Summary of the AAPG–SPE–SEG Hedberg Research Conference · PDF filereservoir...

AUTHORS

Susan Agar � ExxonMobil UpstreamResearch Company, Houston, Texas;[email protected]

Susan Agar is an advisor for emerging and disruptivetechnologies at ExxonMobil Upstream ResearchCompany. She directed the ExxonMobil-Academic(FC)2 Alliance for five years before this HedbergConference on the same research theme. She ob-tained her Ph.D. from Imperial College and has di-vided her research and development career equallybetween academic positions and industry. Her re-search interests include structural geology, geome-chanics, and flow prediction in fractured reservoirs.She also pursues interests in academic-industry-government collaboration.

Sebastian Geiger � Heriot Watt University,Edinburgh, United Kingdom;[email protected]

Sebastian Geiger is the foundation CMG chair for

GEOHORIZON

Summary of theAAPG–SPE–SEG HedbergResearch Conference on“Fundamental Controls onFlow in Carbonates”Susan Agar, Sebastian Geiger, Philippe Léonide,Juliette Lamarche, Giovanni Bertotti, Olivier Gosselin,Gary Hampson, Matt Jackson, Gareth Jones,Jeroen Kenter, Stephan Matthäi, Joyce Neilson,Laura Pyrak-Nolte, and Fiona Whitaker

carbonate reservoir simulation at the Institute ofPetroleum Engineering, Heriot-Watt University, wherehe leads the carbonate research group. He is also thecodirector of the International Centre for CarbonateReservoirs in Edinburgh, a joint research alliancebetween Heriot-Watt University and University ofEdinburgh. His current research interests includemodeling, simulating, and upscaling multiphase flowprocesses in (fractured) carbonate reservoirs, en-hanced oil recovery processes for carbonate reser-voirs, and studying the fundamental transport pro-cesses in carbonates from a pore-scale perspective.Sebastian received a Ph.D. from ETH Zurich andan M.Sc. degree from Oregon State University.

Philippe Léonide � Université Aix-Marseille,Marseille, France; [email protected]

Philippe Léonide is an assistant professor in carbon-ate sedimentology at the Aix-Marseille University(Centre Européen de Recherche et d’Enseignementdes Géosciences de l’Environnement [CEREGE],National Center for Scientific Research [CNRS],Institut de Recherche pour le Développement [IRD],Centre Européen de Recherche et d’Enseignementdes Géosciences de l’Environnement [CEREGE]UM34, France). He received his Ph.D. in sedimen-

A joint AAPG–Society of Petroleum Engineers–Society ofExploration Geophysicists Hedberg Research Conference washeld in Saint-Cyr sur Mer, France, on July 8 to 13, 2012, toreview current research and explore future research directionsrelated to improved production from carbonate reservoirs.Eighty-seven scientists fromacademia and industry (split roughlyequally) attended for five days. A primary objective for theconference was to explore novel connections among differentdisciplines (primarily within geoscience and reservoir engineer-ing) as a way to define new research opportunities. Researchareas represented included carbonate sedimentology and stra-tigraphy, structural geology, geomechanics, hydrology, reactivetransportmodeling, seismic imaging (including four-dimensionalseismic, tomography, and seismic forward modeling), geologicmodeling and forward modeling of geologic processes, petro-physics, statistical methods, numerical methods for simulation,reservoir engineering, pore-scale processes, in-situ flow experi-ments (e.g., x-ray computed tomography), visualization, andmethods for data interaction.

tology from University of Provence in 2007. He was apostdoctoral researcher on stratigraphical architec-ture in carbonates at Total/University of Provencefrom 2008 to 2009. He moved to the sedimentologyand marine geology group of John Reijmer at theVU Amsterdam University in February 2009. Hispresent-day research focuses on evolution of car-bonates through time, which have importance for thecharacterization of petrophysical properties in thecarbonate systems and reservoirs.

Copyright ©2013. The American Association of Petroleum Geologists. All rights reserved.

Manuscript received September 6, 2012; revised manuscript received January 14, 2013; final acceptanceDecember 17, 2012.DOI:10.1306/12171212229

AAPG Bulletin, v. 97, no. 4 (April 2013), pp. 533–552 533

Juliette Lamarche � Université Aix-Marseille,Marseille, France; [email protected]

Juliette Lamarche is an assistant professor in struc-tural geology at the Aix-Marseille University (CEREGE,CNRS, IRD, CEREGE UM34, France). She receivedher Ph.D. in structural geology from Paris VI Uni-versity in 1999. In 1999 to 2000, she taught structuralgeology at the Lille 1 University (France). She thendid postdoctoral research on three-dimensional basinmodeling at the GeoForschungsZentrum Potsdam(Germany) from 2000 to 2003. Her present-dayresearch focuses on fractured reservoir analogin carbonates and three-dimensional structuralbasin modeling.

Giovanni Bertotti � Delft University of Tech-nology and Vrije Universiteit Amsterdam;[email protected]

Giovanni Bertotti received his M.Sc. degree in fieldgeology at the University of Pisa (Italy). He thenobtained a Ph.D. from the Swiss Federal Institute ofTechnology developing a project on the tectonicsof the South Alpine passive continental margin.From 1991 to 2010, he has been working at VUAmsterdam on the tectonics of basins in Carpa-thians, Turkey, Morocco, and elsewhere. Since 2010,he also holds the position of a full professor inapplied geology at the Delft University of Technol-ogy where he is working on the geology of fracturedreservoirs.

Olivier Gosselin � Total/Imperial College, Lon-don, United Kingdom; [email protected]

Olivier Gosselin has been a principal reservoirengineer, with Elf and Total for more than 30 years.He is now a full-time visiting professor of petroleumengineering at Imperial College London. Hisresearch interest and expertise concerns mathe-matical and numerical methods applied to char-acterization, modeling, and flow simulation ofreservoirs, especially carbonates and fracturedreservoirs, and also identification of problems andassisted history-matching using four-dimensionalseismic data.

Gary Hampson � Imperial College, London, Uni-ted Kingdom; [email protected]

Gary Hampson is a reader in sedimentary geologyin the Department of Earth Science and Engineer-ing, Imperial College, London. He holds a B.A.degree in natural sciences from the University ofCambridge and a Ph.D. in sedimentology andsequence stratigraphy from the University ofLiverpool. His research interests lie in the under-standing of depositional systems and their pre-served stratigraphy, and in applying this knowledgeto reservoir characterization.

534 Geohorizon

MEETING FORMAT

The conference was organized into four thematic sessions onthe first two days (fundamentals, measurement and detectionof flow on laboratory to field scales, uncertainty and prediction,and novel modeling and simulation techniques); a field trip onthe third day was preceded by a dedicated poster session thatintroduced the geology of the area, whereas the ice breakerfeatured guest lectures on innovation and complex adaptiveleadership, as well as a panel discussion. Given the challengeof cross-disciplinary communication, delegates were encour-aged to adopt a beginner’s mind, challenging the status quoand exploring basic questions that the establishment mighthave overlooked. Stepping back and slowing down to pro-mote effective conversations among different disciplines wasemphasized upfront. Several delegates noted that technicaljargon was a significant barrier to novel thinking in the waythat it impeded effective communication among disciplinesduring the meeting. Cross-disciplinary interactions were en-couraged by several further mechanisms, representing a shiftfrom more common Hedberg Conference formats. Overall,the conference started with substantial guidance to promoteengagement. As the week progressed, the conference formatbecame less structured as delegates learned more about eachother and pursued the development of research ideas. For agroup of free-thinking research scientists, early deliberate or-chestration of interactions was an unusual experience. The ob-jective, however, was to ensure that delegates did not fall backon established connections, to promote new connections and toengage all delegates (particularly non-native English speakers).Some of the techniques used resemble methods used in OpenSpace Technology to promote self-organization in a frameworkof simple guidelines (Owen, 2012). Nick Obolensky (VthDimension) and Julian Birkinshaw (London Business School)helped to kick off the meeting with introductions related toinnovation. Birkinshaw discussed “Where Ideas Come From,”while Obolensky led sessions related to “Complex AdaptiveLeadership” and “Self-Organization.” Delegate feedback com-mented on the value of these nonscientific contributions, rec-ommending that similar efforts might be worthwhile for futureconferences. The techniques encouraged delegates not only tothink about the science and technology but also to consider howsituations and interactionswere impacting their ability to connectand innovate. Several of the approaches had been tested pre-viously through the ExxonMobil-Academic (FC)2 Alliance(Agar, 2009), but this Hedberg Conference provided an oppor-tunity to evaluate thesemethods in a large group of academic and

Matt Jackson � Imperial College, London, Uni-ted Kingdom; [email protected]

Matthew Jackson is total chair of geological fluidmechanics in the Department of Earth Science andEngineering, Imperial College, London. He holds aB.S. degree in physics from Imperial College anda Ph.D. in geologic fluid mechanics from the Uni-versity of Liverpool. His research interests includesimulation of multiphase flow through porousmedia, new techniques for reservoir characteriza-tion and modeling, and monitoring and inflowcontrol in advanced wells.

Gareth Jones � ExxonMobil Exploration Com-pany, Houston, Texas;[email protected]

Gareth Jones is a geoscientist with ExxonMobil. Hisresearch interests include carbonate reservoir char-acterization and modeling with an emphasis onprocess-based predictions of diagenesis. He has aB.Sc. degree from the University of London (RoyalHolloway), an M.Sc. degree in hydrogeology fromthe University of Birmingham, and a Ph.D. from theUniversity of Bristol.

Jeroen Kenter � Statoil, Bergen, Norway;[email protected]

Jeroen Kenter is a member of the Carbonate Pro-ducibility Group at Statoil Company in Bergen,Norway. He received his Ph.D. from Vrije Universiteitin Amsterdam. He worked as a postdoctoral re-searcher at Rosenstiel School of Marine and Atmo-spheric Science (RSMAS) in Miami, returned as ajunior researcher to Vrije Universiteit in Amsterdam,joined the Chevron Energy Technology Company in2005, and moved to Statoil in 2011. His researchincludes carbonate reservoir characterization and, inparticular, rock typing and upscaling workflows, andthe development of new geomodeling techniques andfast-track reservoir modeling and simulation.

Stephan Matthäi � Montan Universität Leoben,Leoben, Austria; [email protected]

Stephan Matthäi is chair of reservoir engineering atthe Montan University of Leoben, Austria, and aconsultant to the oil and gas industry. His previousacademic appointment was as a senior lecturer ofcomputational hydrodynamics at Imperial CollegeLondon, United Kingdom. He holds an M.Sc. degreefrom Tübingen University, Germany, and a Ph.D.from the Research School of Earth Sciences at theAustralian National University. Furthermore, he hasconducted postdoctoral research at Cornell Uni-versity and Stanford University in the United States,as well as the Swiss ETH Zürich. His publicationsrange from the formation of hydrothermal golddeposits to the upscaling of multiphase flow innaturally fractured hydrocarbon reservoirs.

industry researchers, many of them meeting for the first time. Apanel discussion, led by Professor Martin Blunt (Imperial Col-lege London), Donatella Astratti (Schlumberger), and BrodieThompson (ExxonMobil) then emphasized why innovation isneeded for carbonate reservoirs and what keeps the scientificcommunity from achieving it.

From the start of the week, delegates were asked to considernew research opportunities in the form of proposals for collab-orative multidisciplinary research involving academic and in-dustry representatives. A strong emphasis was placed on the de-velopment of boldnew ideas regardless of budget andpresent-daytechnical feasibility. The motto for the week was “Build yourspaceship to Mars!,” encouraging delegates to think beyond in-cremental developments on their current line of research and toexplore new and unfamiliar areas. During the first three daysof the conference, delegates could propose a potential “ven-ture group” by posting an idea or research direction and invitingothers to sign up. Essentially, the conference provided a marketplace to sell and buy into ideas with a view to consolidating alimited number of teams later in the week to discuss forefrontresearch proposals (Figure 1). Although many high-quality andinformative presentations and posters were presented, the em-phasis was very much on the future. To support the developmentof venture groups, 13 keynote presentations were delivered overthe first 2 days to introduce provocative thinking, novel research,and case studies related to each of the four half-day themes. Thepresentations were followed by poster sessions (∼15–20 postersper session). Nearly all delegates who were not giving keynotepresentations contributed a poster, creating a rich library of on-going research primarily related to geoscience, flow prediction,and production in carbonate reservoirs. Abstracts for the talks arenow available on AAPG Search and Discovery and were dis-tributed electronically to the delegates. In addition, many del-egates volunteered to make their talks and posters available inelectronic format after the conference via a secure Web site atHeriot-Watt University.

Posters (Figure 2) provided a starting point to identify con-nections and future research opportunities. Speed dates wereused at the start of each poster session as a means to identifyconnections. These 8-min exchanges required the poster pre-senter to deliver key research messages, needs, future directions,and opportunities. Some delegates noted that posters commonlyhave more writing on them than can be easily absorbed and thatthe focus on a few key points helped to drill quickly into whatreally mattered. After 4 min, the audience (that had self-organized around the posters in each session) was asked to pro-vide feedback on common interests and connections that they

Agar et al. 535

Joyce Neilson � University of Aberdeen, Aber-deen, United Kingdom; [email protected]

Joyce Neilson is currently a lecturer in carbonates andpetroleum geology at the University of Aberdeen.Since receiving her Ph.D. from Imperial College in1988, she has also worked for BP and as a consultant.Her research interests include the effects of diagen-esis and faulting and fracturing on reservoir qualityin carbonates.

Laura Pyrak-Nolte � Purdue University, WestLafayette, Indiana; [email protected]

Laura Pyrak-Nolte is a professor in physics at PurdueUniversity with courtesy appointments in the De-partment of Earth, Atmospheric and Planetary Sci-ences and the School of Civil Engineering. She re-ceived her B.S. degree in engineering science fromState University of New York at Buffalo, M.S. degreein geophysics from Virginia Polytechnic Instituteand State University, and Ph.D. in material sciencesand mineral engineering from the University ofCalifornia, Berkeley. Her research interests includeapplied geophysics, experimental and theoreticalseismic wave propagation, rock mechanics, micro-fluidics, particle swarms, and fluid flow through earthmaterials.

Fiona Whitaker � University of Bristol, Bristol,United Kingdom; [email protected]

FionaWhitaker is a senior lecturer in earth sciences atBristol University. Her research focuses on water-rock interactions in carbonate, evaporite, and volca-nic systems, integrating field studies of modern en-vironments with reactive-transport modeling. Shehas a B.Sc. degree in physical geography from theUniversity of Bristol and a Ph.D. on the hydrochem-istry of modern Bahamian carbonate platforms.

ACKNOWLEDGEMENTS

The conveners thank AAPG, the Society of PetroleumEngineers, and the Society of Exploration Geophysi-cists for their support of this Hedberg Conference.The following companies generously sponsored theconference: BG Group, Chevron, ConocoPhillips,ExxonMobil, Maersk Oil, Petrobras, Saudi Aramco,Shell, Statoil, and Total. We thank the GeosciencesResearch Program Office of Basic Energy SciencesU.S. Department of Energy Grant No. DE-FG02-12ER16287 for a generous award to support travelfor students, postdoctoral students, and early careerfaculty. We also thank Laura Wegener (ExxonMobil)for her substantial contributions to field safety, con-ference preparation, and networking data compila-tion. We thank Debbi Boonstra (AAPG) for all herefforts during the two years involved in conferencepreparation. Frederic Jan and the staff of the DolceFregate hotel provided excellent logistical support.

536 Geohorizon

could identify and ways that they might help. For some, this felt alittle different because their usual mode of interaction is to inter-rogate the science as opposed to identifying opportunities. Thefundamental science was still important, but delegates were beingasked to think on their feet about potential mutual gains. In theopen format poster session that followed, delegates wereasked to identify three to five posters that could offer op-portunities for research collaboration on a common theme. Asimple numbering system enabled the rapid acquisition ofdata andmapping of networks (Figure 1). Delegates agreed toshare information on connections to posters, although it wasrecognized that large differentials in the apparent popularityreflected might exist. It was emphasized that network maps ofconnections did not represent any evaluation of the scientificor technical merit of a given poster. First, delegates might noteven recognize the opportunities associated with cutting-edgeresearch. Second, to a large extent, connections tend to reflectthe interests of delegates instead of the opportunities for com-mercialization. Some biases also arose from the position ofposters in the meeting room and timing of sessions. Recognizingthat the process was far from perfect, the network maps stilloffered some early insights to potential hubs for venture groups.Several poster presenters also commented that the networkingprocess helped to identify novel connections that they had notpreviously considered. Following a spectacular introductoryboat trip along the coast between La Ciotat and Cassis on thefirst day of the conference, a mid-week field trip provided thetransition from orchestrated presentations and poster sessionsto more informal interactions. Led by the faculty and studentsat the Université Aix-Marseille, the trip focused on Cretaceouscarbonate outcrops nearOrgon and Cassis. The diverse group ofdelegates was introduced to analogs for Middle East carbonatereservoirs, examples of fault zones in carbonates, facies in var-ious ramp settings, outcrop fracture studies, and multi-kilo-meter–scale overviews of carbonate sequences.

The fourth morning of the conference provided additionaltime for delegates to review the venture group proposals and toconsolidate into 12 or less groups. Starting with more than 20proposals, self-organized mergers and acquisitions condensedfairly quickly down to just five venture groups (see below).Breakout discussionswith interim report outswere used to shaperesearch proposals by each venture group for the next day withfinal presentations delivered on the last morning (Figure 1).

The following section summarizes key points from each ofthe sessions including all keynote oral presentations and posters.To avoid unnecessary repetition, some posters are discussed in adifferent session from the one in which they were presented.

EMERGING THEMES RELATED TOFUNDAMENTAL PROCESSES

Joyce Neilson (University of Aberdeen) and Oliv-ier Gosselin (Total/Imperial College) led the firstsession on “fundamentals.” Key messages includedthe following:

• Future research needs to emphasize the funda-mental physics and chemistry controlling fluidmovements in carbonate rocks. Some of this re-search is currently limited by technical capabilities(e.g., imaging).

• Expansion of pore-scale simulation research isstimulating further discussion related to multi–scale-modeling approaches, more rigorous ap-proaches to upscaling, and improvements forreactive transport modeling (RTM).

• The creation and destruction of porosity involvefundamental processes affecting all carbonates.Significant advances are still needed to developreliable simulations of porosity evolution.

• Insights to first-order processes and their relation tooriginalmineralogy and stratigraphymayofferwaysto simplify porosity and permeability prediction.

• Knowledge creation related to the processes con-trolling stylolite formation and fracture aperturedevelopment is opening up paths to improvedprediction and representation in flow models.

• Many alternatives available for flow and mechan-icalmodeling tools that canoffer improvements onthe current state of industry technology exist. Astate-of-the-art summary would benefit many re-searchers (see also Session 4 below).

Methods to Predict and Control Wettability

Discussions emphasized the need for more researchon the fundamental physics and chemistry control-ling fluidmovements in carbonate rocks. Inparticular,presentations highlighted the relationship of wetta-bility to pore structures, how wettability varies withpore surface, and howwe can use this knowledge andsmart water to change the wettability of pore throatsto increase permeability. One approach discussed inMartin Blunt’s (Imperial College) keynote presenta-tion introduced pore-networkmodeling as ameansto study the impact of wettability and connectivityonwaterflood relative permeability. This approachinvolves computing flow through binarized images,solving the Stoke’s equation for slow viscous flow,tracing streamlines through the pore space, andmoving particles combined with a random displace-ment. Agreement between simulation results andexperimental neutron magnetic resonance (NMR)measurements reinforces confidence in the method.A further experimental study of wettability, thatof Kristian Mogensen and Søren Frank (Maersk),looked at different scales to constrain the impact ofheterogeneities. Their results highlighted the func-tion of surface reactivity and roughness on wet-tability and the potential to modify pore-throatroughness (and wettability) through carefully de-signed acid jobs. Surface electrical charge was pro-posed as another influence on the wetting behaviorof carbonates by Matt Jackson (Imperial College).He discussed the use of the streaming potentialcoupling coefficient as a way to probe the surfaceelectrical charge properties of carbonate samples

Figure 1. Example network of individual delegates (black dots)and their connections to posters in all four poster sessions for thefirst two days. The posters common to a given poster session arerepresented by dots with a common color (i.e., red, green, blue,or orange). The size of the circles represents the in-degree, that is,the number of delegates who identified a connection to a givenposter. These displays were used to provide early indications ofresearch interests and potential links that could underpin ven-ture groups. Data presentation by Laura Wegener.

Agar et al. 537

saturated with brine and crude oil. Themethod alsooffers a way to explore whether changes in surfacecharge and wetting state are responsible for im-proved oil recovery during controlled salinity waterfloods. The importance of being able to visualize di-rectly the distribution of residual phases in the rockpore space based on plug nanometer-scale studieswas discussed by Mark Knackstedt (AustralianNational University). His multiscale imaging ap-proach not only provides insights to recovery mech-anisms but also can support the development ofupscaling methods for flow properties from pore toplug to core scales. Masa Pradonovic (University ofTexas at Austin) also revealed submicron porositycharacteristics of carbonate rocks through her ionbeam microscopy studies (see more below). Anidentified challenge for the wettability theme wasthe need for a set of recommendations or guidelinesto represent relative permeabilities for (fractured)multiporosity carbonates with complex diageneticevolution. Furthermore, current image resolutionfor tomography imposes limitations on pore-scalevisualization andmodeling. This said, the potentialfor pore-scalemodeling approaches to advance RTMwas highlighted as a promising future research area.

Processes Controlling Porosity Development

The creation and destruction of porosity involvefundamental processes affecting all carbonates andrepresent a cornerstone of carbonate geoscience re-search. Without appropriate constraints for mod-

538 Geohorizon

eling parameters, simulations of porosity evolutionremain exploratory at best. Starting with the fun-damental assumptions used in numerical models,Simon Emmanuel (Hebrew University) discussedthe shortcomings of empirical rate laws of disso-lution for dissolution in heterogeneous carbonates.Suggested improvements have been developed basedon combined atomic forcemicroscopy and numericalmodeling studies of dissolution rates and mecha-nisms. Linking to comments on pore-scale modelingabove, these results are now being used to proposetechniques to integrate pore-scale heterogeneitywith reactive transport models. The large numberof complex interactions involved in diagenesiswerenoted as a key challenge for simulations. However,a contribution by Adrian Immenhause (Ruhr Uni-versity Bochum) emphasized the first-order controlsof originalmineralogy on dissolution (the aragonitevs. calcite seas issue). The recognition of carbonatemineralogy as one of a limited number of factorsthat exerts an overriding influence on diagenesismay offer a way to simplify diagenetic models. Inthe absence of robust simulations of porosity evo-lution, a tendency to fall back on common assump-tions (rules of thumband anecdotes) for porosity andpermeability development also exists.Whereas somemay argue that sequence boundaries tend to localizesignificant porosity and permeability development,Robert Goldstein (University of Kansas) presented adifferent view: reductions in porosity and perme-ability at sequence boundaries are also predictable onthe basis of depositional facies and thus can improve

Figure 2. Speed dating at the posters.

subsurfacemodels.Cementation as aprocess that candestroy overall permeability was further highlightedby the concept of diagenetic backstripping (RachelWood, University of Edinburgh) as a means toidentify diagenetic tipping points during the evolu-tion of carbonate reservoirs during burial that canbe linked to fluid migration histories. Discussionsextended to porosity and cementation in stylolitesand fractures. Einat Aharonov (Hebrew University)identified categories of stylolites; the mechanismscontrolling their evolution; and their impacts onlarge-scale strains, spatial distributions of porosity,and potential flow. A proposed upscaling approachalso offers a new way to incorporate core-scale ob-servations of stylolites into field-scale models. Theprocesses controlling fracture aperture developmentwere also addressed. As a key impact on fractureflow predictions, appropriate definitions of fractureaperture populations remain elusive. In responseto this challenge, novel approaches to evaluate theimpacts of stress and fluid reactivity on apertureswere shared by Derek Elsworth (Pennsylvania StateUniversity). His experimental results suggest thatboth the fluid chemistry and characteristics of thefracture surfaces influence whether the fractureaperture is likely to increase or reduce with time.The direction of flow along fractures can also im-pact whether their apertures decrease or increase.Gareth Jones (ExxonMobil) showed the strongimpacts of thermal gradients on dissolution andcementation rates in convective flow models (var-iable density fluid flow coupled with thermodynamicreactions) for nonmarine carbonates. Weak gradi-ents along strata have little impact; however in faultzones, stronger thermal gradients promote fasterdissolution when flow is directed up the fault butmore cementation when flow is directed down thefault. These thermal gradient effects have the po-tential to drive locally an order-of-magnitude dif-ference in permeability.Whereas each of the studiesabove focused on a particular aspect of carbonateporosity, RudySwennen (KULeuven) promoted theneed to integrate and coordinate studies of variouscarbonate pore networks to realize gains for bothacademia and industry. He highlighted the ad-hocnature of reservoir studies undertaken to meet in-dustry needs for the short term, the small size of ac-

ademic research groups, and the lack of a full spec-trum of required expertise in any given group. In hisview, the coordination of a consortium of multipleresearch groups to develop a structured databaseon this theme can offer significant advances interms of standardization of data and broader access.

Simulation of Flow and FundamentalGeologic Processes

In addition to direct investigations of the funda-mental physical and chemical processes that impactflow and rock properties in carbonate reservoirs, thefirst session introduced some of the challenges re-lated to the development of appropriate proxies andtheir representation at different scales in models(pore scale to full-field scale). Issues surroundingthe scaling rock and fluid flow behavior, as well asmultiscale modeling techniques, were raised com-monly. As an example of a way to capture thefundamental physics while reducing computa-tional cost, Masa Pradonovic (University of Texasat Austin) demonstrated a novel two-scale networkmodel to connect flow on pore and microporosityscales.Moving to a coarser scale, Jim Jennings (Shell)sought a way to simplify. He proposed a generalizedapproach for permeability averaging through theuse of power averaging. This approach providesflexibility to range between harmonic and arithmeticmeans, using exponents estimated by the Ababouconjecture and applying these, stepwise, to randompermeability fields containing different anisot-ropies. At an even coarser scale, Cedric Griffiths(Commonwealth Scientific and Industrial ResearchOrganization) addressed multiscale forward strati-graphic modeling as a way to gain insights to ap-propriate upscaling techniques for rock propertiesand to predict rock properties between wells.

Highlighting the numerous improvements need-ed to advance flow simulation for carbonate reser-voirs, Olivier Gosselin’s (Total/Imperial College) key-note presentation challenged delegates to pursueimprovements or alternatives to dual porositymodelsfor fracture-flow simulations. No resistance to thisproposal was voiced at the conference, and severalalternativeswere highlighted in subsequent talks andposters (discrete fracture-matrix modeling, lattice

Agar et al. 539

Boltzmann, cellular automata (CA), network (balland stick model), Langrangian, and continuous timerandom walk (CTRW) [see below]). Ensuing dis-cussion of discrete fracture network (DFN) modelsraised questions concerning the current state of geo-mechanical modeling as a means to generate frac-ture populations, their connectivity, and the valueof coupling fluid flow to geomechanics. Reinforcingthe need for a review of geomechanical modelingtools, Gosselin also highlighted new approaches thatare under development (e.g., incorporating fluid flowinto geomechanical simulations and attempts to in-corporate fracture propagation and realistic frac-ture aperture distributions in discrete fracture matrix[DFM] models). The problem, however, is that,while numerous geomechanicalmodeling efforts arebeing pursued, interested parties are challenged tofind the time and resources to evaluate all of them,and little consensus on the most promising researchavenues appears to exist.

Although many people are familiar with the ap-plication of DFN simulation approaches for fracturedcarbonate reservoirs, the application of the CTRWmethod,which is well established in the groundwatercommunity and has been used in physics since theearly 1970s,was less familiar tomanyof thedelegates.The CTRW approach has been applied in many dif-ferent fields, but examples of its application to sub-surface flow in hydrocarbon reservoirs are morelimited. In his keynote presentation, Ruben Juanes(Massachusetts Institute of Technology) introducedthe CTRW method as a novel way to look at anom-alous flow and particle velocity, emphasizing flow infractured porous media. In his example, the CTRWwas being used to validate a spatial Markovmodel ona lattice network that explicitly captures the multi-dimensional effects associated with changes in direc-tion along the particle trajectory. Potential applica-tions of this approach include forecasting and riskassessment of the drained reservoir volume or timeto breakthrough in fractured reservoirs directly fromDFN models. This approach avoids the need forlengthy dynamic simulations on conventional cornerpoint grids and thedifficult taskofupscaling theDFN.

A wide-ranging conversation on various aspectsof modeling permeated discussions throughout theconference. Comments reinforced the need to im-

540 Geohorizon

prove the integration of static and dynamic data andthe types of dynamic data needed to calibratemodelsfor a given scale with a given recovery process. Inaddition, the relative merits of more simulations andmore sophisticated simulations versus a move tosimpler approaches were debated. The fact that con-clusions from simulations may arise from an inputthat is not being tracked carefully was duly noted.Discussions also drew attention to the fact that res-ervoir simulationmodels step forward linearly in timewhereas much of the inherent flow physics operatesonmultiple time scales, that is, it happensmuch fasterthan a single simulation time step (and hence wouldbe represented incorrectly) or is much slower (andhence the simulator calculates that nothing happenedin part of the reservoir volume). This not only makesthe use of computational hardware inefficient butalso simplifies the flow physics, potentially leadingto erroneous results and production forecasts.

EMERGING THEMES RELATED TOMEASUREMENT AND DETECTION OFFLOW AND ROCK PROPERTIES

Laura Pyrak-Nolte (Purdue University) and MattJackson (Imperial College) chaired a session onvarious approaches tomeasure and detect flow androck properties fromplug to play scales. Key pointsfrom this session were as follows:

• Petrophysical experiments are revealing furthercomplexities in terms of rock and fluid impacts onacoustic signatures—there is much more to do.

• Petrophysical assumptions for clastic rocks do notnecessarily apply in carbonates caused in part bythe multiple scales of heterogeneities that exist incarbonates. Novel experiments are highlightingadditional insights to controls on distinct paths forporosity and permeability development in car-bonates and offer improvements for establishedpetrophysical parameters.

• An opportunity exists to pursue more subsurfaceexperiments to validate modeling (flow and seis-mic) and interpretation of seismic signatures, tolearn more about what happens between wells,and to measure flow properties directly.

• Linking different measurements over differentscales is a major challenge for monitoring anddetection of flow. Opportunities exist to developbetter tools for subsurface monitoring and bettertools for seamless data integration.

• Various inversionmethods and integration of allavailable data types (geophysical, geologic, andproduction and/or engineering data) can help toreduce uncertainties. No single approach cantell us what we need to know.

• We need to remember that operations also im-pact flow—it is not just geology and fluids.

INSIGHTS FROM NOVELPETROPHYSICAL EXPERIMENTS

The overlap in length scales of discrete elements incarbonate rocks together with changes induced byfactors such as stress, fluid content, and reactive fluidflowmake it particularly challenging to interpret thegeophysical signatures of flow behavior and to tiepetroacoustic responses to rock properties. In herkeynote presentation, Laura Pyrak-Nolte (PurdueUniversity) used results from time-lapse imaging ofAustin chalk samples to promote three key areas forfuture research on seismic wave behavior: (1) in lay-ered media in the transition zone between ray theoryand effective medium theory, (2) for layered sys-tems with two competing anisotropic sources, and(3) in layered fractured systems that are geochemicallyaltered over time. The experimental results high-lighted several issues related to the influence of dif-ferent rock and fluid characteristics on acousticproperties. Fracture-specific stiffness will change asa consequence of precipitation and reduction of afracture aperture,whereas the locations of fractures ina layered medium will impact interpretations of spe-cific stiffness. Changes in flow paths, fluid saturation(velocity dispersion), and fluid substitution (seismicanisotropy) will also influence the seismic response.

Further laboratory studies of petrophysicalproperties introduced several novel approaches toimprove the link between petroacoustic signaturesand the rock properties while highlighting the dis-tinct approaches required for carbonate rocks. Ef-forts to improve constraints on porosity and poten-

tial permeability were targeted by Elizabeth Bemer(Institut Français du Pétrole) via a micromechanicalmodel. By capturing microstructural characteristics,Bemer is able to compute theoretical velocities andcompare these with experimental petroacoustic mea-surements. This, in turn, enables inversion for op-timal parameters such as pore aspect ratio. Thelimitations of Archie’s law when applied to car-bonate rocks were further highlighted by two studiesfrom Geosciences Montpellier: using a synchrotronfacility to image connected porosity and percolationclusters, Charlotte Garing (Geosciences Montpellier)illustrated a flow dependency on the critical pore sizeconnecting the percolating network instead of theelectrical formation factor or tortuosity. Discussionreinforced the need to integrate three-dimensional(3-D) imaging with core-scale geophysical proper-ties as a consequenceof the fine-scale heterogeneities(below that of the integration volume of the bore-hole geophysical methods) and the strong influencethat these heterogeneities have on the hydrody-namic properties of the rock. In addition to poresizes, geometry, and connectivity, reactive surfacearea was introduced as a new controlling parameter.PhilippeGouze (GeosciencesMontpellier) presentedcontrolled dissolution experiments to show how thepore structure changes in different ways with dis-solution regimes (representing different rates ofdissolution). One consequence of this is the de-velopment of distinct porosity-permeability rela-tionships within different dissolution regimes. Theimportance of links between chemical and physicalprocesseswas further emphasized byTizianaVanorio(Stanford University) in a discussion of time-varyingelastic parameters. Laboratory-based time-lapse ex-periments with high-resolution imaging illustratedthe changes in P- and S-wave velocities resultingfrom injection-induced dissolution. By recognizingdistinct styles of porosity-permeability modificationfor each carbonate facies during dissolution, anopportunity to reduce the number of parameters tomodel permeability and velocity trends exists. Acase study from the Campos Basin, Brazil, also tar-geted ways to distinguish porosity and permeabilityassociated with different facies. Abel Carrasquilla(Universidade Estadual do Norte Fluminense) dis-cussed the integration of laboratory petrophysical

Agar et al. 541

measurements with conventional logs as a means toquantify permeability and classify electrofacies. Fo-cusing on ways to reduce the costs associated withNMR logs, the study attempted to simulate theNMRprofile through the use of other conventional logs andartificial intelligence (neural network approachesproved to work better than other methods). Resultsfor porosity were shown to be reasonable, but per-meability prediction was more challenging.

Seeing More

The issue of linking different measurements overdifferent scales was raised again in the context of di-rectly imaging faults and fractures or predicting theirpresence from bulk volume seismic attributes. Sam-pling, scaling, and resolution issues limit any ability tocross-validate seismically derived fracture attributeswith geologic observations of fractures. A keynotepresentation by Donatella Astratti (Schlumberger)introduced various seismic attributes and time-lapseseismic data as ameans to capture information on theconnectivity of fracture networks. Using a chalk res-ervoir example, she illustrated the need to differ-entiate distinct generations of structures caused bytheir significantly different impacts on flow and theirdifferent responses to stress. Integration of the pro-duction history with the comparisons of repeatedsurveys was used to link changes in fault images toqualitative interpretations of changes in fault-flowbehavior. A related keynote by André Revil (Colo-rado School of Mines) discussed time-lapse jointinversion of geophysical data as a way to reduce thenonuniqueness of the inverse problem. Using a com-bination of two inversion methods (active time-constrained and structural time-lapse inversion)to simulate the inversion of cross-hole data, Revilshowed the potential advantages for monitoringchanges in partial saturation during the produc-tion of oil from carbonate reservoirs. Recognizedbenefits were the reduction in spatial artifacts inthe tomograms relative to other inversion methodsas well as improvements for the use of time-lapseinversion of seismic and resistivity data performedindependently. Enru Liu (ExxonMobil) alsomade a strong case for further measurements toexamine the interwell space while ensuring a full

542 Geohorizon

understanding of the physics, limitations, andcomplementary nature of tools and techniques usedto acquire information on different scales. In an ef-fort to link laboratory data to field-scale seismic ve-locity variations and interpretations of fracture po-pulations, RichardGibson (TexasA&M)presenteda model for effective seismic velocities in mediawith isotropic or aligned fracture sets. His methodexpresses the stress dependence of fracture com-pliances to the increasing contact area of rough-surfaced fractures. This provides a way to representchanges in seismic anisotropy caused by variations instress fields while relating fracture distributions tochanges in seismic amplitudes. Discussion noted,however, thatwhereas the presence of fracturesmaybe determined from bulk attributes, the precise lo-cation of a given low-offset fault or large-openingmode fracturemay be needed to understand specificimpacts of discrete structures on flow. Ralf Opper-mann (OPPtimal Exploration and Development)addressed this challenge through new workflows forautomated fault extraction that integrate very high-resolution 3-D seismic image processing results withdetailed calibration. A key paradigm shift here isthe move from subjective interpretation to ob-jective measurements, which can highlight faultsin seismic data and decrease a reliance on sto-chastic approaches. He showed examples wherehigh-resolution fault extraction enabled the iden-tification of multiple seismic fault penetrationsin wells that were ground-truthed with image logdata and directly linked with productivity and/orsweet spots or unfavorable fluid flow effects (drillingfluid losses, water channeling, well-to-well short-cuts, and compartmentalization). Four-dimensionalground-penetrating radar (GPR) as a technique toimage fractures in near-surface settings were re-ported by Mark Grasmueck (University of Miami).One of his previous studies focused on the Solvayquarry (Cassis) that was visited during the confer-ence field trip. In a further quarry study, time-lapseGPR was used to show the impacts of deformationbands on near-surface flow of water. The presence ofbaffles caused the water to spread, but some defor-mation bands provided connections between strataon meter to decimeter scales. The potential use ofdiffractions for imaging fractures and karst was also

considered by Grasmueck. It was suggested that acombination of 3-D GPR and ray-born syntheticmodeling can be used to decipher the signatures ofunmigrated diffractions. Seemingly incomplete andasymmetric diffraction circles visible on time slicesactually contain dip information of crosscutting frac-ture systems. Diagenetic impacts on the elasticproperties of carbonates were also included. Usingseismic-scale examples of carbonate reservoir analogsfrom the Southeast Basin in France, Renaud Toullecand Francois Fournier (Université Aix-Marseille) pre-sented forward seismic models of depositional anddiagenetic heterogeneities. Following a program ofdetailed sampling and petrophysical measurements,their study shows that sequence boundaries andunconformities will not necessarily correspond tochanges in the seismic signal. Furthermore, a diage-netic overprint can generate nondepositional reflectorterminations and abrupt lateral polarity changes.

The Integration of Geophysical Monitoringwith Production Data

Several delegates soughtways tomaximize the valueof insights from seismic data through integrationwith other monitoring techniques and routine pro-duction data (e.g., more common use of cross-welltomography andwells thatmonitor above andbelowreservoirs).One approachproposed byMatt Jackson(Imperial College) discussed the use of spontaneouspotential (SP) in hydrocarbon reservoirs during wa-ter flooding to detect andmonitorwater encroachingon a well through the use of SP and electrodes in-stalled permanently downhole. The technique hasthe potential to detect increasing water saturationseveral meters to tens to hundreds of meters awaybut is still looking for developments of appropriatehardware and interpretation methods and a betterunderstanding of the coupling coefficients involved(these relate gradients in water phase pressure,salinity, and temperature to gradients in electricalpotential). In addition to novel monitoring tech-niques, considerable energy exists around the needto acknowledge the uncertainty in 3-D and four-dimensional seismic data and ways to reduce thisuncertainty through joint inversion with other geo-physical data. As a method to capture 3-D petro-

physical properties from inverted prestack seis-mic data, Andrew Curtis (University of Edinburgh)showed a neural network approach for fully prob-abilistic inversion techniques. A key advantage ofthis approach was an ability to represent the un-certainty associated with rock and fluid propertymaps derived from seismic (such as variations ineffective pressure, bulk modulus, density of hydro-carbons, random noise in recorded data, and thepetrophysical forward function) while realizing sig-nificant computational efficiencies.

Frequent calls for better tools to integrate all thedifferent types of data across multiple scales empha-sized the need for smoother mechanisms to updatemodels with monitoring and survey data. Moving onfrom geophysical insights, an interesting case studyfrom Saudi Arabia and Kuwait (Nicole ChampenoyandScottMeddaugh,Chevron)was used to broadenthe picture by drawing attention to the less com-monly recognized variables that impact fluid flow.These include well deliverability, historical opera-tions, completions, facility constraints, and reactivity.Champenoy and Meddaugh emphasized that, with-out appropriate measurements and monitoring, itcan be hard to determine which of these has themost impact. Moreover, these factors are not ne-cessarily at the front of a geologist’s mind whenconsidering controls on flow. A further case studyshared by Rick Wachtman (ExxonMobil) showedhow a comprehensive measurement and surveil-lance program combined with geologic modelingwas used on theMeans field residual oil zone to assessrecovery efficiency and potential flow streams. In thiscase, repeated simulations identified key factors suchas proxies for fractures, high-permeability leachedzones, and ratios between vertical and horizontalpermeabilities as requirements to obtain a match toproduction histories. Complementarymodelswereused to estimate fieldwide flow streams of a water-alternating-gas flood, providing an estimated ex-tension of field life by 20 yr.

Several delegates wanted to learn more aboutthe extent to which fracture and fault patternsare validated by modeling and/or monitoring data.The following presentations helped to fill in someknowledge gaps while highlighting limitations andopportunities to do more. In a more data-limited

Agar et al. 543

case study than the preceding examples, StephenSmart (Hess) emphasized the importance of earlyconceptual models to develop ideas for the 3-Ddistribution of fracture intensity. Subsequent in-tegration of robust data sets across various scalesand several iterations with reservoir performancedata were used to construct and refine a dual-porosity simulation model for offshore East Java.In an assisted history matching example, ArnaudLange (Institut Français du Pétrole) demonstratedthe use of connectivity information from produc-tion data to characterize seismic and subseismicfault networks. By examining possible correlationsbetween water breakthrough time and connectivity,Lange was able to identify the most probable faultnetwork realizations to match the production data.Given the sensitivity of flow simulation results todifferent fault network realizations, the method canhelp to focus on the most likely scenarios. ThomasFinkbeiner (Baker Hughes) provided insights tofracture-flow properties on production time scalesthrough geomechanical modeling of a carbonatereservoir. Key developments in this study empha-sized permeability changes associated with depletionand/or injection, fracture property variations (i.e.,weak vs. strong fractures), as well as the impact thatwould be predicted had all fracture sets been as-signed the same mechanical properties and stresssensitivity. It was noted that, in carbonates wherefractures may be stiffer and less stress (pressure) sen-sitive, stress impacts on production and injectionmay be far less pronounced relative to reservoir rockscontainingmore stress-sensitive fractures. The impactof a single fracture on well-test responses was ex-plored by Bander Al-Quaimi (Saudi Aramco). Nu-merical simulation (dual porosity), inspired by a realfield example, was used to generate a spectrum ofwell-test responses for different scenarios related toa fracture located between two wells. The resultsshowed the impact of permeability contrasts betweenthe fracture and the matrix in different layers as wellas the connection of the fracture to different layers.

Monitoring Flow on Local to Regional Scales

Whereas most of the discussion focused on pro-duction time scales, a novel contribution by Apollo

544 Geohorizon

Kok (Maersk) illustrated the concept of an “oil-on-the-move system” inwhichhydrocarbons are neitherstructurally nor dynamically trapped but still rep-resent viable accumulations as they continue to mi-grate. This work has supported the development ofa regional oilmigration atlas based on oil expulsion,vertical migration, aquifer flow, and residual oil sat-urations. By performing numerical simulations of oilmigration and comparing the results with knownaccumulations, several opportunities and potentialleads were identified in the Danish North Sea Chalk.

EMERGING THEMES RELATED TOUNCERTAINTY AND PREDICTION

Giovanni Bertotti (TU Delft), Gareth Jones(ExxonMobil), and Jeroen Kenter (Statoil) chairedthe third session on uncertainty and prediction. Keypoints from this session included the following:

• First principle and robust geologic concepts arelacking in reservoir models for reasons such as (1)poorly defined integration of geologic attributesand static anddynamicproperties (multiscale poresystem) and resulting conversion to rock types, (2)inadequate nongeologic geostatistical simulationtechniques and fear to deviate from hard data indata-poor scenarios and, (3) lack of techniques tofast trackmodel building anddynamic simulationofa wider range of models in a shorter period of time.

• Workflows need to identify early the function ofdiagenetic modification on static and dynamicproperties. Consequently, improved knowledgeof diagenetic processes and related spatial trendsas well as diagenetic modeling capabilities areneeded to reduce uncertainty in matrix char-acteristics and propertydistributions.Organizationof the fewexistingdata sets andaconcertedeffort toacquire new multiscale diagenetic and/or poresystem data sets will be required to validate modelcapabilities and realizations. Geologic databasescapturing depositional rock-type assemblagesfrom analogs, their spatial juxtaposition rulesand morphometric trends, will support the de-tection of diagenetic modification and help toconstrain pre-drill scenarios.

• Reservoir (or petrophysical) rock typing needs togo beyond basic rock classifications (e.g., textureand fabric) and incorporate many more geologicfactors (e.g., diagenetic attributes, certain fracturetypes, juxtaposition rules, and spatial trends)while integrating static and dynamic data.

• One size does not fit all—local (andwhenneeded,refined) models may still be needed to explainflow behavior even with substantial geologic dataand insights across a producing field. This is be-cause heterogeneity varies spatially and generallyincreases with data quality and quantity.

• Variations in fracture densities are unlikely to befully captured by well data or properly predictedfromanalogs. Fracturepredictionneeds to includean understanding of the evolution of mechanicalproperties as a function of primary depositionaland diagenetic factors. Mechanical modeling ofcarbonate rocks is still limited by the identifica-tion of appropriate mechanical properties to as-sign to models at different scales. Representa-tion of depth-dependent fracture mechanismsand the evolution of rock strengths during plat-form development provide examples of the typesof model improvements needed.

• Current geostatistical techniques and practicestend to obscure the relationships between geo-logic concepts and permeability distributions inreservoir models. Significant opportunities to gobeyond entrenched methods for geologic model-ing and to invest in new and innovative tech-niques andworkflows exist. In addition, a need fora wider range of models to be tested and/or othertechniques to fast track simulation exists.

• A clear need to take the art out of reservoir qualitypredictions and to develop more rigorous andconcept-driven workflows exists. Expert opinionsare rarely objective, but subjectivity can be good if itis recognized and used appropriately. The key is tobe aware of the factors influencing expert opinions.

Uncertainty in the Matrix

Several presentations and posters addressed a rangeof characteristics in different carbonate facies, topredict them and to capture key attributes in geo-logic models and flow simulations. As a way to

“take the art out of reservoir quality predictions,”DaveCantrell (SaudiAramco) issued the challengeto develop quantitative process-based tools thatwould allow the prediction of reservoir qualityahead of the bit. Based on a pilot in the Sha’aiba, heoutlined a multimodel approach to generate theinitial reservoir quality (forward stratigraphic model-ing) including environmental constraints (e.g., waterdepth, initial bathymetry, temperature, sedimentaccretion rates, and wind speeds) and the super-imposed diagenetic modifications (calibrated kineticcementation model). Although initial results for po-rosity were within two porosity units, the project hasnot evolved yet to the point where predicted dy-namic properties and trends can be contrastedwithsubsurface data. Related discussion reemphasizedthat the origin of multiscale carbonate pore systemsremains poorly understood and requires improve-ments through research on diagenetic modeling andthe development of guidelines.

Anita Csoma and Hesham El Sobky (Conoco-Phillips) developed the diagenetic theme further topredict anhydrite cementation of the karst system inthe San Andres Formation. When compared withdeterministic petrophysical methods and other sta-tistical approaches, a modular neural network meth-od proved to be superior for the determination ofanhydrite abundance. Given the potential impact ofcemented karst features on recovery, predicted vol-umes and distribution of anhydrite were used in ageocellular model to delineate anhydrite-filled karstnetworks via multiple-point geostatistics with cus-tomized training images. Two case studies providedfundamental observations related to the distributionand origin of dolomitized reservoir intervals and theircommercial significance. From Brazil, Mary RaigosaDiaz (Baker Hughes) focused on dolomites thatform the best reservoir units in the Sergipe subbasin.Detailed paragenesis identified the top of high-energycarbonate banks that were subjected to completedolomitization as the prime reservoir candidates.Reporting on the characteristics of a less commonlyencountered environment in carbonate reservoirs,Ray Mitchell (ConocoPhillips) pointed out that pro-duction from the Bakken petroleum system comesmainly from interbedded, mostly dolomitic carbon-ate intervals interpreted to be of mostly continental

Agar et al. 545

origin. The mixed siliciclastic and carbonate sedi-ments in the Three Forks Dolomite were depositedmainly by eolian processes with heterolithic bedding(dolomite silt and mudstone) formed during wetperiods. Further characterization efforts used sa-tellite images of modern isolated carbonate plat-forms (Philippe Ruelland, Total) to derive lateralvariations in environments of deposition to generatetraining images for multiple-point geostatistics. Di-rect sampling was used to develop facies modelssynchronouslywithmodels ofmatrixporosity beforediagenesis. Overall, this cluster of posters demon-strated the increasingly sophisticated use of data tocharacterize matrix properties, together with theeffective use of modern and recent carbonates toinform our understanding of the distribution ofcarbonate heterogeneities over a range of scales.

Mark Skalinski (Chevron) and Jeroen Kenter(Statoil) discussed several shortcomings in the clas-sification and use of carbonate rock types, includingthe need to incorporate diagenetic attributes andmodification; integrating multiscale andmultimodalpore types, including fractures; integrating dynamicdata; and the lack of appropriate geostatistical tools.Examples from Tengiz and First Eocene (Wafra)reservoirs were used to illustrate the application of anew workflow designed to optimize petrophysicalrock typing and the generation of carbonate reservoirmodels. Petrophysical rock types are defined as (1)the category of rocks characterized by specific rangesof petrophysical properties, (2) exhibiting distinctrelationships relevant for flow characterization, (3)identified by logging surveys, and (4) linked to geo-logic attributes like primary texture or diageneticmodifications. The objective of this approach is todetermine the petrophysical rock types that controlthe dynamic behavior of the reservoir while opti-mally linking the geologic attributes (depositionaland diagenetic attributes and their hybrid combina-tions) and their spatial interrelationships and trends.Michel Rebelle and Cecile Pabian Goyheneche(Total) also showcased an approach to integrate res-ervoir geology, seismic data, engineering, and petro-physics as a more sophisticated workflow for res-ervoir rock typing. JimMarkello and RickWachtman(ExxonMobil) showed a new sequence-stratigraphic–based reservoir architecture for the Lisburne field

546 Geohorizon

that was developed in the context of Late Pennsyl-vanian regional and global controls on tectonics, cli-mate, eustasy, ocean circulation, and geologic history.The improved framework helped to guide the con-tent of geologic models and simulations to achievereasonable performance matches. However, evenwith substantial geologic and production data, thesingle framework could not capture local differ-ences that impacted specific flow directions, con-nectivity lengths, and rates on the sector scale. Akey message was “one size does not fit all.”

Complementary outcrop studies of the Urgo-nian carbonate platform in southern France by Phi-lippe Léonide, Francois Fournier, and Jean Borgo-mano (Université Aix-Marseille) suggest that earlycementation influenced the preservation of tightand/or microporous units that compartmentalizethe platform vertically and laterally. An associationbetween the early diagenesis and major sequenceboundaries has been recognized. By combining pe-trographic, diagenetic, and isotope geochemistry,they have been able to identify links among pore-type distributions, micrite diagenetic patterns, andsequence stratigraphy in microporous-dominatedcarbonate reservoir analogs thatmay offer predictivecapabilities. A further example of outcrop modelingwas presented by Maria Mutti (University of Pots-dam) based on a Jurassic carbonate ramp inMorocco.In this case, the focus was the development of a geo-statistical database of geobodies and the choice ofappropriate statistical modeling algorithms to rep-resent the spatial organization of different hierar-chical scales of heterogeneity. A truncated Gaussiansimulation algorithm was used to represent deposi-tional environments because of the gradational andlinear trends observed between geobodies. However,the sequential indicator simulation was used forlithofacies distributions because of its flexibility inhandling spatially independent lithofacies elements.

Uncertainty in Fractures

Delegates continued to wrestle with long-standingissues related to the prediction of fracture networksand ways to capture uncertainty in their character-istics and distributions in the subsurface. BertrandGauthier’s (Total) keynote presentation focused on

the need to know more about fracture networksbetween wells and at the scale of a reservoir modelcell. Outcrop studies can complement informa-tion on fractures at a well by providing insights tothe factors controlling fracture populations, whichcan then underpin qualitative concepts or quanti-tative relationships. A detailed quarry study in theSoutheast Basin, France, used to construct a digitalfracture network, provided several useful lessons,including the following: (1) fracture data from wellsmay not really be hard data because they cannotcapture the full spectrum of variability in fracturedensities and (2) identified relationships amongone-dimensional, two-dimensional (2-D), and 3-Drepresentations of the same fracture network maysimplify the extrapolation of well data to 3-Dproperties in the subsurface. A broader evaluationof fracture populations across the Southeast Ba-sin of France was reviewed by Juliette Lamarche(University of Provence) and Bertrand Gauthier(Total). The study offered a departure from moretraditional mechanical stratigraphy, indicating thatgeographic position was more important for themechanical properties of the carbonates than de-positional facies, with early diagenesis potentiallylocking in mechanical differentiation of the rocks.Regional fracture patterns were also considered tobemostly unrelated to large-scale structural events.In contrast, sedimentologic controls on fractures werethe focus of Chris Zahm’s (Bureau of EconomicGeology) presentation. Nine vertical mechanicalfacies associations were linked between core andoutcrop studies of facies in transgressive- andhighstand-systems tracts. Both rock fabric and po-rosity were found to be key influences on rockstrength. The vertical mechanical facies associa-tions constrained a mechanical framework for sub-surface dual-porosity simulation models and ulti-mately supported a pressure match to well testsand fieldwide production. In another fractured car-bonate reservoir case study, Alex Assaf and RichardSteele (BG Group) addressed uncertainty in a se-verely heterogeneous carbonate field in NorthAfrica. They developed multiple models (fully com-partmentalized, fully open faults, and partly com-partmentalized) to explore a spectrum of scenarios.Further reductions in uncertainty were realized by

integrating pressure transient analysis and numer-icalmodeling of nearwellbore effects that providedcritical feedback and led to a geologically appro-priate history match. Michael Welch (Rock De-formation Research Ltd.) reported on his quest topredict fractures based on outcrop studies in chalk.Examples from southeastern and northeasternEngland provided insights to the larger structuralinfluences on the locations of fracture corridors andemphasized the way that rock strength (reflectingdifferent porosities in chalk) and pore fluid pres-sure will impact fracture failure modes (shear ortensile). Prediction of fracture populations in flat-topped carbonate platforms was addressed byGiovanni Bertotti (TU Delft). In this case, fracturegeneration scenarios (stress and mechanisms) wererepresented by first-order 3-D finite-element mod-eling. Key uncertainties included (1) the stressconditions that control the formation of stylolitesand transitions from mode 1 to mode 2 fractures,(2) the appropriate bulk mechanical properties fora platform-scale model, and (3) the difficulty of pre-dicting the number and dimensions of fractures.Important factors represented by this work werethe depth dependence of fracture formation andlarge sensitivities to assumed paleostress scenar-ios. A further geomechanical study related to asteep-rimmed carbonate platform was presentedby Vincent Heesakkers (Chevron). Two-dimensionalfinite-element modeling was used to represent step-wise carbonate platform development with appro-priate constitutive models to reflect the differentstrength of facies during synsedimentary fracturedevelopment. Based on studies of the Canning Basinand the Guadalupe Mountains, such models offerinsights to early fractures in large carbonate resourcessuch as Tengiz and Karachaganak. Dave Healy (Uni-versity of Aberdeen) shared insights to the variabilityof fault-zone properties based on outcrop analogyfrom Malta. The overall objective of this ongoing re-search is to constrain the natural statistical distribu-tions in all of the pore-system attributes, as well astheir spatial variation with respect to depositionalfaces and tectonic damage. In a related study, JoyceNeilson and Dave Healy (University of Aberdeen)showed how effective medium theory is being ap-plied to translate the frequency range from ultrasonic

Agar et al. 547

data from fractured rock to seismic scales. As such,this work supports a way to link the fracture porosityand fault properties in the Malta study to acousticsignatures and to determine how property variationsare manifested in petrophysical attributes. Based onthe preceding presentations, an interesting discus-sion developed surrounding the importance of pro-duction data as a way to provide a check on thevalidity of the initial geologic predictions and in-terpretations and, possibly, to identify their flaws.However, the time-lag between insights from pro-duction data and the development of a geologicmodelmakes such validation less feasible. A possiblesolution lies in the definition of proxies to signal thequality of themodel as early as possible. The soonera shortcoming in the model is identified, the lessdamaging are the consequences: fail fast!

Uncertainty, Statistics, and Modeling

Brodie Thomson (ExxonMobil) provoked the audi-ence by addressing the failure of carbonate reservoircharacterization and modeling to define the dis-tribution and continuity of permeability extremesand to represent our geologic concepts adequately.The current practice of geostatistical methods, heargued, tends to obscure the relationship betweengeologic concepts and the final (and noisy) perme-ability distribution in the model. The effects of aver-aging and stacking multiple geostatistical steps canobscure flow pathways, thin baffles, and many othersubtle geologic features (e.g., thin-bedded and mi-croporous intervals and stylolites). The presentationstirred considerable discussion, dividing the dele-gates into those who sought greater simplificationand those who sought more (appropriate) geologicinfluence or concepts in the model. More unifiedsupport developed around the need for a widerrange of models to be tested and other techniquesto fast track simulations. In addition, it was rec-ognized that this was an area of considerable en-trenchment and that significant opportunities tothink outside of the box exist.

Comments on the need to improve commu-nication and integration across groups of experts re-inforced the overall thinking behind the confer-ence. In a relatedposter presentation,AndrewCurtis

548 Geohorizon

(University of Edinburgh) drew attention to theinformation scale gap that exists as a result of thetools and approaches available for subsurface sam-pling. Given the large geologic uncertainties thatresult from this gap, the function of expert opinionswas reviewed, using examples to highlight a lackof objectivity that emerges because of group dynam-ics. Examples of the ways that opinions evolve in re-sponse to group dynamics have been tracked bysoftware during discussions and raise concerns forconsensus-driven outcomes. Expert elicitation, hence,is potentially a low-cost method to reduce overalluncertainties by improving the quality of how pre-vious information is obtained and parameterized.

EMERGING THEMES RELATED TO NOVELMODELING AND SIMULATION TECHNIQUES

The fourth session, chaired by Gary Hampson(Imperial College), Fiona Whitaker (University ofBristol), and StephanMatthäi (Montan UniversitätLeoben) addressed novel modeling and simulationmethods. Discussions returned to some of theinitial comments related to the simulation of fun-damental processes at the start of themeeting. Keymessages from this session included the following:

• Models can serve to integratedifferent data sourcesacrossmultiple scales, but techniques for upscalingacross several orders of magnitude in a singlemodel remain challenging. Multiscale models of-fer an alternative approach that allows significantfine-scale details to be capturedwhilemaintainingcomputational efficiency.

• Recognizing the caveats related to uncertainties inthe previous session, it was still emphasized thata large amount of data are available to pursuemodeling in a larger, more integrated, and stra-tegic way, with strong opportunities to link fieldobservations and hypothesis testing via numer-ical models and laboratory experiments.

• Many new (or less commonly used) modelingtools are available or on the horizon (discussedin this and other sessions). We need to developthe most effective ways to use them and to seekclever and more creative applications.

• An ability to compare different models throughstandardization approaches, to use common mod-els as a basis for further analysis, and to conductcollaborative research on common reservoirs andoutcrops can serve to increase the overall value ofmodeling.

• Outcrop studies are perceived to have waned inpopularity, but these still have a function to playin geologic modeling and flow simulations. Theyprovide low-cost opportunities to test out datahandling and modeling techniques for differentstratal and structural geometries. They can alsoprovide reasonable geologic scenarios and as-sumptions for characteristics that are not easilyconstrained by subsurface data (e.g., fracture sizedistributions and effective fracture permeability).

• Fracture-flow simulations would benefit fromguidelines to determinewhen fractures and similarsmall pervasive heterogeneities (e.g., stylolites andkarst) shouldbe explicitly representedversus beingimplicitly represented by effective properties.

• Fracture-flow simulations mostly ignore the im-pacts of fracture-associated diagenesis on sweepand fracture-matrix fluid exchange and struggleto assign appropriate aperture distributions. Fur-ther developments in RTM need to extend tofracture diagenesis as well as the matrix.

• Thecouplingofprocesses inmodels is recognizedasimportant but has yet to be fully realized (e.g., in-tegrated sedimentologic DFM-RTM geomechan-ical models).

• Developments in computational graphics andvisualization offer ways to truly interact with dataand models and provide opportunities to rep-resent the associated uncertainty.

• The essentials of geologic heterogeneity and evolv-ing flow patterns must be captured in a reservoirsimulation for better production forecasting;however, this is normally not achieved with thecurrent, industry-standard reservoir simulators.

• In light of the above,manygeologic and simulationmodels constructed using standard tools andworkflows are unnecessarily complex in someregard, simplistic in others, and their constructionis too time intensive to allow assessment of mul-tiple scenarios and uncertainty. New modelingand visualization tools can help to tackle these

issues, but their effective exploitation probablyrequires a shift in the mindset of the user. It iscommonly more useful to generate a suite ofsimple models that encompass different scenariosand uncertainty (while representing key hetero-geneities and flow processes realistically) than togenerate a small number of detailed models an-chored to a single scenario, which may fail to rep-resent key aspects of the system of interest.

SimulatingMatrix Properties Over Different Scales

Further reinforcing the need to integrate differentdata sources across multiple scales, Chris Nichols(Shell) focused on inputs for upscaling based on in-formation fromcore-plug towhole-core scales. Threecase studies were used by Nichols to show how dif-ferent data (core plug, logs, and core) can lead todifferent impressions of porosity and permeability.A key message here was the need to examine rocktypes in both the petrophysical and geologic space.This integration canhelp to determine approaches tohandle different types of heterogeneities for a givenrock type while shaping guidelines to upscale fromcore to log to cell scales. Michael Sukop (Florida In-ternational University) very effectively demonstratedhow dense data–driven variograms from boreholeimages of relatively young carbonates in Floridaappear to capture high-frequency stratigraphic cy-cles and can be used to generate 3-D volumes forborehole-scale lattice Boltzmann flow simulationsand thereby extend the scale of application of di-rect flow simulation. Limitations of readily avail-able geostatistical software to accommodate thecomplex variogram structure led to simulations thatoverrepresented the horizontal continuity and un-derestimated vuggy porosity. By expanding applica-tions of the lattice Boltzmann method to borehole-scale simulations of flow for vuggy carbonates (muchlarger than the usual pore-scale applications), Sukopconfirmed reasonable agreement with other exper-imentally derived estimates of permeability.

Advances for Fracture-Flow Simulation

Robin Hui (Chevron) introduced this theme witha keynote presentation on an in-depth sensitivity

Agar et al. 549

analysis using a DFM, where fractures and matrixare both represented in a dynamic model using anunstructured grid. The study highlighted the chal-lenge of applying appropriate numerical approachesto the simulation of flowon a geologically driven gridstructure. Whereas the DFM technology enablesthe inclusion of aperture and length-displacementscaling from outcrop-analog data, solid guidelinesto determine when to draw the line between ex-plicitly represented fractures and those representedby effective properties do not exist yet. Such deci-sions can be influenced by gridding and other soft-ware considerations and, also, by geologic rationale.Questions were raised concerning the value of run-ning a DFM as opposed to conventional approachesand if different business decisions would have re-sulted from using more traditional dual porosityand/or dual-permeability simulation approaches. Theneed for a comparative study to determine thesefactors was also discussed. Wayne Narr (Chevron)generated much interest in his work characterizingsyndepositional fracturing from the Devonian Can-ning Basin, features similar to those seen in the re-cent as well as reservoirs such as Tengiz but verydifferent from fracturing that occurs after burial.Supporting the value of outcrop-based studies forthe subsurface, the Canning Basin data compilationsof fracture sizes and their relationships to stratig-raphy have been shown to complement well dataand provide useful guidelines to constrain flow sim-ulations of the Tengiz field. A further case studypresented by Jim Sylte (ConocoPhillips) demon-strated how dynamic data have been integrated for aperiod of 25 yr to monitor the influence of fracturesand stylolites during waterflooding of the Ekofiskchalk reservoir. The duration of the study reinforcesthe value of continuing to reevaluate and integratedata with simulations as new technology brings fur-ther insights and as enhanced oil recovery projectspose new challenges. IrynaMalinouskaya (UniversitéPierre et Marie Curie) demonstrated the use of 2-Doutcrop data from a Jurassic carbonate ramp in Mo-rocco to calculate the 3-D tensor for fracture per-meability. The approach is being used to explore theimpact of different fracture network characteristicsoneffectivepermeability. In somecases, differences infracture populations will have a substantial impact

550 Geohorizon

on the effective permeability, but in others, the de-tails may not make that much difference. Ole PetterWennberg (Statoil) showed the implications for fluidflowof thedevelopmentof a cemented zone aroundafracture andpatchymatrix cement alone. Preliminaryresults indicate that cement distribution exerts a pri-mary influence on simulation outcomes and that thepresence of cement at the matrix-fracture interfaceshould be factored into history matching and upscal-ing efforts.The studyelegantly createdanticipation forthe Notre Dame de Beauregard outcrop, which wasseen during the field trip. Following previous sessionsin which fracture impacts were discussed, this sessionreinforced that the simplifications we commonlymake about the effect of fractures on flow are pro-blematic. This occurs not only on the reservoir scalefor reservoir characterization and flow simulation, butalso where fractures are represented in 2-D, ignoringfluid circulation in the fracture plane and incorpor-ating simplistic assumptions about fracture apertures.

Reactive Transport Modeling

A keynote presentation by Nicole Champenoy andScott Meddaugh (Chevron) discussed advancedmethods to characterize permeability heterogene-ity and to handle steamflood RTM. Based on 2-DRTMs, steam injection was predicted to drive cal-cite and brucite precipitation, dissolution of dolo-mite, and conversion of gypsum to anhydrite. Dis-cussions emphasized the need to anticipate potentialchanges in flow behavior, the need to understandthe physics of displacement, the need for technicalknowledge to guide the use of various softwaretools, and the impact of grid design. Opportunitiesto develop more sophisticated RTMs were illus-trated using examples of replacement dolomitiza-tion by FionaWhitaker (University of Bristol), whoalso explored some of the hurdles that need to beovercome to generate moremeaningful simulations.Such challenges include the feedbacks between de-positional and diagenetic sediment texture, perme-ability and reactivity, the relative importance of var-ious forcing mechanisms, and the sensitivity of anenvironmental system to changes in any of theseforcing processes. Scenarios for dolomitization werealso discussed by Conxita Taberner (Shell) in the

context of density-driven flow. Two-dimensionalRTM simulations were used to predict geometries(layered dolomite bodies vs. irregular fingerlikebodies) resulting from (1) hypersaline brine refluxand (2) thermally driven flow. Potentially beneficiallinks were evident within this cluster of posters, forexample, RTM simulations offering a route to gen-erate rules and to describe diagenetic geobodiesthat could usefully feed into geologic models.Enrique Gomez-Rivas (University of Tuebingen)proposed a crustal-scale mechanism for the emer-gence of a self-organized flow system that mayexplain the development of localized alteration suchas hydrothermal mineralization along fault zones.

Advances in Geologic Modeling, DataVisualization, and Interaction Methods

Contributions related to static modeling of bothmatrix and fracture properties introduced significantdevelopments. In this cluster of posters, GregoryBenson (ExxonMobil) exemplified the workflow forcollecting and interpreting data from detailed fieldstudies and LIDAR imaging to construct a geologicmodel of aMiocene outcrop in southeastern Spain. Inongoing efforts to ensure compatibility of various flowsimulation studies, Benson introduced a “standardproperty calculator” as a means to standardize as-signment of reservoir properties. Gary Hampson(Imperial College) demonstrated the principles andapplication of a pragmatic surface-based modelingapproach. The approach is still limited by severalfactors, including selection of an appropriate levelin the hierarchy, gradations in geologic character-istics, and the incorporation of fracture and dia-genetic heterogeneities. Nevertheless, the technologyoffers improvements related to the next generation ofunstructured mesh simulators. The surface-basedmodeling approach was applied by Peter Fitch(Imperial College) to a Jurassic ramp system to sys-tematically investigate controls on patterns of multi-phase fluid flow. In combination with experimentaldesign techniques, the objective is to develop insightsto the impact of heterogeneities on flow in carbon-ates as a means to support the prioritization ofeffort during geologic model construction. A com-pletely different approach presented by Claude-

Alain Hassler (Shell) used the numerically efficientcellular automata (CA) method. The techniqueprovides a way to incorporate simple diagenesis inreservoir models. Although the CA method iswidely used, its application to reservoir modelinghas been very limited. Perceived benefits includeimprovements on classic variogram-based meth-ods through the application of stochastic rules,straightforward conditioning to existing data, andcapabilities to represent complex geometries.

Providing a link to the field trip, JeanBorgomano(Université Aix-Marseille) introduced work in thecarbonates group of the University Aix-Marseilleand their research on generic learning from outcropsthat can be translated to relevance in the subsurface(e.g., correlation length scales and rules). Usingmainly Cretaceous carbonates in the Provence re-gion, the group has developed an impressive suiteof sedimentologic, petrophysical, LIDAR, and seis-mic data integrated across multiple scales. It wasemphasized that the required level of detail is notalways obvious at the outset of a study and, echoingpoints raised in the first session, opportunities to linkmultiple models at different scales exist. This pre-sentation also served to introduce several postercontributions by students and postdoctoral studentsat the University of Provence for a special sessionthat recognized their contributions to field-trip or-ganization and local logistics for the conference.

A strong contender for the most glamorouspresentation, Mario Costa Sousa (University of Cal-gary) shared novel approaches to visualization anddata interaction. Examples emphasized the need tointeract with data as opposed to simply observingthem, showing ways to tear apart or zoom into res-ervoir simulation model results and to construct 3-Dobjects from 2-D sketches. Discussions noted thatsuch impressive visualization methods still have thepotential to mask significant uncertainty and thatgeologists and engineers may be awed by the visual-ization, masking the underlying data; however, moreflexible visualization would actually help to querysimulation results more robustly. In addition, gen-eralists (and others) examining data by thesemethodswould benefit from simultaneous representations ofuncertainty. The concept of a Google Earth styleinterface for zooming in and out of models was

Agar et al. 551

proposed during the discussion as a way to link dataand models across different scales.

Venture Groups

Recharged fromthe field trip, delegates self-organizedto form five venture groups from the 20 venturegroups initially proposed. Each venture group com-prised 10 or more delegates representing differentdisciplines (geology vs. engineering) and background(industry vs. academia). For the next day, each ven-ture group came up with a well-defined researchproject with a clearly identified hypothesis, researchgoal, and research plan, including an idea for an end-product that could be rolled out to the industry.

The topics proposed were as follows:

1. Diagenetic and structural controls on flowmodels:The aim was to build a multidimensional matrixthat will allow isolation of variables driving theresulting 3-D distribution reservoir quality prop-erties useful both at exploration and produc-tion stages.

2. Geoprinting:Theaimwas todevelop3-Dprintingtechnology to create large-scale (tens of meters),integrated dynamic analog models for carbonatereservoirs to experiment with geology, fluid flow,geophysics, and geomechanics.

3. Disconnect between geology and reservoir char-acterization: The aim was to develop an app tofast track the creation and validation of 3-D res-ervoir models and test multiple flow and geologicscenarios.

4. Multiscale field experiment (from 10–6 to 102m2):The aim was to study flow processes over morethan eight orders of magnitude in length scale—from laboratory to field, including the excava-tion of the field site for reconstructing the 3-Dgeology—to revolutionize flow simulation toolsused for reservoir predictions.

5. Wettability engineering: The aimwas to develop aflexible toolkit that accurately predicts reservoirwettability at the pore scale and suggests the bestrecovery mechanism, based on the fundamentalphysics and chemistry, to increase productionfrom carbonate reservoirs.

552 Geohorizon

Crossing the Academic-Industry Divide

As noted above, this Hedberg Conference achievedmany of its success measures, defining promisingnew research directions that are being shared withthe global scientific community through this articleand a special conference volume anticipated laterin 2014. New connections were formed, and de-legates developed interesting ideas for collabora-tions and research proposals. However, there ismuch that remains to be done to strengthen in-dustry-academic collaboration. Discussion drewattention to the fact that industry-academic colla-boration may be limited by the extent to whichacademics are aware of routine industry applica-tions and the awareness of novel research advancesin academia by industry representatives. Theseshortcomings limited the abilities of delegates toidentify opportunities for research advances throughcollaboration. Several academics wanted to learnmore about the basic modeling assumptions used inindustry, the knowledge gaps, and the opportunitiesthat might exist to contribute fundamental geologicdata. Industry researchers needed opportunities tolearn about the assumptions andmethods implicit innovel modeling techniques and advances in funda-mental science. In the experience of the conveners,this is a problem that extends far beyond the re-search needs for carbonate reservoirs. There was acall to be more organized and united in our datacollection. Significant valuable data sets have beencollected through the years both by industry andacademia. Given this, it would be beneficial for thecombined industry and academic community toadopt a coordinated approach in which the datawould be more accessible and comparable, as perrecommendations of Rudy Swennen (KU Leuven).

REFERENCES CITED

Agar, S. M., 2009, The (FC)2 Alliance: An innovation portal forresearch on the fundamental controls on flow in carbon-ates: International Petroleum Technology Conference Pa-per MS-13175, Doha, Qatar, December 7–9, 2009, 9 p.

Owen, H., 2012, Open space technology: San Francisco,Berrett-Koehler Publishers Inc., 200 p.

Summary of the AAPG–SPE–SEG Hedberg Research Conference · PDF filereservoir...

Documents

Transcript of Summary of the AAPG–SPE–SEG Hedberg Research Conference · PDF filereservoir...