AMPLITUDE INVERSION OF FAST AND SLOW CONVERTED WAVES FOR FRACTURE CHARACTERIZATION OF THE MONTNEY FORMATION IN
POUCE COUPE FIELD, ALBERTA, CANADA
c
Copyright by Tyler L. MacFarlane, 2014 All Rights Reserved
A thesis submitted to the Faculty and the Board of Trustees of the Colorado School of Mines in partial fulfillment of the requirements for the degree of Master of Science (Geo-physics). Golden, Colorado Date Signed: Tyler L. MacFarlane Signed: Dr. Thomas L. Davis Thesis Advisor Golden, Colorado Date Signed: Dr. Terence K. Young Professor and Head Department of Geophysics
ABSTRACT
The Montney Formation of western Canada is one of the largest economically viable gas resource plays in North America with reserves of 449TCF. As an unconventional tight gas play, the well development costs are high due to the hydraulic stimulations necessary for economic success. The Pouce Coupe research project is a multidisciplinary collaboration between the Reservoir Characterization Project (RCP) and Talisman Energy Inc. with the objective of understanding the reservoir to enable the optimization of well placement and completion design. The work in this thesis focuses on identifying the natural fractures in the reservoir that act as the delivery systems for hydrocarbon flow to the wellbore.
Characterization of the Montney Formation at Pouce Coupe is based on time-lapse mul-ticomponent seismic surveys that were acquired before and after the hydraulic stimulation of two horizontal wells. Since shear-wave velocities and amplitudes of the PS-waves are known to be sensitive to near-vertical fractures, I utilize isotropic simultaneous seismic in-versions on azimuthally-sectored PS1 and PS2 data sets to obtain measurements of the fast and slow shear-velocities. Specifically, I analyze two orthogonal azimuths that are parallel and perpendicular to the strike of the dominant fracture system in the field. These volumes are used to approximate the shear-wave splitting parameter (γ(s∗)) that is closely related to crack density. Since crack density has a significant impact on defining the percolation zone, the work presented in this thesis provides information that can be utilized to reduce uncertainty in the reservoirs fracture model.
Isotropic AVO inversion of azimuthally limited PS-waves demonstrates sufficient sensi-tivity to detect contrast between the anisotropic elastic properties of the reservoir and is capable of identifying regions with high crack density. This is supported by integration with spinner production logs, hydraulic stimulation history of the field, and microseismic. Results also show significant fracture network heterogeneity that is not typically accounted for in
engineering-driven development despite a strong link to production.
The main value of this work lies in the integration of fracture characterization with preceding RCP theses that defined the geomechanical model and composition of the reservoir at Pouce Coupe. Geophysical attributes that relate to the composition and natural fractures enable a more complete understanding of the reservoir and indicate that a successful well is dependent on both the hydrocarbon storage capacity of the matrix and a large permeable network of natural fractures.
TABLE OF CONTENTS
ABSTRACT . . . iii
LIST OF FIGURES . . . viii
LIST OF TABLES . . . xiv
LIST OF ABBREVIATIONS . . . xv ACKNOWLEDGMENTS . . . xvi DEDICATION . . . xvii CHAPTER 1 INTRODUCTION . . . 1 1.1 Montney Geology . . . 3 1.1.1 Stratigraphy . . . 3 1.1.2 Reservoir Characteristics . . . 5 1.1.3 Regional Tectonics . . . 8
1.2 Pouce Coupe data set . . . 10
1.2.1 Field Development . . . 10
1.2.2 Time-Lapse Multicomponent Seismic . . . 13
1.2.3 Microseismic Data . . . 15
1.3 Previous Pouce Coupe Research . . . 16
1.4 Research Objective . . . 24
CHAPTER 2 SEISMIC MODEL OF FRACTURES . . . 26
2.1 Natural Fracture Background . . . 27
2.3 Fracture Compliance Methodology . . . 30
2.4 Numeric Modeling . . . 33
2.5 Pouce Coupe Seismic Implications . . . 39
CHAPTER 3 MULTICOMPONENT SEISMIC DATA PROCESSING . . . 43
3.1 Fixed Receiver Rotation . . . 44
3.2 Bin Size and COV Interpolation . . . 47
3.3 Azimuthal Sectoring . . . 49
3.4 Processing Conclusions . . . 52
CHAPTER 4 CONVERTED WAVE SEISMIC INVERSION . . . 53
4.1 Inversion Theory . . . 54
4.1.1 Constrained Sparse Spike Inversion . . . 56
4.2 Available Data and PS Seismic Interpretation . . . 58
4.3 Low Frequency Model . . . 61
4.4 PS1 Inversion . . . 62
4.4.1 Data Conditioning . . . 66
4.4.2 Well Tie and Wavelet Extraction . . . 69
4.4.3 Inversion Parameters . . . 75
4.4.4 Inversion Results and Quality Control . . . 77
4.5 PS2 Inversion . . . 80
4.5.1 Data Conditioning . . . 82
CHAPTER 5 FRACTURE NETWORK INTERPRETATION AND INTEGRATION . 95
5.1 Shear Velocity Fracture Characterization . . . 95
5.1.1 Interpretation Limitations . . . 96
5.1.2 Fracture Map . . . 97
5.2 Fracture/Production Correlation . . . 99
5.3 Rock Composition Analysis . . . 106
5.4 Microseismic . . . 106
5.5 Comparison to SWS Analysis . . . 108
CHAPTER 6 CONCLUSIONS AND RECOMMENDATIONS . . . 112
6.1 Recommendations . . . 113
LIST OF FIGURES
Figure 1.1 Map of the Peace River embayment that bounds the Montney Formation . 4 Figure 1.2 Paleogeographic map of North America during the Early Triassic (245Ma) . 5
Figure 1.3 Montney type log and stratigraphic chart of the Triassic . . . 6
Figure 1.4 Depositional model of the Montney Formation . . . 6
Figure 1.5 East-West modeled stratigraphic section of the Montney Formation . . . . 7
Figure 1.6 Map of maximum horizontal stress derived from well bore breakouts. Pouce Coupe field is highlighted by a star . . . 9
Figure 1.7 Pouce Coupe Field layout and timeline of operations . . . 11
Figure 1.8 Pouce Coupe Average Daily Gas Production for Montney wells . . . 13
Figure 1.9 Source and Receiver layout of the Pouce Coupe seismic survey . . . 14
Figure 1.10 Offset/Azimuth distribution in individual bins of the Pouce Coupe Seismic . . . 16
Figure 1.11 Microseismic events: a) from the 02/07-07 stimulation recorded by the 08-07 toolstring , b) from the 02/07-07 stimulation recorded by the 09-07 toolstring c) from the 02/02-07 well recorded by the 08-07 toolstring d) from the 02/02-07 well recorded by the 09-07 toolstring . . . 17
Figure 1.12 Downhole microseismic acquisition geometry . . . 18
Figure 1.13 Model of SWS . . . 19
Figure 1.14 Baseline SWS magnitude and PS1 polarization orientation . . . 21
Figure 1.15 Monitor 1 SWS magnitude and PS1 polarization orientation . . . 21
Figure 1.18 Inverted P-Impedance cross section through well 02/02-07 with
geobodies of best clusters . . . 24 Figure 2.1 Schematic of an orthorhombic model . . . 26 Figure 2.2 Mohr-Coulomb graphical representation of stress fields producing
tensional and shear rock failures . . . 28 Figure 2.3 Mohr-Coulomb graphical representation of fracture failure produced by
increasing pore pressure . . . 28 Figure 2.4 Simulated hydraulic fracture in the presence of perforations that act as
permeable conduits and glass plates that represent cemented natural
fractures . . . 30 Figure 2.5 Schematic model of a medium with multiple sets of parallel fractures . . . 32 Figure 2.6 Workflow for modeling the seismic response of fractures. . . 34 Figure 2.7 PS reflection coefficients of the four fracture models associated with the
Montney C/D interface along azimuths aligned with the x1 and x2 axes . 38 Figure 2.8 Comparison between anisotropic and isotropic PS reflection coefficient
equations used to represent fracture model 3 . . . 41 Figure 3.1 Previous processing of the Pouce Coupe data set. PS1 is displayed on
the left, and PS2 is on the right . . . 44 Figure 3.2 New AVO compliant processing of the Pouce Coupe data set. PS1 is
displayed on the left, and PS2 is on the right . . . 45 Figure 3.3 Schematic of field components H1/H2, and rotated PS1/PS2 coordinates . 46 Figure 3.4 PS1 orientation as determined by SWS analysis within a time window
of 700-900ms . . . 46 Figure 3.5 Overburden shear-wave splitting for three analysis windows . . . 48 Figure 3.6 Common offset vector diagram and its relationship to offset and azimuth . 49 Figure 3.7 Old and New AVO compliant PS1 pre-migration stacks . . . 50 Figure 3.8 PS1 and PS2 limited azimuth stacks sorted in 10◦ azimuth sectors from
Figure 3.9 Limited azimuth ranges allowed in PS1 and PS2 volumes . . . 52
Figure 4.1 Deterministic and stochastic inversion workflow overview . . . 54
Figure 4.2 Representation of the convolutional model . . . 55
Figure 4.3 Jason’s workflow for the simultaneous CSSI . . . 57
Figure 4.4 Pouce Coupe basemap with the location of wells 13-12 and 02/07/07 displayed . . . 59
Figure 4.5 Cross-section of the PS1 seismic with inlayed seismograms from wells 13-12 and 02/07-07. Horizon interpretation is also displayed . . . 59
Figure 4.6 PS1 time structure map of the a) Triassic b) Montney E c) Belloy . . . . 60
Figure 4.7 Bandwidth of the Pouce Coupe seismic with the lowpass filter applied to the models shown in red and the bandpass filter applied to the inversion result in cyan . . . 61
Figure 4.8 Inverse distance interpolation technique . . . 62
Figure 4.9 Cross section of the shear velocity model with a) 30Hz high cut filter applied b) 9Hz high cut filter applied . . . 63
Figure 4.10 Cross section of the compressional velocity model with a) 30Hz high cut filter applied b) 9Hz high cut filter applied . . . 64
Figure 4.11 Cross section of the density model with a) 30Hz high cut filter applied b) 9Hz high cut filter applied . . . 65
Figure 4.12 PS1 substack 10◦-25◦ . . . 67
Figure 4.13 PS1 substack 18◦-33◦ . . . 67
Figure 4.14 PS1 substack 26◦-41◦ . . . 68
Figure 4.15 PS1 substack 34◦-49◦ . . . 68
Figure 4.16 All PS1 substacks overlaid to QC data alignment. The red box highlights a misaligned region, and the blue box highlights a area of high data quality . . . 69
Figure 4.18 Well 02/07-07 well tie with the full PS1 substack . . . 70
Figure 4.19 Well 02/07-07 well tie with the 10◦-25◦ PS1 substack . . . 71
Figure 4.20 Well 02/07-07 well tie with the 18◦-33◦ PS1 substack . . . 71
Figure 4.21 Well 02/07-07 well tie with the 26◦-41◦ PS1 substack . . . 72
Figure 4.22 Well 02/07-07 well tie with the 34◦-49◦ PS1 substack . . . 72
Figure 4.23 Well 13-12 well tie with the full PS1 substack . . . 73
Figure 4.24 Well 13-12 well tie with the 10◦-25◦ PS1 substack . . . 73
Figure 4.25 Well 13-12 well tie with the 18◦-33◦ PS1 substack. The overall correlation is 0.66 . . . 74
Figure 4.26 Well 13-12 well tie with the 26◦-41◦ PS1 substack . . . 74
Figure 4.27 Well 13-12 well tie with the 34◦-49◦ PS1 substack . . . 75
Figure 4.28 Angle dependent wavelets for each of the four PS1 substacks . . . 76
Figure 4.29 Final multi-well angle dependent wavelets for each of the four PS1 substacks . . . 76
Figure 4.30 Cross section of the inverted V(1)S (0-30Hz) . . . 78
Figure 4.31 Cross section of the bandpass filtered inverted V(1)S (9-30Hz) . . . 79
Figure 4.32 Cross section of a) low frequency model of VS (0-9Hz) b) inverted V (1) S (0-9Hz) . . . 79
Figure 4.33 Pseudo V(1)S logs extracted from the PS1 inversion in comparison to the VS log from well 02/07-07 . . . 80
Figure 4.34 Seismic residuals associated with the four PS1 substacks . . . 81
Figure 4.35 PS2 substack 10◦-25◦ . . . 83
Figure 4.36 PS2 substack 18◦-33◦ . . . 83
Figure 4.38 PS2 substack 34◦-49◦ . . . 84
Figure 4.39 All PS2 substacks overlaid to QC data alignment . . . 85
Figure 4.40 Histogram of the RMS(PS1)/RMS(PS2) extraction . . . 85
Figure 4.41 Angle dependent wavelets for each of the PS2 substacks in comparison with the PS1 multi-well wavelet . . . 86
Figure 4.42 Well 02/07-07 well tie with the 10◦-25◦ PS2 substack . . . 87
Figure 4.43 Well 02/07-07 well tie with the 18◦-33◦ PS2 substack . . . 88
Figure 4.44 Well 02/07-07 well tie with the 26◦-41◦ PS2 substack . . . 88
Figure 4.45 Well 02/07-07 well tie with the 34◦-49◦ PS2 substack . . . 89
Figure 4.46 Cross-section of the PS2 inverted V (2) S (0 − 30Hz) . . . 89
Figure 4.47 Cross-section of the PS2 bandpass filtered inverted V (2) S (9 − 30Hz) . . . . 90
Figure 4.48 Pseudo V(2)S logs extracted from the PS2 inversion . . . 90
Figure 4.49 Cross section of a) low frequency model of VS (0-9Hz) b) PS2 inverted V(2)S (0-9Hz) . . . 92
Figure 4.50 Seismic residuals associated with the four PS2 substacks . . . 93
Figure 4.51 Comparison of pseudo V(2)S logs extracted from PS2 inversions that use the original and scaled LFM’s . . . 94
Figure 5.1 Subset of the Pouce Coupe seismic survey used for interpretation . . . 97
Figure 5.2 Cross section of γ(s∗) along the well bore of 02/02-07 . . . 98
Figure 5.3 Cross section of γ(s∗) along the well bore of 02/07-07 . . . 99
Figure 5.4 γ(s∗) time slice at 2108 ms through the Montney D subunit . . . 100
Figure 5.7 γ(s∗) % time slice maps through a) the Montney D b) Montney C. Stage by stage gas flows obtained from a spinner log are presented as a
percentage of the total flow . . . 102 Figure 5.8 Oblique view of the well layout and frac stages highlighting the
proximity of wells 00/07-07 and 02/07-07 . . . 103 Figure 5.9 Average daily water production of well 00/07-07 . . . 104 Figure 5.10 γ(s∗) time slice at 2140ms through the Montney C subunit highlighting
the large fracture signature surrounding the three stimulation stages of
well 00/07-07 . . . 105 Figure 5.11 Cross section of P impedance (derived from PP seismic) through well
02/02-07 including geobodies of clusters 1 and 2 . . . 107 Figure 5.12 Cross section of P impedance (derived from PP seismic) through wells
00/07-07 and 02/07-07 including geobodies of clusters 1 and 2 . . . 107 Figure 5.13 γ(s∗) timeslices integrated with microseismic . . . 109 Figure 5.14 Comparison of the shear-wave splitting and azimuthal velocity inversion
LIST OF TABLES
Table 1.1 General reservoir properties of the Montney Formation within Pouce Coupe . 8 Table 1.2 Hydraulic stimulation parameters . . . 12 Table 1.3 Stage-by-stage production data from spinner logs as percent of total flow
volume . . . 13 Table 1.4 Pouce Coupe seismic survey acquisition parameters . . . 15 Table 2.1 Properties of unfractured Montney C/D reservoir units . . . 35 Table 2.2 Tsvankin parameters of the Montney C associated with the four fracture
models . . . 37 Table 2.3 Shear-wave splitting coefficients for the four Montney C fracture models . . 39 Table 4.1 Misfit functions and their influence on the Inversion . . . 58 Table 4.2 PS1 constrained sparse spike inversion parameters . . . 77
LIST OF ABBREVIATIONS
Ultra Sonic Borehole Image . . . UBI Amplitude Variation with Offset . . . AVO Amplitude Variation with Angle . . . AVA Shear-Wave Splitting . . . SWS Fast Converted Wave Mode . . . PS1 Slow Converted Wave Mode . . . PS2 Limited Azimuth Stack . . . LAS Common Offset Vector . . . COV Pre-Stack Time Migration . . . PSTM Horizontal Transverse Isotropy . . . HTI Normalized Root Mean Squared . . . NRMS Constrained Sparse Spike Inversion . . . CSSI Low Frequency Model . . . LFM National Energy Board . . . NEB Gamma Ray . . . GR
ACKNOWLEDGMENTS
First and foremost, I would like to thank my advisor Dr. Tom Davis for his guidance, friendship, and life changing opportunities he has provided. You are a one of a kind professor who I am privileged to have shared the same penalty box with. I also need to express gratitude to the rest of my committee: Dr. Ilya Tsvankin, Dr. Jeff Grossman, and Dr. Robert Benson for their wisdom and support. A special thanks needs to go to Talisman Energy for acquiring the unique Pouce Coupe data set and graciously providing it to the RCP.
The last 2 years at the Colorado School of Mines have proved to be an spectacular experience that stems from the amazing people that I have been fortunate to surround myself with. As a whole you are all the most talented and friendly group of people I could have expected to encounter. I truly hope we are able to stay in touch no matter where this worldly group disperses to after graduation. I am especially grateful to the students who preceded my time on the Pouce Coupe project. Collaborations between Jared Atkinson, Heather Davey, Chris Steinhoff, Matthew Lee, and Claudia Duenas have always pushed me and broadened my understanding of reservoir characterization.
I have also received advice from numerous people throughout my program. Most notably is David D0amico at Talisman Energy who has been leading Talisman’s interaction with the RCP. I also worked closely with Tom Bratton who is a spectacular mentor and scientist. Peter Mesdag, Brad Bacon, Walt Lynn, Michael O0Brien have also provided significant guidance throughout my time at Mines.
Lastly, I need to deeply thank my wife Jackie for her love, support, and putting up with a long distance relationship immediately after our wedding. You are my world, and I can
CHAPTER 1 INTRODUCTION
Research presented in this thesis is part of a joint effort between the Reservoir Character-ization Project (RCP) at the Colorado School of Mines and Talisman Energy Inc. The Pouce Coupe project began in 2008 with the objective of improving the economics of tight gas de-velopment of the Montney Formation by understanding how hydraulic stimulations interact with the reservoir. This multidisciplinary project focuses on defining reservoir components critical to a wells success, and utilizes time-lapse multicomponent seismic to demonstrate integrated technology that can be used to optimize future well locations.
The onset of horizontal drilling and multistage hydraulic fracturing led to the economic development of unconventional resources in the early 2000s. Production from unconvention-als has risen sharply and now contributes 23.1% of the total gas produced in the United States (Zhongmin and Krupnick (2013)). While significant technological advancements have enabled rapid unconventional growth, there is still substantial uncertainty in predicting reservoir ”sweet spots”.
A consequence of this lack of geologic/geophysical information is engineering-driven field development, which typically assumes uniform/isotropic reservoir properties in heteroge-neous/anisotropic fields. This results in lost profits due to misplaced wells and unsuccessful stimulations. In 2010, $30 billion was spent on hydraulic fracturing within the US, and approximately 25% of all completions failed to meet performance expectations (Machnizh (2013)). Clear examples of both a successful and unsatisfactory stimulation will be exhibited within Pouce Coupe.
of magnitude and strongly impact how a hydraulic stimulation interacts with the reservoir (Olson et al. (2012)). Results in this thesis aim to improve upon previous Pouce Coupe fracture characterizations performed by Atkinson (2010) and Steinhoff (2013). Vertical res-olution of fractures is expected to improve from the reservoir scale (∼300m) to subunit scale (∼40m) due to a shift from previous travel time based measurements of seismic anisotropy to pre-stack amplitude measurements.
While many interrelated reservoir properties are responsible for the success of an uncon-ventional well, this thesis focuses on modeling the anisotropic nature of fractures, processing converted-wave seismic data for the purpose of azimuthal amplitude analysis, and provid-ing an interpretational tool for fracture characterization. This fracture characterization methodology utilizes two separate constrained sparse spike inversions to predict azimuthally dependent shear velocities in the study area’s vertical symmetry planes that are parallel and perpendicular to the expected dominant fracture set. It is important to note that the inversion utilizes an isotropic PS-wave AVO-inversion algorithm, which is currently the state of the art for commercial availability. Therefore, many anisotropic effects are not taken into account, which limits the interpretation to qualitative observations.
Input into these inversions include azimuthally sectored pre-stack converted wave seismic representative of fast and slow wave modes. Fast wave mode is acquired from source-receiver (S/R) azimuths parallel to the dominant fracture set, and the slow wave mode represents S/R azimuths perpendicular to the fracture set. The elastic properties of a fractured rock are shown to be azimuthally dependent, and therefore converted waves in the fast and slow mode will exhibit different seismic amplitudes for the same spatial location. The inversions utilize Amplitude Variation with Offset (AVO) to determine shear impedances that can be related back to the fracture model.
Fractional difference between the fast and slow shear velocities is defined as the shear wave splitting parameter (γ) and provides a predictive and quantitative measurement of reservoir anisotropy due to fractures and the prevailing stress field. Integration of multicomponent
seismic with microseismic, image logs and cluster analysis enables calibration of the seismic response and a more complete interpretation of the reservoir.
1.1 Montney Geology
One of the largest economically viable gas resource plays in North America is the Montney Formation located in Northern Alberta and British Columbia. In 2013, marketable reserves from the Montney were estimated by Canada0s National Energy Board (NEB) at 449tcf gas, 14.5 billion bbl of natural gas liquids (NGL), and 1.125 million bbl of oil (British Columbia Oil and Gas Commission, 2013). Despite this play frequently being called new, hydrocarbon exploration of the Montney began with the exploitation of conventional oil from sandstones on the eastern margin in the 1950s (ERCB, 2012). A resurgence of this formation began in 2005 with the onset of horizontal drilling and multistage hydraulic fracturing that produced economic drivers that opened up the majority of the Montney that is rich in gas and condensate. Current drilling operations and this research project are focused on the unconventional tight siltstones and shales that make up the greatest volume of rock within the Montney.
1.1.1 Stratigraphy
The Montney Formation is a Triassic age unconventional reservoir confined to the Peace River Embayment (Figure 1.1). This sub-basin developed during early Carboniferous and Permian when subsidence occurred as a broad downwarp with a large central half-graben (Edwards et al. (1994)). This produced a marine to marginal-marine continental shelf with water depths increasing toward the west (Figure 1.2). Montney deposition began after a major transgression eroded Carboniferous/Permian strata and subsequent regression that transported multi-cycled sediment from the craton in the east and north (Edwards et al. (1994)). Erosion has led to unconformable boundaries above and below Montney that can
high gamma ray (GR) response.
Figure 1.1: Map of the Peace River embayment that bounds the Montney Formation. Various Montney oil and gas fields are shown in green and red respectively. Pouce Coupe Field highlighted by a black star (Zonneveld et al. (2011)).
Six Montney subunits (A-F) associated with lower order transgression/regression cycles are frequently observed (Figure 1.3). Units A-C are commonly called the Lower Montney, and units D-F are linked to the upper Montney. The lower Montney typically consists of shoreface facies and coarse siltstones deposited more proximal to the sediment source. Within the confines of this Pouce Coupe study area, Montney C was classified by Derder (2012) as a finely laminated siltstone, and production data has proven it to be the best producing subunit. Montney D is also frequently targeted for exploitation, and expected to have been deposited in a more distal environment.
Figure 1.2: Paleogeographic map of North America during the Early Triassic (245Ma) which is representative of the time during Montney deposition. A black box highlights the approx-imate bounds of the formation and a red star shows the Pouce Coupe fields paleolocation (Blakey (2011)).
Figure 1.4 shows the general depositional model of the Montney as a continental ramp with progressive parasequences off-lapping to the west (Zonneveld et al. (2011)). This model demonstrates the complexities associated with the Montney, and how reservoir properties can change dramatically depending on where deposition occurred in the model. Deep water successions in the west can include turbidite channels and fan complexes, while deltaic or estuary influenced upper shoreface facies exist to the east. Montney Formation within Pouce Coupe is expected to have been deposited on the proximal slope where turbidite potential exists (Derder (2012)).
1.1.2 Reservoir Characteristics
As a self-sourcing petroleum system, the Montney Formation acts as the source, reservoir, and seal largely due to the formations high organic content, and low porosity/permeability.
Figure 1.3: Montney type log and stratigraphic chart of the Triassic. (Davey (2012)).
Figure 1.4: Depositional model of the Montney Formation. Environments range from shoreface in the East, to congenital slope and distal marine in the West (Davey (2012)).
voir, the organic content undergoes a volume expansion as hydrocarbons are generated. This volume expansion combine with very low matrix permeability leads to three desirable traits within the Montney: 1) reservoir overpressure, 2) naturally occurring fractures, and 3) low water saturations (Meissner (1978)). All of these characteristics make the Montney and other unconventional plays like it exceptional hydrocarbon targets.
While the Montney is frequently referred to as shale, the most accurate description of the reservoir rock at Pouce Coupe is an organic-rich argillaceous siltstone interbedded with shale (Davey, 2012). Siltstones dominate the central-west, and shales become more prominent in the far west as the depositional environment becomes more distal (Figure 1.5). Reservoir rocks within this research project are clastic-rich, composed of >60% quartz and feldspar, and contain lower concentrations of clay minerals and dolomite (Derder (2012)).
Figure 1.5: East-West modeled stratigraphic section of the Montney with approximate loca-tion of Pouce Coupe highlighted by a star (Davey (2012)).
commercial production of hydrocarbons. Samples with natural fractures are observed to increase permeability by orders of magnitude and range from 1.7 to 24 Ka in mD (ERCB, 2012).
High values of total organic carbon (TOC) within the Montney range from 1-5% and are one of the key reasons for the formation0s success as an unconventional reservoir. Due to the monocline slope of the Peace River Embayment, Montney reservoir depths gradually increase from outcrop in the east to over 4000m. Depth to the reservoir top within Pouce Coupe Field ranges from 1700-2200m. This overburden is sufficient for thermal maturity of the type II/III kerogen in the gas generation window. Formation thickness follows a similar east-west trend that ranges from 0 to 300m, with an average of 250m for unconventional assets. A summary of relevant reservoir properties for Pouce Coupe Field is shown in Table 1.1.
Table 1.1: General Reservoir Properties of the Montney Formation Within Pouce Coupe (Steinhoff (2013))
Property Approximate Range
Permeability (md) 20 Porosity(%) 7-9 TOC (%) 0-1.5 Thermal Maturity (%R0) 1-2 Thickness (m) 170-350 Burial Depth (m) 1700-2200 1.1.3 Regional Tectonics
Tectonic activity was much more pronounced in the Paleozoic prior to deposition of the Montney. This tectonic activity caused variations in structure and accommodation space that explain local thickness variations and location of turbidity flows. Some of these faults were reactivated in the Triassic and thus could have an impact in generating fractures within the reservoir. However, the magnitude of such events was diminished and varied spatially with the strongest stresses acting in the western foothills (Edwards et al. (1994)). The Laramide Orogeny extending from the Late Cretaceous into the Tertiary was the major
tectonic event responsible for the current compressional stress state observed throughout the field where strike-slip and reverse failures are expected.
Within Pouce Coupe Field, the stress field indicates that vertical strike-slip failure will be the most common due to the overburden stress (Sv) having a lower magnitude than the max-imum horizontal stress (SH), but less then the minimum horizontal stress (Sh) (SH>Sv>Sh). It is important to note that a large differential in horizontal stress will tend to produce linear fractures that propagate in the direction of SH. According to the World Stress Map and field data observations, the maximum horizontal stress orientation is N40◦E (Figure 1.6).
Figure 1.6: Map of maximum horizontal stress orientation derived from well bore breakouts. Pouce Coupe field is highlighted by a star (Heidbach et al. (2008)).
1.2 Pouce Coupe data set
Hydraulic stimulations are necessary for unconventional tight gas development to be economic, and is often the single largest expense associated with bringing a well online. The Pouce Coupe Field is data rich and based around a time-lapse multicomponent seismic survey acquired by Talisman Energy in 2008 to monitor and evaluate hydraulic stimulations of two horizontal wells.
1.2.1 Field Development
Wells of specific interest are the 02/02-07 well drilled into the Montney C, and the 02/07-07 well drilled into Montney D. After well stimulation, both wells were allowed to flow just long enough to recover the treatment balls at the surface before being shut in to maintain maximum reservoir pressure. This ensures that stimulated fractures remain propped open due to the artificially high pore pressure. Figure 1.7 displays the field layout and timeline of field operations, which took place in rapid succession. Over an 11 day interval, three time-lapse seismic surveys were acquired before and after each of the two hydraulic stimulations. It is also important to note that well 00/07-07 produces from the Montney C directly below well 02/07-07 and hence does not show up clearly on the basemap. This third important well was drilled and hydraulically stimulated 10 months prior to acquisition of the baseline seismic survey. This rich data set also includes:
1. Several well logging suites (two wells have shear sonic) 2. Downhole and surface microseismic
3. Spinner production data
4. Ultrasonic Borehole Image (UBI) 5. Vertical Seismic Profile (VSP)
Figure 1.7: Pouce Coupe Field layout and timeline of operations. Wells highlighted in red contain sonic logs, wells shown in orange enclose downhole microseismic tool strings, and well 08-07 was logged with a UBI. Modified from Atkinson (2010).
The stimulations in each well being monitored were comprised of 5 stages using an open hole packer system. Amount of proppant, proppant size, and H2O load were identical in both the 02/02-07 and 02/07-07 wells. The primary difference between these two completions is the interval between stage spacing. Well 02/02-07 has stages spaced 200m, and the stages in well 02/07-07 are separated by 250m. Consistent wellbore and completion design enable the significant production differences observed in these wells to be attributed to heterogeneity in the reservoir. These heterogeneities need to be predictively mapped for the full potential of a unconventional reservoirs to be realized. Table 1.2 summarizes hydraulic stimulation parameters for all three Montney producing wells.
Table 1.2: Hydraulic Stimulation Parameters
Well
Stimulation Date
(mm/dd/yy) FluidType
# of
Stages ProppantAmt of
Proppant Size
H2O Load
(m3) Closure Pressure(MPa)
02/02-07 12/12/08 Clear Frac 5 100T 20/40 1328 30
02/07-07 12/17/08 Clear Frac 5 100T 20/40 1300 28
00/07-07 02/10/08 Clear Frac 3 100T 20/40 360 N/A
Cumulative flow volume and spinner log production data demonstrate large differences in the production profiles between both Pouce Coupe study wells. Figure 1.8 shows average daily production per month for the three Montney producing wells within Pouce Coupe Field from initial production to November 2013. The most notable observation is that well 02/02-07 produced 84% more gas than well 02/07-07 in the first two months of production despite the wells being separated by ≈500m. This indicates significant differences in rock quality and level of hydraulic fracturing success. Well 00/07-07 produces from the Montney C with a comparable production profile to well 02/02-07 despite the application of a less intensive stimulation.
The success of individual hydraulic fracture stages is assessed by spinner production logs acquired for wells 02/02-07 and 02/07-07 on Jan 13, 2009 and Jan 15, 2009 respectively (Table 1.3). Well 02/02-07 shows consistent production from all stages, demonstrating a successful completion and resultant high volume of net production. In contrast, 43% of
Figure 1.8: Pouce Coupe Average Daily Gas Production for Montney wells
production from well 02/07-07 comes from one stage. Operational challenges prevented the spinner tool from reaching the toe of well 02/07-07, leading to commingled production measurements from stages 1 and 2.
Table 1.3: Stage by Stage Production Data From Spinner Log as Percent of Total Flow Volume Stage 02/02-07 (%) 02/07-07 (%) 1 20 16 (commingled) 2 13 16 (commingled) 3 25 43 4 17 10 5 24 14
sources buried 5.5m. This permanent recording system enabled the fast timeline necessary to monitor the immediate seismic response of hydraulic stimulation prior to production. High data quality resulted, and ensured high time lapse repeatability. Use of three-component geophones (2 horizontal and 1 vertical axis), enable compressional (PP) and converted (PS) wave modes to be recorded. PS converted wave seismic refers to a down going P wave generated by a conventional source at the surface, and an up going Sv wave that is generated by P wave incidence upon a reflection interface. This process is called mode conversion.
The 3000m by 1600m survey was acquired using megabin geometry with parallel source and receiver lines (Figure 1.9). This acquisition technique provides the advantage of rapidly building fold along the survey edges, uniform wave field sampling, and was cost effective (Goodway and Ragan (1996)). Both source and receiver lines were spaced 200m apart, and a typical patch includes 9 receiver lines, each with 31 stations. Table 1.4 summarizes acquisition parameters.
Figure 1.9: Source and Receiver layout of the Pouce Coupe seismic survey
One consequence of this source and receiver line spacing is a large natural bin size of 100m x 50m. The rectangular geometry also creates an azimuthally biased range of offsets
Table 1.4: Pouce Coupe Seismic Survey Acquisition Parameters
Survey geometry Megabin
Source Line Interval 200/400m
Source Interval 100m
Number of Sources 1241
Charge Size: Baseline = 0.5kg
Monitor 1 = 0.2 kg Monitor 2 = 0.2 kg
Receiver Line Interval 200m
Receiver Interval 200m
Number of Receivers 162
Geophones 135 32 CT-Single, 144 OYO Geospace 3C Nails, 61 DSU 3C Number of Recording Channels 340 (144 permanently buried)
Recording Instrument Sercel 408 XL, 2ms sample Rate
displayed by Figure 1.10. For this study, the consequences of this azimuth bias are limited because the processing workflow requires the data to be sectored in orientations parallel and perpendicular to the maximum horizontal stress (N40E), which happens to lie along the survey0s diagonal. Thus, far offsets are available in both necessary azimuths.
Baseline survey was acquired on December 8-10, 2008. Two subsequent monitor surveys were performed immediately after wells 02/02-07 and 02/07-07 were stimulated. For the purpose of this study, only the baseline seismic survey was used.
1.2.3 Microseismic Data
Microseismic data were acquired from surface and downhole arrays to monitor the inter-action of two 5 stage hydraulic stimulations within the reservoir. This thesis only integrates microseisms recorded by downhole sensors in the 09-07 and 08-07 wells to interpret fracture geometry (length, height, and orientation) (Figure 1.11). Failure mechanisms derived from amplitude ratio analysis by Lee (2014) are also incorporated in the final interpretation of the fracture network in Chapter 5. Failure mechanisms derived from microseism amplitude ratio analysis have the benefit of being on similar investigative scales as surface seismic, which
Figure 1.10: Offset/Azimuth distribution in individual bins of the Pouce Coupe Seismic showing a significant offset bias dependent on azimuth.
Well 08-07 contains a tool string of 10 3C geophones. The tool string in well 09-07 encloses 50 3C geophones spaced by 15m (Figure 1.12). Rapid attenuation of low magnitude microseisms combined with an increasing observation bias associated with increasing source-receiver distance introduce data quality limitations. Within the Pouce Coupe data set, these sources of error are mitigated through the use of a multi-well array. Geophone locations near the heel and the toe of production wells provide reasonable coverage of all hydraulic stimulation stages in the 02/02-07 and 02/07-07 wells.
1.3 Previous Pouce Coupe Research
RCP research on the Pouce Coupe data set began with Atkinson (2010) and was pro-gressed by Davey (2012), Steinhoff (2013), Lee (2014), and Due˜nas (2014). Atkinson (2010) demonstrated that the stress conditions of the reservoir are greatly altered in a hydraulic stimulation through reservoir modeling of fluid flow and stress regime analysis. However, the time-lapse compressional seismic response was determined to be too small for observa-tion due to the low permeability of the reservoir which prevented fluid from flow from the
Figure 1.11: Microseismic events: a) from the 02/07-07 stimulation recorded by the 08-07 toolstring , b) from the 02/07-07 stimulation recorded by the 09-07 toolstring c) from the 02/02-07 well recorded by the 08-07 toolstring d) from the 02/02-07 well recorded by the 09-07 toolstring (Lee (2014)).
Figure 1.12: Downhole microseismic acquisition geometry (Lee (2014)).
induced fractures into the background rock. This initiated the use of converted wave seismic and shear wave splitting (SWS) analysis to monitor the hydraulic stimulation.
SWS analysis is a measurement of seismic anisotropy based on time delays associated with fast and slow shear wave propagation through anisotropic media. As a reflected shear wave passes through a preferentially aligned fractured media, it splits into a fast (PS1) and slow (PS2) mode with orthogonal sets of particle propagation (Figure 1.13) (R¨uger (2001)). At vertical incidence, the PS1-wave will have particle motion in the direction of the fracture strike, and the PS2-wave will be polarized orthogonally. The arrival time delays between the fast and slow wave modes are associated with fracture density (equation 1.1). The PS1 orientation, which can be used as a proxy for fracture strike is also determined by SWS analysis. Grossman et al. (2013) provides a detailed a detailed overview of the processing sequence that developed the SWS analysis maps in the Pouce Coupe data set.
SW S = tP S2− tP S1 tP S2
Figure 1.13: Model of SWS. As a shear wave passes through uniform fractured media it splits in to the fast (S1) and slow (S2) modes with perpendicular polarizations (Hardage et al. (2011))
Steinhoff (2013) advanced the SWS analysis performed by Atkinson (2010) utilizing a detailed seismic reprocessing effort by Sensor Geophysical to increase the NRMS repeatability of the time lapse surveys and thereby preserve time delays within the reservoir. A significant improvement in data quality resulted, and enabled a reasonable correlation between observed time delays at the base of the reservoir and production data. Key improvements in the processing workflow included;
1. Careful overburden layer stripping
2. Improved alignment of the two horizontal components of the 3C geophone with fast and slow shear wave particle polarizations
3. Detailed velocity analysis in a simultaneous time lapse processing workflow
SWS analysis on baseline survey was used to characterize natural fractures within the reservoir prior to stimulation (Figure 1.14). The dominant fracture orientation was observed to shift from the expected maximum horizontal stress orientation of N40◦E in the east part of the survey, towards an orthogonal direction in the west. Within the reservoir interval, maxi-mum magnitude of SWS was 3% with most of the survey having values close to 0%. Monitor 1 was acquired immediately after well 02/02-07 was stimulated and used to determine the induced hydraulic fracturing effect. SWS analysis detected 3 areas within the survey with increased time delays when compared to the baseline survey (Figure 1.15). SWS magni-tude reaches 8% near stage 3 and 4 of well 02/02-07 with the build up constrained to the south part of the wellbore indicating a potential asymmetric frac. Monitor 2 was acquired to study the hydraulic fracturing effect of well 02/07-07 and produced an interesting result. SWS magnitude did not have a significant increase surrounding the 02/07-07 well that was stimulated, however the anomaly surrounding well 02/02-07 expanded (Figure 1.16).
A wellbore scale geomechanical characterization of the Montney Fm. performed by Davey (2012) provided a model that defined reservoir characteristics prone to successful hydraulic stimulations. Davey developed the modified rock quality index (RQI) to define best parts
Figure 1.14: Baseline SWS magnitude (color) and PS1 polarization orientation (needle). Weak splitting anomaly are observed (Steinhoff (2013)).
Figure 1.15: Monitor 1 SWS magnitude (color) and PS1 polarization orientation (needle) (Steinhoff (2013)).
Figure 1.16: Monitor 3 SWS magnitude (color) and PS1 polarization orientation (needle). After two hydraulic stimulating, the splitting anomalies appear to have dispersed throughout the survey (Steinhoff (2013)).
of the reservoir in an effort to understand how composition and rock fabric relate to stress anisotropy, fracturing, and reservoir properties. Principal factors determined to affect reser-voir quality of the Montney Fm. included natural fractures, orientation and magnitude of the stress field, and rock brittleness (Davey (2012)). Hydraulic fracture propagation was also observed to be most effective in homogeneous zones, and not necessarily the regions with the most brittle rock. Heterogeneity in the rock was observed to be a detriment to frac-ture growth and the best locations occurred where homogeneous and heterogeneous intervals intersect (Davey (2012)).
Due˜nas (2014) produced a seismic-based characterization of rock fabric and rock compo-sition through the combination of cluster analysis and elastic property inversion of compres-sional wave seismic. Several well logging suites were analyzed to define six multidimencompres-sional clusters. Two clusters associated with the best producing intervals were isolated when cross plotted in λρ vs µρ space. This provided a crucial link between wellbore scale investigations and seismic which can now be used to predictively identify desirable rock fabric properties defined by Davey’s RQI. A constrained sparse spike inversion was performed on the base-line PP seismic data set to produce λρ and µρ volumes (Figure 1.17). Geobodies defined by λρ, µρ polygons mapped homogeneous zones of brittle rock throughout the survey and
demonstrated a methodology that could be used to predictively identify optimal rock facies (Figure 1.18).
Figure 1.17: Inverted λρ and µρ cross sections used to perform compositional analysis (Due˜nas (2014)).
Detailed work with the downhole microseismic data set was performed by Lee (2014). Amplitude ratio analysis of microseisms provided a relatively accurate method of deter-mining composite focal mechanism solutions for clouds of microseismic data. The data set demonstrated that the Montney reservoir is dominated by strike-slip failures during hydraulic stimulation, and concluded that the success of a hydraulic stimulation is largely dependent on naturally occurring weak planes such as fractures
Figure 1.18: Inverted P-Impedance cross section through well 02/02-07 with geobodies of best clusters displayed in blue (Due˜nas (2014)).
1.4 Research Objective
This thesis expands the multidisciplinary work already performed in Pouce Coupe by enhancing the vertical resolution of fracture characterization through the use of amplitude differences in the PS1 and PS2 seismic volumes. The importance of fractures is well defined by the modified RQI and microseismic where fracture propagation has been interpreted to occur along weak planes such as fracture within the reservoir. This previous research produced a need for a robust and detailed fracture mapping methodology. Inversion of the converted wave data will be the primary tool utilized in my research to derive anisotropic elastic properties of the reservoir can be related to crack density. This type of work is based on amplitude variation-with-offset (AVO), which has the advantage of extracting information from the Montney subunit reflectors. Vertical resolution is the most significant limitation of previous traveltime based interpretations that look at the cumulative fracture effect of the entire 300m reservoir. Operators typically target individual subunits, which are expected to have different fracture characteristics due to their heterogeneous nature. This fracture
characterization workflow will be integrated with microseismic, production data, composition analysis, and knowledge of the fields development history to verify its accuracy. Mapping fractures on the subunit scale enables integration with seismic-based compositional analysis to identify the best reservoir targets, as defined by geomechanical analysis.
CHAPTER 2
SEISMIC MODEL OF FRACTURES
The focus of this chapter is to provide background theory on how natural fractures form, their influence on hydraulic stimulations, and develop an anisotropic elastic model of fractured rock that can be used to evaluate seismic based observations within Pouce Coupe. Effective elastic models will be generated to represent a range of geologically realistic fracture sets we expect to see in Pouce Coupe. To accomplish this, we will look into the compliance tensor0s ability to represent different sources of seismic anisotropy to model the effective elastic properties of orthorhombic media that contain up to two fracture sets (Figure 2.1). As previously discussed, a dominant fracture set parallel to σH is well documented in Pouce Coupe, while a secondary orthogonal set is hypothesized.
Figure 2.1: Schematic of an orthorhombic model that combines vertical cracks and horizontal layering. Vertical symmetry planes are determined by the crack orientation (R¨uger (2001)).
Since the inversions performed in the following chapters will utilize amplitude variation with angle (AVA) trends in the seismic, it is important to understand how the effective elastic properties associated with various fracture models impact the seismic response. Reflection
coefficient modeling provides the link between the elastic properties of individual media, and the amplitudes observed in seismic. This anisotropic modeling will demonstrate how the presence of vertical fractures causes well known azimuthal anisotropy that provides the basis for fracture characterization. Due to limitations of Pouce Coupe0s seismic survey design, AVA analysis is restricted to two azimuths that align with the vertical symmetry planes. Therefore, modeling will also focus on these specific cases.
This chapter will conclude with a section on seismic implications that will discuss the challenges, pitfalls, and sources of error associated with the fracture characterization tech-nique performed in this thesis. The largest source of error is due to differences between the anisotropic response of the earth that is recorded by seismic, and the isotropic AVO equations utilized in the inversion. To address this issue, isotropic and anisotropic reflection coefficient modeling will be compared.
2.1 Natural Fracture Background
The occurrence of natural fractures is closely associated with mature over-pressured source rock reservoirs like the Montney due to basic failure theory of porous brittle-elastic rocks (Meissner (1978)). Controls on the stress conditions that cause a rock to break are defined by Mohr0s stress circle, and the failure envelope that is related to a rocks tensile (σ) and shear (τ ) strength (Figure 2.2). On a Mohr diagram, rock failure occurs when the stress circles become tangent to the failure envelope. Rock failures of particular value are tensile fractures, which occurs when the point of tangency is on the negative side of the origin. Elevated pore pressures caused by the thermal maturity of kerogen or fluid injection have the benefit of shifting the stress circles toward tensile failure by reducing the effective stress (Figure 2.3). Equation 2.1 defines effective stress as a function of externally applied total stress (σ) and pore pressure (PP).
Figure 2.2: Mohr-Coulomb graphical representation of stress fields producing tensional and shear rock failures (Meissner (1978)).
Figure 2.3: Mohr-Coulomb graphical representation of fracture failure produced by increasing pore pressure (Meissner (1978)).
Stratigraphic architecture and composition of sedimentary layers also have an impact on controlling the fracture density within a unit (Zahm and Hennings (2009)). Observations in many reservoirs have shown that fracture spacing is proportional to the thickness of the bed (Davey (2012)). However, extent of this stratigraphic control on fractures decreases with an increase in tectonic activity. Since Pouce Coupe is located on the fringe of the Alberta foothills where a high differential in horizontal stress was caused by the Larimide orogeny, we expect fractures to be largely controlled by the regional stress field (Wikel (2011)). 2.2 Induced/Natural Fracture Interaction
In the early development of hydraulic stimulations, treatments were designed using ide-alized models that assume an homogeneous and isotropic earth. Accordingly, well spacing, stage spacing, lateral length, and injected volumes of fluid/proppant were applied uniformly to a wide range of reservoir conditions. The density, orientation, and connectivity of natural fractures is known to be heterogeneous throughout fields and it is becoming increasingly apparent that the interaction between natural and induced fractures is complex (Olson et al. (2012)) (Chuprakov et al. (2011)). Therefore, proper characterization of the natural fracture network and understanding their role during a hydraulic stimulation is crucial for optimal completion design. The end result of a proper fracture characterization is use in a geome-chanical model to predict future fluid flow patterns, and whether new tensile fractures will be created in stimulation.
Geomechanic analysis of induced fracture interaction with natural fractures demonstrates how discontinuities can change the path of fluid flow associated with hydraulic fractures (Chuprakov et al. (2011)) (Olson et al. (2012)) (Davey (2012)) (Figure 2.4). Natural fractures often act as planes of weakness in a reservoir, but can also act as barriers to induced fractures. The extent to which fractures act as a conduit or barrier frequently depends on the degree of diagenetic cementation/mineralization, prevailing stress conditions, and the orientations
Figure 2.4: Simulated hydraulic fracture in the presence of perforations that act as permeable conduits and glass plates that represent cemented natural fractures (Olson et al. (2012)).
2.3 Fracture Compliance Methodology
The fracture-characterization model used in this thesis is based on effective media theory to derive the overall impact of fractures on the elastic properties, even though fracture size is small compared to a typical seismic wavelength (Tsvankin and Grechka (2011)). This theory assumes that the displacement field for a long seismic wavelength is nearly constant inside a representative volume element (RVE) that may contain numerous small-scale heterogeneities such as fractures (Tsvankin and Grechka (2011)). This enables the exact compliance tensor (s) that is heterogeneous on the micro scale, to be represented as an effective stiffness tensor (se) that has the same average elastic properties of the original medium on the scale of seismic wavelength (Tsvankin and Grechka (2011)).
Equation 2.2 demonstrates the relationship of the effective compliance tensor to stress (τ ) and strain () by Hooke0s law. Since fractures can be represented as sources of extra strain in relation to the background rock, the effective compliance tensor is equivalent to the sum of the background rock compliance (sb) and the fracture compliance (∆s) (equation
2.3)(Tsvankin and Grechka (2011)).
ij = se ijklτkl (2.2)
se= sb+ ∆s (2.3)
While many different fracture theories can be used for modeling fracture compliance, we will focus on the “scalar crack” model (Tsvankin and Grechka (2011)). This technique has the benefit of simplicity since ∆s is defined by three scalars (α1, α2, α3) that represent the elastic properties of three orthogonal fracture sets (equation 2.4) (Schoenberg and Sayers (1995)). Equation 2.4 represents the “crack-density tensor”, where α1 corresponds to a set of vertical fractures normal to the x1 axis, α2 to a set of vertical fractures normal to the x2 axis, and α3 to a set of horizontal fractures normal to the x3 axis which is defined to be vertical. Figure 2.5 show a schematic of how this method utilizes the α0s to effectively model a medium with orthogonal sets of fractures by assuming each fracture set to be independent from the others.
∆s = α1 0 0 0 0 0 0 α2 0 0 0 0 0 0 α3 0 0 0 0 0 0 α2 + α3 0 0 0 0 0 0 α1 + α3 0 0 0 0 0 0 α1 + α2 (2.4)
Application of this modeling technique to Pouce Coupe Field will utilize VTI symmetry of the background rock due to the horizontally laminated nature of Montney and two α co-efficients that represent vertical fracture sets. Within Pouce Coupe, stress conditions, image logs, and microseismic give no indication that horizontal fractures are present. Therefore α3 can be set to zero.
Equations presented below will utilize both the compliance tensor (s) and the stiffness tensor (c), which have a simple inverse relationship (2.5). The stiffness tensor proves to be useful due to well known equations in linear elastic theory that relate the anisotropic velocity
Figure 2.5: Schematic model of a medium with multiple sets of parallel fractures (A) as the sum of independent parallel fracture sets (B and C) (Schoenberg and Sayers (1995)).
defined by equation 2.6, with equations 2.7 through 2.12 showing how each element in the tensor is related to density (ρ), vertical velocities, and anisotropic coefficients. Note that VP 0 and VS0 represent the vertical P and S wave velocities respectively, and the Thomsen parameters (γ, , and δ) characterize the strength of velocity anisotropy. γ is close to the fractional difference between horizontal and vertical velocity of SH-waves. is the fractional difference between the horizontal and vertical compressional velocity. δ is a function of several stiffness tensor coefficients that describes P and S wave anisotropy at oblique incidence angles (Thomsen (1986)). S = C−1 (2.5) [S(b)]−1 = C(b)= C11 C12 C13 0 0 0 C12 C11 C13 0 0 0 C13 C13 C33 0 0 0 0 0 0 C55 0 0 0 0 0 0 C55 0 0 0 0 0 0 C66 (2.6) C11(b) = (1 + 2)ρVP 02 (2.7) C33(b) = ρVP 02 (2.8) C55(b)= ρVS02 (2.9) C66(b) = (2γ + 1)ρVS02 (2.10) C12(b)= C11− 2C66 (2.11) C13(b) = −C55+ q C2 55− (2δ + 2)(C33C55) − (2δ + 1)(C332 ) (2.12) 2.4 Numeric Modeling
The objective of numeric modeling is to obtain the azimuthal PS reflection coefficients of the Montney D/Montney C interface for various geologically realistic fracture models.
seismic, and illustrate the project0s feasibility despite notable pitfalls. Figure 2.6 summarizes the numeric modeling workflow which is based on the concept of scalar fractures and effective compliance presented in the preceding section.
Figure 2.6: Workflow for modeling the seismic response of fractures.
The first step in the workflow is to populate the stiffness tensor of the unfractured back-ground rock (sb) using well logs, estimates of the Thomsen parameters and equations 2.7 through 2.12. Since sonic pulses recorded by logging tools follow the fastest path from the source to the receiver, and fractures have the effect of lowering wavefield velocities, sonic logs are representative of the unfractured background rock (Asquith and Krygowski (2004)). To
obtain the slow shear velocity the full wave-field of a dipole sonic log needs to be processed, and this was not performed in Pouce Coupe. Since well 13-12 is the only well in the field with density, shear sonic, and compressional sonic logs through the entire reservoir, it will be used to obtain average reservoir properties of the Montney C and D subunits. Thomsen parameters can be obtained through a combination of P-wave NMO velocity analysis, core analysis and modern logging tools such as Schlumberger0s sonic scanner. Note that velocity anisotropy measurements from core and logs are not at the seismic scale. Unfortunately, Pouce Coupe does not have the necessary data available to determine the Thomsen param-eters, so values are estimated from literature where the Ft. Union siltstone will be used as a proxy for the Montney (γ=0.040, =0.045, δ=-0.045) (Thomsen (1986)). Thomsen parameters found in literature demonstrate a significant amount of variability between silt-stone/shale formations, and the values used in this thesis may underestimate the anisotropy associated with the background rock. It is therefore instructive that further information be obtained to ensure accurate measurements of the VTI parameters are used in the modeling to reduce uncertainty. Table 2.1 summarizes the unfractured reservoir properties of Montney C and D used to generate the background compliances.
Table 2.1: Properties of the unfractured Montney C/D subunits obtained from well 13-12 logs and literature
Property Montney D Montney C
ρ (g/cc) 2.67 2.60 VS0 (m/s) 3066 2959 VP 0(m/s) 5166 4861 0.045 0.045 γ 0.04 0.04 δ -0.045 -0.045
Four fracture models are generated using various combinations of α to demonstrate a range of geologic scenarios that can be encountered within the Montney. These α-values
understand, fractures will only be simulated in the Montney C half-space.
Model 1: No fractures (α1 = α2 = 0)
Model 2: One parallel set of fractures with low crack density (α1 = 0.03, α2 = 0.0) Model 3: One parallel set of fractures with high crack density (α1 = 0.07, α2 = 0.0) Model 4: Two orthogonal fracture sets of equal crack density (α1 = α2 = 0.07)
At this point, we have sufficient input to calculate the effective compliance tensors uti-lizing equation 2.3. Since the background compliance has VTI symmetry, the addition of at least one vertical fracture set generates an effective medium with orthorhombic symme-try. To condition these effective compliance tensors for the reflection coefficient modeling software, the Tsvankin parameters must be calculated. We previously used the widely ac-cepted notation developed by Thomsen (1986) to represent the VTI elastic properties of the Montney background (VP 0, VS0, γ, , and δ). Tsvankin (1997) extended this notation to orthorhombic symmetry by defining two vertical velocities (VP 0 and VS0) and seven dimen-sionless anisotropy parameters (γ(1), (1), δ(1), γ(2), (2), δ(2), and δ(3)). The superscripts (1) and (2) refers to the x1 and x2 axes, which defines the normal direction of the [x2, x3] and [x1, x3] symmetry planes respectively. The definition of these parameters is presented below in terms of the stiffnesses cij and density (equations 2.13 - 2.21).
VP 0 = s C33 ρ (2.13) VS0= s C55 ρ (2.14) (1) = c22− c33 2c33 (2.15) δ(1) = (c23+ c44) 2− (c 33− c44)2 2c33(c33− c44) (2.16) γ(1) = c66− c55 2c55 (2.17)
(2) = c11− c33 2c33 (2.18) δ(2) = (c13+ c55) 2− (c 33− c55)2 2c33(c33− c55) (2.19) γ(2) = c66− c44 2c44 (2.20) δ(3) = (c12+ c66) 2− (c 11− c66)2 2c11(c11− c66) (2.21) Table 2.2 represents the Tsvankin parameters calculated for the four Montney C fracture models. Note that models 1 and 4 are effectively VTI since γ(1) = γ(2) = γ, (1) = (2) = , δ(1) = δ(2) = δ, and γ(3) =0.
Table 2.2: Tsvankin Parameters for the Four Montney C Fracture Models
Parameter Model 1 Model 2 Model 3 Model 4
VP 0(m/s) 4861 4839 4820 4787 VS0(m/s) 2959 2823 2667 2667 γ(1)) 0.040 0.036 0.032 -0.057 (1) 0.045 0.041 0.038 -0.171 δ(1) -0.045 -0.049 -0.051 -0.212 γ(2) 0.040 -0.012 -0.068 -0.057 (2) 0.045 -0.074 -0.170 -0.171 δ(2) -0.045 -0.136 -0.211 -0.212 δ(3) 0.000 0.135 0.316 0.000
The final and most important step in the modeling workflow is to calculate the PS re-flection coefficients for each fracture model in azimuths parallel to the x1 and x2 axes. This will demonstrate that seismic amplitudes are highly sensitive to azimuthal velocity varia-tions associated with vertical fracture systems (Xu and Tsvankin (2006)). In models where fractures are present, these axes coincide with the two orthogonal symmetry planes of the system. Reflection coefficients provide insight into understanding the amplitude signatures in reflected wave seismic in the presence of fractures. Modeling was accomplished using code
for arbitrarily anisotropic media.
Figure 2.7 presents the reflection coefficients calculated at the interface between the Montney D halfspace with VTI symmetry, and the various representations of the Montney C halfspace that exhibits VTI or orthorhombic symmetry depending on the assumed fracture model. Each model is uniquely colored, and uses circles and squares to represent reflection coefficients associated with S/R azimuths in the x1 and x2 directions respectively.
Figure 2.7: PS reflection coefficients of the four fracture models associated with the Montney C/D interface along azimuths aligned with the x1 and x2 axes. Note that the reflection coefficients from Model 3 (x1)=Model 4 (x1)=Model 4 (x2) are nearly overlain.
Two key observations can be made from Figure 2.7 that provide insight into how the presence of fractures will impact the seismic response:
1: There is no azimuthal variation in the reflection coefficients of models 1 and 4, even though both models exhibit different AVA behaviour. Since the fracture
character-ization methodology applied in this thesis requires azimuthal variation in the AVA signature, the inversion performed in chapter 4 will not be able to identify orthogonal fractures sets.
2: Models 2 and 3 demonstrate that media with one dominant fracture set display greater amounts of azimuthal variation of PS-wave reflection coefficients with increasing values of crack density. This concept is well established in literature and provides the basis for the fracture characterization presented in chapters 4 and 5.
Another way to quantify these modeled results is through use of the shear wave splitting coefficient γ(S), which can be defined in terms of the Tsvankin parameters (γ(1) and γ(2)) or the fractional difference between the fast (VS1) and slow (VS2) vertical shear wave velocities (equation 2.22) (Tsvankin (2012)). This parameter has several benefits that include being directly related to previous Pouce Coupe SWS analysis where up to 8% SWS was observed (Steinhoff (2013)). γ(S) is also noted by Bakulin et al. (2000) to be close to the crack density and therefore provides an excellent way to quantify the fractures we plan to characterize. Table 2.3 summarizes the γ(S) results for the four fractures models.
γ(S) = |γ (1)− γ(2)| 1 + 2γ(2) ≈ VS1− VS2 VS2 (2.22)
Table 2.3: Shear-Wave Splitting Coefficients for the Four Montney C Fracture Models
Fracture Model γ(S)
Model 1 0
Model 2 4.7%
Model 3 8.8%
Model 4 0
limited to the two vertical symmetry planes associated with a fracture network with one dominant orientation. This inversion will generate two azimuthally dependent shear velocity volumes whose differences are expected to be related to the fracture network. It is therefore necessary to demonstrate how the anisotropic AVA response of each seismic volume relates to the isotropic AVA equations used in the inversion. This will help us understand the final results and the associated error/uncertainty.
Without going into detail about the inversion algorithm that will be presented in chapter 4, isotropic reflection coefficient equations are utilized in the inversion to determine elastic properties that provide the best match to the input AVA response. Since isotropic equations are unable to precisely match the anisotropic AVA character over the entire incident angle range, the inversion will generate isotropic properties (Vp, Vs, and ρ) that provide the closest match. Therefore, the isotropic equations must be capable of producing AVA trends that are reasonably close to the anisotropic reflection coefficients calculated in both the [x1, x3] and [x2, x3] symmetry planes for the fracture signatures defined in the previous section to be identified.
Figure 2.8 compares the anisotropic reflection coefficients for fracture model 3 (red) with isotropic modeling that only uses the average log values from well 13-12 as input (cyan). We can see that this isotropic equation does not provide a good match to the anisotropic character in either symmetry plane. However, when we scale the shear velocity in the isotropic equation by 0.97 (blue) and 1.02 (green), we obtain a very reasonable match to the anisotropic AVA character in both symmetry planes up to an incidence angle of ≈ 30◦. This indicates that the isotropic equations are capable of reasonably approximating the initial trend of the anisotropic reflection coefficients with sufficient accuracy. The deviation of the isotropic coefficient after 35◦ degrees is partially related to increasing errors in the approximations used to calculate the reflection coefficients (J`ılek (2001)).
It is very important to note that the elastic properties in the isotropic models used to fit the anisotropic data do not correspond to any parameter previously used to represent an
Figure 2.8: Comparison between anisotropic and isotropic PS reflection coefficient equations used to represent fracture model 3.
isotropic, VTI, or orthorhombic medium. They are simply elastic properties that provide the best fit of an isotropic equation to anisotropic response. To avoid confusion, these parameters will be defined as V(1)P , V(1)S , and ρ(1) to define isotropic elastic properties derived in the [x
2, x3] symmetry plane (parallel to the fractures) and V
(2) P , V
(2)
S , and ρ(2) to define isotropic elastic properties derived in the [x1, x3] plane (orthogonal to the fractures).
Modeling results also indicate that the use of isotropic equations will tend to underes-timate the difference in the inverted fast and slow shear velocities due to the anisotropic reflection coefficient responses of both symmetry planes converging at large angles and the inability of the isotropic equations to match this behavior. Table 2.3 demonstrated that model 3 exhibited 8.8% SWS. If we use the percentage shear velocity difference between V(1)S and V(2)S as a proxy for the SWS parameter, we can roughly compare how sensitive the application of isotropic equations in the two symmetry planes is to azimuthal anisotropy. Since there is only a 5.1% difference in shear velocity, we can see that the isotropic equations provide a lower resolution measurement of the fracture density, but can still help identify large fracture anomalies.
CHAPTER 3
MULTICOMPONENT SEISMIC DATA PROCESSING
A need to reprocess the Pouce Coupe multicomponent seismic data set was demonstrated when interpretational focus was shifted from traveltime based techniques such as SWS to-wards amplitude sensitive methodologies including AVO inversion. The primary objective of previous processing efforts was to preserve time delays associated with the difference between fast and slow shear wave modes. SWS analysis performed by Steinhoff (2013) was enabled by this last effort, which resulted in a significant improvement of data quality, and a high NRMS repeatability necessary for time lapse work. A consequence of this previous processing effort was the alteration of relative amplitudes by non AVO friendly processes, and the PS2 wave field was not captured above the Montney. This is apparent simply by looking at the large amplitude and seismic character variations within the fast and slow converted wave data set (Figure 3.1). This level of amplitude variation would severely inhibit any AVO work.
This chapter will summarize basic PS converted wave processing theory related to key steps performed by Sensor Geophysical to produce AVO compliant PS1 and PS2 converted wave seismic volumes. Multicomponent seismic processing is a complex subject that has significant deviations from workflows used on conventional compressional seismic. The com-plexity is due to several unique problems related to: binning traces, polarity reversals, re-ceiver misalignment, statics, non hyperbolic move out, lack of S/R reciprocity, high noise levels, shear wave splitting, and the vector alignment of horizontal geophones with SV wave particle displacement. Readers are referred to Hardage et al. (2011) for a more thorough de-scription of a processing workflow. Significant advancements in multicomponent processing have been made in recent years to address these challenges and produce high quality data
