestatusandriskofglobaloildepletiontobeadequatelyassessed. the globallevel. constrained oncetheresource issubstantiallydepleted.Thisincreasinglythecaseat available to the appropriate data and the range of ‘possible futures’ becomes more str affect conventional oil production, anticipating a forthcoming peak is far from Given the complex mix of geological, technical, economic and political factors that regions aroundtheworld. peaking of conventional oil production can be observed in an increasing number of discover andproducethesefieldsrelatively early. Thisprocesscanbemodelledandthe concentration ofresources inasmallnumberoflargefieldsandthetendencyto ultimately decline.Thesefeaturesincludetheproductionprofileofindividualfields, make itinevitablethatproductioninaregionwillrisetopeakorplateauand aightforward. However, supplyforecastingbecomesmorereliableonceaccessis ersies such as the source and magnitude of ‘reserves growth’. e discoveries. fields are relatively Most of these ‘giant’ old, many are well past v ailable from commercial vii Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Prelims_Intro:UKERC 23/09/2009 10:31 Page viii Page 10:31 23/09/2009 Prelims_Intro:UKERC viii Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production • Methods for estimating resource size and forecasting future supply have important 4. • • • • • methods such as ‘discovery process modelling’. But they are best applied to well- level dataisnotavailable andalso have much incommonwithmoresophisticated ‘curv of confidence when explor technical factorsasmuchgeologyandcanonly beestimatedtoareasonabledegree The ultimatelyrecoverable resources(URR)ofaregion dependuponeconomicand limitations that need to be acknowledged. historical experience are likely to require careful justification. been produced. Supply forecasts that assume or imply significant departures from this regions have reachedtheirpeakwellbeforehalfofrecoverable resourceshave from a region has rarely exceeded 5% of the remaining recoverable resources and most region and the pattern of production over time. For example, the annual production engineering andeconomicconstraints uponboththerate ofdepletionafieldor Oil reserves cannotbeproducedatarbitrarily highrates. Therearephysical, from falling. At best, this is likely to prove extremely challenging. oil production capacity may need to be replaced by 2030, simply to prevent production lead tomorerapid post-peak decline.Asaresult,morethantwothirdsofcurrent crude shifts towards smaller, younger and offshore fields and as changing production methods Decline rates are on an upward trend as more giant fields enter decline, as production current levels - equivalent to a new Saudi Arabia coming on stream every three years. mb/d of new capacity must be added each year, simply to maintain production at all currently-producing fieldsisatleast4%/year. Thisimpliesthatapproximately 3 production is at least 6.5%/year globally, while the corresponding rate of decline from existing fields. The av The oilindustrymustcontinuallyinvest toreplacethedeclineinproductionfrom which theymay beappliedremainkey areasofuncertainty. The suitability of these techniques for different sizes and types of field and the rate at prices may stimulatethemorewidespreaduseofenhancedoilrecovery techniques. mayterms. At decreaseinbothpercentageandabsolute thesametime,higheroil production shiftstowards newer, smallerandoffshorefieldstherate ofreserve growth Reserveto begreaterforlarger, growthtends olderandonshorefields,soasglobal of conservative reporting. different fieldsandregions,‘reserve growth’doesnotappeartobeprimarilytheresult so inthefuture.Whilecontributionofdifferentfactorsvaries widelybetween the pastdecadethandiscovery ofnewfieldsanditseemslikely tocontinuedo recoverable reserves. This processappearstohave addedmoretoglobalreserves over changes ineconomicconditionsandrevisionstoinitiallyconservative estimatesof grow over time as a result of improved geological knowledge, better technology, Estimates oftherecoverable resourcesofindividualfieldsarecommonlyobserved to e-fitting’ techniquesfor estimating URR have an importantroletoplay whenfield- er age rate ofdeclinefromfieldsthatarepasttheirpeak ation is well advanced. Although widely criticised, simple Prelims_Intro:UKERC 23/09/2009 10:31 Page ix Page 10:31 23/09/2009 Prelims_Intro:UKERC • • Largeresourcesofconventional oilmay beavailable, buttheseareunlikely tobe 5. • • • relatively narrowwindowinterms oftheleadtimetodevelop substitutefuels.Inthis this range appearswideinthelightofforecasts ofanimminentpeak,itmay be a the date of peak production can be estimated to lie between 2009 and 2031. Although about theglobalURRofconventional oil andtheshapeoffutureproductioncycle, assumptions aboutthesize oftheglobalresource.For awiderange ofassumptions The timingoftheglobalpeakforconventional oilproductionisrelatively insensitive to the mostpromisingareas. been underestimatedandtherearecontinuing restrictionsonexploration insomeof discoveries is lower than implied by the USGS, the size of these discoveries may have appears to be matching the USGS assumptions and although the rate of new that theUSGSestimatesare‘discredited’at bestpremature.Globalreserve growth uncertainty inglobalsupplyforecasts.Butdespitetheirapparentoptimism,assertions production throughto2007of1,128Gb. Thiswiderange leadstoacorresponding now fallwithintherange 2,000-4,300billionbarrels(Gb),comparedtocumulative represent a substantial departure from the historical trend. Contemporary estimates for the last 50 years, the most recent estimates from the US Geological Survey (USGS) Although estimates of the global URR of conventional oil have been trending upwards accessed quickly and may make little difference to the timing of the global peak. subject torapidly diminishingreturns. precision. Increasing model complexity does little to address this problem and is economic, or technological disruptions, no model can provide estimates of great and nosignificantdisruptionstotheoilmarket. Butgiven thepotentialforpolitical, estimated towithindecadalaccuracy assumingaparticularvalue fortheglobal URR The timingofaglobalpeak(orplateau)inconventional oilproductionmay be the presentationofuncertaintiesremainexception rather thantherule. economic variablesSensitivity testingand for multipleassumptions. andrequirement hampered by their reliance on proprietary datasets, lack of transparency, neglect of reliable basisforneartomedium-termforecasts,butmany existingmodelsare in all circumstances. Bottom-up models using field or project data provide a fairly approach hasitsstrengthsandweaknessesnosingleshouldbefavoured their inclusionofdifferentvariables andtheirlevel ofaggregationandcomplexity. Each Methods of forecasting future oil supply vary widely in terms of their theoretical basis, has contributedtoexcessively pessimisticforecastsoffuturesupply. production ordisco frequent neglectoffuturereserve growthandtheinabilitytoanticipatefuturecycles of techniques, suchasthesensitivityofestimatestochoice of functionalform,the Many analystshave paidinsufficientattentiontothelimitationsofcurve-fitting Since many regions do not meet these criteria, errors are likely to result. explored andgeologicallyhomogeneousareaswithaconsistentexploration history. very. This has led to underestimates of regional and global URR and ix Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Prelims_Intro:UKERC 23/09/2009 10:31 Page x Page 10:31 23/09/2009 Prelims_Intro:UKERC

x Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production • • • Theriskspresentedbyglobaloildepletiondeserve muchmoreseriousattentionbythe 6. • given seriousconsideration. to bothdevelop substitute fuels andimprove energyefficiency, thisriskneedstobe likely andthereisasignificant riskofapeakbefore2020.Given theleadtimesrequired evidence we suggest that a peak of conventional oil production before 2030 appears met orwhy theconstraints identifiedinthisreport donotapply. Onthe basisofcurrent implausible. Such forecasts need to either demonstrate how these assumptions can be 2030 restuponseveral assumptions thatareatbestoptimisticandworst we considerthatforecastsdelay the peakofconventional oilproductionuntilafter making precise forecasts of the timing of peak production unwarranted. Nev For medium to long-term forecasting, the number and scale of uncertainties multiply when demandrecovers. result ofthe2008economicrecessionandconsequentrisksupplyshortages primary issuefortheshorttermiscancellationanddelay oftheseprojectsasa term forecasts for any region can be constructed using widely available public data. The because theprojectswhichwillraise supplyarealreadycommitted.Reasonable short The shorttermfutureofoilproductioncapacity, toabout2016,is relatively inflexible, outcomes has not been adequately assessed. ph security andfailstoeitherassessoreffectiv Much existingresearchfocusesupontheeconomicandpoliticalthreatstooilsupply research and policy communities. resources lack the incentive or ability to invest. growth may also be constrained if the national oil companies that control much of these supply constraints may inhibitdemandgrowthatarelatively earlystage.Demand seems likely) theseresourcescanonlybeproducedrelatively slowlyathighcost, plausible, much of this is located in smaller fields in less accessible locations. If (as Although moreoptimisticestimatesoftheglobalURRconventional oilappear forecasts. following thepeak.Theseconsiderations constrain therange ofplausible globalsupply demand growth prior to the peak and/or a relatively steep decline in production assumptions aboutthesize oftherecoverable resourcecombinedwithaslowrate of is approximately 24 Gb). Delaying the peak beyond 2030 requires optimistic production byonlyafewdays (forcomparison,thecumulative productionfromtheUK model, increasing the global URR by one billion barrels delays the date of peak ysical depletion.Thishasmeantthattheprobabilityandconsequencesofdifferent ely integrate theriskspresentedby ertheless, - Prelims_Intro:UKERC 23/09/2009 10:31 Page xi Page 10:31 23/09/2009 Prelims_Intro:UKERC • • three general comments may be made. The evaluation of different mitigation options is beyond the scope of this report. However, Policy implications • dioxide (CO into liquidfuelswouldresultinemissionsofmorethan800billiontonnescarbon carbon non-con mitigate theeffectsofoildepletion,therewillbestrongincentives toexploithigh Second, althoughmany measuresassociatedwithclimatechangepolicywillhelpto developing anappropriatepolicyresponse. environmental constraints. Hence, even 2030 may not be a distant date in terms of assumes a relatively modest post-peak decline rate (2%/year) and ignores many importantmitigationoptions(e.g.publictransport, electricvehicles) italso liquid fuelssupplyaretobeavoided (Hirsch, reduction needtobeinitiatedatleast20years beforethepeakifseriousshortfallsin Department ofEnergyarguesthatlarge-scaleprogrammessubstitution anddemand of inv First, it seems likely that mitigation will prove challenging owing to both the scale of presented byglobaloildepletion. requires bothimproved understandingandmuchgreaterawareness oftherisks currently envisaged could potentially be required. For this to become politically feasible within nationalclimateprogrammes. Butgreaterandmorerapid changethanis This investment canbeencouraged bymeasurescomparable tothosebeingestablished uncertainty andvolatility andseemsunlikely tooccurwithoutsignificantpolicysupport. Third, investment inlarge-scalemitigationeffortswillbeinhibitedbyoilprice alternatives toconventional oilisofconsiderable importance. warming istobekept to2 future emissionsshouldbelessthan1,800billiontonnesifthemostlikely global through carboncaptureandstorage. Thiscomparestorecommendationsthattotal estment required and the associated lead times. For example, a report for the US 2 ), with less than half of these emissions being potentially avoidable v entional fuels.Converting onethirdoftheworld'sproved coalreserves o C (Allen, et al , 2009).Hence,earlyinvestment inlow-carbon et al,2005 ). Whilethisreportoverlooks xi Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Prelims_Intro:UKERC 23/09/2009 10:31 Page xii Page 10:31 23/09/2009 Prelims_Intro:UKERC xii Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Prelims_Intro:UKERC 23/09/2009 10:31 Page xiii Page 10:31 23/09/2009 Prelims_Intro:UKERC uuaiePouto Total production from a field or region since production began. Decline Rate Production Cumulative BERR API Gravity All-oil Depletion Total discoveries inthefieldorregionat aparticularpointin Cumulative Discoveries CTLs Crude Oil Conventional Oil All-liquids Glossary Condensate Biofuels Collective term used to include Collective term used to include heavy oil as <10° API. medium oil as 20-30° API, heavy oil as 10-20° API, and extra- Definitions vary, butlightoilisoftentaken as>30°API, crude oil UK DepartmentforBusiness,EnterpriseandRegulatory NGLs. usually weighted by production. are in decline. Aggregate estimates of decline rates are post-peak decline rate which refers to the subset of fields that fields thathave yet topasstheir peak ofproduction,andthe distinguish betweentheoverall declinerate whichincludes declines. Whenappliedtoaregion,itis importantto Annual rate at which oil production from a well, reserves time. Given by the sum of gasification of coal followed by a Fischer-Tropsch process. Coal-To-Liquids. Synthetic liquidfuelderived throughthe atmospheric temperature and pressure. underground reservoirs and which remain liquid at A mixtureofhydrocarbons thatexistinliquidphasenatural (non-conventional oils). NGLs, and to exclude Taken in this report to include heavier hydrocarbons. CTLs, GTLs and gasprocessingplants.Includespentanes(C temperatures andpressures.Producedatnatural gaswells oil), commonlyreferstobio-ethanolandbio-diesel. now taken over by Reform. Therelevant responsibilitiesofthisdepartmentare which hasbeenproduced. The portion of the estimated Very light oil which condenses from natural gas at surface Synthetic fuelsmadefrombiomass(suchascornorvegetable The AmericanPetroleum Institutesstandardisedmeasureof . density. APIgravity is measuredindegrees. and biofuels. DECC. oil sands, shale oil and ultimately recoverable resource crude oil, condensate, NGLs, cumulative production crude oil,condensate crude oil, condensate extra-heavy oil field or region 5 ) and and and and xiii

Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Prelims_Intro:UKERC 23/09/2009 10:31 Page xiv Page 10:31 23/09/2009 Prelims_Intro:UKERC xiv Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Depletion Rate EROI EOR EIA DECC Discovery Cycle Discovery Field Fallow Field Extra-heavy oil Heavy Oil GTL FSU Enhanced OilRecovery, alsocalledtertiary recovery. Typically of a If thelatteris annual productiontosomeestimateofrecoverable resources. Energy Return On Investment. A measure of the ratio of factor change thepropertiesofoilandincrease microbes, directionalboreholesorheatintoa involves theintroductionofgas,solvents, chemicals, Administration. US Department of Energy’s Energy Information UK Department of Energy and Climate Change. profile. begins towhenitends.Analternative termis The annual rate at which the inverse of the owing to the phenomenon of period tothenext.Thesemeasuresmay notbethesame period; orb)thechangeincumulative discoveries fromone fields thatarenewlydiscovered withinaparticulartime An areaconsistingofasingle scheduled for development. An oilfield injection. high viscosity, extra-heavy oil has to be produced using steam Crude oil recoverable from the produced fuel. energy expendedinoilexploration andproductiontoenergy Either: a)theeconomicallyrecoverable resourcescontainedin ‘extra heavy’. This definition is not consistent, however, with 20°. OilwithAPIgravity lessthan10° is oftenreferredtoas Commonly definedas methane using the Fischer-Tropsch process. Gas-to-liquids. Synthetic fuelderived fromtheliquifactionof one to thousands. and thenumberofwellsinaproducingfieldmay range from be discovered, underdevelopment, producing orabandoned all relatedtoasinglegeologicalstructure.Fields may either A graph of Former Soviet Union. field . or region are being produced. Defined as the ratio of having anAPIgravity lessthan10°.Becauseofits that has been discovered but is not presently discovery R/P ratio. proved reserves,thedepletionrate is the crude oil against time,fromwhen eann recoverable resources remaining reserve growth. reservoir having aAPIgravity lessthan or multiplereservoirs, reservoir discovery discovery recovery to Prelims_Intro:UKERC 23/09/2009 10:31 Page xv Page 10:31 23/09/2009 Prelims_Intro:UKERC Production Primary Recovery Peak OPEC Oil Sands OOIP OGJ NOCs NGLs Natural Gas MMS IOCs IEA Hydrocarbons Play Plateau Petroleum Basin Petroleum Quantity of oil recovered from a field or region over a Organisation ofPetroleum ExportingCountries. atoms. specified periodoftime. Alsotermedrate of production,or boundary. attributes and lie within some well-defined geographic can be mined and processed to produce Sandstone impregnatedwithheavy or extra-heavy oilthat reservoir, Original Oil In Place. Total quantity of oil contained within a Oil and Gas Journal National Oil Companies (i.e. State owned). the applicationofmoderate pressure. pressures, or can be relatively easily turned into a liquid with natural gas Natural GasLiquids.Light hydrocarbons foundassociatedwith International Oil Companies. International EnergyAgency. example, oxygen, nitrogen, sulphur, vanadium etc. molecules, butitmay alsocontainsmallamountsof, for 25°. Venezuela includingoilupto22°asheavy, andCanadausing petroleum An areaforpetroleumexploration,of containingacollection is higher than a specified percentage of peak production. progressively buried,heatedandcompressed. source of petroleum, as a result of organic carbon being to containhydrocarbons. Sedimentarybasinsaretheprimary sedimentary orvolcanic rocksandwhichisknownorexpected including gases, liquids and solids (bitumen). Methane found naturally occurring in reservoir rock. US DepartmentoftheInteriorMinerals ManagementService. Period surroundingproduction The highestannual The recovery of oil under its own natural pressure. A singleareaofsubsidencewhichfilledupwitheither General nameforallnaturally occurring Any moleculeconsistingentirelyofcarbonandhydrogen Petroleum field prospects that areeitherliquidatnormaltemperatures and or region before production begins. production is primarily a mixture of hydrocarbon which sharecertaincommongeological peak of oil from a field or region. where annualproduction hydrocarbon syncrude. species, xv

Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Prelims_Intro:UKERC 23/09/2009 10:31 Page xvi Page 10:31 23/09/2009 Prelims_Intro:UKERC xvi Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production osbe(P eevs have alow Thequantityofoilinknownfieldswhichisconsideredto possible (3P) Reserves Proved, probable and Reserves Recovery Factor Province Production Cycle Reserves Thequantityofoilinknownfieldswhichisconsideredto(2P) (2P) probable and Proved Reserve Growth The economically recoverable resources that have yet to be Resources Remaining Recoverable Refinery Gains rvd(P eevs Thequantityofoilinknownfieldswhichisconsideredto have Reserves (1P) Proved Prospect economically recovered. probable reserves of confidence,suchas under defined conditions. May be estimated to different levels be technicallypossibleandeconomicallyfeasible toextract yet-to-find. May beestimatedtodifferinglevels ofconfidence. for resourceassessment. solely onthebasisofgeologicalconsiderations thatisrelevant petroleum petroleum formation. May contain a single or several recovered. A graph of on a daily (mb/d) or annual (Gb/year) basis. rate ofchangecumulative production.Normallymeasured recovered. more oil than was initially estimated as reserves. reserves reserves specific gravity than the crude oil which was refined. the productionofproductswhich,onaverage, have alower products and the volumetric input of crude oil. Attributed to with current or anticipated technology. An areawithcommongeologicalpropertiesrelevant to a high suitable targetforexploration. pools ofrecoverable A geologicalanomalythathassomeprobabilityofcontaining profile. begins to when it ends. An alternative term is have a The phenomenonbywhich many fieldsultimately produce Those quantitiesofoilinknownfieldswhichare consideredto produced fromafieldorregion.Definedas thesumof The differencebetweenthevolumetric outputofrefinery The percentageof , anticipatedfuture probability (e.g.>90%)ofbeingeconomically (3P). medium basins. Aprovince isthelargestentitydefined production probability (e.g. >10%) of being economically original oilinplace (2P); hydrocarbons probability (e.g.>50%)ofbeing against time,fromwhen proved reserves or proved,probableandpossible reserve growth and isconsideredtobea that canberecovered (1P); and anticipated proved and production production Prelims_Intro:UKERC 23/09/2009 10:31 Page xvii Page 10:31 23/09/2009 Prelims_Intro:UKERC Reservoir (or‘pool’) Resource R/P Ratio Synfuels Syncrude Therecovery ofoilusingwater orgasinjectiontomaintain Supply Secondary Recovery SEC YTF WEO URR UAE single natural pressure system. A single physically separated fromotherreservoirs andwhichhasa A subsurfaceaccumulationofoiland/orgaswhichis Ratio of some measure of oil The totalquantityof many reservoirs. undiscovered fields. considered economically feasible to extract as well as those in region, includingthoseinknownfieldswhicharenot production. Normally defined with respect to much as conventional sands strategic reserves or lost through accident. slightly differentfromproductionsinceoilmay bestoredas Volume of produced oil that reaches the market. May be pressure. US Securities and Exchange Commission. frame. is expectedtobediscovered ina region inarelevant time Yet-To-Find. The amount of economically recoverable oil that IEA World Energy Outlook. recovery (EUR). it finallyends.Analternative termisestimatedultimate region over alltime–fromwhenproductionbeginsto estimated to be economically extractable from a field or Ultimately Recoverable Resource.amount ofoilthatis The Liquid fuels made from coal ( Synthetic crudeoil madefromthebitumeninCanadian United Arab Emirates. . Syncrude canbe handled, pumped,pipedandrefined crude oil. hydrocarbons CTL) or gas ( reserves estimated toexistina GTL). field proved reserves. to annualoil may contain oil xvii

Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Prelims_Intro:UKERC 23/09/2009 10:31 Page xviii Page 10:31 23/09/2009 Prelims_Intro:UKERC xviii Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production t m boe Gb mb/d mb kb b Units 3 Tonnes Billion barrels Cubic meters Barrel of oil equivalent (6.1 GJ) Million barrels per day Million barrels Thousand barrels Barrels. 42 US Gallons or 158.76 litres Prelims_Intro:UKERC 23/09/2009 10:31 Page xix Page 10:31 23/09/2009 Prelims_Intro:UKERC 4 Looking beneath - methods of estimating ultimately recoverable resources recoverable ultimately estimating of methods - beneath Looking 4 .43 ...... growth reserve Enduring controversies – field sizes, and decline rates 3 Readingthe fuel gauge measuring- oil 2 supply and resources ..15 Introduction ..1 1 Contents . icvr vrefr ehius. . .88 techniques effort over Discovery 4.5 . ..84 techniques time over Discovery ...... 79 4.4 techniques time over Production 4.3 . .75 techniques of Overview . . .73 4.2 resources recoverable ultimately of importance The 4.1 . umr ...... 70 Summary 3.5 ...... rates. .67 Depletion 3.4 ...... 58 Rates Decline 3.3 ...... 50 Growth Reserve 3.2 ...... 43 distributions size Field 3.1 ...... 41 Summary ...... 39 2.7 schemes classification Reserve 2.6 ...... 36 estimates reserve in uncertainty Assessing 2.5 ...... 28 reserves reporting and Estimating ...... 26 2.4 figures key and data of Sources 2.3 ...... 15 reserves and discovery production, of Measures .2.2 . WhatisOil?...... 16 2.1 . . .12 report the of Structure ...... 10 1.6 conducted was assessment the How ..8 1.5 report the of scope and Objectives ...... 5 1.4 peaking oil of Mechanisms ...... 2 1.3 oil? of out running we Are ..1 1.2 oil’ ‘peak and resources oil of Depletion 1.1 . . .73 .. einladgoa vrg eln ae ...... 63 rates averagedecline global and Regional ...... 59 3.3.2 decline production of Analysis 3.3.1 ...... 56 growth reserve global Estimating ...... 55 3.2.4 growth reserve Variationsin ...... 53 3.2.3 growth reserve forecasting and Estimating 3.2.2 ...... 50 growth reserve of Sources 3.2.1 ...... 46 fields small of contribution The ...... 3.1.2 . . . .44 fields large of importance The 3.1.1 ...... 25 depletion oil Measuring ...... 24 2.2.6 recoverableresources Ultimately ...... 23 2.2.5 Discoveries 2.2.4 ...... 21 discoveries Cumulative ...... 19 2.2.3 Reserves ..19 2.2.2 Production 2.2.1 . xix

Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Prelims_Intro:UKERC 23/09/2009 10:31 Page xx Page 10:31 23/09/2009 Prelims_Intro:UKERC xx Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Possiblefutures – a comparison of global 7 supply forecasts ..139 How much do we have? – global estimates of the ultimately recoverable 6 Lookingahead methods- of forecasting 5 future oil supply ..103 . Theassumedorimpliedglobalultimately recoverable resourceofconventional ...... 144 7.4 forecasts oil current of Overview 7.3 ...... 140 Forecasts Oil Historical ...... 139 7.2 Approach 7.1 ...... 136 Summary ...... 134 6.5 supply global future for Implications 6.4 . ..130 estimates URR global Recent 6.3 ...... 124 2000 WorldAssessment PetroleumUSGS The . . . . . 6.2 ...... 123 estimates URR Trendsglobal in 6.1 resourcesof conventional oil . . .123 ...... 122 Summary 5.6 ...... 116 models depletion on thoughts Synthesizing 5.5 ...... 112 depletion oil of models Economic 5.4 ...... 110 Bottom-up models: building up oil depletion from lower levels 5.3 ..109 technologies and rates, discovery resources, simulation: Systems 5.2 Simplemodelsofoildepletion:reserve-to-production ratios andcurve-fitting 5.1 ...... 101 Summary ..97 4.8 econometrics with curve-fitting Reconciling ..93 4.7 techniques curve-fitting of Consistency 4.6 i ...... 152 oil ...... 132 provinces additional of Inclusion ...... 130 6.3.2 2008 Outlook WorldEnergy IEA The 6.3.1 . . .129 Evaluation ...... 126 6.2.3 Results ...... 124 6.2.2 Methods 6.2.1 ...... 121 modelling depletion oil improving forward: Moving ...... 120 modelling 5.5.5 of purpose the and Complexity ...... 118 5.5.4 prediction of problem The ...... 117 ability predictive of 5.5.3 indicator an as fit Model ..116 types model 5.5.2 across trends and classification Model 5.5.1 ...... 115 models economic with Difficulties ...... 114 5.4.3 depletion oil of models Econometric ...... 113 5.4.2 theory depletion Optimal 5.4.1 ...... 112 models bottom-up with Difficulties 5.3.1 ...... 110 models simulation with Difficulties 5.2.1 ..106 models curve-fitting with Difficulties ...... 104 5.1.2 models Curve-fitting 5.1.1 ...... 103 Prelims_Intro:UKERC 23/09/2009 10:31 Page xxi Page 10:31 23/09/2009 Prelims_Intro:UKERC AnnexTechnical2: . ..198Reports AnnexProject1: team, Expert Group and contributors References ..197 ..175 .167 ...... implications policy andneeds research Conclusions, 8 . umr ...... 164 ...... 162 Summary 7.8 Discussion 7.7 The impacts of rates of discovery, reserves growth and depletion on the date 7.6 The interaction between ultimately recoverable resources, the aggregate 7.5 . oiyipiain ...... 173 implications Policy 8.3 ...... 171 ...... needs ...... Research...... 167 8.2 Conclusions 8.1 7.6.4 Bottom-up modelling ...... 162 Bottom-up 7.6.4 ...... 163 rates Depletion 7.6.3 ...... 160 rule’ ‘PFC The ...... 159 7.6.2 peaking ‘Mid-point’ 7.6.1 ...... 159 peak of ...... 154 production peak of date the and rate decline xxi

Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Prelims_Intro:UKERC 23/09/2009 10:31 Page xxii Page 10:31 23/09/2009 Prelims_Intro:UKERC xxii Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production iue32Oladgsfedsz itiuinfrteDne ai n15 . . . .47 Oil and gas field size distribution for the Denver basin in 1958 Figure 3.2 The estimated contribution of giant oilfields to global crude oil production ..40 System Management ResourcePetroleum The Figure 3.1 2.21 Figure Probability and cumulative probability distribution of recoverable reserves ...... 34 producers post-peak of group expanding The Figure 2.20 2.19 Figure Change in all-oil production over the last 5 years for top 49 producing ...... 33 regions. selected for production all-oil world of Share Figure 2.18 ..32 country by production all-oil world of Share 2.17 Figure ...... 32 reserves proved of distribution Regional 2.16 Figure ...... 31 sources data different from trends reserve proved Global 2.15 Figure ...... 31 sources data different from trends production oil Global 2.14 Figure 2.13 Figure Figure 2.12 Two ways ofpresentingUKcumulative discoveries -backdated 2P versus ..26 (URR) resources recoverable ultimately of Components 2.11 Figure ...... 25 discoveries backdated and production in trends Global 2.10 Figure . .24 Global trends in backdated discoveries and cumulative discoveries 2.9 Figure Figure 2.8 ...... 22 reserves oil proved in trends Global ...... 21 Figure 2.7 reserves and resources Oil ...... 21 2.6 Figure production oil capita per in trends Global 2.5 Figure ...... 20 production oil in trends Global 2.4 Figure ...... 17 liquids hydrocarbon of classification Proposed ...... 17 2.3 Figure production fuels liquid 2008 of Breakdown ...... 8 2.2 Figure field by UKCS the in production Oil ...... 7 2.1 Figure production oil in peak regional a of model Stylised ...... 6 1.6 Figure Sea North the in field Thistle the of cycle Production ...... 5 1.5 Figure States United the of history production Oil 1.4 Figure ...... 3 fuels hydrocarbon liquid potential of base resource global The ...... 2 1.3 Figure 2009 March to 2002 January - production liquids Worldtotal 1.2 Figure 1.1 Figure Figures onre.. . .33 countries...... 27 discoveries Effect ofreserve growthonbackdatedestimatesofcumulative ...... 27 1P current ..23 revisions reserve of treatment Current versus backdatedestimatesofcumulative discoveries – ...... 45 ...... 37 Prelims_Intro:UKERC 23/09/2009 10:31 Page xxiii Page 10:31 23/09/2009 Prelims_Intro:UKERC iue41 Creamingcurve forRegion E fitted withone(top)andtwo(bottom) Figure 4.13 Creaming curves for Region Ausingexponentialand hyperbolic functional ..96 J Region for curves decline Production 4.12 Figure 4.11 Figure Cumulative discovery projectionforRegion Eusinglogistic model– ..94 results projection discovery Cumulative ..91 Figure curve 4.10 effort per yield a of Example 4.9 Figure ...... 90 curve ‘creaming’ a of Example ...... 87 4.8 Figure projections discovery on growth reserve of impact The 4.7 Figure ...... 85 Hubbert’s idealised lifecycle model of an oil-producing region 4.6 Figure Linearisation of exponential production decline for the UK Forties field84 ..83 Figure 4.5 production oil US of Linearisation’ ‘Hubbert . Figure 4.4 . . . .81 A fit of the logistic model to US production of conventional oil 4.3 Figure ...... 78 time of function a as region a in discovery Cumulative Figure 4.2 ..67 2030 to production all-oil global of forecast IEA 4.1 Figure 65 Evolution of production-weighted giant oilfield decline rates over time 3.15 Figure Decline curves for UK offshore fields entering peak at 5 year intervals63 Figure 3.14 Figure 3.13 ProductionfromUKoffshorefieldswhichpeaked before1997,stacked by 3.12 Figure ProductionfromfourUKoilfieldsfittedbythreeexponentialdecline ..60 field oil an of cycle production Stylised Figure 3.11 3.10 Figure Contribution of large producers to global reserve growth between 2000 Figure 3.9 Reserve growthintheIHSEnergy PEPSdatabasebetween2000and ..54 data reserve 1P US for functions growth reserve Cumulative Figure . 3.8 . . . . .53 Reserve growth in oil fields smaller than 0.5 Gb in the UKCS 3.7 Figure ...... 52 Reserve growth in oil fields larger than 0.5 Gb in the UKCS Figure 3.6 Figure 3.5 Log-logplotofcumulative frequencyversus fieldsize, illustrating the HowtheundersamplingofsmallfieldsmayFigure leadtoalognormalfrequency 3.4 3.3 Figure yebl ...... 98 hyperbola ...... 97 forms ...... 95 series data of length to sensitivity ...... 62 year peak ...... 62 models ...... 59 2007 and ...... 2007 ...... 58 ...... 49 distribution sample difference betweentheoreticalpopulationdistributionandobserved ...... 48 fields discovered of size the of distribution xxiii

Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Prelims_Intro:UKERC 23/09/2009 10:31 Page xxiv Page 10:31 23/09/2009 Prelims_Intro:UKERC xxiv Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production iue77Mapping globalsupplyforecastsaccordingtotheimpliedURRof Figure Theeffectonthedateofpeakvarying theURRandpost-peak 7.7 Figure Constituentsandrange ofuncertaintyinthemodelassumptionsfor 7.6 ...... 148 2030 to production all-oil of forecasts ‘Peaking’ Figure 7.5 . . . . forecasts of . all-oil and all-liquids to 2030 ‘Quasi-linear’ . . . .147 7.4 Figure . . . Comparison of thirteen forecasts of all-oil . production to 2030 .144 Figure 7.3 Figure 7.2 Pre-1973logisticforecastofglobal oilproductioncomparedtoactual Figure Illustration of the asymmetric production model of Kaufmann and Shiers 7.1 Figure 6.7 Thepeakingofglobalconventional oilproductionunderdifferent Figure IEA2008:componentsoftheestimatedglobalURRforconventional oil 6.6 Figure 6.5 Comparinghistoricaltrendsinbackdated2Pdiscoveries withthose Figure USGS 2000: components of the estimated global URR for conventional oil 6.4 125 USGS estimates of undiscovered crude oil resources outside the US 6.3 Figure ..124 years 70 last the over estimates URR Global ...... 116 Figure 6.2 dimensions three in classified types model Different 6.1 Figure 5.4 Figure Six model types applied to regions where they were found to be the best 5.3 Differencebetweenaggregateexponentialdecline(left)anddisaggregate Figure ...... 104 regions post-peak in ratios production to reserve Proved 5.2 Figure 5.1 Figure grgt eln ae ...... 156 rate. decline aggregate conventional oil,thedateofpeak productionandthepost-peak ..155 rate decline aggregate ...... 153 oil conventional of URR global ...... 142 production . .135 model logistic simple - URR global the about assumptions ..128 1995-2025 period the for 2000 USGS the by implied ...... 127 values) (mean ...... 108 model fitting ...... 105 (right) decline exponential ...... 136 ...... 132 Prelims_Intro:UKERC 23/09/2009 10:31 Page xxv Page 10:31 23/09/2009 Prelims_Intro:UKERC al . Applying the PFC 60% rule to forecasts of global cumulative production Table Summary of the main parameters and assumptions used by each 7.6 . . . .146 Comparison of thirteen forecasts of all-oil production to 2030 ...... 145 study this in Tablecompared Forecasts 7.5 Table 7.4 Selectedforecastsofglobaloilproductionthatforecastnopeakbefore Table7.3 Selectedforecastsofglobal oilproduction,madebetween1956and Table 7.2 IEA 2008 WEO: mean estimates of global URR for petroleum liquids Table 7.1 USGS 2000: mean estimates of global URR for conventional oil (Gb)127 Table 6.2 Classificationofcurve-fitting techniquesbytheirchoiceofexplainedand Table 6.1 Estimateddepletionatpeakandannualrate atpeakforgiant Table4.1 Table 3.7 IEAestimatesofaggregateproduction-weighted declinerates for Estimates of production-weighted aggregate decline rates for samples of Table 3.6 ..64 studies rate decline global of Comparison Table3.5 Contribution of large producers to global reserve growth between 2000 Table3.4 ..46 world the in fields oil largest ten The Table 3.3 Ivanhoe andLeckie’s estimatesofthesizeworld’s distributionofthe Table3.2 Deterministicandprobabilisticterminologiesassociatedwithoilreserves Table 3.1 . .29 Comparison of global data sources on oil production and reserves Table2.2 Table 2.1 Tables n uuaiedsoeis...... 161 discoveries cumulative and ...... 150 forecast ...... 143 2030 ...... 141 peak the for date a gave which 2005, ...... 131 (Gb) ...... 79 variables explanatory ...... 68 fields oil ...... 65 (%/year) field post-peak of sizes different ..65 (%/year) fields post-peak large ...... 59 2007 and ..44 oilfields ...... 37 estimation. xxv

Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Prelims_Intro:UKERC 23/09/2009 10:31 Page xxvi Page 10:31 23/09/2009 Prelims_Intro:UKERC xxvi Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Chap1:UKERC 22/09/200909:51Page1 demand wellintothe21 production willbesufficienttomeetglobal ‘optimists’ whobelieve thatliquidfuels timescale required. On the other side are being unable to ‘fill the gap’ on the alternative and non-conventional sources to substantialeconomicdislocation,with Schindler, 2007).Thisisexpectedtolead 1997; Deffeyes, 2005;Zitteland production ofconventional oil(Campbell, subsequent terminal decline in the global forecast an imminent peak and and contentious. On one side, ‘pessimists’ The debateover oildepletionispolarised resources and‘peakoil’ Depletionofoil 1.1 depletionphysical before 2030? ‘conventional oil’ will be constrained by propos What evidence isthere to support the addresses the following question: that is useful to policymakers. This report reports, whileexplainingresultsinaway to provide rigorous and authoritative of the current state of knowledge. It aims field through comprehensive assessments address key controversies intheenergy Assessment (TPA) functionwas setupto The UKERCTechnology andPolicy Introduction 1. technology. Similarly, pessimistsclaim importance ofinvestment andnew supply, while optimists emphasise the factors will largely determine future oil 2004). Pessimists claimthatgeological 2003; CERA, 2005; Mills, 2008; Odell, resources such as oil sands (Adelman, the development ofnon-conventional enhanced recovery of conventional oil and oil prices stimulate new discovery, the i tion thatthe global supply of st century, asrising forthcoming. the required investment would be expressed seriousconcerns aboutwhether conventional oilbefore2030,the IEA not envisage a peak in the production of preceding 23 years. While this scenario did than had been achieved over the output ofSaudiArabia or20mb/dmore 2030, equivalent to six times the current need tobeputintoproductionbefore barrels/day (mb/d)ofnewcapacitywould the IEAestimatedthat64million existing fields. In its reference scenario, accelerating declineinproductionfrom included a detailed examination of the International Energy Agency (IEA) which Energy Outlook (WEO) from the security. Amilestonewas the2008World analysts has increased concerns about oil and ominouswarnings frommarket prices, declining production in key regions China, IndiaandtheMiddleEast),rising strong demand growth (especially in But beginningin2003,acombinationof between nowand2020.” (BERR,2008). eventuality of crude oil supplies peaking plansspecificallyforthe contingency and “…..doesnotfeeltheneedtohold mentions the issue in official publications example, the UK government rarely conventional policy discourse. For has had relatively little influence on The growingpopulardebateon‘peakoil’ price (Gately, 2004). enough toavoid significantincreasesin little incentive toexpandcapacityfast investment, withOPECnationshaving argue thatproductionwillbelimitedby ground is provided by commentators who ‘end of oil’ (Lynch, 1998). Some middle the longhistoryoffailedpredictions global situation while optimists point to that thereisnoprecedentforthecurrent 1 Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Chap1:UKERC 22/09/200909:51Page2

2 Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production gas-based oilsubstitutesandmethane-based blendingcomponents. Note: Source: March 2002to 2009 January - production World liquids total 1.1 Figure poor. understanding ofthistopicremainsfairly consequence, thegeneral level of analysts remainsceptical.Asa dismissive and the majority of energy several oilcompanieshave beenpublicly little concernaboutphysical depletion, unresolved. Most governments exhibit resource depletion and peak oil remains outlook has changed, the dispute about (IEA, 2009a). But while the short-term and price spikes when demand recovers next five years and the risk of shortfalls likely result is constrained supply over the the longleadtimeonthoseprojects, delay ofmany upstream projects.Given January 2009)andthecancellationor $150/barrel in July 2008 to $40/barrel in (Figure 1.1), tumbling oil prices (from major reduction in global oil demand worsening economic recession leading to a publication ofthe2008WEO, withthe The supplyoutlookhaschangedsincethe Includes crudeoil,condensate,natual gasliquids,refinerygains,oilsands,heavy oil, oilshales,coal-basedandnatural

IEA Production (mb/d) 75000 77000 79000 81000 83000 85000 87000 89000

2002

2003

2004

2005 sources of supply may be required. In which demand reduction and alternative following the peak and hence the rate at production may beexpectedtodecline equally importantistherate atwhich has beenafocusofdebate,whatappears (or plateau)inconventional oilproduction be made.Whilethetimingofafuturepeak anticipated if required investments are to security implications, which need to be have major economic, environmental and substitute sourcesofenergyislikely to the transition from conventional oilto (Rubin and Buchanan, 2007). In addition, importing countriescouldbemagnified production, theeconomicimpactsin capacity declines more rapidly than global major economic impacts. If global export production ofconventional oilcouldhave of energy, a decline in the global demand reduction and substitute sources because withoutsufficientinvestment in The debateisnevertheless important Year

2006

2007

2008

2009 Chap1:UKERC 22/09/200909:51Page3 hydrates are ignored due to a lack of reliable data. less certainresources.Results arebasedonconversion efficienciesofcurrenttechnologiesavailable intheopenliterature. and coalareusedasfeedstockfor liquidfuelsandnoneforotherpurposes.Thelightlyshadedportions ofthegraph represent are gas-and coal-derived syntheticliquidfuels.TheCTLandGTLquantitiesare theoreticalmaximabecausetheyassumeallgas Note: Source: The globalFigure resource 1.2 base of potential liquid hydrocarbon fuels biofuels or for replacing liquid fuels with without consideringthepotentialfor fuels intheforeseeablefuture,even very of liquid little risk of ‘running out’ But asFigure1.2demonstrates, thereis concern is precisely when this will occur. running dry, whichimpliesthatthemain being slowlydrained andeventually ‘running out’. The image is one of a tank often characterised as concerns about Concerns aboutglobaloildepletionare oil? Arewerunningoutof 1.2 2008; Reynolds, 1999b). timely fashion (Kaufmann and Shiers, upon tosignaloildepletioninasufficiently extent towhichthemarket may berelied addition, there are uncertainties over the Global resourcesoffossilhydrocarbons thatcouldbe converted toliquidfuels.EORisenhanced oil recovery, GTLandCTL Farrell andBrandt (2006). Production cost (2000 $ per b) consumed Already ,0 ,0 ,0 ,0 8,000 6,000 4,000 2,000 1,000 100 20 40 60 80 0 oil Conv. consumed Yet to be EOR Potential forliquidhydrocarbon production (Gb) and heavy oil and heavy Tar sands Reserves GTL synfuels exploitation of non-conventional resources Third, compared to conventional oil, the could constrain theiruseforliquidfuels. abstraction and carbon emissions, which including landscape destruction, water severe environmental consequences, of many oftheseresourceswillhave come toanend.Second,theexploitation implies that the era of cheap oil may have which have beenusedtodate,which extract, transport and/orrefinethanthose frequently more expensive to locate, First, theremainingresourcesare At least four other factors are relevant. future oil supply, nor the most important. resource is neither the only constraint on But theabsolutesize ofthehydrocarbon conventional fuels is very much larger. oil andtheresourcebaseofnon- than half of its endowment of conventional electricity. To date,theworldhasusedless Increasingly uncertain resources uncertain Increasingly 10,000 2001,0 60018,000 16,000 14,000 12,000 CTL synfuels CTL Oil shale Gas 3 Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Chap1:UKERC 22/09/200909:51Page4

4 Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production 2 Excluding biofuels and refinery gains. Finally, the society islikely tobereduced(Box 1.1). net energy processing chain,withtheresultthat consumption atallstagesofthe generally requires more energy investment thatisrequired.Andthislast and/or location,togetherwiththescaleof low asaresultoftheirphysical properties remaining resourcescouldberelatively Box 1.1: Net energy and the future supply of liquid fuels liquid of supply future the and energy Net Box 1.1: in 1930 to around 20:1 in the mid-1990s, while Gagnon equivalent EROIforpetroleumextraction intheUSoilproductionfellfromaround 100:1 conventional sources.For example,Cleveland (1992a;2005)estimatesthatthethermal accessing smallerfieldsinmoredifficultlocations(e.g.deepwater) andshiftingtonon- The energycostsofoilproductionhave increasedover timeasaconsequenceof and processing has been taken into account. gain fromtheproductionofoilandotherresources,onceenergyusedinextraction as drilling rigs. The energy return on investment (EROI) is a measure of the net energy the energythatisrequiredtoproduceandmaintainrelevant capitalequipmentsuch consumption ofenergyateachstage–forexample,inpumpingwater intoawell–and transport ittotherefineryandproduceoilproducts.Thisincludesbothdirect neglects theenergythatisrequiredtofindresource,extract itfromtheground, Oil productionisnormallymeasuredbyvolume (barrels)orenergycontent(GJ).Butthis implications that should not be overlooked (Hall, But the anticipated decline in the net energy available to society has important by mostenergyanalystswhoprefertomeasure scarcityonthebasisofpriceorcost. Changes intheaggregateEROIforliquidfuels aredifficulttoestimateandignored the laying of a pipeline may reduce the energy required to exploit small, nearby fields. result inanetlossofenergy(Gilliand,1975). EROIalsovaries withtime:forexample, such as coal, into a more valuable liquid fuel may make financial sense even though they et al into accountthe‘quality’ofenergycarriersas reflected intheirmarket price(Cleveland, Estimates ofEROIvary withthemethodandsystemboundary adoptedandshouldtake fuels. the EROI for conventional oil is much greater than that for the majority of substitute costs of converting crude oil to road fuels lowers this further to around 10:1. However, EROI forglobaloilandgasproductionfellfrom26:1in1992to18:12006.Theenergy ., 2000; Kaufmann, 1994).Projectswhichconvert alessvaluable energycarrier, available forproductive usesin rate of productionthe thousand barrels a second (a barrel is 159 day (mb/d),orapproximately one fuels In 2008,theglobalproductionofliquid eventually decline. and thereasonswhy thatrate must but the is not so much the siz point isthekey to thepeakoildebate:it et al., 2009; 2008). 2 averaged 82.3millionbarrelsper rate ofproduction et al . (2009) estimate that the e of the resource, of thatresource Chap1:UKERC 22/09/200909:51Page5 3 The2007World EnergyOutlookforecast116mb/d in2030. exploit the resource and the various resource, thetechnologyavailable to influenced by the physical features of that The rate of production of a resource is peaking Mechanismsofoil 1.3 IEA forecasts, significant downward revisionofprevious more thandouble.Butdespitethisbeinga – withtheglobalcarfleetprojectedto demand forpersonalautomotive transport in thedeveloping worldandtheexpanding 26%), largelyasaresultofincomegrowth 103.8 mb/dby2030(anincreaseof litres). TheIEAforecaststhisincreasingto of time. whether it can be sustained for any length whether 100mb/d is achievable, or if so, Note: Source: States United the of history production Oil 1.3 Figure Annual productionofcrudeoil,condensate, natural gas liquidsandextra-heavy oilfromallUSstates.

IHS Energy Production (Gb) 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 8016 8010 9014 9018 002020 2000 1980 1960 1940 1920 1900 1880 1860 1840 3 many analystsquestion Year debate. become acentral themeofthepeakoil analytical approachhassubsequently earlier byM.KingHubbert(1956),whose peak was famously anticipated 15 years history is illustrated in Figure 1.3. This the United States, whose production countries toexperiencesuchapeakwas the world(Brandt, 2007). Oneofthefirst in an increasing number of regions around production cannevertheless beobserved significant -degrees,the‘peaking’ofoil pattern to varying – and sometimes While numerous factors can modify this region risingtoapeakandthendeclining. oil resource leads to production from a argue thatthenatureofconventional involved. Those concerned about ‘peak oil’ the behaviourthe organisations of economic and political factors that affect 5 Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Chap1:UKERC 22/09/200909:51Page6

6 Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production entire world.For example,theIEA(2008) aggregation fromindividualbasinstothe pattern can be observed at all levels of much larger number of small fields. This fields, withthebalancebeinglocatedina to belocatedinasmallnumberoflarge Second, mostoftheoilinaregiontends Thistle field in the North Sea. 1.4 shows the production history of the generally observed. Asanexample, Figure managed, the same broad pattern is manner inwhichitisdeveloped and both itsphysical characteristics andthe smooth) productionprofileasaresultof will have aunique(andnotnecessarily breakthrough ofwater. Whileeachfield a result of falling pressure and/or the peak orplateau,afterwhichitdeclinesas from individualfieldsnormallyrisestoa economics andpolitics.First,production influence is invariably mediated by regional oil production, although their resource contribute to the peaking of Three ‘physical’oil the of features Source: ProductionFigure 1.4 of cycle the Thistle field in the North Sea UK DepartmentofEnergyandClimate Change Production (kb/d) 100 120 140 20 40 60 80 0 1975 1980 1985 1990 Year the world’s fieldswerediscovered ‘giant’ exploration. Asanillustration, over halfof restrictions ontheareasavailable for political andeconomicfactors,suchas although itisalways modifiedbytechnical, pattern can be observed at all levels, more efforttolocate.Again,thisbroad progressively smallerandoftenrequire Subsequent discoveries tendtobe a largersurfacearea. because theyoccupy exploration history of a region, in part discovered relatively early in the Third, these large fields tend to be crude oil that has ever been discovered. account foraroundtwothirdsofallthe Saudi Arabia. fields Around 500 ‘giant’ derived fromasinglefield-Ghawar in Indeed, as much as 7% of production much asonefifthfromonly10fields. one quarterfromonly20fieldsandas production derived from only 110 fields, 2007 approximately halfofglobal oilfields in production worldwide, but in estimates thattherearesome70,000 1995 2000 2005 2010 Chap1:UKERC 28/09/2009 11:47 Page 7 Page 11:47 28/09/2009 Chap1:UKERC contributory factor. But the price rises in the first decade of the 21st century have not reversed the production decline. 4 Thecollapseinoilpricesthe late1990sandthesubsequentreductioninexploratory drillingmay alsohave beena Source: model of Stylised aregionalFigure peak in 1.5 oil production produced. Sinceitalsooccurswhenthere the resourcesinregionhave been the peak occurs when around one third of in production.Undertheseassumptions, relatively early, leadingto a regionalpeak from thelargefieldsthatwerediscovered compensate for the decline in production relatively late becomes insufficient to the small fields that were discovered some point,theadditionalproductionfrom than the previous. The result is that, at size, with eachfieldbeing10%smaller fields are developed in declining order of production eachyear. Itisassumedthat field, withonefieldbeingbroughtinto represents the production from a single (Bentley, help of a simple model (Figure 1.5) oil resourcecanbeillustrated withthe The implications of these features of the 1990 (Robelius, 2007). one tenthhave beendiscovered since more thanfiftyyears ago, whilelessthan

Annual Production Bentley, et al., 2000). Here, each triangle et al.(2000) Time declining size of newly discovered fields. peak in1999was largelydriven bythe safety work that followed, the subsequent Piper Alpha disaster and the remedial the mid-1980swas linked inparttothe Figure 1.6.Whilethepeakinproduction UK continental shelf (UKCS), shown in illustration istheproductionhistoryof complications intervene. Agood frequently thecase,althoughnumerous Empirical observation suggests that this is developed relatively early (Stark,2008). assumed that the larger fields are brought inproduction, distribution and the rate at which fields are of individualfields,thefieldsize assumptions abouttheproductionprofile This simplemodelisrobustto peak may not necessarily be anticipated. fields arecontinuingtobediscovered, the (R/P) ratios arerelatively stableandnew producing fields, reserve to production are largequantitiesofreserves inthe provided it is 4 7 Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Chap1:UKERC 22/09/200909:51Page8

8 Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production • to: The specific objectives of this report are the report Objectivesandscopeof 1.4 this question is far from straightforward. political factorsthataffectoilproduction, of geological,technical,economicand oil resources.Butgiven thecomplexmix is knownaboutthesize anddistributionof production canbeanticipatedgiven what whether theregionalandglobalpeaksin The peak oil debate therefore hinges upon data areavailable inthepublicdomain. few countries where accurate field-by-field analysts, despite the UK being one of the The peakwas notanticipatedbymany Source: field by UKCS the in production Oil 1.6 Figure ‘peak oil’ debate; methodological issues relevant to the clarify theconceptual, definitional and

DECC Production (mb) 1,000 100 200 300 400 500 600 700 800 900 0 1975 1980 1985 1990 Year • forecasts offuturesupply (includingthe this is likely to occur. While ‘optimistic’ a peakandultimately decline, butwhen global supplyofconventional oilwillreach the maincontroversy isnotwhether the The period to 2030 was chosen because • • • • the size of oil resources and to of different approaches to estimating identify thestrengthsandweaknesses physical depletion before 2030. con assess therisk of the global supplyof gaps; and identify themainresearchanddata reasons fortheirdifferentconclusions; forecasts ofoilsupplyandidentifythe summarise andcomparecontempor their relativ associated with key issues and assess highlight thedegreeofuncertainty forecasting future oil supply; 1995 v entional oil being constrained by 2000 e importance; 2005 ary Chap1:UKERC 22/09/200909:51Page9 Box 1.2 PoliticalBox 1.2 and ec ‘physical’ factorsthatmay restricttherate The reportalsofocusesonthebroadly discussed furtherinSection2. The definitions used in this report are used bythoseconcernedaboutpeakoil. distinction ismeaningful,itroutinely analysts question whether such a others amovable one. Whilemany definitions implyingafixed boundaryand method ofextraction, withsome economic viabilityofextractionthe and/or resource, thelocationof the physical characteristics ofthe sources is variously drawn on the basis of conventional andnon-conventional report. Theboundarybetween importance, itisbeyond thescopeofthis non-conventional fuelsisofcritical fashion. Whiletheeconomicpotentialof unable tosubstituteinasufficientlytimely supply if ‘non-conventional’ sources are associated with a peak in liquid fuels peak inconventional oilsupplywillonlybe resource baseissubstantiallydepleted.A the period to 2030 and because its bulk oftheglobalsupplyliquidfuelsin because thisisanticipatedtoprovide the The reportfocuseson occurred. arguing thatthepeakhasalready peak before2020withseveral analysts most ‘pessimistic’forecastsanticipatea IEA) do not anticipate a peak before 2030, • • disruptions (e.g. from weather events or terrorist activity). expand thatcapacityand theconsequentvulnerability oftheglobalmarket tosupply Erosion ofexcess productioncapacityinOPECcountries, thelimitedincentives to at the rate forecast by the IEA and other bodies. Eastern andNorthAfricancountrieswhomay lacktheincentive toexpandproduction Increasing concentration of reserves and production in a small number of Middle onomic constraints on expanding global oil supply conventional oil , are priorities for future research. mitigating such shortages, although both feasibility of different approaches to consequences ofsupplyshortagesorthe Neither doesitassessthepotential their importancecannotbeoverstated. as thoseidentifiedinBox 1.2,although the broaderriskstosupplysecurity, such debate. Thereportdoes since theyarethefocusofpeakoil this reportfocusesprimarilyonthelatter and difficult to separate. Nevertheless, ‘below ground’risksareinterdependent In practice, these‘above ground’and political conditions prove more favourable. the near-term, even if economic and also however, iswhetherphysical depletionis constraints (Box 1.2). What is disputed, in theabsenceof‘below-ground’ challenge toglobalenergysecurity, even several ofthesemay poseasignificant political andgeopoliticalfactors upon a much wider range of economic, greatly contested. Oil supply also depends overcome the physical constraints is factors, theextenttowhichthesecan by economic, technical and political produced. While each of these is mediated sequence inwhichtheyarediscovered and distribution ofthosefieldsandthetypical individual fields,theskewed size including theproductionprofileof at whichconventional oilcanbeproduced, likely to constrain global production in not investigate (contd) 9 Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Chap1:UKERC 22/09/200909:51Page10 10 Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Sources: is As withallTPA assessments,theobjective audience. more accessibletoanon-technical ‘opposing camps’andmake theissues constructive dialogue between the for the diverging views, encourage more evidence could help to clarify the reasons that a careful review of the relevant potential importance.Itwas considered relative neglect of the issue and its views and the mismatch between the topic, theexistenceofwidelydiverging persistence ofcontroversy aboutthis research questions.TheGroupnotedthe the TPA function addresses policy-relevant sector. TheGroup'sroleistoensurethat government, academia and the private comprised ofseniorenergyexpertsfrom by theTPA AdvisoryGroupwhichis The topic for this assessment was selected was conducted Howtheassessment 1.5 thorough reviewofthe currentstateof depletion, but instead to provide a • • • • • not Continuing political instability in key producing regions such as Nigeria and Ir of equipment, materials and skilled manpower. Increased in products. increasing proportion of heavy and sour crudes and the changing demand for oil Refinery bottlenecks and the mismatch between available processing capacity, the workforce. Growing skills shortage resulting in part from the age profile of the oil industry resources and potentially lead to less efficient exploitation of those resources. ma The riseof‘resourcenationalism’inseveral producingcountries(e.g.Russia) which Jesse andvan derLinde(2008);Stevens (2008);IEAGately(2004) y further restrict the ability of independent oil companies (IOCs) to access low cost to undertake new research on oil v estment inexploration andproductionbeingtaken upbytherisingcost Note circulated to key stakeholders. This established (Annex1) andtheScoping An AdvisoryGroupfor theprojectwas designed with these issues in mind. objectives oftheassessmentwere methodological approaches.The appropriate use of different and the disputes over the validity and uncertainty over the data that is available; lack ofaccesstorelevant data;the interpretation ofreserve estimates;the definitions, suchastheappropriate economists); the confusion over key (e.g. geologistsandpetroleum between differentdisciplinaryperspectives controversy including:theconflict This identifiedseveral sources of assessment could make (Sorrell, 2008). potential contribution that a TPA on globaloildepletionandidentifiedthe Scoping Note model, theassessmentbeganwitha other fields(Box 1.3).Following this techniques prominentinmedicineand informed by the systematic review knowledge. Thegeneral approachis that summarised the debate aq. Chap1:UKERC 22/09/200909:51Page11 modelling teams.The agreedapproach direct contactswitha number ofenergy supplied by IHS Energy, together with analysis usingacountry-level database It was also agreed to include some data the inclusion of a broader range of studies. nature ofthecurrenttopicnecessitates priority to peer-reviewed literature, the evidence. WhileallTPA assessmentsgive to begiven todifferentsourcesof assessment, includingtherelative weight appropriate scopeandfocusofthe led tofurtherrecommendationsonthe OverviewBox 1.3 of the TPA approach • • • • • • • • • • The process carried out for each assessment includes the following components: to any narrowlydefinedmethodortechnique. UKERC hasthereforesetupaprocessthatisinspiredbythisapproach,butnotbound criticised forexcessive methodologicalrigidityinsomepolicyareas(Sorrell,2007). of challengesfortheapplicationsystematicreviewsandapproachhasbeen upon high quality studies when drawing conclusions. Energy policy presents a number approach includeexhaustive searchingoftheavailable literature andgreaterreliance encourage higherresearch standardsandidentifyresearchgaps.Corefeaturesofthis more robustevidenceforpolicymakers andpractitioners, avoid duplicationofresearch, policy andpractice, including thepractice ofsystematicreviews.Thisaspirestoprovide The TPA approach is informed by a range of techniques referred to as evidence-based Peer reviewoffinaldraft. Drafting ofsynthesisreport. Expert feedbackontechnicalreports. Review anddrafting oftechnicalreports. Categorisation and assessment of evidence. Systematic searches ofclearlydefinedevidencebaseusingkeywords. Stakeholder consultation. Convening an Expert Group with a diversity of opinions and perspectives. Establishment ofaprojectteamwithdiversity ofexpertise. Publication of Scoping Note and Assessment Protocol. individual tasks.The full resultsofthe were assigned to who consultants mix of in-house staff and external A projectteamwas formed comprisinga relevance andnotallofwhichwereused. of which was categorised and assessed for This revealed over 900 publications, each the assessmentquestion(SeeAnnex1). search forreportsandpapersrelatedto The assessment began with a systematic (Sorrell andSpeirs,2008) was setoutinan Assessment Protocol 11

Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Chap1:UKERC 22/09/200909:51Page12 12 Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production highlights the important distinction estimates may best beinterpreted.It reserves and discusses how reserve some basicdataonproductionand sources relevant tothistopic,presents introduces the key definitions and data • • • • namely: 5 http://www.ukerc.ac.uk/support/tiki-index.php?page=TPA%20Overview” Section 2 ( follows: This synthesis report is structured as Structureofthereport 1.6 download from the UKERC website Technical Reports assessment arereportedinseven further research. the policy implications and priorities for findings oftheassessmentandhighlights The present report presents the main • • • Technical Report4: importance of reserve growth Technical Report 3: interpretation ofreserv Technical Report 2: issues Technical Report1: global supplyforecasts Technical Report7: forecasting futureoilsupply Technical Report6: resources depletion r estimating ultimatelyrecov Technical Report5: ates Reading the fuel gauge that areavailable to Data sourcesand Decline rates and e estimates Comparison of Definition and Nature and Methods of Methods of erable 5 ) , methods, the level of confidence that can the strengthsandweaknesses ofthese significantly different. It again assesses fiercely debatedandtheresultsareoften relative meritsoftheseapproachesare forecasting futureoilproduction.The compares thevarious approachesto Section 5( the results. level of confidence that can be placed in weaknesses of these methods and the that arenot.Itassessesthestrengthsand peak oil and regularly criticised by those are widelyusedbythoseconcernedabout production anddiscovery. Suchtechniques extrapolation of historical trends in a region, focusing in particular on the ultimately recoverable resources(URR)in methods available forestimatingthe Section 4( field or region can be produced. how rapidly the remaining resources in a field may beexpectedtodecline;andd) production fromdifferentcategoriesof continue inthefuture;c)howrapidly the time andwhetherthismay beexpectedto the size of fields are found to grow over different sizes offield;b)why estimatesof in aregionaredistributedbetween controversy, namely: a) how oil resources and which continue to be a focus of important roleinshapingfutureoilsupply discusses fourissuesthatplay an Section 3( examining oildepletion. available dataisnot‘fitforpurpose’in estimates andconcludesthatthepublicly with thedifficultiesinaggregatingthose probable (2P)reserve estimates,together between proved (1P) and proved and Looking beneath Looking ahead Enduring controversies ) describesand ) examinesthe ) Chap1:UKERC 22/09/200909:51Page13 for global supply forecasts. over timeandexaminestheirimplications shows howsuchestimateshave increased that have updated their estimates. It Survey (USGS) and some recent studies particular ontheworkofUSGeological global URRofconventional oil,focusingin examines the various estimates of the Section 6 ( debate. implications ofthisforthepeakoil be placedintheresultsand How muchdowehave? ) implications. highlights some important policy identifies the main research needs and conclusions fromtheassessment, Finally, Section 8 summarises the key near-term peak in global production. draws some conclusions on the risk of a peak productiondeclinerate. Itthen conventional oil and the aggregate post- assumptions regarding the URR of the importanceofexplicitorimplicit forecasts of global oil supply. It highlights evaluates fourteencontemporary Section 7 ( Possible futures ) comparesand 13

Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Chap1:UKERC 22/09/200909:51Page14 14 Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Chap2:UKERC 22/09/200910:11Page15 7 Sandrea and Sandrea (2007) estimate that 29% of proved oil reserves are light crude, 57% medium and 14% heavy. lightest. 6 TheAPIscaleranges from10°(equaltothedensity ofwater) fortheheaviest conventional oilstomorethan45°for the while Section2.5looks more closelyatthe factors thatinfluencereportedfigures produced and highlights the multiple 2.3 examines how reserve estimates are information on global oil supply. Section reserves andprovides somebackground and data onproduction sources of depletion. Section2.3comparesthe estimates inexaminingresource ‘proved andprobable’(2P)reserve the relative usefulnessof‘proved’ (1P)and production anddiscoveries andassesses some ofthemeasuresassociatedwithoil ‘conventional oil’. Section2.2introduces working definition of the term hydrocarbon liquids and proposes a Section 2.1 clarifies the different types of Reports 1 information is contained in of thedifficultiesthatmay result.Further different data sources and highlight some identify thestrengthsandweaknessesof clarify someoftherelevant definitions, set ofnumbers.Thissectionseeksto substantial riskinrelyingonany particular both fuelsthepeakoildebateandcreates in agiven year. Theresultingconfusion amount ofoilproducedbyagiven country level, such as uncertainties over the they alsoapplyatamuchmorebasic the oilreserves ofOPECcountries.But greatest where they matter most, namely data thatisavailable. Thedifficultiesare corresponding uncertaintysurroundingthe third-party auditingofthatdataandthe reliable data,thefrequentabsenceof of relevant variables, the paucity of are createdbytheinconsistentdefinitions which tomeasureoildepletion.Problems The world lacks a reliable gauge with Readingthefuelgauge-measuringoilsupplyandresources 2. and 2. Technical has a higher ratio of hydrogen to carbon dense oil. Gravity, withhigherAPIindicatingless of AmericanPetroleum Institute(API) classified by its density, measured in units sulphur and metals. Crude oil is normally includes unwanted impuritiessuchas one regiontoanotherandtypically The compositionofcrudeoilvaries from crude oil, defined as: data. Themostimportantsubcategoryis not always distinguishedintheavailable between themarenotfixed andtheyare commonly defined,buttheboundaries sources. Several subcategoriesof oilare hydrocarbons derived from a range of a heterogeneousmixofliquid What iscommonlytermed‘oil’represents WhatisOil? 2.1 making thesedistinctions. different sourcesusecriteriain 30°API) and heavy (<20°API) crude, but between light(>30°API),medium(20°- Section 2.7 concludes. harmonisation in reserve reporting. schemes and the prospects for greater currently used reserve classification estimates. Section2.6reviewsthe difficulties of aggregating reserve treatment of uncertainty and the 2006). temperature and pressure.” (EIA, remain liquidatatmospheric underground reservoirs and which exist in liquid phase in natural “...a mixtureofhydrocarbons that 6 It iscommontodistinguish 7 Light crudeoil 15

Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Chap2:UKERC 22/09/200910:11Page16 16 Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production associated gas at oil wells is normally gas. Condensate produced from the oil wellsalsoproduceassociatednatural temperatures andpressures)whilemany condense to liquids at surface of gas wellsalsoproduceliquidsintheform complicated by the fact that many natural Data oncrudeoilproductionis those liquids. somewhat fromdataontheproductionof changes andoperating fuel)sodiffers into accountthesegains(aswellasstock Data onthesupplyofliquidfuelstakes (2008). Refinery gainsareexcluded. 10 Globalproductionisderived byadding0.6 mb/dofbiofuelstothe82.3liquidfuelsproductionreported bytheIEA 9 We estimate that, on an energy equivalent basis, the current ratio of NGLs to gas production is approximately 15% heavy fueloil;andspecialities suchasbitumen,lubricantsandcoke. 8 Including:lightproductssuchas gasolineandnaphtha;middledistillatessuchaskerosene anddiesel;residualfuelssucha hundred differentproducts, Crude oilcanberefinedintomorethana Texas Intermediate. standard referencecrudes,suchasWest emissions. Pricingisnormallybasedupon regulatory restrictions on sulphur increasing demand for light products and crudes trade at a premium, due to both sulphur crudes‘sweet’. Lightandsweet andlow crudes beingtermed‘sour’ by its sulphur content, with high sulphur and kerosene. Crudeoilisalsoclassified more valuable productssuchasgasoline refine andallowinggreaterproductionof atoms, makingiteasiertotransport and 2007) beingreferredtoas refinery, withthedifference(2.1mb/din exceeds thevolume ofcrudeinputstoa met. The volume of products normally the demandforaparticularproductcanbe crude supplyprovides noguarantee that processes. Asaresult,thevolume of the crude and the available refinery depending upon both the composition of condensate (i.e. hydrocarbons that 8 refinery gains. with themix that gas. production and the average ‘wetness’ of opportunities available for expanding gas be determined by the demand for gas, the production, thefuturesupplyofNGLswill they are a by-product of natural gas at natural gas processing plants. Since as butaneandpropanewhichareliquified condensate and lighter hydrocarbons such moderate pressure.NGLs compriseboth turned intoaliquidwiththeapplicationof and pressures, or can be relatively easily are eitherliquidatnormaltemperatures (NGLs). Thesearelighthydrocarbons that included inthedatafor while thecondensatefromgaswellsis combined withcrudeoilinproductiondata crude oil but the US Energy Information classify Venezuelan extra heavy oilas different ways: forexample,theIEA subdivide these different liquids in Commonly useddatasourcesgroupand biofuels (see Box 2.1 and Figure 2.1). liquids (CTLs), gas-to-liquids (GTLs) and of extra heavy oil,oilsands,coal-to- remaining 2.2mb/d(2.7%)was madeup 10.5 mb/d,or12.7%ofthetotal.The mb/d). global productionofliquidfuels(82.9 per day (mb/d) in 2007, or 84.6% of the condensate averaged 70.2millionbarrels global production of crude oil and According totheIEA(2008;2009b), transport fuels. only aportionofNGLscanbeblendedinto constraint onfutureNGLproduction.But physical depletion should be less of a developed than crude oil resources, so 10 Production of NGLs averaged 9 Gas resources are less natural gasliquids s Chap2:UKERC 22/09/200910:11Page17 Source: ProposedFigure 2.2 classification of hydrocarbon liquids Source: Breakdown production fuels of2008liquid 2.1 Figure sands, oil shales. Non-conventional oil: and natural gasliquids (NGLs). Conventional oil: following terminology(Figure2.2): liquids’. Wherepossible,weusethe Administration (EIA) classifies it as ‘other IEA (2008) IEA (2008) crude oil, condensate CONVENTIONAL Condensate extra-heavy oil,oil Crude Oil NGLs OIL Condensate, 70.2 Crude oiland ALL OIL Non Conventionals, ALL LIQUIDS NGLs, 10.5 Extra HeavyOil Oil Sands Oil Shale biofuels. All-liquids: oil. All-oil: biofuels. conventional oil plus GTLs, CTLs and Non-conventional liquids: 1.98 CONVENTIONAL LIQUIDS NON- conventional andnon-conventional all-oil plusGTLs,CTLsand Biofuels GTLs CTLs Oil Sands,1.2 Biofuels, 0.6 GTLs, 0.05 CTLs, 0.13 non- 17

Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Chap2:UKERC 22/09/200910:11Page18 18 Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production to contribute less than 10% in 2030 (8.8 2007 (1.6 mb/d) and the IEA expect them than 2% of global all-liquids production in conventional liquids accounted for less (IEA, 2008). Excluding biofuels, non- mb/d ordoubletheproductionin2007) with NGLs providing another fifth (~20 mb/d) ofgloballiquidsproductionin2030, still provide aroundthreequarters(75.2 projects that crude oil and condensate will increase inthefuture.However, theIEA over timeandisexpectedtocontinue liquids to total production has increased The contributionofnon-conventional Box 2.1 Non-conventional liquid hydrocarbon fuels fuels hydrocarbon liquid Non-conventional Box 2.1 available. Major GTL projects are currently underway in Qatar. used andtheeconomics maybe morefavourable whennopipeline facilities are hydrocarbons and water. The product mix depends upon the temperature and catalyst monoxide and hydrogen followed by catalysed chemical reactions to produce liquid Tropsch process. This involves steam reforming of natural gas to produce carbon Gas-to-Liquids (GTLs) production over the next 20 years. environmental termsandis unlikely tomake significantcontributiontogloballiquids 2008). But the resource remains costly to exploit in economic, energy and could both improve the recovery factor and reduce environmental impacts (Brandt, Shell isdeveloping anin-situprocessinvolving underground electricalheatingwhich crushing therock,heatingtoahightemperature, drivingoffavapour and distilling. a solid mixture of hydrocarbons from which liquids can be extracted by mining and Oil shale China, Romania, Nigeria and the US. global productionderives fromCanada,butlargedepositsarealsofoundinRussia, directly orupgraded toasyntheticcrudetransportable bypipeline( being developed toaccessdeeperdeposits.Theresultingbitumenmay bemarketed production is through open-cast mining, but in-situ methods using steam injection are Oil sands China andRussia. the OrinocobeltinVenezuela, butlargedepositsarealsofoundinotherregionssuchas steam injection,whichiscapitalandenergyintensive. Mostcurrentproductionisfrom this definitionisnotconsistent.Becauseofitshighviscosityithastobeproducedusing Extra heavyoil is afine-grained sedimentaryrockcontainingsignificantamountsofkerogen - (or is commonlydefinedasoilhaving anAPI gravity lessthan10°,although tar sands are derived through the liquefaction of methane using the Fischer- ) are sandstone impregnated with bitumen. Most current (Hirsch, rapid expansion of production challenging and environmental constraints make a However, numeroustechnical,economic liquids may beexpectedtoincrease. absolute contribution of non-conventional than the IEA anticipates, the relative and mb/d). Ifcrudeoildepletesmorerapidly 6% of projected global production. mb/d by2030whichrepresentslessthan Canadian oilsandscoulddeliver only5 ‘crash programme’ todevelop the Söderbergh et al et al. (2007) estimate that a ., 2005). For example, syncrude). Most Chap2:UKERC 22/09/200910:11Page19 Box 2.2 Units ofMeasurement Units Box 2.2 basis (barrelsofoil)butthiscanbe are commonlyreportedonavolumetric Oil production,discoveries andreserves 2.2.1 Production andreserves discovery Measuresofproduction, 2.2 always clearwhichdefinitionisbeingused. released bycondensing thewater generated duringcombustion.Unfortunately, itisnot with the7-9%difference betweenthetwocorrespondingtoheat thatcouldbe into asinglemeasure,butheatcontentmay eitherbemeasuredonagrossornetbasis content of a barrel of oil (6.1 GJ). This is commonly used to combine oil and gas data definition, whichisaunitofenergymeasure correspondingtoastandardisedheat 1700kWh for conventional oil. This forms the basis of the barrel of oil equivalent (boe) The gross heat content of a barrel of oil is similarly variable, but typically lies around estimates of the volume or weight of oil production. crude oil, but the use of different conversion factors can lead to widely different (Karbuz, 2004).Many analystsuseastandardconversion factorof7.33b/tonnefor 8.0 barrels per tonne. For NGLs, the corresponding figures are 10.0 and 13.5 b/tonne depends on the source and hence composition of the oil and can vary between 6.0 and Oil may alsobemeasuredonaweight basis inmetrictonnes.Theweightofabarrel commonly used. (thousand barrels),mb, MMbbl(millionbarrels)andGborGbbl(billionarealso (m Oil istypicallymeasuredonavolumetric basisineitherbarrels(borbbl)cubicmeters processes using non-food feedstocks offer greater promise in the longer term. crops, orbiodieselderived fromseedoilsorrecycled oils.Secondgeneration cellulosic either ethanolproducedthroughtheyeast fermentationofsugarorstarch-richarable Biofuels carbon emissions. development inChinaandtheUSbutamajordrawback istheirhighcapitalcostsand dissolution ofcoalinasolvent followedbycatalyticcracking. CTLplantsareunder Tropsch process. Research is also being carried out on direct conversion through Coal-to-Liquids (CTLs) 3 ). A barrel is equivalent to 42 US gallons or 159 litres. The abbreviations kb or Mbbl are transport fuelsderived frombiologicalsources.Commonlythisconsistsof are derived throughthegasificationofcoalfollowedbyaFischer- data. 1.1), but are not visible in the available investment may alsobeimportant (Box Changes in the energy return on from agiven volume ofliquidssupply. the formerwillreduceenergyavailable volume than crude oil, so a shift towards have a lower energy content per unit misleading (Box 2.2).For example,NGLs (contd) 19

Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Chap2:UKERC 29/09/2009 09:59 Page 20 Page 09:59 29/09/2009 Chap2:UKERC 20 Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Note: Source: production oil in trends Global 2.3 Figure year to year), with one quarter of (albeit withconsiderable variation from has grownatanaverage of1.5%/year (IEA, 2008)Since1995,globalproduction production was approximately 1128Gb. Gb/year, while global cumulative averaged 80.7mb/din2007,or29.5 conventional oil (as defined above) billion barrels(Gb).Globalproductionof cumulative productionismeasured in barrels per year (Gb/year) while million barrelsperday (mb/d)orbillion global level is normally measured in Oil productionatthecountry, regionalor difficulties account for some of the discrepancies between different data sources. aggregate estimatesofthevolume, weightorheatcontentof oil production.These differences in the assumed conversion factors can make a very large difference to dissimilar liquids through the use of boe. Karbuz (2004) has shown how relatively small the useofinconsistentandsometimesinaccurate conversion factorswhenaggregating aggregation andinterpretationofproductionreserve data.Thisiscompoundedby The variationsignificant impact on the oil hasa and densityofcrude inthecomposition Includes crudeoil,condensate,NGL, LPG,heavy oilandsyncrudefromsands.

IHS Energy Annual Production (Gb) 10 15 20 25 30 35 0 5 8016 8010 9014 9018 002020 2000 1980 1960 1940 1920 1900 1880 1860 1840 Annual Production Year Cumulative AnnualProduction capita. currently consumes around 25 barrels per (Figure 2.4).For comparison,theUS barrels perpersonsincethemid-1980s 1979 andhasremainedaround4.5 peaked at 5.5 barrels per person in per capita basis, global oil production produced (226Gb)ineightyears. Ona months and as much as the US has ever UK hasever produced(24Gb)inonlyten world usesasmuchconventional oilasthe 2.3). At current rates of production, the last tenyears and59%since1980 (Figure cumulative productionoccurringover the 0 200 400 600 800 1,000 1,200

Cumulative Production (Gb) Chap2:UKERC 29/09/2009 09:59 Page 21 Page 09:59 29/09/2009 Chap2:UKERC Source: resources Oil and reserves Figure 2.5 distinguished from conditions. Reserves must be feasible to extract under defined technically possibleandeconomically known fields which are considered to be Oil reserves arethosequantitiesofoilin 2.2.2 Reserves Note: Source: trends Global in per capitaFigure oil production 2.4 Includes crude oil, condensate, NGL, LPG, heavy oil and syncrude from oil sands. McKelvey (1972) IHS Energy;USCensusBureau Commercial Commercial Barrels per capita 0 1 2 3 4 5 6 1940 Sub- 1950 resources P2 3P 2P 1P Reserves Discovered 1960 which are Resources 1970 Year (Figure 2.5). illustrated bytheso-called‘McKelvey Box’ fields. Thisdistinctioniscommonly extract aswellthoseinundiscovered not consideredeconomicallyfeasibleto including thoseinknownfieldswhichare the total quantities estimated to exist, 1980 Undiscovered 1990 2000 2010 21

Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Chap2:UKERC 22/09/200910:11Page22 22 Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production smaller thantheBPStatisticalReview’s forUAE,Libya andKuwait butgreaterforRussia and Saudi Arabia. 11 Thishideslargediscrepancies in theestimatesforindividualcountries.For example,theIHSEnergyestimatesaresignific Source: trends Global in provedFigure oil reserves 2.6 reduced by addition, both may be either increased or affected bymergersandacquisitions.In while company reserves willbefurther increased bythediscovery ofnewfields reserves willbereducedbyproductionand in differentways (see below). Regional oil (3P) althoughthesetermsareinterpreted confidence, namelyproved commonly quoted to three levels of oil production.Reserve estimatesare resource extraction and theeconomicsof features of an oil field, the technology of and assumptions about the geological uncertain sincetheyrelyuponinformation Estimates ofoilreserves areinherently proved, probable and possible proved andprobable reporting practices. Changesinreserves economic conditions or changes in extraction technology, variations in better geologicalunderstanding,improved estimates of known fields as a result of BP (2008) Global Proved Reserves (Gb) 1000 1200 1400 200 400 600 800 0 1975 revisions 1980 reserves (2P)and to thereserve reserves 1985 reserves (1P), 1990 Year public domain. the data is not readily available in the trends in 2P reserves are less clear, since increasing production(Figure2.6).Global steadily since the mid-1980s, despite reported 1P reserves have increased are discussedfurtherbelow. Global The possiblereasonsforthisdiscrepancy approximately thesame(IEA,2008). industry source(IHSEnergy)theyare that this,butaccordingtoanauthoritative probable (2P)reserves shouldbelarger production. Inprinciple,globalproved and or slightlymorethancumulative end of2007wereapproximately 1240Gb, (1P) reserves ofconventional oilatthe According to BP (2008), global proved additions toproduction. replacement ratio headline financial indicator is the could beeitherpositive ornegative. A reserve additions, although the changes over timearecommonlyreferredtoas 1995 2000 , or the ratio of reserves 2005 2010 antly 11 Chap2:UKERC 22/09/200910:11Page23 discovery estimates for all intervening years. The treatment of newly discovered fields is the same in both cases. estimates, theserevisionsarebackdated totheyear inwhichtherelevant fieldwas discovered andhenceincreasethecumulati Note: revisions reserve versus backdated of estimates Current cumulative discoveries – treatment ofFigure 2.7 referred toas known fields. The latter is commonly revisions tothereserve estimatesfor and may beeitherincreasedorreducedby increased bythediscovery ofnewfields However, cumulative discoveries willbe another (cumulative production). resources from one category (reserves) to production since this merely transfers discovery estimateswillnotbechangedby Unlike reserve estimates, cumulative do notcoincide. reserve definitions and coverage of liquids to be around 2370 Gb, although their cumulative discoveries of conventional oil and IHSEnergyestimateglobal discoveries. Both the BP Statistical Review estimate cumulative 1P, 2Por3P the dataavailable, itmay bepossibleto cumulative discoveries reserves is commonly referred to as The sumofcumulative productionand Cumulativediscoveries 2.2.3 With currentestimates,reserve revisionsincrease thecumulative discovery estimatesinthecurrent year. With backdated reserve growth Current estimates of cumulative discoveries Current estimatesof cumulative Backdated estimatesofcumulativediscoveries . Depending upon field discovery Year of since Time estimate was revised and therefore should become ‘available’ for production until the first approach is that the reserves did not discovered (Figure2.7).Thelogicofthe in which the relevant fields were others adjustment tothedataforearlieryears, year inwhich theyaremadeandmake no sources record reserve revisions in the discovery estimates.Whilesomedata common practice of A major source of confusion is the Section 3. oil supplyandisdiscussedfurtherin Reserve growth is a critical issue for future recovered fromthefieldorregion. estimates of what will ultimately be growth’ sincewhatisgrowingarethe alternative termsis‘ultimaterecovery being depletedbyproduction.An growth accurate term is rather than downwards. However, a more estimates arenormallyrevisedupwards , sincereserves arecontinually reserve revision backdate Year of the revisionstoyear backdating cumulative discovery cumulative ve 23

Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Chap2:UKERC 28/09/2009 11:52 Page 24 Page 11:52 28/09/2009 Chap2:UKERC 24 Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production the discovery of California to the landing of Columbus in the Bahamas in 1492. neither knownnoraccessibleuntil theyweredrilledin1991.For Mills(2008),backdatingtheserevisionsto1971islike ascr 12 For example,theworld’s largestgasfield(inQatar)was discovered in1971butthereserves intheIranian sectionwere • may mean: The term 2.2.4 Discoveries appreciated. difference betweenthemisnotalways approaches have their merits, but the recovered from that field. Both of these that time as well as what will ultimately be indication ofwhatwas ‘actually’foundat backdating provides amoreaccurate was discovered many years earlier, so the reservesfield that arecontainedina The logic of the second approach is that only appear at the time of the revision. Note: Source: trends Global in backdatedFigure discoveries and 2.8 cumulative discoveries particular timeperiod;or fields thatarenewlydisco the recoverable resourcescontainedin Includes crude oil, condensate, NGL, LPG, heavy oil and syncrude. Based upon backdated 2P reserve estimates. IHS Energy

Discoveries (Gb) discovery 10 20 30 40 50 60 70 80 0 8018 9012 9016 9020 2020 2000 1980 1960 1940 1920 1900 1880 1860 is ambiguous since it v ered withina Backdated Cumulative2PDiscoveries Backdated 2PDiscoveries(3yrmovingaverage) 12 Year reserv These arenotnecessarilythesame,since • over a 35 year period from 1946 through world's conventional oilwas discovered definition. This suggests that most of the ‘discoveries’ based upon the second together with a time series of backdated backdated cumulative discoveries, Figure 2.8 shows a time series of existing fields. distinguished fromreserve growthat contained innewlydiscovered fieldstobe data sourcesdonotallowtheresources which definition is being used and most found. Unfortunately,always it isnot clear second definitioneven ifnonewfieldsare contribute to‘discoveries’ underthe from one period to the next. the changeincumulative discoveries e growthatexistingfieldswill 0 500 1,000 1,500 2,000 2,500

Cumulative Discoveries (Gb) ibing Chap2:UKERC 22/09/200910:11Page25 ‘heights’ and shapes for the backdated discovery data (e.g. Campbell, 2002b). for 2006 include only one year. This helps explain why comparable graphs published at different times have slightly different estimated on a consistent basis. For example, the estimates for 1957 include 50 years of reserve growth, while the estimates While discoveries have fallen over time, the graph is potentially misleading since the discoveries for different years have not been Note: Source: discoveries backdated and production in trends Global 2.9 Figure are distributed between the earlier years since revisions made in the current year This cannot be established from Figure 2.9 fields (BP, 2008; Stark and Chew, 2005). revisions to the estimates for existing to acombinationofnewdiscoveries and production over most of this period owing additions have exceeded annual Indeed, annual1Pand2Preserve than wehave addedtoreservesyear. each not mean that we have produced more oil have discovered each year. But this does 1980 wehave producedmoreoilthanwe This widelycitedgraph suggeststhatsince with global‘discoveries’ asdefinedabove. Figure 2.9 compares global production century. with an upturn around the turn of the and have fallensteadilysince,although to 1980.Discoveries peaked inthe1960s Includes crudeoil,condensate,NGL, LPG,heavy oilandsyncrude.Discoveries baseduponbackdated2Preserve estimates. IHS Energy Volume (Gb) 10 20 30 40 50 60 70 80 0 1900 1920 Backdated Global2PDiscovery(3yrmovingaverage) Global Production 1940 1960 Year 2008; USGS, 2000). off date of 2025 (Campbell and Heapes, Geological Survey (USGS)imposesacut- while awidelycitedstudybytheUS relatively small volumes ofproduction estimates in order to remove a ‘long tail’ of imposes a cut-off date of 2070 for his URR shorter timeframe. For example,Campbell estimates of regional URR relate to a ends. Inpractice, however, many when productionbeginstoitfinally from a field or region over all time – from estimated tobeeconomicallyextractable represent theamountofoilthatis Ultimately recoverableresources Ultimatelyrecoverable resources 2.2.5 somewhat misleading. relevant fields.Asaresult,thegraph is according tothedate of discovery ofthe 1980 2000 2020 (URR) 25

Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Chap2:UKERC 22/09/200910:11Page26 26 Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Figure 2.10 Components of ultimately recoverable resources (URR) resources recoverable ultimately of Components 2.10 Figure .. Measuringoildepletion 2.2.6 commonly termed the extractable from undiscovered fields - oil estimatedtobeeconomically growth at known fields and the volume of cumulative discoveries, futurereserve region, theURRrepresentssumof estimates offuturereserve growth.For a the sum of cumulative discoveries and For individual fields, the URR represents resources areavailable fromanumberof Estimates ofthesefourcategoriesoil larger. global URR for all-oil are several times illustrated inFigure1.2,estimatesofthe cumulative production through to 2007. As to 4300 Gb or two to four times conventional oilfallwithintherange 2000 uncertain. EstimatesoftheglobalURRfor categories areprogressively more yet-to-find (Figure 2.10). These three of reserves, future reserve growth and the that have yet tobeproduced,orthesum region are all the recoverable resources The remaining recoverableresources Discoveries Cumulative Remaining Resources yet-to-find Cumulative production Future reservegrowth (YTF). Yet tofind(YTF) for a Reserves 2.11) suggests that more than half of the The backdated2Pcurve fortheUK(Figure allow for future reserve growth. discovery curve must be ‘corrected’ to estimate theURR,cumulative time (Figure2.12).Hence,toaccurately discovery curve is likely to change over ofthecumulative future, the‘height’ more reserve growthmay occurinthe regional URR(seeSection4).Butsince commonly taken as an estimate of the an asymptoteof30-35Gbwhichis This curve appears to be trending towards diminishing size ofnewlydiscovered fields. exploration, largelyas aresultofthe curve suggestsdiminishingreturnsto variables fortheUK.The‘backdated2P’ Figure 2.11 which compares these two Laherrère, 1995). This is illustrated in Bentley, cumulative 1Pdiscoveries depletion thancurrentestimatesof provide moreinformationaboutresource estimates ofcumulative2Pdiscoveriescan that areused.Inparticular, be derived willdependuponthedefinitions data sources, but the information that can et al ., 2007; Campbell and Recoverable Ultimately Resources (URR) (Bentley, 2009; backdated Chap2:UKERC 22/09/200910:11Page27 production may be expected, even in the suggests that a near-term peak in Given the simple model of Section 1, this URR hadalreadybeenproducedby1999. of growth reserve Effect Figure 2.12 on backdated of estimates cumulative discoveries Source: current 1P TwoFigure 2.11 ways of presenting UK backdated cumulative 2P discoveries - versus IHS Energy;BP(2008) Cumulative Discoveries (Gb) 10 15 20 25 30 35 0 5 1970 discoveries Cumulative 1975 Cumulative discoveries 1980 Cumulative discoveries 2P CumulativeDiscoveries 1P CumulativeDiscoveries Year X Year X+N 1985 Year the ‘current1P’curve sincethisdoes not information ismoredifficult toobtainfrom peak actually occurred in 1999). The same absence ofdataonindividualfields(the 1990 between XandX+N New discoveries fields betweenXand Growth atexisting Reserve 1995 X+N 2000 Time 27

Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Chap2:UKERC 22/09/200910:11Page28 28 Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Box 2.3 PubliclyBox 2.3 sources available of global on data oil production and reserves sources (Bentley, in part to reliance upon different data over anear-term peakarethereforelinked analysts using current 1P data. Disputes forthcoming peakinproductionthen data should be better placed to forecast a field. Hence, analysts using backdated 2P backdated to the date of discovery of the because these estimates are not recoverable reserves (see below) because the 1P data underestimates for estimatingtheURR.Thisis and hencedoesnotprovide agoodbasis indicate diminishingreturnstoexploration reserves data. the relevant National Oil Companies (NOCs) (OPEC, 2008). OPEC also provides global and theseareusedby most otherdatasourcessincetheydonothave direct accessto OPEC Secretariat assessments (BP, 2008). sources, other data providers (particularly OGJ and World Oil) and independent energy statistics,gatheredthroughacombination ofprimarysources,thirdparty BP Statistical Review every year (WorldOil, 2007). World Oil other sources. issue of each year (Radler, 2007). The OGJ reserve estimates are commonly quoted by Oil andGasJournal(OGJ) stocks (EIA,2008b). Monthly whichprovides dataonregionalandglobalproduction,oildemand, trade and provides internationalstatistics.ThemostusefuloftheseistheInternational Petroleum (EIA) Energy InformationAdministration forecasting purposes,theIEAreliesonthirdpartyassessmentsofrecoverable reserves. Report (OMR)andtheannualWorld EnergyOutlook(WEO)(IEA,2008;2009b).For (IEA) International EnergyAgency publishes globaldataonproductionandreserves inAugust orSeptember et al., 2007). maintains itsownfiguresregardingMember’s reserve andproduction provides a comprehensive and widely quoted compilation of global provides globalproductionandreserve estimatesinthefinal publishes productiondatainthemonthlyOilMarket both and focuses mostly on the US oil market but also compares their coverage with that of the sources ofglobaldata,whileTable 2.1 Box 2.3summarisesthepubliclyavailable cost. commercial organisations at considerable The mostusefuldataisonlyavailable from same primarysources(Karbuz, 2004). despite compilinginformationfromthe generally provide different estimates, These usedifferentdefinitionsand country, regionalandindividual fieldlevel. production andreserves attheglobal, A variety ofsourcesprovide dataonoil figures Sourcesofdataandkey 2.3 Chap2:UKERC 22/09/200910:11Page29 ESdtbs rbbeNL,LG ev i,snrd)biofuels NGLs, LPG, heavy oil, syncrude) Proved and probable PEPS database IHS Energy Journal Oil andGas Review E i aktNo IEA OilMarket Report the InternationalPetroleum Encyclopaedia andseveral nationalgovernments. 14 Including the American Petroleum Institute (API), the American Association of Petroleum Geologists (AAPG), Baker Hughes provided the basis for an influential publication by Campbell and Laherrère (1995). 13 PEPSisthecontinuationofstatistics maintainedby‘Petroconsultants’its purchasebyIHSEnergy in1996.Theforme before Notes: Source: Petroleum Monthly No EIA International Source Comparison of globalTable 2.1 data sources on oil production and reserves on drilling activity and other relevant 180 countries,togetherwithinformation oil andgasproductionreserves for contract andprovides time-seriesdataon The PEPSdatabaseisavailable under (PEPS) database provided by IHS Energy. Petroleum Economics and Policy Solutions Report and OilMarket Statistical Bulletin OPEC Annual Magazine W Prov BP Statistical orld Oil 5. NGLsreportedseparately forOPEC. 4. Biofuels, CTLs, non-oil inputs to MTBE, orimulsion and other hydrocarbons (EIA, 2008b) 3. BiofuelsfromsourcesoutsideBrazil andUS. 2. Including biofuels, oil sands, oil shales, CTLs, GTLs and blending components such as MTBE. 1. NGLsreportedseparately forOPEC. Precise definitionandcoverage ofliquidsisnotalways madeclear. IEA (2009b),EIA(2008b),BP(2008),IHSEnergy, Oil&GasJournal,World OilMagazine eevs Grouping of liquids in Reserves Proved Proved Data production data production Data Pro ved ed • • • • • • • • • • • iud cueol odnae CTLs,GTLs, Liquids (crudeoil,condensate, rd i,cnest,snrd NGLs, CTLs, Crude oil, condensate, syncrude CTLs, GTLs, condensate, NGLs (aggregated) Crude oil, oil shale, oil sands, Other Biofuels Refinery gains Crude oil, condensate, NGLs rd i,cnest,NL,None Crude oil, condensate, NGLs, Refinery gains (aggregated) rd i,cnest,snrd NGLs, CTLs, Crude oil, condensate, syncrude NGLs reported separately Crude oil, condensate, NGLs non-conventional other liquids (aggregated) (aggregated) a variety ofsources, Uppsala University maintain a database on maintain adatabaseofUSfieldsand fields. Inaddition,NehringAssociates governments publish data on individual UK, NorwegianandAustralian variables. 4 , refinery gains 3 2 13 Regional dataisavailable from 1 5 None , 14 Liquids excluded CTLs, GTLs, biofuels GTLs, biofuels GTLs, biofuels biofuels but onlytheUS, r 29

Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Chap2:UKERC 22/09/200910:11Page30 30 Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Oil includeonly21Gb, representingthoseresources whichare‘underactive development’. 19 In 2002, the OGJ included 174.8 Gb of Canadian oil sands reserves for the first time. But the BP Statistical Review and Worl 18 The World Oil figures reflect a change from one reserve reporting system to another and back again. successive editionsoftheindustrydatabases. 2Preservesimportant consequenceofbackdating reserve canonlybeconstructedus revisionsisthatatime-seriesof‘current’ result, theindustrydatabasesarelikely tounderestimateglobal2Preserves (theUSaccountsfor2.4%ofthe total).Also, an of theSecuritiesandExchangea conservative Commission (SEC)andreflect interpretationof1Preserves3). Asa (seeSection 17 The US is treated differently in the industry databases, since the reserve estimates are based upon the reporting requiremen 16 TheIHSfield-level databaseformsthebasisofPEPScountry-level database. over ayear. 15 Definedasfieldswithultimatelyrecoverable resourcesexceeding 0.5Gborfieldsthathave producedmorethan100kb/dfor 2007). ~500 of the world's ‘giant fields’ (Robelius, IHS Energy are maintainedbyWood Mackenzie and covering themajorityofworld'sfields • • that: sources andtheindustrydatabasesare Key differencesbetweenthepublicdomain upon IHSEnergydata. aggregate estimates that are based in part expensive. The IEA and USGS publish sources ismuchdebated(Bentley, The relative meritsofthesedifferentdata • Miller, 2005;Smith,2008).Sincethese parties (Campbell and Heapes, 2008; which areeven lessaccessibletothird rely upontheirown,proprietarydatabases difficult. Many analystsofglobaloilsupply the industrydatabasesmakes comparison 2007), butthecostandconfidentialityof cumulative 2P discoveries; latter provide backdatedestimatesof cumulative 1P disco the formerprovide currentestimatesof estimates; estimates whilethelatterpro the formerprovide 1Preserve attempt toverify theiraccuracy. a widerrange ofsources,withmore while thelatterderive informationfrom supplied by national gov the former rely heavily upon data 15 Comprehensive databases 16 but these are very v eries while the 17 ernments, and vide 2P et al ., countries accounted for more than half of all-oil production in 2007, seven forty nine countries accounted for 98.9% Around 100countriesproduceoil,but importance of the Middle East producers. proved reserves, highlightingtheglobal 2.15 showstheregionaldistributionof previously beenunderestimated.Figure reserves of someofthesecountrieshad (Campbell, 2002a),itispossiblethatthe connection to physical resources being politicallymotivated withlittleorno commentators discount such revisions as exploratory drilling.Whilemany increased by 195% despite little or no increased by 57% in 1986 while the UAE’s the 1980s. For example, Iraq’s reserves estimates forseveral OPECcountries in the upward revisions to the reserve of the Canadian oil sands after 2002 the early1990s, differing treatment of Russian reserves in sources ofdiscrepancyincludethe reserve classification schemes. Major owing in part to the use of competing case of 1P reserve estimates (Figure 2.14) The differencesaresomewhatlargerinthe differing coverage ofliquids(Table 2.1). differences are largely explained by the oil productionfromsixdatasources.The Figure 2.13comparesestimatesofglobal estimates of cumulative 2P discoveries. databases, they contain backdated are based in part on the industry 18 the differingtreatment 19 and ing ts d Chap2:UKERC 28/09/2009 11:57 Page 31 Page 11:57 28/09/2009 Chap2:UKERC (Figure 2.17) and it is anticipated that the 2.16). OPEC is growing in importance and Iran) accounted for 39% (Figure while four(SaudiArabia, Russia, theUS Source: sources data different from trends reserve proved Global 2.14: Figure NGLs. IEA and EIA also include refinery gains, CTL, GTL and biofuels. Notes: Sources: sources data different from trends production oil Global 2.13 Figure OGJ andWorld Oilonlyincludecrudeoil,condensateandsyncrude.BPStatisticalReview andIHSEnergyalsoinclude BP StatisticalReview, OGJ, World Oil Monthly andIHSEnergyPEPSdatabase

Volume (Gb) Production Volume (mb/d) 50 55 60 65 70 75 80 85 90 9017 9018 9019 0020 2010 2005 2000 1995 1990 1985 1980 1975 1970 1000 1100 1200 1300 1400 600 700 800 900 9518 9519 9520 052010 2005 2000 1995 1990 1985 1980 1975 BP World Oil BP OGJ WorldOil Year over thelast five years, whiletwenty eight eight countrieshave increased production global production(IEA, 2008).Twenty Middle Eastwillincreasinglydominate IEA OMR Year EIA IPM OGJ IHS PEPS 31

Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Chap2:UKERC 22/09/200910:11Page32 32 Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production (Janssen, 2005).Itis frequently claimed investment and political constraints expansion may be obstructedbyphysical, production over thisperiod,but future for the most of the growth in non-OPEC declines (Figure2.18).Russia accounted countries have experienced production Source: of world Share Figure 2.16 production all-oil c by Source: reserves proved of distribution Regional 2.15 Figure BP (2008) BP Statistical Review, OGJ, World Oil Gb Gb

10000 15000 20000 25000 Saudi Arabia

Saudi Arabia 100 150 200 250 300 5000 50

Russian Federation 0 0

Iran

Iraq US United Arab Emirates

Kuwait Iran

China

Russian FederationVenezuela Mexico

United Arab Emirates Canada

Libya Kazakhstan

primarily the resultofeconomic and Senegal), insomecases thepeakmay be contribution toglobal production(e.g. many of these countries make only a tiny handful inthe1970s(Figure2.19).But peak ofproductioncomparedtoonlya that around 60 countries are past their Kuwait

Venezuela ountry Nigeria

Norway

US

Nigeria Canada

Qatar Iraq

Algeria China

Libya Brazil Rest of World Rest of World Chap2:UKERC 22/09/200910:11Page33 Source: countries. production in all-oil Change for 5 overthe last years topFigure 49 2.18 producing Note: Source: of world Share Figure 2.17 production all-oil for selected regions. 1-. 0 -0.5 -1 European Union excludes Estonia, Latvia and Lithuania prior to 1985 and Slovenia prior to 1991. BP (2008) BP (2008) Production (mb/d)

1965 10 20 30 40 50 60 70 80 90 0

1970 Production changeinlast5years(mb/d)

1975 Rest ofWorld US FSU

0.5 1980

1985 Year 1 European Union OPEC

1990 1.5

1995

2 2000 Russian Federation Angola Iraq Azerbaijan Kazakhstan Libya China Qatar Canada United ArabEmirates Kuwait Brazil Saudi Arabia Iran Sudan Algeria Equatorial Guinea Chad Ecuador Nigeria Thailand Venezuela Tunisia Peru Cameroon Rep. ofCongo(Brazzaville) Italy India Colombia Turkmenistan Trinidad &Tobago Gabon Romania Brunei Malaysia Vietnam Egypt Uzbekistan Denmark Australia Oman Argentina Yemen Syria Indonesia Mexico US United Kingdom Norway

2005 33

Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Chap2:UKERC 22/09/200910:11Page34 34 Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production (e.g. Iraq). could potentiallyexceed theirearlierpeak appropriate investment someproducers political factors (e.g. Poland) and with than physical depletionandsomecountries (e.g.Iraq) may subsequentlybeabletoincreaseproductionabove thepreviouspeak. date ofpeak.Insomecases(e.g.Poland), peakproductionmay beprimarilytheresultofeconomicandpoliticalfactorsrather Note: Source: producers post-peak of group expanding The 2.19 Figure Country and Year of Peak Production Trinidad andTobago1978(26.4%) Shows year ofpeakproductionandestimatedpercentagetheUSGS(2000)estimateURRthatwas produced bythe Papua NewGuinea1993(52.7%) BP (2008);USGS(2000) United Kingdom1999(43.4%) United States1970(28.6%) Czech Republic2003(31%) South Africa2004(34.8%) Turkmenistan 1973(6.3%) Netherlands 1986(24.1%) Kyrgyzstan 1961(24.8%) Uzbekistan 1999(10.7%) Venezuela 1970(19.6%) Cameroon 1985(32.2%) Tajikistan 1979(46.4%) Indonesia 1977(16.9%) Colombia 1999(27.6%) Myanmar 1981(35.9%) Germany 1967(26.1%) Denmark 2004(51.7%) Australia 2000(24.3%) Morocco 1963(55.1%) Romania 1976(36.3%) Hungary 1982(24.9%) Bulgaria 1969(10.1%) Barbados 1999(0.5%) Argentina 1999(52%) Georgia 1985(34.2%) Albania 1974(12.5%) Belarus 1974(29.6%) Austria 1955(24.1%) Tunisia 1980(10.2%) Greece 1984(15.4%) Turkey 1991(29.1%) Mexico 2003(39.3%) Yemen 2001(17.7%) Slovakia 1995(4.3%) France 1988(25.8%) Poland 2004(19.4%) Senegal 1991(0.1%) Gabon 1997(12.7%) Brunei 1979(17.9%) Egypt 1993(35.2%) Oman 2001(40.2%) Norway 2001(26%) Spain 1983(36.3%) Ukraine 1970(14%) Croatia 1977(44%) Cuba 2004(13.4%) Syria 1996(39.7%) Chile 1982(32.5%) Peru 1982(17.3%) Russia 1987(18%) Benin 1985(4.6%) Libya 1970(8.3%) Chad 2005(7.2%) Italy 2005(8.5%) Iran 1974(8.9%) 0 0010 0020 0030 004500 4000 3500 3000 2500 2000 1500 1000 500 0 following sectionsexplore thisinmore definition and size of oil reserves. The controversy in thepeakoildebateis Perhaps thegreatestsourceof Peak Production (mb/y)

Chap2:UKERC 22/09/200910:11Page35 Sources: fields and Reservoirs Box 2.4 and Dromgoole,1992).Initialestimatesof computer simulation(Barss,1978;Speers (e.g. viscosity, porosity, permeability) and measurements ofreservoir properties features), direct and indirect seismic surveys (to map geological originally inplace The firststageistoestimatethe same field (Mitchell, 2004b). estimates arelegitimatelypossibleforthe uncertainties, significantlydifferent publicly as reserves. Given the multiple viable toextract andhowmuchtodeclare is technically possible and economically how muchoilisintheground, The processinvolves judgmentsabout then aggregatedtocompaniesorregions. wells, reservoirs orfields(Box 2.4)and Reserves are estimated for individual reserves Estimatingandreporting 2.4 reported. expressed andhowsuchestimatesare are produced,howuncertainties detail, includinghowreserve estimates large field, and larger fields being broken down into smaller ones. of afieldmay change over time,withpreviouslydistinct fieldsbeingmergedintoone abandoned andthenumberofwellsmay range fromonetothousands.Theclassification contiguous area.Oilfieldsmay eitherbediscovered, underdevelopment, producingor projected tothesurface,reservoirs withinthefield canformanapproximately gas, all related to a single geological structure and/or stratigraphic feature. When Field: natural pressure system. An alternative term for reservoir is ‘pool’. or not,whichisphysically separated fromother reservoirs andwhichhasasingle Reservoir: Klett (2004);Drew(1997) A field is an area consisting of a single reservoir or multiple reservoirs of oil and A reservoir is a subsurface accumulation of oil and/or gas whether discovered (OOIP) throughamixof oil particular fields. techniques may beappropriatefor assumed, butonlyasubsetofEOR method andhencerecovery factor estimates aresensitive totherecovery mb/d) of global production. Reserve currently accountingforsome3%(2.5 traditionally appliedinsequence,withEOR improving themovement ofoil.Theseare reducing viscosity, displacing water and/or added with the aim of raising pressure, where gas,solvents, chemicalsorheatare normally made between and Dromgoole,1992).Adistinctionis anticipated technology(IEA,2008;Speers that canbeproducedwithcurrentor recovery factor The secondstageistoestimatethe improves as fields are developed. heterogeneous fields,buttheaccuracy with geologicallycomplicatedand OOIP can be highly uncertain, especially tertiary is pumpedintomaintainpressure,and physical liftisusedorwater ornatural gas pressure; where oilisproduced underitsown or secondary recovery nacdolrecovery enhanced oil , orthefraction ofOOIP primary recovery, , where (EOR) 35

Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Chap2:UKERC 22/09/200910:11Page36 36 Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production act whenitbecamecleartheestimates wereunrealistic. investigation foundthatexecutives in the exploration andproduction divisionhadexaggerated thesize ofreserves andfailed led toanimmediate7%fallin itssharepriceandthesubsequentresignationofchieffinancial officer. Asubsequent 20 InJanuary2004,Shellannounced thatitwas revisingitsreserve reporting andreduceditsproved reserves by3.9Gb. This have since remained unchanged for negotiations over productionquotas and by 80% (~300 Gb) in response to 1P reserves of OPEC countries increased For example, in the 1980's, the reported particular definitionsandinterpretations. have political motives for choosing sensitive tomarket conditionsbut can National oilcompanies(NOCs)areless The finalstageis 2004a). commercial negotiations(Mitchell, public domain since they may prejudice generally willnotbe releasedintothe above. These will be used internally, but 1P, 2Pand3Pclassificationsintroduced reserves, whichbroadlycorrespondtothe and‘high’estimatesofrecoverable‘best’ economic uncertainties to produce ‘low’, combine these geological, engineering and other factors.Companiestraditionally assumptions abouttaxtreatmentand and operating costs ofextraction, future priceofoil,theestimatedcapital depends onthecurrentandanticipated economic viability The thirdstageistoestimatethe project financing (Arnott, 2004). affect stock valuations, debt ratings and (IOCs) are closely scrutinised and can reports ofindependentoilcompanies from onecompany toanother. The reserve widely fromonecountrytoanotherand approaches to reserve reporting vary international standard, the definitions and consistent and widely enforced 1P estimatesandintheabsenceofaclear, It isunusualtoreportanything otherthan regulatory, accountingorotherpurposes. reserve reporting of recovery which for 20 recommended byseveral international The probabilisticinterpretationis different way. 2.2) butthelattermay beproducedina estimates ofrecoverable reserves (Table low (1P),best(2P)andhigh(3P) are often equated with the ‘deterministic’ has a10%probability(SPE,2007).These being exceeded, and a quantity whichhasa50%probabilityof probability ofbeingexceeded, a of a Probabilistic estimatesallowidentification increase over time (reserve growth). the measures of central tendency tend to proceeds andinformationimproves and distribution tends to narrow as production (Schulyer, 1999). The estimated these within a Monte Carlo simulation and economic parameters and combining probabilities tothegeological,engineering outcomes (Figure2.20)byassigning probability distribution better approachistoestimatea quantify thedegreeofuncertainty. A error in reserve estimates but fail to provide someindicationofthemargin The traditional 1P, 2Pand3Pcategories estimates reserve Assessinguncertaintyin 2.5 holdings. and distortedpictureofactualreserve reserve declarations provide onlyapartial 2007; Salameh,2004).Hence,public periods ofuptoadecade(Bentley, P90 quantity whichhasa90% P10 quantity which of possible et al P50 to ., Chap2:UKERC 22/09/200910:11Page37 High Best Low multiplicative rather than additive combination of independent variables which may be the case in petroleum formation. 21 Arandom variable (x)islognormallydistributedif ln(x) isnormallydistributed.Thiscommonlytheoutcome of the estimate, a statistical definition of a ‘best’ that isused.Someschemesdonotspecify particular reserve classification scheme interpretation willdependuponthe estimates, but their appropriate be easier to produce than low or high estimatesshould (Laherrère, 2001).‘Best’ P60 while a study of US data suggests P65 interpretation of1Preservesis closerto (1997) foundthattheCanadian P90 interpretation. For example, Jung actual estimatesmay notcoincidewiththe they should be exceeded in 90% of cases) conservative natureof1Pestimates(i.e. 2004) and while it highlights the organisations (SEC,2008;SPE,2007;UN, Deterministic terminology ofType estimate Deterministic and probabilistic terminologiesTable 2.2 associated with oilestimation. reserves probability densityfunction.Thevertical scalereferssolelytothecumulative distribution. P50 (median)andP10estimatesallrepresentpointsonthecumulative distributionfunction(blue),whichistheintegral ofth of reserves being somewhere between 0.5 and 2 units, and a small probability of there being a much large amount. The P90, In thecontextofreserve estimates,thereisnoprobabilityofbeinganegative volume ofoil, but thereisahighproba Note: ProbabilityFigure 2.20 and cumulative probability distribution of recoverable reserves 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 0 1 0 The probabilitydensityfunction(redline)representsastatisticaldistributionwhichinthisexampleisskewed totheleft. 0.5 P 9 0 Cumulative ProbabilityDistribution rvd rbbeadpsil 3)P10 Proved, probable and possible (3P) Proved (1P) Proved and probable (2P) M o d e P M 5 e 0 d 1 i a M n e a n 1.5 P 1 0 and Ruthrauff, 2006). left (Bloomfield, lognormal distribution iscommonlyfoundtotake a Figure 2.20).Inpractice, theprobability significantly differentifitisskewed (see normally distributed and may be coincide if the probability function is 2004). Thesethreemeasureswillonly median, modeormeantobeused(UN, median (P50) and others allow either the others implicitly assume it to be the mean. median which in turn is less than the circumstances, themodeislessthan 2 Probability DensityFunction emnlg description terminology P90 rbblsi Statistical Probabilistic P50 form which is skewed to the 2.5 et al ., 1979; Quirk 90th percentile 10th percentile Median 3 21 In these 3.5 bility e 37

Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Chap2:UKERC 22/09/200910:11Page38 38 Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production of being exceeded) is not well-defined and could in fact be any number between 1.0 and 1.9999. 22 Althoughhelpful,thisexample isstrictlyincorrectbecausetheP83ofonedie(thenumberwhich hasan83%or5in6chanc Aggregation of probabilistic estimates reserve Box 2.5: dice is thrown, the probability of the with theexampleoftwodice.Ifasingle Pike (2006) illustrates the logic behind this median isequaltothemean. be acorrectestimateoftheP50if overestimate fact an have reserves (40,20,2)Gb-thisisin generally truethatthe newcompany will the twocompaniesmerge.Itis reserves R2,alsowith(20,10,1)Gb, and 10, 1)Gbandanothercompany owns estimated tohave (P10,P50,P90)=(20, company owns reserves R1 which are Ross, 1998). For example, suppose one (Cronquist, 1991; Jung, 1997; Pike, 2006; estimates canleadtosignificanterrors The commonpractice ofaddingreserve probability of total reserves exceeding 2Gb is: 2Gb byhaving onefieldperformslightly worseandtheotheralotbetter. Hence,the probability ofthesecondexceeding 1Gb. However, thesumoftwocouldalsoexceed that thereisa90%probabilityofthefirstexceeding 1Gbandanindependent90% Assume thattheprobabilitydistributionsofreserves intwofieldsareindependentand the twodistributionsare linked. case theabove calculationiscomplicated byaweightingfactorindicatinghowclosely of one is likely to be associated with an under- or over-estimation of the other. In this the samegeologicalformationoraredeveloped inparallel) anunder- orover-estimation Alternatively, ifthetworeserves arenotcompletelyindependent(e.g.theypartof If a normal distribution is assumed, it will be an underestimate. be anover orunderestimateofthe trueP90accordingtotheshapeofdistributions. the P90estimates(which gives theprobabilityofR1+R2exceeding 2Gbas0.9) may probability of the sum exceeding 2Gb will be greater than this. A simple summation of probability of both fields independently exceeding 1Gb is 0.9 x 0.9 = 0.81, but the The second term represents the degree of overlap of the two distributions. The underestimate ∫ [P(R1 exceeds 1 Gb by X)* P(R2 falls short of 1 Gb by less than X)]dX of the true P10, and will only of thetrueP90,an P(R1 exceeds 1 Gb)*P(R2 exceeds 1 Gb) + not then combined (seealsoBox 3.1). the actual1Pfigurefortwofields of two fields would be an the 1P (P90) estimates of the oil reserves 83%). probability is actually 97% and not be significantlyunderestimated(the the combinedscoreexceeding twowould individual P83figures,theprobabilityof bysimplyaddingthe figures. Hence, aggregation of the two individual P83 out of36),ortwicethesimplearithmetic 2.0. The corresponding P83 figure is 4 (6 is 97% (35 out of 36). So the P97 figure is probability of the outcome exceeding two But iftwodicewerethrown,the 6). Inotherwords,theP83figureis1.0. outcome exceeding oneis83%(5outof 22 In asimilarmanner, thesumof underestimate of e Chap2:UKERC 22/09/200910:11Page39 dataset toanother. and isunlikely tobeconsistent fromone should be smaller than with 1P reserves aggregate 2Preserves. However, theerror could also lead to an underestimate of mean (e.g.Figure2.20),simpleaddition median (P50).Iftheliesbelow common tointerpret2Pestimatesasthe represent mean estimates, but it is more Simple addition is only valid if they of the relevant probability distributions. the mean,medianormode)andshape estimates (i.e.whethertheycorrespondto probabilistic interpretationofthese outcome will depend upon both the the aggregation of 2P estimates since the The sameconclusionsmay notapplyto (Figure 2.6). also morelikely toincreaseover time produced (Pike, 2006).Thesenumbersare understate theamountofoillikely tobe to beasetofnumberswhichsignificantly is normallyemployed. Theresultislikely reinforced bytheaggregationprocessthat degree ofconservatism isfurther of likely recoverable resources, but the estimates provide a conservative estimate the most biased. Hence, not only do 1P that the global estimates are likely to be degree of underestimation, with the result 1P figure. Each addition increases the systematic underestimation of the actual data toglobalestimates,therewillbea whole company or country and national aggregated toawholefield,fielddata used. When1P(P90)reservoir datais estimates sincethesearethemostwidely Aggregation isespeciallyimportantfor1P World Petroleum Council (WPC), and the Society of Petroleum Evaluation Engineers (SPEE). 23 JointlyproposedbytheSociety ofPetroleum Engineers(SPE), theAmericanAssociationofPetroleum Geologists(AAPG),the 2.21) Management System (PRMS, Figure step forward was thePetroleum Resources with limitedsuccesstodate.Animportant attempts tostandardisedefinitions,but The petroleum industry has made several different sources. difficulty ofaggregatingestimatesfrom consistent. This disparity highlights the ‘possible’ werebothlessfrequentand of beingexceeded, whiledefinitionsof identified witha90%and50%probability ‘Proved’ and ‘probable’ were commonly terms andthesedidnotalways agree. used probabilistic definitions of these ‘probable’ and 13 ‘possible’. Only nine that 19usedtheterm‘proved’, 16used (1991) compared25schemesandfound the McKelvey Box (Figure2.5).Cronquist reserves, most of which are based upon There aremultiplesystemsforclassifying schemes classification Reserve 2.6 classification has been proposed by the UN A competing,three-dimensional used. depending upon the particular method hence couldleadtodifferentresults summation andstatisticalaggregation and P10.Italsoallowsbotharithmetic probabilistic estimatesbasedonP90,P50 based on qualitative guidelines and but allowsbothdeterministicestimates PRMS uses the 1P, 2P and 3P terminology, level ofuncertainty (WPC,2007).The each being subdivided according to the resources andprospective resources,with recoverable reserves, contingent 23 which distinguishesbetween 39

Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Chap2:UKERC 22/09/200910:11Page40 40 Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production scheme isprovided by the USSecurities An importantandinfluentialclassification by thepetroleumindustry. this scheme seems unlikely to be adopted classification of other mineral resources, clear whetherthisisactuallythe case. 25 AnumberofNOCs(e.g.Saudi Aramco) alsoclaimto followSECguidelinesbutintheabsenceofthird-partyauditingit isno 24 For example,awell-developed, economicallyviableprojectwithwell-characterised geology wouldbeE1F1G1. for each. assessment (G),withfurthersubclasses (E), fieldprojectstatus(F)andgeological distinguishes betweeneconomicviability (Ahlbrandt, recoverable from undiscovered accumulations by application of future development projects. considered matureenoughforcommercialdevelopment duetooneormorecontingencies. Prospective resourcesarepotentially Contingent resources are potentially recoverable from known accumulations, but where the applied projects are not yet Note: Source: System Management Petroleum Resource The 2.21 Figure Reserves arecommerciallyrecoverable bytheapplicationofprojectstoknownaccumulationsunderdefined conditions. SPE (2007)

24 TOTAL PETROLEUM INITIALLY-IN-PLACE (PIIP) Although widelyusedforthe et al

., 2003; UN, 2004). This DISCOVERED PIIP UNDISCOVERED PIIP SUB- COMMERCIAL COMMERCIAL RVDPOAL POSSIBLE PROBABLE PROVED Estimate 1 1 Low C P PROSPECTIVE CONTINGENT UNRECOVERABLE UNRECOVERABLE RESOURCES RESOURCES PRODUCTION Uncertainty RESERVES Range of Estimate ES 2 2 Best majority of IOCs. of globalproductionandincludethe (SEC, 2008).Theseaccountforaquarter reserve estimatesbyUS-listedcompanies requires thedisclosureofaudited1P and Exchange Commission(SEC)which to probabilisticestimatesanddonotallow revenue streams.Theymake noreference provide confidence to investors in future conservative sincethey areintendedto C P Estimate 25 High 3 3 C P The SECdefinitionsare scale Not to

Increasing Chance of Commerciality t Chap2:UKERC 22/09/200910:11Page41 • as follows: The main conclusions from this section are 2.7 Summary remain handicapped by inadequate data. assessment ofglobaloildepletionwill economic and political considerations, the 1P estimatesandbothareinfluencedby continue toreportonlyhighlyconservative estimates using their own definitions, IOCs that NOCs continue to report un-audited comprehensively adopted.To theextent drawbacks andarealong way from being the currentproposalshave several schemes arebecomingmoreharmonised In sum, while reserve classification optional. permitted, butunfortunatelythisremains Declaration of2Preserves willalsobe absence of contact to a well (SEC, 2008). allow reserves to be estimated in the incorporate probabilistic definitions and substantial revisions and from 2010 will The SEC requirements are undergoing the reportingconvention. reserve growththatispartlyanartefactof more wells are drilled, contributing to well, reserve estimates appear to grow as reserves within the production range of a conditions. Also, sincetheyonlyinclude changes in economic and operating for future technical developments or conventional oil is anticipated to However, whatiscommonlytermed their differingcoverage ofthosefuels. supply forecastsresultinpartfrom between both data sources and oil consistent way andthedisagreements Liquid fuelsarenotsubdividedina • • • • economic and political influences on conserv confined tothereporting ofhighly requirements andthisislargely reserv standardisation. Only a subset of global only limited progress towards schemes inusearoundtheworldand There aremultiplereserve classification but remaininsufficientlyused. evaluation oftheaccuracy ofestimates) (including allowingretrospective estimates offer significant advantages different w they are defined and interpreted in and ‘possible’ reserves are widely used, While termssuchas‘proved’, ‘probable’ very uncertain. the othertwocategoriesarenecessarily level noneof theseagree.Estimatesfor number ofsources,butattheglobal two categoriesareavailable froma yet-to-find resources. Data on the first reserv cumulative production,declared for aregionrepresentthesumof The ultimately recoverable resources production peaked in 1979. On a per capita basis, global oil produced) vary between28%and56%. resources that have already been proportion of ultimately recoverable con the level of depletion of global Gb) inonlytenmonths.Estimatesof much oilastheUKhasproduced(24 preceding centuryandnowusesas half asmuchoilwas producedinthe Since 1993, the world has produced next 20 years. dominate globalliquidssupplyover the ventional oilresources(i.e.the es is subject to formal reporting es, futurereserve growthand ative 1Pestimates.Various a ys. Probabilisticreserve 41

Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Chap2:UKERC 22/09/200910:11Page42 42 Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production • • av different data sources. Publicly production result in part from the use of Disputes over a near-term peak in oil the reliability of published figures. reserve reporting further compromise underlying probabilitydistributions. the estimatesandshapeof depending upontheinterpretationof either positive, negative orzero introduce lesserrorbutthismay be Aggregation of 2P estimates should underestimate actual1Preserv total is likely to significantly estimates to form a regional or global The common practice of adding 1P sources remaininsufficientlyexplained. disagreements betweendifferentdata insufficiently understoodandthe between theseapproachesremain production. At present, the differences and theriskofanear-term peakin information about resource depletion region and can provide more ultimately recoverable resourcesofa latter is more suited to estimating the Both approaches have merits, but the date ofdiscovery oftherelevant fields. reserves and‘backdate’revisionstothe databases provide estimatesof2P they are made, while the industry reserve revisions in the year in which estimates of1Preserves andrecord ailable datasourcespro vide es. • • aggregation, global1Preserves could As aresultofinappropriate world's reserves. countries whichholdthemajorityof uncertainties aregreatestforthose proportional to their importance. The whose level ofconfidenceisinversely forecasts mustrelyuponassumptions the absence of such data, supply tr most oftheworld'soilfieldswould Audited estimates of 2P reserves for to strict confidentiality requirements. regard, theyareexpensive andsubject industry databasesarebetterinthis assessing globaloildepletion.While poorly suited to the purpose of Publicly available 1Preserve datais treated with considerable caution. to another, theglobaltotalsshouldbe in magnitudeandsignfromonecountry between thesedatasourcesvary both overestimated. Since the discrepancies underestimated or the latter that theformerhave been global 1Preserves, suggestingeither the BPStatisticalReview estimatesof reserves are approximately the same as the industry estimates of global 2P and Heapes, 2008). At the same time, claimed by some authors (Campbell ov BP (2008) – potentially offsetting the be larger than the 1240Gb reported by ansform thepeakoildebate.Butin erestimation ofOPECreserves thatis Chap3:UKERC 22/09/200910:12Page43 geographic boundary. 26 A‘play’ isanarea forpetroleumexploration thathascommongeologicalattributesandlieswithin somewell-defined • • • • namely: more detail at four of these issues, already beenmade.Thissectionlooksin number ofareasandsomeprogresshas increasing thedegreeofconsensusina Nevertheless, thereispotentialfor available intheforeseeablefuture. but this seems unlikely to become towards resolving such disagreements, individual fields could go a long way interpretations. Improved dataon creates forcompetingviewsand publicly available dataandthescopethis made worsebytheinadequacyof unsurprising. However, thesituationis existence ofsuchdisagreementsis multi-dimensional nature of this topic, the technical issues.Given thecomplexityand disagreements over asetofmore supply isunderpinnedbyequallypolarised ‘pessimists’ over the future of global oil The disputebetween‘optimists’and 3. field orregioncanbeproduced. remaining reco Depletion rates: expected to change in the future. field is declining and how this may be production fromdifferentcategoriesof Decline rates: expected inthefuture. how muchgrowthmay reasonablybe siz Reserve growth: small fields. contained withinvery largeandvery relativ among differentsizes offieldandthe resources inaregionaredistributed Field size distributions: e offieldstendtogrowover timeand Enduring controversies –fieldsizes,reserve growth anddeclinerates e proportionofresources v erable resourcesina why estimatesofthe how rapidly the how rapidly the how oil Section 3.5 concludes. constraints onviablesupplyforecasts. and showshowthelatterprovide between declinerates anddepletionrates Section 3.4clarifiestherelationship estimates ofglobalaverage declinerates. rates, illustrative examplesandcurrent including definitionsandmodelsofdecline state of knowledge on decline rates, 1995. Section3.3summarisesthecurrent growth observed in industry databases since forecasting ofreserve growthandthe different factorstoreserve growth,the growth, including the relative contribution of oil supply. Section3.2 examinesreserve contribution ofsmallfieldstofutureglobal size distribution of discovered fields and the including theimportanceoflargefields, Section 3.1examinesfieldsize distributions, Reports 3,4 these topics can be found in More comprehensive examinations of from individual ‘exploration plays’ seems broadlyapplicableatlevels ranging discovered relatively early. Thisrule number oflargefieldswhichtendtobe of resourcesarecontainedwithinasmall It isgenerally assumedthatthemajority refers to the estimated URR of each field. oil fieldswithinaregion,where‘size’ assumptions about the size distribution of supply forecasting frequently rely upon and assessment Methods ofresource Fieldsizedistributions 3.1 distribution varies fromone region to However, the precise form of the size entire world. and 5. Technical 26 to the 43

Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Chap3:UKERC 22/09/2009 10:12 Page 44

another and is a long-standing focus of exceeding 5 Gb (i.e. >73 days current dispute (Drew, 1997; Kaufman, 2005; global production) with the largest Laherrère, 2000a). Of particular interest is (Ghawar in Saudi Arabia) having a URR of the proportion of resources contained ~140 Gb (~5 years current global within very large and very small fields, production). The 1300 fields with a URR since the former dominate current oil exceeding 0.1 Gb represented only 3% of production and the latter are expected to the total but accounted for 94% of become increasingly important in the cumulative discoveries. The remaining future. 39000 fields accounted for less than 6% of the total and individually contributed only 3.1.1 The importance of large fields a tiny fraction of global production.

One of the first global surveys of crude oil Similar results were obtained by Robelius fields was by Ivanhoe and Leckie (1993) (2007), who provided an updated analysis who grouped fields into ten size categories of the world's ‘giant’ oilfields using data on the basis of their estimated URR (Table from a variety of sources. Robelius 3.1). The 370 fields with a URR exceeding estimates that there are ~47500 oil fields 0.5 Gb (i.e. >7 days current global in the world, 73% of which are in the production of crude oil) represented less United States.27 Only 507 (~1%) of these than 1% of the total number of fields but are ‘giants’, with an estimated URR of accounted for three quarters of cumulative more than 0.5 Gb, of which 430 are in discoveries. Of particular importance were production and 17 under development. the 42 ‘super-giant’ fields with a URR Robelius estimates that the 100 largest

Table 3.1: Ivanhoe and Leckie’s estimates of the size distribution of the world’s oilfields

Category Estimated URR (mb) No. in world Megagiant >50,000 2

Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Global Oil Depletion: An assessment of the evidence for a near-term Supergiant 5000-50,000 40 Giant 500-5000 328 Major 100-500 961 Large 50-100 895 Medium 25-50 1109 Small 10-25 2128 Very small 1-10 7112 Tiny 0.1-1 10849 Insignificant < 0.1 17740 Total 41164

Source: Ivanhoe and Leckie (1993)

27 These figures demonstrate that much more exploration has taken place in the US compared to other regions of the world. This has implications for the relative suitability of different resource assessment methodologies (Section 4) and suggests there could be unexploited potential outside of the US.

44 Chap3:UKERC 22/09/200910:12Page45 definition. 28 Höök, Source: The estimatedFigure contribution 3.1 of giant oilfields to global crude oil production oil). 0.14% currentglobalproductionofcrude those producingmorethan100kb/d(i.e. Simmons (2002)definesgiantfieldsas years ago. giants were discovered more than 50 cumulative discoveries. Halfofthese approximately two thirds of global while the giants as a whole account for production ofcrudeoil(seeFigure3.1) fields account for 45% of the global Ivanhoe's classificationsystemandrely provided bytheIEA(2008)whouse The mostup-to-dateestimatesare 14 accountedforover 20%. only 12% of production while the largest smallest 62ofthesefieldsaccountedfor global production of crude oil. The accounted forapproximately halfthe giants underthisdefinitionwhichin2002 28 Robelius (2007) et al. Daily production (mb/d) He estimates that there are 116 (2009b) estimatethereare20giant fieldsunderSimmons’definitionwhicharenotgiant Ivanhoe's 10 20 30 40 50 60 70 80 0 9513 9514 9515 1960 1950 1945 1940 1935 1930 1925 itrclPouto l insTop 20Giants AllGiants Historical Production World CrudeOilProduction1925-2005 9516 1975 1965 1955 Year fields account for up to half of the global uncertain, itisclearthat around 100oil Hence, whiletheprecisenumbersmay be by ~70%. and itsproductionhassincedeclined largest oilfield,Canterell,peaked in2003 peak ofproduction.Theworld’s second- decades and16ofthemarepasttheir have beeninproductionforseveral (Table 3.2).Mostofthe20largestfields with Ghawar accounting for a full 7% and asmuch20%fromonly10fields, only 110fields,25%from20fields half ofglobalproductionderived from development or‘fallow’. Approximately additional 84 giant fields were either under fields (54supergiantand320giant).An crude oil production derived from 374 in production2007,butaround60%of They estimatethat70,000oilfieldswere largely upon the IHS Energy database. 9018 1990 1980 1970 9519 2005 1995 1985 2000 45

Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Chap3:UKERC 22/09/200910:12Page46 46 Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production 29 These results indicate that skewed size distributions apply as much to reservoirs within fields as to fields within larger r Note: Source: words, the frequency distribution of the typically took a lognormal form - in other were among the first to observe that this discovered fields. Arps and Roberts (1958) the size distributionofthesample oil fields in a region can be inferred from The size distribution of the population of Thecontributionofsmallfields 3.1.2 global supply. therefore of critical importance for future and the potential for reserve growth are these fields, their future production profile decade or so. The remaining reserves at will begin to decline within the next peak of production and most of the rest relatively old,many arewellpasttheir discoveries. Most of these fields are fields account for two thirds of cumulative production ofcrudeoil,whileupto500 Table 3.2 The ten largest oil fields inTable the world 3.2 Azeri-Chirag--- Priobskoye Safaniyah Samotlor Canterell Guneshli Data as of 2007, so ‘peak’ in 2007 may indicate that production is still increasing. Rumalla Ghawar AHWAZ Burgan Zakum

Field IEA (2008) Total Saudi Arabia Saudi Arabia Azerbaijan Abu Dhabi

Country Mexico Kuwait Russia Russia Iran Iraq

Onshore/

offshore On/off Off Off Off On On On On On On

discovery

Year of Year 1982 1985 1964 1958 1960 1938 1953 1951 1977 1948 1993). Inparticularsmallfieldsarelikely whole (Bloomfield,etal., 1979;Kaufman, be unrepresentative ofthepopulationasa relatively small sampleoffieldsthatmay difficulties indrawing inferencesfroma Wang, 1983).Thiswas despitethe Hinde, 1985; Kaufman, 1975; Lee and ‘discovery processmodels’(Forman and basis ofsomehighlysophisticated and 1970ssubsequentlyformedthe conventional wisdomduringthe1960s lognormal size distributionbecame The notion that oilfields followed a Griffiths (1965). US databyKaufman (1963)andDrew sizes McCrossan’s (1969)analysisofreservoir supported byseveral studies,including 3.2). This observation was subsequently resembled a normal distribution (Figure logarithm ofdiscovered fieldsizes 29

production

Year of Year in Western Canadaandstudiesof 2007 2007 1998 1977 1980 1972 1979 1998 2003 1980

peak

production 1082 3435 2415 1493 2128 2054 5588

Peak

kb/d 652 658 795

production 14260 1170 1250 1408 1675 5100

2007

kb/d 652 658 674 770 903 egions. Chap3:UKERC 22/09/200910:12Page47 truncation’ beingimposed. acceptable modelforthesample size distribution over awiderange ofmeasuresexploratory efforteven without‘economic increased, thesamplesize distributionevolved towards thepopulationdistribution.However, lognormality was foundtobean progressively morefieldsineachofthesmaller size classes(Figure3.3).Hefoundthat,asthenumberofexploratory wells 31 Power (1992)simulatedthediscovery offieldsfromatheoreticalpopulationthathad‘power-law’ size distributionwith region toanother(DrewandSchuenemeyer, 1993;Drew, 30 Comparable inferences can be drawn from cross-sectional data, since the minimum viable field size varies widely from one Drew, 1997). (Cramer Barton and La Pointe, 1995; including the Permian basin in Texas observed in many oil-producing regions, fields to fall. This process has been modal size of the sample of discovered economic to find and develop, causing the these fieldsshouldbecomeincreasingly improves, oilpricesriseand/orcostsfall, exploration proceeds, technology sampling of small fields (Figure 3.3). As may therefore result from the under- Drew, 1983). Thelognormaldistribution Drew, Roberts, 1958; Attanasi and Drew, 1984; - so-called ‘economic truncation’ (Arps and because theyarelesseconomictodevelop to be underrepresented in such samples Source: 1958 in basin Denver forthe distribution size field gas and Oil 3.2 Figure

Adapted fromArpsandRoberts (1958)andDrew(1997). Number of Fields Found et al., 1988;Schuenemeyer and 10 20 30 40 50 60 0

0-2 30

2-4 Additional sampling bias

4-8

8-16

16-32

32-64

Estimated URR(kb) 64-128

et al. 128-256 , 1982).

256-512 result ofbiasedsampling. observed inpractice, thisispartlythe While curved plotsaremorecommonly lognormal, thisplotwouldbecurved. In contrast, ifthesize distributionis approximate astraight line(Figure3.4). size (V)onlogarithmicscalesshould number (N)offieldsexceeding aparticular 3.1). Ifthisisthecase,aplotof ‘power-law’, or‘Pareto’ form(seeBox distribution was morelikely totake a proposed thatthepopulationfieldsize observations, Drewandcolleagues On thebasisoftheseandsimilar 1992). discover thelargerfieldsfirst(Power, may be introduced by the tendency to

512-1024 31 1024-2048

2048-4096

4096-8192

8192-16,38

16,384-65,53 4

6 47

Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Chap3:UKERC 22/09/200910:12Page48 48 Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Box 3.1: Power-law field-size distributions and Zipf’s law Zipf’s Power-law and distributions field-size Box 3.1: size distribution ofthesamplediscovered fieldsattimet Note: Source: distribution of the size of discovered fields HowFigure 3.3 the undersampling of small fields may lead to a lognormal frequency at t V shape ofthedistribution.Hence,aplotnatural logof size distributionisgiven by: Let states thattheproductofrank andsize isapproximately constant( oil fieldsareranked indescendingorderofsize sothat the largestisrank 1,Zipf’s law the size and‘rank’ Power-law distributionsarerelatedto‘Zipf’s law’whichdescribesarelationshipbetween Mast, 1989). been widelyusedtoassessthepetroleumresources oftheUS(Houghton, US oil fields. Davies and Chang (1989) have criticised the power-law model, but it has resources couldbemodelled with a power-law anddemonstrated thisbyananalysis of Mandlebrot (1977). Indeed, Mandlebrot (1962) was the first to propose that oil distributions known as ‘probabilistic fractals’ which have their roots in the work of (1897) who represented income distribution in a similar way. They belong to a family of Power-law distributions are also termed ‘Pareto distributions’, after Vilfredo Pareto account for increasing proportion of the URR. decrease asthesize classitselfdecreases,butas of Barton andScholz(1995)fittedthisequationtodatafromsixregionsfoundvalues plotting cumulative frequency( field rank ( applicability ofZipf’s lawisusuallyinvestigated by plottingfieldsize ( Huberman, 2002).The two approachesareequivalent. 1 should approximate astraight linewithslope >t a Green line indicates ‘power-law’ size distribution of the population of fields. Blue line indicates the approximately lognormal N(V) 0 ranging from0.8to1.0.If when changesineconomicsandtechnologyhave loweredthesize thresholdforeconomicallyviablefields. Drew (1997) Number of Fields represent thenumberoffieldsthatexceed aparticularsize N), whiletheapplicability ofaPareto distributionisusually investigated by (N) S 1 of discretephenomena(Merriam, N(V) a a S N(V 0 = ≤ 1.0 AV )) asafunctionoffield size ( Size ofField -a 0 . Red lineindicatessize distributionof thesampleofdiscovered fields , where , thevolume ofoilineachfieldsize classwill -a A a is ascalingfactorand : tends towards 1.0,smallfieldswill l nN(V) = et al. N , 2004;Zipf, 1949).When against thenatural logof f (S f (S Parent Population l nA 0 1 ) ) -a V. V A power-law field l ) asafunctionof V N(V nV ) (Adamicand a (Figure 3.4). )*V et al. defines the ~k). The , 1993; Chap3:UKERC 22/09/200910:12Page49 a parabolic fractal and smallerstillwitha power law distribution, smaller with The proportion will be greater with a and extrapolating tosmallerfieldsizes. to thesize distributionofdiscovered fields estimated byfittingoneofthesefunctions smaller, undiscovered fieldsmay be The proportionoftheURRcontainedin small fields are found. plots typicallyreducesover timeas more provide abetterfit.The curvature ofthese distribution (a ‘parabolic fractal’) can often quadratic cumulative frequency curved, not linear plots.” and shows how a argues that “….natural data gives rise to However, Laherrère(1996a;2000a) single exploration play totheentireworld. data fromsixregionsranging froma power-law modelprovides agoodfitto Barton andScholz(1995)showhowthe difference between theoretical population distribution and observed sample distribution Log-log plot illustrating of the cumulativeFigure frequency 3.4 versus field size,

Log of number of fields exceeding size – N(V) sample ofdiscoveredfields Size distributionof Size distributionof Population offields recovery from the giant fields. supply thanthepotential forincreased of lesssignificanceto futureglobaloil of the number of small fields are therefore offshore regions. The competing estimates contribute to global supply, especially in result, many small fields will never by theenergyreturnoninvestment.As a there willalways bealowerlimitimposed should make moresmall fields viable, technical improvements andhigher prices minimum size thresholdassumed.While level, but are likely to be sensitive to the estimates arenotavailable attheglobal smaller than30kb).Corresponding 31% oftheregionalURR(excluding fields small fieldscontainedbetween9%and regions andestimatedthatundiscovered Scholz (1995) fitted a power law to six lognormal. For example, Barton and Log offieldsize-V 49

Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Chap3:UKERC 22/09/200910:12Page50 50 Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production recovery factorofonly27%. factors shouldbetreatedwithcaution (IEA,2005).For example, Laherrère(2006)usesIHSdatatoestimateaglobalaverage 32 Therecovery factoristheratio oftheestimatedURRtoOOIP. Estimates ofregionalandglobalaverage recove some cases, smaller fields that were use of seismic and other techniques. In reservoirs may be obtained through the the volume, shape and characteristics of be discovered orabetterunderstanding of or extensionstopreviousreservoirs may knowledge. For example, new reservoirs region asaresultofimproved geological oil inplace (or decrease) in the estimates of factorsGeological technological anddefinitional. three headings,namelygeological, reserve growth canbegroupedunder The multiplefactorsthatcontributeto Sources ofreserve growth 3.2.1 understood andthesubjectofcontroversy. However, itremainsbothpoorly importance for future global oil supply. future. It is therefore of considerable is expectedtocontinuedosointhe additions inmostregionsoftheworldand accounts forthemajorityofreserve reserves. Reserve growthcurrently which is growing rather than declared ultimate recovery growth since it is this cumulative discovery growthorestimated describe thisphenomenonwouldbe Section 2,amoreaccurate termto Schmoker, 2003a). As described in discovered fieldsthroughtime.” (Klettand recoverable resources in previously “...the commonly observed increase in Reserve growth has been defined as: Growth Reserve 3.2 (OOIP) forareservoir, fieldor represent anincrease original average isaround34%(IEA,2008). reservoirs, though theestimatedglobal high as80%forpermeability technologies. Recovery factors can be as from theapplicationofimproved recovery reservoirs and optimised drilling as well as result frombettercharacterisation of economically viabletorecover. Thiscan OOIP thatisbothtechnicallypossibleand average recovery factorcouldaddsome each percentage increase in the global other regions of the world. In principle, is considerable scope for application in higher thantheglobalaverage andthere has ledtorecovery factorsthatare ~5% reservoir. Extensive useof EOR intheUS accessibility andcharacteristics ofthe whose suitabilitywillvary withthetype, oil recovery factor through the application of for substantiallyimproving therecovery surface. Inmany casesthereisalsoscope pressures andtheflowofoilto injection is employed to increase reservoir where active pumpingorwater/gas own pressure;and recovery is commontodistinguishbetween recovery factor, ortheproportionof activities whichincreasetheestimated Technological factors individual field. not representthegrowthofOOIPforan reserve growth is estimated, but it does affect thehistoricaldatafromwhich proceeds (Drew, 1997).Thisprocessmay merged into larger fields as exploration previously classified as separate may be where oil is recovered under its (EOR) techniques(Box 3.2) secondary recovery represent those enhanced primary 32 It ry Chap3:UKERC 22/09/200910:12Page51 potential may be large, improving the years ofcurrentproduction.Butwhilethe 80 Gbtoglobalreserves, ornearlythree Box 3.2 Enhanced oil recovery techniques techniques recovery oil Enhanced Box 3.2 • • There are three broad groups of EOR techniques (NPC, 2007): drilling followed by CO what canbeachieved withEOR–inthiscaseadditionalvertical andhorizontal The IEA (2008) use the example of the Weyburn field in Canada (see below) to illustrate • reproduced. Source: sustain pressureandmobilisealargerproportionoftheoil.CO reduce viscosity, achieve ‘miscibility’ (ahomogeneoussolution),displacewater, Gaseous heavy oil but their use has declined since the mid-1980s. partially ‘crack’ heavy oiland/orincreasepressure.Theyareparticularlysuitablefor Thermal complicated, unpredictable, costly and sensitive to reservoir characteristics. between oilandinjectedw Chemical storage (CCS)technologies. use natural sourcesofCO fastest growing form of EOR and is very effective for light oil. While many applications

IEA (2008) Thousand barrels per day 10 20 30 40 50 0 9016 9018 9020 0022 2030 2020 2010 2000 1990 1980 1970 1960 1950 methods introduceheat,typicallyintheformofsteamtoreduceviscosity, methods inject carbon dioxide, nitrogen or other gases at high pressure to methods injectvarious compoundstoreducethe‘interfacial tension’ 2 injection. Butitisnotclearhowwidelythisexamplecanbe 2 , futureprojectsmay belinked tocarboncaptureand ater . These are not widely used and tend to be 2008). more thantwodecades’toachieve (IEA, global average to50%couldtake ‘much Original -verticalwells Infill -verticalwells Infill -horizontalwells CO 2 injection 2 injection isthe 51

Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Chap3:UKERC 22/09/2009 10:12 Page 52

Definitional factors comprise a mix of The amount of reserve growth should definitional, legal, economic and political depend upon the particular definition of influences which influence reserve reserves on which the cumulative estimates but are independent of either discovery estimates are based (e.g. 1P or the OOIP or recovery factor. These include 2P). Reserve growth may be particularly changes in reserve classification schemes, high for cumulative discovery estimates such as occurred in Russia in the 1990s based upon 1P reserves since these and are currently underway in the US, and typically underestimate the URR. Reserve the practice of excluding reserves at growth should be smaller for cumulative discovered fields that have yet to receive discovery estimates based upon 2P production sanction. There may also be an reserves and if these correspond to implicit shift in definitions over time as a median (P50) estimates of URR, we would result of changes in personnel, operators expect cumulative discovery estimates to and reporting cultures. All reserve be downgraded as frequently as they are definitions require assessment of upgraded. However, analysis suggests that economic viability, so changes in this is not the case. With respect to 1P, 2P technology, oil prices and other economic and 3P reserve estimates Drew (1997) conditions may also affect the volume of notes that: “…the irony …….is that all three declared reserves. sets of numbers are pessimistic – they all

Figure 3.5 Reserve growth in oil fields larger than 0.5 Gb in the UKCS

350%

300%

250% Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Global Oil Depletion: An assessment of the evidence for a near-term 200%

150%

100%

50%

0% 0 5 10 15 20 25 30

Years Since Discovery

Cumulative Discovery as % of Initial URR Estimate MEAN Beryl Brent Claymore Cormorant Forties

Fulmar Magnus Ninian Piper Schiehallion

Source: BERR Note: Horizontal axis represents years after first production. Vertical axis is cumulative discoveries for each field, expressed as a percentage of the initial declared reserves. The heavy black line is the simple arithmetic means of percentages, un-weighted by volume. This data series is no longer published by the UK government, in part because of inconsistencies in reporting between different operators.

52 Chap3:UKERC 22/09/200910:12Page53 reserve decrease(Figure3.6). fields discoveredhave since1980 showna by only20%over thisperiodwhilemany However, smaller fieldsintheUKCSgrew growth intheindustrydatabases. world, wemay expectsignificantreserve case for2Pdatafromotherregionsofthe estimates decreasedinsize. Ifthisisthe years, whilenoneoftheindividual increased byapproximately 50%over 27 estimate ofcumulative discoveries be similarto2P. For largefields,themean the UKgovernment andwetake themto 3.5). Theseestimateswerepublishedby in theUKContinentalShelf(UKCS)(Figure in cumulative discovery estimatesforfields Figure 3.5and3.6showthechange 1997). grow withthepassageoftime”(Drew, different operators. volume. This data series is no longer published by the UK government, in part because of inconsistencies in reporting between a percentageoftheinitialdeclaredreserves. Theheavy blacklineisthesimplearithmeticmeansofpercentages,unweightedb Note: Source: Reserve growthFigure 3.6 in the UKCS in Gb oil fields smaller than 0.5

Horizontal axisrepresentsyears afterfirstproduction.Vertical axisiscumulative discoveries foreachfield,expressedas Cumulative Discovery as % of Initial URR Estimate BERR 100% 150% 200% 250% 300% 350% 50% 0% 0 Curlew Captain Bruce Rob Roy Eider Hutton Beatrice Thistle MEAN 5 Harding Saltire Osprey Tern Clyde Maureen Auk Montrose 10 Years SinceDiscovery 15 studied in the United States, where it Reserve growth has beenmostclosely about their completeness prior to 2000. years arerequiredandtherequestions Also, industrydatabasesfromconcurrent inaccessible to most analysts (Section 2). most oftheworld'sfields,buttheseare estimates of cumulative 2P discoveries for world. Theindustrydatabasescontain a limited number of regions around the required dataisonlypubliclyavailable for reserves for subsequentyears. Butthe cumulative productionanddeclared for aspecificfield,withthesumof comparing the initial booking of reserves Reserve growthcanbeestimatedby growth Estimatingandforecasting reserve 3.2.2 Tiffany Arbroath Alwyn North Buchan Murchison Dunlin Mungo Nelson 20 25 Scott Miller Balmoral Hutton Northwest Tartan Heather Foinaven Gryphon 30 y 53

Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Chap3:UKERC 22/09/2009 10:12 Page 54

accounted for 89% of the additions to US of fields (Root and Mast, 1993). Both proved reserves over the period 1978 to annual and cumulative growth functions 1990 (Attanasi and Root, 1994).33 While can be calculated and used to convert reserve growth also occurs in other current estimates of cumulative regions of the world, the evidence base is discoveries into future estimates for a much thinner.34 But despite being specified year, with the amount of growth systematically investigated more than 40 depending solely upon the age of the field. years ago (Arrington, 1960), reserve Figure 3.7 shows two growth functions growth was relatively neglected before the estimated for onshore US oil fields 1980s (Drew, 1997).35 An important (Attanasi and Root, 1994; Verma, 2005). stimulus to further investigation was the Both show rapid growth in the years’ retrospective examination of discovery immediately following discovery and forecasts for the US, which were found to although growth subsequently slows, it is have systematically underestimated future still continuing some 80 years later. Such discoveries as a result of neglecting findings are typical for US 1P data: for example, Lore, et al. (1996) found that reserve growth at known fields (Drew and the estimated size of offshore fields in the Schuenemeyer, 1992).36 Gulf of Mexico doubled within six years of Future reserve growth can be estimated discovery and quadrupled within 40 years, through the creation of reserve growth while a later study by Attanasi (2000) functions based upon the measured suggested an eight-fold growth in 50 growth of a statistically significant sample years.

Figure 3.7 Cumulative reserve growth functions for US 1P reserve data

10

9 1995 U.S National Assessment

8

7 Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Global Oil Depletion: An assessment of the evidence for a near-term 6

5 Modified Arrington 4

3

2 multiple of initial estimate

Oil Cumulative growth factor 1 0 10 20 30 40 50 60 70 80 90 100 Years Since Discovery Source: Verma (2005)

33 Relevant references include Verma (2003; 2005), Schmoker (2000), Nehring (1984), Attanasi and Root (1994), Root and Mast (1993), Schuenemeyer and Drew (1994) and Klett (2005). 34 Relevant references include Klett (2005), Klett and Gautier (2005), Gautier and Klett (2005), Gautier, et al (2005), Klett and Schmoker (2003b), Klett and Verma (2004), Verma (2000; 2003; 2005), Verma and Ulmishek (2003), Verma, et al. (2004; 2001), Watkins (2002), Sem and Ellerman (1999) and Odell (1973) 35 Exceptions in this intervening period included the work of Hubbert (1967), Arps, et al. (1971), Marsh (1971), Pelto (1973) and several Canadian studies (OGCB, 1970). 36 The forecasting methodology relied upon estimates of the size of known fields but failed to adjust these to allow for future reserve growth. Since these fields subsequently doubled in size within less than ten years, the volume of new discoveries was greatly underestimated.

54 Chap3:UKERC 22/09/200910:12Page55 and Root (1994)foundthat‘lowquality’ significant change(Grace, 2007).Attansi one fifth shrank and the rest showed no half grewover theperiod1975to 2002, Gulf ofMexicofoundthatapproximately example, ananalysisof934fieldsinthe between fieldswithinthesameregion.For First, reserve growth varies widely obtained from the existing literature. available, someusefulpointerscanbe While there is little systematic evidence different ages,sizes andtypesoffield. between different regions and between Reserve growthmay beexpectedtovary Variations inreserve growth 3.2.3 growth. measured by water depth) and reserve correlation betweenoperating costs(as of Mexico, together with a negative and reserve growth in gas fields in the Gulf positive correlation betweengasprices Zampelli (2009) who found a strong factors was demonstrated by Forbes and drilling. Theimportanceofeconomic modifying the incentives for development could change the rate of growth by varying factorssuchasoilpriceswhich reserve growth and hence neglect time- as the sole explanatory variable for Watkins, 2002). Both approaches use age Zampelli, 2009; Sem and Ellerman, 1999; the dateoffirstproduction(Forbes and authors estimategrowthfunctionsfrom years later(Klett,2005).Hence,several has commenced–whichmay beseveral reserve growthoccursafterproduction development work that contributes to (e.g. Attanasi, 2000),muchofthe functions from the date of field discovery While many US studies estimate growth (Beliveau andBaker, 2003).Growthis influenced by changes in recovery factors OOIP whilelaterstage growthismore growth is more influenced by growth in seems likely thatearlystage reserve growth varies over thelifeofafield.It Third, thesourceandextentofreserve same size. growing much slower than US fields of the contributed to West Siberian fields (2003) found that lack of investment over time.Similarly, Verma andUlmishek conditions bothbetweencountriesand of NGLsandeconomicregulatory definitions, reporting practices, treatment field development practices, reserve explanations forthisincludedifferencesin Ellerman, 1999; Watkins, 2002). Possible fields (KlettandGautier, 2005;Semand offshore fieldsthanineitherUKorDanish greater reserve growth in Norwegian For example,studiesshowsignificantly even when theyaregeologicallysimilar. significantly from one region to another, Second, reserve growth also varies is invariably positive with 1P data. cumulative result forlargegroupsoffields be eitherpositive ornegative, the reserve growthforaparticularfieldmay problematic. Nevertheless, whilethe global average growth functions disparity makes theuseofregionalor been employed (Tennyson, 2002). This hydraulic fracturing andCO techniques suchassteaminjection, heavy oilsandlowpermeabilityinwhich confined to fields with solution gas drive, significant reserve growthwas largely study of 300 US fields showed that more than conventional fields, while a (notably heavy oil) fields grew five times 2 injection had 55

Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Chap3:UKERC 22/09/2009 10:12 Page 56

frequently very rapid immediately after significant correlation between field size field discovery reflecting continuing and reserve growth (Forbes and Zampelli, delineation of the reservoirs, but once 2009; Klett and Gautier, 2005).38 production is well-established growth Finally, some evidence suggests that derives more from implementation of EOR, onshore fields grow by more than offshore optimisation of well spacing and improved fields (Watkins, 2002) and older fields understanding of reservoir characteristics grow by a greater proportion than more (Verma and Ulmishek, 2003). recent discoveries. For example, Forbes Fourth, large fields grow more than small and Zampelli (2009) find a shift to a lower fields (see Figure 3.5 and Figure 3.6). For ‘reserve growth regime’ in the Gulf of example, Verma and Ulmishek (2003) Mexico after 1987, following the more found that Siberian fields with a URR widespread use of 3-D seismic techniques exceeding 1 Gb doubled in size in 19 that allowed more accurate estimation of years, while smaller fields increased by the size of newly discovered fields. Again, these results suggest that reserve growth only 19% over the same period. The may decline in the future as a greater production-weighted average for all fields share of production derives from newer was a 95% increase, since larger fields fields that are more likely to be located dominate total reserve additions. offshore. Similarly, Grace (2007) found that growing fields contained 80% of the discovered hydrocarbons in the Gulf of Mexico and 3.2.4 Estimating global reserve growth were on average six times larger than Most analysis of reserve growth has taken fields that shrank. One possible place in the US, where the SEC rules explanation is that smaller fields are more require the reporting of a particularly completely explored before the conservative interpretation of 1P reserves confirmation of reserves, leaving less that is confined to oil that is in contact scope for growth in the estimated OOIP, or Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Global Oil Depletion: An assessment of the evidence for a near-term with a well. As a result, authors such as that reserve reporting practices differ Laherrère (1999a) argue that the primary between large and small fields. But Grace source of observed reserve growth is also found that the dominant mechanism conservative reporting.39 In contrast, of reserve growth was the discovery of ‘optimists’ such as Mills (2008) highlight new reservoirs which is more likely to the historic and potential contribution of occur in large fields.37 These results improved technology. suggest that reserve growth may decline in the future (in both absolute and This disagreement was brought to a head percentage terms) as the average size of by the publication of the USGS World new discoveries declines. However, other Petroleum Assessment in 2000 which studies have found no statistically provided an authoritative estimate of the

37 There were 78 single reservoir fields, of which 33 shrank in size, 27 were static and 18 grew. 38 Forbes and Zampelli (2009) study the same region as Grace (2007), but focus on gas fields. 39 Reserve growth in the US may also be influenced by the production restrictions imposed in the 1950s and 60s (‘pro-rationing’) and by the underestimation of field size in the early days of exploration owing to inferior geophysical techniques.

56 Chap3:UKERC 22/09/200910:12Page57 growth in NGL resources. 40 Thiscorrespondstoa44%growth intheestimatedURRofcrudeoilfieldsdiscovered before1995andacorresponding56% suggests that 28% of the mean USGS additions throughnewdiscoveries. This 2003, ormorethantwicethereserve growth atnon-USfieldsbetween1995and Gb had been added through reserve Klett worked remarkably well. For example, suggests thattheUSGSassumptionshave resources, butsubsequentexamination to anoverestimate ofrecoverable a 1Pgrowthfunctionto2Pdatacouldlead associated uncertainty. Theapplicationof probability distributiontoaccountforthe this approach and assumed a triangular The USGSacknowledgedthelimitationsof estimated yet tofindresources. URR whichwas equivalent insize tothe 654 Gb to the mean estimate of the global Gautier, of the world (Attanasi, lack ofadequatedatafromotherregions on cumulative 1Pdiscoveries, owingtothe a growthfunctionestimatedfromUSdata field. The latter in turn were derived from factors thatdependedupontheageof estimates for non-US fields by growth multiplied the cumulative 2P discovery growth for the first time. The USGS included explicitallowance forreserve (USGS, 2000).The2000studytherefore increased by 26% between 1981 and 1996 fields outside the US were found to have cumulative 2Pdiscoveries for186giant was incorrect.For example,the evidence indicatedthatthisassumption of discovered fields, but a growing body of provided areasonableestimateoftheURR that cumulative 2P discovery data 6). TheUSGShadpreviouslyconsidered global URR of conventional oil (see Section et al et al . (2005)foundthatatotalof171 ., 1995).Thisprocessadded et al., 1999; 40 fields (Figure3.8). growth inthemorerecentlydiscovered discovered before 1986, with very little of the growth derives from fields and Iran (Figure3.9).Interestingly, most absolute terms deriving from Saudi Arabia (Table 3.3)withthelargestcontributionin varies widely fromonecountrytoanother years. Thepercentagereserve growth reserve growth hasincreasedinrecent 2007, suggesting that the rate of non-US have grown by 13.9% between 2000 and removed, pre-2000 fields are estimated to based upon1Preserves. IftheUSdatais with therestofdatabasesinceitis Canadian data which is not comparable The globalfigureincludesUSand 2000 fieldsgrewby11%over thisperiod. that cumulative 2P discoveries for pre- database (Figure 3.8). The results suggest 2007 iterations of the IHS Energy PEPS maintained, wecomparedthe2000and growth observed byKlett To check whether the rate of reserve data is particularly uncertain. from MiddleEastfieldswherethereserves biggest growthinabsolutetermsderived fields wherethedatawas poor. The databases andfromrevisedestimatesof previously omitted fields in the industry derive from factors such as the inclusion of of the apparent reserve growth could data’. This distinction suggests that much and theremainderto‘new and revised was attributed to ‘classic’ reserve growth between 1995and2003,ofwhich175Gb global totalof465Gbreserve growth Similarly, Stark and Chew (2005) found a assessment timeframe (1995-2025). been addedtointhefirst27%of estimate for non-US reserve growth had et al.isbeing 57

Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Chap3:UKERC 22/09/2009 10:12 Page 58

Figure 3.8 Reserve growth in the IHS Energy PEPS database between 2000 and 2007

2000 iteration

2007 iteration Cumulative Discovery

1950 1960 1970 1980 1990 2000 2010

Year Source: IHS Energy Note: Y-axis labelling withheld on grounds of confidentiality.

In summary, though the actual rates may suitability of such techniques for different be contested, it is clear that significant types of field and the rate at which they reserve growth is observed in cumulative may be applied remain key areas of discovery estimates based upon both 1P uncertainty. and 2P reserves. With 2P data, the global average reserve growth observed since 1995 seems broadly in line with the 3.3 Decline Rates assumptions made by the USGS (2000).

Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Global Oil Depletion: An assessment of the evidence for a near-term However, the global average is strongly An important determinant of future influenced by reserve growth in countries investment needs is the rate of decline of with the largest reserves where there is production from currently producing fields. less confidence in the accuracy of the Supply forecasts are more sensitive to data. Also, it is far from clear that the assumptions about the rate of decline than observed trend in reserve growth will be to assumptions about future oil demand, maintained in the future. Reserve growth but the former have generated appears to be greater for larger, older and controversy owing to lack of data onshore fields, so as global production (Simmons, 2000). Fortunately, three shifts towards newer, smaller and offshore fields the rate of reserve growth may recent studies have put a great deal of decrease in both percentage and absolute data into the public domain. This section terms. At the same time, higher oil prices examines the nature of production decline, may stimulate the more widespread use of summarises evidence on the rate of EOR techniques that have the potential to decline from different categories of field substantially increase global reserves. The and highlights the implications.

58 Chap3:UKERC 22/09/200910:13Page59 which they are developed (Figure 3.10). geology andlocation themannerin can vary widelydependingupon their The production cycle of individual fields Analysisofproduction decline 3.3.1 growth reserve global to between producers oflarge 2007 2000and Contribution 3.9 Figure Source: Table 3.3 Contribution of large producers to global reserve growth reserve global to between producers oflarge 2007 2000and Contribution 3.3 Table Abu Dhabi Kazakhstan Russia Nigeria Iraq Kuwait Qatar Venezuela Iran Saudi Arabia Country IHS Energy

% of World Total Reserve Growth

Saudi Arabia 10 15 20 25 30 35 40 0 5

Iran

Venezuela rwh20-07% growth reserve global % growth 2000-2007

Qatar 102.4 20.1 15.8 12.5 22.5 23.5 27.5 3.2 3.0 8.5

Kuwait facilities and/orthesteady development of capacity of pipelines and other surface plateau asaconsequence ofthelimited which may extendintoamulti-year production typically rises rapidly to a peak As afieldisbroughton-line,itsrate of

Iraq

Nigeria

Russia

Kazakhstan 11.5 15.6 37.0 1.0 2.9 3.3 3.5 4.5 4.7 9.0

Abu Dhabi 59

Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Chap3:UKERC 22/09/2009 10:13 Page 60

the field through additional drilling.41 The subset of fields that are in decline. Since length of plateau tends to be greater for the production cycle of individual fields is large fields and the production cycle can rarely smooth (e.g. a ‘bumpy’ plateau), be complicated by interruptions and the the point at which decline begins can be introduction of new technology. But at ambiguous. Some analysts also estimate some point, the rate of production will natural decline rates which exclude the begin to decline as a result of falling effects of capital investment. pressure and/or the breakthrough of water.42 Typically, more than half of the Production from individual wells, recoverable resources of a field will be reservoirs and fields is usually assumed to produced during the decline phase. decline exponentially at a constant rate, although there is no physical law requiring The term ‘decline’ is loosely applied at this and the rate of decline often falls various levels of aggregation, including during the later stages of the production single wells, reservoirs, fields, basins and cycle. Empirical equations to model countries. When applied to a region, it is production decline were first developed important to distinguish between the over a century ago and have since seen overall decline rate which refers to all wide application (Box 3.3).43 The currently producing fields, including those exponential model is the most widely that have yet to pass their peak, and the used, but it can underestimate production post-peak decline rate which refers to the during the later stages of a field’s life.

Figure 3.10 Stylised production cycle of an oil field

Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Global Oil Depletion: An assessment of the evidence for a near-term First Oil Discovery Well Plateau Abandonment Build Appraisal Up Well Decline Rate of Production

Economic Limit Time

Source: Höök (2009)

41 This involves trade-offs. For example, larger piplelines increase up-front costs, but speed production which allows rapid recovery of those costs. However, decline then begins earlier, thereby reducing the period the pipeline can operate at capacity and increasing operating costs. Conversely, smaller pipelines reduce up-front costs, lengthen the plateau, increase the time the pipeline can be operated at a capacity and slow the conversion of oil to cash. 42 In mature fields, the ‘water-cut’ may represent 90% or more of the volume of produced liquids. 43 See for example Chaudhry (2003), Porges (2006) and Guo, et al. (2007).

60 Chap3:UKERC 22/09/2009 10:13 Page 61

Box 3.3 Empirical equations to model production decline

Production decline from oil wells was first modelled by Arnold and Anderson (1908) and subsequently by Cutler (1924) and Larkey (1925) among others. These early studies were consistent with primary recovery being driven by the expansion of natural gas. Contemporary decline curve analysis has its roots in Arps (1945), who introduced

empirical curves defined by three variables: the initial rate of production (Q’(t0)), the curvature of decline ( ß ) and the rate of decline ( λ ).

Q’(t0) The general hyperbolic equation for the rate of production is: Q’(t) = λ 1/ß (1+ ß(t - t0))

λ - (t-t0) If ß = 0, this reduces to exponential decline: Q’(t) = Q’(t0)e

Q’(t0) If ß = 1, this reduces to harmonic decline: Q’(t) = λ (1+ (t - t0))

Decline models have since been developed in a variety of ways, including linearised curves (Li, 2003; Luther, 1985; Spivey, 1986), and the econometric analysis of residuals (Chen, 1991). Kemp and Kasim (2005) found that a logistic curve provided a better fit for UKCS fields. Decline curves are commonly used to estimate the URR of a field (Section 4.3).

Exponential Harmonic Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Global Oil Depletion: An assessment of the evidence for a near-term Hyperbolic Rate of production (Q’(t))

Time (t)

61 Chap3:UKERC 28/09/2009 13:23 Page 62 Page 13:23 28/09/2009 Chap3:UKERC 62 Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Source: year peak by stacked which Production from fields UKoffshore 1997, before peaked 3.12 Figure Source: ProductionFigure 3.11 from four UK oil fields bythree exponential fitted decline models (Wytch Farm). Forties canbe Ninian) and its largest onshorefield largest offshorefields(Forties, Brentand production profiles for the UK’s three As an illustration, Figure 3.11 shows BERR BERR

Production (b/d) production (b/d) 1000000 1500000 2000000 2500000 3000000 3500000 4000000 4500000 5000000 500000 100000 200000 300000 400000 500000 600000 0

1 8-6 -8 -10 1975 0

1980

1985 -4

1990 Years Beforeand AfterPeak Year -2 fields, but more oil was produced pre-peak. decline rate is similar to the off-shore poor fitofaround13%.Wytch Farm’s decline, Ninianby11%/year, andBrenta approximated witha9%/year exponential

1995 0 2 2000 4

2005 6 8 p.a. Decline at9% WYTCH FARM BRENT NINIAN FORTIES 13% p.a. Decline at 11% p.a. Decline at 10 Chap3:UKERC 28/09/2009 13:23 Page 63 Page 13:23 28/09/2009 Chap3:UKERC relates to both the size and age of the field. smaller onaverage, theobserved trend rates, but since these fields are also younger UKCSfieldshave steeper decline producing. Figure3.13suggeststhat decline phase and these fields are still cumulative production occurred during the peak declinerate. At least60%of of ~12.5%/year for the aggregate post- interrupted exploration andcontributedtotheearlierpeak inUKCSproduction. 1988. TheentireUKCSindustry was requiredtocarryoutasignificant programme of safety inspectionandupgrading which 44 Thisexcludes thePiperAlphafieldwhereproductionwas interruptedforseveral years followingadisastrousexplosionin from agloballyrepresentative sample of Three studieshave estimateddeclinerates decline rates Regionalandglobalaverage 3.3.2 in or before 1996. for agroupof77UKCS fieldswhichpeaked Figure 3.12showstheannual production 2005: 70 mb Note: Source: Decline curvesfor UKFigure offshore 3.13 fields entering peak intervals 5 year at Mean fieldsize: 1976-80:874mb;1981-85:5821986-90:4051991-95:881996-2000:802000- BERR

% of Peak Production 100 10 20 30 40 50 60 70 80 90 0 01 21 41 61 81 20 19 18 17 16 15 14 13 12 11 10 9 8 7 6 5 4 3 2 1 0 44 This gives anestimate Years PostPeak more slowly. The studies also agree that: production tend to be larger and decline decline because fields with higher weighted declineislessthantheaverage (CERA) (Table 3.5).Theproduction- 5.5%/year (Hook post-peak fieldstobe5.1%/year (IEA), weighted decline rate of their sample of These studiesestimatetheproduction- weighting. different approachestoproduction studies usecompetingdefinitionsand global production.Unfortunately, the fields whichaccountforaroundhalfof different sample,theyallincludethegiant fields (Table 3.4). While each study uses a et al 20% Decline 8% Decline Peaking 2001-2005 Peaking 1996-2000 Peaking 1991-1995 Peaking 1986-1990 Peaking 1981-1985 Peaking 1976-1980 .) and5.8%/year 63

Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Chap3:UKERC 22/09/2009 10:13 Page 64

Table 3.4 Comparison of global decline rate studies

IEA Hook et al. CERA No. of fields in sample 651 331 811 (54 supergiant, 263 (all giant) (400 large and giant, 334 large) above) No. post-peak fields 5801, 2 2613 - % of total production ~58% ~50% ~66% of crude oil in sample

Cumulative discoveries 1241 Gb 1130 Gb 1155 Gb of crude oil in sample Definition of plateau Production >85% of Production >96% of Production >80% of peak peak peak Definition of onset of After year of peak After last year of After last year of decline production plateau plateau Production weighting Cumulative Annual production Annual production production4

Source: IEA(2008), CERA (2008) and Höök, et al.(2009; 2008; 2009a; 2009b). Notes: 1 101 fields in plateau (production >85% of peak), 117 fields in ‘phase 1 decline’ (production >50% of peak), 362 fields in ‘phase 3’ decline (production <50% of peak) 2 387 onshore, 264 offshore, 185 OPEC and 466 non-OPEC. 3 261 onshore, 214 offshore, 143 OPEC and 188 non-OPEC. 4 IEA weights by annual production when estimating historical trends in decline rates.

• Decline rates are lower for OPEC fields greater proportion of their URR during and particularly for Middle East fields the decline phase (IEA, 2008). (Table 3.5). This partly reflects Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Global Oil Depletion: An assessment of the evidence for a near-term Importantly, both the IEA and Höök, et al. differences in average size, but also find decline rates to be significantly higher quota restrictions and disruptions from for newer fields (Figure 3.14). The IEA political conflict. argues that newer fields build up more • Decline rates are higher for offshore quickly to a higher plateau that is fields (Table 3.5). These tend to be maintained over a shorter period of time, produced at higher rates in order to but Höök, et al. (2009a) show that the recover their higher fixed costs, leading length of plateau for giant fields has to higher peaks, shorter plateaus and increased together with the proportion of steeper declines. the remaining recoverable resources produced prior to peak. They argue that • Decline rates are lower for larger fields new technology allows the plateau to be and are particularly low for the super- maintained for extended periods of time, giant fields in the Middle East (Table but at the cost of more rapid decline 3.6). Large fields reach their peak later following the peak (see also Gowdy and than small fields, but also produce a Roxana, 2007). The collapse of production

64 Chap3:UKERC 28/09/2009 12:52 Page 65 Page 12:52 28/09/2009 Chap3:UKERC post-plateau fieldsis5.8%. The production-weighted average forphase1isstronglyinfluencedbyGhawar. Theproduction-weighted sampleaverage for Note: Source: Source: sizes of post-peak field(%/year)sizes of post-peak IEATable 3.6 of estimates aggregate production-weighted decline for rates different fields (%/year)post-peak of production-weighted Estimates Table 3.5 aggregate decline for rates samples of large have access to the full study. Note: Source: Evolution of production-weightedFigure 3.14 giant oilfield decline over time rates Note:Figures for most recent decade less certain since sample of fields is much smaller All fields OPEC Non-OPEC Offshore Onshore Total OPEC Non-OPEC Offshore Onshore Parameter Studies use different data sets, definitions and methods of production weighting. Details missing forCERAsincewedonot Studies usedifferentdatasets,definitionsandmethodsofproductionweighting.Details The production-weighted decline rate is 1.4% in decline phase 1, 3.6% in decline phase 2 and 6.7% in decline phase 3. IEA(2008) IEA(2008), CERA(2008)andHöök, Höök,

et al. Production-weighted average decline rate (%) 10 12 14 16 (2009b) 0 2 4 6 8 Pre-1960 Total 5.1 3.1 7.1 7.3 4.3 Onshore 1960s et al.(2009; 2008; 2009a; 2009b). IEA 5.1 3.1 7.1 7.3 4.3 Supergiant Offshore 1970s Decade ofPeak 3.4 2.3 5.7 3.4 3.4 Höök, Höök, Non-OPEC 1980s 5.5 3.4 7.1 9.7 3.9 et al. et Giant 6.5 5.4 6.9 8.6 5.6 1990s OPEC 2000- CERA 5.8 Other - - - - 10.4 10.5 11.6 9.1 8.8 65

Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Chap3:UKERC 22/09/200910:13Page66 66 Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production requirements incorrectly. they appear to calculate the capacity figures areabsentfromtheIEAstudyand fields, includingthoseinbuild-up. Both weighted aggregatedeclinerate ofall post-peak fields,ortheproduction- either the proportion of production from required eachyear itisnecessarytoknow But toestimatetheadditionalcapacity discovery anddevelopment ofnewfields. development of‘fallow’fieldsorthe producing fields(Box 3.2),the replaced by investment in EOR at This declineinproductionneedstobe may decline by more than this in 2009. recession, productionfromthesefields reduced asaresultofthe2008economic 4.5%/year. compares to CERA’s estimate of the latterfiguretobe~4.1%/year which a globalaverage. Theresultappearsconsistentwiththe IEA’s graphs (Figure3.15). 47 TheIEAprovide thisfigureforOPEC(3.3%)andnon-OPEC fields (4.7%)sowesimplyweightby2007productiontoobtain decline rate ofallfields,includingthoseinbuild-up. mb/d) toestimateanannualloss ofoutput4.7mb/d.Butthecorrectprocedureistouseproduction-weighted aggregate 46 The IEA appear to multiply the production-weighted decline rate of post-peak fields (6.7%) by global crude production (70 estimated tohave grownby1%over the lastfive years. 45 Theyfurtherestimatethatthisis2.3%lessthantheproduction-weighted natural declinerate forthesefields.Thelatter average declinerate of IEA estimate a production-weighted global for thesampleoflargefields(10.4%), rate forsmallerfields is thesameasthat the optimisticassumptionthatdecline than thatoftheglobalpopulation.Under size of each sample of fields is greater rate for all post-peak fields since the mean underestimate theglobalaverage decline The figuresonthesepagesarelikely to nitrogen injection is a notable example. at Canterellfollowingextensive useof post-peak fields. added by new investment each year, approximately 3 mb/d of capacity must be 47 45 This implies that With capitalinvestment 6.7%/year 46 We estimate for all post-peak decline.Significantinvestment production methodsleadtomorerapid and offshore fields and as changing production shiftstowards smaller, younger trend asmoregiantfieldsenterdecline, least 4%/year. Both areonanupward rate of all currently producing fields is at 6.5%/year and thecorrespondingdecline rate ofpost-peak fieldsisatleast In summary, the global average decline (Höök and Aleklett, 2008). converge onthe (higher) average rate production-weighted declinerate to and theobserved tendencyforthe increasing declinerates inthegiantfields optimistic, given thetrendtowards et al. (43 mb/d)(Figure3.15).However, Höök, estimated lossof61%currentcapacity to 8.5%/year by2030,leadingtoan decline rate ofpost-peak fieldstoincrease production-weighted globalaverage offshore fields. The IEA expects the production fromyounger, smallerand this period, with a growing proportion of Most existingfieldswillenterdeclineover expected todevelop intheperiodto2030. how globalaverage declinerates may be A critical question for supply forecasting is growth (Figure3.15). year tomeetthe IEA’s forecast ofdemand additional 1mb/dwillberequiredeach coming onstreamevery threeyears. An levels –equivalent toanewSaudiArabia simply tomaintainproductionatcurrent (2009a) consider this estimate to be is Chap3:UKERC 22/09/200910:13Page67 confused with year to the next. They should not be in theratea fieldfromone ofproduction Decline rates are a measure of the change Depletionrates 3.4 favourable. even if‘above-ground’ conditionsprove 2.8), thiswillpresentamajorchallenge term decline in new discoveries (Figure keep productionconstant.Given thelong- need tobereplacedby2030,simply current crude oil production capacity will optimistic. Ifso, more thantwothirdsof be made that the IEA’s assumptions are rates aredifficulttoforecast,acasecould increase. Whilefuturetrendsindecline the economic slowdown) decline rates will forthcoming (forexample,asaresultof natural decline rates and if this is not is neededsimplytooffsettheunderlying Source: IEA production of 2030 forecast to globalFigure 3.15 all-oil measure oftherate atwhichthe mb/d 100 120 20 40 60 80 IEA (2008) 0 1990 depletion rates 2000 which area 2010 the URR. He then shows the close links cumulative productionfromanestimate of the latter is calculated by subtracting remaining recoverable resources,where the ratio of annual production to Höök (2009) defines the depletion rate as rates. leading tolowerestimatesofdepletion over time - with higher resource estimates estimates whichvary betweensourcesand are based upon uncertain resource can be measured precisely, depletion rates production (R/P)ratio. Whiledeclinerates inverse ofthemorefamiliarreserve to reserves, the depletionrate issimply the URR. Whendefinedinrelationto1P allowing forfuturereserve growth)orthe the remaining recoverable resources (i.e. could bethe1Preserves, the2Preserves, recoverable resources,wherethelatter annual production to some estimate of individual fieldisdefinedastheratio of are being produced. The depletion rate of recoverable resourcesofafieldorregion 2020 2030 producing fields Crude oil-currently to bedeveloped Crude oil-fieldsyet to befound Crude oil-fieldsyet additional EOR Crude oil- Non-conventional oil Natural gasliquids 67

Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Chap3:UKERC 22/09/200910:13Page68 68 Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Table 3.7 Estimated depletion peakTable at 3.7 and annual depletion peak at rate for giant oil fields other thanphysical depletionandinsomecasesthepeakmay subsequentlybeexceeded. Russia. Itisimportanttonotethattimingofthepeakproduction formany ofthesecountriesmay beinfluencedbyfactors estimated URRexcluding reserve growth.Post-peak countriesforwhichURRestimateswerenotavailable wereexcluded, aswas 48 SincetheUSGSonlyestimate reserve growth atthegloballevel, thiswas allocatedbetween countriesinproportiontotheir Source: Höök, for OPECfields. rate ishigherforoffshorefieldsandlower rates, themaximumobserved depletion 7.2%/year (Table 3.7). As with decline band, withaproduction-weighted meanof typically fallswithinarelatively narrow maximum depletionrate ofgiantoilfields Höök, reached justpriortotheonsetofdecline. hyperbolic the maximum depletion rate is equals the decline rate, while if decline is decline isexponentialthedepletionrate URR estimate remains unchanged. If remains constant or falls - provided the decline begins, the depletion rate either phase asreserves areproduced.Once increases duringthebuild-upandplateau rate ofafield.Thedepletion rate generally between the depletion rate and the decline average beinghigherforoffshorefields sample of giant fields is 37%, with the The production-weighted meanoftheir URR produced at the onset of field decline. depletion at peak, or the proportion of All figures production-weighted Depletion = Ratio of cumulative production to estimated ultimately recoverable resources Depletion rate =Ratio of annualproductiontoestimatedremainingrecoverable resources Notes: All fields OPEC Non-OPEC Offshore Onshore Höök, et al. et al. et al. (2009b) alsoestimatethe (2009b) (2009b) show that the Depletion peak at 36.8% 31.5% 37.4% 44.0% 34.1% peak of22%,aproduction-weighted mean estimate asimplemeanfordepletionat post-peak countries(seeSection6),we the USGS estimates of regional URR for 37 peak ofproduction(Figure2.19).Using for the countries that have passed their particular interest are the values at peak greater sincetheyalsoincludetheYTF. Of resource estimates will necessarily be the uncertaintyonrecoverable estimated attheregionallevel, although Depletion anddepletionrates canalsobe 78% in North America. varying from37%intheMiddleEastto 48% depleted, with the regional average estimates thatgiantfieldsareonaverage resources have been produced. than onethird)oftheirrecoverable decline whenlessthanhalf(andoften production frommostfieldsbeginsto fields. Theseresultsdemonstrate that large fieldsto25%forsmalloffshore who findvalues ranging from15%for analysis isconductedbytheIEA(2008) and lowerforOPECfields.Asimilar Depletion peak at rate 11.0% 7.2% 5.3% 8.7% 5.8% The IEA Chap3:UKERC 22/09/200910:13Page69 estimates totheassumedURR. estimate adepletionrate atpeakof6.9%assuminga lowerURRof35Gb. Thisdiscrepancy highlightsthesensitivityofthese 49 We estimatethe depletion rate atpeakfortheUKtobe4.4%assuming a URR of 43 Gb. Incontrast, Aleklett, of 24% and a maximum of 52%. accuracy oftheestimatesresourcesize. thumb’ depends very much upon the However, theusefulness ofthese‘rules-of- URR ofthatregionhasbeenproduced. until significantlymorethanhalfofthe that delay regionalpeaksofproduction justification. Thesameappliestoforecasts producing regions will require careful previously experiencedinotheroil- are significantly higher than those forecast thatimpliesdepletionrates that (Aleklett, useful ‘reality check’ on supply forecasts peak. Hence,bothmeasurescanprovide a the URR that can be produced prior to the for afieldorregionandtheproportionof constraints on both the rate of depletion physical, technicalandeconomic This analysis suggests that there are approximately 1.2%. global average depletionrate is than the maximum rate. At present, the production cycle is typically much lower average depletion rate over the full been much less than 5%/year. Also, the depletion rate foraregion has typically outlier). to be5.2%/year (forBulgariawhichisan mean to be 1.9%/year and the maximum to be 2.1%/year, the production-weighted depletion rate atpeak forthesecountries In a similar manner, we estimate the mean produced. their recoverableresourceshavebeen have reachedtheirpeakwellbeforehalfof other words, 49 et al. In other words, the maximum most countries appear to , 2009).Specifically, a 48 In engineering and economic limits. while thelatterissubjecttophysical, trend of declining discoveries (Figure 2.8) However, the former runs counter to the which theyaredepletedneedstoincrease. added fromthesesources,ortherate at term, eithertherate atwhichreserves are rates increase in the medium to long- required. Asdemandgrowsanddecline reserve additions of ~80 Gb/year are current globalaverage forallproduction), of these resources is only 1.2%/year (the maintained. Ifinsteadthedepletionrate these sourcesifglobalproductionistobe for ~20Gb/year ofreserve additionsfrom ~5.0%/year, thisleadstoarequirement Using apeakdepletionrate of maintain productionatcurrentlevels. fields andnewdiscoveries simplyto fallow fields, reserve growth at existing replaced by a combination of developing capacity. This capacity needs to be mb/d or ~1.0 Gb/year of production rate of4.1%impliesan annuallossof2.9 fall. For example,aglobalaverage decline production inaregionmay beexpectedto production decline, then aggregate capacity anticipated to be lost through the productoftwoislessthan multiply by an assumed depletion rate. If rate ofproduction it isnecessary to Gb/year, totranslate thisintoafeasible common to estimate reserve additions in and therate ofproduction. Whileitis reserve growthand/ornewdiscoveries bridge betweenestimatesoftherate of Depletion rates can also provide a useful et al .(2009) 69

Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Chap3:UKERC 22/09/200910:13Page70 70 Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production • • • 3.5 Summary cumulative discovery estimates based growth continues to be observed in some ‘pessimists’, significant reserve reporting. Contrary to the claims of not primarilytheresultofconservativ Reserve growth is real, significant and the giantfields. potential forincreasedrecovery from significance to future supply than the small fieldsshouldbeofless estimates oftheresourcescontainedin returns. Asaresult,thecompeting will besubjecttorapidly diminishing develop and theexploitationofrest the smallest will remain uneconomic to make moresmallfieldsviable,many of improvements andhigherpricesshould distribution. Moreover, whiletechnical better describes the population size or ‘parabolicwhether a‘linear’ fractal’ insufficient evidencetoconclude the resultofsamplingbias,thereis distribution ofdiscovered fieldsispartly the observed lognormalsize fields continuestobedisputed.While contained withinsmall,undiscov The proportionoftotalresources future global supply. therefore ofcriticalimportancefor and thepotentialforreserve growthis fields, their future production profile or so. The remaining reserves at these begin todeclinewithinthenextdecade production andmostoftherestwill many are well past their peak of Most ofthesefieldsarerelatively old, two thirdsofcumulative discov oil, whileupto500fieldsaccountfor half of the global production of crude Around 100oilfieldsaccountforupto eries. ered e • • line with the contro observed since 1995 seems broadly in The globalaverage reserve growth contribution to future global oil supply. techniques shouldmake amajor geological knowledge and recovery reserve growthfromimproved upon 2P reserves. This suggests that more rapid post-peak decline. More changing productionmethods leadto younger and offshore fields and as production shiftstowards smaller, as moregiantfieldsenterdecline, 4%/year. Bothare onanupward trend currently producingfieldsisatleast corresponding declinerate ofall is at least 6.5%/year and the a fields. Theproduction-weighted global decline ofproductionfromexisting needs andfuturesupplyistherate of A criticaldeterminantofinvestment uncertainty. be appliedremainkey areasof of fieldandtherate atwhichtheymay techniques for different sizes and types reserves. The suitability of such to substantiallyincreaseglobal EOR techniques that have the potential stimulate the more widespread use of the same time, higher oil prices may both percentageandabsoluteterms.At rate ofreserve growthmay decreasein newer, smallerandoffshorefieldsthe as globalproductionshiftstowards for larger, olderandonshorefields,so reserve growthappearstobegreater and thepoorestqualitydata.Also, countries withboththelargestreserves influenced by reserve growth in the global average is strongly made bytheUSGS(2000).However, verage declinerate ofpost-peak fields v ersial assumptions Chap3:UKERC 22/09/200910:13Page71 • • resources have been produced. Hence, peak wellbeforehalfofthereco countries appeartohave reachedtheir To date,mostgiantfieldsand justification. rates are likely to require careful that assumeorimplyhigherdepletion lower thanthis.Hence,supplyforecasts the productioncycle istypicallymuch ~5%. The average depletion rate over regions the maximum observed rate is ~11% foroffshoregiantfields,while been ~6%foronshoregiantfieldsand maximum observed depletionrate has recoverable resources. Historically, the upon uncertain estimates of the size of estimates oftheserates arecontingent depletion ofafieldorregion,although economic constraints upon the rate of There arephysical, technicaland likely toprove extremelychallenging. production constant.At best,thisis replaced by2030,simplytokeep production capacity may need to be than two thirds of current crude oil verable • further. figure will need to increase even grows and decline rates increase, this to be required. Moreover, as demand higher rate ofreserve additionsis likely lower depletionrates, asignificantly production from existing fields. With compensate for the decline in at least20Gb/year willberequiredto producing region,reserve additionsof rate previously seen in any oil- is as high as the maximum depletion depletion rate for these new resources at existingfields.Iftheaverage of fallowfieldsandenhancedrecovery from new discoveries, the development the requiredr Depletion rates can be used to estimate decline. associated with more rapid post-peak patterns that delay the peak may be justification. Inaddition,development of declinewillalsorequirecareful resources of a region prior to the start the productionofmorethanhalf supply forecasts that assume or imply ate ofreserve additions 71

Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Chap3:UKERC 22/09/200910:13Page72 72 Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Chap4:UKERC 22/09/200910:14Page73 and viceversa (Bartlett,2000). While optimistic forecasts for future oil supply, larger estimates of URR lead to more is because,otherthingsbeingequal, resource play a very prominent role. This despite this,disputesover thesize ofthe absolute size of the oil resource. But rate The peak oil debate is primarily about the resources ultimately recoverable Theimportanceof 4.1 concludes. they may be addressed. Section 4.8 statistical issuesraised andshowshow while Section4.7highlightssomeofthe techniques produce consistent results, producing regionstoassesswhetherthese uses illustrative datafromanumber ofoil- strengths andweaknesses.Section4.6 contemporary applicationandmajor detail, includingtheirhistoricalorigins, extrapolation of historical trends in more investigatebased uponthe thetechniques estimate theURR.Sections4.3to4.5 summarises the methods available to to futureoilsupply, while Section4.2 Section 4.1 discusses the relevance of URR Report 5 techniques is contained in comprehensive examinationofthese criticised by those that are not. A concerned about‘peakoil’andregularly techniques are widely used by those trends inproductionordiscovery. Such particular ontheextrapolation ofhistorical resources (URR) in a region, focusing in estimating theultimatelyrecoverable This section examines the methods for 4. of global production rather than the Looking beneath-methodsofestimatingultimatelyrecoverable resources . Technical characteristics oftheregionand will dependuponthe in turn marginal increasesmay beexpected.This resources fromaregion,orwhetheronly significantly increasetherecoverable such developments can be expected to inaccessible resources.At issue iswhether factors and allow access to previously reduce production costs, improve recovery as inducing technical improvements that marginal resourcesmoreprofitableaswell example, increasingpricesshouldmake URR isanendogenousvariable: for 1997). Whilenotalways treatedassuch, much oftheongoingcontroversy (Rogner, produce singlevalue estimates,explains issues, together with the tendency to uncertainty. Lackofclarityover such assumptions and the associated range of the relevant technicalandeconomic timeframe forwhichtheestimateismade, of thehydrocarbons covered, the Any estimateofURRrequiresspecification estimate fitting’ techniquescanalsobeusedto region isrelatively advanced, such‘curve- 2008). However, provided exploration ina reach theirpeakofproduction(Caithamer, especially forregionsthathave yet to would be more difficult to perform, Without thisconstraint, suchprojections the area under the curve (see Section 5). estimate oftheregionalURRtoconstrain and projectingthisforward usingan curve to historical data on oil production Hubbert forecastfuturesupplybyfittinga 1999b). Campbell (1997)andLaherrère(2003; and of subsequent authors such as they arecentral tothe workofHubbert play a useful role in supply forecasting, reject thenotionthatsuchestimatescan some economists(e.g.Adelman,1991) the URR. 73

Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Chap4:UKERC 22/09/200910:14Page74 74 Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Box 4.1 Geological levels of levels aggregation Geological Box in petroleum 4.1 resource assessment or less suitable for different levels. 4.1), with different techniques being more individual wellstothewholeworld(Box levels of aggregation ranging from Estimates ofURRmay bedeveloped for the rate ofpost-peak decline. production, althoughtheycouldreduce influence onthetimingofpeak long-term, theymay have littleorno that resources are only accessible in the technical oreconomicconstraints mean timeframe under consideration. If being progressively buried, heated and compressed. Sedimentary basinsare theprimarysourceofpetroleum,asaresult oforganiccarbon sedimentary orvolcanic rocksandwhichisknownorexpected tocontainhydrocarbons. Petroleum Basin: defined geographic boundary. prospects whichsharecertaincommongeological attributesandliewithinsomewell- Petroleum Play: economic considerations, such as the availability of leases for exploration. target forexploration. Theboundariesofaprospectmay alsobeinfluencedbylegaland containing reservoirs of recoverable hydrocarbon and is considered to be a suitable Petroleum Prospect: producing field may range from one to thousands. discovered, under development, producing or abandoned and the number of wells in a approximately contiguousareawhenprojectedtothesurface.Oilfieldsmay eitherbe reservoirs in asinglefieldmay beseparated vertically orlaterally butforman all relatedtoasinglegeologicalstructureand/orstratigraphic feature.Individual Petroleum Field: has a single natural pressure system. whether discovered or not, which is physically separated from other reservoirs and which Petroleum Reservoir/Pool: using standard formulae for decline curves (Arps, 1945; Chaudhry, 2003). URR of a producing well is typically calculated by extrapolation of its past performance, some wells being drilled to inject fluids to enhance the productivity of other wells. The Petroleum Well: A play isanareaforpetroleumexploration, containingacollectionof A fieldisanareaconsistingofasinglereservoir ormultiplereservoirs, A wellmay bedrilledtofind,delineate and producepetroleum,with A basinisasingleareaofsubsidencewhich filledupwitheither A prospectisageologicalanomalythathassomeprobabilityof A reservoir isasubsurfaceaccumulationofoiland/orgas used (Charpentier, 2003). assessments whenonlyaggregatedatais boundaries cangreatlycomplicateresource lack ofgeologicalhomogeneitywithinsuch neighbouring countries or regions. The 4.1) thatmay alsoextendinto several geologicallydistinctareas(seeBox level, but these typically encompass frequently made at the country or regional arithmetically (Section2).Estimatesare only meanestimatescanbesummed estimates developed atlowerlevels, but Aggregate estimatesmay bederived from Chap4:UKERC 22/09/200910:14Page75 other informationand themethodsare There isrelatively littleuseofgeological or regions such as an oil-producing country. cumulative productionforaggregate data oncumulative discovery or extrapolation ofcurves fittedtohistoric producing single-value estimates from the Hubbert may becharacterised as Most ofthemethodsassociatedwith (Ahlbrandt and Klett, 2005; Divi, 2004). from acombinationofmethods reliable estimates are likely to be derived human resources available. The most region understudyandthedata choice depends upon the nature of the the basictechniques.Theappropriate estimating URRandmany variations on There areavariety ofmethodsfor oftechniques Overview 4.2 Sources: resources. were known to contain petroleum and 76 were estimated to contain 95% of discovered resource assessment.Globally, theUSGS(2000)identified937provinces, ofwhich406 largest entitydefinedsolelyonthebasisofgeologicalconsiderations thatisrelevant for single basinorpetroleumsystemseveral similarbasins/systems.Aprovince isthe considered separate because theyhave differenthistories.Aprovince may containa to petroleum formation. Adjacent provinces might have the same original rocks, but be Petroleum Province: or the latter may be broken down into several AUs. resource assessment methodology. An AU may coincide with a single petroleum system, exploration considerations, accessibility and risk to be examined with a particular petroleum system that is sufficiently homogeneous, both in terms of geology, Petroleum Assessment Unit: the USGS. forms thebasisofresourceassessmentsconductedby introduced byDowandnow whose provenancea singlepodofactive is sourcerock”.Theconceptwas first well asallgeneticallyrelatedhydrocarbons thatoccurinpetroleumaccumulations Petroleum System: Energy Information Administration (1990); Klett (2004); Magoon and Sanchez (1995) A petroleumsystemis“….theessentialelementsandprocessesas A province isanareawithcommongeologicalpropertiesrelevant An assessment unit (AU) is a volume of rock within a seismic andotherdata. Atraditional rely uponthegeological analysisof For less-explored regions,estimatesmust and Schuenemeyer, 1993). stage ofexploration andproduction(Drew regions thatareatarelatively mature overlaps between the two, especially for however, asthere areconsiderable characterisation isanoversimplification, inaccessible tothirdparties.This upon extensive data sources that are often complex andresource-intensive andrely (USGS, 2000). These methods are information andstatisticaltechniques regions, withextensive useofgeological geological assessments of disaggregate produce probabilistic estimates from the methods used by the USGS and others available in the public domain. In contrast, simple to apply using data that is often 75

Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Chap4:UKERC 22/09/200910:14Page76 76 Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production some methods(e.g.Pareto size distribution)beingconsistentlyoptimistic. obtained. Ahlbrandt and Klett (2005) found considerable variation in the results of seven methods applied to seven regions, wit extrapolation ofhistoricaltrendsmay bemore conservative. However, fewstudieshave systematicallycomparedtheresults 50 Geologicalassessmentsareoften consideredtoproducerelativelywhile techniquesbaseduponthe estimates, optimistic the fullrange ofopinions(Gautier, 2004; into aprobabilitydistributionthatreflects relevant factors whicharethencombined a probability distribution for each of the and thenestimateseitherasinglevalue or geologist reviewstherelevant information (Baxter, the expert judgment of several geologists Estimates may alsobemadebycombining view: large confidence bounds. In Hubbert’s such estimates must necessarily have and forunexploredareas(e.g.theArctic), be available forasubsetofthesevariables saturation (Capen,1976).Datamay only such asporevolume, porosityandoil estimates ormeasurementsofvariables use MonteCarlomethodstomultiply lower levels ofaggregationandtypically available. Other approaches are applied at more informationis regions where measurements fromgeologicallysimilar calculations are based upon For unexplored areas, the values for such Weeks, 1952; White and Gehman, 1979). per cubic kilometre (Gautier, 2004; volume byanestimatedyieldinbarrels by multiplyingtheestimatedsedimentary level, istoestimatehydrocarbon volumes approach, commonly applied at the basin of magnitude.” (Hubbert, 1982) uncertainty oflessthanseveral orders primary areathathasarange of recoverable oilobtainablefroma permit an estimate to be made of the than thatprovided bydrilling,thatwill geological information exists, other “.… it is easy to show that no et al., 1978). Typically, each on discovered fields. estimates canbeobtainedbyusingdata For well-exploredregions,morereliable of the individual assessors. heavily ontheknowledgeandobjectivity However, itlackstransparency andrelies methods (Charpentier, that may bepoorlyhandledbyother exploration constraints andotherfactors availability andcanaccommodate for alllevels ofaggregationanddata White, 1981).Thismethodisappropriate • approaches: distribution (Charpentier, etal., 1995; viable field size and appropriate size as to assumptions about the minimum curtailed whenfittingthecurve, aswell which theobserved size distributionis This estimateissensitive tothepointat (Cramer BartonandLaPointe, 1995). calculating theareaundercurve this tosmallerfieldsizes and fitting alinearregression,extrapolating frequency distribution on a log scale, estimated by plotting a cumulative undiscovered resourcesmay be law size distributionisassumed, (Section 3).For example,ifapower- of the underlying population of fields assumptions aboutthesize distribution on thesampleofdiscovered fieldswith URR may bederived bycombiningdata Field-size distributions: estimates theURRfromasymptote (where the largest field is rank 1) and as afunctionoftherank ofthefield approach plots cumulative discoveries Laherrère, 2000a).Analternative 50 There arethree et al Estimates of ., 1995). h Chap4:UKERC 22/09/200910:14Page77 discovery projectiontechniqueswouldbeidentical. 51 Ifthediscovery rate was constantandiffieldswerediscovered preciselyindescendingorderofsize, thefield-rank and • • fitting techniques is that they do not drilled. Amajoradvantage ofcurve cumulative numberofexploratory wells exploratory effort, such as the may either be time or some measure of (‘yield’), whiletheexplanatoryvariable discoveries ortherate ofdiscoveries the rate ofproduction,cumulative variable may becumulative production, estimate theURR.Theexplained region andextrapolate thecurves to trends indisco techniques tofitcurves tothehistoric Curve-fitting: Fuller, 1992a; b). rate of exploratory drilling (Power and together with the anticipated success and sequenceoffuture discoveries to provide forecasts of the number, size of new field discovery and can be used probabilistic lawgoverning theprocess Several ofthesemodelssimulate a (Schuenemeyer and Drew, 1994). information about field location field size distribution and/or combined withassumptionsaboutthe Demirmen, 1981).Thisissometimes 1985; Kaufman, 1975;Meisnerand Roberts, 1958;Forman andHinde, of exploratory wellsdrilled (Arpsand exploratory effort,such asthenumber sequence orsomemeasureof a functionofeithertime,thediscovery number andsize ofdiscovered fieldsas involv Discovery process modeling: described below. ‘discovery projection’ technique approach has much in common with the to whichthecurve istrending.This e statisticalanalysesofthe v These use regression ery orproductionina 51 These can leadtoinconsistencies inthetime- exploration (e.g.newplays withinabasin) of fieldsortheopeningupnewareasfor done, the mixing of different populations for legal or political reasons). If this is not without areasbeingclosedtoexploration unrestricted exploration history(e.g. areas thathave hadarelatively applied togeologicallyhomogeneous The extrapolation techniques are best upon patterns such as this. All of the curve-fitting techniques rely taken as an estimate of the regional URR. trends towards anasymptotewhichcanbe newly discovered fieldsfalls,thecurve a functionoftime.Astheaverage size of plots cumulative discoveries inaregionas production cycle. For example,Figure4.1 and the‘shape’ofdiscovery or the size distributionofdiscovered fields These featuresshouldbereflectedinboth product of increasingly greater effort. progressively smaller and often the subsequent discoveries being discovered relatively early, with fields; and that large fields tend to be being locatedinasmallnumberoflarge highly skewed, withthemajorityofoil assume that:thefieldsize distributionis These three‘extrapolation’ techniquesall Evans, 2008). et al Cleveland andKaufmann, 1991;Imam, about ‘peakoil’(Campbell,2002a; including inparticularthoseconcerned developed bynumerousanalysts, subsequently been adopted and (1956; 1959; 1962; 1982) and has fitting was pioneeredbyHubbert require dataonindividualfields.Curve ., 2004; Laherrère, 2003; Mohr and 77

Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Chap4:UKERC 22/09/200910:14Page78 78 Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production of Mexico, leadingtosteadyincreasesintheyieldfrom exploratory drilling over thelast35years. 1977). structural breaks in a time-series (Harris, horizontal drilling in Texas ) can lead to and politicalfactors (e.g.theadvent of homogeneous, various technical, economic even if the regionisgeologically 54 For example, Managi, areas intheRockies areoff-limits forenvironmental reasons (Mills,2008). countries, buttheArcticNational WildlifeRefuge, theeastern Gulf ofMexico, much ofthewesternoffshoreandmany onshore 53 Exploration isvery rarely unrestrictedinaggregateregions.For example,theUSimposesfarfewerrestrictions thanmost 1977). so shallowfieldsmay beover-represented anddeeperfieldsunder-represented inthedistributionsofdiscovered fields(Harris 52 Thesameappliestoexploratory depth.Shallowfieldsareusuallymoreextensively exploredandexploitedbeforedeepfields (Wendebourg andLamiraux, 2002). extrapolating historicaltrends estimating size distributions and series and undermine the basis for Note: Source: in discovery aregion Cumulative asafunctionFigure of 4.1 time oil-producing regions future. This condition willnotapply inall influences - and continues to do so in the if physical depletion outweighstheother expected toprovide reliableURRestimates extrapolation ofthosetrends canonlybe political andinstitutional influences, the technical changeandnumerouseconomic, the neteffectofphysical depletion,

Cumulative Discovery (Gb) Name of region withheld on grounds of confidentiality. IHS Energy 10 12 14 16 53 0 2 4 6 8 1900 Since allhistorical trendsreflect et al 1920 54 . (2005) shows how technical improvements have more than offset resource depletion in the Gulf although depletion 1940 52 But 1960 Year discuss each group of techniques in turn. (Section 4.7).Thefollowingsections the specification,butgenerally arenot variables can and should be included in explanatory variables (Table 4.1).Other choice anddefinitionoftheexplained aggregate data,buttheyvary intheir extrapolating historicaltrendsin data. ThesetechniquesestimateURRby and therelative availability oftherelevant most widelyusedowingtotheirsimplicity curve-fitting techniqueswhicharethe The remainder of this section focuses upon methods (Drew andSchuenemeyer, 1992). major influence ontheresultsofallthree future reserve growthcanalsohave a exploration proceeds.Assumptions about should become increasingly important as 1980 2000 2020 , , Chap4:UKERC 29/09/2009 09:50 Page 79 Page 09:50 29/09/2009 Chap4:UKERC time-series dataoncumulative asymptote to which the curve is trending (Bentley, 2009). discovery cycle iswell advanced (asinFigure4.1),itispossibletoestimatethe URR throughvisualidentificationofthe such as the linear transformation of the functional form followed by a linear regression (Hubbert, 1982). If the production or 55 Non-linearregressionisstraightforward withmoderncomputertechnology, buttheearlierliterature usessimplermethods non-linear regression reliable, methodofestimatingURRuses The simplest, although not the most techniques Productionovertime 4.3 • • Notes: explanatory variables techniques of their by curve-fitting choice Classification of explainedTable 4.1 and effort exploratory over Discovery over time Discovery over time Production ru Technique Group the derivative may bewithrespecttoeithertimeorexploratory effort. change of cumulativ Rate of production is the first derivative of cumulative production with respect to time. Alternative terms are the rate of The termsusedtolabelthesetechniquesarenotstandardised. (effort) Discovery decline curve Yield pereffortcurve Creaming curve (time) Discovery decline curve Discovery projection projection Cumulative discovery Production declinecurve Production projection projection Cumulative production e production, or more simply production. Similar comments apply to the rate of discovery, although here 55 to fit a curve to Cumulative discovery Cumulative discovery Rate of discovery Rate of discovery Rate of production Rate of production xlie aibeExplanatory variable Explained variable exploratory effort exploratory effort Rate of discovery Rate of discovery assumed a corresponds to the URR. Hubbert (1982) three ormoreparameters, oneofwhich of forms with its shape being defined by production. This curve may take a variety initially growexponentially, buttherate of implies thatcumulative productionwill Cumulative production logistic model wrt wrt Cumulative production wrt Cumulative discovery Cumulative discovery Exploratory effort Exploratory effort exploratory effort (Box 4.1)which Time Time Time Time 79

Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Chap4:UKERC 29/09/2009 09:50 Page 80 Page 09:50 29/09/2009 Chap4:UKERC 80 Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production obtained fromthefirstdifferentialofthis trajectory, withtheproductioncycle being the curve andthemidpointofgrowth representing theURR,‘steepness’of curve is defined by three parameters, more irregular in shape is the curve likely to be....” (Hubbert, 1982). 57 “…There is no necessity that the curve….have a single maximum or that it be symmetrical. In fact, the smaller the region the he firstreferredtothelogisticmodel (Hubbert,1959;SorrellandSpeirs,2009). to derive alogisticequationforcumulative production over time.Butthisformalderivation came more thantwentyyears after 56 Hubbert(1982)beginswithan assumedparabolic relationship betweenproductionandcumulative productionandusesthis depletion oil of model logistic Hubbert The Box 4.2 zero as theURRisapproached. growth willfallandeventually declineto n n n n The key features of the Hubbert logistic model are that:

Cumulative production (Gb) Where cy The diagram shows the resulting cumulative production cycle (left) and production cumulative production reaches one half of the URR the ‘steepness’ of the cumulative production curve; and Hubbert’s model for production is defined mathematically as follows: assumptions to allow tractable analysis. Hubbert (1959; 1982) noted frequently that these were only simplifying • • • • 100 200 300 400 500 600 700 0 cle (right) 01 02 03 04 50 45 40 35 30 25 20 15 10 5 0 Production increases and decreases in a single cycle without multiple peaks. the curve); resource is half depleted and its functional form is equivalent on both sides of The production profile is symmetric (i.e. maximum production occurs when the Production is modelled with the first derivative of the logistic function; Cumulative production is modelled with a logistic function; Q (t) = Q High Med Low (t) is cumulative production; ( 1 +e Q -a(t- t Year ∞ m ) ) 56 The Q Q’ ’(t)

Production (Gb/y) = (t) = stated that it need not take this form. as a‘Hubbertcurve’, Hubbertrepeatedly production cycle iscommonlyreferredto (Carlson, 2007).Whilea‘bell-shaped’ although theresultmay wellbedifferent fitting acurve totheproductiondata, curve. TheURRmay alsobeestimatedby is production; 10 20 30 40 50 60 70 80 0 0 dQ(t) 5 dt 10 t m 15 Q specifies the time when ∞ 20 ( aQ 1 is the URR; +e Year ∞ 25 e -a(t- t -a(t- t 30 m m ) ) ) 35 a 40 defines High Med Low 45 57 50 Chap4:UKERC 22/09/200910:14Page81 Note: Source: AFigure of fit the 4.2 logistic model production to US of conventional oil revolutionary changesintechnologyand wars, several recessions,twooilshocks, covering aperiodthatincludestwoworld model relatively well (Figures 3.4) despite production trend).USdatafitsthelogistic (i.e. thepointofinflectiononrising increase ofproductionhaspasseditspeak its peakandisonlyviableiftherate of be more reliable if production has passed Curve-fitting to production trends should

Data forallUSstates,includingcrude oil,NGLs,condensateandheavy oils(<100API) Production (Gb) Cumulative Production (Gb) IHS Energy 100 150 200 250 300 0.5 1.5 2.5 3.5 4.5 50 0 0 1 2 3 4 1800 1800 1850 1850 0.993 1900 R 95% CI±2Gb 2 1900 R URR =257 = 2 =0.999 (Gb) 1950 1950 Year Year 2000 and asymmetriccurves thatarewidely The logisticisoneofafamilysymmetric (Brandt, 2007). occurs over ashorterperiodoftime producing regions,especiallywhenthis to theproductioncycle formostotheroil- provides a relatively poorly approximation regions. Incontrast, the logisticmodel the openingupofnewoil-producing 2000 2050 2050 2100 2100 2150 2150 81

Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Chap4:UKERC 22/09/200910:14Page82 82 Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production depletion is the of thefewthathasbeenappliedtooil R 60 Cleveland andKaufmann (1991) fitted a logisticcurve toUSproductiondatathrough1988and found thattheadjusted well butledtosignificantlydifferent URRestimates(445Gband235respectively). depending upontheparameters chosen)withacumulative Weibull andfoundthattheyfitUScumulative productiondataequally 59 Wiorkowski (1981)compared a ‘Generalized Richards’model (whichcantake anexponential,logistic,orGompertzform (Meyer, 1994) aswellthecumulative normal,lognormal,Cauchy andWeibull distributions(Meade,1984;Wiorkowski, 1981). 58 Alternatives including the generalised logistic (Nelder, 1971), Bass (Bass, 1969), Gompertz (Moore, 1966) and bi-logistic 1984; Tsoularis andWallace, 2002). used tomodelgrowthprocesses(Meade, many published examples of curve-fitting do not meet this criterion (e.g. Campbell and Heapes, 2008; Laherrère, 2002a; 2004). 61 Ausefulruleofthumbistoensure thereareatleasttendatapointsperparameter (MotulskyandChristopoulos,2004), but 2000b). For example,Illinoisexperienced opening upofanewregion(Laherrère, technical or political changes or the one peak as a result of economic, Production cycles oftenhave morethan estimates oftheURR comparably wellbutgiveverydifferent functional forms often fit the data fitting techniques,namely: highlights agenericweaknessofcurve Cleveland and Kaufmann (1991). obtained by Wiorkowski (1981) of fit. Very similar results were later logistic model for a comparable goodness almost twiceaslargethatfromthe data heobtainedaURRestimatethatwas an asymmetric appropriate. But when Moore (1962) fitted that an asymmetric model may be more decline inover 90%ofcases,suggesting production increaseexceeded therate of regions andfoundthattherate of Brandt (2007)analysed 74oilproducing more slowly if EOR techniques are used. large fieldsaredepleted,orcoulddecline could declinerapidly after thepeakas asymmetric. For example, production expecting the production cycle to be there are a variety of reasons for basis forchoosingbetweenthesemodels, data. Whilethereisnorobusttheoretical Bartlett (2000) fitted to US production 2 changed onlyfrom0.9880to0.9909 asthevalue ofURRvaried from160to250billionbarrels. Gompertz cumulative normal (Ryan, 1966). function toUS different 59 , which 60 58 This One and This straightforward techniquewas cumulative productionaxis(Figure4.3). identifying theintersectionwith may beestimatedbyextrapolating and linear regressionisfittothisdata,theURR approximately linear(Hubbert,1982).Ifa cumulative production should be cumulative production asafunctionof a plotoftheratio ofproductionto If cumulative production grows logistically, production in aggregate regions anticipate futurecyclesofdiscoveryand techniques, namely: a secondgenericweaknessofcurve-fitting are expectedinthefuture.Thishighlights results will be unreliable if further cycles may notbejustifiedstatistically, the betterfitofamorecomplexmodel logistic curves (Meyer, decomposed intoanarbitrary numberof production trend could in principle be Kolodziej, 2008) and any cumulative Evans, 2007;Patzek, 2008;Reynolds and approach (Imam, curves. Several authorshave followedthis sum ofestimatesfromtheindividual aggregate URRmay bederived fromthe geologically homogeneous region, the production of resources from a curves. If each curve represents the and arebestmodelledbytwoormore several cycles ofdiscovery andproduction (2004) arguesthatmostcountrieshave technology (Hubbert,1956).Laherrère early developments inexploration two productioncycles asaconsequenceof et al., 2004;Mohrand their inabilityto et al ., 1999). But . 61 and the Chap4:UKERC 22/09/200910:14Page83 related tothe The linearisation technique is closely logistic model. cumulative production departsfromthe will beunreliableif(asisusuallythecase) curve tocumulative productionandhence forward, it is equivalent to fitting a logistic (HL). Whilemethodologicallystraight sometimes termed‘HubbertLinearisation’ popularised by Deffeyes (2005) and is production oil US of Linearisation’ ‘Hubbert 4.3 Figure Horne, 2007).Nevertheless, a ‘production URR (Kemp and Kasim, 2005; Li and exponential modelcan underestimate the rates shouldalsobeinvestigated sincethe alternative functionalformsfordecline This technique is widely used, but cumulative productionaxis(Figure4.4). extrapolating this until it crosses the fitting alinearregressionand production againstcumulative production, field may be estimated by plotting is assumed(Section3),theURRfor fields (Section3.3).Ifexponentialdecline future productionofindividualwellsor by reservoir engineers to project the

Production/Cumulative Production 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1 0 0.00 decline curveanalysis 50.00 used Cumulative Production(Gb) 100.00 below. discovery-based techniques described drawbacks arealsoshared by the shape theproductioncycle. These that have shaped and will continue to of economic,politicalandothervariables future productioncycles; andtheneglect cycle models; the inability to anticipate that choice; the risk of ‘over-fitting’ multi- form; thesensitivityofestimatesto a robust basis for the choice of functional important drawbacks including:thelackof past theirpeakofproduction,theyhave reliable estimates in regions that are well techniques may sometimes provide growth (seebelow).Butwhilethese and freefromthecomplicationsofreserve that isreadilyavailable, relatively accurate is straightforward andreliesupondata In sum,curve-fitting toproductiontrends technique for an oil-producing region. estimates of URR than an equivalent be expectedto lead tomorereliable decline curve’ for an individual field may 150.00 95% CI±18Gb URR= 270Gb 200.00 250.00 83

Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Chap4:UKERC 22/09/200910:14Page84 84 Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production 4.4 Discovery overtime Discovery 4.4 (e.g. 1P or 2P). estimated to different levels of confidence production data and, unlike the latter, is is less accessible and reliable than more advanced. However, discovery data the URR because the discovery cycle is should provide more reliable estimates of the extrapolation of discovery trends set ofissuesandconcerns.Inprinciple, described above andraise acomparable have much incommonwiththose 2004; 1999b;2005).Thesetechniques other authors, including Laherrère (2003; 1982) and has since been employed by introduced by Hubbert (1962; 1966; Curve-fitting to discovery trends was first techniques in Canada (Box 3.2) provides a useful counter-example. decline rates inlateryears. Whetherthisconclusionappliesmoregenerally isatopicofconsiderable dispute.TheWeyburn fi the Yates fieldinTexas andatPrudhoeBay inAlaska,whereEORappearstohave increasedproductionattheexpenseofsteeper production inthisfieldwithout having asignificantimpactontheURR.Gowdy andRoxana (2007)observe similar patternsin Source: Figure 4.4 LinearisationFigure of exponential 4.4 production decline for the UK Forties field

Annual Production (b) Gowdy andRoxana (2007)Note:TheintroductionofEORtechniquesin1986appearstohave onlytemporarily increased 100,000 200,000 300,000 400,000 500,000 600,000 0 ,0,0 ,0,0 ,0,0 ,0,0 ,0,0 ,0,0 ,0,0 8,000,000 7,000,000 6,000,000 5,000,000 4,000,000 3,000,000 2,000,000 1,000,000 0 y= -0.106x+788553 R 2 = 0.966 Cumulative Production(b) data andfoundtheyled towidelydifferent curves to US production and discovery example, Ryan (1965; 1966)fittedlogistic questioned Hubbert'sresults.For 1979; 1982)butseveral authors estimate (Hubbert,1966;1968;1974; Subsequent studies confirmed this used thistoestimateaURRof170Gb. US dataoncumulative 1Pdiscoveries and Hubbert (1962) fitted a logistic curve to could formabasisforpredictingthelatter. in production,identificationoftheformer peak rate ofdiscovery precedesthepeak latter bysometimeinterval. Sincethe logistically, withtheformerpreceding discovery and cumulative production grew Hubbert assumedthatbothcumulative illustrated inFigure4.5(Hubbert,1959). based upon the idealised life-cycle model Hubbert’s discovery projectionswere eld Chap4:UKERC 22/09/200910:14Page85 original datasetandfound thattheR Cavallo (2004) recreated Hubbert’s addition ofonlyafewmoreyears ofdata. estimates increasedrapidly withthe with equaljustificationandthatthe that muchlargerestimatescouldbecited estimates for the URR. He also showed their maximum value, production (which is still increasing) is equal to discovery (which is decreasing). Note: Source: Hubbert’sFigure idealised model lifecycle 4.5 of an oil-producing region 0.9991 to0.9146as the value ofURR the best-fitting models changed only from Cumulative discoveries arecalculatedfromthesumofcumulative productionanddeclaredreserves. Whenreserves reach Hubbert (1982) Gb Gb 100 200 300 400 500 600 700 -30 -20 -10 0 10 20 30 40 50 1 0 0

6 5

Rate ofdiscovery 11 10 Cumulative discovery

15 16

20 21 2 25 26 for Rate ofproduction

30 Cumulative production 31 Year Year plausible -althoughitworks fairlywellfor cycle appearsneithernecessary nor takes the same form as the production The assumptionthatthediscovery cycle the length of data series chosen. Hubbert’s results to be highly sensitive to Cleveland and Kaufmann (1991) found varied from 150 to 600 Gb. Similarly,

35 36

40 41

45 46 Reserve additions Reserves 50 51

55 56

60 85

Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Chap4:UKERC 22/09/2009 10:14 Page 86

the US when 1P data are used.62 The discovery cycle based upon backdated factors influencing discovery at different estimates will be a different shape from points in time are likely to be different one based upon current estimates and will from those influencing production at a have a different date for the peak in later point in time and the skewed field- discoveries. size distribution would be expected to Backdated estimates provide a more (and frequently does) lead to a sharply accurate picture of what was found at a rising cumulative discovery cycle and an particular time and are also more suitable asymmetric discovery cycle (Nehring, for estimating the URR since the 2006b). For similar reasons, there is cumulative discovery curve is more likely unlikely to be a predictable time-lag to trend to an asymptote. But they can be between the peaks in discovery and misleading since the sizes of fields production. discovered at different times will not have While Hubbert used 1P reserves to form been estimated on a consistent basis (i.e. his cumulative discovery estimates, they will reflect differing amounts of subsequent authors use 2P reserves from reserve growth). For the same reason, industry data sets (Bentley, et al., 2007; both the height and shape of the curve will Campbell, 1997; Laherrère, 2004). Since change over time and it will not represent these are generally larger than 1P the ultimate resources that were found estimates, they should lead to a higher since more reserve growth can be estimate of the URR. Also, cumulative anticipated in the future (Figure 4.6). discovery estimates tend to increase over Hence, to provide reliable estimates of the time even if no new fields are found as a URR, backdated discovery data should be result of reserve growth at existing fields. adjusted to allow for future reserve growth To take account of this, Laherrère (1996; (Drew and Schuenemeyer, 1992; Root and 2002) and others use backdated estimates Mast, 1993). While there will be which are typically larger than those made uncertainty regarding the appropriate

Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Global Oil Depletion: An assessment of the evidence for a near-term at the time of field discovery as a result of growth function to use, the failure to do reserve growth in the intervening period.63 this may lead to underestimates of the Hubbert (1967) was one of the first to URR.64 Campbell and Laherrère claim that develop backdated estimates, but did not such adjustments are unnecessary, since use these for discovery projection. A 2P estimates should not change much

62 “…..Because petroleum exploration in the US began very early, because the initial exploration and discoveries occurred in what has proved to be relatively minor basins, because early drilling technology was very limited in its drilling depth capabilities, and because discoveries in the major basins only hit their stride between 1910 in 1950, the US comes closest to a symmetric discovery curve of any major oil producing country or region.” (Nehring, 2006b) 63 Backdated cumulative discoveries may be represented by B(td,t), where td is the time of field discovery and t is the time when

∂B(td,t) the estimate is made. The backdated rate of discovery is given by: B’ (t ,t) = while the rate of change of backdated td d ∂td ∂B(td,t) cumulative discovery estimates with respect to the time of the estimate is given by: B’t (td,t) = . A plot of B’td (td,t) versus ∂t

td for a particular value of t represents a backdated discovery cycle, while a plot of B’t (td,t) versus t for a particular value of td represents a growth function. See Technical Report 5 for a full analysis of the relationships between these variables.

64 Lynch (2002) likens the neglect of future reserve growth to: “…. comparing old orchards with newly planted saplings and extrapolating to demonstrate declining tree size.”.

86 Chap4:UKERC 22/09/200910:14Page87 through to2000,theURR estimateis37% uncorrected data. But when using data compared toonly19 Gbwiththe Permian BasinsuggestsaURRof27.5Gb, cumulative discovery curve forthe using data through to 1964, the corrected discovered ineachtimeinterval. When estimate the ultimate resources Hubbert’s (1967)growthfunctionto estimates andcorrectsthesewith backdated 1Pcumulative discovery (Nehring, 2006a;b;c).Nehringemploys producing oil for more than 80 years valley in the US - which have both been of the Permian Basin and San Joaquin growth are illustrated by Nehring’s study The complicationsintroducedbyreserve should be no advantage in backdating. if 2P estimates are relatively stable there (Section 3.2)andalsocontradictory, since inconsistent withtheavailable data following field discovery. But this is Figure 4.6 The impact of reserve growth on discovery projections growth ondiscovery reserve of impact The 4.6 Figure discoveries Cumulative Cumulative discoveries Year X Cumulative discoveries however. Proved reserve estimateswould This examplecouldbe unrepresentative, Nehring comments: occurred -especiallyfortheolderfields. underestimates thegrowththatactually substantial reserve growth, it nevertheless Hubbert’s growth function predicts discovery hasmoved backintime.While larger and the estimated date of peak Year X+N Estimated ultimate made.” (Nehring, 2006b) fields……….no suchguarantee canbe ultimate recovery ofdiscovered will benofurtherincreasesinthe valid ifone canguarantee that there cumulative discovery curves areonly ultimate recovery derived from ultimate recovery. Estimatesof useless as a tool for predicting the discovery curve makes thiscurve in the [corrected] cumulative “…the continuousupward movement discoveries discoveries Future Reserve growth Future Time 87

Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Chap4:UKERC 22/09/200910:14Page88 88 Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production 65 The backdated rate of discovery with respect to exploratory effort is given by should provide abetterexplanatory If data is available, exploratory effort techniques overeffort Discovery 4.5 techniques that rely upon discovery data. important problem that is common to all regions. But Nehring highlights an suitable nor available in other fields and and such techniques may neither be injection in large fields of low permeability reserve growthderives primarilyfromCO 2003; 2005).Inaddition,theobserved growth functions being available (Verma, years of data – despite more recent and isonlyappliedtothemostrecent30 growth functionthatisnearly40years old estimates and Nehring relies upon a be expected to grow by more than the 2P cumulative exploratory effort at the time of field discovery. Box 4.3 Discovery process modelling Discovery Box 4.3 wells andusedthistoestimatethesize distribution of undiscovered fields,thefuture the numberofdiscoveries ineach size class andthecumulative numberofexploratory surface area of such fields. They estimated a negative exponential relationship between product of the number of undiscovered fields of that size remaining and the average an exploratory wellfindingafieldofparticularsize classmustbeproportional tothe relationship betweentheirURRandsurfacearea. Theypostulatedthattheprobabilityof Arps andRoberts groupeddiscovered fieldsinto15size categoriesand estimateda appraisal, most notably by the USGS. Versions oftheirapproachhave subsequentlyenjoyed widespread useinresource a lognormalfield-size distributionandhighlightedtheimportanceofsamplingbias. introduced discovery process modelling, provided one of the first pieces of evidence for the areaanexcellent candidateforstatisticalexamination.Theirclassicpaper region, unrestricted exploration and a relatively large sample of wells and fields made Denver-Julesburg basin in Colorado. The combination of a geologically homogeneous Arps andRoberts (1958)investigated thehistoryofoildiscovery intheeastflankof 2 individual fields. growth, butonlythelatterneedsdataon method forestimatingfuturereserve backdated discoveries andrequiresome 1992a). Bothmethodsrelyupon Schuenemeyer, 1993;Power andFuller, organisations (Drew, 1997;Drewand been widelyusedbytheUSGSandother Roberts (1958)andhassubsequently which was first introduced by Arps and discovery processmodelling 2002a). This approach is also the basis of 1992b; Ivanhoe, 1986; Laherrère, authors (Campbell,1996;Cleveland, subsequently beenemployed byother effort and variants of this approach have discovery dataasafunctionofexploratory (1967) was one of the first to fit curves to such aschangesintaxregimes.Hubbert be lessaffectedbytime-varying factors corresponding rate of discovery variable than time, since the B’ e d (e d ,t) = ∂B(e ∂e d d ,t) where e d (Box 4.3), represents the 65 should Chap4:UKERC 22/09/200910:14Page89 between the first exploratory well to be drilled (‘new field wildcats’ or NFWs) and subsequent wells. the cumulative lengthofallwells(i.e.both exploratory anddevelopment) (Cleveland, 1992b). Adistinctionmay alsobemade the numberofsuccessfulexploratory wells(Moore,1962), thecumulative lengthofsuccessfulexploratory wells(Stitt,1982)o 67 Including the cumulative length of exploratory drilling (Hubbert, 1967), the total number of exploratory wells (Ryan, 1973), (1983; 1985;1986),Lee(2008), Fryer and Greenman(1990)andPower andFuller(1991). 66 For example, Meisner and Demirmen (1981); Rabinowitz (1991), Arps et al. (1971), Forman and Hinde (1985), Lee and Wang measuring exploratory effort, There arenumberofdifferentways of cumulative numberof‘newfieldwildcat’ availability. Themostcommonmetric(the the choice will be largely dictated by data and political influences on exploration (Power and Jewkes, 1992; Walls, 1992). assumptions aboutfuturereserve growthandpaying insufficientattentiontoeconomic They alsosharemany ofthesamedrawbacks assimplecurve fitting, such as requiring significant numberoffieldspresentsabarriertotheirmorewidespreaduse. methodological sophistication of these models and the need for data on a statistically America and there is a lack of comparative studies on their relative performance. The variations on the basic themes, area and the existing literature appears patchy. While numerous authors have developed be applied. However, there has yet to be a comprehensive synthesis of research in this may potentiallyprovide morereliableestimatesofURRfortheregionsinwhichtheycan Discovery processmodelshave astronger theoreticalbasisthansimplecurve-fitting and discoveries thanothercompetingmodels. (1992a; b)foundtheprobabilisticapproachprovided moreaccurate forecastsoffuture 1984; SmithandWard, 1981).InastudyofCanada'sScotianShelf, Power andFuller problems ofsamplingbiasinthesize distributionofdiscovered fields(Smith,1980; Sea, althoughthisisnotageologicallyhomogeneousregionandtheauthorsneglect and Ward (1981)usethistechniquetosuccessfullyforecastdiscoveries intheNorth maximises the likelihood of the observed discovery sequence having occurred. Smith (Barouch andKaufman, 1975).Theassumedfieldsize distributionisthatwhich being proportionaltothesize ofthefielddividedbysumsizes ofremainingfields replacement from an underlying population of fields, with the probability of discovery the discoveryA moresophisticatedapproachmodels processassamplingwithout for future reserve growth (Drew and Schuenemeyer, 1992; 1993). sampling bias(Root andAttanasi, 1993)andcorrectingthefieldsize estimatestoallow (Drew, 1997),assumingapower-law size distributionforsmallfieldstocorrect 1993). Modifications include specifying discovery efficiency as a function of field size assessment oftheoilandgasresourcesUnitedStates(DrewSchuenemeyer, itself (Attanasi, et al., 1981). The USGS has employed versions of the model in their number ofexploration plays, basinsandprovinces, includingtheDenver-Julesburg basin structure, the Arps-Roberts model has subsequently produced accurate forecasts for a discovery rates fordifferentsizes offieldandtheregionalURR.Althoughsimplein 67 66 although the techniquesappeartobelittleusedoutsideNorth There arealsodifficultieswithaccounting exploratory drilling(Cleveland, 1992b). derives fromdevelopment rather than however, since much reserve growth wells -NFWs)may notbethebest, r 89

Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Chap4:UKERC 22/09/200910:14Page90 90 Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production the latterorcorrespondingparameters the asymptote of the former, the integral of estimate of the URR may be derived from yield declinesasexploration proceeds,an the creaming curve) (Figure 4.8). Provided exploratory effort (i.e. the first derivative of backdated rate of discovery against (YPE) curveeffort’ is a plot of the the USoffshoresuccessrate by8.3%/year. econometric analysis, Forbes and Zampelli (2000) estimate that, over the period 1986-1995, technological progress increased and 2002Forbes andZampelli (2000)reportthattheUSoffshoresuccessrate doubledbetween1978and 1995.Inan wells to one in three. Similarly, Lynch (2002) reports that the average success rate in the US increased by 50% between 1992 69 TheIEA(2008)reportsthat,over thelast50years, theglobalaverage successrate hasincreasedfromoneinsixexplorato average fieldsizes andassumesalognormalfieldsize distribution(MeisnerandDemirmen,1981). (and notwidelyused)discovery process modelthatreliesuponMonteCarlosimulationoftrendsinbothsuccessrates and 1980). Two employees of Shell published a paper on ‘the creaming method’ in 1981, but this describes a highly sophisticated used in the oil industry for much longer (Arps, 68 Laherrère(2004)statesthatthecreamingcurve was invented byShellinthe1980s,butvariants ofthisapproachhave been Note: Source: effort (Figure4.7), cumulative discoveries against exploratory A ‘creaming curve’ is a plot of backdated resources (Byrd, the searchforoilandgas additions andindistinguishingbetween for thedelays betweendrillingandreserve Figure 4.7 Example ofFigure a‘creaming’ 4.7 curve

Name of region withheld on grounds of confidentiality. Cumulative Discovery (Gb) IHS Energy 10 12 14 16 0 2 4 6 8 0 0 0010 0020 0030 004500 4000 3500 3000 2500 2000 1500 1000 500 et al., 1985). 68 while a ‘yield per Cumulative numberofnewfieldwildcats et al ., 1971;ArpsandRoberts, 1958;Harbaugh, occupy a larger surface area, they tend to 1981). Zampelli, 2000;MeisnerandDemirmen, number ofundiscovered fields(Forbes and success rate as a result of the declining wholly offsettheanticipateddeclinein exploration technology have partially or if at all, indicating that improvements in regions has declined only relatively gently, suggests that the success rate in most average size ofdiscovered fields.Evidence quantities ofoil)andchangesinthe wells drilled that yield commercially viable success rate (the fraction of exploratory represent theneteffectofchangesin in thefittedcurves. Changesinyield 69 But sincelargefieldsgenerally et al ., 1995;OdellandRosing, ry Chap4:UKERC 29/09/2009 09:51 Page 91 Page 09:51 29/09/2009 Chap4:UKERC expected from a random search because oil companies avoided areas that they (erroneously) thought could not contain oil. 70 Although Menard and Sharman (1975) found that early exploratory activity in the US performed worse than would be random (ArpsandRoberts, 1958). be foundrelatively earlyeven ifdrillingis Note: Source: exponential decline. past trends, which showed a negative future yielduponadetailedanalysisof Hubbert (1967) based his forecast of would declinelinearly. Incontrast, (1965), whosimplyassumedthat the yield subsequently adopted by Hendricks URR (590Gb).Zapp’s approachwas an unrealistically large estimate for the US yield would remain unchanged, leading to explanatory variable butassumedthatthe who usedexploratory effortasan developed in responsetoZapp (1962), Hubbert’s investigation ofYPEcurves was field-sizes. likely tobetheresultoffallingaverage exploration. Hence,decliningYPEismost order ofmagnitudesincetheearlydays of fields inmany regionshasfallenbyan contrast, the average size of discovered Figure 4.8 Example of a yield per effort curve effort per yield a of Example 4.8 Figure

Name ofregionwithheldongroundsconfidentiality. Discovery (mb 3 year moving average) IHS Energy 100 200 300 400 500 600 700 800 900 0 0 0 0010 0020 0030 4000 3500 3000 2500 2000 1500 1000 500 Cumulative numberofnewfieldwildcats 70 In estimates fromcurves fittoregional data. data will be different from the sum of the estimated from a curve fit to aggregate are exponential. As a result, the URR even ifthetrendsforindividualregions US willnotnecessarilybeexponential, curve for an aggregate region such as the temporary. Harris also showed that a YPE considered to be both anomalous and increased -somethingwhichHubbert work was that the discovery rate had estimate was consistentwithhisearlier estimates. TheonlyreasonHubbert’s data pointandledtosystematicallybiased weight on the last (and most uncertain) statistical procedures, placed excessive Hubbert's method violated standard projection ButHarris(1977)showedthat estimates fromproductionanddiscovery URR of ~170Gb, consistent with his in thelower48USstatesandestimateda exponential curve to his estimates of YPE Hubbert (1967) fitted a negative 4500 91

Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Chap4:UKERC 22/09/200910:14Page92 92 Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production the data, with interpretation in terms of exploration history being made ex post. subsequent paper this has increased to four (Laherrère, 2004). The choice appears to be largely determined by the ‘shape’ of 72 For example, Laherrère (2003) models the oil and gas resources of the Middle East with two creaming curves but in a 71 Although the case for such plateaus is not always supported by rigorous statistical analysis (Cleveland and Kaufmann, 1997). geographical region,themoresignificant estimate of46Mt.Thelargerthe through to 1996 leads to a much larger similar curve estimatedusingdata 1986 leads to a URR estimate of 15Mt, a curve estimatedusingdatathroughto cycles intheParis basin.Whileacreaming Lamiraux (2002),whofindtwoexploration are reportedbyWendebourg and (Charpentier, 2003).Similarphenomena basin beingdeveloped relatively late led tothelargestplay intheMichigan improvements inexploration technology combination ofgeologicalaccessibilityand (Charpentier, 2003).For examplea development rather than size in order of ease of exploration and because regionsarefrequentlydeveloped observed at largergeographical scales level, the same may notalways be effort arewidelyobserved attheplay While diminishingreturnstoexploratory reasons for why this may be the case. three plateausandprovide sometechnical exploration play typically exhibits two or (2003) show how the YPE in an however. For example,Sneddon the exception rather thantherule, Smooth curves (e.g.Figure4.7)may be functional formorthegoodnessoffit. data, but rarely provides either the large fields.Hefits‘hyperbolas’ tothis reflecting thediscovery ofsmallnumber in theearlystagesofexploration, world and found they tend to rise steeply creaming curves forallregionsofthe (2002a; 2004;2002b)hasestimated curves rather than YPE curves. Laherrère More recent work has used creaming et al 71 . data, the uncertainty about future reserve include the inaccuracy of the relevant used with care. Common difficulties time, curve-fitting techniquesmuststillbe provide a better explanatory variable than In sum,whileexploratory effortshould a significant proportion of the global URR. are precisely the regions that account for regions suchasIraq. Unfortunately, these world, it remains problematic for key reasonable formany regionsaroundthe small. Whilethisjudgementmay be the discovered resourceswillberelatively there will be few such cycles or because aggregate resources-eitherbecause cycles will have only a small impact on assumption thatany newexploration therefore depends heavily on the higher. Thereliabilityofcurve-fitting restrictions) the probability may be much geographical remoteness, or political owing tothedepthofdrillingrequired,or regions, whicharepartlyunexplored(e.g. low, while forlarge,politically-defined probability ofnewcycles may berelatively exploration iswell advanced, the geologically-defined regions where and exploration history. For small, detailed evaluation of geological potential analysis ofhistoricaldata,butonlyfroma cannot beestablishedfromthestatistical The potentialforfurtherexploration cycles number is unclear. curves andinsomecasestheappropriate little statistical support for his choice of use ofmulti-cycle models,butprovides (2003; 2004)addressesthisthroughthe this problemcouldbecome.Laherrère 72 Chap4:UKERC 29/09/2009 09:52 Page 93 Page 09:52 29/09/2009 Chap4:UKERC comparing ‘non-nested’models, such asalogisticversus aGompertz(Kennedy, 2003). 74 Goodness of fit was measured using simple 73 Thenamesoftheregionsare withheldduetodataconfidentiality. regions. illustrative datafromtenoil-producing fitting techniqueswiththehelpof We investigated the reliability of curve techniques curve-fitting Consistencyof 4.6 to underestimatesoftheregionalURR. these difficultiesappearmorelikely tolead exploration cycles in the future. Overall, cycles andtheinabilitytoanticipatenew the existence of multiple exploration estimates to the choice of functional form, growth, the apparent sensitivity of produced byeachtechnique.For curves, andalsocomparedtheestimates functional formanddifferentnumbersof of dataseries,differentchoices estimates obtainedwithdifferentlengths investigated theconsistencyof the creaming curves. Ineachcase,we cumulative discovery projectionsand using productiondeclinecurves, We estimated the URR for each region allow forfuturereserve growth. did currently used(e.g.Laherrère,2004),we reliability ofcurve-fitting techniquesas Since the objective was to test the and five past theirpeakofproduction. apparently pasttheirpeakofdiscovery countries, withallbutone(Region B) individual countries and groups of wildcat wells. The regions include both 2P discoveries and the number of new field annual estimates of production, backdated IHS EnergyPEPSdatabaseandincludes not correct thediscovery estimatesto 73 The data was taken from the R 2 . However, we recognise that this is not the best measure to use when URR. Thisisillustrated in(Figure4.9) provide substantially different estimates of cumulative discoveries ( can befoundin different. Thefullresultsfromthesetests broadly consistent or substantially most estimates were found to be either significantly change the results, since definition ofconsistencywouldnot through to2007.Amoreorlessstringent found tofitthedataequallywell, Different functional forms were often mature regions as well. results werefrequentlyobtainedfor and/or productioncycle, inconsistent regions atalaterstageof their discovery were more likely to be consistent for frequently very large.Whileestimates inconsistency intheURRestimateswas consistent results; and c) the degree of to inconsistent results more often than functional formsandnumberofcurves led b) variations inthelengthoftimeseries, between allthreecurve-fitting techniques; the meanURRestimatesconsistent in only one of the regions examined were growth. Inparticular, weobserved that:a) has not been corrected for future reserve country orregionallevel withdatathat is usually the case) they are applied at the curve-fitting techniques,atleastwhen(as reliability oftheURRestimatesfrom First, the results raise concerns about the below. of themainfindings aresummarised cumulative production( estimates differedbylessthan20%ofthe of results to be consistent if the mean URR illustrative purposes,wejudgedtwosets Technical Report5 D 2007 ) in a region Q 2007 74 . Some ) or but to 93

Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Chap4:UKERC 22/09/200910:14Page94 94 Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Figure 4.9 Cumulative discovery projection results projection discovery Cumulative 4.9 Figure Backdated Cumulative Discovery Backdated Cumulative Discovery Backdated Cumulative Discovery Backdated Cumulative Discovery Backdated Cumulative Discovery R URR =99% Gompertz R URR =109% Gompertz 2 R URR =104% Gompertz R URR =98% Gompertz R URR =156% Gompertz 2 = 0.993 2 = 0.998 2 2 = 0.999 = 0.981 = 0.996 Region G Region I Region E R URR =95% Logistic Region A Region C Time Time Time 2 Time Time = 0.988 R URR =127% Logistic R URR =97% Logistic R URR =111% Logistic 2 2 2 = 0.996 R URR =99% Logistic = 0.971 = 0.999 2 = 0.999 Logistic Gompertz Observed Logistic Gompertz Observed Logistic Gompertz Observed Logistic Gompertz Observed Logistic Gompertz Observed

Backdated Cumulative Discovery Backdated Cumulative Discovery Backdated Cumulative Discovery Backdated Cumulative Discovery Backdated Cumulative Discovery R URR =236% Gompertz R URR =100% Gompertz R URR =96% Gompertz R URR =519% Gompertz R URR =144% Gompertz 2 2 2 2 2 = 0.968 = 0.992 = 0.993 = 0.988 = 0.999 Region F Region H Region J Region D Region B Time Time Time Time Time R URR =175% Logistic 2 R URR =97% Logistic URR =93% Logistic R URR =108% Logistic = 0.987 2 R URR =135% Logistic 2 = 0.99 = 0.999 2 = 0.97 Logistic Gompertz Observed Logistic Gompertz Observed Logistic Gompertz Observed Logistic Gompertz Observed Logistic Gompertz Observed Chap4:UKERC 22/09/200910:14Page95 (ranging from 1% to 362%), but the mean difference in 75 ThemeandifferenceintheURR estimatesfromthetwomodelswas 59%ofthecumulative discoveries through to2007 by athird. is only0.003buttheURRestimatesdiffer can biastheresults(e.g.Gompertz in alltheregionsexamined,butchoice be moreappropriatethananexponential or Gompertz functional form was found to 4.10). Contrary toexpectations,alogistic significantly differentresults(seeFigure stages inthediscovery cycle canleadto As aresult,estimatesmadeatearlier curve isclearlyapparent(e.g.Region I). discovery cycle when the asymptote of the are at a relatively late stage in their The estimatesonlyconverge whenregions cumulative discoveries (e.g. Region J). estimates that are less than the simple curve fitting can lead to mean URR sensitivity to lengthsensitivity of series data – model logistic using E Region for projection discovery Cumulative 4.10 Figure difference betweenthe Taking Region Easanillustration, the logistic and Gompertz functional forms. projections for each region using both which shows cumulative discovery

URR Estimate as % of D2007 75 100 110 120 130 140 150 160 170 180 1985 The results also show that R 1990 2 for eachmodel Last YearinTimeSeries 1995 R 2 estimates was only0.001. observed for sixofthetenregionsusing URR. Trend breaksofthisformwere significant underestimateoftheregional the trend break) it would have led to a stage of the production cycle (i.e. prior to this technique had been used at an earlier cycles ofexploration andproduction. If breaks’ thatmay resultfromadditional and theregionasawholeexhibits‘trend the corresponding data for both offshore be modelled by a single linear regression, approximate linearrelationshipthatcan While theonshoredata‘settles’intoan Figure 4.11 shows the results for Region J. underestimate the URR. As an illustration, suggest asystematictendencyto particularly unreliableandtheresults Linearisation’) techniquewas foundtobe The productiondecline(‘Hubbert cases). model provides higher estimates in all 2000 2005 2010 Confidence Interval Confidence Interval Interval 95

Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Chap4:UKERC 22/09/200910:14Page96 96 Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Figure 4.11 ProductionFigure 4.11 decline curvesfor Region J

Production / Cumulative Production Production / Cumulative Production Production / Cumulative Production URR= 51% URR= 55% Cumulative Production Cumulative Production Cumulative Production Region JOnshore Region JOffshore Region JTotal URR= 177% URR= 124% URR= 124% Chap4:UKERC 22/09/200910:14Page97 forms curvesfor Creaming RegionFigure 4.12 A using exponential and hyperbolic functional results isthatthetechniquesarebeing The primary reason for these inconsistent to justify one choice over the other. explorationregion, itisdifficult of a history Without a detailed knowledge of the significantly different(seeFigure4.13). corresponding URRestimateswere one ortwocreamingcurves, the three of the regions could be fit with either form assumed(seeFigure4.12)andwhile were sensitive totheparticularfunctional was asymptotic. Once again, the results the corresponding discovery projection behaviour, althoughintwoofthesecases the regionsdonotexhibitasymptotic Notably, thecreaming curves forfourof reliable than discovery projections. that creamingcurves aregenerally more Our results also do not support the claim will remainstableinthefuture. little confidencethatthedeclinecurves both). The frequency of such breaks gives using eitheronshoreoroffshoredata(or aggregate data and for all of the regions

Backdated Cumulative Discovery URR =102% Exponential R 2 =0.988 Cumulative NFW URR =141% Hyperbola R 2 =0.981 data andthatthisshapewillnotbe cycle canbe estimatedfromthehistorical ‘shape’ of the production or discovery Curve-fitting techniquesassumethatthe with econometrics Reconcilingcurve-fitting 4.7 should be treated with caution. suggest thattheassociatedURRestimates fitting techniqueascurrentlyusedand demonstrate the limitations of curve- Nevertheless, theresultsaresufficientto as fartheavailable datapermits. therefore addresseachoftheseproblems applications ofcurve-fitting should allow for future reserve growth. Future we didnotcorrectthediscovery datato exploratory holesaseither oilorgas,and regions, thedatasourcedidnotclassify distinguish betweenonshoreandoffshore history. Inaddition, we didnotalways regions thatlackaconsistentexploration applied to large and geologically diverse Exponential Hyperbola Observed 97

Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Chap4:UKERC 22/09/200910:14Page98 98 Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production production when resources are abundant, political constraints may restrict may result.For example,lowoilpricesand variables, despite the potential errors that been a tendency to neglect these relevant variables. Asaresult,therehas changes in prices, technology and other significantly affected by any future hyperbola with (top)two fitted and one E Region for (bottom) curve Creaming 4.13 Figure

Backdated Cumulative Discovery Backdated Cumulative Discovery R 2 =0.996 R 2 =0.983 Cumulative NFW Cumulative NFW R 2 =0.996 parameters (includingURR), lead tobiasedestimates ofmodel correlation oftheerror terms.Thiscould missing variable bias and/or serial techniques are likely to suffer from a result,many applications of curve-fitting production whendepletionisadvanced. As while high prices may stabilise or increase URR estimate=196% URR Estimate=286% Hyperbola 2 Hyperbola 1 Observed Hyperbola Observed Chap4:UKERC 22/09/200910:14Page99 of Texan oil-producing capacity. 77 For example,between1957and1968,the‘prorationing’ decisionsoftheTexas Railroad Commissionshutinmorethan50% 76 For example, Epple and Hansen (1981), MacAvoy and Pindyck (1973), Walls (1994) and Mohn and Osmundsen (2008). formulation assumes that geologic and predicted and actual production. This account forthedeviationsbetween then usesaneconometricmodelto model toUScumulative productionand Kaufmann (1991), who fitsalogistic A good example of a hybrid model is production (Power andFuller, 1992b). determinants ofdiscoveries and biased sincetheyignorethegeological estimating URR and are also likely to be forecasting discoveries rather than but thesehave mostlyfocusedupon variables. Many variants have followed, discovery size and lagged dependent function ofoilprices,pastaverage the average size ofdiscovered fieldsasa for exploratory activity, successrate and Fisher (1964), who estimated equations 1992; 1994).Thelatteroriginatedwith to estimatingfutureoildiscoveries (Walls, estimating URRandeconometricapproach a ‘hybrid’ ofacurve-fitting approachesto model specification.Thiseffectively gives discovery and/orproductionwithinthe the economic and political determinants of promising approach is to include some of itself totheestimationofURR.Amore specified model may not necessarily lend in oneormorepreviousyears), butthere- the currentyear afunctionofproduction specification (e.g.makingproductionin of thedependentvariable withinthemodel addressed by including one or more ‘lags’ These problems may potentially be are provided in goodness of fit. Several examples of this errors and overestimates of the model underestimates oftheassociatedstandard Technical Report5 . 76 restrictions onproduction) economic and political variables (e.g. legal rise and fall over the long-term, while physical factorscause oil productionto steeply rising costs after 1970 indicating depletion and technical change – with and representstheneteffectofresource mirrors the bell-shaped production curve estimated U-shaped average cost curve production cycle. Theempirically assume a particular functional form for the variable and hence remove the need to production costsasanexplanatory Cleveland (2001)includeaverage US In a significant innovation, Kaufmann and which seems implausible. (e.g. the same before and after the peak) both symmetricandindependentoftime However, theoilpriceelasticityofURRis leads toahigherestimateoftheURR. production over the period 1948-1990 and explains over 98% of the variation in US problems of serial correlation. Their model and re-formulatingittoeliminate number of economic and other variables allow for the dependence of URR on a this bymodifyingthelogisticmodelto analysis in the second stage. They avoid economic factors which only enter the not take into account the effect of estimation of URR in the first stage does Kaufmann’s modelisbiasedbecausethe Pesaran and Samiei (1995) argue that relevant economicandpoliticalvariables. assumptions about the future values of the to 1985,butestimatesoftheURRrequire fit toUSproductionover theperiod1947 trend. This model provides a much better term variations aroundthisunderlying 77 lead toshort- 99

Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Chap4:UKERC 22/09/200910:14Page100 100 Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production 78 Cavello (2004; 2005a; b) reaches a very similar conclusion through a more qualitative argument. discovery or production cycle: variables indeterminingthe‘shape’of importance ofeconomicandpolitical 1938 in1991andhighlightsthe variation inUSoilproductionbetween This modelaccountsformostofthe accelerating depletion (Cleveland, 1991). techniques. For example,Cleveland and and improve any ofthecurve fitting similar approaches can be used to modify If therequireddataisavailable, very difference totheestimatedURR. economic variables will make a significant is muchlessclearwhetheraccountingfor forecasting thedateofpeakproduction,it Hubbert may have ‘got lucky’ in in production costs. In addition, while assumption is required about future trends estimate theregionalURR,some costs. Moreover, if it is to be used to owing tothelackofdataonproduction may not be applicable for other regions the arbitrary choiceoffunctionalform),it weakness ofcurve-fitting techniques(i.e. But whilethismodelovercomes akey Cleveland, 2001) got lucky.” (Kaufmann and shaped curve.....In effect,Hubbert significantly differentfromabell production thatwaspattern of variables have could produceda evolutionary path for any of these production over time.Adifferent symmetric bell-shapedcurve for traced what appears to be a the TRC coevolved in a way that cost of production, and decisions by because realoilprices,average real peak inUSproductionaccurately “….Hubbert was able to predict a 78 standard curve fitting,theymay bemore In sum,whilehybrid models improve upon is normally assumed. the approximate declining order of size that As aresult,fieldshave notbeenfoundin changing licensingregimes(Priest,2007). difficulties of deep-water drilling and owing toacombinationofthetechnical been geographically restrictedinthepast, a morelikely reasonisthatexploration has the GulfofMexicocontributestothisresult, previously. Whilethegeologicaldiversity of comparable tothatachieved 50 years and leadingtoaYPEin2000thatis greatly expanding the area of exploration technical change has increased since1975, 1995). Theresultsshowthatthepaceof importance (Managi, offshore industry, weightedbytheirrelative number ofinnovations adoptedbythe technical changebyanindexoftheannual depletion bycumulative discoveries and Mexico, Managi, example, intheirstudyofYPEtheGulf 1999; Power andJewkes, 1992).For resource depletion(IledareandPulsipher, effect oftechnicalchangefromthat Some authorsattempttoseparate the Cleveland, 1991). estimates ofURR(Kaufmann and model may notsignificantlychangethe the long-term,implyingthatrevised prices. Butagain,depletiondominatesin changes intherates of drillingand/oroil increases in YPE were associated with shows howperiodsofrelative stabilityor much betterfittohistoricaltrendsand rate ofdrilling.Their equation provides a short-term changesinoilpricesandthe exponential modelofYPEtoaccountfor Kaufmann (1991;1997)modifyHubbert's et al. et al (2005) model ., 2004;NPC, Chap4:UKERC 22/09/200910:14Page101 79 Anotableexception isDées, • as follows: The main conclusions from this section are 4.8 Summary States. studies arelargelyconfinedtotheUnited few ‘hybrid’ studiesandwhy theavailable these may partly explain why there are so disaggregation required.Problemssuchas is rarely available at the level of frequently eitherlackingorunreliableand (Lynch, 2002). Also the required data is production/import/export quotasetc.) restrictions onexploration, tax rules,leasingdecisions,geographical production and/ordiscovery trends(e.g. a subsetofthevariables thatcouldaffect since itisimpractical toincludemorethan be vulnerable to missing variable bias consistent exploration history andmay still either geologicalhomogeneityora conclusions ifappliedtoregionsthatlack Hybrid modelsmay stillleadtomisleading and otherfactorstobedirectlyexplored. the dependence of URR on energy prices estimates of URR. However, they do allow models lead to substantially different historical data it is not obvious that hybrid variables and despite their better fit to values oftherelevant explanatory requires assumptionsaboutthefuture than forestimatingURR.Thelatter suitable for short-term supply forecasting relativ techniques aremoreappropriatefor the basictechniques.‘Geological’ estimating URRandmany variations on There are a variety of methods for 79 ely unexplored regions while et al.(2007) • • pessimistic forecasts of oil supply. probably contributedtoexcessively underestimates oftheURRandhave appear more likely to lead to variables. Ingeneral, theseweaknesses neglect of economic, political and other production ordiscovery; and the inability toanticipatefuturecycles of ov choice of functional form; the risk of sensitivity oftheestimates to the inadequate theoreticalbasis;the of this technique, including: the insufficient accountoftheweaknesses Many applicationsofcurve-fitting take history. Thisisfrequentlynotthecase. a relatively unrestricted exploration geologically homogeneous and has had technical changeandiftheregionis depletion outweighs the effect of these assumptions will only hold if fields beingfoundrelatively early. But returns to exploration, with the large field size distribution and diminishing aggregate data.Allassumeaskewed curve-fittingsimple onlyrequires require data on individual fields, while and discovery processtechniques difference isthatfield-size distribution weaknesses. Butakey practical man degree rather thankindandshare The extrapolation techniques differin lead to quite different results. different techniques show that they can large andthefewstudiesthatcompare these estimates are commonly very advanced.on Theconfidencebounds appropriate whereexploration is ‘extrapolation’ techniques are more erfitting multi-cycle models;the y of the same strengths and 101

Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Chap4:UKERC 22/09/200910:14Page102 102 Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production • • • illustrativ Tests ofcurve fittingtechniquesusing of resource size. further contributed to underestimates such adjustments is likely to have growth. Thecommonfailuretomak estimates to allow for future reserve estimates andtheneedtoadjust such as the uncertainty in reserve introduces additionalcomplications Curve fittingtodiscovery data may beo Some ofthelimitationscurve fitting in the future as exploration matures. uncertainty may beexpectedtodecline of suchestimates,thedegree this raises concerns about the reliability lead to inconsistent results. But while time series and numbers of curves can techniques, functionalforms,lengthof regions ha v e data from a number of ercome with the use of hybrid ve shownhowdifferent e • in the results obtained. must alsoacknowledgetheuncertainty exploration historyoftheregion.They geological characteristics and and beinformedbyknowledgeofthe multiple techniquesandsourcesofdata resource assessments should emplo commonly assumed.Wherepossible, on theresultsarewiderthanis limited andthattheconfidencebounds do imply that its applicability is more curv These limitations do not mean that URR. substantially differentestimatesofthe such modelsmay notleadto despite their better fit to historical data, economic and political variables. But models thatincorporate relevant e fittingshouldbeabandoned,but y Chap5:UKERC 22/09/200910:16Page103 by Wicks (2009) to support the argument that the level of remaining resources presents no risk to future global oil production. hydrocarbons. Notso. Theworldhasampleresources,withmorethan40years ofproven oilreserves.” Thesamefigureisused 80 For example,in2008Tony Hayward, theCEOofBP, commentedthat:“….Myth No. 2isthattheworldrunning outof confidence in future oil supply, are widely quoted as a measure of current production. Although such ratios resources (typically1Preserves) by estimate ofremainingrecoverable exhaustion iscalculatedbydividingan models, thenumberofyears untilreserve production (R/P)ratios. Withthese production are based upon reserve-to- The simplest models of future oil ratios andcurve-fitting depletion: reserve-to-production Simplemodelsofoil 5.1 synthesizing thoughtsandcritique. Lastly, Section5.5concludeswith econometric analysisofhistoricaltrends. depletion theory, togetherwiththe 5.4 describes models based upon optimal regional andfield-level production.Section projections using data-rich models of bottom-up models,whichbuildaggregate complex functions.Section5.3discusses of oil discovery and production with more models, whichrepresentthemechanisms models. Section5.2describessimulation production modelsandcurve-fitting production, includingreserve-to- Section 5.1outlinessimplemodelsofoil in full examinationofthistopiciscontained results are often significantly different. A approaches are fiercely debated and the oil production.Therelative meritsofthese various approachestoforecastingfuture This sectiondescribesandcompares 5. provide little useful information about the Technical Report6. Looking ahead-methodsofforecasting future oilsupply 80 they .Fittheconstrained modeltohistorical 3. Includeconstraints toimprove the 2. Define a mathematical function to 1. approach is as follows: variety of models exist, but their general which have beenusedsincethe1950s.A curve-fitting modelsofoilproduction, Somewhat more complex and realistic are longevity offuturesupply. than providing assurances aboutthe production fromafieldorregionrather in thisway, R/Pratios constrain therate of limits onR/Pratios (Höök,2009).Defined from afieldorregion,implyinglower economic limits to the rate of depletion including the YTF).There are technical and resources foranoil-producingregion(i.e. estimates of remaining recoverable reserves for an individual field or but mostcommonlydefinedusing2P inverse oftheR/Pratio (i.e.theP/Rratio), of depletionrates. These aresimplythe described inSection3,namelytheconcept A morelegitimateuseofR/Pratios isthat region peaks and declines (Figure 5.1). during periodsinwhichproductiona often remain constant or even increase ratios is indicated by the fact that they to zero. ThelimitedusefulnessofR/P the point of exhaustion and then collapse that productioncanremainconstantuntil future productionprofilesincetheyimply production. data and use this to project future quality of model fit. data. statistically fittohistoricalproduction 103

Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Chap5:UKERC 22/09/200910:16Page104 104 Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Note: Source: regions post-peak in ratios production to Proved reserve 5.1 Figure 1970 depending upon the value of the URR production wouldpeak between 1965and curve (Box 4.2). HeforecastthatUSoil US oil production, utilizing a bell-shaped his, nowwell-knownprojectionoffuture In March 1956, Hubbert (1956) produced Curve-fittingmodels 5.1.1 or multiplecurves. asymmetrical) andinthechoiceofsingle (including whetherthisissymmetricalor mathematical function that is assumed cases, themodelsvary intheparticular the URR and/or other parameters. In both further constrained by assumed values for while inforecasting,thecurve fittingis functional formandthehistoricaldata, constrained only by the assumed production projection, the curve-fitting is estimating URR,describedinSection4.In production projection techniques for These approaches are very similar to the The USpeaked in1970,theUK1999andNorway in2001. BP (2008) R/P Ratio 10 11 12 5 6 7 8 9 0020 0220 0420 062007 2006 2005 2004 2003 2002 2001 2000 Norway Year UK the next 25 years, it was not until 1980 used variants ofthelogisticmodel over exogenous assumptions. Butwhilehe estimate theURRrather thanrelyupon comparable curve-fitting techniquesto Hubbert subsequentlydeveloped United Statesproductionpeaked in1970. subsequently showntobeaccurate when opposition, Hubbert’s forecast was Despite encounteringwidespread 1952). 1963 and 1967 in two scenarios (PMPC, Resources forFreedom Gb respectively), whilethe1952study the level of ultimate recovery (100 or 200 would occurin1960or1970dependingon peak ofoilproductionintheUnitedStates example, Ayres (1953)forecastthatthe were studying oil depletion at the time. For projection was unprecedented, others some have argued that Hubbert’s Gb, respectively) (Hubbert,1956).While used to constrain the curve (150 and 200 US predicted peaksin Chap5:UKERC 22/09/200910:16Page105 Note: exponential decline (right) Difference betweenFigure 5.2 aggregate exponential decline (left) and disaggregate model, Hallock post-peak production declines. In another unrealistically sharppeaksandvery rapid exponential depletionmodelandproduces USGS (2000)withinaglobally-aggregated recoverable resources taken from the probabilistic estimatesofremaining (Wood, Energy InformationAdministration (EIA) form was madebyanalystsfromtheUS production usinganexponentialfunctional A notable projection of global oil altering someofitsassumptions. original Hubbert method while relaxing or These models share properties of the functions toforecastfutureproduction. use Gaussian,exponentialandother models have sincebeendeveloped which A variety of Hubbert-like curve-fitting (Hubbert, 1982). that hepublishedafullderivation 2021 to 2112, Hallock indicates datesfortheglobalpeakfrom severe declines. While the EIA model resulting inaroundedpeakandless smoothes productionnearthepeak, modified exponential methodology that Decline rate is same in each case, but applied individually to a number of sub-regions in the right figure. Production (Gb/y) 10 15 20 0 5 et al.,2000).Thisutilizes 05101520253035404550 et al. et al. concludethat Year (2004) usea production increasesuntilthe (Cavallo, 2002). In the EIA model, between aggregationanddeclinerates exponential models is the interaction (Pickering, 2008).Butakey problemwith decline within larger aggregated regions there issomeevidenceforexponential declines exponentially(Section3.3)and production from individual fields often It haslongbeenobserved thatthe decline before2037. conventional oil will begin an irreversible Production (Gb/y) remaining recoverable resources. model istheUSGS(2000)estimateof production, while the numerator in the EIA ratio of proved reserves to annual inappropriate sincethelatterrelatestoa experience. Butthisanalogyis which is justified on the basis of US depletion rate of10%followingthepeak, 5.2). Also, theEIA modelassumesa according tothesamerule(seeFigure country, basinorfieldbehaves individually is very different fromamodelwhereeach which productionisforcedtodecline.This production reachesthetargetvalue, after of remainingrecoverable resourcesto 10 15 20 0 5 05101520253035404550 Year global ratio 105

Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Chap5:UKERC 22/09/200910:16Page106 106 Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production 82 For example, Lynch (2003) notes that Campbell increased his value of global URR from 1,575 Gb in 1989 to 1,950 Gb in 2002. divergent futures become increasingly narrow. for useinprojections.Thissituation improvesthe productioncycle thefurtherinto becomes, asthepossibilitiesfo aregion 81 Caithamer(2008)arguesthat thereisnotenoughinformationinpre-peakproductiondatatogenerate stablevalues of URR magnitude ofany futureproduction cycles. order to estimate the probability and geological andtechnicalknowledge,in techniques needtobecombinedwith charitably, itsuggeststhatcurve-fitting renders [predictions] useless”. More whether any given peakisthefinal one of the bell curve” because “not knowing technique “destroys theexplanatoryvalue injection. Lynch (2003)arguesthatthis production technologies such as CO as drillingatgreaterdepths,orbynew advances inexploration technology, such unpredictable. Theymightbeenabledby additional cycles are inherently complexity. Anotherproblemisthat always justified by the additional model better fit of a more complex model is not Multi-cycle models sufferbecausethe another. disruptive changes from one function to built multi-functionmodelsthatexhibit production. Mohr and Evans (2007; 2008) differentiated from the main body of of a well-defined resource that can be additional cycle represents the production curves toproductiondata.Intheory, each sum of a number of independent logistic (1999b) andPatzek (2008),whofit the models have beendeveloped byLaherrere profiles observed empirically. Multi-cycle recreate thenon-smoothproduction and multi-function models attempt to rises and falls in a single cycle, multi-cycle In contrast tomodels whereproduction 2 that theseestimatesareunlikely to discovery estimatesareavailable) and oil (provided backdated 2P cumulative regional andglobalURRforconventional argue that we can accurately estimate the difficulties.Some seriousness ofthese Researchers disagreeaboutthe on investment. ultimately belimitedbytheenergyreturn However, the amount of growth will physical propertiesoftheresource. and URR will grow with no change in the willing topay moreforaunitofpetroleum materialize, future consumers would be URR. Alternatively, ifoilsubstitutes failto resource depositswouldresultinalower valuable, thesamephysical endowmentof 1997). If oil suddenly became less economic conditions(AdelmanandLynch, variable that depends in part upon while inpractice itisanendogenous fixed constraint incurve-fitting models is that URR is frequently interpreted as a Using exogenous estimates Difficulties withcurve-fittingmodels 5.1.2 c). (Lynch, 1999;2003;Nehring,2006a;b; the useofinappropriategrowthfunctions been ignored or underestimated through and/or futurereserve growthhaseither use ofsimplecurve-fitting techniques resources has been underestimated by the This is typically because the volume of YTF and regionalURRhave oftenbeentoolow. problematic becauseestimatesofglobal constrain curve-fitting modelscanbe 82 Another more fundamental problem 81 of URRto r Chap5:UKERC 22/09/200910:16Page107 oil fields is determined in part by the they arenotindependent. Productionfrom field-level productioncurves aresummed, summed oraveraged. While regionalor that areindependentofoneanother a normaldistributionwhendistributions McCabe, 1998).TheCLT actstogenerate (Babusiaux, applying theCLT tooilproductioncurves models, but there are difficulties in theorem (CLT) to justify bell-shaped Other authorsusethecentral limit production. greatly alterthetimingofinvestment and economic andpoliticalconstraints can discovery and production. But in reality, resulting inafairlyconsistentlagbetween will bepromptlybroughtintoproduction, reasonable toassumethatoildiscovered conditions under ample demand, it is market aswell.Infree logistic path that physical depletion, but it does not follow interaction ofgeologicalinformationand logistic develop a simple modelthatgenerates that is made. Rehrl and Freidrich (2006) can besensitive totheparticularchoice any other functional form and the results theoretical basisforchoosingalogisticor 4). For example, there is no robust techniques toestimatetheURR(Section analogous tothosefacedinusingsuch curve-fitting toforecastproductionare Most of the other difficulties in using (McCabe, 1998; Mills, 2008). currently excluded fromestimatesofURR produced fromresourcesthatare that hydrocarbons will increasingly be estimates willcontinuetogrowandalso et al significantly changeinthefuture(Bentley, production ., 2007).OthersarguethatURR discovery et al., 2004; Brandt, 2007; will necessarily follow a behaviour fromthe are likely toincreaseover time,estimates inherently uncertain.Since URRestimates estimates oftheregional URRwhichare production profile, but relies upon produced. Thissuggestsanasymmetric recoverable resources have been their peakwellbeforehalfof most countries appear to have reached of decline(Section3).Inotherwords, level ofdepletiononly25%atthestart post-peak countries and found an average recoverable resources, we analysed 55 the USGS (2000) estimates of remaining supported byempiricalobservation. Using symmetric productioncycle isonlypoorly Lastly, the common assumption of a tested models. were eachbestfittedbyoneofthesix 5.3 showsregionalproductioncurves that to thesedisparate modelscanbe,Figure regions. To illustrate howexcellent thefit model tothedatafrom139oilproducing model, alinearmodelandanexponential asymmetric versions of a bell-shaped (2007) compared symmetric and Laherrere, 2003; Patzek, 2008). Brandt have multiple peaks (Hubbert, 1956; 2007; Hirsch, 2005), while many others more pointedthanthebell-curve (Brandt, from thebellshape.Somearesignificantly shaped profiles,agoodnumberdeviate curves are well-approximated by bell- that whilesomeregionalproduction smoothing powerislimited.Theresult historical evidencesuggeststhatthis smoothing will occur with aggregation. But independence between producers, some Given thatthereissomedegreeof transport costs;and nationalpolitics. prices; local, regional, and global face commonstimulisuchasglobalcrude part bythedecisionsofproducerswho physical conditionsinthosefieldsand 107

Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Chap5:UKERC 29/09/2009 09:55 Page 108 Page 09:55 29/09/2009 Chap5:UKERC 108 Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production fitting model (Brandt, 2007) model (Brandt, fitting where were regions to best the they be applied to found types model Six 5.3 Figure increase –asmay beexpectedfromthe slower rates ofdecline thanrates of tend to be slightly asymmetric, with analysis suggeststhatproductionprofiles the medianrate ofincrease (7.8%).This (2.6%) was approximately 5%lessthan found that the median rate of decline to 67 out of 74 post-peak regions and Brandt (2007)fittedan exponentialmodel to fall. of the level of depletion at peak are likely 100000 120000 140000 160000 180000 Production peryear(kb/y) Production peryear(kb/y) 100000 20000 40000 60000 80000 10000 20000 30000 40000 50000 60000 70000 80000 90000 0

0 1862 1862

1872 1872 Guassian modelofWyomingprod. Wyoming prod.data

prod. Exponential modelofCameroon Carribbean prod.data 1882 1882

1892 1892

1902

C. Exponential C. 1902 A. Gaussian A.

1912 1912

1922 1922

1932 1932

1942 1942

1952 1952

1962 1962

1972 1972

1982 1982

1992 1992

2002 2002 Production peryear(kb/y) Production peryear(kb/y) 10000 12500 15000 17500 20000 22500 25000 27500 30000 10000 12500 15000 17500 20000 22500 25000 2500 5000 7500 2500 5000 7500 pattern. may not follow the same production regions that have yet to reach their peak economic conditionsmay meanthat larger regions.However, changing also suggest that decline is slower in weighted mean(1.9%),Brandt’s results significantly lessthantheproduction- Since themeanrate of decline(4.1%)was stylised modelpresentedinSection1. 0 0 1862 1862

1872 1872 Alabama prod.data prod. Asymmetric GaussianmodelofAlabama D. Asymmetric ExponentialD. prod. Asym. exponentialmodelofAlbania Albania prod.data 1882 1882 B. Asymmetric Gaussian B.

1892 1892

1902 1902

1912 1912

1922 1922

1932 1932

1942 1942

1952 1952

1962 1962

1972 1972

1982 1982

1992 1992

2002 2002 Chap5:UKERC 29/09/2009 09:55 Page 109 Page 09:55 29/09/2009 Chap5:UKERC buried atdifferentdepthsinthe sand. 83 InReynolds’ model,theagentsareshipwrecked islandersandtheresourceisdefinedastins of provisions washed ashoreand model years. searching for resources over a number of extraction process involves agents (1999) inwhichtheresourcefindingand model based on the work of Reynolds fitting. For example,Bardi(2005)builta scarcely morecomplexthansimplecurve- The simplestsimulationmodelsare 1998). and theexploitationofcrudeoil”(Taylor, effect relationship exists between time models, namelythat“…nocause-and- remedies a key problem of curve-fitting functional form beforehand. This approach mechanisms rather thanspecifyingthe production cycle emerge from these extraction, thereby letting theshape of the mechanisms thatgovern oildiscovery and underlying physical and/or economic Simulation modelsexplicitlyrepresentthe and technologies rates, resources, discovery Systemssimulation: 5.2 cumulative discoveries and a technological resource discovery basedonthelevel of 100000 110000 Production peryear(kb/y) 10000 20000 30000 40000 50000 60000 70000 80000 90000

1862 0

1872 prod. Linear modelofSouthernEurope Southern Europeprod.data

1882

1892

1902 83

1912 E. Linear E.

Probabilities govern 1922

1932

1942

1952

1962

1972

1982

1992

2002 10000 20000 30000 40000 50000 60000 70000 80000 Production peryear(kb/y) quantity, but is the product of the models, undiscovered oilisnota static Naill, 1973).Inthemoresophisticated oil extraction system(anearlyexampleis feedbacks betweenthemany partsofthe importance of rates of change and dynamics, Later models used methods of 1973. in USoilproductionbetween1964and Davis’ modelpredictedapeakanddecline peaks ofexponentialmodelsdonotoccur. constrained productionpath,thesharp reserves are discovered over timewitha discovery andproduction.Because incorporated intoaneconomicmodelof fixed estimateofURR,itisfully being constrained byanexogenous and the reserves. Thus, instead of the function assumed a fixed annual depletion rate of additions based on exploratory effort, and development. Heprojectedfuturereserve complex simulationofoildiscovery and Davis (1958) produced the earliest over time. function thatimproves findingcapabilities

1862 0

1872 Asym. linearmodelofCameroonprod. Cameroon prod.data

1882

F. Asymmetric LinearF. 1892

1902 which focuses on the

1912

1922

1932

1942

1952

1962

1972

1982 system

1992

2002 109

Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Chap5:UKERC 22/09/200910:16Page110 110 Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production 85 Theclassicexampleof“overshoot andcollapse”istheWorld3 model(Meadows, 84 Thelogisticmodelisconsistent withCleveland's (1991)empiricalestimatesofUSproductioncosts. and finely balanced between positive and Also, simulationmodelscanbeunstable impossible. empirically derived, butthisislikely to be Such arelationshipwouldideallybe investment inexploration technology. defined as a function of cumulative theoretically discoverable at any time, fraction of oil deposits that are al. discoverable’ parameter oftheDavidsen Consider, forexample,the‘fraction face of conflicting or nonexistent data. relationships andcorrelations,ofteninthe quantification ofalargenumber Simulation models require the Difficulties withsimulationmodels 5.2.1 al., 1990;Sterman, called ‘fraction discoverable’ (Davidsen, estimated resource base and a parameter built inthisframework sufferfromsimilarproblems,such asthegeneral natural resourceutilization modelofBehrens(1973). cost as a resource becomes depleted. which causes an increase in production cost is modelled with a logistic function transition to substitute fuels. Production model, and resulting in a smooth simultaneous solutionofallyears ofthe recursive demandfunction,allowing project demandforliquidfuelsusinga 1983). For example Greene and Papin, 1981;EdmondsandReilly, used afterapeakinoilproduction(Basile substitutes forconventional oilthatwillbe Other modelsattempttosimulatethe technologies. companies invest revenues inexploration fraction discoverable increasesasoil (1990) model.Thisrepresentsthe et al. , 1988).The et al . (2006) 84 et et in a statistically meaningful fashion. rarely comparedwiththehistoricalrecord validation, in that simulated production is simulation modelsgenerally lackempirical Basile and Papin (1981)). Finally, (as seeninbothSterman the easeofatransition tosubstitutefuels caused simulation models to overestimate inertia andnegative feedbackshasalso are absentfromsuchmodels.Thislackof feedbacks) thatexistintherealworldbut mitigating factors (e.g. negative world productiondataaredueto The moregentledeclinesobserved inreal- bottom-up modelisthatofCampbelland regional widely published The most Energy (BentleyandBoyle, 2008). data sourcessuchasthatprovided byIHS relies upondetailedandoftenconfidential an increasinglyprominentmethodbut world). Bottom-upmodellinghasbecome larger regions(suchasanationorthe ’build up’ projections of production from of project,fieldorregionalproductionto Bottom-up modelsusedetailedmodelling lower levels building upoildepletionfrom Bottom-upmodels: 5.3 dynamics modelsofresourcedepletion. (1973) results,andiscommoninsystem decline. ThiseffectisevidentinNaill’s be rapid growthand sudden, very rapid in the model are too strong, the result can negative feedbacks.Ifpositive feedbacks et al ., 2004). Otherresourcedepletionmodels et al. (1988) and 85

Chap5:UKERC 22/09/200910:16Page111 otherwise onavariety ofdatasources level datawherethisisavailable, and (2008) maintainsamodel basedonfield- regional Formodels. example, Smith obtained fromfield-level rather than More robustresultscouldpotentiallybe generic weaknessinthisapproach. Heapes, 2008).Thistendencysuggestsa production after 2010 (Campbell and more recentversions predictpeak projected beforetheyear 2000,while over time:inthe1995version, thepeakis production has shifted steadily forward Campbell’s predicteddateofpeak (Jakobsson, does have some empirical support decline. Whilethisapproachissimplistic,it rate risesto~3%,-followingwhichthey their depletion production constantuntil East producers are assumed to keep mid-point is reached, while five Middle produce atafixed growthrate untilthe remaining countriesareassumedto resources producedeachyear. Mostofthe fraction of the recoverableremaining current depletion rate, defined as the production is assumed to continue at the produced morethan50%ofitsURR.Ifso, determine whetheracountryhas subtracts cumulative productionto individual producing countries and techniques toestimatetheURRof Campbell uses a number of curve-fitting Energy (BentleyandBoyle, 2008). was subsequentlypurchasedbyIHS database fromPetroconsultants SA that originally produced using a proprietary and Heapes,2008).Thismodelwas (Campbell, 1995;1996;2004;Campbell co-authors, producedsincethemid-1990s et al. , 2009).However, Technical Report 7. givenother bottom-upmodelsis in more detailed discussion of these and for medium to long-term forecasts. A capacity addition, but is much less suitable uncertainty inthe3-5year periodof Skrebowski’s approach removes some of productiondeclinefromexistingfields. forecasts and assumptions about the rate capacity. Theseareoffsetagainstdemand into thelikely short-term increasesin approach provides considerable insight the majorityofnewoiloutput,this projects have longleadtimesandprovide size (‘megaprojects’).Becauselarge development projects above a threshold who maintainsadatabaseofoilfield Skrebowski (2004;2005;2006;2007), A thirdapproachisrepresentedby fields. comprehensive databasesofindividual and PFCEnergyusingglobally approaches are adopted by Miller (2005) assumption isnotused. Very similar growth. Thus, a simple ‘mid-point peak’ assumed toincorporate futurereserve the announcedURRoffieldwhichis of facilitiesmodifiedwhereappropriateby announced plateaulevels orthecapacity fields, productionisbasedupon expected EORprojects.For undeveloped field decline rates, including allowance for existing fields is projected using historical country level (Box 7.2).Productionat aggregated attheoperator, basin,or 111

Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Chap5:UKERC 22/09/200910:16Page112 112 Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production model theeffectofprojectsthatcould field? Also, itisdifficulttoanticipateand appropriate declinerate foreachmodelled assumptions. For example,whatisthe results intheneedformanymore that gives these models their strength also field level. In fact,theproliferation ofdata future isnotremoved bymodellingatthe Unfortunately, alluncertaintyaboutthe data. provided accessisavailable totherelevant modelled withgreaterconfidence, projections whichcangenerally be individual field-level production emerges directlyfromthesummingof aggregate productioncycle. Instead,this the properfunctionalformfor these modelsremove uncertaintyabout decline andinvestment atindividualfields, future production. By accounting for short- tomedium-termprojectionsof hold considerable promise for accurate contrast, thefield-level modelsseemto consistently proved tobepessimistic.In Campbell’s supplyforecastshave point peakingassumption.Asaresult, 88 Kaufman (1983)andWalls (1992)provide useful,iflessrecent,reviewsoftheeconometric literature. 87 See Krautkraemer (1998) for a comprehensive review of the ODT literature. no dataonfunctionalforms,confidence intervals orgoodnessoffit. relation totheactualproduction data. Similarproblemsareapparentinmany oftheregionsexaminedandreportprovides extrapolation doesnotfollowtheactualdatapointsinany obviousway and thefirstderivative ofthelogisticcurve bearsno 86 Asanillustration, CampbellandHeapes(2008)use‘HubbertLinearisation’to estimate theURRforLibya. Butthe techniques used; statistical weaknessoftheparticular underestimate theURR(Section4); tendency of curve-fitting techniques to a narrowdefinitionofconventional oil;the number ofimportantlimitations,including Campbell’s bottom-upmodelhasa Difficulties withbottom-upmodels 5.3.1 86 and the simple mid- econometric modelsofoildepletion. theory (ODT) prices. We firstdescribeoptimaldepletion extraction pathsandtheeffectsofoil economists focusoninvestment, optimal decline rates and field-size distributions, focusing on physical aspects such as or reservoir engineers.Rather than and depletion differently from geologists Economists understandnatural resources depletion Economicmodelsofoil 5.4 2003). difficult or impossible to reproduce (Lynch, susceptible tocriticismandrendersthem experience. Thismakes themodels with themodeller’s judgmentand rely on proprietary databases augmented are notpeerreviewed,partlybecausethey not always ofhighquality. Mostarticles The literature describing these models is too pessimistic. resulting supplyforecastsarelikely tobe such projects is underestimated, the prices. If, asseemslikely, thepotentialfor fields totechnicalchangeandhigheroil response across tens of thousands of oil capture becauseitisadistributed projects. This ‘stealth oil’ is difficult to such as infill drilling, workovers and EOR lessen thedeclinerate atexistingfields, 87 and then describe some 88 Chap5:UKERC 28/09/2009 12:29 Page 113 Page 12:29 28/09/2009 Chap5:UKERC Box 5.1 Extensions of optimal depletion theory depletion optimal of Extensions Box 5.1 have been extended in a variety of ways (Krautkraemer, 1998), but ODT models reproduce historically observed behaviour years. Such a path obviously does not present value of production in future declines over timeduetothedeclining typically startsatitshighestpointand depletion. Theoptimalproductionpath production coststhatdonotvary with production of a single resource with The simplestODT modelsfocusonoptimal interest. extraction costs, should rise at the rate of resource intheground,lessmarginal This suggeststhatthevalue ofaunit was extracted and the profitsinvested. value theywouldreceive iftheresource of aunitresourceinthegroundwith producers shouldequatethefuturevalue (ODT), namely: rational resource underlying optimaldepletiontheory Hotelling (1931)formalized thekey insight resources now or consume them later? over time. That is, resource economicsisallocation The primaryconcernofexhaustible Optimaldepletiontheory 5.4.1 • • Optimal depletion models have been extended as follows: the US. cost increases.Cleveland (1991) provides empiricalevidence forthisinthecaseof production costs rapidly in the early years of an industry, after which depletion causes production profileswithpeaks.Thisoccursbecause technologicalchangebringsdown declines with time, this model results in price paths that are ‘U-shaped’ and Production costvaries withtime,asduetotechnologicalchange(Slade,1982).Ifcost to increaseatlessthantherate ofinterest(Fisher, 1979). Depletion increasesthemarginalcostofproduction, causingthevalue oftheresource should we consume has occurredinreality. suggests iseconomicallyoptimalandwhat arguably bridgesagapbetweenwhatODT exist intheoilindustry, thisdevelopment Given that all fourofthesecausalfeatures Production moves to new sites or new 4. Exploration fornewdepositsoccurs 3. Technological changeresultsin 2. Demandincreasesover time,inducing 1. behavior: ODT model features can cause peaking with peaks. He argues that at least four circumstances resultinproductionpaths extended ODT models can in some (Box 5.1). Holland (2008) illustrates that resource types over time. (e.g., Pindyck, 1978). due to discounting (e.g., Slade, 1982). temporarily outweighthedisincentive decreases in production cost that of futureprofits disincentive resultingfromdiscounting time periodseven given the producers to delay extraction to later 113

Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Chap5:UKERC 28/09/2009 12:29 Page 114 Page 12:29 28/09/2009 Chap5:UKERC 114 Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production 90 Walls (1994)andDahl(1998)survey thesemodelsindetail. 89 See the reviews of Kaufman (1983), Dahl (1998) and Walls (1989) for more information on Fisher-type models. regulatory ones.Thesemodelsvary inthe variables with geological, physical, or models which combined economic Later camemoresophisticated‘hybrid’ Kaufman, 1983). revaluated (Dahl and Duggan, 1998; results wereoftennotrobustwhen Spann, 1971; Walls, 1992), and the magnitudes andsigns(Erickson poor: someparameters hadunexpected performance of these models was often the field size distribution. In addition, the geological natureofoilresources,suchas they did not include information about the A key weakness of earlier models was that numerous variants have since followed. were firstdeveloped byFisher(1964)and Econometric modelsofoilandgassupply trends inoildiscovery andproduction. through regressionanalysisofhistorical variablesare estimatedempirically 1998). Therelationshipsbetweenthese (Kaufman, 1983;PindykandRubinfeld, such as price and extraction cost demand primarilyfromeconomicvariables statistical modelsthatprojectsupplyand Econometric models are data-rich depletion Econometricmodelsofoil 5.4.2 • • 1973). economic jargon,thesesubstitutesarecalled“backstop”resources.(Nordhaus, Depletion canbemodeledasatransition tosubstitutefuelssuchasoilsands.In increases and then decreases after a peak. U-shaped, andiftheinitialreserves aresmall(asintheoilindustry),extraction first model ofPindyck (1978)andsuccessors).Theresultingoilpricepathcanbe Producers explore for and deplete a number of deposits of uncertain size (see the 89 they continuetoassumeaparticular reproduce historicalproductiontrends, While alloftheseapproachesaccurately problems of serial correlation. specifying thismodeltoeliminate variables as used by Kaufmann and re- of thesameeconomicandpolitical parameter inthelogisticmodelafunction improved uponbothbymakingtheURR US production. Pesaran and Samiei (1995) deviations betweenpredictedandactual economic andpoliticalfactorstomodelthe logistic model in a different way, by using static. Kaufmann (1991)augmentedthe the economiccritiquethatURRisnot technological change,therebyaddressing make URRafunctionofoilpricesand who modifiedHubbert’s logisticmodelto first suchattemptswerebyUri(1982), limited numberofeconomicvariables. The augment curve-fitting modelswitha Moroney andBerg,1999),whileothers traditional econometric models (e.g., comparatively minormodifications of Some hybrid models are structured as described earlierinSection4. application toestimatingURRwas and non-economicfactors.Their extent towhichtheyrepresenteconomic 90 Chap5:UKERC 22/09/200910:16Page115 price canincreasevery rapidly afterthe uncertain exploration suggeststhattheoil contrast, Reynolds’ (1999) model of years (Howarth andNorgaard,1990). In optimise net present value over all model ODT requiresresourceproducersto This assumptionisimplicitinthefactthat extent and nature of resource deposits. producers areknowledgeableaboutthe of ODT istheassumptionthatresource grounding. Anotherfundamentaldifficulty that thesemodelslackempirical analytically. Aresultofthissimplicityis simple so as to permit their being solved Optimal depletionmodelsareintentionally Difficulties witheconomicmodels 5.4.3 which introduce uncertainties of its own. the futureevolution ofproductioncosts for forecastingrequiresassumptionsabout Kaufmann and Cleveland model as a basis remains unclear. Moreover, to use the - althoughthelikely size ofthedifference production cycle and the date of the peak have altered both the shape of the prices or production restrictions would 1970 peak,becauseadifferentpathofoil “Hubbert got lucky” in his prediction of the and Cleveland usethismodeltoarguethat historical data.Asnotedearlier, Kaufmann cycle andprovides anexcellent fitto other functional form for the production removes theneedtoassumealogisticor before thepeakinproduction1970) production costs(whichrosesharply variables.The inclusionofaverage term historicalanalysisas one oftheinput production fromCleveland’s (1991)long- including theaverage realcostof Cleveland (2001)improves upon this by The morerecentmodelbyKaufmann and functional formfortheproductioncycle. (Lynch, 2002). While modern parameters estimatedfromhistoricaldata stymieing predictions basedon variables will change in future years, unavailable. Thevalues oftheseomitted variables for whichdataisgenerally swamped byalargenumber ofomitted complex econometric model can be data. Thisisbecauseeven themost poorly onlyafewyears beyond thefitted made with these models typically fare models is generally fragile and predictions The fidelityofeven thebesteconometric et al. process of technological diffusion (Managi, adequate measuresforthecomplex technical change, it is difficult to find models have attemptedtoincorporate to compensatefordepletion.Whilesome technological change is often not included “abysmally bad”, inpart because their long-term forecasting performance Lynch (2002),ontheotherhand,calls exploration, discovery andproduction.” closely to the physical character of econometric modelsdonotconform “the functionalformsemployed inmost few years in advance, mostlikely because well” at predicting production more than (1983) arguesthatthey“have notdone have theirownshortcomings.Kaufman historical data. However, these models provide impressive agreement with the empirical grounding than ODT and can Econometric modelshave amuchfirmer (Kaufmann and Shiers, 2008). has some important policy implications representation ofreal-world behaviour and availability. This seems a more accurate incomplete knowledge of resource ‘sneak up’ontheoilmarket becauseof peak inproduction,asdepletioncan , 2004). 115

Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Chap5:UKERC 22/09/200910:16Page116 116 Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production of ov of Technical Report6 • technological, orphysical aspectsoftheoilproductionprocess. with timeasthekey independentvariable. Incontrast, mechanisticmodelshave functions representingdetailedeconomic, 91 Theoretical models rely on general, simple functions relating oil production to a small number of input parameters, generall • • • the modelsvary, namely: models, acrossfourdimensionsinwhich a numberofways. We classifythese The modelsreviewedabove vary widelyin across modeltypes Modelclassificationandtrends 5.5.1 on depletionmodels Synthesizingthoughts 5.5 accuracy offorecasts. omitted variables can still undermine the problem tosomeextent,futurechangesin cointegration techniques can mitigate this models donotnecessarilyfitclearlyinone models tendtofallontheleftside.Also, figure, while simulation and econometric tend toclusterontherightsideofthis 5.4. For example,curve fittingmodels all quadrants ofthespaceshowninFigure The modelsdiscussedabove occupynearly dimensions along with example models. while Figure 5.4 plots three of these Overall model complexity. mechanistic); Model formulation(e.g.theoreticalor global); Scale of model (e.g. field-level or economic); reasoning applied(e.g.physical or Intellectual orientation and type of er 40 models along these dimensions, 91 and includes a comparison Model Dstraddles economicandphysical portionsofthediagram. builds upfromafield-level basistoprojectglobalproduction. Note: classified in three dimensions types model Different 5.4 Figure also vary significantlyintheircomplexity: models Although notplottedinFigure5.4, Greene this might be the simulation model of a mechanistic framework. An example of physical characteristics simultaneously in Model D includes both economic and projection ofoilavailability. Alternatively, field-level data tobuildupaglobal Smith’s bottom-up model,whichuses quadrant toanother. Anexampleis on theotherhand,‘builds’ fromone model ofoilandgasexploration. ModelC, such as Fisher’s sub-national econometric model thatoperates at aregionalscale, Model Bisaneconomicandmechanistic logistic model.Intheoppositequadrant, production, such as Hubbert’s global is based on physical aspects of oil Model Aisaglobal,theoreticalmodelthat Turning totheexamplesinFigure5.4, combine economicandphysical reasoning. example, anumberofmodelsexistthat classification along a given dimension. For Mechanistic Physical Models AandBresideinonemodelquadrant. ModelC et al (2006). Global Field Economic Theoretical y Chap5:UKERC 22/09/200910:16Page117 .Extendedeconomicoptimal depletion 4. Probabilistic finding models show 3. Oil transition simulations under perfect 2. historical datamore closely willbetter It is often assumed that models which fit across modeltypes Modelclassificationandtrends 5.5.1 System dynamicsofresourceextraction 1. assumptions. Specifically: fundamental physical andeconomic production over time from other, more models generate a rise and fall in particular functionalform.Thatis,these forcing the production cycle to fit a production profiles without explicitly underpinnings have produced peaked of models with divergent intellectual Another importanttrendisthatanumber in example Model D. physical andeconomiccharacteristics -as complex over timeandtoincorporate both tendency formodelstobecomemore Technical Report6 The fullreviewandclassificationin project the production of multiple fuels. models require numerous data inputs and simulation modelsandsomeeconometric R/P models are the simplest, while with roundedpeaks(Holland,2008). models show optimal production cycles (Bardi, 2005; Reynolds, 1999). declining probabilityofnewdiscoveries peaking behavior resultingfromthe 2006). (Basile and Papin, 1981; Greene, foresight show non-bell-shaped peaks (Naill, 1973;Sterman, and depletionshow rounded peaks shows ageneral et al. , 1988). et al. , • • model fitsinclude: Some additionalcharacteristics ofgood depletion literature. measures are not widely used in the oil 2004). Althoughwellestablished,these models (MotulskyandChristopoulos, comparison of model fit across divergent Criterion (AIC) the model,or accounts for the number of parameters in • such as entirely withstandardstatisticalmeasures the fitofamodelcannotbejudged predict futureproduction.Unfortunately, model, such as Adjusted R take intoaccountthecomplexityofeach measures used to compare models must better. For thesereasons,statistical model, allowingittofitthedatapoints each cycle addsmoreflexibilitytothe better thanasinglecycle modelbecause multi-cycle logisticmodelwillalways fit (NIST/SEMATECH, 2008).For example,a complex by adding additional parameters increase when a model is made more being fit),subjecttoph between the model and the data points Small residuals (i.e. the difference 2002). or experience.(BurnhamandAnderson modelling intermsofscientifictheory constr when theycan model, becausemodelsaremostuseful priori A with aconsistentstandard-deviation Residuals thatarenormallydistributed, sum of squared residuals. between modelanddata,suchasthe measure oftheoverall divergence Curve-fitting proceduresminimize a R aints on the free par 2 scientific justificationforusinga because this will nearly always , whichallowsforthe Akaike’s Information explain ysically sensible what theyare 2 ameters. , which 117

Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Chap5:UKERC 22/09/200910:16Page118 118 Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production of therelative qualitiesoffittothedataor of thepredictive accuracies ofpossible alternative models.” 92 Kaufman (1983)notesthatinthe1950sand 1960s“[no]comparison[was] made,basedonacceptedstatisticalprinciples, many predictions of a global peak have forecasting global oil production, and Existing models have been very poor at Theproblem ofprediction 5.5.3 econometric) inthepublishedliterature. model types(e.g.curve-fitting vs. comparisons empirical between broad As farasweareaware,are no there not generally applicabletoallregions. useful ofthesymmetricmodels,butwas shaped) model was found to be the most exponential model.TheGaussian(bell- shaped model,alinearmodelandan asymmetric versions of a Gaussian bell- used AIC to compare symmetric and countries andmulti-countryregions).He variety of scales (US states and regions, models to139oilproductioncycles ata simple (3 and 4 parameter) curve-fitting Brandt (2007) compared the fit of six projecting resourceavailability. suggest a superior general modelfor production datacouldnot,ontheirown, estimating theURR.Hefoundthat cumulative Weibull function for use in compared a flexible function to the analyses was byWiorkowski (1981),who work. initial controversy surroundingHubbert’s model fitwerenotperformedduringthe surprising that empirical analyses of Given thecomplexitiesinvolved, itisnot each of these characteristics. exist to determine whether a fit has no ‘serialcorrelation’).Formal tests runs ofpositive ornegative values (i.e. over timeandwithfewconsecutive 92 One ofthefirstcomprehensive .Simplecurve-fitting modelscan provide 1. much currentdebate): (noting that these topics are the source of value ofthereviewedmodelsisasfollows judgment withrespecttothepredictive neither hopeless nor obvious. Our intermediate caseswheretheanswersare We requirepredictive modelsin mathematical model is needed. century ofpeakoilproduction,thenno Gb, 10 Gb, or 100 Gb?), or only the next year (e.g. willitbeapproximately 1 magnitude of global crude oil production one wishes only to predict the order of model can provide a useful answer. But if hydrocarbon products in the year 2050, no predict basin-level productionofmultiple type ofpredictiondesired:ifonewishesto feasibility of prediction varies with the One important consideration is that the (see Box 5.2). failure ofapredictive effortisnoteasy clear interpretationofthesuccessor historical data. Even after the fact, the more difficultthanfittingfunctionsto but fromthefactthatforecastingis much from the problems of oil supply models, Such pooroutcomesresultnotsomuch eventually prove tobeequallywrong). (although forecastsofadistantpeakmay come and gone without confirmation given thesecaveats. peak productionforanestimateofURR, be sufficienttopredictthedecadeof system. URR andnosignificantshockstothe production, a roughunderstandingoffuture This understandingislikely to assuming agivenlevelof Chap5:UKERC 22/09/200910:16Page119 Box 5.2 The caseofBox Hubbert’s 5.2 prediction of apeak oil production in US • d c b a • • • • • Predictions of peak US oil production of the 1950s. ofthe production oil US ofpeak Predictions dates. It is difficult to determine precise peak years from the figure. 1965 and1975,implying thattheov year ofproductionfromthefigures. Davis Pogue Ion Hubbert Ayres Ayres PMPC Aut In the article text, Hubbert states, “about 1965” and “about 1970”, so we use these T p. 103. These datescomefromasingleprediction,but itisdifficulttodeterminetheprecise There aretwopossibleinterpretationsofthisfact:onecouldargue that thispro peak (see T the 1950s predicted a peak in US production within ≈ 10 years surrounding the actual models from Hubbert’s 1956 prediction difficult at best. 1950s. Thesedifferencesmake drawing conclusionsaboutthepredictive abilityof more heterogenousontheglobalscalethantheywereinUnitedStates The level of exploration, accessibility to reserves, and governance factors are much asymmetric. authors above) thoughtthatitwould(Jackson2006),andtheUScurve issomewhat still remain.Thus,productionhasnotdroppedasquicklyHubbert(ortheother oil. Cumulative USproductionhasalreadyexceeded 200Gbandsignificantreserv An important fact is that all of these studies underestimated the URR for conventional mechanistic detail, and are therefore unappealing from a scientific perspective. an interpretationiscontroversial becausecurve-fitting modelscontainlittlecausalor be generally outlinedsome 10-15years beforehandusingquantitative methods.Such The morefavorable interpretationimpliesthatthefutureofUSoilproductioncould methods produced projections clustered in a band around the actual peak date. came as close to predicting the peak date. Or, one could emphasize that a variety of that Hubbert’s method was not extraordinary, as other authors using other methods Although Hubbert’s analysisisthemostwell-known,nofewerthan7estimatesfrom of URRwhichheconsideredtobelesslikely. Hubbert’s prediction of a peak in US production in 1970 was based on his high value able onp.Total1975. 83shows productionin1955,1965,and is equalin production hor able). Year of estimate 1958 1956 1956 1956 1953 1952 1952 Early peakEarly date er all peak is between these years. 1964 1970 1965 1965 1960 1962 1963 Late peak date peak Late 1968-1970 1972 1975 1964 1967 1970 1973 b d b a c 150 / 200 100 / 200 URR (Gb) URR 100 165 NA NA NA ves es 119

Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Chap5:UKERC 22/09/200910:16Page120 120 Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production peak production. than decadalerrorsinthe forecastdateof at thegloballevel) couldcause greater estimated URR (e.g. many hundreds of Gb 1974 and1979),ormajorerrorsinthe major disruptions (e.g. the oil crises of the date ofpeak(Bartlett,1978).However, unlikely tosignificantlyaffect theforecast due to political or economic factors are estimated URR,orminorhiccupsdelays production thatminorvariations inthe is sohighduringtheyears ofpeak uncertainty (Section4).Butconsumption URR, whichisitselfsubjecttoconsiderable they are contingent on the estimate of interference oreconomicdisruption.Also, and production are unhindered by political assumptions includethatdemandgrowth restrictive assumptions. These defined resource type is reasonable under using abell-shapedmodelforwell- regarding the first point: decadal accuracy Note that there are numerous caveats Thereisnojustificationformaking 3. More-detailed mechanistic models (e.g. 2. are clearly expressed. sensitivities and confidence intervals limited value unlesstheuncertainties, involved make suchpredictionsof models. The many uncertainties any of the surveyed mathematical the exact year of peak production) with precise anddetailedpredictions(e.g. uncertainties. specified and‘brittle’withrespectto forecasts because they are over- this advantage wanes for long-term useful fornear-term predictions.But data andarethereforelikely tobemore greater fidelity in reproducing historical bottom-up, econometric), exhibit number. Though most such effects will be equations – are theoretically infinite in world thatarenot included inthe Out-of-model effects – features of the necessarily be greatly simplified. included inamodelmust what is necessity, beleftoutofany model,and many aspectsoftherealworld must,by forecasts (Smil,2003).Thisisbecause relied ontoproducemoreaccurate shortcomings. Thisapproachcannotbe add additionalfeaturestoaddressits level ofcomplexity, thereisatendency to Once amodelsurpassesrelatively low modelling Complexityandthepurposeof 5.5.4 appropriate. that asimplermodelmay bemore rates, prudenceandparsimony suggest future discoveries orfield-level recovery hundreds oflong-termassumptionsabout increased detail.Ifamodelrequires counteract thebenefitofmodel’s the more these proliferating assumptions critique. The more complex the model is, assumptions are visible and open to ready assumptions. Inasimplemodel,these increasing a model’s reliance on time scaleoftheforecastincreases,thus Uncertainty necessarilyincreasesasthe advantages holdforlong-runforecasts. Unfortunately, itisnot clearthatthese more closely match predictions to data. freedom toallowthefittingalgorithm because they have more degrees of fluctuations inhistoricaloilproduction advantages inreproducingshort-term econometric models)have clear mechanistic models(e.g. Complex Chap5:UKERC 22/09/200910:16Page121 on theunderstandingand insightsgained. can have significantdetrimental impacts to improve the reliabilityofforecasts,and diminishing returns.Itgenerally doeslittle complexity isanactivitysubjecttorapidly Increasing modelspecificityand unavailable. which tobaseabetterassumptionis an accurate assumption, but the data on all regionsandfuels.Thisisunlikely tobe costs asafunctionofdepletionisequalfor (2006): therate ofincreaseextraction availability isgiven byGreene, world behaviour. Anexampleofpoordata be built,buttheseonlyapproximate real relationships toallowtractable modelsto typically assumelog-linearorlinear cost andreserves.” Econometricians in therelationshipbetweenunitsupply with whatappeartobebasicnonlinearities econometric models “have not coped well example, Krautkraemer (2003) notes that be in a highly simplified form. For and datalimitationsoftenrequirethemto as functionsareadded,modelformulation complexity asamodellingstrategy isthat The secondfundamentalproblemwith and arecertaintohappenagain. deviations have occurredofteninthepast actions of the market, but significant Many sucheffectsaresmoothedbythe regulation, or demand discontinuities. changes in institutional structure or disruption, warpolitical andconflict, together with ‘human factors’, such as features of discovery and production, behaviour. These include technical between modelforecastsandobserved model tocausesignificantdeviations unimportant, enough will exist in any et al. production (Hirsch, the rate ofdeclineconventional oil conventional liquidscanbebuiltrelative to rate atwhichproductioncapacityfornon- important factorstomodelbecomethe production. Recognizing this, the mb/d, or more than 2.5% of all-liquids produced in quantities in excess of 2.0 Non-conventional liquids are already depletion-and-substitution world. depletion-or-substitution world, but a alternative resources. We donotlive ina induce investment in a variety of significant oil price increases, which will conventional productionwillresultin economists, whoarguethatapeakin already occurringisemphasized by using a variety of technologies. That this is produce oil from a wide range of deposits defined) neglectsthefactthatwealready (regardless ofhow‘conventional’ is entirely. Thisfocuson conventional oil Many modelsleave thisquestionaside production? demand afterthepeakinconventionaloil resources willbeusedtomeetliquidfuel arises from asking the question: techniques. Theneedforthisintegration physical andeconomicmodelling modelling exists in better integration of A key opportunitytoimprove oilsupply depletion modelling Movingforward: improving oil 5.5.5 integrated models. economic aspectsof theproblemin modelling: include both physical and a clearstrategy forimproving supply substitution inoilsupplymodelspointsto to account for both depletion and these liquidsintorefinedfuels.Thisneed costly itistoaccess,extract, and upgrade et al. , 2005),andhow what 121

Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Chap5:UKERC 22/09/200910:16Page122 122 Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production • • follows: The mainconclusionsofthissectionareas 5.6 Summary because theyattempttoreproducethe Simulation modelsareattractive also have thedisadvantages ofboth. econometrics offer promise, but can Hybrids of curve-fitting and accurate forecastsoffutureproduction. that thisabilitytranslates tomore match historicaldata,butitisnotclear models exhibit a superior ability to perform poorlyasaresult.Econometric form, neglect key variables and can sensitive to the choice of functional lack anadequatetheoreticalbasis;are man they builduponempiricalexperiencein straightforw Curve-fitting modelsare model over time. to comparetheresultsfromsame the performanceofdifferentmodels,or work hasbeencarriedouttocompare all circumstances.Unfortunately, little single approachshouldbefavoured in strengths andweaknessesno and complexity. Each approach has its different variables, level ofaggregation their theoretical basis, inclusion of future oilsupplyvary widelyintermsof Existing approaches to forecasting y oil-producing regions. But they ard andwidelyused • • already beguninthisdirection. understanding provided. Work has technologies) inordertoimprove the as resource characteristics or substitution) and physical aspects (such economic aspects(suchasresource Future modelsmustincorporate both diminishing returns. this problemandissubjecttorapidly model complexitydoeslittletoaddress estimates of great precision. Increasing disruptions, nomodelcanprovide political, economic,ortechnological market. Butgiven thepotentialfor significant disruptions to the global oil assumed level of URR, assuming no accur that peak projections of decadal Basic mathematical analysis suggests require alargenumberofassumptions. lack oftransparency and necessarily proprietary datasetsandassociated are hampered by their reliance on term forecasts,buttheexistingmodels provide themostreliablebasisfornear- models usingfieldorprojectdatamay parameterisation. Finally, bottom-up and sufficientdatafor frequently lackbothempiricalvalidation be overcomplicated andunstable that govern oil production, but they can physical andeconomicmechanisms acy arelikely tobepossibleforan Chap6:UKERC 22/09/200910:17Page123 93 Thisfigureincludesconventional oil,extra-heavy oilandsands. Section 6.5concludes. the implications for future global supply. range ofuncertaintyover thisvariable and (2009). Section6.4identifiestheimplied updated bytheIEAandAguilera, the USGS estimates have recently been estimates. Section6.3summariseshow since 2000 is consistent with these and evaluates whether the experience USGS World Petroleum Assessment 2000, the mostinfluentialstudytodate, summarises the methods and results of changed over time.Section6.2 the past and illustrates how these have some estimatesthathave beenmadein global oil supply. Section 6.1 compares assesses their implications for future evaluation of global URR estimates and This sectionprovides anoverview and oranges’. considerable riskof‘comparingapplesand the resourceswillbedeveloped, thereisa viability andthetimehorizon over which assumptions abouteconomicandtechnical conventional oil,togetherwithdifferent frequently usedifferentdefinitionsof consensus emerging.Sinceanalysts grown over time,thereislittlesignofany systems. Whilesuchestimateshave geological characteristics of petroleum curve-fitting to in-depth evaluations of the these estimates can range from simple future supply. The methods used to derive underpinning moreoptimisticforecastsof the peakoildebate,withlargerestimates conventional oilplay aprominentrole in recoverable resource(URR)of Global estimatesoftheultimately 6. How muchdowehave? et al. – globalestimatesoftheultimatelyrecoverable resources ofconventionaloil (IEA, 2008). and cumulative 2P discoveries of 2369 Gb production throughto2007of1128Gb 3000 Gbwhichcompareswithcumulative estimates clusterintherange 2000Gbto illustrated inFigure6.1.Mostrecent summarised in (Salvador, 2005).Theseestimatesare the sameinstitutionsorindividuals including several repeated estimates by published fromavariety ofsources, estimates of the global URR have been Since this time, at least one hundred estimate tosome600Gb(Pratt, 1942). 1942, Stebinger had increased his oil tobeonly43Gb(White,1920).By estimated theglobalURRofconventional In 1920, Eugene Stebinger of the USGS estimates Trends inglobal URR 6.1 global URRandusesthesetosupport the mostpessimisticestimatesof Campbell (1991;1997)provides someof include NGLswhiletheearlieronesdonot. particular, most of the recent estimates given the difficulties of comparison. In attribute muchsignificancetothistrend is very poor and it would be wrong to between 1942and2007.However, thefit they have increasedbyaround17Gb/year estimates (Figure6.1)tosuggestthat A linearregressioncanbefittothese overestimated. OPEC countriesinthebeliefthattheseare exclude someofthedeclaredreserves of definition of conventional oil and/or although theseoftenrelatetoanarrower than some contemporary URR estimates 93 This latterfigureishigher Technical Report 5 and 123

Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Chap6:UKERC 22/09/2009 10:17 Page 124

Figure 6.1 Global URR estimates over the last 70 years

4500

4000 1988 - 2008 R2= 0.103 3500

3000 1942 - 2008 R2= 0.118 2500

2000

URR Estimate (Gb) 1500 1970 - 2008 R2= 0.007 1000

500

0 1940 1950 1960 1970 1980 1990 2000 2010 2020

Year

forecasts of an early peak in global referred to as the ‘global oil system’. With production. He uses multiple techniques to this model, the total oil in place is estimate the URR for individual countries estimated from the balance between oil and then aggregates these to the global generation, oil seepage to the surface, level. Campbell relies upon backdated thermal cracking and loss from the source estimates of cumulative 2P discoveries rocks over geological time. The derived originally from the recoverable resource is estimated as a Petroconsultants (now IHS Energy) proportion of the oil in place, using database and does not adjust these to assumptions about recovery factors. Miller reflect future reserve growth. He also Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Global Oil Depletion: An assessment of the evidence for a near-term concludes that a URR of 3960 Gbs is significantly downgrades the reserve plausible, excluding heavy oil and tar estimates for OPEC countries and uses a sands. restrictive definition of conventional oil (‘regular oil’) that excludes ‘deepwater’ (>500m), ‘polar oil’ and NGLs. Campbell’s most recent estimate for regular oil is 6.2 The USGS World 1900 Gb (Campbell and Heapes, 2008) Petroleum Assessment 2000 which is less than two thirds of the USGS 2000 estimate for conventional oil (see The most comprehensive and authoritative below) and 469 Gbs less than cumulative estimates of global URR are from the US 2P discoveries of conventional oil (IEA, Geological Survey (USGS) which has 2008). published five global assessments since 1980 (USGS, 2000). Each uses a By contrast, one of the largest estimates of the global URR is by Miller (1992) who combination of methods to estimate the adopts an unconventional approach, URR of geologically homogeneous regions 124 Chap6:UKERC 22/09/200910:17Page125 term) and economically accessibleinthelonger resources may onlybetechnicallyand underestimate implies thattheresultscouldboth about technical and economic viability and combined these with previous estimates using a common methodology and The USGSassessedseven regionsindepth mind. important caveats should be borne in as the global URR in what follows, these Although theestimateswillbereferredto from beingaccessedandexploited). other constraints may prevent resources availability up to 2025 (since political and and imposed a minimum field size, which ranged from 1 to 20 million boe depending upon the individual AU. on a world scale. The were study judged excludedto be ‘significant’ non-conventional resources such as the Canadian oil sands 95 The128provinces weredividedinto270AssessmentUnits(AUs, seeBox 4.1)andassessments weremadeof246AUsthat 94 PreviousUSGSassessmentsdid notspecifyaparticulartimespan. technology. between 1995and2025usingexisting had thepotentialtobeaddedreserves (USGS 2000)consideredresourcesthat The World Petroleum Assessment 2000 6.2.1 Methods Information Administration (EIA,2008a). Agency (IEA, 2008) and the Energy bodies such as the International Energy underlie projections of global oil supply by (Laherrère, 2001).Theynevertheless commentators have disputedtheirvalidity previous USGS estimates and several The resultsweresignificantlylargerthan five years (Ahlbrandt, 2002;USGS, 2000). team of 41 geoscientists over a period of following 100person-years ofeffortbya assessment was published in 2000, the world as a whole. The most recent which are then aggregated to the level of 94 the globalURR(sincesome This requiredassumptions overestimate resource year time horizon of the study. contribute to global oil supply over the 30 they were considered unlikely to from theassessment,presumablybecause 528 provinces wereexcluded altogether petroleum (USGS, 2000). contained 95% of the world's discovered 128 non-US provinces, 76 of which 409. Formal assessmentsweremadeof the East Greenland Rift) giving a total of assessment (NorthBarents,Provence, and only threeofthesewereincludedinthe considered likely to contain petroleum, but US). An additional five provinces were (354 outsidetheUSand52within known tocontainpetroleumresources petroleum provinces, 406 of which were whole. Itdividedtheworldinto937 1995) toobtainafigurefortheworldas for the United States (MMS, 1996; USGS, the US the outside resources oil crude undiscovered of estimates USGS 6.2 Figure Note: Source:

Frequency of Occurrence Crude oilonly-excludes NGLs.BBO=Gb. 0 0 0 ,0 1,300 1,000 700 400 100 0 USGS (2000) World UndiscoveredPetroleum Resources (excludingU.S.) Oil Volume (BBO) Mean Oil 649 95 F F F By implication, 5 50 95 =1,107BBO =607BBO =334BBO 125

Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Chap6:UKERC 28/09/2009 12:34 Page 126 Page 12:34 28/09/2009 Chap6:UKERC 126 Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production • are: URR of conventional oil. The main points USGS 2000meanestimatesoftheglobal Table 6.1andFigure6.3summarisesthe 6.2.2 Results the study was published. had alreadybeendiscovered bythetime 1996 and 2025. Hence, a portion of this potential tobeaddedreserves between resources refer to resources that had the Similarly, theestimatesof‘undiscovered’ date bythetimestudywas published. discoveries were already five years out of January 1996, the estimates of cumulative Since thestudyusedabaselineof1st (334 Gbto1107Gb). the 95% confidence interval is very large in turnisgreaterthanthemode,andthat is greater than the median (607 Gb) which notable thatthemeanestimate(649Gb) US are illustrated in Figure 6.2. It is oil in the assessed provinces outside the (Charpentier, 2005).Theresultsforcrude shifted lognormal field size distribution undiscovered resources assuming a probabilistic estimates of the size of Monte Carlo simulation which derived maximum size. These provided inputs to a with their minimum, median and number ofundiscovered fields,together the minimum,modeandmaximum modelling. These provided estimates of assessments and discovery process was baseduponamixtureofgeological The assessmentofundiscovered resources The meanestimateofURRwas which derived fromboththeinclusion of previous USGSestimate(2273 Gb) represented a47%increase onthe crude oil and the remainder NGLs. This Gb , ofwhich90.3%(3021Gb)was for 3345 • • • • through toJanuary1996. Howev estimate of URR had been produced Only 21.4% (717 Gb) of the mean reserve additions as new discoveries. to contribute almost as much to future growth atexistingfieldswas expected recoverable resources.Hence,reserve URR and27.8%oftheremaining w The mean estimate for reserve growth a role. In practice, eachislikely tohave played disputed (Bentley, 2002; Mills, 2008). reductions in exploration activity is promising areas or price-induced depletion, restrictedaccesstothemost this declineisaresultofphysical the mid1960s(Figure6.4).Whether been decliningfairlycontinuouslysince global exploration success,whichhas achieved through a major turnaround in reserve additions could only be 1995). Inotherwords,thesepotential Gb in the previous twenty years (1976- previous tenyears (1986-1995)and22 compares toanaverage of14Gbinthe year between 1996and2025.This av the mean estimates imply that an Assuming a constant discovery rate, skewed (Figure 6.2). probability distributionispositively large (495Gbto1589Gb)andthe Gb), but the 95% confidence interval is in the previous USGS assessment (471 is 48%largerthanthemodeestimate remaining recoverable resources.This of the estimated URR and 35.6% of the conv The meanestimateforundiscovered size of NGL resources. significant increaseintheestimated reserve growthforthefirsttimeanda as 730Gb, or21.8%oftheestimated erage of31Gbcouldbefoundeach entional oil was 939 Gb, or 28.1% er , Chap6:UKERC 22/09/200910:18Page127 Source: (mean values) oil URR forconventional global estimated ofthe components 2000: USGS 6.3 Figure Note: Source: (Gb) oil URR forconventional ofglobal estimates mean 2000: USGS 6.1 Table resources recoverable Remaining URR resources Undiscovered Reserve growth reserves Remaining 2P production Cumulative All figuresrefertoJanuary1996. USGS (2000) USGS (2000) Undiscovered resources Future reservesgrowth (939 Gb) US conventional US (730 Gb) 191 362 171 83 76 32 oil World (non-US) crude oil crude 2120 2659 649 612 859 539 World (non-US) Cumulative production NGLs 2P Reserves(959Gb) 317 324 207 42 68 7 (717Gb) conventional oil conventional World Total 2628 3345 939 730 959 717 127

Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Chap6:UKERC 22/09/200910:18Page128 128 Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Source: 1995-2025 period 2000forthe USGS the Comparing historical trendsFigure in 6.4 backdated 2P discoveries with those implied by about reserve growth(Laherrère, 2001). USGS 2000studywere theassumptions The mostcontroversial aspectsofthe • additions at the rate that is required. discoveries fromcontributingtoreserve prevent reserve growth and new More importantly, various factors may estimate ofpotentialreserve additions. growth deliver lessthanthemean if newdiscoveries and/orreserve rapidly. Itcouldalsobereachedearlier more slowlyorearlierifitgrew would bereachedlaterifdemandgrew production at103mb/d).Themidpoint would be reached around 2024 (with the meanestimateofURR(1672Gb) av If consumptioncontinuestogrowatan 1.5%/year. growth in annual production averaging has nowincreasedto33.7%,withthe December 2007of1128Gb, thisfigure with cumulative productionthroughto er IHS Energy;USGS(2000) Discovery (Gb) age of 1.5%/year, the midpoint of 10 20 30 40 50 60 70 80 0 1900 1920 1940 USGS WPA2000 Global Discovery(3yrmov.avg.) 1960 Year relatively crude.Sincetheavailable data for estimating reserve growth was undiscovered resources, the methodology In contrast totheassessmentof reserve growth no longer seemed viable. 1P data. Hence, the neglect of non-US while theUSfunctionwas estimated from database containing2Preserves data fields, despite the Petroconsultants growth functions estimated from US oil have been predictedbythereserve and 1996. This was greater than would US had increased by 26% between 1981 discoveries for 186 giant fields outside the (2000) foundthatthecumulative 2P Petroconsultants database,theUSGS reserve additions. Usingthe for an increasing proportion of global reserve growth appeared to be accounting increasingly inappropriate, given that data. However, this neglect was becoming global assessmentsowingtoinsufficient This hadbeenexcluded fromprevious 1980 2000 2020 2040 Chap6:UKERC 22/09/200910:18Page129 value of zero. probability distribution,withaminimum assumed a symmetrical triangular uncertainty intheseestimates,theUSGS discovered before 1996. To reflect the 2025) of each field that had been This gave estimatesofthe‘grownsize’ (by and gasfieldsthroughouttheworld. single US reserve growth function to all oil outside theUS, theUSGSchosetoapplya reserve growthfunctionsforregions was consideredinadequatetoestimate at the global level. room forarange ofviewsontheneteffect age. Intheabsenceofgooddata,thereis developed as fastasUSfieldsofthesame function, or if these fields have not been determined thehistoricalUSgrowth better technologythanthatwhich next 30 years if non-US fields benefit from underestimate At thesametime,approachcould reducing the potential for future growth). accurate initialreserve estimates (thus likely), or if non-US fields provide more overstated insomecountries(asseems US (astheyare),ifreserves are reserves was lessrestrictive thaninthe growth if the criteria for reporting non-US its approach may functions inthisway. TheUSGSnotesthat the validity ofapplyingUSgrowth considerable debatehas sincearisenover has prevented a rapid increase in exploration capacity (IEA, 2008). partly taken up by price inflation (driven in part by shortages of equipment) and the long lead times for producing new equipme large priceincreasesbetween2000 and2008have stimulatedincreasedexploration activity, buttheincreased investment hasbe (Mills, 2008).Globally, thenumberofdrillingrigspeaked ataround6000in 1981 butsubsequentlyfelltoaround1000in 1999. The 100 For example,some46exploration wellsweredrilledintheUKCS1998,butfollowing thepricecrash thisfelltoonly15 in1999 99 Although,Iran hadthelargestamountofdiscoveries between 1993and2002. has beenfoundinthelasteightyears. Butagain,CampbellandHeapesareusinganarrowerdefinitionofconventional oil. of theargumentthatCampbellandHeapes(2008)underestimate globalYTF, sinceitsuggeststhat60%oftheirtotal(114Gb 98 ThisisoftencitedinsupportoftheargumentthatIHSoverestimated theglobalYTF. Butitcouldequallybeusedinsu 97 Arguably, a minimum value of less than zero should have been used to reflect the possibility that 2P estimates could fall over time 1999; Gautier, 96 Thefunctionwas aweightedaverage oftheoilandgas fieldfunctionsusedinthe1995USnationalassessment(Attanasi, et al ., 1995). 97 reserve growth over the As noted in Section 3, overestimate reserve 96 Mexico. such as Greenland and the Mexican Gulf of and noexploration atallinkey regions historically lowrates ofexploratory drilling during the late 1990s leading to Central Asianrepublics,andlowoilprices economic instability in Russia and the over this period. were lessthanhalfofwhatwas ‘expected’ real-world oil discoveries outside the US undiscovered resource.Inotherwords, discovered by2003or27%ofthe rate, atotalof173Gb shouldhave been provinces. Assuming a constant discovery undiscovered resources (649 Gb) for those than 11% of the mean estimate of 128 assessednon-USprovinces, orless Gb/year) of oil had been discovered in the They estimatedthatonly69Gb(8 assessment period)(Klett, 2003 (i.e. 27% of the 1995-2025 their assessment through to December The USGShave evaluated the‘accuracy’ of 6.2.3 Evaluation as Iraq, Iran access tothemostpromisingregionssuch possible reasonsforthis,includinglimited Klett growth. Ifthisadjustmentwas made,the has notbeenadjustedforfuturereserve of thepreviouseightyears andtherefore estimate ofthecumulative 2Pdiscoveries the 69Gbfigurerepresentsa2003 apparently low discovery rate may be that et al 100 . (2005)highlightanumberof But anadditionalreasonforthe 99 98 and Libya, political and et al ., 2005).

et al pport en nt ., ) 129

Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Chap6:UKERC 22/09/200910:18Page130 130 Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production unexplored regions of the world, such as 2000 assessmentwere forrelatively The mostuncertainestimates intheUSGS could diminishinlateryears. been ‘front-loaded’ byinvestment and strategy. Ifso, reserve growth may have represents a low-cost, minimal-risk growth rather thanexploration, sinceit may have preferredtoinvest inreserve previously omittedfields.Also, companies reporting practices or the inclusion of database may resultfromchanges in the apparentreserve growthintheIHS least atthegloballevel. However, someof relatively well whenappliedto2Pdata-at that theUSreserve growthfunctionworks (Section 3)confirmsthisandsuggests reserve growth between2000and2007 through new discoveries. Our analysis of and morethantwicethereserve additions what was expectedover thefull30years reserve growth by2003,whichis28%of total of171Gbhadbeenaddedthrough well. Klett tracking theUSGSprojectionsrelatively growth atexistingfieldsappearstobe In contrast tonewdiscoveries, reserve importance. future reserve growthisofcritical estimates shouldbeadjustedtoreflect whether and by how much the discovery 2003, despitethelowoilprices.Hence, what was ‘expected’ between 1995 and discoveries weresignificantly Gb whichwouldimplythatreal-world discoveries estimatefrom69Gbto~200 (Verma, 2005).Thiswouldincreasethe increase occurring in the first ten years in 20 years, with the majority of the quadrupling of the initial estimate of URR function (Figure 3.7) projects a ‘modified Arrington’reserve growth discoveries.‘forecast’ For example, the ‘actual’ discoveries may be closer to the et al . (2005) conclude that a more than 2007 to2030was presentedbytheIEAin of conventional oil usingatimeframe from An updatedassessment oftheglobalURR 2008 6.3.1 TheIEAWorld EnergyOutlook estimates RecentglobalURR 6.3 occurring atanearlierdate. could contributetosupplydifficulties resource is actually there. This in turn be realisedbefore2025,even ifthe additions identifiedbytheUSGSwillnot latter continue,thepotentialreserve economic andtechnicalconstraints. Ifthe regions was restrictedbyvarious political, while exploration inmany oftheincluded altogether fromtheUSGSassessment, and Mexico. Many ofthesewereexcluded of Argentina,Colombia,Peru, Venezuela and the Middle East, and offshore regions Kenya and Namibia, large parts of Libya parts ofAfrica,theoffshoreregions that remain poorly explored, including There arealsolargeareasoftheworld remaining recoverable resources by 5.5%. petroleum liquidsby3.7%andthe would increasetheglobalURRfor and 44GbNGLs)(USGS, 2008).This undiscovered petroleumliquids(90Gboil provided ameanestimateof134Gb Arctic assessmentin2008,theUSGS before 2025. However, in an updated unlikely tocontribute to global supply presumably becauseitwas considered was excluded fromtheassessment, exploited. Similarly, much of the Arctic the rate at which resources can be natural difficultiesare likely toconstrain the East Greenland Rift where formidable Chap6:UKERC 22/09/200910:18Page131 recoverable resources). The estimates of estimated at805Gb(33%ofremaining that fromundiscovered resourcesis remaining recoverable resources),while growth isestimatedat402Gb(16%of remaining contribution from reserve Gb ofproductionover thisperiod.The discoveries have more than offset the 411 indicates thatreserve growthandnew have increasedby29%since1996which Remaining 2Preserves areestimated to resources is2448Gb, or68%oftheURR. estimate ofremainingrecoverable than theearlierUSGSestimate.Themean oil tobe3577Gb estimate the global URR for conventional The mostnotablepointisthattheIEA in Table 6.2andFigure6.5. for totalpetroleumliquidsaresummarised own databases and analyses. The results information fromtheUSGSandIEA’s assessment (Klett, recent evaluation oftheUSGS2000 with informationfromIHSEnergy, a 2008). ThisupdatedtheUSGS2000study its World Energy Outlook 2008 (IEA, Source: (Gb) liquids URR forpetroleum ofglobal estimates mean IEA 2008WEO: 6.2 Table resources recoverable Remaining URR resources Undiscovered Reserve growth reserves Remaining 2P production Cumulative IEA (2008) et al which is6.9%larger OECD 307 670 185 363 ., 2007), additional 27 95 Non-OECD 2142 2907 1147 620 375 765 remaining recoverable resources: and gives the followingestimatesof cost curve which includesthiscategory IEA alsoprovides along-termoilsupply definition of this category is unclear. The unconventional means’, but the precise ‘conventional oil produced by The estimatesinTable 6.2exclude and the Former Soviet Union. located in the Middle East, North Africa outside the OECD, with the majority being recoverable resourcesareestimatedtolie notable that the 75% of the remaining production and/or 2P reserves. It is of bothhave alreadybeenconverted into 2000 study, in part because a proportion resources arelowerthanintheUSGS reserveundiscovered growthand • • • • ultra deepwater: recovery (EOR): Arctic: Deepw Enhanced oil Conventional oil: World 2448 3577 1241 1128 805 402 ater and % diff from USGS 2000 USGS -52.8% -14.3% -44.9% 29.4% 57.3% 6.9% 90 Gb 160 Gb 400 - 500 Gb 2100 Gb OECD as % OECD of total 24.7% 18.7% 23.0% 32.2% 6.7% 7.7% 131

Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Chap6:UKERC 22/09/200910:18Page132 132 Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production found before2030. Gb) oftheundiscovered resources willbe Similarly, itforecaststhatonly14%(114 than 12% of the estimated potential. 24 GbfromEORbefore2030whichisless forecasts acumulative contributionofonly estimates seentodate.However, theIEA supply curve is one of the largest 6.2. Indeed,theURRimpliedbyIEA significantly largerthanthefigureinTable conventional oilof resources (3148 Gb) and implies a URR for estimate oftotalremainingrecoverable contribution fromNGLsleadstoahigher estimates andadding the implied the intervening 12years. 101 Thisislargerthanthecorresponding estimateintheUSGS2000study(317Gb),despitesignificant productionofNGLsin oil URR forconventional global estimated ofthe components IEA 2008: 6.5 Figure latter is348Gb. so, theimpliedmean estimateforthe omission ofNGLsfromthecostcurve. If is unclear, but could be due to the in Table 6.2 (2448 Gb). The reason for this conventional oil(2100Gb)thanindicated the remainingrecoverable resourcesof The cost curve gives a lower estimate of Future reservesgrowth (402 Gb) Undiscovered resources 101 (805 Gb) 4276 Gb Combining these which is 2P Reserves(1241Gb) unassessed provinces. Thisisequivalent estimate of593Gboilcontainedinthe the USGSestimatestoobtainamean an estimate of the global URR and subtract provinces, extrapolate thiscurve togive distribution’ model to the assessed provinces, they fit a ‘variable size Africa. To estimatetheresources inthese the Arctic,Antarcticandsub-Saharan excludes 528provinces inregionssuchas underestimates totalresourcesbecauseit First, they argue that the USGS study increase the USGS estimates in two ways. reserve growth since 1996. Instead, they allow forproduction,discoveries and they donotupdatetheUSGSfiguresto the USGS2000studybut(unlike theIEA) (2009). Their estimate is also based upon global URRisprovided byAguilera, A comparably optimisticassessmentofthe Inclusionofadditionalprovinces 6.3.2 Cumulative production (1128 Gb) et al. Chap6:UKERC 22/09/200910:18Page133 growth multiplierstothe meanestimates yield. Third,theapplicationofreserve viability ofextraction andthenetenergy serious questionsaboutboththeeconomic than theassessedprovinces, raising significantly smaller(inresourcesize) Second, the unassessed provinces are for either technical or political reasons. relatively inaccessibleinthemedium-term many ofthosethatdoarelikely toremain provinces contain recoverable liquids and unreasonable toassumethatallofthe937 in a number of respects. First, it may be Aguilera URR. more optimisticestimatefortheglobal IEA’s assumptions could lead to an even combination of their approach with the no explicit assumptions about EOR, a different route. Since Aguilera, above), but derived through an entirely implied by the IEA’s supply curve (see Aguilera, 102 SinceAguilera, recoverable resourcesand an estimateof3516Gbforremaining The netresultofthesetwoadjustmentsis unassessed provinces. much as50%forboththeassessedand they adjusttheseestimatesupwards byas Using the USGS reserve growth functions, also applies to undiscovered resources. Second, they assume that reserve growth estimate of the ‘long-term’ global URR. provinces, theypotentiallyobtainabetter study. Byincludingestimates forthese resources identified by the original USGS to onefifthoftheremainingrecoverable URR. 102 et al’s estimate of remaining recoverable resources to the USGS figure for cumulative production through to 1996. et al This figure is comparable to that .’s approachisquestionable et al.areworkingwiththeoriginalUSGS data(whichappliesto1996),theURRisestimatedherebyadding 4233 Gb et al. make for the detail. and flows (global production) in more relationship betweenstocks(globalURR) The followingsectionexploresthis may slow the rate of post-peak decline. of the global oil supply peak, although it may notmake any differencetothetiming optimistic URRestimatesarecorrect,this to medium-term.Hence,even ifthemore contribute to global oil supply in the near- will be accessed sufficiently quickly to the resource is there, but whether it can or because theissueisnotsimplywhether of global oil supply is questionable. This is medium-term forecasts(e.g.upto2030) URR estimates, their relevance to IEA have ‘pushedtheenvelope’ ofglobal While thestudiesbyAguilera these resources. ‘double counting’futurereserve growthfor ‘grown’, Aguilera their undiscovered volumes arealready technology. SincetheUSGSstatethat about recovery factorsthatreflectmodern and are implicitly based upon assumptions confidence intervals toreflectuncertainty information, alreadycontainwide resources are based largely on geological contrast, theestimates ofundiscovered technology over thepast50years. In reporting andimprovements inrecovery factors suchasconservative initial exploratory drilling, andtheyreflect estimates of field size derived from old) fields,theyrelatetooriginal from studiesofexisting(frequentlyvery to justify. Thesemultipliers werederived of undiscovered resourcesseemsdifficult et al.appeartobe et al and the 133

Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Chap6:UKERC 22/09/200910:18Page134 134 Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Brecha, 103 Earlierinvestigations alongtheselineshave beenconductedbyBartlett (2000),Carlson(2007a;b)andBrecha(2008; with thismodel,a125%increaseinthe level of42Gb/year (115mb/d).Hence, Gb, it gives peak production in 2032 at a (82 mb/d), while for an estimate of 4500 production in 2009 at a level of 30 Gb/year of 2500Gb, themodelgives peak from 2000Gbto4500Gb. For anestimate assumptions for the global URR ranging Figure 6.6 shows the results using size oftheglobalresource. the date of global peak production to the URR, we can investigate the sensitivity of assuming arange ofvalues fortheglobal logistic model of oil production. By this uncertainty with the help of a simple It isusefultoexploretheimplicationsof projections of future global oil supply. leads toacorrespondinguncertainty in the high fortheforeseeablefutureandthis degree ofuncertaintyislikely toremain potential of enhanced oil recovery. The undiscovered resources and/or the future assumptions about reserve growth, end may resultfromexcessively optimistic definition ofconventional oil,theupper this range results in part from an narrow lowest estimate. While the lower end of the resources isfourtimeslargerthan highest estimateofremainingrecoverable 3170Gb 870 to recoverable resourcesfallwithintherange estimates ofthequantityremaining Gb 2000-4300 for conventional oil fall within the range Contemporary estimates oftheglobalURR global supply Implicationsforfuture 6.4 104 Thisislikely tobeanoverstatement, since thediscovery figurehasnotbeencorrected forfuturereserve growth. et al., 2007) amongothers. , whilethecorresponding . In other words, the 103 production cycle. about thefutureshapeofglobal URR are compounded by uncertainties uncertainties aboutthesize oftheglobal size and longevity of revenues. Hence, producers restrict exports to maximise the and enhancedoilrecovery andas major provide incentives fordemandreduction decline more slowly as price signals fields. Alternatively, production could production shifts towards much smaller the giant fields are depleted and that productioncoulddeclinerapidly as logistic model. For example, it is possible is very unlikely to follow the symmetric scenarios andtheglobalproductioncycle the levels indicated in the ‘late peak’ However, future demand may not reach uncertainty over the size of the resource. production must be less than the range of uncertainty over thedateofpeak These resultssuggestthattherange of years. delay the global peak by less than four those oftheentireUnitedStateswould the discovery ofresourcesequivalent to that year. times greaterthanglobaldiscoveries in production in2007andalmostseven is two and half times greater than global 78 billion barrels to the global URR, which year wouldrequiretheadditionofsome delay the date of peak production by one of peak production by only 4.7 days. To URR by one billion barrels delays the date Put another way, increasing the global date ofpeakproductionbyonly23years. size ofthe remainingresource),delays the size ofthe URR(ora260%increaseinthe 104 To putthisinperspective, Chap6:UKERC 28/09/2009 12:36 Page 135 Page 12:36 28/09/2009 Chap6:UKERC and/or arelatively steeppost-peak decline production increase prior to the peak (e.g. at least 3000 Gb), slow rates of requires acombinationoflargeURR in global oil production beyond 2030 Their resultsimplythatdelaying thepeak on thedateofpeak(Figure6.7). and declinerates hasrelatively littleeffect value of URR, changing the initial growth between 2009and2031. scenarios, thedateofpeakproductionlies production. Inparticular, differences inthetimingofpeak the assumed URR lead to relatively small model, theyfindthatlargedifferencesin conventional oil.Aswiththesimplelogistic scenarios for the global production of (Figure 6.7),theygenerate 64possible production before and after the peak assumptions abouttherate ofchange assumptions about the global URR with testing. By combining different these uncertainties through sensitivity Kaufmann and Shiers (2008) address about the global URR simple logistic model - assumptions different under production oil conventional global of peaking The 6.6 Figure

Production (Gb) 10 15 20 25 30 35 40 45 0 5 8010 9020 0020 102200 2150 2100 2050 2000 1950 1900 1850 2001 2009 2016 2023 For any given for 53of64 2028 2034 Year extent be considered a ‘best case’. the singlepeakscenario may tosome extended into a multi-year plateau. Hence, rapid decline may also result if the peak is very rapidly afterthesecondpeak.More the firstpeak,orthatproductiondeclines second peakismuchlowerthanthatfor the annual rate of production for the but the URR constraint implies either that also consideredscenarioswithtwopeaks, fall morerapidly. Kaufmann andShiers be developed faster and/or demand must which in turn implies that substitutes must Later peaks imply faster rates of decline to the current output of Saudi Arabia. 10 years after thepeakwhichisequivalent production fallingby10mb/dwithin5to majority of scenarios show annual implications oncethepeakispassed.The unlikely andcouldalsohave serious asymmetry intheproductioncycle seems assumptions tohold.Thisformof peak after 2040 and these require all three rate. Only three of their scenarios give a URR = URR = URR = URR = URR = URR = Global Production 4500Gbbls 4000Gbbls 3500Gbbls 3000Gbbls 2500Gbbls 2000Gbbl 135

Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Chap6:UKERC 22/09/200910:18Page136 136 Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production as follows: The mainconclusionsfrom thissectionare 6.5 Summary constraints (IEA,2009a). which couldlead tonear-term supply cancellation of many upstream projects However, the recession has also led to the did theearlier‘oilshocks’of1970s. be todelay thepeakbyseveral years - as conventional oilresource,theresultmay constraints imposedbythesize ofthe possibly beyond. Consideringonlythe anticipated tofallfurtherin2009and a result of the economic recession and is The production of all-liquids fell in 2008 as to considerlowerrates ofdemandgrowth. Kaufmann andShiersstudyisthefailure However, animportantweaknessofthe (0.51, 0.165),(0.545,0.545)and(0.575,2)percentforthelow, mediumandhighcases,respectively Note: asymmetric ofthe Illustration 6.7 Figure Cumulative productionsumsto~600Gbinallcases.Theinitialrate ofgrowthandrate ofdeclinebeforeandafterthepeak ar Production (Gb/y) 10 20 30 40 50 0 01 02 03 04 50 45 40 35 30 25 20 15 10 5 0 High Med Low production model ofKa model production Year • • y been trendingupwards forthelast50 Although global URR estimates have variables. constraints tobeplacedonthelasttwo available doesnotallowstrong growth. The information currently inconsistent treatmentofreserve surrounding OPECreserves andthe of URR,togetherwiththeuncertainty differing time-frames for the definition competing reserve definitions and difficulties arisefromtheuseof particularly narrowdefinition.Further the globalURRresultinpartfroma and themorepessimisticestimatesof differing definitionsof‘conventional oil’ Comparison is complicated by the methods, assumptions and results. conv Estimates of the global URR for ears, the estimate of 3345 Gb by the entional oil vary widely in their ufmann and Shiers (2008) Shiers ufmann and e Chap6:UKERC 22/09/200910:18Page137 • • contribution fromEOR thatthey estimate reliesupon alarge reserve growth.However, theIEA generous assumptions about future from the USGS study together with contained within provinces excluded Gb) by including the resources comparably optimistic estimate (4233 Aguilera, increases this figure to 4276 Gb. produced by unconventional means’ Gb, buttheinclusionof‘con estimates theglobalURRtobe3577 The 2008IEAWorld EnergyOutlook premature. study is‘discredited’areatbest repeated assertionsthattheUSGS for futurereserve growth.Hence,the adjust the discovery estimates to allow exploration activity and the failure to reductionsin regions, price-induced exploration in the most promising consequence ofrestrictionson by the USGS, this is partly a appears tobelowerthan‘anticipated’ while therate ofnewdiscoveries observed in theIHSdatabase.Also, consistent with the reserve growth in that they are broadly ‘correct’ indicates that these have proved growth. Howev (2008) andAguilera, more recent estimates by the IEA yet torepeattheirglobalassessment, USGS estimate. While the USGS have being 47%largerthantheprevious departure fromthehistoricaltrend– USGS (2000) represented a substantial its assumptionsaboutglobalreserve The USGS has been widely criticized for URR inexcess of4000Gb. been even more optimistic – implying a et al er , subsequentanalysis (2009) arrive ata et al.(2009)have v entional oil • • global URRforcon Overall, contemporary estimatesofthe assumptions appear very questionable. realised while some of Aguilera anticipate willtake decadestobe constraints in the near-term. investment which could lead to supply recession has inhibited upstream be delayed. At the same time, the ‘resource-constrained’ peakislikely to reduced demand,thedateof the globaleconomic recessionhas evaluated on this basis. However, since predict no peak before 2030 should be conventionally assumed.Forecasts that rates ofdemandgrowththanare post-peak decline rate and/or slower URR combinedwitharelatively steep optimistic assumptions about the global the peakbeyond 2030requires lie between 2009 and 2031. Delaying the date of peak production is found to production beforeandafterthepeak, global URRandtherate ofchange assumptions aboutthesize ofthe and Shiers, 2008). For a range of the global production cycle (Kaufmann for varying degreesofasymmetryin sophisticated modelisused,thatallows substantially changedifamore by only4.7days. Thisresultisnot would delay thedateofpeakproduction the global URR by one billion barrels In asimplelogisticmodel,increasing and the date of peak production. projections offutureglobaloilsupply corresponding uncertainty in the 3170 Gb. This wide range leads to a resources fallwithintherange 870to quantity ofremainingrecoverable the corresponding estimates of the within the range 2000-4300 Gb, while v entional oilfall et al’s 137

Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Chap6:UKERC 22/09/200910:18Page138 138 Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production • inhibit demandgrowth.Furthermore,if high cost, supply constraints could only beaccessedrelatively slowlyat particular level. Iftheseresourcescan maintain globalproductionata accessed at the rate required to that theresourcecanorwillbe correct, it does not necessarily follow Even ifthelargerURRestimatesare produced. resource canandwillbeaccessed to 2030isthe rate atwhich the Hence, the primary issue for the period especially intheimportingcountries. could beconstrained further– maximize production, demand growth key producerslacktheincentive to Chap7:UKERC 22/09/200910:19Page139 international oilcompanies. 105 NotableomissionsincludePFC Energy, CambridgeEnergyResearch Associates(CERA),theWorld EnergyCouncilandsome before 2030, and those which forecast no insurmountablesupplydifficulties dichotomy betweenmodels which forecast oil demand.Therehaslongbeena multiple determinantsofoilsupplyand/or complexity that seek to capture the by models of varying degrees of Forecasts ofglobaloilsupplyareproduced 7.1 Approach concludes. lessons fromthecomparison.Section7.8 while Section7.7summarisesthemain on theforecasttimingofpeakproduction, discovery and reserve growth could impact rate ofreserve additionsfrom new compared. Section 7.6 discusses how the assumptions behind each forecast can be 7.5 provides aframework inwhichthe implied bytheseforecasts,whileSection URR ofconventional oilthatisassumedor supply. Section7.4discussestheglobal contemporary forecastsofglobaloil provides anoverview offourteen between 1956and2005,whileSection7.3 summarises some of the forecasts made Following an introduction, Section 7.2 in comparison of these forecasts is provided production declinerate. Amoredetailed conventional oil and the post-peak ultimately recoverable resource(URR)of implicit assumptionsregardingthe highlights theimportanceofexplicitor provide anhistoricalperspective. It and also reviews earlier forecasts to contemporary forecasts of global oil supply This sectioncomparesandevaluates some 7. Technical Report7. Possible futures –acomparisonofglobalsupplyforecasts overall conclusions. that theseomissionsmateriallyaffectthe reasons. collaborate forcommercialorother contacted orbecausetheywereunableto individuals ororganisationscouldnotbe are notincluded,eitherbecausethe completeness. Several significantmodels statements ontheiraccuracy or models rather thanproviding definitive level, focusing on general aspects of the The comparisons are at a relatively high totals wherever possible. the latterfromforecastproduction conventional liquids.We have deducted oil sandsandsomeincludeothernon- bitumen andsyncrudesfromtheCanadian However, mostofthemodelsalsoinclude be challenging(Hirsch, development of substitutes - which could either demandreductionorthemorerapid conventional oilsupplyimplytheneedfor 2030. More pessimistic forecasts of forecasts of The comparison focuses primarily on Section 5. than oneofthecategoriesidentifiedin several of the models encompass more While bottom-upmodelspredominate, combinations ofapproaches)isused. wide range of approaches (or is produced by a different model and a while five do not. Each of these forecasts conventional oilproductionbefore2030, these forecasts predict a peak in of globaloilsupplyupto2030.Nine comparing fourteenforecastsofthefuture some lightonthisdichotomy by production. Thissectionseekstoshed that the world is at or near the peak of oil 105 conventional However wedonotconsider oil supplyupto et al ., 2005). 139

Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Chap7:UKERC 22/09/200910:19Page140 140 Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Box 7.1 Examples of optimistic overly forecasts ofBox oil supply 7.1 realised byimproving recovery factors. reserves, the oil yet-to-find, and that to be then discovered but classed as probable this ignored the large volume of oil already proved reserve toproductionratio. But years was largelybasedontheglobal in30 The 1970sviewofoil‘runningout’ al., 2002) suggests the need for humility (Craig, of long-term energy supply forecasting well (Box 7.1). In general, the poor record the latterhave frequentlyproved wrong as sooner than optimistic forecasts and that wrong by the historical record much forecasts will,bydefinition,beproved remembered thatpessimisticsupply for the global peak. But it must be others that have given premature dates the forecasts from Campbell, Deffeyes and of oil in about 30 yearsout’ and partly by the 1970sthatworldwas likely to‘run partly informedbypronouncementsfrom forecasts have beenincorrect’. Thisviewis frequently asserted that ‘all past oil In thediscussionofoildepletionitis HistoricalOilForecasts 7.2 • • • • reach $12-$19. Actual production was 10.2 mb/d with a higher price lev (1971) toforecastthatUSproductionwouldreach 19mb/din1980ifnominalprices In 1974Adelman to thelevels impliedbythiselasticity. 2.9-3.7%. Given thelargeincrease inrealoilprices,disco Eysell (1978)estimated thata1%riseinrealoilpriceswouldincrease discoveries by level and the opening up of the Alaska North Slope. reached $12-$19.Actualproductiondeclined to10.5mb/d,despiteahigherprice production wouldclimbfrom11.2mb/din 1971 to13.4mb/din1985ifprices In 1971,theNationalPetroleum CouncilusedDelphitechniquestopredictthat USoil estimate of the US URR (including Alaska) is 362 Gb (USGS, 2000) estimated aURRof587-620Gb. For comparison,themostoptimisticcontempor Zapp (1961) estimated a URR of 590 Gb for the Lower 48, while Moore (1971) et al (MIT energyLab1974)usedamodelbyErikson and Spann et occurring around 2010. global production of conventional oil as would have predictedthepeakfor about 2000Gb of fitting modelwithaURR these been factored in, a simple curve- 7.1). Campbell(2005)arguesthat,had 1973 and1978oilpriceshocks(Figure and fuel substitution that followed the largely becauseofthedemandresponses type peak, did not take place. This was production increase,leadingtoaHubbert- In the event, the anticipated global production. techniques to forecast a peak in global oil of the forecasts that have used such 2000 (Figure7.1).Table 7.1listsanumber until reachingapeakaroundtheyear continue toincreaseforsome30years supply, thisimpliedthat productioncould Assuming asimplelogisticmodelofglobal conventional oilofaround2000Gb. general consensuswas foraURR estimated globalURR.Inthe1970s, fitting’ techniquesconstrained bythe forecasting was tousesimple‘curve- recognised that a better approach to Not surprisingly, even inthe1970sitwas v eries have notcomeclose el. ary Chap7:UKERC 22/09/200910:19Page141 Source: which adatefor gave the peak between made 2005, 1956and production, oil ofglobal forecasts Selected 7.1 Table 2005 2004 1998 1996 1995 1981 1979 1979 1977 1977 1976 1972 1972 1969 1956 Date Bentley and Boyle (2008) Petroconsultants Ehrlich World Bank PFC Energy Conference Report: UN Deffeyes Ivanhoe Hubbert Hubbert Hubbert UK DoE Author Shell Esso IEA BP et al. Conventional oil Conventional oil Conventional oil Conventional oil Conventional oil Conventional oil Conventional oil conventional oil conventional oil conventional oil conventional oil conventional oil conventional oil conventional oil conventional oil Liquids covered excluding NGLs Conventional & Conventional & Probably Probably Probably Probably Probably Probably non- non- (reference URR (Gb) URR ~ 2000 case) 2500 2100 2100 1350 1250 1900 2000 2300 1800 1900 n/a n/a n/a plateau to2035ifproductionflat. “plateau ~ turn of the century.” “about the year 2000” [at 35 Peak (non-communistworld): 1996 if‘unconstrained’ logistic; “plateau withinthenext25 “increasingly scarce from “likely peak by 2000.” 2000 [at100mb/d] Date ofDate global peak 1990 [at 65 mb/d] “about 2000” About 2010 About 2005 ~ 2000.” years.” mb/d] 1985 2000 2005 2018 2014 141

Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Chap7:UKERC 22/09/200910:19Page142 142 Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production attention toeconomicvariables. of oildemandand pay insufficient most continuetoneglect thedeterminants output rather than an input. However, and estimatingtheglobalURRasan explicit assumptionsaboutdeclinerates field orregionalforecasts,relyingupon derive globalforecastsfromthesumof disaggregate approaches.Thesetypically increasingly being replaced by more by anassumedglobalURRare Moreover, curve-fitting models constrained the uncertaintyinsuchforecasts. horizon, andmostauthorsacknowledged assumptions over a thirty year time consequences ofsomevery simple the models was to assess the mechanistic approach.Butthepurposeof conventional oilandtheirexcessively assumptions about the global URR of higher prices, their overly pessimistic demand-side and supply-side responses to criticised fortheirinabilitytocaptureboth These early models have been variously production actual to compared production oil ofglobal forecast logistic Pre-1973 7.1 Figure

Gb/y 10 15 20 25 30 35 40 45 50 0 5 9012 9016 9020 0024 0028 2100 2080 2060 2040 2020 2000 1980 1960 1940 1920 1900 Curve fittopre-1973dataassumingURRof2100Gb Actual Production Year dispute. these assumptionshas beenafocusof 2000 estimatesforthe global URR.Eachof declared reserves andrelyupontheUSGS experience inthe US, accept OPEC’s reserve growth on the basis of historical reserves data,calculatefutureglobal not. Several of them mix 1P and 2P recoverable resources, while others do forecasts make explicitassessments about constraints in particular. Some of these supply ingeneral andgeological less attentiontothedeterminantsofoil detailed modellingofenergydemandand models’ andtypicallyincludemore in anumberofrespectsfromthe‘peaking assumptions usedbythesemodelsdiffer methodologies and peaking’ forecasts.The some ofthemorerecentthese‘non- 2020 or 2030). Table 7.2 summarises the foreseeable future (generally out to be sufficienttomeetforecastdemandfor were forecasts that oil production would Opposed tothese‘peaking’calculations Chap7:UKERC 22/09/200910:19Page143 examine theoilresource base,onthe A number of analysts rule out any need to reach aplateauby2030. conventionalof expect thesupply oilto being describedas“daunting”. TheIEAnow required tomeettheforecastdemandnow future supply, withthelevel ofinvestment contributed toamorepessimisticviewof behaviour of individual fields. This has does muchmoredetailedmodellingofthe significant departure,incorporating asit Energy Outlook2008 by the USGS (2000). The most recent the globalURRofconventional oilpresented result ofthemoreoptimisticassessment Energy Outlook2000 (Table 7.1).Thisviewchangedin peaking intheir who recognisedthepossibilityofglobaloil An importantexampleisthatoftheIEA, Note: Source: 2030 before nopeak forecast that production oil ofglobal forecasts Selected 7.2 Table 2007 2005 2004 2003 2002 2002 2001 2000 1998 Date Conv: conventional. NonConv.: non-conventional. Bentley andBoyle (2008) IEA: IEA: IEA: Deferred Invest. WEC/IIASA-A2 Shell Scenario Reference Sc. Reference Sc. WETO study US DoEEIA ExxonMobil US DoE Author WEO 2005 WEO 2000 WEO 2007 World Energy Outlook 1998 World EnergyOutlook (Table 7.2),partlyasa NonConv. Oil NonConv. Oil NonConv. Oil NonConv. Oil represents a Conv. Oil Conv. Oil Conv. Oil Conv. Oil Conv. Oil covered Conv. & Conv. & Conv. & Conv. & Liquids Liquids World World ~4000 (Gb) 4500 3303 3345 URR URR modelling. Each oftheserequires analysisand significant portionofcurrent production. lead timesrequiredtodisplacea reduction andsubstitutefuelsthe and economic potential for demand behaviour of oil markets, the technical conventional oil production, the involves judgementsaboutthefutureof the relative probabilityofthesescenarios cause disruptionandshortage.Assessing or sufficientlyrapid andunpredictableto predictable toalloweconomiesadjust, the pricerisewillbesufficientlyslowand at someprice,thequestioniswhether true thatsupplywillalways meetdemand (Bentley andBoyle, 2008).Butwhileitis end-use efficiencyandsubstitutefuels a relatively smoothtransition to greater that supply meets demand and encourage grounds thateconomicforceswillensure Forecast date 2025 - 2040 2016 / 2037 Plateau: No peak No peak No peak No peak No peak No peak No peak No peak

of peak of

Global production (mb/ production Global 2020 2030 2020 0 105 100 115 105 118 114 120 102 105 100 - 109 Various - 103 100 90

116 - d) 143

Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Chap7:UKERC 22/09/2009 10:19 Page 144

7.3 Overview of current oil summarised in Table 7.5.107 It is important to note that all the forecasts forecasts reviewed here were developed prior to the global economic recession in 2008 and Now we turn to contemporary forecasts of hence do not reflect the subsequent global oil production. The fourteen models reductions in upstream investment and analysed in this study are listed in Table global oil demand. 7.3, together with summary data about the liquids that are covered, whether Figure 7.2 plots thirteen of the global demand is modelled explicitly, the general forecasts of all-oil while Table 7.3 approach to modelling oil supply, and compares the estimates of cumulative whether a peak in the production of production through to 2030 (BGR is conventional oil is forecast before 2030.106 omitted owing to insufficient data). In the Full descriptions of each model are few cases where alternative scenarios are provided in Technical Report 7, and the provided only the ‘base case’ is shown. main parameters and assumptions (in so However, two scenarios are shown for far as they can be identified) are Shell.

Figure 7.2 Comparison of thirteen forecasts of all-oil production to 2030

120

110

100

90

80

70

60

Production (mb/d) 50

40

30 Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Global Oil Depletion: An assessment of the evidence for a near-term 20 2000 2005 2010 2015 2020 2025 2030 Year BP Statistical Review ExxonMobil: All-oil US EIA: Conventional oil, reference case IEA WEO 2008: Conventional oil OPEC 2008: Conventional oil (rebased) Total 2008: All-oil Miller 2000: All-oil (rebased) Meling 2006: Base case, all-oil Shell: All-oil (Blueprint scenario) Shell: All-oil (Scramble scenario) Energyfiles 2009: All-oil Uppsala: All-oil excluding YTF Peak Oil Consulting 2008: All-oil Campbell 2008: All-oil Notes: LBST: All-oil • Annual global production from 2000 to 2007 taken from BP (2008). • Forecasts refer to all-oil as far as possible, but coverage of liquids does not always coincide. • The OPEC and Miller forecasts exclude NGLs. These forecasts have been ‘re-based’ here to match the BP production figure for 2007. Since the estimated production of NGLs is assumed to remain fixed until 2030, these forecasts may be downwardly biased.

106 The data for each model were either gathered from published literature or by approaching the author(s). Descriptions were written and submitted to the authors for comment and approval (see the Annex of Technical Report 7). Feedback and corrections were received for all but one of the models which is much appreciated. 107 We also obtained the views of ENI and BP on future global oil supply. ENI does not publish quantitative forecasts, but it rules out the possibility of a peak in the global production of oil ‘for the foreseeable future’. BP acknowledges an eventual peak in conventional oil supply, but gives neither a date nor a level of production for this. The company implies that the peak is beyond 2030, and forecasts global all-liquids production in 2030 to be in the range 95-105 mb/d, saying that this level is ‘sustainable’. BP notes a conventional oil YTF of at least 300-400 Gb, and agrees with the USGS assessment of global reserves growth, citing a figure of up to 700 Gb.

144 Chap7:UKERC 22/09/200910:19Page145 Table 7.3 Forecasts comparedTable in 7.3 this study

Category Model Liquids covered Detailed Approach to supply modelling Conventional oil demand peak forecast modelling? before 2030?

International IEA All-liquids Yes Bottom-up; incremental supply No constrained by investment OPEC All-liquids Yes Top-down No National US EIA All-liquids Yes Top-down, some individual country No forecasts BGR (2006) Conventional oil No Assumes mid-point peaking Yes

Oil companies Shell All-liquids Yes Bottom-up by field or country, Yes demand constrained (demand driven)

Meling (Statoil All-oil No Bottom-up by country Yes Hydro)

Total All-oil No Bottom-up by field or basin Yes Exxon Mobil All-liquids Yes Top-down (little detail provided) No Consultancies Energyfiles All-oil No Bottom-up by field or basin Yes LBST All-oil No Bottom-up by field or region. Simple Yes curve-fitting for pre-peak countries Peak Oil Consulting All-oil, GTL and No Top-down for current production, Yes biofuels, bottom-up for new production Universities Colin Campbell All-oil No Bottom-up by country. Assumes Yes and mid-point peaking and constant individuals post-peak depletion rate University of All-oil No Bottom-up for giant fields, top-down Yes Uppsala for other sources Richard Miller Crude oil No Bottom-up by field Yes (excluding NGLs and some condensate) 145

Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Chap7:UKERC 22/09/200910:19Page146 146 Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production 108 BPandENIareinbroadagreement withtheseforecasts. oil andforMillerOPEC thecontribution crude oil,butnotallincludeextra-heavy forecasts includelight,mediumandheavy forecasts includedifferentliquids.Allthe One causeforthisrange isthatthe conclusions. detailed studiescanreachsuchdifferent future. It may seem remarkable that the immediatepasttoindefinite range offorecastpeakdateranges from and-a half times the lowest, while the production ofall-oilin2030isover two- As Figure 7.2 shows, the highest forecast • • • • Notes: Comparison of thirteenfo 7.4 Table Miller Uppsala Campbell Peak OilConsulting LBST Energyfiles ExxonMobil Total Meling (StatoilHydro) Shell (Scramble) Shell (Blueprint) US EIA OPEC 2008 IEA WEO2008 IEA (2008) estimate the global URR of conventional oil to be 3577 Gb, but many of the forecasts use different assumptions. Miller andOPECforecasts‘re-based’toincludeanestimatedcontributionfromNGLs. Forecasts refertoall-oilasfarpossible,butcoverage ofliquidsdoesnotalways coincide. Cumulative globalproductionto2007taken from BP(2008). Production in 105.2 102.8 /d b m 2030 91.5 67.1 60.0 65.0 39.3 78.6 93.1 94.1 85.6 91.4 93.5 95.0 recasts of all-oil ofrecasts all-oil Cumulative 2008-2030 2008-2030 production /d b m 888 697 609 670 504 748 796 773 764 724 748 780 738 747 OPEC and Exxon forecastsarefromthe IEA,USEIA, linear’ the endoftheirforecast.These‘quasi- the modellersforeseeapeakitisbeyond production of all-oil to 2030, such that if approximately lineargrowth inthe 7.2. The first group indicates an There aretwogroupsofforecastsinFigure of the difference. only explains a relatively small proportion of NGLshasbeenestimated.However, this shows thecorresponding forecastsforall- more clearlyinFigure7.3,whichalso production to 2030 to production Total cumulative production to 2038 1847 1759 1820 1654 1898 1946 1923 1914 1874 1898 1888 1930 1897 Gb 108 2030 and are illustrated % ofIEA (2008) estimate of URRestimate 57.0 51.6 49.2 50.9 46.3 53.1 54.4 53.8 53.5 52.4 53.0 53.9 52.8 53.0 Chap7:UKERC 22/09/200910:19Page147 Figure 7.3 ‘Quasi-linear’ forecas ‘Quasi-linear’ Figure 7.3 that time.Shell’s forecastsofall-liquids ‘Blueprints’ where production levels off at production declinesafter2020,and scenarios, ‘Scramble’ where all-oil constrained supply. Shellofferstwo and substitutionrather thanby demand duetoefficiencyimprovements that theall-oilpeakisdriven bydeclining demand, andtheirscenariosareuniquein one of this group to explicitly model oil separately in Figure 7.4. Shell is the only by a decline. These are illustrated some formofpeakbefore2030,followed The secondgroupofforecastsindicates assumed demand from 2011 onwards. modelled supplyfailstomatchthe similar, reaching a peak in 2028, but the demand. Meling’s modelissuperficially allocates sourcesofsupplytofillthis primarily bydemandmodellingand biofuels). Eachoftheseforecastsisdriven liquids (i.e., including CTLs, GTLs and

Production (mb/d) 100 110 120 60 70 80 90 2000 2005 OPEC 2008:Conventionaloil(rebased) IEA WEO2008:Conventionaloil US EIA:Conventionaloil,referencecase ExxonMobil: All-oil BP StatisticalReview ts of all-oil and ofts all-oil 2010 Year 2015 falling away. capacity beyond likely demand,before model showsaninitialriseofpotential developed immediately. Consequently this fallow fields and new discoveries to be be achieved, regardlessofcost,wereall maximum productionthatcouldpossibly forecast of actual production, but of the Miller’s forecast is specifically not a 1.6%/year demandgrowthbeyond 2011. but doesnotmeettheassumed falling away. Meling’s forecastpeakslate initially followtherisingdemandbefore Uppsala, Total and Energyfiles forecasts LBST, Campbell,Peak OilConsulting, project level toforecastfuturesupply. The bottom-up methodsatthecountry, fieldor combination ofsimplecurve-fitting and assumptions fordemandgrowthanda The remaining models use exogenous appear to reach a peak until after 2050. production arenotwell-definedbutdo all-liquids to 2030 to all-liquids 2020 Meling 2006:Basecase,all-oil OPEC 2008:All-oilplusbiofuels IEA WEO2008:Allliquidsexceptbiofuels US EIA:Allliquids,referencecase ExxonMobil: Allliquids 2025 2030 147

Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Chap7:UKERC 22/09/2009 10:19 Page 148

The annual rate of post-peak decline of Despite the multiple inconsistencies in the global all-oil production (the ‘aggregate’ coverage of liquids and the wide decline) is variously forecast by the differences in methodology and peaking models to be 0.2% (Total); 2.1% assumptions, it is possible to compare the (Campbell); 2.0% in 2025 rising to 2.3% forecasts by focusing on a limited number in 2030 (Peak Oil Consulting); 2.0% in of key variables. The following two 2022 rising to 3.0% in 2029 (Energyfiles); sections do this by examining: first, the 2.6% (Meling); 3.3% from 2025 (Miller); assumed or implied global URR of and 3.5-4.0% (LBST). The URR of these conventional oil; and second, the peaking forecasts is variously defined,109 interaction between the URR, the forecast but ranges from 1840 Gb (LBST) to 2450 date of peak production and the aggregate Gb (Campbell), 2800 Gb (Miller) and 3149 decline rate following the peak.110 Gb (Meling).

Figure 7.4 ‘Peaking’ forecasts of all-oil production to 2030

120

110

100

90

80

70

60

Production (mb/d) 50 Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Global Oil Depletion: An assessment of the evidence for a near-term 40

30

20 2000 2005 2010 2015 2020 2025 2030 Year

BP Statistical Review Total 2008: All-oil Miller 2000: All-oil (rebased) Meling 2006: Base case, all-oil Shell: All-oil (Blueprint scenario) Shell: All-oil (Scramble scenario) Energyfiles 2009: All-oil Uppsala: All-oil excluding YTF Peak Oil Consulting 2008: All-oil Campbell 2008: All-oil LBST: All-oil

109 It is usually for conventional oil, but may sometimes exclude NGLs or include a portion of the oil sands resource. 110 As described in Section 3, this is the net outcome of the average decline of post-peak fields, the production from fields in plateau and the contribution of new fields which are coming on-stream.

148 Chap7:UKERC 22/09/2009 10:19 Page 149

Box 7.2 Illustration of a ‘peaking’ forecast – Energyfiles

Energyfiles maintains a detailed model for forecasting regional and global petroleum supply, based upon field-level data where available and otherwise on data aggregated at the operator, basin or country level. Demand is not modelled and is simply assumed to grow at a fixed rate.

Detailed profiles of cumulative production are created at the field, basin or country level according to data availability, and are extrapolated by balancing the constraints of past decline rates, regional decline rates and estimates of ‘most likely remaining reserves’. For producing fields, the latter are obtained by extrapolating past decline rates while for undeveloped fields they are based upon announced expected plateau levels or the capacity of facilities, modified where appropriate by announced expected reserves. NGL production is estimated directly from similar profiles of future gas production. This approach is assumed to incorporate future reserve growth.

Yet-to-find resources are estimated through the analysis of the geology and discovery history of basins and plays, with onshore and offshore treated separately. The current estimate of 250Gb is only one third of the USGS mean estimate, in part because some areas (e.g. the Arctic) are considered unlikely to contribute to global supply in the medium-term. Energyfiles assumes that YTF resources in countries past or near their peak will have little influence on the timing of the global peak. However they may reduce the rate of post-peak decline.

In the forecast reviewed here, Energyfiles projects a global all-oil production peak of 95 mb/d around 2017 assuming a relatively high demand growth of 1.8%/year The conventional oil URR is estimated at 2685 Gb and the aggregate post-peak decline rate is estimated at 2.0%/year by 2022 and 3.0% by 2030.

GLOBAL: Oil forecast (regions) 100000 Refinery gain

NB: Includes OPEC capacity from 2010 peak in global oil production Global Oil Depletion: An assessment of the evidence for a near-term 90000 Asia-Pacific off

Asia-Pacific on 80000 Middle East off

70000 Middle East on

Africa off 60000 Africa on

50000 E Europe/FSU off

E Europe/FSU on 40000 W Europe off

30000 W Europe on 000s of bbls oil per day L America off 20000 L America on

10000 N America off

N America on 0 1945 1955 1965 1975 1985 1995 2005 2015 2025

© Energyfiles Ltd Year

149 Chap7:UKERC 22/09/200910:19Page150 150 Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Table 7.5 Summary of Summary the main parametersTable and 7.5 assumptions used each by forecast

Category Model Oil demand Date of global Production in URR (Gb) YTF (Gb) Reserve Decline rates growth peak 2030 (mb/d)111 growth (Gb)

International IEA 1.3%/year 2008- No peak 106.4 mb/d all 3577 Gb 805 Gb 402 Gb Detailed modelling organisations 2015 Conventional oil liquids excluding gives average 0.8%/year 2015- plateau by 2030 biofuels production-weighted 2030 managed decline rate of 6.7%/year for post-peak fields

OPEC 1.14%/year after No peak 113.6 mb/d all- 3345 Gb. Accepts Accepts Natural field decline 2012 liquids USGS USGS rate 4-5%/year

National US EIA 1.16% No peak 112.5 mb/d of all- Not given Not given Not quantified Not given organisations liquids, 109.8 mb/d excluding biofuels

BGR (2006) Not given ~2020 Not given 2979 Gb 632 Gb Not quantified Not given

Oil companies Shell No growth after No peak for all- 85.6 mb/d all-oil Modelled in- Modelled in- Assumes Modelled in-house 2020 (Blueprints). liquids (Scramble) house but house but improvements but not given Decline after 2020 not given not given in recovery (Scramble) ~2030 91.4 mb/d all-oil factors, but not (Blueprints) and (Blueprints) quantified ~2020 (Scramble) for all-oil

Meling 1.6%/year 2028 (base case) 94.1 mb/d all-oil Unstated, 309 Gb 520 Gb We calculate (StatoilHydro) but probably aggregate post-peak 3149 Gb decline of 2.6%/year

Total 1.4%/year 2020 93.1 mb/d all-oil Not given 200-370 Gb Raising mean We estimate global recovery aggregate post-peak factor by 5% decline of 0.2%/year

Exxon Mobil ~1.4%/year No peak 105.2 mb/d all-oil Implicitly Not given Not given Not given follows USGS - 3345 Gb Chap7:UKERC 22/09/200910:19Page151 Table 7.5 Cont. 7.5 Table

Category Model Oil demand Date of global Production in URR (Gb) YTF (Gb) Reserve Decline rates growth peak 2030 (mb/d)111 growth (Gb)

Consultancies Energyfiles 1.8%/year 2017 78.6 mb/d all-oil 2685 Gb 250 Gb. Not given but Aggregate global excluding refinery incorporated production decline gains in model 2%/year by 2022 and methodology 3%/year by 2029. Field decline 5- 30%/year depending on size and location LBST Not given 2006 39.4 mb/d all-oil 1840 Gb Not given Not given We calculate an aggregate post-peak decline rate of 3.5- 4.0%/year Peak Oil Not used 2011-2013 65 mb/d all-oil Not Not Not given Currently 4.5% from Consulting estimated estimated existing producing fields. Aggregate post-peak decline 2.0%/year by 2025 and 2.3%/year by 2030

Universities Campbell Not given 2008 57.0 mb/d all-oil 2425 Gb, 114 Gb, Not given, We calculate an and all-oil, ‘regular’ oil assumed small aggregate post-peak produced by in terms of decline rate of Individuals 2100. impact on peak. 2.1%/year 1900 Gb regular oil University of Not modelled 2008-2018 67.1 mb/d all-oil Not given Not given Assumed to Field decline rates of Uppsala contribute little 6-16%/year

Miller Not modelled 2013-2017 (2019 91.5 mb/d all-oil 2800 Gb 227 Gb 0.2%/year Aggregate post-peak given unlimited excluding NGLs cumulative production decline of investment) increment in 3.3%/year by 2025 production

111 Includes refinery gains where these are distinguished (typically 2-3 mb/d) 151

Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Chap7:UKERC 22/09/200910:19Page152 152 Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production 113 135Gbwerediscovered morethan10years ago, 104Gbmorethan20years agoand64Gbmorethan30years ago. 112 133GbinOPECcountriesand 125Gbinnon-OPEC. • These areasfollows: differing levels ofuncertainty(Figure7.5). five constituents thatareassociatedwith The globalURRcanbebroken downinto peak. production declinesvery rapidly after the only ifconventionalURR, but oil compatible withlowerestimatesofthe below, the ‘non-peaking’ forecastsmay be values of3345-3577Gb. Asdescribed estimates for the URR suggest larger forecast nosuchpeakandalsoprovide 3150 Gb, while the three models which conventional oil to be in the range 1840- before 2030estimatetheURRof Those models which forecast a peak follows. we will refer to conventional oil in what liquids does not always coincide. However, straightforward becausethecoverage of conventional oil, the comparison is not While mostoftheseassumptionsarefor these cover awiderange (Table 7.4). assumptions about the global URR and Only eightofthemodelsmake explicit resource ofconventionaloil global ultimatelyrecoverable Theassumedorimplied 7.4 • C v of cumulative production.We usea scope forlargedifferencesinestimates 1332 Gb, buttheliquidscovered and Reserves: global reserves range from 734 Gb to alue of1150GbforDecember2008. umulative production: The model estimates for There islittle • • estimates thatthereare257Gb for development. TheIEA(2008) fairly conservative values forreserve such asEnergyfilesandMiller, take URR ofindividualfields.Somemodels, provide fairlyaccurate estimatesofthe holding that the industry databases reserve growthasessentiallyzero Reserves Growth: 20 years ago. fields thatwerediscovered morethan contained insmall,isolatedorcomplex not be developed, since they are that uptohalfofthesereserves may these fieldsby2030.ButMillerargues cumulative production of220Gbfrom undeveloped fields and they forecast of thetotal)in1874knownbut conventional oil2Preserves (i.e.20% discov Fallow Fields: smallest reserve estimateis210Gb. difference betweenthelargestand If Campbell’s estimateisexcluded, the deepwater fields,polarfieldsandNGLs. also excludes extra-heavy oil,oilsands, Russian and Middle East reserves and who discountsmuchofthereported estimate (734 Gb) is from Campbell, ambiguity in some cases. The smallest 2P reserves, althoughthereis all themodelsimplicitlyorexplicitlyuse treatment of NGLs and oil sands. Nearly misreporting andthediffering downwards tocorrectforpotential the adjustmentofregionalestimates coincide. Sourcesofvariation include reserve definitionsdonotalways ered but not currently scheduled 113 These fieldsare Campbell takes 112 of , Chap7:UKERC 28/09/2009 12:37 Page 153 Page 12:37 28/09/2009 Chap7:UKERC • Notes: URR of conventional oil global the for assumptions model the in uncertainty of range and Constituents 7.5 Figure • • • • 805 Gb (all conventional oil). For 309 Gb; BGR implicitly 623 Gb; and IEA (crude oil); Energyfiles 250 Gb; Meling oil);Miller227Gb 114 Gb(‘regular’ estimates include(inorder):Campbell referring onlytotheglobalURR.Direct models donotspecifyYTFdirectly, difference of some 690 Gb. Some YTF range from114Gbto805–a Y interim. because ofcumulative productioninthe gives a smaller value of 402 Gb, in part 2008 updateoftheUSGSestimates a referenceyear of1995.TheIEA’s the USGSmeanestimateof730Gbfor growth, while others implicitly accept Compares theassumptionsofeightmodels,togetherwithreserve estimatesfromvarious sourcesthatareusedbythemode reflects global‘all-oil’productionas thebulkof‘non-regular ‘Produced’ column for ‘Campbell regular oil’ sums Campbell’s production data to 1980 and BP (2008) data thereafter. This largel Miller argues that a significant portion of the fallow fields will not be developed. Some authorsassumezero reserves growthwhileothers anticipategrowthbutdonotquantifyit. Gb et-to-Find: 1000 1200 1400 1600 -200 200 400 600 800 0 rdcdFlo ed eevsRsre rwhYTF Reservesgrowth Reserves Fallow fields Produced The model estimates for IEA Total CERA O&GJ, EIA Meling Campbell regularoil IHS ’ oil has been produced since 1980. (LBST) whilethelargestis3577Gb(IEA). smallest (mean) URR estimate is 1840 Gb contribute approximately 100 Gb. The while the uncertainties over fallow fields contribute an uncertainty of some 210 Gb different assumptionsforglobalreserves Campbell's lowestimateisexcluded, the global URR of conventional oil. If uncertainty of1100Gbintheassumed YTF. Taken together, these give an their assumptionsforreserve growthand In conclusion, the models differ largely in IEA. comparable tothat assumed bythe USGS estimatesimplyaYTF 400 Gb. Forecasts relyinguponthe comparison, BPestimatesaYTFof300- BGR Miller BP USGS Energyfiles World Oil ls y 153

Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Chap7:UKERC 22/09/2009 10:19 Page 154

The interplay between the URR, peak date 7.5 The interaction between and post-peak decline rate is shown in ultimately recoverable Figure 7.6. Here all the other parameters resources, the aggregate have been fixed at values typical of the forecasts studied, specifically: decline rate and the date of • Production climbs exponentially to a peak production peak and then declines exponentially at a different rate, producing a sharp peak Here we introduce a way of comparing the (this production cycle is unlikely in different forecasts in terms of the key practice, but serves as a simple assumptions they make, explicitly or approximation). implicitly, for conventional oil production. • Production continues for 100 years after Production forecasts for conventional oil peak, and the cumulative production by vary in two basic respects, namely the area then is the effective URR. Figure 7.6 beneath the curve and the ‘shape’ of the shows URRs of 2600, 2800 and 3000 Gb. curve. Since conventional oil is a finite resource, production forecasts must rise • The growth rate to peak is 1.3%/year over time to a peak or plateau and then fall • The decline from peak is shown for away.114 Even forecasts of quasi-linear values of 2%, 4% and 6%/year production growth up to 2030 must eventually peak and decline. For forecasts • Production in 2007 is 85 mb/d and extending sufficiently far into the future, cumulative production by end 2007 is the area under the curve represents the 1150 Gb. URR. The assumption of a 1.3%/year growth rate The production forecast can be divided into up to the peak is particularly important a growth phase and a decline phase, and since (other things being equal) a faster perhaps a plateau phase. A change in the rate of demand growth should lead to an height and date of the peak implies a Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Global Oil Depletion: An assessment of the evidence for a near-term earlier peak and vice versa. The 1.3% change in either the area under the curve assumption is consistent with the assumed or the shape of the curve. The differences or modelled growth rates in the majority of between the various forecasts can the forecasts (Table 7.4), but these were therefore be viewed as differences in the developed prior to the global economic assumed URR, the growth rate and/or recession of 2008. The recession has decline rate and the shape of the peak or reduced global oil demand which could plateau. Some of the forecasts use an delay the peak in a similar manner to the oil explicit URR and/or shape of the production shocks of the 1970s (Figure 7.1). At the curve as input parameters, while others same time, the recession has led to the generate one or both as an output. But all cancellation or delay of many upstream can in principle be assessed according to investment projects which could lead to how these parameters compare and near-term supply constraints when demand whether they can be considered realistic. recovers (IEA, 2009a).

114 While multiple peaks are possible, a more plausible scenario is a multi-year plateau in which production fluctuates by a few percentage points (Hirsch, 2008).

154 Chap7:UKERC 22/09/200910:19Page155 assumptions made. constraints of which are set by the albeit in the form of a region, the forecast can also be put in this space, making reasonable assumptions the forecast provides onlyone parameter, by be plottedinthisspace.Even wherea specifies two of the three parameters can held fixed. Thus any forecast which increase inproductionpriortothepeakis third isdetermined-provided therate of two of these parameters, the value of the lines. Oncevalues areassumedforany URR values areshownbyanarray ofiso- rate againstpeakyear, wherepotential Figure 7.7,whichplotspost-peak decline rate, andthedateofpeak.Thisisdonein the URR,aggregatepost-peak decline on threeofthekey parameters, namely Figure 7.6 can be re-formulated to focus forces anearlierpeakandvice-versa. rate of1.3%/year, aURRof2800Gbanddeclinerate of4%/year For agiven growthrate andURR,asloweraggregatedecline Note: decline rate aggregate post-peak URR the and the varying of peak of date onthe effect The 7.6 Figure

The circled point, for example, indicates the date of the productionpeak(in2027)that resultsfromanassumedgrowth date of The circledpoint,forexample,indicatesthe Production (mb/d) 100 120 140 20 40 60 80 0 0721 0722 0723 0724 0725 2057 2052 2047 2042 2037 2032 2027 2022 2017 2012 2007 URR =2800Gb,post-peakdecline6%p.a. URR =3000Gb,post-peakdecline4%p.a. URR =2800Gb,post-peakdecline4%p.a. Year sufficiently robust to be useful. detail, theoverall pictureremains assumptions donotmatcheachforecastin production. Whiletherequired conventional oil,pluscurrentoilsands as possible,theforecastsarelimitedto summarised inBox 7.3and Box 7.4.Asfar in constructingtheseellipsesare appear to the left. The assumptions made right of 2030, while the ‘peaking’ forecasts conventional oil production appear to the forecasts that show no peak in forecast tobelocated).The‘quasi-linear’ insufficient informationtoallowtheir as ellipses in this space (Exxon provides Figure 7.7showsthirteen of theforecasts URR =2600Gb,post-peakdecline4%p.a. URR =2800Gb,post-peakdecline2%p.a. 155

Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Chap7:UKERC 29/09/2009 09:58 Page 156 Page 09:58 29/09/2009 Chap7:UKERC 156 Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production of ~6.7%/year expected (although thisis thisgives(Section 3.3), anupperbound production-weighted estimate for 2007 onstream.Takingfields coming theIEA's because there will always be some new ratemanaged decline ofpost-peak fields decline rate. shouldbelessthan the This placed upontheaggregatepost-peak Box 7.4). However, some bounds may be is notalwaysinformation provided (see difficult tolocate,sincetherelevant The quasi-linearforecastsaremore date of peak within this group. accounts forthedifferencesin the forecast different assumptionsfortheURRthat Apart from Total, it is primarily the and sometimes the effect of an early peak. output. Thesearesometimesthecause decline rates, whether as an inputor an of theseforecastsaremarked bylow to locate on Figure 7.7 (see Box 7.3). All The peakingforecastsarerelatively easy is 1.3%/year. Mappingofindividualforecastsontothisgraph involves somejudgment Note: the dateof peak production and aggregate the post-peak decline rate. MappingFigure 7.7 global supply forecasts according to the implied URR of conventional oil, Iso-lines representtheassumedorimpliedglobalURRofconventional oil.Assumesrate ofincrease production priortothe

Forecast post-peak decline rate 0% 1% 2% 3% 4% 5% 6% 7% 8% 0020 0021 0022 0023 0024 2050 2045 2040 2035 2030 2025 2020 2015 2010 2005 2000 URR =3400Gb URR =2400Gb Campbell LBST Uppsala URR =3600Gb URR =2600Gb Energyfiles Total Forecast dateofpeak BGR Miller URR =2800Gb Shell Meling beyond 2030. 2.5%/year tojustify bedifficult would ratesconsider thatdecline of lessthan beisopentodebate.Wewhat thatshould although considered precisely reasonable, minimum declinerate thatcanbe these considerations constrain the subject to constraints. Taken together, and asdescribedinSection3,thisis the rate theycanbeproduced– atwhich eachyearto beadded upon willdepend The volume that of needs new resources fields or the development of fields. fallow discoveries, EOR projects at existing new projects. These could be either new productionfrom be metbyincremental decline rate to needs totalproduction of post-peak fields and the aggregate declineratebetween themanaged of URR. Furthermore,thedifference rates theglobal largerestimatesof imply to increase). Lower aggregate decline URR =3000Gb (see Box 7.3 and Box 7.4). OPEC IEA URR =3200Gb USEIIA peak Chap7:UKERC 22/09/200910:19Page157 Box 7.3 AssumedBox 7.3 or implied decline and rate URR, of date peak for the peaking forecasts production within20years. the lossoftwothirdsconventional oil example, a6%/year declinerate implies of itslikely effectsuponsociety. For (Section 3)anddisturbinglyhighinterms inconsistent withthecurrentevidence the requireddeclinerate appearsboth if thisissettomoreconservative levels URR (e.g.3000-3600Gbfor3%/year), but more optimistic assumptions for the global important role. Lower decline rates imply aggregate decline rate also plays an or impliedURR,butthepost-peak of forecasts differ largely in their assumed This analysisindicatesthatthetwogroups • • • • • • • decline whichfinallylevels offtojust under2%/year no effectonthedateof thepeak.Theirforecastindicatesaninitial rapid aggregate or an aggregate post-peak production decline rate and assume that the YTF will ha Uppsala University forecast a peak between 2008 and 2018. They do not state a URR Alternatively, the aggregate decline rate may steepen after 2030. a URRofatleast4500Gbwhichmay beconsistentwithTotal’s assumptions. not stated,wecalculatethepost-peak aggregatedeclinetobe0.2%/year Thisimplies T a peakin2020.Theimplieddeclinerate mustthereforebeabout2.5%/y The BGR forecast has very little information except a URR of just under 3000 Gb and However, Miller questions whether all of the fallow fields will be developed. the potential excess before 2018 is likely to be deferred, leading to a later peak. and thepeakisaround2018.Becausethismodelsmaximumpossibleproduction, production that can be achiev Miller’s bottom-upfield-by–field modelisunique inestimatingtheabsolutemaximum Energyfiles forecasts a peak in 2017 with a URR of 2685 Gb (Box 7.2). because those which will come on-stream within this period must already be committed. existing fields.Near-term supplyistightlyconstrained bytheleadtimeofmajorprojects, supply ofoilfromidentifiednewprojectsisoffsetbythedeclineinproduction The Peak Oil Consulting model focuses on near-term production up to 2016, where the except forTotal andisonereasonfortheearlypeak. aggregate declinerate isabout2%/year, whichisthelowestofallforecasts Campbell forecasts a peak in 2008 with an all-oil URR of 2450 Gb. The post-peak otal forecastsapeakofall-oilat2020.Although thedeclinerate and theURRare ed, regardless of cost or demand. The URR is 2800 Gb much less likely in the near-term. consensus ontheglobalURRappears others (Section 3). Unfortunately, a the IEA (2008), Höök, scale, building upon the recent studies by understanding of decline rates at every We anticipate an improvement in the data which constrain these parameters. should thereforebegiven toobtainingthe of these assumptions. Some priority decline rate and to assess the plausibility or implied global URR and aggregate supply forecastsistoidentifytheassumed Hence, one way of comparing global et al . (2009b) and ear ve 157

Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Chap7:UKERC 22/09/200910:19Page158 158 Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production rate is much larger than experienced in the US and other oil-producing regions. (2000) estimateofremainingrecoverable resources(i.e.URRminuscumulative production).Asaresult,the assumeddepletion since theUSdepletionrate ismeasuredwithrespect to proved reserves whiletheEIAdepletionrate isappliedtotheUSGS rate isequaltothedepletionrate (Section3.4).Whiletheyjustifythe10%figure withreferencetoUSexperience,thisisi 116 Wood higher if(asseemslikely) thedepletion rate was lower. Theseassumptionsseemvery optimistic. need tobeaddedeachyear tomaintainanaggregatedeclinerate of2.5%/year. Therate ofreserve additionswouldneedtobe growth and new discoveries. If these resources were to be produced at an average depletion rate of 5%/year, then ~41 Gb would 115 Thedifferenceofsome6%(andrising)globalproduction in 2030,or2.1Gb/year, wouldhave tobefoundfromreserve Box 7.4 Assumed or implied URR, decline rate and date of peak for the quasi-linear forecasts quasi-linear the for peak of date and rate decline URR, Assumed implied or Box 7.4 • • • • • • global URR(Wood, etal., 2003). an aggregate decline rate of 10%/year and endorsed the USGS estimates of the aggregate declinerate. Howev The USEIAforecastdoesnotquantifytheconventional oilURRandpost-peak decline rate of 6%/year. centred on3600Gbandextendingbetweenapeakyear of2030andamaximum be higher or the URR larger. We show this forecast as a narrow, slanted ellipse, 8.5%/year by 2030. reconcile with the IEA’s estimate of a global average managed field decline rate of which is less than the decline rate of the super-giant fields and seems difficult to of 3577Gb. Apeakdateat~2030requiresanaggregatedeclinerate of<2.5%/y The IEA forecast reaches a plateau by 2030 of un-stated duration and assumes a URR is between3.5and4%/y estimates thatthepeakhasalreadyoccurred.Theaggregatepost-peak decliner The LBST forecast uses the smallest explicit value of the URR (1840 Gb), and Meling’s forecast has a peak in 2028 and a URR of about 3150 Gb. 5%/year, butlower inOPECstateswhichmay dominate future production.We show Gb production ofsome100mb/dconventional oilin2030.TheURRis statedas3345 liquids. IfweestimateOPECNGLproductionat 7mb/dby2030,theforecastimplies It isnotpossibletoaccurately decomposetheOPECforecastinto itscomponent peak year of 2030 and a maximum decline rate of 8%/year. forecast asnarrow, slantedellipse,centredon3600Gbandextendingbetweena of non-conventional fuelsbecomingmorecompetitive. We thereforeshowthis high pricescenarioconventional supplyhaspassedpeakby2030,partlyasaresult conventional production reaches 102.8 mb/d in 2030 and no peak is forecast. In the rate andameanestimateof~3600GbfortheURR.InEIA’s referencescenario, widespread introductionofelectricvehicles. uniquely forecastadecreasingdemandforconventional oil,byassuminga Blueprints scenario, andaround91mb/din2020theScramble scenario. Shell Shell suggests that the peak for all-oil production is around 94 mb/d in 2030 in the , andOPECestimatestheglobalaggregate productiondeclinerate tobe4- et al . assumeproductiondeclinesexponentially atadepletionrate of10%/year. Withexponentialdecline,thedecline 115 But if the peak were delayed, the decline rate would need to ear er 116 , anarticlepublishedbytheUSEIAin2003assumed Here weuse8%astheupperlimitfordecline ear nvalid ate Chap7:UKERC 22/09/200910:19Page159 Box 7.5 Different ways Different Box 7.5 of estimating the global aggregate decline rate discoveries and reserve growth.Building the historicaltrends inbothnew conventional oil and thedateofpeakgiven realistic istheassumedorimpliedURRfor plausibility oftheforecastsistoaskhow An alternative approachtojudgingthe peak and depletiononthedateof discovery, growth reserves Theimpactsofrates 7.6 • • • • • • of howproductionrates willev Some modelsestimatethedepletionrate ofaregionandusethistosupportforecasts data on their natural decline rates at all. controlled upanddowntofulfilOPECquotas,sothattheremay bevery littlereliable of thesefields.Theproductionfrommostthemhasalsobeendeliber of super-giant fields. Unfortunately there is almost no public data on the beha Some authorsarguethattheglobalfielddeclinerate willconverge onthedeclinerate the aggregatedeclineinproductionisanoutput,notinput. of post-peak fields in a region to all pre-peak and y Some fieldbymodelssuchasEnergyfilesextrapolate thehistoricaldeclinerates rate mustbe. peak willcomeafteraspecifieddate.Thisthendefineswhattheminimumdecline Some quasi-linearmodelsappeartouseanestimateforURR,andassertthatany fields. This isthenoffsetbymodellingthedevelopment offallowfieldsanddiscovery ofnew their expectations of EOR to give an aggregate global decline rate from existing fields. OPEC and the IEA measure or estimate actual field decline rates and adjust these for unconstrained. We thereforeomititfromthediagram. the peakyear andURR,thelocationofthisforecastonFigure7.7isrelatively and there is no consideration of post-peak decline. In the absence of estimates for This includes some 105 mb/d of conventional oil. No other robust data are quoted, ExxonMobil’s forecast reaches 116 mb/d by 2030, the highest of all those reviewed. an aggregate4-5%declinerate. this forecastasanarrow, slantedellipse,centredon3345Gbandextendingbetween olv e and decline as the rate of discovery declines. investigate this rulebyestimatingthe estimates from the USGS (2000), we can has been produced. Using the URR in aregionpeakswhen50%oftheURR established ruleofthumbthatproduction One approachistousethelong- ‘Mid-point’peaking 7.6.1 of looking at this issue. Section 6.4, we present some simple ways upon the earlier analysis in Section 3 and et -to-find fields. In these models, viour ately 159

Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Chap7:UKERC 22/09/200910:19Page160 160 Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production of technology and expenditure, but to our knowledge, no such study has been attempted at the global level. growth would increase to 38 Gb/year. A better approach would be to assess the recovery potential of individual fields as a func compared to the 2771 Gb assumed by the IEA). If an average of 50% recovery can be achieved within 30 years, annual reserve leads toanestimatedannualreserves growth fromdiscovered fieldsof22Gb/year (andURR fromthesefieldsof3770Gb oil-in-place (OOIP)fordiscovered fieldsof7540Gb. Assumingthatanaverage of50%recovery canbe achieved in 50years 2369 Gb, withthe standard (althoughuncertain)estimateforglobalaverage recovery factorof35%gives anestimatedoriginal- 120 Forecasts ofreserve growthmay alsobemadewithreferencetorecovery factors.Combiningcumulative 2Pdiscoveries of 119 AsdescribedinSection2,thiscouldbeanunderestimatesince itdoesnottake intoaccountfuturereserve growth. discovery estimateswerenotavailable. 118 Cumulative productionand2Pdiscovery estimatesweretaken fromtheIHSdatabase. TheUSwas excluded since2P other thanphysical depletion. Russia. Itisimportanttonotethat timingofthepeakproductionformany ofthesecountriesmay beinfluencedbyfactors estimated URRexcluding reserve growth.Post-peak countriesforwhichURR estimates werenotavailable wereexcluded, aswas 117 SincetheUSGSonlyestimatereserve growthatthegloballevel, thiswas allocatedbetweencountriesinproportiontothei empirical ruleproposedbyPFCEnergy An alternative approachistoapplythe The‘PFCrule’ 7.6.2 note to the quasi-linear forecasts. simple calculationsoundsacautionary the mid-pointrulemay beoptimistic,this Since historicalexperiencesuggeststhat production growth of around 1.3%/year. estimate ofURRandmostassume forecasts assumeorimplyalargermean the possible exception of Total, none of the rapid growthwouldbringitforward. With point whilesmallerresourcesand/ormore production growth would delay the mid- Larger resourcesand/oraslowerrate of peak is 2027. then thedateof‘mid-point’ assume productiongrowthof1.3%/year, mean URR estimate of 3577 Gb and 80.7 mb/d. If we take the IEA (2008) Gb andtheannualproductionin2007was production ofconventional oilwas 1128 At end 2007, the global cumulative URRhasbeenproduced. of) well before50%oftheir(USGSestimate their peakofconventional oilproduction countries to date appear to have reached maximum of52%.Inotherwords,most production-weighted mean of 26% and a This gives asimplemeanof22%, have passed their peak of production. depletion at peak for 37 countries that 117 years, approximately 15Gb/year over thelast5 of conventional oilcanbeestimatedat IHS Energydatabase,therate ofdiscovery discoveries and reserve growth. Using the 2P reserve additionsthroughnew we need an estimate of the future rate of To applythePFCruleatgloballevel, produced. cumulative 2Pdiscoveries have been production wellbefore60%ofthe date appear to have reached their peak of 36%. In other words, most countries to 38% and a production-weighted mean of 2007. average of33Gb/year between2000and our analysis in Section 3 suggests an 22 Gb/year between1996and2003,while that reserve growthaddedanaverage of data source, Klett, annual rate of production. Using the same discovery rate of15Gb/year (i.e., no approximately 48%.Assuming anewfield production tocumulative discoveries was In 2007,theglobalratio ofcumulative production. countries thathave passedtheirpeakof cumulative discoveries atpeakfor54 ratio ofcumulative productionto can investigate thisrulebyestimatingthe or lessofcumulative 2Pdiscoveries. We when cumulative productionreaches60% that productionpeaksinmostregions 120 119 or roughly half thecurrent or roughlyhalf 118 This gives a simple mean of et al. (2005) estimate tion r Chap7:UKERC 22/09/200910:19Page161 production to remaining recoverable depletion rates (i.e. the ratio of annual A third approach is to investigate the Depletionrates 7.6.3 note tothequasi-linearforecasts. optimistic, this also sounds a cautionary suggests thatthe‘PFCrule’may be (Section 3) and historical experience over thefuturescopeforreserve growth trend (Figure2.8),therearequestions Since discovery is on a long-term declining annual reserve additionsandalargerURR. requires correspondinglyhigherrates of with higher rates of demand growth a URRofatleast3200Gb. To dothesame demand growth (1%/year). It also implies Gb/year, even for low rates of annual annual 2P reserve additions around 35 peak beyond 2030requiressustaining calculations suggestthattopostponethe reserve growth(Table 7.5).These demand growth and the contribution from rule usingdifferentassumptionsfor production canbeestimatedfromthePFC further decline) the expected date of peak cumulative discoveries Applying theTable PFC 7.5 60% rule to forecasts of global cumulative production and (%/year) Demand growth 2.0 2.0 2.0 1.5 1.5 1.5 1.0 1.0 1.0 discoveries year a e /y b G ( New 15 15 15 15 15 15 15 15 15 year) r a e /y b G ( Reserve growth 30 20 30 20 30 20 0 0 0 these resources by 2030, which far depletion rate of10%(andrising)from et al cumulative production of 46 Gb (Aleklett, fields producing19mb/dby2030with 2030 (5 Gb/year) and forecasts these projects 114 Gb of new discoveries by different categoriesoffield.TheIEA allow depletion rates to be estimated for forecast thatprovides sufficientdetailto The IEA(2008)istheonlyquasi-linear Jakobsson, (Aleklett, depletion occurringinoffshorebasins been lessthan5%,withthemostrapid maximum depletion rates have typically peak of production. At the regional level depletion rate typically occurringnearthe fields and regions, with the maximum constraints ontherate ofextraction ofoil reflect physical, technicalandeconomic described in Section 3, depletion rates consistent withhistoricalexperience.As forecasts and to assess whether these are resources) impliedbythedifferent ., 2009). This implies an aggregate Date of peakDate production et al et al., 2009). 2034 2027 2020 2039 2029 2020 2044 2032 2021 ., 2009; Höök, 2009; discoveries at discoveries Cumulative peak (Gb) peak 3611 3089 2577 3841 3161 2577 4071 3269 2593 161

Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Chap7:UKERC 22/09/200910:19Page162 162 Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production rates implied(Dées, incentive todepletetheseresourcesatthe whether OPECproducerswouldhave the physical constraints, itisquestionable forecasts. likely toapplytheotherquasi-linear implication, the same conclusions are discovery and reserve growth. By earlier peakorrequirehigherrates of depletion rates would either imply an 2004). But the use of more realistic demonstration ofsuchconstraints. extraction fromoilreservoirs”. But the standardfieldproduction profilesestimatedbytheIEA(2008)provide aclear 121 Aleklett, caution to the quasi-linear results. before 2030provides a furthernoteof these modelsforecastapeakoccurring explored. Nevertheless, thefactthatallof those assumptions more carefully more transparent andthesensitivityto specific assumptionsneedtobemade bottom-up models do this, although their from improved technology. Several ofthe discovery rates andlikely recovery gains realistic assumptions of production cycles, existing andyet-to-find fieldsunder model theproductionexpectedfrom caution. Abetterapproachmay be to they should be used with considerable Since alltheserulesarevery approximate Bottom-upmodelling 7.6.4 rising) by2030. higher depletionrates of12-13%(and analysis forfallowfieldssuggestseven oil-producing region.Acomparable exceeds the historical experience of any et al . (2009) argue that “…. the IEA do not seem to be aware that there are physical constraints on the rate of 121 et al In additiontothe ., 2007;Gately, non-conventional liquids,theuncertainties While theabove analysislargelyexcludes on plausible oil supply futures. available, leadingtoagreaterconsensus continue as better constraints become We anticipatethatthisconvergence will conventional oilproductionaround2030. models nowforeseea levelling off of parameters. Even some of the quasi-linear secondary or are components of these two other differences are arguably either aggregate production decline rate. The for theURRand/orpost-peak global different explicitorimplicitassumptions the final peak can be closely linked to approaches is wide, the range of dates for reviewed. Althoughtherange ofmodelling convergence appearinginthe forecasts impressions, thereisadegreeof Nevertheless, contrary to initial rather thantherule. assumptions itremainstheexception apart from high and low oil price over many oftherelevant variables, but essential given thedegreeofuncertainty Comprehensive sensitivitytestingis plausible socioeconomicscenarios. been employed toexploreawiderrange of ‘single-value’ forecastsandnonehave Shell, allofthemodelsareusedfor are used. With the notable exception of range of methods and assumptions that coverage of different liquids and the wide inconsistencies in the definition and transparency ofseveral ofthemodels, straightforward owing to the lack of The comparison of forecasts is far from 7.7 Discussion Chap7:UKERC 22/09/200910:19Page163 • major offshore fields, the delay can be significantly longer. 122 Thedelay between amajornewdiscovery beenannounced andthefirstproductionflowscurrentlyaverages 6.5years. For on majornewprojects. appears fairly robust given the lead times of groups such as Peak Oil Consulting For short-term forecasting,theapproach study. lead timesdeserve urgentandcareful conventional liquids and the associated both the economic potential of non- substitution that is likely to be required, fuels. Given thespeedandscaleof also theavailability andcostofsubstitute timing ofthepeakconventional oilbut for policy-makers concernnotonly the • all) ofthefollowingconditions: require some combination (although not until after2030.Thesewouldappearto that delay thepeakofconventional oil judgment of the plausibility of forecasts Nevertheless, itispossibletoformsome peak production unwarranted. making precise forecasts of the date of and ‘above-ground’ uncertaintiesmultiply, number andscaleofboth‘below-ground’ For mediumtolong-termforecasting,the investment rather thanphysical depletion. largely betheresultofdelayed upstream spikes (IEA,2009a).However, thesewill supply may welloccurleadingtoprice next twotofouryears, constraints onoil recession. If demand recovers over the account the impact of the global economic of the forecasts reviewed take into estimate of the USGS (~3600 Gb); and possibly greater than the mean a global URR that is at least 3000 Gb currently assumed(e.g.1%/y lower rates ofdemandgrowththanare 122 However, none ear); • • • • • • • following the peak (e.g. 3%/y fairly rapid declineinproduction regions. so oninallthemajoroil-producing prospective areas, political stability and in including appropriateincentives for favourable ‘above-ground’ conditions, in any oil-producing region; and the maximumrate previouslyachieved av depletion oftheseresourcesatan have lesspotentialforreserve growth); newer, smaller and offshore fields that decade (despite the growing share of exceeds that achieved over the last the periodto2030thatequalsor an annualrate ofreserve growth over of newly discovered fields); last 40years, despitethedecliningsize decade (i.e.reversing thetrendof exceeds that achieved over the last the period the 2030 that equals or an annualrate ofnewdiscoveries over post-peak regions); previously observed in the majority of 2P discoveries (i.e. much greater than peak that exceeds 60% of cumulativ cumulative production at the date of regions); observ URR (i.e.muchgreaterthanpreviously peak that exceeds 50% of the global cumulativ more); v er estment, sufficientaccessto age rate that is much higher than ed inthemajorityofpost e production at the date of ear or -peak e 163

Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Chap7:UKERC 22/09/200910:19Page164 164 Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production tripled, but then increased more than between 1936and1970 asproduction costs remainedsteady ordeclined an instructive precedenthere:extraction 2008; Reynolds, 1999).TheUSprovides enter rapid decline(Kaufmann andShiers, small numberoflargefieldsthatmay soon dependence ofcurrentproductionona also deserves tobequestioned,given the oil market tosignalanimpending peak industrialising countries.Theabilityofthe demand in China, India and other term, given theanticipatedgrowthinoil have asignificantimpactinthemedium- reasons, climate policy seems unlikely to be morethanafewyears. For similar relatively soon thedifferenceisunlikely to supply constraints, but if demand recovers point atwhichphysical depletionleadsto current economic recession may delay the a sharp peak into a ‘bumpy plateau’. The the likely consequence of which is to turn interactions betweensupplyanddemand, models do not capture the complex the rate that is required. Also, most of the these resourcesfrombeingdeveloped at below ground constraints could prevent pessimistic. However, amixofabove and some ofthe‘peaking’forecastsoverly reasonable andhencetheassumptionsof estimates for the global URR may be difficult. Section6concludedthatlarger peak productionintheinterimismore Forming a judgement on the timing of production before2030appearslikely. conclusion that analysis in Section 6 leads us to the combination. This,togetherwiththe appear optimistic and more so in judgment, most of these conditions met or why they do not apply. In our demonstrate howsuch conditionscanbe forecasts need to either ‘Quasi-linear’ a peakofconventionaloil • • follows: The mainconclusionsofthissectionareas 7.8 Summary serious consideration to this risk. many countriesthatarefailingtogive As noted in Section 1, the UK is one of conventional oilproductionbefore2020. there isasignificantriskofpeakin Given thesecomplexities,wesuggestthat likely to be unprepared. occurs at the global level, the world is (Cleveland,pattern 1991).Ifasimilar fourfold withinadecadeafterthepeak forecasts reviewedhere restlargely r now becomingapparent. Althoughthe forecasts, adegreeofconv between ‘peaking’and‘non-peaking’ Despite theapparentdivergence socioeconomic scenarios. exploration ofwiderrange ofplausible forecasts toward the broader need tomove beyond single-value models ofoildemandandthereisa need tobebetterintegrated with presenting results. Models of oil supply consideration ofuncertaintieswhen those assumptionsandbyexplicit exploring the sensitivity of forecasts to key assumptions,bysystematically consensus byrevealing andcomparing is considerable scope for improving methods andassumptionsused.There co the inconsistency in the definition and transparency ofmany ofthemodels, forecasts ishamperedbythelackof Comparison ofglobaloilsupply ange ofapproachesis very wide, the verage ofliquidsandtherange of ergence is Chap7:UKERC 22/09/200910:19Page165 • • the numberandscaleofuncertainties For medium to long-term forecasting, when demandrecovers. supply shortages consequent riskof the economicrecessionand and delay oftheseprojectsasaresult for the short term is the cancellation available publicdata.Theprimaryissue region canbeconstructedusingwidely Reasonable short-term forecastsforany will raise supply are already committed. inflexible, becausetheprojectswhich capacity, toabout2016,isrelativ The short term future of oil production than atpresent. these parameters to a greater degree priority should be given to constraining values fortheseparameters. Hence, results if they used or implied similar that mostmodelswouldgive similar rate ofconventional oil.Itseemslikely post-peak aggregateproductiondecline assumptions fortheglobalURRand/or upon differentexplicitorimplicit ely • given urgent consideration. develop alternatives, thisriskshouldbe supply constraints and the lead times to potentially serious consequences of oil productionbefore2020.Given the significant risk of a peak in conventional balance, wesuggestthatthereisa ‘below’ ground factors involved. On interim, giv on the timing of peak production in the It ismoredifficulttoformajudgment before 2030mustbeconsideredlikely. peak of conventional oil production On currentevidence,wesuggestthata conditions identifiedheredonotapply. assumptions can be met or why the demonstrate either how these implausible. Such forecasts need to that are at best optimistic and at worst 2030 rest upon several assumptions forecasts thatdelay thepeakuntilafter Nevertheless, weconsiderthat date ofpeakproductionunw multiply, making precise forecasts of the en the multiple ‘abo arranted. v e’ and 165

Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Chap7:UKERC 22/09/200910:19Page166 166 Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Chap8:UKERC 22/09/200910:20Page167 .The mechanisms leading to a‘peaking’ 1. follows: The mainconclusionsofthisreportareas 8.1 Conclusions Conclusions,research needs andpolicyimplications 8. • the regional and global level. constraints on future its supply both at understood and provide identifiable of conventional oil production are well • ‘abo complex and interdependent mix of Oil supplyisdeterminedbya constrained oncetheresourceis ‘possible futures’ becomes more appropriate dataand the range of once accessisavailable tothe forecasting becomesmorereliable str forthcoming peak is far from production, anticipatinga factors thataffectconventional oil technical, economic and political Given the complex mix of geological, the world. increasing numberofregionsaround production canbeobserved inan and the peaking of conventional oil early. This process can be modelled and producethesefieldsrelatively fields and the tendency to discover resources in a small number of large individual fields,theconcentration of include theproductionprofileof ultimately decline.These features will risetoapeakorplateauand inevitable thatproductioninaregion conventional oilresourcemake it fundamental featuresofthe over the other. Nevertheless, emphasising one set of variables factors and little is to be gained by aightforward. However, supply v e-ground’ and‘below -ground’ 2. Despite large uncertainties Despite in the 2. • assessed. global oil depletion to be adequately toavailable allow the status and risk of sufficient information is data, available • databases a insufficiently appreciated.The depletion and their limitations are poorly suited to studying oil Publicly available data sources are improved understanding ofthe depletion would benefit from countries. Hence,the debate onoil of thosereserves bykey producing potentially offset any overestimation regional andgloballevel whichcould underestimation ofreserves atthe Doing somay leadtoan fields toobtain aregionaltotal. of ‘proved’ reserves fromdifferentoil incorrect tosimplyaddtheestimates F standardisation inreserve reporting. slow progresstowards by errorsininterpretationandthe Data uncertaintiesarecompounded reserves. hold the majority of the world's including key OPEC members, that being lowestforthosecountries, proportional totheirimportance– level of confidence is inversely must rely upon assumptions whose reserve estimates, supplyforecasts regions. Intheabsenceofaudited not necessarilyreliableforall are alsoexpensive, confidentialand sources arebetterinthisregard,but level. increasingly the case at the global substantially depleted.Thisis or example,itisstatistically v ailable fromcommercial 167

Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Chap8:UKERC 22/09/200910:20Page168 168 Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production .There ispotential for improving 3. • growth’. ‘reserves of magnitude and source standing controversies such asthe long- and onimportant consensus • resources between different siz The distributionofconventional oil commonly observed togrowo resources ofindividualfieldsare Estimates of the recoverable future supply. therefore ofcriticalimportancefor the potential for reserve growth are their future production profile and remaining reserves atthesefields, fields are expected to be found. The next decadeorsoandfewnewgiant rest willbegintodeclinewithinthe their peakofproduction,mostthe relatively old, many are well past fields are Most of these ‘giant’ thirds of cumulative discoveries. and up to 500 fields account for two fields accountforhalfofproduction global productionofcrudeoil,100 fields accountforonequarterofthe fields intheworld,approximately 25 Although therearearound70,000oil field is increasingly well understood. available data. nature andlimitationsofthe it seems likely to continue to do so in than thediscovery ofnewfieldsand global reserves over the pastdecade appears tohave addedmoreto recoverable reserves. This process conservative estimates of conditions andrevisionstoinitially technology, changesineconomic geological knowledge,better time as a result of improved es of ver • • • more rapid post-peak decline. As a changing productionmethods leadto y production shiftstowards smaller as moregiantfieldsenterdecline, Decline rates areonanupward trend coming onstreamevery threeyears. equivalent toanewSaudiArabia production atcurrentlevels - each year, simply tomaintain mb/d of new capacity must be added This impliesthatapproximately 3 producing fieldsisatleast4%/year. rate of decline from all currently- globally, whilethecorresponding production isatleast6.5%/year that are past their peak of a production fromexistingfields.The inv The oilindustrymustcontinually key areas of uncertainty. at whichtheymay beappliedremain sizes andtypesoffieldtherate of thesetechniquesfordifferent recovery techniques.Thesuitability more widespread use of enhanced oil higher oil prices may stimulate the absolute terms. At the same time, decrease inbothpercentageand fields the rate of reserve growth may tow so asglobalproductionshifts for larger, olderandonshore fields, Reserve growthtendstobegreater conservative reporting. be primarily the result of ‘reserve growth’doesnotappearto between differentfieldsandregions, different factors varies widely the future.Whilecontributionof ounger andoffshorefields andas verage rate ofdeclinefromfields est toreplacethedeclinein ards newer, smallerandoffshore , Chap8:UKERC 22/09/200910:20Page169 4. Methods for estimating resource size resource estimating for Methods 4. • acknowledged. important limitations thatneed to be and forecasting future supply have • confidence whenexplor estimated toareasonabledegreeof much as geology and can only be economic andtechnicalfactorsas (URR) of a region depend upon The ultimatelyrecoverable resources extremely challenging. falling. At best,thisislikely toprove simply to prevent production from may needtobereplacedby2030, current crudeoilproductioncapacity result, more than two thirds of modelling’. But they are best applied methods such as ‘discovery process common with more sophisticated not available andalsohave muchin role to play when field-level data is estimating URR have an important simple ‘curve-fitting’ techniques for adv careful justification. experience are likely to require departures from this historical assume or imply significant produced. Supplyforecaststhat recoverable resourceshave been peak well before half of their and mostregionshave reachedtheir the remaining recoverable resources a region has rarely exceeded 5% of example, theannualproductionfrom pattern ofproductionover time.For depletion of a field or region and the constraints upon both the rate of ph arbitrarily high rates. There are Oil reserves cannot be produced at ysical, engineering and economic anced. Although widely criticised, ation iswell • • • production may beestimated to plateau) inconventional oil The timing of a global peak (or rather than the rule. uncertainties remain the exception testing and the presentation of for multipleassumptions.Sensitivity economic variables andrequirement lack oftransparency, neglectof their relianceonproprietarydatasets, existing models are hampered by medium-term forecasts,butmany fairly reliablebasisfornearto using fieldorprojectdataprovide a circumstances. Bottom-upmodels should befavoured inall weaknesses andnosingleapproach approach has its strengths and aggregation andcomplexity. Each different variables andtheirlevel of theoretical basis, their inclusion of supply vary widely in terms of their Methods offorecastingfutureoil future supply. excessively pessimistic forecasts of global URR and has contributed to underestimates ofregionaland discovery. This has led to future cycles of production or growth andtheinabilitytoanticipate frequent neglectoffuturereserv choice offunctionalform,the sensitivity of the estimates to the fitting techniques,suchasthe attention to the limitations of curve- Many analystshave paidinsufficient criteria, errors are likely to result. many regionsdonotmeetthese consistent exploration history. Since homogeneous areas with a to well-explored and geologically e 169

Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Chap8:UKERC 22/09/200910:20Page170 170 Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production .Large resources of conventional oil may 5. • global peak. differencelittle to the timing of the be accessed quickly and may make but these areunlikely to be available, of con Although estimatesoftheglobalURR in some of the most promising areas. continuing restrictions on exploration underestimated and there are these discoveries may have been implied bytheUSGS, the size of of newdiscoveries islowerthan assumptions andalthoughtherate appears tobematchingtheUSGS premature. Globalreserve growth are ‘discredited’atbest assertions thattheUSGSestimates despite their apparent optimism, in global supply forecasts. But leads toacorrespondinguncertainty 2007 of1,128Gb. Thiswiderange cumulative production throughto barrels (Gb), compared to within therange 2,000-4,300billion Contemporary estimatesnowfall departure from the historical trend. (USGS) represent a substantial from the US Geological Survey years, the most recentestimates trending upwards for the last 50 diminishing returns. problem andissubjecttorapidly complexity does little to address this precision. Increasingmodel can provide estimatesofgreat technological disruptions, no model for political,economic,or oil market. Butgiven thepotential and nosignificantdisruptionstothe particular value fortheglobalURR within decadal accuracy assuming a v entional oilhave been • • lik accessible locations.If(asseems located insmallerfieldsless appear plausible, much of this is of the global URR of conventional oil Although moreoptimisticestimates plausible global supply forecasts. considerations constrain the range of production followingthepeak.These and/or arelatively steepdeclinein demand growthpriortothepeak combined withaslowrate of size oftherecoverable resource optimistic assumptionsaboutthe peak beyond 2030 requires approximately 24Gb).Delaying the production fromtheUKis (for comparison, the cumulative peak productionbyonlyafewdays billion barrels delays the date of increasing the global URR by one substitute fuels. In this model, in terms of the lead time to develop may be a relatively narrow window of forecastsanimminentpeak,it this range appears wide in the light between 2009and2031.Although production can be estimated to lie production cycle, thedateofpeak oil andtheshapeoffuture about theglobalURRofconventional For a wide range of assumptions about the size of the global resource. relativ conventional oil production is The timingoftheglobalpeakfor companies that control much of constrained ifthenationaloil stage. Demand growth may also be demand growth at a relatively early cost, supply constraints may inhibit produced relatively slowly at high ely) these resources can only be ely insensitive to assumptions Chap8:UKERC 22/09/200910:20Page171 6. The risks presented by global oil global by presented risks The 6. • communities. attention the by research and policy much serious more deserve depletion • • assess ortoeffectiv oil supply security and fails to either the economic and political threats to Much existingresearchfocusesupon optimistic andatworst implausible. assumptions that are at best after 2030restuponseveral conventional oil production until forecasts that delay the peak of Nev adequately assessed. different outcomes has not been probability andconsequencesof depletion. Thishasmeantthatthe is relatively inflexible, because the production capacity, toabout2016, The short term future of oil risks presented by ph ability to invest. these resourceslacktheincentive or peak productionunwarr precise forecasts of the timing of uncertainties multiplymaking forecasting, thenumberandscaleof For medium to long-term when demandrecovers. consequent riskofsupplyshortages economic recession and the projects asaresultofthe2008 cancellation anddelay ofthese issue for the short term is the available publicdata.Theprimary can be constructed using widely short-term forecastsforany region already committed. Reasonable projects whichwillraise supplyare ertheless, weconsiderthat ely integrate the anted. ysical .Key parameters thatinfluence global oil 1. future research. following highlightssomeprioritiesfor knowledge in a number of areas. The there isconsiderable scopeforimproving environmental consequences. Hence, the potential economic, social and near-term peakinglobaloilproductionor research examiningeithertherisksofa There is remarkably little contemporary Researchneeds 8.2 • ofa greater level confidence to estimated be should and can supply con Estimates oftheglobalURR serious consideration. efficiency, this risk needs to be given substitute fuelsandimprove energy times requiredtobothdevelop a peakbefore2020.Given thelead likely andthereisasignificantriskof oil production before 2030 appears suggest thatapeakofconventional On thebasisofcurrentevidencewe identified inthisreportdonotapply. can bemetorwhy theconstraints demonstrate howthese assumptions Such forecasts need to either reserve growth. global trendsinnewdiscoveries and estimates inthelightofregionaland key regionsandtoevaluate these need to update the assessments for unlikely in the near term, there is a revised globalassessment may be increasingly outofdate.Whilea authoritative estimatesare confidence boundsandthemost v entional oil haveentional oil very wide 171

Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Chap8:UKERC 29/09/2009 10:00 Page 172 Page 10:00 29/09/2009 Chap8:UKERC 172 Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production 2. Different approaches to studying oil studying to approaches Different 2. • be more integrated. effectively to need depletion resource and supply • • • assessments shouldemplo Where possible,resource acknowledge the uncertainty in the techniques andsources ofdataand over theperiodto2030. fields may be expected to change average decline rate of post-peak Of particular importance is how the studies require further clarification. within andbetweentheavailable considerably, buttheinconsistencies depletion rates has improved Understanding ofdeclinerates and great value. the leadtimesinvolved, wouldbeof of differenttechnologies,including economic evaluation ofthepotential supply. Adetailedtechnicaland consider at thegiantoilfieldsisof The potentialforenhancedrecovery forecasts. together withtheassociatedsupply projections offuturereserve growth, used to underpin more accurate conditions. Thisinformationcouldbe how theyareinfluencedbyeconomic different sizes andtypesoffield reserve growthvary between possible to examine how patterns of regional studies, it should be combination withindividualfieldand different regions of the world. In experience withreserve growthin detailed examinationofthe sufficient informationtoassistina The industry databases contain able importanceforfuture y multiple • • • a sufficientlytimely fashion and upon to signal global oil depletion in whether oilmarkets canberelied There aregroundsforquestioning for export. impact onthevolume ofoilavailable significant andpotentiallyrapid producing countries could have a and rising internal demand within incentives for upstream investment resource depletion,insufficient For example,acombinationof assessed in an integrated fashion. political dimensionswhichneedtobe interrelated physical, economic and Threats toenergysecurityhave available. assuming thatsupplywillbe as modelling demand and simply and assumeddemandisasunhelpful ‘gap’ betweenthemodelledsupply supply and demand. Forecasting a integrate thedeterminantsofboth determinants ofoilsupplyandto reflect botheconomicandphysical assumptions. Modelsshouldseekto sensitivity of forecasts to those systematically exploring the assumptions of different models and may beimproved bycomparingthe socioeconomic scenarios. Consensus examination of competing forecasts to a more exploratory to mo Projections ofglobaloilsupplyneed economic andpoliticalvariables. techniques to incorporate relevant models that use econometric overcome withtheuseofhybrid limitations ofcurve-fitting may be results obtained.Someofthe ve beyond ‘single value’ Chap8:UKERC 29/09/2009 10:00 Page 173 Page 10:00 29/09/2009 Chap8:UKERC 3. Urgent consideration should be given to given be should consideration Urgent 3. installed capacityofwindpower (assuming a30%loadfactor). least 2%.Thisistheenergyequivalent of100nuclearpowerstationsthesize ofSizewell B(1GW)orthreetimes theglobal 123 Asanillustration, mostoftheforecastsreviewed in Section7projectannualreductionsconventional oilproductionof a • mitigation. for options the and depletion the potential consequences of global oil • con liquid fuelssupplyif‘non- only beassociatedwithapeakin A peak in conventional oil supply will a needformorecarefulassessment for carbon emissions. Hence, there is positive and negative consequences to besmoothandcouldhave both to non-conventional fuelsisunlikely But thetransition fromconventional assumptions about future oil prices. frequently relyuponoptimistic and reference tooildepletion usually conductedwithlittleorno Climate policy assessments are require careful study. and theassociatedleadtimes potential of non-conventional fuels required, the technical and economic of substitutionthatislikely tobe potential. Given thespeedandscale and good reasons to question their impacts of non-conventional fuels availability, costandenvironmental uncertainties regardingthe fashion. Buttherearemajor substitute inasufficientlytimely examination. investment deserves close relation todepletionandnew the behaviour of oil markets in fuels is likely to be induced. Hence, improved efficiencyandsubstitute whether adequateinvestment in ventional’ sources areunableto made. However, threegeneral commentsmay be options is beyond the scope of this report. The evaluation ofdifferentmitigation Policyimplications 8.3 • • will prove challenging owing to both First, it seems likely that mitigation their effects. prevent their occurrenceormitigate measures thatcanbetaken toeither plausibility ofsuchscenariosandthe analysis is required to assess the and climatepolicyscenarios.Further currently consideredwithinenergy experienced, as well as that well beyond thatpreviously steeply (e.g.3-6%/year). Thisis con includes scenarios in which The envelope ofpossiblefutures growth. fuel substitution and economic prices on efficiency improvements, includes the impact of higher oil climate policyandviceversa. This of the impact of oil depletion on in liquidfuelssupplyaretobe before thepeakifseriousshortfalls need tobeinitiatedatleast20years substitution anddemandreduction large-scale programmes of Department ofEnergyarguesthat example, a report for the US the associated lead times. the scaleofinvestment requiredand ventional oilproductionfallsvery 123 For 173

Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Chap8:UKERC 22/09/200910:20Page174 174 Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production • carbon capture and storage. This being potentiallyavoidable through less than half of these emissions tonnes ofcarbondioxide (CO in emissionsofmorethan800billion reserves into liquidfuelswouldresult one third of the world's proved coal non-conv incentives to exploithighcarbon of oil depletion, there will be strong policy willhelptomitigatetheeffects associated with climate change Second, althoughmany measures response. developing an appropriate policy not beadistantdateintermsof constraints. Hence, even 2030 may and ignoresenvironmental post-peak decline rate (2%/year) also assumesarelatively modest public transport, electric vehicles) it important mitigation options (e.g. this reportoverlooks many avoided (Hirsch, entional fuels. Converting et al ., 2005).While 2 ), with • presented byglobaloildepletion. greater awareness oftherisks improved understandingandmuch politically feasible requires both required. For thistobecome envisaged couldpotentiallybe rapid changethaniscurrently programmes. Butgreaterandmore established withinnationalclimate measures comparable tothosebeing in significant policysupport.This and seemsunlikely tooccurwithout 2 likely globalwarming istobekept to than 1,800billiontonnesifthemost total futureemissionsshouldbeless compares torecommendationsthat oil priceuncertaintyandvolatility mitigation efforts will be inhibited by Third, investment inlarge-scale considerable importance. alternatives to conventional oil is of early investment inlow-carbon o vestment canbeencouraged by C (Allen,etal., 2009).Hence, Refs_App:UKERC 23/09/2009 10:32 Page 175 Page 10:32 23/09/2009 Refs_App:UKERC Greenwich JAIPress. discovery process," inJ. Ramsay ed. Attanasi, E. D., L. J. Drew, and D. H. Root. (1981). "Physical variables and the petroleum World Oil, Attanasi, E. D. (2000). "Growth in conventional fields in high cost areas: a case study." exploration programs." Arrington, J. R. (1960). "Predicting the size of crude reserves is key to evaluating flank ofDenver-Julesberg basin" Arps, J. J. and T. J. Roberts. (1958). "Economics of drilling for Cretaceous oil on the east Energy Outlook2008." "The peakoftheoilage:reviewingreferencescenarioworldoutlookinIEAWorld Aleklett, K., M. Höök, K. Jakobsson, M. Lardelli, S. Snowden, and B. Söderbergh. (2009). and Exploratory Effort." Arps, J. J., M. Mortada,andA.E.Smith.(1971)."Relationship BetweenProved Reserves Arps, J. J. (1945)."Analysisofdeclinecurves." BP 04/02. Royal Institute of International Affairs: London. Arnott, R.(2004)."Oilandgasreserves: Communicationwiththefinancialsector." SDP Geological SurveyBulletin Arnold, R.andAnderson.(1908)."PreliminaryreportonCoalingaoildistrict." ASPO Conference Andrews, S. and R. Udall. (2003). "Oil prophets: looking at world oil studies over time." trillionth tonne." Meinshausen. (2009). "Warming caused by cumulative carbon emissions towards the Allen, M.R., D. J. Frame, C.Huntingford,D. Jones,J. A.Lowe,M.Meinshausen,andN. Geology Review Ahlbrandt, T. S. (2002)."Futurepetroleumenergyresourcesoftheworld." Geology Review Ahlbrandt, T. S. (2002)."Futurepetroleumenergyresourcesoftheworld." availability of petroleum resources." Aguilera, R.F., R.G.Eggert,Lagos,andJ. E.Tilton.(2009)."Depletionandthefuture pessimism." Adelman, M.A.andC.Lynch. (1997)."Fixed viewofresourcelimitscreates undue Adelman, M. A. (1991). "Oil Fallacies." Glottometrics, Adamic, L. A. and B. A. Huberman. (2002). "Zipf, power laws and the Internet." References 221:4, pp. 63-68. Oil and Gas Journal 3 , , , pp. 143-50. Nature 44:12, pp. 1092-104. 44:12, pp. 1092-104. : Paris, France. in submission Oil andGasJournal , Journal of Petroleum Technology 458:7242, pp. 1163-66. , 357. , 95:14, pp. 4. AAPG Bulletin, Energy Journal, . The economicsofexploration forenergyresources Foreign Policy, , 58:9, pp. 130-34. Trans AIME, pp. 160-247. 42:11, pp. 2549-66. 30:1, pp. 141-74. 82, pp. 3-16. , pp. 671–75. International International US . 175

Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Refs_App:UKERC 23/09/2009 10:32 Page 176 Page 10:32 23/09/2009 Refs_App:UKERC 176 Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production of EvidenceforGlobalOil Depletion.UKEnergyResearch Centre:London. Bentley, R.W. (2009)."Aprimeronoil peaking." BackgroundnotefortheUKERCReview pp. 189-205. Bentley, R.W. (2002). "Globaloilandgasdepletion:Anoverview." 60. Society of Petroleum Engineers (SPE): Denver, CO, United States. Beliveau, D. andR.Baker. (2003). "Reserves Growth:Enigma,Expectation orFact?": 851- Wright-Allen Press Inc. and D.H. Meadowseds. Behrens, W. W. (1973)."Thedynamics ofnatural resource utilisation,"inD.L. Meadows Mathematical Geology, "Workshop on the Delphi method." Baxter, G.G., A.H.C.Cargill,P. E.Hart,G.M.Kaufmann, andF. Urquidi-Barrau. (1978). Science Bass, F. M. (1969). "A new product growth model for consumer durables." Basile, P. S. andA.Papin. (1981)."World oil:Along-termlook." Geology andEarthProcessesSpringer hydrocarbon accumulations," inC.BartonandP.R. LaPointe eds. Barton, C.andH.Scholz.(1995)."Thefractal size andspatialdistributionof style curves." Bartlett, A.(2000)."AnanalysisofUSandworldoilproductionpatternsusingHubbert- Physics, Bartlett, A. A. (1978). "Forgotten fundamentals of the energy crisis." to exploration." Barss, D. L. (1978). "The significance of petroleum resources estimates and their relation Energy: mathematics and models Barouch, E. and G. M. Kaufman. (1975). "Probablistic modelling of oil and gas discovery." Energy Policy, Bardi, U. (2005). "The mineral economy: A model for the shape of oil production curves." costs, andcontracts Babusiaux, D., S. Barreau,andP. R.Bauquis.(2004). Ayres, E. (1953). "U.S. oil outlook: How coal fits in." Bulletin, Attanasi, E. D. and D. H. Root. (1994). "The enigma of oil and gas field growth." wishful thinkingorreality?" Attanasi, E. D., R. F. Mast, and D. H. Root. (1999). "Oil, gas field growth projections: , 46:9, pp. 876-88. 78:3, pp. 321-32. 15, pp. 215-27. Mathematical Geology, 33:1, pp. 53-61. Bull .Can.PetGeol . Paris: Editons TECHNIP. 10:5, pp. 581-87. Towards globalequilibrium:collectedpapers . Cambridge,MA Oil and Gas Journal SIAM InstituteforMathematicsandSociety, pp. 133-52. , 26:2, pp. 275-91. 32:1, pp. 1-17. . Springer. Journal of the International Association for theInternationalAssociation Journal of , 97:14, pp. 3. Coal Age Oil and gas production: Reserves, , 1953:August, pp. 70-73. Energy Fractals inPetroleum Energy Policy American Journalof , 6 :6, pp. 529-41. Management , AAPG 30:3, Refs_App:UKERC 23/09/2009 10:32 Page 177 Page 10:32 23/09/2009 Refs_App:UKERC Campbell, C.J. (1997). Exploration &Exploitation , Campbell, C.J. (1996)."Statusofworldoildepletion attheendof1995." depletion." Campbell, C.J. (1995)."The nextoilpriceshock - theworld’s remainingoilandits Springer. Campbell, C.J. (1991). Petroconsultants S.A.: Geneva. Campbell, C. and J. Laherrère. (1995). "The world's supply of oil 1930-2050." production: anotherlook." Caithamer, P. (2008)."Regression andtimeseriesanalysisoftheworldoilpeak of Technology, and development costs 1973-1982, lower 48 onshore and offshore." Byrd, L.T., R.M.Kumar, A.F. Williams,andD. L.Moore.(1985)."USoilandgasfinding society tests." Brecha, R.J., R.Berney, andB. Craver. (2007)."Revisiting Hafemeister'sscienceand 36:9, pp. 3492-504. Brecha, R.J. (2008)."Emissionscenariosinthefaceoffossil-fuelpeaking." Technology, gas emissionsoftheshellin-situconversion process." Brandt, A.R.(2008)."Converting oilshaletoliquidfuels:energyinputsandgreenhouse Brandt, A. R. (2007). "Testing Hubbert." BP. (2008). "Statistical Review of World Energy 2008 ". BP plc: London. Department of Statistics and Geology, Princeton University: Princeton, NJ. distribution inestimatingcrudeoilresources."Resource EstimationandValidation Project, Bloomfield, P., M.Hudson,K.Kim,andG.S. Watsion. (1979)."Theuseofthelognormal Enterprise andRegulatory Reform: London. BERR. (2008). "Response to FoI request, Ref 08/0091." Department of Business, oil peak:theneedtouse2Preserves." Bentley, R. W., S. A. Mannan, and S. J. Wheeler. (2007). "Assessing the date of the global Environment and Planning B: Planning and Design Bentley, R.W. andG.Boyle. (2008)."Globaloilproduction:forecastsandmethodologies." 18:2-3, pp. 147-206. Whitfield. (2000)."Perspectives onthefutureofoil." Bentley, R.W., R.H.Booth,J. D. Burton,M.L.Coleman,B. W. Sellwood,andG.R. Energy Exploration & Exploitation 31, pp. 2040-48. 42:19, pp. 7489-95. American JournalofPhysics The ComingOilCrisis The goldencenturyofoil1950:2050:thedepletion ofaresource: Journal of Mathematical Geosciences 14:1, pp. 63-81. Energy Policy, Energy Policy, 35:5, pp. 3074-88. , 75:10,pp. 916-30. : Multi-sciencepublishing &Petroconsultants. , 13:1, pp. 19-46. , 34, pp. 609-26. Energy Exploration&Exploitation, 35:12, pp. 6364-82. Environmental Scienceand , 40:6, pp. 1874-. Journal of Petroleum Energy Policy Energy , 177

Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Refs_App:UKERC 23/09/2009 10:32 Page 178 Page 10:32 23/09/2009 Refs_App:UKERC 178 Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production undiscovered conventional oil andgasresources." Charpentier, R. L. (2005). "Guiding principles of USGS methodology for assessment of London. CERA. (2008). "Finding the Critical Numbers." Cambridge Energy Research Associates: Journal Cavallo, A.J. (2005b)."Hubbert'smodel:Uses,meanings,andlimits-2." Journal Cavallo, A. J. (2005a). "Hubbert's model: Uses, meanings, and limits - 1." 211-21. implications forworldoilproductionforecasts." Cavallo, A.J. (2004)."Hubbert'spetroleumproductionmodel:anevaluation and ed. Chaudhry, A.U. (2003)."Application ofdeclinecurve analysismethods,"inA.U. Chaudhry Research methodology forassessmentofundiscovered oilandgasresources." Charpentier, R.R., G.L.Dolton,andF. Ulmishek. (1995)."Annotatedbibliography of else?" USGS Open File Report 03-137. US Geological Survey. Charpentier, R.(2003)."Thefutureofpetroleum:optimism,pessimism,orsomething 175-86. Research (International Association for Mathematical Geology) Cavallo, A.J. (2002)."Predictingthepeakinworldoilproduction." pp. 421-28. function and its derivatives." Carlson, W. B. (2007). "Analysisofworldoilproductionbasedonthefittinglogistic Technology, Capen, E.(1976)."Thedifficultyofassessinguncertainty." Jeremy Mills Publishing. Campbell, C.J. (2004)."ASPOOilDepletionModel." 193-207. Campbell, C. J. and S. Heapes. (2008). Campbell, C. J. (2005). Campbell, C.J. (2002b)."Petroleum andPeople." Exploration &Exploitation, Campbell, C.J. (2002a)."Theassessmentandimportanceofoildepletion." Gas WellTestingHandbook , , 103:22, pp. 20-25. 103:21, pp. 22-24. , 4:2,pp. 154-86. 28, pp. 843-50. Oil Crisis :6-1, pp. 407-35. 20-1 Energy Sources Part B: Economics, Planning and Policy . Burlington Gulf Professional Publishing, pp. 637-63. . Brentwood: Multi-Science Publishing. An atlas of oil and gas depletion Natural ResourcesResearch Population andEnvironment Natural ResourcesResearch Association for the Study of Peak Oil , 11:3, pp. 187-95. Journal ofPetroleum Natural Resources Natural Resources . Huddersfield: Oil andGas Oil andGas , , , 13:4, pp. 24:2, pp. 14:3, pp. Energy , 2 :4, . Refs_App:UKERC 23/09/2009 10:32 Page 179 Page 10:32 23/09/2009 Refs_App:UKERC petroleum provinces." Davies, J. C. and T. Chang. (1989). "Estimating potential for small fields in mature models ofUSoilandgassupply." Dahl, C. and T. E. Duggan. (1998). "Survey of price elasticities from economic exploration curves." Bulletin 228. Department of the Interior. Cutler, W. C. (1924). "Estimation of underground oil reserves by oil-well production System DynamicsReview for theUnitedStateswithendogenoustechnology, exploration, recovery, anddemand." Davidsen, P. I., J. D. Sterman, andG.P. Richardson.(1990)."Apetroleumlifecycle model 69. Petroleum Technology, Cronquist, C.(1991)."Reserves andprobabilities. Synergism oranachronism?" Processes Cramer Barton,C.andP. R.LaPointe. (1995). Review of Energy and the Environment, retrospective examinationoflong-termenergyforecastsfortheUnitedStates." Craig, P. P., A. Gadgil, and J. G. Koomey. (2002). "What can history teach us? A energy intheeconomy." Cleveland, C. J., R. K. Kaufmann, and D. I. Stern. (2000). "Aggregation and the role of technology stretchtheresourcebase." Cleveland, C. J. andR.K.Kaufmann. (1997)."Natural gasintheUS:howfarcan Energy Journal, rate of production: incorporating economic forces into the models of M. King Hubbert." Cleveland, C. J. and R. K. Kaufmann. (1991). "Forecasting ultimate oil recovery and its States." Cleveland, C.J. (2005)."NetenergyfromtheextractionUnited the ofoilandgasin fuels intheU.S." Cleveland, C. J. (1992a). "Energy surplus and energy quality in the extraction of fossil 163-88. oil supply in the lower 48 United States, 1936-1988." Cleveland, C.J. (1991)."Physical andeconomicaspectsofresource quality:thecostof Dallas, Texas, USA: Dallas,Texas, USA. Residual Plots." Chen, Z. (1991). "A Detecting Technique for Production Rate Decline-Curve Analysis with 48 united-states,1946-1989." Cleveland, C.J. (1992b)."Yieldpereffortforadditionstocrude-oilreserves inthelower Energy Springer. , 12:2, pp. 17-46. Society ofPetroleumEngineers(SPE)eLibraryPaper 30:5, pp. 769-82. Ecological Economics AAPG Bulletin, 43:10, pp. 1258-64. Ecological Economics , 6 :1, pp. 66-93. AAPG Bulletin, Journal ofEnergyFinance&Development 73:8, pp. 967-76. , The Energy Journal, 6 27, pp. 83-118. :2, pp. 139-62. , 76:6, pp. 948-58. 32, pp. 301-17. Fractals inPetroleumGeologyandEarth Resources andEnergy 18:2, pp. 89-108. , Vol. No. 22313. , 3 :2, pp. 129- , Journal of 13:2, pp. Annual The 179

Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Refs_App:UKERC 23/09/2009 10:32 Page 180 Page 10:32 23/09/2009 Refs_App:UKERC 180 Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Deffeyes, K.S. (2005). 91. oil market: assessmentofaquarterlyeconometricmodel." Dées, S., P. Karadeloglou, R.K.Kaufmann, andM.Sánchez.(2007)."Modellingtheworld U.S." Davis, W. (1958)."Astudyofthefutureproductive capacityandprobablereserves ofthe the USA." Drew, L. J. and J. C. Griffiths. (1965). "Size, shape and arrangement of some oil fields in controversy Drew, L.J. (1997). Williston basin."AAPG Bulletin, Dow, W. G. (1972). "Application of oil correlation and source rock data to exploration in Hill and Wang, New York, NY, United States (USA). of natural gas." Erickson, E. W. and R. M. Spann. (1971). "Supply response in a regulated industry: the case Theory and an Application. Final Report." 63p. United States. Epple, D. andL.P. Hansen.(1981)."Econometrics ofExhaustibleResource Supply:A EIA. (2008b). "International Petroleum Monthly." Energy Information Admin. Information Administration: Washington, DC, USA. EIA. (2008a)."InternationalEnergyOutlook 2008"DOE/EIA-0484(2008). Energy Information Administration. EIA. (2007)."AnnualEnergyReview 2007."Report DOE/EIA-0384 (2007).Energy Information Administration: Washington DC. EIA. (2006)."DefinitionsofPetroleum ProductsandOtherTerms." U.S. Energy information Administration. EIA. (1990). "US oil and gas reserves by year of field discovery." DOE/EIA-0534. Energy Energy Edmonds, J. andJ. Reilly. (1983)."Globalenergyproductionandusetotheyear 2050." Survey. of discovery ofoilandgasintheGulfMexico." Professional Paper 1252.USGeological Drew, L.J., J. H.Schuenemeyer, andW. J. Bawiec.(1982)."Estimationofthefuturerates models at the US Geological Survey." Drew, L. J. and J. H. Schuenemeyer. (1993). "The evolution and use of discovery process revisited-the problemoffieldgrowth." Drew, L. J. and J. H. Schuenemeyer. (1992). "A petroleum discovery rate forecast Pennsylvania StateUniversity. Oil & Gas Journal , 8 :6, pp. 419-32. 1965 symposium,Computerapplicationsinthemineralindustries Springer. The Bell Journal of Economics and Management Science Undiscovered PetroleumandMineralResources:assessment , Beyond oil:theviewfromHubbert'speak. United States (USA): 1958:February 24th, pp. 105-18. 56, pp. 615. AAPG Bulletin, Nonrenewable Resources 77:3, pp. 467-78. Energy Policy , 1 :1, pp. 51-60. , 2 , :1, pp. 94-121. 35:1, pp. 178- : FF-1-31: Refs_App:UKERC 23/09/2009 10:32 Page 181 Page 10:32 23/09/2009 Refs_App:UKERC two mega-oilfields."Energy Gowdy, J. andJ. Roxana. (2007)."Technology andpetroleumexhaustion:Evidencefrom 56. Gilliand, M.W. (1975)."Energyanalysisandpublicpolicy." 25. U. S. Geological Survey, Reston, VA, United States (USA): United States (USA). Gautier, D. L., T. R.Klett,andB. S. Pierce.(2005)."Globalsignificanceofreserve growth." United Kingdom (GBR): United Kingdom (GBR). Petroleum GeologyConference Petroleum geology; north-west Europe and global perspectives; proceedings of the 6th Gautier, D. L.andT. R.Klett.(2005)."Reserve growthinchalkfieldsoftheNorthSea" data." US Geological Survey Digital Data Series DDS-30. assessment ofUnitedStatesoilandgasresources-results,methodologysupporting Gautier, D. L., G. L. Dolton, K. I. Takahashi, and K. L. Varnes. (1995). "National C.J. Cleveland ed. Gautier, D. L.(2004)."Oilandnatural gasresourceassessment:geologicalmethods,"in 75-96. Gately, D. (2004)."OPEC'sincentives forfasteroutputgrowth." return onenergyinvestment forglobaloilandgasproduction." Gagnon, N., C.A.S. Hall,andL.Brinker. (2009)."Apreliminary investigation ofenergy continental shelf." Fryer, M. J. and J. V. Greenman (1990). "Estimating the oil reserve base in the UK undiscovered petroleum resources." Forman, D. andA.L.Hinde.(1985)."Improved statisticalmethodforassessmentof known fields." Forbes, K. F. and E. M. Zampelli. (2009). "Modelling the growth in gas reserves from the U.S. offshore." Forbes, K.F. andE.M.Zampelli. (2000)."Technology andtheexploratory successrate in Hopkins University Press. Fisher, F. M.(1964). and GrowthReconsidered Fisher, A. C. (1979). "Measures of natural resource scarcity," in V. Kerry Smith ed. Research Letters Farrell, A. E. and A. R. Brandt. (2006). "Risks of the oil transition." Business Economics Eysell, J. H. (1978). "The supply response of crude petroleum-new and optimistic results." Energy Journal, , 1 The Journal of the Operational Research Society The Encyclopedia of Energy Energy Journal, , pp. 1-6. , 1393, pp. 338-46. Supply andcostsintheUSpetroleumindustry . Baltimore, MD Johns Hopkins University Press. , 32:8, pp. 1448-54. , Vol. 6:169-75.TheGeologicalSocietyofLondon, 30, pp. 201-13. 21:1. AAPG Bulletin, . Cheltenham Edward Elgar. 69:1, pp. 106-18. Science Energy Journal, Energies, , , 189:4208, pp. 1051- 41:8, pp. 725-33. . Baltimore:John 2 Environmental :490-503. 25:2, pp. Scarcity 181

Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Refs_App:UKERC 23/09/2009 10:32 Page 182 Page 10:32 23/09/2009 Refs_App:UKERC 182 Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production the limitstoavailability anddiversity ofglobalconventional oilsupply." Hallock Jr, J. L., P. J. Tharakan, C.A.S. Hall,M.Jefferson,andW. Wu. (2004)."Forecasting Environmental and Energetic Issues the economy inanuncertainfuture,"D. Pimenteled. Hall, C. A. S., R. Powers, and W. Schoenberg. (2008). "Peak oil, EROI, investments and analysis fromanoptimist'sperspective." Greene, D. L., J. L. Hopson, and L. Jia. (2006). "Have we run out of oil yet? Oil peaking money?". Bergan et al, Inc.: Rockport, Texas. Grace, J. D. (2007)."Acloserlookatfieldreserve growth:Science,engineering,orjust sustainable societymusthave?" Hall, C.A.S., S. Balogh,andD. J. Murphy. (2009)."WhatistheminimumEROIthata Production Engineering Guo, B., W. C.Lyons, andA.Ghalambor. (2007)."ProductionDeclineAnalysis." estimates, methodforestimationandpolicyconsiderations." Harris, d. P. (1977). "Conventional crude oil resources of the United States: recent prospects Harbaugh, J. W., J. C. Davies, and J. Wendebourg. (1995). pp. 1673-96. influence onworldoilproduction." Höök, M., R.L.Hirsch, and K.Aleklett.(2009a)."Giantoilfielddecline rates andthe Energy Policy, Höök, M. and K. Aleklett. (2008). "A decline rate study of Norwegian oil production." thesis, UppsalaUniversity. Höök, M.(2009)."Depletionanddeclinecurve analysisin crudeoilproduction."Licentiate Holland, S. P. (2008). "Modeling peak oil." Technologies Laboratory. impacts, mitigation,&riskmanagement."91. USDepartmentofEnergy, NationalEnergy Hirsch, R.L., R.Bezdek,andWendling. (2005)."Peaking ofworldoilproduction: Energy Policy, Hirsch, R.L.(2008)."Mitigationofmaximumworldoilproduction:Shortagescenarios." Hirsch, R. L. (2005). "Shaping the peak of world oil production." and theworld."USGeologicalSurvey Circular522. Hendricks, T. A. (1965). "Resources of oil, gas and natural gas liquids in the United States 263-86. . Oxford: Pergamon Press. 36:11, pp. 4262-71. 36:2, pp. 881-89. . Burlington Gulf Professional Publishing, pp. 97-105. Energies, Energy Policy, . LondonElsevierpp. 113-36. Energy Policy, 2 Energy Journal, , pp. 25–47. 37:6, pp. 2262-72. 34:5, pp. 515-31. 29:2, pp. 61-79. Renewable Energy Systems: Materials and Society World Oil, Computing riskforoil Energy 226:10. Petroleum , , 29:11, 1 , pp. Refs_App:UKERC 23/09/2009 10:32 Page 183 Page 10:32 23/09/2009 Refs_App:UKERC IEA. (2008). "World Energy Outlook 2008." International Energy Agency, OECD: Paris. the Future."International EnergyAgency:OECD, Paris. IEA. (2005)."Resources toReserves -OilandGasTechnologies fortheEnergyMarkets of modeling, Washington,DC,UnitedStates,June 18-20 gas; Oilandgassupplymodeling;proceedings ofasymposium." Hubbert, M. K. (1982). "Techniques of prediction as applied to the production of oil and Grenon ed. Hubbert, M. K. (1979). "Hubbert estimates from 1956 to 1974 of US oil and gas," in Michel 2nd Session, Serial 93-40 (92-75). Washington: U.S. Government Printing Office. prepared fortheU.S. SenateCommitteeonInteriorand InsularAffairs,93rdCongress, Hubbert, M.K.(1974)."U.S. EnergyResources: AReview asof1972"Backgroundpaper American AssociationofPetroleumGeologists States; Natural gasesofNorthAmerica;Part 4,Papers ofgeneral scope." Hubbert, M. K. (1968). "Degree of advancement of petroleum exploration in United States." Hubbert, M. K. (1967). "Degree of Advancement of Petroleum Exploration in United Hubbert, M. K. (1966). "Reply to J.M. Ryan." States. Resources of the National Academy of Sciences-National Research Council." 153p. United Hubbert, M.K.(1962)."EnergyResources: AReport totheCommitteeonNatural Shell Development Company: Dallas,TX. industry." Hubbert, M.K.(1959)."Techniques ofpredictionwithapplication tothepetroleum Company: San Antonio, Texas. Dsitrict, Divisionofproduction,AmericanPetroleumInstitute Hubbert, M. K. (1956). "Nuclear energy and the fossil fuels." and socialoptimality." Howarth, R.B. andR.B. Norgaard.(1990). "Intergenerational resourcerights,efficiency, Bulletin, Geological Survey estimation procedure for accumulation size distributions by play." Houghton, J. C., G.L.Dolton,R.F. Mast,C.D. Masters,andD. H.Root. (1993)."US Economy, Hotelling, H. (1931). "The economics of exhaustible resources." oil fieldproductionbehaviour." Höök, M., B. Söderbergh,K.Jakobsson, andK.Aleklett.(2009b)."Theevolution ofgiant AAPG Bulletin, 77:454-486. 44th AnnualmeetingoftheAmericanAssociationPetroleumGeologists 39:2, pp. 137-75. Methods andModelsforAssessingEnergyResources Land Economics 51, pp. 2207-27. Natural ResourcesResearch , 66:1, pp. 1-11. AAPG Bulletin, , 9 :2, pp. 2149-70. , 1980:631, pp. 16-141. , 18:1, pp. 39-56. 18, pp. 284-86. . OxfordPergamon Press. Meeting of the Southern The Journal of Political . ShellDevelopment Oil andgassupply Memoir - AAPG : 43. 183

Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Refs_App:UKERC 23/09/2009 10:32 Page 184 Page 10:32 23/09/2009 Refs_App:UKERC 184 Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production and econometricmodels." Kaufmann, R.K.(1991)."Oilproductionin the lower48states:Reconciling curve fitting Research Kaufman, G. M. (2005). "Where have we been? Where are we going?" Karbuz, S. (2004)."Conversion factorsand oilstatistics." resources." Kaufman, G.M.(1993). "Statisticalissuesintheassessmentofundiscovered oilandgas Cambridge, MABallingerPublishingCompany. Resources inanUncertainFuture:Coal,Gas, OilandUraniumSupplyForecasting Adelman, John C. Houghton, Gordon M. Kaufman and Martin B. Zimmerman eds. Kaufman, G.M.(1983). "Oilandgas:Estimationofundiscovered resources,"inM.A. in GeologySeries1 Methods ofEstimatingtheVolumeUndiscoveredOilandGasResources:AAPGStudies Kaufman, G.M.(1975)."Aprobabilisticmodelofoilandgasdiscovery," inJ.D. Hauned. exploration Kaufman, G. M. (1963). global conventional gasoutputpeakingin2019." Imam, A., R. A. Startzman, and M. A. Barrufet. (2004). "Multicyclic Hubbert model shows Economics exploration and development in the U.S. onshore Louisiana, 1977-1994." Iledare, O. andA.G.Pulsipher. (1999)."Sourcesofchangeinproductivitypetroleum IEA. (2009b). "Oil Market Report." 10 April. International Energy Agency. investment." International Energy Agency: Paris. IEA. (2009a)."Theimpactofthefinancialandeconomiccrisisonglobalenergy Petroleum Technology, Jung, H.(1997)."Reserves definitions- Clarifying theuncertainties." Energy Programme: The Netherlands. high oilpricesinasupply-constrained world."Report 2008/03.ClingendaelInternational Jesse, J.-H. andC.van derLinde.(2008). "Oilturbulenceinthenextdecade:Anessay on International Energy Program. Janssen, E. (2005). "Can Russian oil growth be sustained?" Working Paper. Clingendael Global Energy Systems Group, University of Uppsala. Rate Modelforforecastingoilproduction:itsusesandmisuses."Report No. PP-09.01. Jakobsson, K., B. Söderbergh,M.Höök,andK.Aleklett.(2009)."TheMaximumDepletion & GasJournal Ivanhoe, L.F. andG.Leckie.(1993)."Globaloil, gasfields,sizes tallied,analyzed." , 14:3, pp. 145-51. . Englewood Cliffs, New Jersey: Prentice-Hall. 21:3, pp. 261-72. Energy Journal, , 1993:February 15, pp. 87-91. , pp. 113-42. 36:4, pp. 26-30. Resources andEnergy, Statistical decisionandrelatedtechniquesinoilgas 14:1, pp. 183-213. Oil & Gas Journal 13:1, pp. 111-27. Energy Policy, , 102:31. Journal ofCanadian Natural Resources 32:1, pp. 41-45. Energy Energy Oil . Refs_App:UKERC 23/09/2009 10:32 Page 185 Page 10:32 23/09/2009 Refs_App:UKERC Literature Krautkraemer, J. A.(1998)."Nonrenewable resourcescarcity." supply." 03-01. 30. Resources for the Future: Washington, D.C. Krautkraemer, J. and M. Toman. (2003). "Fundamental economics of depleteable energy Mexico. Association ofPetroleumGeologistsInternational ConferenceandExhibition Klett, T. R.andM.K.Verma. (2004). "Reserve growthinoilandgasfieldsof Mexico." Giant oilandgasfieldsofthedecade,1990-1999." Klett, T. R.andJ. W. Schmoker. (2003b)."Reserve growthoftheworld'sgiantoilfields; pp. 107-22. in M.T.; Halboutyed. Klett, T. R.andJ. W. Schmoker. (2003a)."Reserve Growth oftheWorld's GiantOil Fields," Geological Survey. Petroleum Assessment 2000: supporting data." USGS Open File a report 2007-1021. US Klett, T. R., D. L.Gautier, andT. S. Ahlbrant. (2007)."Anevaluation oftheWorld Survey World Petroleum Assessment2000." Klett, T. R., D. L. Gautier, and T. S. Ahlbrandt. (2005). "An evaluation of the US Geological Petroleum Geoscience Klett, T. R.andD. L.Gautier. (2005)."Reserve growthinoilfieldsoftheNorthSea." implementation." Klett, T. R. (2005). "United States Geological Survey's reserve-growth models and their terminology," inC.J. Cleveland ed. Klett, T. R. (2004). "Oil and natural gas resource assessment: classifications and Kennedy, P. (2003). the UKcontinentalshelf." Kemp, A.G.andS. Kasim. (2005)."Aredeclinerates reallyexponential?Evidencefrom how quickly, andmarket driven." Kaufmann, R.K.andL.D. Shiers.(2008)."Alternatives toconventional crudeoil:When, Economic, geological,andinstitutionaldeterminants." Kaufmann, R.K.andC.J. Cleveland. (2001)."Oilproductioninthelower48states: the Environment to fail,damagetheeconomy, anddamagetheenvironment." Kaufmann, R. K. and C. J. Cleveland. (1991). "Policies to increase US oil production: likely markets: implications for climate change policy." Kaufmann, R. K. (1994). "The relation between marginal product and price in US energy , 36:4, pp. 2065-107. , Natural Resources Research 16, pp. 379-400. A Guide to Econometrics. Oxford: Blackwell. Giant oilandgasfieldsofthedecade1990-1999.AAPGMemoir78 , 11:2, pp. 179-90. The Energy Journal, Ecological Economics Encyclopaedia ofEnergyElsevier AAPG Bulletin, , 26:1, pp. 27-50. 14:3, pp. 249-64. Energy Economics AAPG Memoir, Energy Journal, , 67:3, pp. 405-11. 89:8, pp. 1033-42. Annual ReviewofEnergyand 78, pp. 107-22. , Journal of Economic , pp. 595-605. 16:2, pp. 145-58. 22:1, pp. 27-49. : Cancun, , 185

Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Refs_App:UKERC 23/09/2009 10:32 Page 186 Page 10:32 23/09/2009 Refs_App:UKERC 186 Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production lognormal or stretched exponential?" Laherrère, J. (2000a)."Distributionoffieldsizes inapetroleumsystem;parabolic fractal, Lee, P. J. (2008). AIME, Larkey, C.S.curves." (1925)."Mathematicaldeterminationofproductiondecline Workshop on Oil and Gas Depletion Laherrère, J. H. (2005). "Forecasting production from discovery." and fromultimates." Laherrère, J. H.(2002)."Modellingfutureliquidsproductionfromextrapolation ofthepast and GasJournal Laherrère, J. H.(2000b)."Learnstrengths, weaknesses tounderstandHubbertcurve." Laherrère, J. (2003). "Future of oil supplies." cyberconference oftheWEC. Laherrère, J. (2001)."IstheUSGS2000AssessmentReliable?" publishedonthe evaluation of petroleum resources." Lee, P. J. and P. C. Wang. (1983). "Probabilistic formulation of a method for the University Press. Petroleum Engineers: Long Beach, California. SPE Western Regional/AAPG Pacific Section Joint Meeting Li, K.H., R.N.;.(2003)."A DeclineCurve AnalysisModelBasedonFluidFlowMechanisms." models." Li, K.andR.N.Horne.(2007)."Comparison andverification ofproductionprediction Mathematical Geology , 15:1, pp. 163-82. it peak?" Laherrère, J. H. (1999b). "World oil supply - what goes up must come down, but when will Laherrère, J. H. (1996). "Discovery and production trends." Groningen, 21November 2006. Laherrère, J. (2006). "Oil and gas, what future?" arithmetic?" Laherrère, J. (1999a)."Reserve growth:technologicalprogress,orbadreportingand Second International Workshop on Oil and Gas Laherrere, J. (2003)."Modelingfutureoilproduction,populationandtheeconomy." models," inC.J. Cleveland ed. Laherrère, J. (2004)."Oilandnatural gasresourceassessment:productiongrowthcycle pp. 227-67. 71, pp. 1322-28. Journal of Petroleum Science and Engineering Oil and Gas Journal Geopolitics ofEnergy, , 98:16, pp. 63-64. Statistical MethodsforEstimatingPetroleum Resources Energy Exploration&Exploitation , 97:5, pp. 57-64. Encyclopaedia of Energy 22:4, pp. 7-16. . Association for the Study of Peak Oil: Lisbon. Marine and Petroleum Geology Journal of the International Association of Energy Exploration and Exploitation : 80: Paris. Groningen AnnualEnergyConvention: , . AmsterdamElsevier. , 20:6, pp. 457-79. 55, pp. 213-20. , Vol. SPE 83470. Society of OPEC Bulletin, pp. 7-11. ASPO IV International , 17:4, pp. 539-46. : Oxford , Trans ASPO 21:3, Oil Refs_App:UKERC 23/09/2009 10:32 Page 187 Page 10:32 23/09/2009 Refs_App:UKERC 100-04. Marsh, R.G.(1971)."How muchoilarewereallyfinding." Freeman. Mandelbrot, B. B. (1977). New York andLondon. and P.R. La Pointe (eds) Fractals in Petroleum Geology and Earth Processes, Plenum Press, International BusinessMachinesResearch NoteNC-146, June29,reprintedinC.CBarton Mandelbrot, B. B. (1962). "Statistics of natural resources and the law of Pareto." and petroleumexploration intheGulfofMexico" Managi, S., J. J. Opaluchb, D. Jinc,and T. A.Grigalunas.(2005)."Technological change 47, pp. 388-409. depletion inoffshoreoilandgas" Managi, S., J. J. Opaluch,D. Jin,andT. A.Grigalunas.(2004)."Technological changeand Bulletin, Magoon, L.B. andR.M.O. Sanchez.(1995)."Beyond thepetroleumsystem." pp. 454-98. the natural gasshortage." MacAvoy, P. W. andR.S. Pindyck. (1973)."Alternative regulatorypoliciesfordealingwith Hubbert modelers'assessment." Lynch, M. C. (2003). "Petroleum resources pessimism debunked in Hubbert model and of EconomicsandFinance Lynch, M. C. (2002). "Forecasting oil supply: theory and practice." sciences: keys andclues." Limpert, E., W. A.Stahel,andM.Abbt.(2001)."Lognormaldistributionsacrossthe http://sepwww.stanford.edu/sep/jon/world-oil.dir/lynch/worldoil.html [Accessed 1 Lynch, M.C.(1998)."CryingWolf: warnings aboutoilsupply." [Online]Available from: decline in reserves." Luther, L.C.(1985)."Linearizationandregressionanalysis techniquepredictshyperbolic Management Service. of Mexico and Atlantic Continental Shelf." OCS Report MMS 96-003, Appendix A. Minerals of the 1995 assessment of the conventional recoverable hydrocarbon resources of the Gulf Lore, G.L., K.M.Ross, B. J. Bascle,L.D. Nixon, andR.J. Klazynski. (1996)."Summary Advances in the Economics of Energy and Resources Lynch, M.C.(1999)."Thewolfatthedoororcryingwolf:Fears aboutthenextoilcrisis." 2008] 79:12, pp. 1731-36. Oil and Gas Journal , Bioscience, The Bell Journal of Economics and Management Science 42:2, pp. 373-89. Fractals: form,chanceanddimension Oil and Gas Journal Journal ofEnvironmentalEconomicsandManagement 51:5, pp. 341-52. :August 261985,pp. 78-79. Energy Policy, , , 101:27, pp. 38-47. 11, pp. 117-42. Oil & Gas Journal 33:5, pp. 619-32. . SanFrancisco: W.H. The QuarterlyReview , 69:14, pp. st , AAPG 4 July :2, , 187

Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Refs_App:UKERC 23/09/2009 10:32 Page 188 Page 10:32 23/09/2009 Refs_App:UKERC 188 Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production 20. province: ThecaseoftheNorwegianContinental Shelf." Mohn, K. and P. Osmundsen. (2008). "Exploration economics in a regulated petroleum outer continental shelf." Minerals Management Service, Report MMS 96-0034. MMS. (1996). "An assessment of the undiscovered hydrocarbon potential of the nation's London. Mitchell, J. (2004b). "Petroleum reserves in question." Chatham House. Mitchell, J. (2004a). "Petroleum Reserves in Question." SDP BP 04/03. Chatham House: Mills, R. M. (2008). Miller, R. G. (2005). "Global oil supply to 2030." half life and the world crude oil resource." Miller, R.G.(1992)."Theglobaloilsystem:therelationshipbetweenallgeneration, loss, Change, substitution: themathematicsoflogletlabsoftware." Meyer, P. S., J. W. Yung, andJ. H.Ausbel.(1999)."Aprimeron logistic growthand pp. 89-102. Meyer, P. S. (1994)."Bi-logisticgrowth." paradigm?" Merriam, D. F., L.J. Drew, andJ. H.Schuenemeyer. law:aviablegeological (2004)."Zipf's Science Menard, H.W. andG.Sharman.(1975). "Scientificusesofrandom drillingmodels." Statistical SocietyA future oilandgasdiscoveries inmatureexploration provinces." Meisner, J. andF. Demirmen.(1981)."Thecreamingmethod:aproceduretoforecast update Meadows, D. H., J. Randers, and D. L. Meadows. (2004). review andappraisal." Meade, N.M.(1984)."Theuseofgrowthcurves inforecastingmarket development -a of Western Canada." Mcrossan, R.G.(1969)."Ananalysisofsizefrequency distributionofoilandgasreserves 60, pp. 32-40. McKelvey, V. E.(1972)."Mineral resourceestimatesandpublicpolicy." 82:11, pp. 2110-34. McCabe, P. J. (1998)."Energyresources-Cornucopiaoremptybarrel?" the United States." US Geological Survey and US Minerals Management Service. Mast, R. F. . WhiteRiver Junction,VT: ChelseaGreenPub. , 61, pp. 247-71. 190, pp. 337. et al. Natural ResourcesResearch (1989). "Estimates of undiscovered conventional oil and gas resources in , The myth of the oil crisis Canadian Journal of Earth Sciences 144:1, pp. 1-31. Journal ofForecasting , Technological ForecastingandSocialChange 13:4. AAPG Bulletin, , . Westport, CT, USA: Praeger Publishers. 3 , pp. 49-451. Unpublished Manuscript Energy Economics , Technological ForecastingSocial 6 76:4, pp. 489-500. Limits to growth: The 30-year :201, pp. 201-11. Journal oftheRoyal American Scientist, . , AAPG Bulletin, 30:2, pp. 303- , 47, Refs_App:UKERC 23/09/2009 10:32 Page 189 Page 10:32 23/09/2009 Refs_App:UKERC industry." NationalPetroleum Council:Washington, DC. NPC. (1995)."Research development anddemonstration needsoftheoiland gas Activity Nordhaus, W. D. (1973). "Allocation of energy resources." Instititue ofStandardsandTechnology. NIST/SEMATECH. (2008)."e-HandbookofStatisticalMethods." Vol. 2008.National pp. 89-110. Nelder, J. A.(1971)."Thefittingofageneralisation of the logisticscurve." production." Nehring, R. (2006c). "Two basins show Hubbert's method underestimates future oil production, EUR." Nehring, R. (2006b). "Post-Hubbert challenge is to find new methods to predict Basin." Nehring, R.(2006a)."HowHubbertmethodfailstopredictoilproductioninthePermian TX, UnitedStates,May20-23,1984 SEPM/EMD/DPA." petroleum resourceassessment;AAPGannualconvention withdivisions; Nehring, R. (1984). "Estimating reserve growth; opportunities and challenges for papers natural gas"inD.L. MeadowsandD.H. Meadowseds. Naill, R.F. (1973)."Thediscovery lifecycle ofafiniteresource:AcasestudyinU.S. Press. and nonlinear regression: a practical guide to curve fitting. New York: Oxford University Motulsky, H.andA.Christopoulos. (2004). Journal Moroney, J. R.andM.D. Berg.(1999)."Anintegrated modelofoilproduction." Washington, DC. Moore, C.L.(1966)."ProjectionsofUSpetroleumsupplyto1980".OfficeOilandGas: domestic crude oil availability." US Department of the Interior. Moore, C.L.(1962)."Methodforevaluating UScrudeoilresourcesandprojecting theoretical modelling." Mohr, S. H.andG.M.Evans. (2008)."Peak oil:testingHubbert'smethodologyvia will peak."Oil and Gas Journal Mohr, S. H. and G. M. Evans. (2007). "Mathematical model forecasts year conventional oil . CambridgeMAWright-Allen PressInc. , , Oil and Gas Journal 20:1, pp. 105-24. 3 , pp. 529-70. Oil and Gas Journal AAPG annual convention with divisions; SEPM/EMD/DPA, San Antonio, Oil and Gas Journal Natural Resources Research , 104:15, pp. 30-35. , 105:17, pp. 45-46. , 104:13, pp. 37-42. , 68:4, pp. 511. , 104:16, pp. 43-46. Fitting modelstobiologicaldatausinglinear , 17:1, pp. 1-11. Toward globalequilibrium:collected Brookings Papers on Economic Biometrics, The Energy 17, 189

Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Refs_App:UKERC 23/09/2009 10:32 Page 190 Page 10:32 23/09/2009 Refs_App:UKERC 190 Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Pindyk, R. S. and D. L. Rubinfeld. (1998). Review, Pike, R.(2006)."Have weunderestimated theenvironmental challenge?"Petroleum pp. 352-70. Pickering, A.(2008)."Theoilreserves productionrelationship." gas: theexampleof Scotianshelf." Power, M. and J. D. Fuller. (1991). "Predicting the discoveries and finding costs of natural consequence ofsamplingbias." Power, M.(1992)."Lognormalityin the observed size distributionofoilandgaspoolsasa (Third Edition) Porges, F. (2006). "Analysis of decline and type curves." The President's Materials Policy Commission: Washington, D.C. PMPC. (1952)."Resourcesfreedom: Volume for I-Foundations forgrowth andsecurity." Boston, MA:McGraw-Hill. Journal of Political Economy Pindyck, R. S. (1978)."Optimalexploration andproductionofnon-renewable resources." International JournalofForecasting Pesaran, M.H.andSamiei.(1995)."Forecasting ultimateresourcerecovery." economics andevaluation,SocietyofPetroleumEngineersAIME Pelto, C.R.(1973)."Forecasting ultimateoilrecovery." 22:Accepted for publication May 3, 2008. of technology." Patzek, T. W. (2008)."Exponentialgrowth,energeticHubbertcycles andtheadvancement Pareto, V. (1897). "Cours d'economie politique." Vol. Vol I and II: Rouge, Laussane. OPEC. (2008). "OPEC Secretariat Web Page." Vol. 2008. OPEC: Austria. Alberta." Oil and Gas Conservation Board. OGCB. (1970)."Reserves ofCrudeOil,Gas,Natural GasLiquidsandSulphur, Province of Publishing Co. interrelationships ofresources,reserves,anduse,1980-2080 Odell, P. R.andK.E.Rosing. (1980). 54. Odell, P. R.(1973)."Thefutureofoil:arejoinder." global oil and natural gas." National Petroleum Council: Washington, DC. NPC. (2007)."Facing theHardTruths aboutEnergy:acomprehensive viewto2030of June , pp. 26-27. . Burlington Gulf Professional Publishing, pp. 1235-337. Archives of mining sciences of the Polish Academy of Sciences , 86:5, pp. 841-61. Mathematical Geology , , 11:4, pp. 543-55. The Energy Journal, The futureofoil:asimulationstudythe Econometric models and economic forecasts Geographical Journal 24:8. Reservoir Engineering Handbook 12:3, pp. 77-93. Symposium onpetroleum Energy Economics . NewYork: Nichols 45-52: Dallas. , 139:3, pp. 436- , 30:2, . , Refs_App:UKERC 23/09/2009 10:32 Page 191 Page 10:32 23/09/2009 Refs_App:UKERC pp. 1713-27. limitations of logistic projections." Ryan, J. M. (1965). "National Academy report on energy resources: discussion of Rubin, J. and P. Buchanan. (2007). "OPEC’s Growing Call on Itself." CIBC World Markets. Bulletin, Root, D. H. and R.F. Mast.(1993)."Futuregrowthinknowoilandgasfields." assessment." Root, D. H. and E.D. Attanasi. (1993)."Smallfieldsinthenationaloilandgas Energy andtheEnvironment, 22,pp. 217-62. Rogner, H.(1997)."Anassessment ofworldhydrocarbon resources."AnnualReview of from UppsalaUniversity. Robelius, F. (2007). "Giant Oil Fields and Their Importance for Peak Oil." 271-89. decline: Granger causalityandmulti-cycle Hubbertcurve." Reynolds, D. B. andM.Kolodziej. (2008)."Former Soviet UnionoilproductionandGDP decreasing scarcity." Reynolds, D. B. (1999)."Themineral economy: howpricesandcostscanfalselysignal Hubbert approach:TheLOPEX model." Rehrl, T. and R. Friedrich. (2006). "Modelling long-term oil price and extraction with a Journal Radler, M.(2007)."Oil,gasreserves inchup, productionsteadyin2007." Mathematical Geology, Rabinowitz, D. (1991). "Using exploration history to estimate undiscovered resources." exploration." Quirk, D. G. and R. Ruthrauff. (2006). "Analysis of reserves discovered in petroleum Mexico." Priest, T. (2007). "Extraction not creation: the history of offshore petroleum in the Gulf of Pratt, W. E.(1942). Change, technological innovation in oil and gas exploration." Power, M. and E. Jewkes. (1992). "Simulating the discovery rate implications of supply." Power, M.andJ. D. Fuller. (1992b)."Areviewofmethods forestimatinghydrocarbon of hydrocarbon deposits." Power, M. and J. D. Fuller. (1992a). "A comparison of models for forecasting the discovery , 105:48, pp. 22-23. 4, 77:3, pp. 479-84. 41:2, pp. 161-75. Enterprise andSociety 2 , pp. 128-45. Journal ofPetroleumGeology AAPG Bulletin, Oil in the earth Ecological Economics 23:2, pp. 257-74. Journal ofForecasting 77:3, pp. 485-90. , 8 :2, pp. 227-67. Bulletin of the Association of Petroleum Geologists : Lawrence, University of Kansas Press. Energy Policy, , , 31:1, pp. 155-66. 29:2, pp. 125-46. , 11, pp. 183-93. Technological ForecastingandSocial 34:15, pp. 2413-28. Energy Economics Licentiate thesis , 30:2,pp. Oil &Gas AAPG , 49, 191

Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Refs_App:UKERC 23/09/2009 10:32 Page 192 Page 10:32 23/09/2009 Refs_App:UKERC 192 Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production additions." Skrebowski, C. (2006). "Prices holding steady, despite massive planned capacity 2005:October, pp. 36-40. Skrebowski, C.(2005)."Pricessetfirm,despite massive new capacity." pp. 18-20. Skrebowski, C.(2004)."Oilfieldmegaprojects 2004." International: Houston, Texas, USA. Simmons, M.R.(2002)."Theworld'sgiant oilfields:."SimmonsandCompany be met?" Simmons, M.R.(2000)."Fightingrisingdemand&declinecurves: canthechallenge Massachusetts InstituteofTechnology: Cambridge. depletion." MIT-CEEPR 99-011.CentreforEnergyEnvironmental Policy Research, Sem, T. andD. Ellerman.(1999)."NorthSeareserve appreciation,productionand DC. Federal Register, 73(132):39526".USSecurities andExchange Commission:Washington, SEC. (2008)."Modernizationoftheoilandgasreportingrequirements;proposedrule Oil and Gas Journal Schulyer, J. (1999)."Probabilisticreserves definitions, practices needfurther refinement " US Geological Survey, Open File Report. procedure toforecastfutureoilandgasincorporating fieldgrowth."ARDSv5.01.98-111. Schuenemeyer, J. H.andL.J. Drew. (1994)."Descriptionofadiscovery processmodelling (USA). resources." U. S. GeologicalSurvey, Reston, VA, UnitedStates(USA): UnitedStates Schmoker, J. W. (2000)."Reserveestimates ofoilandnaturalon growtheffects gas target forEORtechnologies." Sandrea, R.andI.Sandrea.(2007)."Globaloilreserves: recovery factorsleave vast Studies 54 American Association of Petroleum Geologists. Salvador, A.(2005)."Energy:AHistoricalPerspective and21stCenturyForecast. ."AAPG Review:August, pp. 26-29. Salameh, M. G. (2004). "How realistic are OPEC’s proven oil reserves?" 21:219-235. Ryan, J. T. (1973). "An analysis of crude oil discovery rates in Alberta." gas reserves andavailability." Ryan, J. M.(1966)."Limitationsofstatisticalmethodsforpredictingpetroleumandnatural SPE Asia Pacific Oil & Gas Conference Petroleum Review, , 97:22, pp. 26-30. Oil and Gas Journal Journal of Petroleum Technology 2006:April, pp. 28-31. Yokohama, Japan. :November 05. Petroleum Review , , pp. 231–37. Petroleum Review, AAPG Bulletin, 2004:January, Petroleum Refs_App:UKERC 23/09/2009 10:32 Page 193 Page 10:32 23/09/2009 Refs_App:UKERC of NorthSeaoilfields." Smith, J. L. and G. L. Ward. (1981). "Maximum likelihood estimates of the size distribution No. 15293. Spivey, J. P. (1986)."ANewAlgorithmforHyperbolicDecline Curve fitting." eLibrary Paper Middle EastWellEvaluationReview, Speers, R.andP. S. Dromgoole.(1992)."Managing uncertaintyinoilfieldreserves." Society Petroleum Evaluation Engineers. Engineers, American Association of Petroleum Geologists, World Petroleum Council, and SPE. (2007). "Petroleum Resources Management System." Society of Petroleum Energy Research Centre:London. resources." UKERC Review of evidence on global oil depletion: Technical Report 5 UK Sorrell, S. andJ. Speirs.(2009)."Methodsforestimating ultimatelyrecoverable global oildepletion.UKEnergyResearch Centre:London. Sorrell, S. and J. Speirs. (2008). "Assessment Protocol." UKERC Review of evidence for Energy Research Centre:London. Sorrell, S. (2008)."ScopingNote."UKERCReview ofevidenceforglobaloildepletion.UK reviews." Sorrell, S. (2007)."Improving theevidencebaseforenergypolicy:Theroleofsystematic Canadian oilsandsindustry." Söderbergh, B., F. Robelius, andK.Aleklett.(2007)."A crash programme scenarioforthe stratigraphic perspective." Sneddon, J. W., F. Sarg, and X. Ying. (2003). "Exploration play analysis from a sequence Smith, M. (2008). "EnergyFiles Forecasting Model - Oil production." EnergyFiles Ltd. production." Smith, J. L. and J. L. Paddock. (1984). "Regional modelling of oil discovery and Statistics, Smith, J. L. (1980). "A probabilistic model of oil discovery." Cambrige, Massachusetts:MITPress. Smil, V. (2003). domain." Slade, M.E.(1982)."Trends innatural-resource commodityprices:Ananalysisof thetime depletion?" Skrebowski, C. (2007). "New capacity fails to boost 2006 production - Delays or Journal of Environmental Economics and Management Energy Policy, 62, pp. 587–94. Petroleum Review, Energy Economics Energy atthecrossroads:Globalperspectivesanduncertaintites Mathematical Geology, 35:3, pp. 1858-71. Search andDiscovery Energy Policy, , 2007:February, pp. 40-42. 6 :1, pp. 5-13. 12. 35:3, pp. 1931-47. 13, pp. 399-413. :Article 40079. Review ofEconomicsand , 9 :June, pp. 122-37. . 193

Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Refs_App:UKERC 23/09/2009 10:32 Page 194 Page 10:32 23/09/2009 Refs_App:UKERC 194 Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Basin, Russia." Verma, M.K.andG.F. Ulmishek.(2003)."Reserve growthinoilfieldsofWest Siberian Natural ResourcesResearch Verma, M.K.(2005)."Anewreserve growthmodelforUnitedStatesoilandgasfields." States (USA): UnitedStates (USA). model for United States oil and gas fields." 14. U. S. Geological Survey, Reston, VA, United Verma, M. K. (2003). "Modified Arrington method for calculating reserve growth; a new north of the Artic Circle." Factor Sheet 2008-3049. United StatesGeological Survey. USGS. (2008)."Circum-Artic resourceappraisal: estimatesofundiscovered oilandgas CD-ROM set). U.S. Geological Survey: Reston, VA, USA. results byUSGSWorld EnergyAssessmentTeam." USGSDigitalDataSeriesDDS-60(four VA, USA. USGS. (2000). "USGS World Petroleum Assessment 2000: description and USGS. (1995)."Nationaloilandgasassessment."UnitedStatesGeologicalSurvey Reston, Modelling Uri, N. D. (1982). "Domestic crude oil resource appraisal." Nations: NewYork. UN. (2004)."UNFramework ClassificationforFossil EnergyandMineral Resources." United Biosciences, Tsoularis, A.andJ. Wallace. (2002)."Analysisoflogisticgrowthmodels." Association of Petroleum Geologist Annual Conference 1998: influenceofreservoir drive mechanismsandrecovery technology." Tennyson, M.E.(2002)."Regional patternsofgrowthUnitedStatesgiantoilfields1945- much oilisleft." Taylor, P. J. (1998). "Author's reply to discussion of modeling the U.S. oil industry: How 70. supply modelling Stitt, W. (1982)."Currentproblemsinoilandgasmodelling,"S.I. Gassed. Stevens, P. (2008). "The Coming Oil Supply Crunch." Chatham House: London. 33:3, pp. 219-49. petroleum resourcesintheUnitedStates." Sterman, J. D., G. P. Richardson, and P. Davidsen. (1988). "Modelling the estimation of Journal ofEnergyDevelopment Stark, P. H.andK.Chew. (2005)."Globaloilresources:issuesandimplications." with triangular field profiles." SPE: Dallas, Texas, USA. Stark, P. H. (2008). "Peak production in an oil depletion model , 6 :2, pp. 119-23. 179:1, pp. 21-55. Natural Resources Research . MarylandNationalBureauofStandardSpecialPublication631,pp. 142- Journal ofPetroleumTechnology , 14:2, pp. 77-89. Journal of Interdisciplinary Mathematics , 30:2, pp. 159-70. , Technological ForecastingandSocialChange 12:2, pp. 105-19. , 50:11, pp. 82-83. : Houston. Applied Mathematical , 11, pp. 695-706. Mathematical Oil andgas American The , Refs_App:UKERC 23/09/2009 10:32 Page 195 Page 10:32 23/09/2009 Refs_App:UKERC Zipf, G.K.(1949). 36. UnitedStatesGeologicalSurvey: Washington, D.C. Zapp, A.D. (1962)."Futurepetroleum producingcapacityoftheUnitedStates."1142-H. fuels, radioactivemineralsandgeothermalenergy Zapp, A. D. (1961). "World petroleum resources." WPC. (2007)."Petroleum Resources ManagementSystem." World Petroleum Council. WorldOil. (2007). "Operators ride the crest of the global wave." expected to peak in 21st century." Wood, J. H., G.R.Long,andD. F. Morehouse.(2003)."World conventional oilsupply Administration: Washington DC. future isneitherasrosyorbleaksomeassert.".7.EnergyInformation Wood, J. H., G.Long,andD. F. Morehouse.(2000)."Longtermoilsupplyscenarios: The review " Wiorkowski, J. J. (1981)."Estimatingvolumes of remainingfossilfuelresources:acritical of EnergyandClimateChange:London. Wicks, M.(2009)."Energysecurity:anationalchallengeinchangingworld."Department Greenwich JAIPress,pp. 51-68. evaluation," inJ.B. Ramsay ed. White, L. P. (1981). "A play approach to hydrocarbon resource assessment and AAPG Bulletin, White, D. A. and H. M. Gehman. (1979). "Methods of estimating oil and gas resources." Academy of Political and Social Science, White, D. (1920). "The petroleum resources of the world." Annals of the American of theParis Basin,France." Wendebourg, J. and C. Lamiraux. (2002). "Estimating the ultimate recoverable reserves occurence." Weeks, L. G. (1952). "Factors of sedimentary basin development that control oil Quarterly Review of Economics and Finance Watkins, G. C. (2002). "Characteristics of North Sea oil reserve appreciation." of theGulfMexicooutercontinentalshelf." Walls, M. A. (1994). "Using a 'hybrid' approach to model oil and gas supply: a case study existing approaches." Walls, M.A.(1992)."Modelingandforecastingthesupplyofoilgas:Asurvey of Journal of the American Statistical Association AAPG Bulletin, 63, pp. 2183-92. Human Behavior and the Principle of Least-Effort Resources and Energy, 36, pp. 2017-124. Oil and Gas Science and Technology The economicsofexplorationforenergyresources Offshore 89, pp. 111-34. , , 63:4, pp. 90, 92, 94, 150. 42:2, pp. 335-72. 14:3, pp. 287-309. Land Economics :. Domestic and world resources of fossil , 76:375, pp. 534-47. , 70:1, pp. 1-19. , World Oil, 57:6, pp. 621-29. : Addison-Wesley. 228:9, pp. 4. The . 195

Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Refs_App:UKERC 23/09/2009 10:32 Page 196 Page 10:32 23/09/2009 Refs_App:UKERC 196 Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Refs_App:UKERC 23/09/2009 10:32 Page 197 Page 10:32 23/09/2009 Refs_App:UKERC • • • • • • • • • • Peer Reviewers • • • • members were as follows: input totheinitialframing oftheissues,literature search,synthesisanddrafting. The on relevant aspectsofoilsupply. Itmettwiceduringthecourseofproject, providing The Expert Group was chosen for its combination of academic and commercial expertise Expert Group • • Technology (ICEPT).Thecontributorswere: University ofSussexandJamieSpeirstheImperialCollegeCentreforEnergyPolicy and The assessmentwas ledbytheSteve SorrelloftheSussexEnergyGroup(SEG)at Project Team Annex 1:Project team,ExpertGroup andcontributors Da Progr Professor Paul Stevens, SeniorResearch Fellow, Energy, Environment andDevelopment Dr MichaelSmith,ChiefExecutive, Energyfiles Chris Skrebowski, Director, Peak Oil Consulting Simon Wheeler, Independent Consultant ( (Technical Report7 Godfrey Boyle, Director, EnergyandEnvironment Research Unit,TheOpenUniversity Roger Bentley, Department of Cybernetics, University of Reading ( Richard Miller, IndependentConsultant( Dr Lucia van Geuns, Deputy Head, Clingendael International Energy Programme. of Pennsylvania Professor Robert Kaufmann, CentreforEnergyand Environmental Studies,University Progr Professor JohnV. Mitchell,AssociateFellow, Energy, Environment andDevelopment Dr RichardMiller, IndependentConsultant Dr Ken Chew, IHSEnergy Dr Roger Bentley, DepartmentofCybernetics,University ofReading Adam Brandt University of California, Berkley, ( (Technical Reports2and3 Erica Thompson,DepartmentofEarthScienceandEngineering,ImperialCollege vid Strahan, Author, journalist and film maker amme, ChathamHouse amme, ChathamHouse ) ) Technical Reports4and7 Technical Report7 Technical Report6 ) ) ) Technical Report 7 ) 197

Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production Refs_App:UKERC 23/09/2009 10:32 Page 198 Page 10:32 23/09/2009 Refs_App:UKERC 198 Global Oil Depletion: An assessment of the evidence for a near-term peak in global oil production These reports are all available to download from the UKERC website. Roger Bentley, Richard Miller, Godfrey Boyle and Simon Wheeler Technical Report7:Comparisonofglobalsupplyforecasts Adam Brandt Technical Report 6: Methods of forecasting future oil supply Steve SorrellandJamieSpeirs Technical Report 5: Methods of estimating ultimately recoverable resources Richard Miller, Steve SorrellandJamieSpeirs Technical Report4:Declineratesanddepletion Erica Thompson, Steve Sorrell and Jamie Speirs Technical Report3:Natureandimportanceofreservegrowth Erica Thompson, Steve Sorrell and Jamie Speirs Technical Report 2: Definition and interpretation of reserve estimates Jamie SpeirsandSteve Sorrell Technical Report 1: Data sources and issues follows: The results of the full assessment are contained in five in-depth Technical Reports, as Annex 2:Technical Reports Cover:Example_Design 23/09/2009 10:38 Page 2

An assessment of the evidence for a near-term peak in global oil production

A report produced by the Technology and Policy Assessment function of the UK Energy Research Centre

Steve Sorrell Jamie Speirs Roger Bentley Adam Brandt Richard Miller

August 2009

ISBN number 1-903144-0-35 Cover:Example_Design 23/09/2009 10:38 Page 1 An assessment of the evidence for a near-term peak in global oil production Global Oil Depletion

An assessment of the evidence for a near-term peak in global oil production UK Energy Research Centre September 2009 September Centre Research UK Energy UK ENERGY RESEARCH CENTRE

58 Prince’s Gate Exhibition Road London SW7 2PG tel: +44 (0)20 7594 1574 email: [email protected] www.ukerc.ac.uk

UK ENERGY RESEARCH CENTRE