RESERVOIR CHARACTERIZATION

Characterizing fracture and matrix heterogeneities in tight gas fields

Flow simulations using discrete fracture network models and full-field continuum modeling helped explain extremely variable well performance in the Waterton complex, Western Canada.

Keith Rawnsley, Martin de Keijzer, Lingli Wei, Solenn Bettembourg, Wenche Asyee, Jose-LuisMassaferro* and Peter Swaby, Shell E&P; Donna Drysdale and Dan Boettcher, Shell Canada Ltd.

Well performance is extremely variable in the stacked sequence of tight Devonian and Mississippian carbonates in the northern part of the Waterton complex of Alberta, Canada, despite an extensive fracture system present in all the wells. To determine why some wells penetrated more permeable fractures than others, a full reinterpretation of the geophysical, structural, stress, matrix and dynamic data sets was carried out at West Carbondale field in the complex. Flow simulations at sector scales using discrete fracture network models and full-field continuum modeling were used to test a range of geological and dynamic scenarios. For this field, the best-fit dynamic models consist of a major fracture zone, corresponding to either a seismic scale lineament or zone of enhanced curvature, trending through the area of most prolific wells. Outside this zone, the vast majority of the fracture system makes little contribution to flow in the wells.

BACKGROUND

The Waterton gas fields in southern Alberta represent the largest gas accumulations operated by Shell in the Canadian foothills. Gas is contained in a series of northwest-southeast-trending anticlinal structures within an imbricate stack of thrust sheets, Fig. 1.

The Waterton duplex developed as a result of thin-skinned thrusting and folding during the Laramide orogeny, which occurred from Middle Jurassic to Paleogene time.¹

Prior to the study, north- to north-by-northeast-trending fractures were determined to be associated with mud losses and therefore important for flow. Consequently, horizontal wells in West Carbondale and Castle River were drilled in the structural highs parallel to the crests of the anticlines, aiming to intersect many N-trending fractures. In West Carbondale (Fig. 2), vertical wells and horizontal sidetracks penetrate the West Carbondale Sheet III Devonian structure. Each of these wells, taking into account sample bias due to different well orientations, intersected a broadly similar high number of fractures (hundreds over a 1-km interval). The dynamic response, however, was extremely heterogeneous, with well test permeabilities varying by over three orders of magnitude. Thus, the challenge in these fields is not simply to intersect fractures, which are present wherever drilled, but to intersect productive fractures.

Devonian and Mississippian carbonates form the main reservoirs. Porosity in the carbonates is low, rarely exceeding 6%. Matrix permeability is correspondingly low (0.1 mD) and well productivity depends on the intersection of open fractures. The North Waterton fields, such as West and East Carbondale and Castle River, contain leaner gas, and production in these areas is hindered by elemental sulfur dropout. Depletion drive is the principal recovery process. Aquifer support is generally weak.

The objectives of the study were:

• To identify those reservoir parameters that control the short- and long-term productivity of existing wells
• To provide reservoir models that more accurately predict productivity for new wells and that assess the impact of the main risks within a range of uncertainties
• To provide recommendations for narrowing uncertainties and improving well success through such means as data acquisition and improved well design.
• Key parameters of the matrix and fractures were incorporated and tested against dynamic data to recognize the main uncertainties for the Waterton field.

MATRIX CHARACTERIZATION

To provide a framework for the reservoir modeling and fracture analysis, a detailed core study was conducted, both for the fractures and for the matrix. The result of the matrix study revealed the internal architecture of the main reservoir zone, the Crossfield Member, Fig. 3.

Although completely dolomotized, the reservoir could be subdivided into genetically related cycles. This allowed improved correlation of facies and petrophysical properties in each cycle and, in particular, identified a northward improvement of dolo-grainstone-capped cycles toward the north of the field in thin higher porosity layers.

Based on core measurements, the grainstones have average porosity values of 5%, the packstones and wackestones 3% and 2%, respectively, and the mudstones 0.5%. Other property values were also tested in simulation, as discussed below.

SEISMIC FAULT ANALYSIS

The main objective of the seismic evaluation was the visualization and definition of subtle features in the field that could be used to explain the difference in well behavior. The complex geological setting (large horizontal and vertical velocity contrasts) and the high mountains (terrain effects) make seismic acquisition and processing a challenge. Before any interpretation was carried out on the seismic volume, a structural oriented noise filter was applied to allow smoothly picking events that served as the base for attribute mapping.

The resulting interpreted lineaments consist of mainly NNE-SSW- and NNW-SSE-trending lineaments, Fig. 4. One feature, the “main lineament,” was consistently present on all the attributes, striking NNE-SSW. This lineament intersects the area of the best producer, CB3-09, and can be seen in the background of Fig. 2.

Despite the existence of these lineaments in the seismic data, the poor seismic quality prevented definite confirmation that these are not seismic artifacts.

FRACTURE CHARACTERIZATION

Fracturing, as interpreted from borehole images, is intense throughout West Carbondale. Typically, multiple fractures are intersected per meter by the horizontal wells, and multiple fracture orientations point to a geometrically well-connected fracture system, Fig. 5. The most productive well, CB3-09, is not the most intensely fractured.

The seismic top map in the vicinity of CB3-09 appears to be more disturbed with both axial parallel and north-south-trending “escarpments.”

Curvature analysis of the top structure map shows that this well is located in the area of highest curvature when compared with the less productive wells to the north and south. Fractures sub-parallel to the crest of the anticline are present in all the wells.

Because all the wells are positioned in or near the crest, it is not possible to determine if the axial parallel fractures are also present away from the crest. In both core and borehole images the presence of fractures was noted in all lithologies. In the core, all of which is from vertical wells, a slight increase in fracture frequency could be noted in the finer lithologies, mostly associated with small, closely spaced, cemented fractures. A fracture frequency track is displayed in Fig. 3. No clear relationship was observed with bed thickness. Borehole image data was compared with the loss zones during drilling. Despite large numbers of electrically conductive fractures, only a small subset of these corresponds to losses.

The effective fracture permeability and any variation between layers are explored further in the reservoir engineering history matching, described below.

IN SITU STRESS ANALYSIS

An evaluation of the in situ stress field using the borehole image and caliper data set was conducted. The key results (Fig. 5 inset) suggest that the maximum horizontal stress is oriented NNE, which is parallel to the main lineament trend. The in-situ stress field may provide information as to which structures may be expected to be preferentially reactivated and/or opened up, and to have enhanced permeability.

ANALOG CONSTRAINTS

To better understand the spatial distribution, structural styles and scales of fractures and faults, an analog outcrop study was initiated in the Sawtooth Range, northern Montana. The Montana outcrops have a comparable stratigraphy and structural setting, with Mississippian and Devonian carbonate rocks comprising a regional duplex of imbricate thrust faults and thrust-related anticlines.² Various compressional anticlines were examined, including the Swift Anticline. The Swift Anticline is a kink- or chevron-type fold, with a relatively narrow (about 20-30 m) hinge zone, a shallow dipping backlimb (about 8-15° WSW), and a steeply dipping forelimb (ENE-dipping 70-85°). A combination of helicopter and ground-based photography enabled mapping of a wide variety of fracture scales. In all parts of the fold, fractures occur, including dip parallel, axial parallel and oblique. In the forelimb, the fractures are locally reactivated with shear displacements. Localized zones of intense fracturing occur in the oblique lineaments at a range of scales. There is also intense axial fracturing in the hinge zone.

As an analog outcrop, the Swift Anticline has provided valuable information supplementing the limited subsurface data in terms of geometry and connectivity; types and scales of fractures and faults; and the geometric evolution, reactivation and resulting fracture patterns.

INCORPORATING PRESSURE TRANSIENT DATA IN SECTOR MODELS

Five wells in West Carbondale have been tested for flow after drilling. These wells are, from north to south: CB7-20, CB11-09, CB3-09 (horizontal sidetrack of 11-09), CB7-03 and CB13-35 (horizontal sidetrack of 7-03). Some wells have repeated pressure buildup tests after production was started. The test results from West Carbondale (Fig. 6) show more than three orders of magnitude of average permeability variation over the total gross reservoir thickness, which is similar across the structure. Combined with the fact that none of the well tests shows simple radial homogeneous flow, the West Carbondale structure can be called “extremely heterogeneous” in terms of flowing features.

Other than CB11-09 and 3-09, the tests were consistent with background matrix permeability. To better assess the consistency of the fracture description work in terms of well test responses, discrete fracture models were built and then transferred to the single-phase flow simulator MaficOil, where simulations were carried out to test different scenarios and assess those fractures that were effectively flowing in the reservoir. From the study of log images, mud loss data and regional geology, fracture sets were created, Fig. 7. In all, five different sets of fractures were generated. The first three sets were generated from the image analysis, and the two remaining sets were generated from the seismic attribute map at the scale of the seismic lineaments.³ By simulating well tests with different combinations of fracture sets and comparing the simulation results with the actual well test data, the seismic scale features were found to be the conductive features at Well CB3-09. In these models, the well is crossing a “fault zone,” modeled as a finite-length fracture that intersects the well. This model provides the best fit for CB3-09, Fig. 7 inset. The “steep” slope at late time in the derivative is explained by the fact that the well sees the end of the fracture zone. The orientation of the fault zone could not be determined from the well test simulation, and although illustrated to be parallel to the main lineament, equally good fits can be obtained with the fracture zone parallel to the crest, or any other orientation. Most conclusively, the surrounding fractures are contributing significantly less to the well test response.

It was concluded from this phase of the study that the major contribution to flow in the wells was linked mostly to large-scale features, either corresponding to the seismic lineaments or to zones of highest curvature and enhanced fracture flow properties. From this perspective, the key uncertainty-and opportunity-was to determine which of the mapped seismic lineaments or curvature zones could also contribute to flow, other than those at CB3-09.

FULL-FIELD SIMULATION

To further address the question of which large scale features in the field could be contributing to flow, and how much contribution was coming from the smaller scale fractures and matrix, the full 20 years of production and pressure data were also analyzed and used to constrain the reservoir modeling.

The pressure history measured from static pressure gradient measurements, pressure buildup tests and interference data from 1979 to 2001 is summarized in Fig. 8. Of particular interest is the pressure response to production from CB3-09. The well to the north sees relatively good communication (as recorded by a downhole pressure gauge), whereas the well to south sees little pressure decline.

In the simulation models, the reservoir is modeled in terms of matrix and fractures. Small-scale fractures (including micro-fractures) are regarded as part of the matrix, and the highly permeable fracture zones are represented explicitly in the dynamic models. A geological scenario for each run contains three components: a matrix porosity model, a matrix permeability model and a fracture zone model. Each component has several alternative models. Each geological scenario is formed from different combinations of these components.

The matrix model has 20 layers following the small-scale cyclicity. Each layer has up-scaled porosities based on the facies porosities given in the “matrix characterization” section above, with sensitivities around these. The main north-south facies trend is reflected in the model.

The matrix permeability model assumes no poroperm relationship, and that the matrix permeability is approximately that determined from the matrix support in the transient pressure analysis. It is a bulk value comprised of the matrix and any small-scale fracture enhancement. In the north of the field, the bulk “effective” matrix permeability was given a value of 5 mD; in the center 1 mD was used and in the south 0.05 mD, with sensitivities around these, Fig. 9. The history matching suggests that there is no strong variation in effective matrix permeability between the layers.

Figure 9 shows an overview of the fracture scenarios studied. Red lines are high permeability, blue lines are moderate permeability and black lines are barriers. The fracture zone models are derived from the reservoir data evaluation; the outcrop studies and the well test simulation results. A number of scenarios were generated to capture and test the range of possibilities.

Model 1: Main lineament continuous. This model assesses the possibility that only the main lineament is significantly more permeable (14 mD from well test results) than the rest of the reservoir. Smaller structures/seismic zones such as the north-trending seismic fault intersected by CB13-35 are considered ineffective for flow, and are included in the matrix/background permeability.

Model 2: Main lineament discontinuous. The primary difference is that in model 2, the main lineament is assumed to be hydraulically discontinuous/compartmentalized (not necessarily, but possibly also geometrically). This scenario is considered because: (a) the main lineament appears in the seismic data to consist of smaller, discrete seismic discontinuities of various orientations, some of which appear disconnected; and (b) the productive fracture zone has limited size (about 200 m) as constrained from well test analyses and simulation.

Model 3: Hinge-parallel fault of limited extent/curvature. Model 3 also assesses a single high-permeability zone of limited lateral extent in the CB3-09 area, this time trending 320° (well tests do not provide information on direction). This direction also coincides with seismic discontinuities. No significant contribution to flow from the main lineament is assumed.

Model 4: Discontinuous high-permeability zones of variable orientation in the CB3-09 area. Model 4 is a variation on models 2 and 3. Model 4 most closely reproduces the more complex fault interpretation shown in Fig. 3 and is characterized by a higher density of laterally restricted seismic disturbance zones trending northwest (i.e., axially), north and east in the CB3-09 area and to the northeast.

Model 5: Axial thrust and disconnected 320° fracture zones. In this model, the potential importance of the axial thrust is considered (not only its northern tip, as potentially in model 3). Structures in other directions are included in the background permeability.

Model 6: Curvature and barrier. Model 6 is similar to model 3, and tests a curvature dependency on well deliverability. It assumes no contribution from potential fracture/fault zones outside the strongest curved area, elongated parallel to the hinge (the curvature is shown colored in Fig. 5). A sealing fault barrier (black feature) is placed across the center of the field to test sensitivity to pressure response.

Model 7: Fracture-enhanced productivity at major fault intersections only. This model addresses the concept that significantly enhanced permeability is obtained only at or near the intersection of two or more large faults. A large fault structure by itself is assumed to be ineffective for flow. An example of this could be the north-trending seismic fault (zone) intersected by CB13-35, which did not cause any mud losses. Permeability enhancement can result from increased connectivity within the intersecting fault damage zones, with or without increased fracturing. All the seismic lineaments are represented in this model.

Other models. In model 8, the axial thrust has permeable damage zones at either end. In model 9 the main lineaments flow while smaller lineaments are barriers. Model 10 features an axial thrust connecting two flowing lineaments and crossing through a barrier.

RESULTS

The history matching is most sensitive to the gas volume in the model stored in the matrix porosity and the permeability contrast between the north, center and south of the field. Although it is less sensitive to the details of the fracture models, it can be said that for a fracture model to be consistent with all the available dynamic data, it must possess the following attributes:

• A large but finite-length fracture zone (high permeability) in the vicinity of CB11-09 and intersecting CB3-09, Fig. 9
• The entire field, from north to south, must be in the same “compartment,” i.e., no sealing faults (this does not exclude faults acting as “baffles”). The volume of this compartment is constrained by the dynamic data
• It is unlikely that there are a large number of well-connected fracture zones with large permeability such as the one intersecting CB3-09. This is due to the 5-mD upper limit of the effective permeability in the better part of the reservoir.

In more detail, the models too well connected to the south relative to the north, or with too much permeability in the south, cause over-depletion in the southern wells (models 5, 7, 8 and 10). Models with barriers cause too little depletion in the south (models 6, 7 and 9). Models with very long permeability zones around the most productive well (CB3-09) are not consistent with the transient buildup data (models 1 and 4). Models 2 and 3 provide equally good matches.

Within models 2 and 3, the total production history is well matched, making them the best fits of all the models. Both have a limited zone of enhanced fracturing in the 3-09 area. Further encouragement for the history match comes from more recently acquired pressure measurements in CB7-09, the sidetrack of 3-09 and CB13-35. The predictions of the pressures in CB7-09 and CB13-35 at the end of the simulation (Dec. 31, 2001) are consistent with measurements.

CONCLUSIONS: PROSPECT RISKING AND REDUCED UNCERTAINTIES.

A new integrated fracture/matrix model for Devonian sheet III West Carbondale has been produced that is consistent with the extreme aerial heterogeneity of well performance. In the model, productive fractures are defined at the seismic scale, and are limited to a single zone crossing the CB3-09 area.

One interpretation is that the main lineament controls fracture flow, primarily because it is aligned with the in situ maximum compressive horizontal stress. Other areas are dominated by matrix flow. This interpretation assumes that the seismic lineaments are not artifacts. Alternatively, fracture flow due to enhanced fracturing in the region depicted by the high curvature near well CB3-09 is also a possible model.

It is recommended to target intersections between north-south lineaments and curvature where it is possible to increase the likelihood of intersecting flowing fractures. In the north end of sheet III West Carbondale, no north-south lineaments the size of the main lineament are present in the seismic, and there are no high-curvature areas comparable to the CB3-09 area. In the north, targeting the relatively high-porosity layers (possibly in the Crossfield) may provide the most reliable drilling strategy provided that the porosity can be found and wells can be made to produce from the matrix.

Given the remaining uncertainties associated with the quality of seismic and limited well coverage outside the crestal area, some updating and refinement of the model is to be expected as future information becomes available.

ACKNOWLEDGEMENT

This article was prepared from “Characterizing fracture and matrix heterogeneities in folded Devonian carbonate thrust sheets, Waterton tight gas fields, Western Canada,” in Lonergam, L., Jolly, R. J. H., Rawnsley, K. and D. J. Sanderson, eds., Fractured Reservoirs, Geological Society, London, Special Publications, 2007, pp. 265-279.

LITERATURE CITED

¹ Monger, J. W. H., Price, R. A. and D. J. Tempelman-Kluit, “Tectonic accretion and the origin of the two major metamorphic and plutonic welts in the Canadian cordillera,” Geology, 10, 1982, pp. 70-75; Price, R. A., “Cordilleran tectonics and the evolution of the western Canada sedimentary basin,” in Mossop, G. and I. Shetsin, eds., Geological Atlas of the Western Canada Sedimentary Basin, 2nd ed., Alberta Research Council and Canadian Society of Petroleum Geologists, Calgary, 1994, pp. 13-24.
² Bloy, G. in Structural Style, Sedimentology and Fracture Characterization of the Castle Reef Fm., Montana, field trip guidebook, Post-4, CSPG-CWLS Joint Technical Convention, 1995; Boyer, S. E., “Geometric evidence for synchronous thrusting in the southern Alberta and northwest Montana thrust belts,” in McClay, K. R., ed., Thrust Tectonics, Chapman & Hall, London, 1992, pp. 377-390; Holl, J. E. and D. J. Anastasio, “Deformation of a foreland carbonate thrust system, Sawtooth Range, Montana,” Geological Society of America Bulletin, 104, 1992, pp. 944-953.
³ The modeling techniques are based on Rawnsley, K. et al., “New software tool improves fractured reservoir characterization and modelling through maximized use of constraints and data integration,” SPE 88785 presented at the Abu Dhabi International Conference and Exhibition, Abu Dhabi, Oct. 10-13, 2004.

THE AUTHORS
	Keith Rawnsley is a structural geologist who has focused on fractured reservoir characterization and modeling. He has a PhD degree in engineering geology from the University of Leeds, UK, and 17 years’ industry experience in Elf, BP and Shell since 2000. He is currently working on a cross posting as the production geologist for the Qarn Alam steam project within Petroleum Development Oman. He has worked on fractured and faulted reservoirs in most major petroleum plays.
	Martin de Keijzer joined Shell E&P as a structural geologist in 2000 and has worked on a number of fractured carbonate reservoirs worldwide, combining research and field development. He is currently working with Shell for the Nederlandse Aardolie Maatschappij (NAM) in the Netherlands on a variety of topics, including reservoir compartmentalization, salt tectonics and drilling hazards. He has a PhD degree in structural geology from the University of New Brunswick, Canada. Dr. de Keijzer can be contacted at m.dekeijzer@shell.com.
	Lingli Wei is a reservoir engineer in Shell International Exploration and Production in The Netherlands. He earned an MSc in hydraulic engineering at Tsinghua University, Beijing, in 1988, and a PhD degree in rock mechanics from Imperial College, London, in 1992. He has worked primarily on geomechanical and reservoir engineering aspects of fractured reservoirs. He gained experience in geotechnical, nuclear waste disposal and the oil and gas industry working for an international consultancy company and then BP before joining Shell in 2001.
	Solenn Bettembourg earned an MSc degree in applied mathematics and computer science from the Université Pierre et Marie Curie, Paris. She joined Shell International E&P in 1998, and worked as a reservoir engineer on various research projects focusing on fractured carbonate reservoir characterization. She has experience in fractured carbonate fields and EOR projects.
	Wenche Asyee graduated in 1989 from Utrecht University in the Netherlands with an MSc degree in structural geology and has worked as a consulting geoscientist since then. She joined the Shell carbonate team in 2000, focusing on seismic interpretation (QI, volume interpretation and visualization) on a large variety of proven oil and gas carbonate fields worldwide.
	Peter Swaby received a first-class honors degree in mathematics in 1983 and a PhD degree in computer science in 1987 from the University of Birmingham, UK. Since then he has spent 20 years in the oil business, initially for BP, then for Elf/Total and now for Shell.