Copyright 2000, Society of Petroleum Engineers Inc.

This paper was prepared for presentation at the 2000 SPE Annual Technical Conference andExhibition held in Dallas, Texas, 1–4 October 2000.

This paper was selected for presentation by an SPE Program Committee following review ofinformation contained in an abstract submitted by the author(s). Contents of the paper, aspresented, have not been reviewed by the Society of Petroleum Engineers and are subject tocorrection by the author(s). The material, as presented, does not necessarily reflect anyposition of the Society of Petroleum Engineers, its officers, or members. Papers presented atSPE meetings are subject to publication review by Editorial Committees of the Society ofPetroleum Engineers. Electronic reproduction, distribution, or storage of any part of this paperfor commercial purposes without the written consent of the Society of Petroleum Engineers isprohibited. Permission to reproduce in print is restricted to an abstract of not more than 300words; illustrations may not be copied. The abstract must contain conspicuousacknowledgment of where and by whom the paper was presented. Write Librarian, SPE, P.O.Box 833836, Richardson, TX 75083-3836, U.S.A., fax 01-972-952-9435.

Abstract

Reservoir characterization and simulation modeling ofnaturally fractured reservoirs (NFRs) presents uniquechallenges that differentiate it from conventional, singleporosity continuum reservoirs. Not only do the intrinsiccharacteristics of the fractures, as well as the matrix, have tobe characterized, but the interaction between matrix andfractures must also be modeled accurately.

Three field case studies have been evaluated combining the“forward” modeling approach, typically used by geo-scientists, with “inverse” techniques, usually incorporated byreservoir engineers. The forward approach examines variouscauses of natural fractures and its’ associated properties (e.g.fracture spacing, height, stress distribution, etc.) while theinverse approach focuses more on the effect created by theNFR (e.g. decline analysis, material balance, productivity,etc.).

This study shows how a more powerful methodology iscreated, for the evaluation of naturally fractured reservoirs,when combining two techniques that have, historically, beenapplied in relative isolation.

Introduction

The development of reservoir modeling and reservoircharacterization for Naturally Fractured Reservoirs (NFRs)has lagged behind simpler matrix flow dominated rocksystems due to the practical difficulty in quantifying bothmatrix and fracture parameters. The complexities of,numerical and mathematical calculations have historically

constrained the development of NFR modeling1. This papershows a number of integrated field studies, which haveaddressed the difficulties in characterizing NFRs, and presentsa proven methodology to characterize and model them.

Reservoir characterization presents a unique challenge inNFRs because of:

1) the need to characterize the fractures as well as thematrix

2) the need to characterize the matrix-fractureinteraction.

Characterization of the fracture includes defining parameterssuch as inter-fracture spacing, length, orientation, porosity,connectivity, aperture and permeability. As well, it isimportant to include realistic areal and vertical heterogeneityin both the matrix and the fracture systems.

A fractured medium represents a highly heterogeneoussystem. Fluid transport and pressure dynamics cannot be fullyreplicated in a model using a homogeneous three-dimensionalsystem.

Recent work has emphasized the need to better characterizeheterogeneities in matrix properties. The same attention, if notmore, needs to be given to the characterization of fractureheterogeneities. Reservoir characterization is highly dependentupon the integration of skills from geologists, geophysicists,petrophysicists and reservoir engineers to an even greaterextent in NFRs than in conventional reservoirs.

The methodology presented ensures compatibility between thegeological and engineering models. Specifically, this paperwill show how NFR parameters were determined for the threefields in which this methodology has been applied.

Geological (Forward) Approach

On the geological side, there have been numerous attempts tocompile fracture statistics on spacing, fracture height,orientation, aperture and length, and to subsequently scalethese up to represent an effective fracture porosity andpermeability. This is often referred to as the "forwardapproach" as it characterizes the reservoir from perspective of

SPE 63286

Reservoir Characterization for Naturally Fractured ReservoirsRichard O. Baker and Frank Kuppe/Epic Consulting Services Ltd

2 RICHARD O. BAKER AND FRANK KUPPE SPE 63286

what caused or created the geological setting, as opposed tothe effects (usually the Engineers' preoccupation) of NFRparameters.

Fracture spacing, aperture, length and connectivity arefunctions of:

a) porosityb) lithologyc) structural positiond) rock brittleness.

One of the objectives of a fracture model is to generateempirical/analytical relationships incorporating fractureparameters and the above controlling factors. To do this, acertain amount of reservoir sampling is required tocharacterize the fractures. Unfortunately, lack of sampling orfracture characterization is a common problem in NFRs.

Outcrops are a good source of data for fracture length andconnectivity. Vertical wells have a low probability ofintersecting vertical fractures. Cored horizontal wells, drilledperpendicular to fracture systems, also provide insight intointer-fracture spacing (or length).

Thus, the relatively limited coverage from cored vertical wellsis unlikely to yield a sufficiently large sample. For many oftodays' developed fields, the data to correlate these variables isinsufficient. Engineering data can sample large numbers offractures (i.e., from the larger-scale pressure transient orproduction data) but this analysis cannot generate specificfracture parameters. Fracture parameters and empiricalrelationships must be derived from outcrop or field analogies.Despite the relatively small sample size, forward methods arecritical because they often are zone or area specific.

Although averages of fracture spacing and aperture are useful,as pointed out by Long2, there is ample field evidencesuggesting that only a few fractures are hydraulically activeand fluid flow may be dominated by extreme values of thefractured media.

Outcrop studies can provide valuable information on fracturespacing, length, direction and connectivity. However, asFriedman et al3 demonstrated, weathering and stress effectsmay affect parameters in the outcrops, making them differentfrom insitu conditions. The dimensions of matrix blocks arecontrolling factors for the recovery process. The inter-fracturespacing or determination of block volume distribution istherefore critical. Thus outcrop studies are important, despitethe limitations.

To increase the number of data control points andareal/vertical coverage, it is often necessary to use engineeringor "inverse" techniques. An inverse technique is one where thedynamic response (i.e. the effect) of the larger scale system ismeasured and is then used to infer smaller scalecharacteristics.

Tracer tests, historical production data and pressure transienttests provide us with the measurements, and the means, togenerate inverse solutions for fracture lengths, fracturepermeability and fracture connectivity.

Engineering (Inverse) Approach

On the engineering side, there have been attempts tounderstand the nature of the fracture systems usingpermeability and fracture storativity derived from well testsand production data.

Unfortunately the production and pressure transient data, istypically characterized with generic “sugar cube" models andother simplifying assumptions. Often reservoir heterogeneityis transparent to these tests and analyses. Most buildup(pressure transient) tests and long term production testing isnot zone specific and therefore have limited effectiveness forcharacterizing fracture heterogeneity.

There is a large volume of engineering literature dedicated tosolving the “dual porosity” pressure transient problem. Intheory, the results from a pressure buildup test can be used todetermine effective fracture spacing, however in 90% ofnaturally fractured reservoirs, pressure build up performancedoes not display dual porosity behavior4. Many NFR fieldexamples in the literature show that a simple single porositysystem can be used to obtain a reasonable match of pressurebuild up behavior5. The match is ultimately obtained with ahigher effective permeability, when compared to matrixpermeability.

The importance of fracture heterogeneity has been neglectedor minimized in engineering simulation studies. Without thisheterogeneity, reservoir simulation models often underpredictwatercuts (and water breakthrough time) and underestimatethe amount of bypassed oil in the reservoir.

In summary, both the geological approach and the engineeringapproach have limitations because of undersampling of insitufracture systems and the use of theoretical models that usesimplistic representations of real fracture systems,respectively. These approaches, independently, do not providethe means to accurately characterize the fracture system. Thisis a major problem in that long term deliverability andreserves are controlled by the effective reservoir permeability.Combining the forward and inverse approaches, however,allows us to narrow the range of uncertainty and build morerealistic models of NFRs.

Integrating Forward and Inverse Techniques

It is critical to identify, early in the reservoir study:1. what created the fracture system (i.e., regional,

faulted or bended environment)2. characteristics of the matrix-fracture system (i.e.,

fracture types, length, height, spacing etc.)

SPE 63286 RESERVOIR CHARACTERIZATION FOR NATURALLY FRACTURED RESERVOIRS 3

3. matrix permeability and porosity4. the degree of communication between the matrix

and fractures.

Determining how the fractures originated provides us withimportant clues for the areal and vertical distribution offractures as well as a likely reservoir recovery mechanism.Table 1 shows how the features of the NFR system varybetween the three different fracture systems and describes theassociated production implications. As with mostgeneralizations, there are exception to the "rules", indicated inTable 1.

The degree of flow between the matrix and fractures dictateswhich of many typical production problems may arise and alsodetermines the level of recovery that may be expected. We'veidentified four varying degrees, or "types", of flow and theirimplications in Table 2.

The two primary parameters that control recovery in NFRs are1) the magnitude and heterogeneity of fracture permeabilityand, 2) the extent of matrix-fracture communication. Thefracture permeability will control well deliverability whilefracture heterogeneity will control the extent of water/gasinflux. Good matrix–fracture communication is essential forlong–term productivity or high recovery factors.

Matrix-to-fracture communication is dependent upon the inter-fracture spacing as well as the matrix permeability. These twoparameters determine the strength of reservoir drivemechanism. If fracture spacing is small and/or matrixpermeability is high, a very efficient waterflood (imbibition)and gravity drainage mechanism develops. Conversely lowmatrix permeability and/or wide fracture spacing often resultsin the bypassing of the matrix by injected fluids or aquiferinflux and yields lower recoveries.

The productive permeability cutoff employed controls theoriginal hydrocarbon in place. From a mechanical perspective,the permeability provides a means of measuring the brittlenessof the rock and thus fracture intensity. The matrixpermeability therefore needs to be carefully considered in allNFR studies.

The key to successful fracture characterization is to (a) focuson key variables that dominate the recovery process and (b)use techniques that combine the more accurate, "micro scale",tests (e.g. core analyses), with large scale tests (e.g. pressurebuildups) that implicitly average the reservoir characteristicsover large volumes of rock.

Fracture-Matrix Interaction. Core analyses and loggingtools such as the formation micro-scanner (FMS) provide anestimate of matrix permeability and fracture spacing. Thisprovides us with sufficient data to at least estimate the“transfer function”, or the degree of matrix-fracturecommunication.

Alternative methods such as long term production decline datasupplemented by pressure transient data, helps us to confirmlarge-scale transfer function parameters.

Decline characteristics may provide important information onfracture volume, connectivity and permeability. Thehydrocarbon volume present in the high permeability fractureswill be produced rapidly. After this “flush oil” production therate will decrease rapidly before stabilizing at a lower declinerate. Fracture spacing and the amount of communicationbetween the fracture and the matrix, as well as the drivemechanism will control the stabilized rate.

Simulation Concerns. Depending, generally, on the contrastbetween matrix and fracture permeability and fracture spacing,the classical single continuum description may not beadequate for the simulation modeling of a fracturedreservoir. For theoretical analysis and reservoir simulation, theirregular fracture distribution must be replaced by a regularmatrix network (primary porosity) floating in the inter-connected fractures (secondary porosity continuum). Theidealized representation of fractures in the simulation model isshown in Figure 1.

It is assumed, in the fracture continuum approach, thatfractures are very long relative to the size of the blocks. Asshown by outcrops, fracture length, especially in regionalfractured reservoirs, can have limited length. Therefore, usingthe discontinuous fracture approach vs. a continuum approachin many NFRs may yield better results2.

Figure 1 is an illustration of the classical dual porosity model.Both the fracture and matrix have non-zero porosity andpermeability. Flow takes place within the fracture network andbetween the matrix and fractures. Each matrix block isassumed to be completely surrounded by fractures and cannotcommunicate directly with matrix adjacent blocks. This is notcompletely realistic because matrix blocks are "floating" whilein reality the fractured media supports rock stresses and allowmatrix blocks to touch. Most commercial simulators have theadded feature of including matrix-matrix connections (i.e.,dual permeability).

Warren and Root (1963) presented an analytical solution ofthe pressure transient based on the dual porosity model ofporous media. The key assumption was that the matrix tofracture flow is in pseudo-steady state conditions at all times(i.e., pressure declines uniformly throughout the matrix block).Fluid exchange between a matrix block and fracture wouldthen be given by:

( ) ( )mfmm

mmm PPk

t

Pcb

t−

µ∝−=

∂∂

Φ=Φ∂∂

(1)

4 RICHARD O. BAKER AND FRANK KUPPE SPE 63286

where Pm is the volume average pressure in a matrix block, Pf

is the pressure in the fractures surrounding the matrix block;cm is the primary (matrix) compressibility given by;

( ) wwworm ScS1ccc +−+= (2)

and a is the "shape factor", dependent on the size and shape ofthe matrix block.

The matrix blocks act as sources or sinks for the fracturesystem, according to equation 1, depending on the changes ofpressure in the fracture system.

Warren and Root obtained an analytical solution for singlephase, radial flow in an infinite and finite reservoir, withconstant well rate, as a function of the followingdimensionless parameter, otherwise known as the transferfunction:

eff

2Wm

k

rk∝=λ

where:

keff = effyeffx kk

rw = wellbore radius,

∝ = shape factor =

++

222 zyxL1

L1

L1

4

Lx, Ly, Lz = fracture spacing in x,y and z direction,respectively

This criteria for single porosity behaviour is usually supportedby pressure transient analysis (i.e., no dual porosity behaviourobserved). Engineers usually prefer, whenever possible tomodel a dual porosity reservoir with a single porosity model,capturing the effective permeability, because it halves thenumber of required gridblocks and shortens run time.Applying a single porosity model in a multi-phase enviromentmay, however, generate erroneous results. The breakthroughtime of the flood front, in a miscible flood or waterflood, isusually more rapid when "fingering" through fractures in thedual porosity model rather than the homogenized effectivepermeability of the single porosity model. Consequently thesustained higher production profile in the single porositymodel would decline more rapidly in the dual porosity model.

Care should therefore be taken to ensure the NFRcharacteristics not only satisfies single porosity, single phasecriteria but that there are no multi-phase consequences whenusing a single porosity simulation model.

Determining Fracture Parameters

As mentioned previously, fracture permeability, fractureconnectivity and fracture distribution, in water drives,waterfloods or gas cap drives, are critical controlling factors

for oil recovery. The key parameter, governing connectivitybetween injectors and producers, is the fracture permeability2.Unfortunately porosity–permeability cross-plots, derived fromcore data rarely has any significance for NFRs as it merelyrepresents the matrix properties. Also, conventional openholelogs are limited when predicting fracture distribution andfracture permeability. The key to successful fracture charac-terization is in developing empirical relationships that can re-late fracture spacing to porosity, lithology, structure position,rock properties or layer thickness as shown in Figure 2.

Initially, a theoretical geological fracture model should beconceived built on a combination of analog data, outcrop coreand FMS/FMI data. The primary objective in these initialfracture models should be establishing the empiricalrelationship between fracture parameters and porosity,lithology, structure and rock brittleness. These initialconceptual geological fracture models usually have limitationsin that fracture connectivity and individual fracture length arenot yet well defined.

Pressure buildup tests and production data can be used todetermine effective permeability but are very sensitive tofracture length and heights as well as connectivity. Combiningthe analysis from the conceptual geological fracture modelswith the engineering data allows us to estimate theseparameters. Production data and pressure data allows us todefine fracture connectivity.

Case Study 1 – The Weyburn Field

The first case study shows how old log data and productiondata were integrated to successfully characterize the reservoir.This work helped increase field oil production rate by morethan 50% using horizontal well technology.

The Weyburn field is located approximately 130km southeastof Regina, Saskatchewan, Canada. The productive portion ofthe field covers some 180km2 and has produced mediumgravity crude oil from the fractured, low permeability, Midalebeds of the Mississippian Charles formation since itsdiscovery in 1954,6

The Midale beds of the Mississippian Charles formation weredeposited on a shallow carbonate shelf in the Williston basin.The reservoir is informally subdivided into the upper Marlyand the lower Vuggy zone. The Marly is a chalky intertidaldolostone with occasional limey or limestone interbeds. TheVuggy zone is a heterogeneous, sub-tidal limestone. Althoughboth zones are fractured, the Marly zone is less intenselyfractured than the Vuggy zone.

The porosity of the Marly ranges from 16% to 38%, averagingapproximately 26%. Matrix permeability ranges from 1md toogreater than 100md, with the average being less than 10md.There is some contribution to effective permeability fromnatural fractures within the Marly.

SPE 63286 RESERVOIR CHARACTERIZATION FOR NATURALLY FRACTURED RESERVOIRS 5

The Vuggy is a substantial limestone, which is moreheterogeneous than the Marly due to the interaction of thedepositional environment and diagenetic over-printing.Carbonate sands (packstones and grainstones) were depositedin the higher energy shoal regions and carbonate mudstonesand wackestones in the quieter intershoal regions. Porosity andmatrix permeability vary substantially, ranging from 3% to18% and <0.01md to >500md, respectively. In addition, theVuggy is extensively fractured. Highly permeable carbonatesand bodies and fractures control the magnitude and directionof the permeability anisotropy in the Vuggy. Injectivity studiesalso show that the majority of the floodwater is injecteddirectly into this horizon.

Waterflood development in Weyburn began in the early1960’s. The waterflood has been very successful. Ultimatesecondary recovery factors ranged from 25% – 35%, based ondecline analysis. An extensive reservoir characterization studywas required to provide reservoir simulation models forwaterflood optimization, horizontal well evaluation and anassessment of miscible-flood potential.

A strong regional fracture system controls the behavior of theMidale beds under waterflooda6. It was therefore, important tocharacterize and quantify fracture system parameters such asfracture spacing, aperture, rock type, reservoir quality anddiagenesis.

The characterization study incorporated reservoir performance(i.e., production and pressure) as well as geological andpetrophysical data. Engineering data sources includedinjection profile logging, pressure transient and vertical pulsetesting. Geologic data sources included both vertical andhorizontal well core and wireline logs, repeat formation tester(RFT) and FMS logging. The neighboring Midale fieldproducing from the same reservoir has been extensivelystudied7, 8,9 and served as a valuable source of information.

Core and FMS observations indicate that the fractures arevertical to subvertical and are oriented approximately N45°E.Core observations clearly reveal that the forces causingfractures have been active on more than one occasion andhave generated at least three ages of fractures. Some fracturesare filled with anhydrite cement and are ineffective fluidconduits, whereas other fractures are very effective in movingfluids.

Initial geological studies showed that there were significantdifferences in fracture intensity between the limestone anddolomite zones. As well, there were large differences infracture intensity within limestone beds between shoal andintershoal areas. It is also interesting to note that fracture

a The Vuggy is most intensely fractured with fracture spacing of 1 ft inintershoal and 10 ft in shoal areas. The Marly has fracture spacing in the 3 ftto 10-ft range but is not fractured in some high porosity areas.

intensity decreased dramatically as porosity increased. Oldlogs could therefore be used as semi-quantitative indicators offracture intensity if the porosity and lithology could beestimated.

The relationship between lithology and fracture intensity wasvery useful when creating the fracture model. Particularconsideration was given to the more heterogeneous limestonelayer (Vuggy), as it was important to capture its heterogeneityto accurately model primary flow conduits.

Six simulation layers were selected to provide sufficientresolution for the Vuggy6. Previous simulation studies usedone to two layers: which smears or over-averages porosity,lithology and the dominant high permeability layers. Inmatching injection breakthrough performance, it is critical topreserve the high permeability; high intensity fractures zonesand not lump them together with lower permeability zones.Conversely, as indicated in early engineering studies, ashallow decline in production (5%-6% annual rate) wasgenerated with long remaining reserve life.

Watercut and breakthrough trends showed strong permeabilityanisotropy behavior. The watercut maps also showed a strongrelationship between the geological shoal and intershoal areasand the anisotropic behavior. This relationship meant that itwas important not only to look at dolomite/ limestonedifferences but to also look at the variations within thelimestone (shoal versus intershoal). Studies identifyinglithology and the geological environment therefore served asthe cornerstone of the fracture characterization work.

Estimates of fracture permeability had to be derived from thefracture spacing data (i.e., from core analyses) because of thelimited number of good quality pressure transient tests.Fracture permeability is dictated by fracture aperture andspacing. After having combined calculated fracture parameterswith the matrix parameters, the effective permeability could becompared with the insitu permeability determined form DST's.

Note that DST's have a limited radius of investigation. Toconfirm calculated frac parameters a history match of thepressure drawdown was important. The pressure drawdowndata was among the little data available to calibrate effectivepermeability.

The magnitude of the anisotropic ratio (ontrend NE-SW toofftrend (NE-SE) was a critical variable in determining theeffectiveness of the waterflood, horizontal wells and CO2

flooding. The anisotropy ratio in the simulation model wastherefore pre-conditioned in the simulator using waterbreakthrough times at respective production wells.

The project involved large potential oil reserves. Beforesimulation was initiated, six man-years of effort towardreservoir characterization had already been completed. Theworkflow diagram is shown in Figure 3. In our opinion, the

6 RICHARD O. BAKER AND FRANK KUPPE SPE 63286

key criteria to achieving successful characterization (i.e. onethat generated an excellent history match) included:

1. The ability to utilize open hole log response to estimatefracture spacing

2. The use of horizontal well data to get significant fracturespacing statistics

3. Using the ratio of water breakthrough times to pre-condition the permeability anisotropy ratio. This wassubsequently altered slightly to achieve a history match inthe Simulator

4. Pressure Transient data was combined with direct fracturedata to characterize the reservoir9

5. Selection of geological and simulation model grids thatprovided sufficient resolution for the intervals with largepermeability contrasts.

In the first simulation run, 70% of all wells were matched in a63 well model without any modification of the data6.Horizontal well watercut forecasts were within 3% of theactual values. Finally, it was found that the key parametersgoverning the waterflood and horizontal well recoveries, werevertical permeability and the distribution (and amount) of oilsaturation. This study is a case where geological (direct) dataand engineering (inverse) data was combined to get a betterrepresentation of the reservoir.

Case Study 2: The Spraberry Field

The Spraberry Trend in West Texas covers an area of morethan 400,000 acres10. The Spraberry Trend was once deemed“The largest uneconomic field in the world,” with reservoirsthat contained some 10 billion Bbls OOIP of which less than10% has been recovered today. The Spraberry Trend Areaproduces nearly 60,000 bopd from more than 7,500 wells andhas produced some 700 million barrels of oil11-16.

The Spraberry Trend Area was first developed in the early1950’s. The areal extent of the reservoir combined with manyhundreds of wells having initial production rates of greaterthan 500 bopd, led some to believe the Spraberry Trend wasone of the most prolific fields in the world. However, wellproductivity diminished rapidly as fracture depletion occurred.

The first waterflood in Spraberry began in 1956. Generally,waterflooding in the Spraberry area has not been successful.Injected water bypassed the matrix and did not effectivelysweep oil to producers10. Despite very similar fracture spacingbetween the Weyburn/Midale fields and the Spraberry field,there is a very large difference in waterflood recovery factorsbetween these fields. The incremental waterflood recovery inWeyburn/Midale is on the order of 16% to 25%, whereas theincremental waterflood recovery is only 2% to 5% for mostareas of Spraberry. Various hypotheses have been proposed toexplain why waterflood recovery is so low, including:

1. Lack of pattern confinement and low injection welldensity

2. Assumption that the primary direction of the fracturetrend is N50ºE throughout the trend, thus leading toincorrect pattern alignment in some locations

3. Low matrix permeability (Kair < 1 md), resulting in slowimbibition rates

4. The reservoir rock may not be strongly water-wet,resulting in low capillary forces and slow imbibition rates

5. Low reservoir pressures during the start up of waterflood,resulting in poor capture efficiency of oil as well as highinitial gas saturations and low oil permeability in thematrix.

Preliminary studies show that perhaps all the listedexplanations play some role in establishing ultimate recovery.

The differences in waterflood performance between theWeyburn/Midale fields and the Spraberry field highlight theimportance of not relying too heavily on analogous reservoirsto characterize the Spraberry NFR and that subtle differencesin fracture/matrix parameters can have a huge impact onrecovery factors, and the rate of recovery.

A number of tests have been completed to characterize theSpraberry reservoir fracture system. These include horizontalcore, pulse/interference tests, interwell tracer tests, buildupand falloff tests, FMI logging, outcrop studies, interwell tracerand mini-frac tests10. Like the Weyburn/Midale reservoirs,the direct approach (core/FMI) applications were very usefulin identifying zone specific fracture trends, whereas pressuretransient analysis was useful in identifying overall fracturepermeability and connectivity. Interwell tracer testsb, core dataand pressure transient analyses were used to identify fractureorientation.

Pulse tests were conducted in the Midkiff Unit, of theSpraberry field in the 1960’s. These tests demonstrated thatfracture permeability changed as the injection pressure andreservoir pressure increased18. The results suggest the effectivefracture permeabilities are in the 30md to 200md range. Crossfractures (offtrend E-W fractures) also seemed to have beencreated from the high injection pressures.

In this study, determining the extent of matrix-fracturecommunication was especially important in the evaluation ofthe CO2 flood efficiency. Since the degree of matrix-fracturecommunication is indicated by the waterflood imbibitionprocess, this was more extensively studied. Laboratoryimbibition experiments were designed to examine late stagedecline rates on waterflood. This correspondence was thenused to infer relationships between fracture spacing and matrixpermeability.18

b The interwell tracer tests in this area show very fast tracer breakthroughtimes (i.e., in the order of days).

SPE 63286 RESERVOIR CHARACTERIZATION FOR NATURALLY FRACTURED RESERVOIRS 7

Successful reservoir characterization in Spraberry dependedupon:

1. Use of horizontal well cores to calibrate fracture spacing2. Re-examination of matrix properties and imbibition data

that, coupled with field production data, promotedconsistency between performance data and derivedfracture/matrix properties.

3. Use of pressure buildup tests, interference tests,production data (i.e. watercuts) and outcrop data to inferpermeability anisotropy ratio.

Combining the direct with the inverse approach was critical.Quantifying the fracture spacing is important to the design ofwaterfloods and CO2 floods in a naturally fractured reservoir.The numerous studies completed have led to the generallyaccepted view that the fracture orientation in Spraberry is in anortheast to southwest trend. A more recent study examinedfracturing in a horizontal well core. The study revealed afracture set, located in the first layer, that was oriented N43° Ewhile in a lower layer, there were two fracture sets orientedN32°E and N70°E. The average fracture spacing of the threesets were found to be 3.2ft, 1.6ft and 3.8ft, respectively. Thisconfirms that more than one fracture direction can prevail andmust be incorporated into the model.

Case Study 3: The Waterton Field

The Waterton field is located in the southwestern corner ofAlberta at the front ranges of the Rocky Mountains and thefoothills disturbed belt. The field consists of a westwarddipping thrust sheet of Mississippian and Devonian carbonateswith hydrocarbons trapped along the leading edge. The pool isextensively fractured, especially along the crest of thestructure, and contains a near-critical, rich gas condensate witha compositional gradient19, 20.

The Waterton field was discovered in 1959 and was put onproduction in 1962 upon the completion of the initial phase ofgas plant construction. Twenty-four wells have been drilled inthe "Sheet III" reservoir with the last well drilled in 1977.Drilling and seismic operations have been restricted to narrowvalleys due to the rugged topography of the area.

Data from well logs, cores, drilling and production records,pressure transient analyses and reservoir engineering analyseswere used to characterize the variations in fracture and matrixproperties in the reservoir, as shown in Figure 4.

Conventional, tectonically induced fracture intensity wasfound to vary with lithology (dolomite versus limestone) andstructural position. However, such “conventional” fracturedescriptions did not account for the large drilling mud lossesin some zones, variable productivities (not correlatable tostructural position) and estimated volumes of initialhydrocarbon in place.

As with case studies one and two the key integation here wascombining the openhole log (geological) data with mud lossand AOF test data (engineering inverse data) to characterizethe fractures. In several key zones, solution enhanced fracturesor karst development was proposed to account for thesevariations. This is consistent with observations made frompressure transient analyses from most of the wells. This studyultimately confirmed the presence of a complex heterogeneousfracture system with extensive karsting in several intervals.This was confirmed through the course of history matching aswell as with pressure transient analysis.

Conclusions1) Forward and Inverse methods do not characterize the

fracture network sufficiently, when used in isolation,because fracture connectivity is unknown.Combining these two techniques provides a morepowerful complementary means of doing so.

2) Fracture heterogeneity is often over-simplified, or"smeared" in simulation models by using aninsufficient number of layers or gridblocks (as wasfound to be the case in previous Weyburn studies). Itis important to preserve the high permeability, highintensity fracture zones to accurately modelbreakthrough trends and ultimate recovery.

3) Accurate modeling of anisotropy, driven by overallfracture orientation, was critical in history matching,and predicting, waterflood movement, horizontal wellperformance and CO2 flood performance, inWeyburn.

4) Large scale tests (i.e., buildup and interferencetesting) and production data (i.e., water breakthroughand watercuts) were coupled with smaller scaleevaluations (i.e., outcrops and horizontal well cores)to generate a consistent (i.e., good history match) andaccurate model for predicting waterflood and CO2

flood recovery in Spraberry.5) The "Inverse" techniques of the pressure transient

analyses and matching production data (in reservoirsimulator) confirmed the presence of karstdevelopment and the associated enhancedpermeability, in the Waterton field.

References

1. Waldren, D. and Corrigan, A.F.: "An Engineering andGeological Review of the Problems Encountered inSimulating Naturally Fractured Reservoirs,” SPE paper13717, 1985.

2. Long J.C.: "Construction of Equivalent DiscontinumModels for Fracture Hydro-Geology," ComprehensiveRock engineer Principles, Practices.

3. Friedman, M. and McKiernan, D.E.: “Extrapolation ofFracture Data From Outcrops of the Austin Chalk I Texasto Corresponding Petroleum Reservoirs at Depth,”Petroleum Society of CIM paper 93-10-103, 1993.

8 RICHARD O. BAKER AND FRANK KUPPE SPE 63286

4. Fetkovich, M.J., Vienot, M.E., Bradley, M.D. and Kiesow,U.G.: “Decline Analysis Using Type Curves CaseHistories,” SPE paper 13169, 1984.

5. Carlson, M.R.: “Reservoir Characterization of FracturedReservoirs in Western Canada,” paper 97-87 presented at48th Annual Technical Meeting of the Petroleum Society ofCIM, June 1997.

6. Elsayed et.al.: "Multidisciplinary ReservoirCharacterization and Simulation Study of the WeyburnUnit," SPE, Oct 1993.

7. Beliveau, D., Payne, D.A. and Mundry, M.: “Waterfloodand CO2 Flood of the Fractured Midale Field,” JPT, pp.881-887, September 1993.

8. Beliveau, D. and Payne, D.A.: "Analysis of a Tertiary CO2Flood Pilot in Naturally Fractured Reservoirs," paper SPE22947 presented at the 1991 SPE Annual TechnicalConference and Exhibition, Dallas Oct 7-9.

9. Beliveau, D.: "Pressure Transients Characterize FracturedMidale Unit," JPT (Dec 1989) 1354, Trans., AIME, 287.

10. Schechter, D.S., McDonald, P., Sheffield, T. and Baker, R.:“Reservoir Characterization and CO2 Pilot Design in theNaturally Fractured Spraberry Trend Area,” SPE paper35469, presented at the SPE Permian Basin Oil and GasRecovery Conference, Midland, Texas, March 27 – 29,1999

11. Barfield, E.C., Jordan, J.K. and Moore, W. D.: “AnAnalysis of Large-Scale Flooding in the FracturedSpraberry Trend Area Reservoir,” JPT, pp.15-19, April1959.

12. Brownscombe, E. R. and Dyes, A. B.: “Water-ImbibitionDisplacement - Can it Release Reluctant Spraberry Oil?”Oil and Gas Journal, pp. 99-101, November 1952.

13. Elkins, L.F.: “Reservoir Performance and Well Spacing,Spraberry Trend Area Field of West Texas,” PetroleumTransactions of AIME, pp. 177-196,1953.

14. Elkins, L.F. and Skov, A.M.: "Cyclic Water Flooding theSpraberry Utilizes "End Effects" to Increase Oil ProductionRate," JPT, 1963.

15. Elkins, L.F. and Skov, A.M.: “Determination of FractureOrientation from Pressure Interference,” PetroleumTransactions of AIME, pp. 301-304, 1963.

16. Howell, W. D., Armstrong, F.E. and Watkins, J.W.:“Radioactive Gas Tracer Survey Aids WaterfloodPlanning,” World Oil, February 1961.

17. “Field Test Results of Surfactant Waterflooding andBalanced High Pressure Waterflooding in SpraberryMidkiff Unit,” Humble Oil Internal Memo, PRRCSpraberry Database, 1968.

18. Baker, R.O. et.al.: " Characterization of the DynamicFracture Transport in an Naturally Fractured Reservoir,"SPE 59690, March 2000.

19. Aguilera, R.: “Advances in the Study of NaturallyFractured Reservoirs,” JCPT, p.5, May 1993.

20. Nutakki, R. et.al.: "Three Dimensional simulation of aFractured Rich Gas Condensate Reservoir: a study ofWaterton Sheet lll.

21. Thomas, M.B. et al.: "A New Interpretation of FractureDistribution in Waterton Sheet lll: An Integrated ReservoirCharacterization Study," SPE 35605, May 1996.

Figure 1: Idealization of fracture reservoir accordingto the model of Warren and Root (1963).

Was initial geological model- compatible with engineering model

- good overall continuity

- relatively good waterflood

- recovery (RF> 30%)

- vertical permeability

- current oil saturation

- distribution of oil saturation

Model Construction

- determined fracture parameters from oil logdata- used ratio (ontrend vs. offtrend)of waterbreakthrough

times to give an initial estimate of

permeability anisotropy ratio

History Match

Weyburn Field Objectives:

Identify waterflood, horizontal well recovery

potential as well as examine CO2 potential

Initial EngineeringStudies

Initial GeologicalStudies

- Decline analysis

-(showed long reserve life)

- strong permeability anisotropic behaviordetermined from watercutsand breakthrough timing

- highest permeability in NE to SW direction

Fracture intensity is

a function of:- Dolomite vs limestone - shoal vs. intershoal porosity

A history match was successfully achieved onthe following critical parameters:- watercut- reservoir pressure- injection pressure- producing bottomhole pressure- saturation distribution at late stage of flood as

flood as measured by vertical wells- RFT pressures

buildup/DST derived permeability

Horizontal well projections, CO2 flood forecast

Key parameters identified forhorizontal well evaluation

waterflood

Figure 3: Workflow Diagram for Weyburn.

SPE 63286 RESERVOIR CHARACTERIZATION FOR NATURALLY FRACTURED RESERVOIRS 9

Figure 2: Data Flow for Characterization of Permeability for Natural Fractures and Reservoirs.

10 RICHARD O. BAKER AND FRANK KUPPE SPE 63286

Allocation of fracture permeability on layer basisand areal basis

Core/log/dataMud Losses

Production/injectionlogging

Seismicdata

Pressurebuildup andproductivity

index

Zonespecific

dataTotal effectivepermeability

Structuralchanges surface

maps faulted areas

Figure 4: Workflow Diagram for Waterton.

NFRParameters

Regional Fracture System FaultedSystem

BendedSystem

Areal distribution More uniform Very localized faulted zones Localized at crest and plungeareas

Vertical distribution Small vertical barriers oftenact to terminate fracturesystems

Shales and lithology changedo not terminate fracture set;very high verticalcommunication

Non-uniform higherfrequency in thinners beds

Sensitivity of fracturespacing to lithology

Very sensitive Not sensitive near fault Sensitive

Sensitivity of fracturespacing to matrixporosity/permeability

Very sensitive Not sensitive Sensitive

Fracture porosity Very smallφf < 0.1%

Can be moderate;Depends upon karstingφf < 5 %

Can be moderate; dependsupon karstingφf < 5%

Recovery mechanisms(limiting factors)

Some of these reservoirs canbe very successfullywaterflooded (see Beliveau1989)

Often wells water out;Water drive is common infaulted systems because ofhigh vertical fracturecontinuity.

Often act like multi-layeredreservoirs

Table 1: NFR System Parameters and Production/Recovery Implications.

SPE 63286 RESERVOIR CHARACTERIZATION FOR NATURALLY FRACTURED RESERVOIRS 12

Reservoir Type Problems and OpportunitiesType 1 Productivity essentially

derived from fracture porosity and permeabilityalone

- It is necessary to have high fracture intensityor high fracture porosity for an economicreservoir.

- May result in early water breakthrough thetiming of which is governed by fracture heightand vertical connectivity.

- Water influx is often accompanied by rapid oildecline.

- Fractures may generate production fromotherwise unproductive rock.

- Determination of fracture porosity is critical indetermining recovery.

Type 2: Fractures provide essentialreservoir permeabilityHydrocarbons stored in matrixand fracture but fractures provides the means (i.e. permeability) to flow

- Primary and secondary recovery efficiency ishighly dependent upon how well the matrix isexposed to the fracture network.

- Possible early water breakthrough and rapidoil decline.

- Development patterns must consider thereservoir heterogeneities (e.g. matrix-fracturecommunication may vary areally).

- Fracture intensity and dip must be knownbefore pursuing development.

- Fractures improve productivity from poordeliverability reservoirs.

- Determination of fracture permeability andheterogeneity is critical in accessing effectiveparameters and recovery potential.

Type 3: Productivity of a permeable matrix is enhanced with the additional fracture permeability

- There can be unusual responses in secondaryrecovery

- Drainage areas can often be elliptical- It may be difficult to recognize or detect the

fracture system- Fractures may enhance already commercial

opportunities- Determination of fracture permeability and

heterogeneity is critical (as for Type 2reservoirs.

Type 4: Fractures do not contribute toporosity or permeability, but barriers act as flow.

- Recovery is poor due to severe reservoircompartmentalization

- If properly planned, field development couldbe optimized

- Can have very poor secondary recoverybecause of compartmentalization

Table 2 – Reservoir Types (Adapted from Nelson).