Uncertainty In Subsurface Interpretation: A New Workflow

special topicfirst break volume 31, September 2013Reservoir Geoscience and EngineeringUncertainty in subsurface interpretation:a new workflowGarrett M. Leahy1* and Arne Skorstad1 explain a new interpretation workflow that focuseson the measurement of uncertainty and combines the interpretation and modelling phases.Current interpretation and geomodelling workflowsare ageing and, constrained by conventional wisdom,developed in an era where reservoirs were structurally less complex. In particular, the conventionalworkflow suffers from an overreliance on a single product: adeterministic reservoir model that is difficult to update anddoes not accurately represent the full spectrum of geologicrisk. In order to meet the increasing technical challengesprovided by the structurally complex reservoirs of today andtomorrow, we propose a new interpretation workflow thatfocuses on the measurement of uncertainty and combinesthe interpretation and modelling phases. This workflowprovides quantitative, risk-based decision-making support tohelp integrated asset teams achieve the best possible reservoirmanagement and recovery.The geosciences community faces significant challengesthrough the rise in increasingly remote and geologically complex reservoirs. However, most upstream organisations andworkflows are optimised around conventional oil paradigmsdeveloped in the 1980s and 1990s. Broadly speaking, thetypical practice in subsurface geoscience is for geophysiciststo handle the data and interpretation, and to hand off theinterpretation product (i.e., a configuration of subsurfacefaults and horizons representing their impression of thedata) to a geomodeller, who then integrates the interpretationproducts with other geologic data (well logs and picks, inversion products, etc.) to produce a reservoir model.The reservoir model is the central focus for decisionmaking in reservoir management – it provides a commonrepresentation of the reservoir to facilitate integration ofall subsurface disciplines and workflows (Figure 1). Thereservoir model can be used to compute a wide range ofquantitative decision support products, from volumetricestimates to simulation and production estimates.With increasingly complex tectonics resulting in morecommercially complex business decisions, the reservoirmodel has never been more critical – and it is becoming clearthat the current paradigm needs to evolve to meet the futureneeds of the oil and gas industry. Conventional interpretationand reservoir modelling workflows fail to meet these challenges for three key reasons.1*First, conventional interpretation and reservoir modellingworkflows are typically disjointed and independent withinorganisations. This leads to a constrictive lack of agility onthe part of the typical asset team, whether it is via a difficultyin adapting and responding to new data, or challenges correcting errors in existing data.Second, conventional workflows are geared to producinga single, deterministic, ‘best estimate’ model. This modelbecomes the quantitative support for all business decisionsfor reservoir management. This mode of thought becomes aliability when developing prospects that are challenging toimage, or when the tectonic setting (or style of faulting) ispoorly constrained.Third, and perhaps most fundamental, is that conventional workflows are not equipped to quantify uncertainty.Every piece of geologic data has an associated uncertainty(whether from resolution, sensitivity, or noise), and for goodor for ill, all of these uncertainties are carried into the reservoir model. As uncertainties in static reservoir properties(i.e., spatial description and volume) tend to be the key driverfor the economics of a prospect, this challenge is becomingincreasingly relevant to operators worldwide.Here, we describe a new workflow for subsurface interpretation and modelling to meet these and future challenges.We focus on the development and quantification of ‘measurement uncertainty’ associated with seismic interp field is calculated, modified and then used to updatethe horizon model. In this case, the new approach allowsthe displacement field to be modified by scaling, adding orsubtracting displacement. Changes can also be distributeddifferently on the hanging wall (HW) and footwall (FW)sides and uncertainty loops then run.Faults can vary within the fault uncertainty envelope andgenerate realistic structural scenarios. Furthermore, horizonscan also be extrapolated to ensure realistic results.Once you have an initial model, interpreters can thenupdate the model with new deviated wells. Figure 5, forexample, illustrates the initial structural model, the new welldrilled and the model then adjusted to include the new wellpath. In this way, the model-driven interpretation is ideal forQC’ing zonation. Other applications include simply how toincorporate uncertainty into the well planning process itself(e.g., Rivenæs et al., 2005).The most important input to reservoir modelling decision making and investment returns, however, is the abilityto quantify volumetric uncertainty. In this case, the inputdata includes a combination of velocity, isochore and faultuncertainties.The workflow consists of running horizon uncertaintymodelling (Abrahamsen, 1993, 2005) with time interpretation,Figure 4 A fault configuration implies a displacement field that can be inspected for matches withother data. Modifying the displacement field canthen be done, with the date being honoured anda direct and consistent update of the combinedfault/horizon model. 2013 EAGE www.firstbreak.org91

special topicfirst break volume 31, September 2013Reservoir Geoscience and EngineeringFigure 5 Using spatially varying uncertainty ininterpretations, an ensemble of structural modelscan be generated that satisfy well zonation data.Figure 6 Volume distribution for 50 realizations ofthe structural model (as shown in Figure 5). Bothrelative frequency for a given volume and a cumulative probability distribution can be obtained,allowing for quantitative, risked decision-making.Figure 7 Parameter sensitivities of GIIP for 50realizations of the structural model (as shown inFigure 5). Blue indicates positive correlation, andred indicates negative correlation. Uncertainty inthe gas-water contacts and net/gross are the mostimportant factors controlling reservoir volume.92www.firstbreak.org 2013 EAGE

special topicfirst break volume 31, September 2013Reservoir Geoscience and Engineeringvelocity, isochores and well paths. Depth convert faults can bedeveloped with updated velocities and uncertainty modellingthen run on these fault positions.The result is that simulated structural and grid models aredeveloped. Gas Initially In Place (GIIP) is identified and net-togross (NTG) uncertainty as well as uncertainties on contactsand porosity are all generated.In the Norwegian Continental Shelf example, 50 realizations of the structural model are run with changing faultpositions (Figure 5). Figure 6 shows the 50 gridded realizationsof bulk volume and Figure 7 the GIIP distribution.uncertainty can also lead to real-time risk assessments andhazard avoidance.In summary, the new approach places the model and riskanalysis at the centre of the decision-making process – whether itis bid valuations, new field development and operational plans,drilling programmes, or production estimates or divestments.ReferencesAbrahamsen, P. [1993] Bayesian Kriging for Seismic Depth Conversionof a Multi-layer Reservoir. In Soares, A. (Ed.) Geostatistics Troia ‘92.Kluwer Academic Publishers, Dordrecht, 385–398.Abrahamsen, P. [2005] Combining Methods for Subsurface Prediction. InConclusionsThe model-driven interpretation approach can have a significant impact on decision-making support in a number of areas.For example, the new workflow can generate a morecomplete representation of the data – no matter what the dataquality. By capturing uncertainty and building models directlyfrom the data, the new model-driven interpretation approachgenerates a more complete representation of the data in lesstime. Model-driven interpretation also does not allow poordata quality to cause unnecessary delays with the uncertaintystaying with the interpretation throughout the modellingworkflow.Similarly, the new workflow can provide early distributions of reservoir volumes. By increasing productivity andstreamlining workflows, model-driven interpretation allowsgeoscientists to quickly build risked models of static reservoirvolumes, providing the best possible estimates to supportcommercial decisions.In addition, the workflow can help to generate riskestimates for drilling decisions. Combined with logging-whiledrilling data and precision steering the captured interpretationLeuangthong, O. and Deutsch, C.V. (Eds.) Geostatistics Banff 2004,Vol. 2, Springer, Dordrecht, 601–610.Abrahamsen, P. and Benth, F.E. [2001] Kriging with InequalityConstraints. Journal of Mathematical Geology, 33(6), 719–744.Aki, K. and Richards, P.G. [1980] Quantitative Seismology. W.H.Freeman and Co.Bond, C.E., Gibbs, A.D., Shipton, Z.K. and Jones, S. [2007] What do youthink this is? Conceptual Uncertainty in geosciences interpretation.GSA Today, 17(11).Georgsen, F., Røe, P., Syversveen, A.R. and Lia, O. [2012] Fault displacement modelling using 3D vector fields. Computational Geosciences,16(2), 247-259, doi: 10.1007/s10596-011-9257-z.Rivenæs, J.C., Otterlei, C., Zachariassen, E., Dart, C. and Sjøholm, J.[2005] A 3D stochastic model integrating depth, fault and propertyuncertainty for planning robust wells, Njord Field, offshore Norway.Petroleum Geoscience, 11, 57–65.Stenerud, V.R., Kallekleiv, H., Abrahamsen, P., Dahle, P., Skorstad,A. and Viken, M.H.A. [2012] Added Value by Fast and RobustConditioning of Structural Surfaces to Horizontal Wells for RealWorld Reservoir Models. SPE Annual Technical Conference andExhibition, San Antonio, USA, Abstract, SPE 159746.PGS MultiClientRESERVOIRGeoscience expertise behind the dataWith access to our comprehensive MultiClient datalibrary, our G&G experts bring their deepunderstanding of the world’s hydrocarbon producingregions and reservoirs. Our data analysis togetherwith our seismic and quantitative interpretationexpertise has helped organizations qualify newleads and identify prospects, lowering theirexploration risk. We’re ready where you are; let PGShelp guide your next exploration venture.Supporting your exploration successglobalmc@pgs.comCC00660-MA128 PGS.indd 1 2013 EAGE www.firstbreak.org05-08-13 15:3293

the interpretation, and therefore that these workflows should be tightly integrated to provide ultimate flexibility to asset teams. We call the concept model-driven interpretation. Theory & Methods Interpretation can be a catch-all term that encompasses a variety of subsurface mapping and