UNDERSTANDING ROCK QUALITY HETEROGENEITY

OF MONTNEY SHALE RESERVOIR, POUCE COUPE

FIELD, ,

by Claudia Due˜nas c Copyright by Claudia Due˜nas2014 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: Claudia Due˜nas

Signed: Dr. Thomas L. Davis Thesis Advisor

Golden, Colorado Date

Signed: Dr. Terence K. Young Professor and Head Department of Geophysics

ii ABSTRACT

Understanding the lateral heterogeneity of unconventional plays prior to hydraulic frac- turing is important for hydrocarbon production and recovery. Lateral and vertical variability can be affected by composition and textural variation of the rock, which define the rock qual- ity. To characterize the lateral and vertical heterogeneity of rock quality (composition) of the Montney Shale reservoir at Pouce Coupe, Alberta at different scales I conducted a multi- attribute analysis of wells logs integrated with post-stack and pre-stack inversion of a baseline multicomponent seismic survey. Cluster analysis was performed in four wells using the well logs that are most affected by composition. The cluster analysis provides more representative upscale input parameters for reservoir characterization that can be compared with seismic results. The result of this cluster analysis has indicated a lateral variation of composition of the unit C to the east side of the area, where six clusters were chosen and two of them have good petrophysical rock properties that were tied with core data. Post-stack and pre-stack inversions of the baseline of the multicomponent seismic data were performed using constrained sparse spike inversion (CSSI). Pre-stack results shows similar results for the P-impedance, however, there is an improvement in the accuracy of the estimated P-impedance from the pre-stack CSSI (compared to well log P-impedance). The results of P-impedance and S-impedance show the same strong change on the east side of the survey that was detected with the cluster analysis. Crossplots of elastic properties such as Lambda-rho and Mu-rho combined with the results of cluster analysis helped to identify the areas of better rock quality in the 3D seismic. The integration of this heterogeneity analysis with the production profile of the two horizontal wells in the area shows that the lithology has a major influence on the rock quality of the Montney interval. The combined interpretation of this work with an understanding

iii of the natural fracture system and the stress state of the reservoir can provide a rock quality index (RQI). This RQI can aid in future exploration and operational development of the Montney play and other shale reservoirs worldwide.

iv TABLE OF CONTENTS

ABSTRACT ...... iii

LIST OF FIGURES ...... viii

LIST OF TABLES ...... xvi

LIST OF SYMBOLS ...... xvii

ACKNOWLEDGMENTS ...... xviii

DEDICATION ...... xix

CHAPTER 1 INTRODUCTION ...... 1

1.1 Objectives ...... 4

1.2 Geology ...... 4

1.2.1 Reservoir units ...... 4

1.2.2 Regional tectonics ...... 7

1.2.3 Petroleum system ...... 9

1.3 Time-lapse multicomponent seismic, microseismic data ...... 10

1.3.1 Surface seismic acquisition ...... 10

1.3.2 Seismic processing ...... 11

1.3.3 Microseismicity ...... 13

1.4 Available thesis data ...... 15

1.4.1 Production and completion data ...... 17

1.5 Previous work ...... 19

1.5.1 Geomechanical characterization ...... 26

v 1.6 Methodology of this work ...... 29

CHAPTER 2 CLUSTER ANALYSIS ...... 31

2.1 Theory and workflow ...... 31

2.1.1 Example in literature ...... 35

2.2 Cluster Analysis of composition in four wells of Pouce Coupe ...... 36

2.2.1 Correction and edition of well logs ...... 37

2.2.2 Definition of clusters in the master well log ...... 38

2.2.3 Integration of cluster analysis and core data ...... 40

2.2.4 Cluster tagging ...... 42

2.2.5 Interpretation ...... 46

2.3 Summary ...... 47

CHAPTER 3 INVERSION ...... 48

3.1 Method and theory ...... 48

3.1.1 Convolutional model ...... 49

3.1.2 Constrained Sparse Spike Inversion (CSSI) ...... 49

3.1.3 Post-stack and pre-stack inversion ...... 50

3.1.4 The wavelet ...... 55

3.2 Available seismic and well log data ...... 56

3.3 PP data interpretation ...... 56

3.4 Building the low-frequency impedance model ...... 61

3.5 Post-stack inversion ...... 62

3.5.1 Well tie ...... 63

3.5.2 Wavelet estimation ...... 68

vi 3.5.3 Filtering the model ...... 68

3.5.4 Inversion parameters ...... 69

3.6 Pre-stack inversion ...... 73

3.6.1 Angle domain ...... 73

3.6.2 Well tie ...... 73

3.6.3 Angle - dependent wavelet estimation ...... 81

3.6.4 Inversion parameters ...... 81

3.7 Results and comparison ...... 82

3.7.1 Comparison of post-stack and pre-stack inversion ...... 94

3.8 Summary ...... 100

CHAPTER 4 PREDICTING ROCK QUALITY FROM PRE-STACK ATTRIBUTES ...... 101

4.1 Lambda-rho and Mu-rho crossplots ...... 102

4.1.1 Well response and clusters ...... 103

4.1.2 Seismic LMR response ...... 104

4.2 Integration with microseismicity and production data ...... 109

4.3 Summary ...... 116

CHAPTER 5 CONCLUSIONS AND RECOMMENDATIONS ...... 117

5.1 Recommendations ...... 118

REFERENCES CITED ...... 119

APPENDIX - CLUSTER ANALYSIS ...... 123

vii LIST OF FIGURES

Figure 1.1 Map of strata in the Western Canada Sedimentary Basin showing the locations of Montney oil fields, conventional gas fields and tight gas plays. Pouce Coupe area is shown as a black start. Modified from Zonneveld (2011)...... 2

Figure 1.2 Reservoir characterization of rock quality of Montney Shale and methodology. Integrated study and analysis at different scales...... 3

Figure 1.3 Triassic in the Peace River Arch region. Pouce Coupe Field is on the border of (BC) and Alberta and is represented by the colored formations in the Talisman BC chart section (courtesy of Talisman Energy)...... 5

Figure 1.4 Schematic simple WestEast section of Lower Triassic Montney Formation facies ...... 6

Figure 1.5 Type log of the Triassic Montney of the southern Pouce Coupe Field. The red curve is the gamma ray log. Modified from Steinhoff (2013). . . . . 7

Figure 1.6 Comparison of predominant mineralogy of shales reservoir worldwide in a ternary plot. VacaMuerta Consortium. Sonnenberg, 2013 ...... 8

Figure 1.7 The stratigraphic framework at Farrell Creek and Pouce Coupe. Maximum regressive surfaces are defined by red lines while maximum flooding surfaces are defined by green lines (courtesy of Lindsay Dunn, Talisman Energy Inc.). Modified from Davey (2012) ...... 9

Figure 1.8 Pouce Coupe time-lapse, multicomponent surface seismic and field operations timeline. The map shows the horizontal wells hydraulically stimulated (blue and green color). Red circles are the vertical wells. Orange shows wells with microseismicity receivers. Modified from Atkinson (2010) ...... 11

Figure 1.9 Multicomponent seismic survey acquisition layout of Pouce Coupe time-lapse. 5km2 patch centered over horizontal wells 02/02-07 and 02/07-07. Modified from Atkinson (2010)...... 13

Figure 1.10 Workflow of seismic processing (2013) PP data of Pouce Coupe. Processed by Sensor Geophysical...... 14

viii Figure 1.11 Pouce Coupe iline 4 baseline survey - PSTM of PP data. Comparison of version 2012 and 2013 PSTM in PP data. Red box show the zone of Montney Formation...... 15

Figure 1.12 Results of the downhole microseismic events on both treatment wells, showing the total combined event location from the individual 5 stage of each treatment...... 16

Figure 1.13 Map of the Pouce Coupe 3D seismic outline (orange), vertical wells used in the cluster analysis (black points) and the important horizontal wells (blue and red)...... 17

Figure 1.14 Pouce Coupe wells used in this project. Some wells are used in different stages of the project...... 17

Figure 1.15 Production data decline of the horizontal wells. The wells drilled in Montney unit C have higher initial production...... 20

Figure 1.16 Difference in sum of monitor 2 negative shifts and monitor 1 negative shifts from 2000-2200 ms, showing areas of increased azimuthal anisotropy. Indicated in red is the outline of the anisotropy present in the baseline. Calculated by Atkinson (2010)...... 21

Figure 1.17 Shear wave splitting map from the baseline survey (Steinhoff, 2013). . . 22

Figure 1.18 Shear wave splitting map from the monitor 1 survey after the 02/02-07 well had been stimulated. (Steinhoff, 2013)...... 24

Figure 1.19 Shear wave splitting map from the monitor 2 survey after the 02/07-07 well had been stimulated. (Steinhoff, 2013)...... 25

Figure 1.20 On the left is monitor 1 minus baseline of the SWS magnitudes. On the right is monitor 2 minus baseline of the SWS magnitudes. Also plotted is the production from each stage for each well indicated by percentage. (Steinhoff, 2013)...... 26

Figure 1.21 Correlation of microseismicity with RQI. First panel is RQI, second panel is Gamma Ray and third panel is mechanical stratigraphy. (Davey, 2012) ...... 28

Figure 1.22 Correlation of RQI log with the SWVA in the baseline. Hydraulic energy will preferentially propagate to more homogeneous areas of the reservoir. (Davey, 2012) ...... 29

ix Figure 1.23 Workflow followed in this thesis research. Numbers after heading denote the chapter...... 30

Figure 2.1 Scatterplot of the first three principal components form a principal components analysis of multivariate log data. Schlumberger ...... 33

Figure 2.2 Box and whisker plot by class of 10 clusters showing the variability of GR log using Techlog software ...... 33

Figure 2.3 Fall-off plot of HRA of 6 clusters showing 200 runs in horizontal axes and distance in vertical axes using Techlog software...... 34

Figure 2.4 Cluster analysis workflow using a 2-D crossplot of neutron and density porosity. A) Determine how many cluster are represented by the data. B) Example of 4 clusters, where the centroids are the black points. The compliance distance would be the distance from the square data point to the centroid of the red cluster. Modified from Silvis (2011)...... 35

Figure 2.5 Map of the Pouce Coupe 3D seismic outline (green), vertical wells used in the HRA analysis (black points) and important horizontal wells (blue, red)...... 36

Figure 2.6 Histogram of NPHI of four wells before and after enviromental and matrix correction in the interval Doig to Belloy...... 37

Figure 2.7 Histogram of GR log of four wells before and after normalization in the interval Doig to Belloy...... 38

Figure 2.8 Histogram of photoelectric log of four wells before and after normalization in the interval Doig to Belloy...... 39

Figure 2.9 Histogram of density log of four wells in the interval Doig to Belloy. This log did not required normalization...... 39

Figure 2.10 Box and whisker plot for 6 clusters in the logs Neutron-Porosity, GR, RHOZ and PEFZ...... 40

Figure 2.11 Master well log 13-12 and HRA reference model of cluster analysis Six clusters. First column is the intervals of Montney Formation. Panels from left to right: Neutron-Porosity, GR, RHOZ, PEFZ and cluster composition. The legend is in the right part, blue to red color. The core interval is located in the red box. The units C and D are highlighted in the dashed black box...... 43

x Figure 2.12 Neutron porosity vs. density plot of the results of master well log 13-12 in the Montney interval. The legend is the color of the six clusters identified...... 44

Figure 2.13 Log correlation of the four wells of Pouce Coupe. Master well is 13-12 and three wells were tagged. The last panel is the compliance...... 44

Figure 2.14 Cluster analysis results in the units C and D of Montney of the four wells. Cluster colors are showed in the legend - lower part of the figure. . 45

Figure 3.1 Categories of seismic inversion. Constrained sparse spike is a trace-based seismic inversion...... 50

Figure 3.2 P-impedance seismic inversion workflow using CSSI. Colors represent input, process and product...... 51

Figure 3.3 Simultaneous inversion of pre-stack seismic data. Colors represent input, process and product. (CSSI algorithm)...... 53

Figure 3.4 Map of the Pouce Coupe 3D seismic grid outline and the three vertical well used in the seismic inversion...... 56

Figure 3.5 Arbitrary line of PP stack of Pouce Coupe data. Montney reservoir is located at 1.2-1.34 seconds of two way time (TWT). Gamma ray log is in black color...... 57

Figure 3.6 Statistical wavelet of PP data in window 1100-1400 ms. Amplitud and phase spectrum are in the lower part...... 58

Figure 3.7 Arbitrary line of PP stack and interval velocities form seismic velocities of Pouce Coupe data. Interval velocity of Montney is 5500 m/s...... 58

Figure 3.8 Arbitrary line of PP stack of Pouce Coupe data. Montney reservoir is located at 1.2-1.34 seconds of two way time (TWT). Green line is Gamma ray log . Seismic interpretation of PP data...... 59

Figure 3.9 Structural time map of Top of Belloy Formation. Arbitrary line is the black line over the map...... 60

Figure 3.10 Structural time map of Top of Montney C unit. Weak peak response in the seismic...... 60

Figure 3.11 Structural time map of Top of Montney D unit. Zero crossing in the seismic...... 60

xi Figure 3.12 Structural time map of Top of Montney E unit. Zero crossing in the seismic very continuous...... 61

Figure 3.13 Structural time map of Top of . Peak response very continuous laterally...... 61

Figure 3.14 Locally weighted interpolation method using to build the low frequency model. Modified from Lorenzen (2000) ...... 62

Figure 3.15 Arbitrary line of P-impedance with a high cut of 80 Hz. Belloy and Doig Formation are the limits...... 63

Figure 3.16 Arbitrary line of P-impedance model with a high cut of a)8 Hz and b)10Hz...... 64

Figure 3.17 Arbitrary line of S-impedance model with a high cut of a)80 Hz and b)8Hz...... 65

Figure 3.18 Arbitrary line of density model with a high cut of a)80 Hz and b)8Hz. . . 66

Figure 3.19 Seismic well tie of wells 13-12 and 5-14 using a Ricker wavelet of 45 Hz. The response of top of Belloy is a trough...... 67

Figure 3.20 Synthetic seismogram of well 13-12. Seismic well tie using PP baseline full offset...... 68

Figure 3.21 Synthetic seismogram of well 09-07. Seismic well tie using PP baseline full offset...... 69

Figure 3.22 Synthetic seismogram of well 02-07. Seismic well tie using PP baseline full offset...... 70

Figure 3.23 Wavelet extraction in PP data full offset. The colors represent the wavelets extraction of the vertical wells used in the seismic inversion. . . 71

Figure 3.24 Filter for low frequency model of post-stack seismic inversion. (10 to 80 Hz)...... 72

Figure 3.25 CDP Gathers at well locations showing the four ranges of angles selected in the stacks. Courtesy of Sensor Geophysical (2013)...... 74

Figure 3.26 Wiggle-trace display of 0-9 degree (purple) , 9-18 degree (green) before horizon alignment and 9-18 (black) after horizon alignment...... 75

xii Figure 3.27 Wiggle-trace display of 0-9 degree (purple) , 18-27 degree (green) before horizon alignment and 18-27 (black) after horizon alignment...... 75

Figure 3.28 Wiggle-trace display of 0-9 degree (purple) , 27-36 degree (green) before horizon alignment and 27-36 (black) after horizon alignment...... 76

Figure 3.29 Cross-correlation coefficient in a window of Doig to Belloy Formation for the seismic alignment of pre-stack data...... 77

Figure 3.30 Shear slowness vs compressional slowness of wells 13-12 and 5-14 in the Fernie to Belloy Formation interval. The color is the Gamma Ray. . . . . 77

Figure 3.31 Seismic well tie of well 13-12 with four stacks of PP data. First panel is the seismic, second panel is the synthetic seismogram, third panel is the cross correlation of seismic and synthetic...... 78

Figure 3.32 Seismic well tie of well 09-07 with four stacks of PP data. First panel is the seismic, second panel is the synthetic seismogram, third panel is the cross correlation of seismic and synthetic...... 79

Figure 3.33 Seismic well tie of well 02-07 with four stacks of PP data. First panel is the seismic, second panel is the synthetic seismogram, third panel is the cross correlation of seismic and synthetic...... 80

Figure 3.34 Angle-dependent wavelet estimation of four stacks of PP data for three wells...... 82

Figure 3.35 Final angle-dependent wavelet of four stacks of PP data. The colors represent the wavelets extraction of the vertical wells used in the seismic inversion...... 84

Figure 3.36 Filter for low frequency model of pre-stack seismic inversion. Bandpass filter values are (8-80 Hz) ...... 85

Figure 3.37 Arbitrary line of P-impedance inverted in post-stack inversion (bandlimited (10-80 Hz))...... 86

Figure 3.38 Comparison in arbitrary line of P-impedance inverted in post-stack inversion (lowpass filtered (0-10 Hz)) and low frequency model of P-impedance...... 87

Figure 3.39 Arbitrary line of P-impedance inverted in pre-stack inversion (bandpass limited (8-80 Hz))...... 88

xiii Figure 3.40 Comparison in arbitrary line of P-impedance inverted in pre-stack inversion (lowpass filtered (0-8 Hz)) and low frequency model of P-impedance...... 89

Figure 3.41 Arbitrary line of S-impedance inverted in pre-stack inversion (bandpass limited (8-80 Hz)) and full band (0-80Hz)...... 90

Figure 3.42 Comparison in arbitrary line of S-impedance inverted in pre-stack inversion (lowpass filtered (0-8 Hz)) and low frequency model of S-impedance...... 91

Figure 3.43 Crossplot of well logs and pseudo logs of P-impedance inverted in post-stack inversion. Bandpass limited and full band...... 92

Figure 3.44 Crossplot of well logs and pseudo logs of P-impedance inverted in pre-stack inversion. Bandpass limited and full band...... 93

Figure 3.45 Comparison of arbitrary line of P-impedance in pre-stack inversion and post-stack inversion full-band frequency (0-80 Hz)...... 95

Figure 3.46 Well log comparison of original logs, post-stack inversion and pre-stack inversion in the wells used in the inversion (bandpass limited (8-80 Hz)). . 95

Figure 3.47 Well log comparison of original logs vs post and pre-stack inversion full-band frequency 0-80 Hz...... 96

Figure 3.48 Well log crossplot of post-stack inversion and pre-stack inversion in the wells used in the inversion in the area of interest bandpass limited and full-band frequency...... 97

Figure 3.49 P-impedance extraction in the Montney D, upper Montney C and lower Montney C in post-stack and pre-stack inversion...... 98

Figure 3.50 S-impedance extraction in the Montney D, upper Montney C and lower Montney C of pre-stack inversion...... 99

Figure 4.1 LambdaRho versus MuRho crossplot comparing various shales and carbonates from Western Canada to the Barnett Shale ()...... 103

Figure 4.2 LambdaRho versus MuRho crossplot of well 13-12 in Montney and Triassic interval. Results of cluster analysis in Montney C and D unit. . 104

Figure 4.3 Arbitrary line with average velocities for depth conversion in Montney zone. Higher velocities are yellow color...... 105

xiv Figure 4.4 2D section view of depth conversion of Montney C and D units. (Modified from Lee, 2014) ...... 106

Figure 4.5 Arbitrary line of Lambda-rho and mu-rho in the horizontal wells 07-07 and 02/07-07...... 107

Figure 4.6 Arbitrary line of Lambda-rho and mu-rho in the horizontal wells 02/02-07. The well 13-12 is colored with Lambda-rho log...... 108

Figure 4.7 LMR crossplot of well log (13-12) and pseudo well log in seismic data. Brittleness area are in the black polygon...... 110

Figure 4.8 Arbitrary line of P-impedance through 02/02-07 well and geobodies of clusters 1 and 2...... 111

Figure 4.9 Arbitrary line of P-impedance through 02/07-07 well and geobodies of clusters 1 and 2...... 111

Figure 4.10 Thickness map of geobodies of clusters 1 and 2 extracted from LMR crossplot...... 112

Figure 4.11 Microseismicity events in horizontal wells 02/07-07 and 02/07-07 in the interval Montney units C and D...... 112

Figure 4.12 Thickness map of geobodies of clusters 1 and 2 extracted from LMR crossplot...... 113

Figure 4.13 Combined map: Thickness map of clusters 1 and 2 from the LMR crossplot and the microseismic events in Montney units C and D; MS events of Stage 1 in blue, Stage 2 in red, Stage 3 in cyan, Stage 4 in orange, and Stage 5 in green...... 114

Figure 4.14 P-impedance inversion results in units C and D of the Montney (RMS extraction) showing a cutoff of P-impendace ¡1.29e7 kgs−1m−2. 13-12 well is the circle in red color in the maps...... 115

Figure A.1 HRA X-Plot showing two principal components and the associated cluster centroids. In total are 6 clusters...... 123

Figure A.2 HRA Silhouette plot of the final choice of 6 clusters of Pouce Coupe. Each color represent a cluster...... 124

Figure A.3 HRA Fall-off of 6 clusters. At 75 runs the cluster converged to a distance of 2275, which explain that 6 cluster is the correct number. . . 124

xv LIST OF TABLES

Table 1.1 Montney reservoir properties compared with other North American shales. Pouce Coupe Field is located in the Montney core area (information from Talisman Energy Inc.) (Steinhoff, 2013) ...... 8

Table 1.2 Pouce Coupe time-lapse, multicomponent seismic acquisition parameters. . 12

Table 1.3 Abbreviation legend of Figure 1.14 ...... 18

Table 1.4 Pouce Coupe Field horizontal well hydraulic stimulation parameters. . . . 19

Table 1.5 Spinner log data from the two horizontal completed wells ...... 19

Table 2.1 XRD analysis for samples of the well 13-12. Mineral by constitution is by weight % ...... 41

Table 2.2 Microprobe analysis for sample of the well 13-12 ...... 41

Table 3.1 Pre-stack seismic inversion parameters ...... 83

xvi LIST OF SYMBOLS

P-wave Velocity ...... VP

S-wave Velocity ...... VS

Density ...... ρ

Ratio between S-wave and P-wave velocity ...... γ

P-Impedance ...... Pimpedance

S-Impedance ...... Simpedance

P-Reflectivity ...... RP

S-Reflectivity ...... RS

Density-Reflectivity ...... RD

Angle Dependent P-wave Reflectivity ...... RPP (θ)

Angle Dependent Converted S-wave Reflectivity ...... RPS(θ)

Lambda-rho:Lam´eparameter ...... λ − ρ

Mu-rho:Lam´eparameter ...... µ − ρ

xvii ACKNOWLEDGMENTS

These two years in Colorado School of Mines have been the most exciting and enjoyable best times of my life. First of all, I would like to extend my deepest gratitude to my advisor Tom Davis. I feel privileged and honored to be one of his students. I thank David D’Amico for his feedback, teaching and encouragement. I really enjoyed every scientific discussion and assurance of every step of this thesis. I thank Tom Bratton for his teaching and conversations about how to innovate in this study. Understanding well logs and how to associate it with seismic are valuable knowledge that I will use for my entire career. I thank my other committee members, Bob Benson and Terry Young for their constructive comments and excellent suggestions. I thank Justine Gomez and Michelle Szobody for taking care of all the administrative details. I want to express my gratitude to RCP and its industrial affiliates for financial support. I thank Fulbright for being one of my sponsors and for creating the right environment for fulfilling my research goals. My colleagues in Montney Team in RCP; Tyler and Matt. Thank you for all the discus- sions and advice in this project. RCP friends and the good Latinos and Canadian friends I have made here in CSM: Carla, Esteban, Carelia, Johanna, Antonio, Jorge, Irene, Kyla, Luiz and Mauricio, thank you for have a nice time. I also want to thank the music; Pandora webpage was always in my long study sessions. Lastly I would like to thank my family: my parents, my beautiful sisters, my nephew Mathias and all of my family in Medellin. Also special thanks to my best friends in Colombia: Paula and Susana. Thank you for keeping me motivated in this phase of my life.

xviii To my nephew Matias.

xix CHAPTER 1 INTRODUCTION

The Pouce Coupe area is located in Northwest Alberta and it is part of the increasingly important Montney Shale play Figure 1.1 (Zonneveld et al., 2011). According to Canada’s National Energy Board, recent reserve estimates (November 2013) suggest that the play could carry Canada’s natural gas needs for the next 145 years. It is believed to contain 449 tcf of marketable natural gas, 14.5 billion bbl of marketable natural gas liquids, and 1.125 million bbl of marketable oil (Oil and Gas Journal, November 2013). This thesis is part of an integrated study by the Reservoir Characterization Project in conjunction with Talisman Energy that explores different scales of investigation to charac- terize the Montney Shale reservoir in Pouce Coupe Field (Figure 1.2). The data include borehole scale geomechanics, engineering data (microseismic and reservoir test), production data that were integrated with a multicomponent time-lapse data in determining the areal extent of induced fracture networks due to the hydraulic fracturing treatment of horizontal wells. Previous studies were focused on understanding the geomechanical analysis at borehole scale of how the Montney is predicted to fail when it is stimulated by hydraulic fracturing (Davey, 2012). Atkinson (2010) and Steinhoff (2013) demonstrated the ability of time-lapse multicomponent seismic to detect the orientation and magnitude of in-situ natural fracturing and also the changes in stress sensitive attributes induced by hydraulic fracturing through the Shear-Wave Splitting Analysis (SWSA). From those previous studies on multicomponent seismic analysis and the geomechanical characterization, this new study is looking to further analyze the available data and increase the understanding of the stimulations. The main purpose of this thesis is to characterize the lateral and vertical heterogeneity of rock quality of the Montney unconventional reservoir at different scales through the use

1 Figure 1.1: Map of Triassic strata in the Western Canada Sedimentary Basin showing the locations of Montney oil fields, conventional gas fields and tight gas plays. Pouce Coupe area is shown as a black start. Modified from Zonneveld (2011). of multi-attribute analysis of wells logs integrated with post-stack and pre-stack inversion of a baseline multicomponent seismic survey (Figure 1). Rock quality variability can be affected by composition and textural variation. Burial, compaction, hydrocarbon generation, diagenesis and tectonics all affect the mechanical integrity and in-situ stress state of the

2 Figure 1.2: Reservoir characterization of rock quality of Montney Shale and methodology. Integrated study and analysis at different scales. reservoir. In this thesis I examine how the composition variability can be associated with seismic parameters and ultimately aid in defining the best zones for future stimulation. Rock quality (composition) should be characterized before the hydraulic fracturing stimulation. The data input for this study is the baseline multicomponent seismic survey and the well logs before production and hydraulic stimulation. In this introductory chapter, I explain the geology, the data acquired in the area, and summarize the previous studies and methodology of this thesis. Chapter 2 explains the theory of cluster analysis and the interpretation of the lateral and vertical variability of the Montney C and D units at well log scale using well logs and core data. Chapter 3 explains the post and pre-stack inversion of the baseline multicomponent seismic survey that was performed using Jason software. Chapter 4 integrates the results of the cluster analysis and the seismic inversion in order to quantify and describe the lateral and vertical heterogeneity of rock quality of Montney Shale. These results are correlated with the microseismicity and production data of the horizontal wells. Chapter 5 contains the conclusion and recommen- dations of this study. The methodology details are contained within this chapter.

3 1.1 Objectives

The objectives of this study are to:

• Quantify and describe the lateral and vertical variability of composition of Montney C and D units at the well log scale through cluster analysis.

• Predict the heterogeneity of the Montney Shale in Pouce Coupe using elastic properties from the results of the seismic inversion of multicomponent data.

• Define the robustness of the pre-stack inversion compared with post-stack inversion in the prediction of the elastic properties.

• Integrate the results at well log scale with seismic to understand the variability of rock quality that can be correlated with engineering data.

1.2 Geology

The Lower Triassic Montney Formation was deposited on the western margin of the Western Canadian Sedimentary Basin (WCSB) (Moslow, 2000). Figure 1.1 shows the dis- tribution of Triassic strata with the location of Montney oil fields. The Montney Formation is one the most active plays in Western Canada and it has been the target of hydrocarbon exploration since 1955 (Kendall, 1999). More recently the exploration interest has focused on fine-grained (tight gas/shale) intervals in both the lower and upper Montney as a tight unconventional gas. This chapter discusses the geologic background and reservoir charac- teristics of the Montney Formation in west-central Alberta. The study area is Pouce Coupe Field, located near the British Columbia - Alberta Border, in the Peace River area of Alberta (Davies et al., 1997).

1.2.1 Reservoir units

The Montney Formation was originally deposited in a large central sub-basin known as the Peace River Embayment, which was developed during the Early and Per-

4 mian time (Edwards et al., 1994). The Triassic stratigraphic framework of the Peace River Arch can be seen in Figure 1.3. Reservoir units of interest in this study are highlighted on this figure. Montney Formation was deposited in a wide variety of depositional environ- ments, from distal offshore including turbidite channel and fan complexes to lower to upper shoreface including deltaic intervals and estuarine successions (Davies et al., 1997; Kendall, 1999; Markhasin, 1997; Mederos, 1995; Moslow, 2000; Panek, 2000; Zonneveld et al., 2010; Zonneveld, 2010).

Figure 1.3: Triassic Montney Formation in the Peace River Arch region. Pouce Coupe Field is on the border of British Columbia (BC) and Alberta and is represented by the colored formations in the Talisman BC chart section (courtesy of Talisman Energy).

In this study the facies of interest are organic-rich argillaceous siltstones and shales. According to Moslow (2000), the deposition of Montney Formation occurred in a ramp setting, and a ramp edge or slope break defines the updip depositional limit of the turbidite facies. The boundary between the lower and upper Montney Formation is a retrogradational shoreface succession and consists of laterally discontinuous dolomitic coquina beds. To the west, the lowstand system is comprised of two sequences of turbidite /siltstone and shale facies Figure 1.4.

5 Figure 1.4: Schematic simple WestEast section of Lower Triassic Montney Formation facies (Moslow, 2000).

The Montney Formation is subdivided into six units (Figure 1.5). The Lower Montney consist of the units A, B and C, unconformably overlain by the Upper Montney that in- cludes units D, E and F. The Lower Montney contains reservoir-quality upward-coarsening shoreface facies and coarse siltstones. It is the major producing interval in the Pouce Coupe area. Montney unit C is characterized by finely-laminated planar and ripple siltstone(Derder, 2012). The maximum thickness of Montney Formation in Pouce Coupe area is 160 me- ters (Davies et al., 1997). The Upper Montney consists of multicyclic coarsening-upward shoreface siltstones and interbedded very fine with hummocky cross-stratification and local developments of thin dolomitized coquina facies. The maximum thickness is 230 meters (Davies et al., 1997). The reservoir units that are important in this study are Mont- ney C and D. The Peace River deposition of the Montney is describing in detail by Atkinson (2010); Davey (2012); Davies et al. (1997). The reservoir property characteristics of the Montney units C and D in Pouce Coupe Field are: low matrix permeability (0.01-0.02 mD) and low porosity (6-10%), which are formally defined as tight gas (Smith et al., 2009). Those values are relatively higher than the values in other shale reservoirs in North America (Table 1.1). In order to produce tight gas it

6 Figure 1.5: Type log of the Triassic Montney of the southern Pouce Coupe Field. The red curve is the gamma ray log. Modified from Steinhoff (2013). is necessary to enhance permeability for economic production. Another characteristic that is important compared with other shale reservoirs is the mineralogy (Figure 1.6)(Sonnenberg, 2014). The Montney Shale shows a higher percentage of quartz and dolomite minerals in general.

1.2.2 Regional tectonics

The Montney was deposited along a NW-SE elongated paleo-continental structure. Its deposition was strongly influenced by paleostructure resulting in large thickness variations

7 Table 1.1: Montney reservoir properties compared with other North American shales. Pouce Coupe Field is located in the Montney core area (information from Talisman Energy Inc.) (Steinhoff, 2013)

Parameters Montney Montney Core Marcellus Utica Muskwa Barnett Permeability (nD) 130 20.000 250 160 230 300 Gas filled porosity (%) 2-5 7-9 1.6-7 2.5 1.6-7 3-5.5 Quartz and calcite (%) 50 60 17-36 20-55 12-69 13-50 Clays (%) 15 15 32 15 15 22 TOC (%) 1-5 0-1.5 1-12 0.2-2.2 1-10 3-8 Young’s Modulus (Gpa) 35-55 40-60 20-30 35-50 28-52 10-38

Figure 1.6: Comparison of predominant mineralogy of shales reservoir worldwide in a ternary plot. VacaMuerta Consortium. Sonnenberg, 2013 corresponding to local structural highs and lows that influenced the direction of the turbidity currents. Both syn-and post-depositional faults controlled the structure of the Montney, because the reactivation of extensional faults occurred contemporaneously with formation of the Dawson Creek graben complex (Moslow, 2000). The present day regional stress regime of the Montney Shale is compressional with the regional maximum horizontal stress direction approximately N4OE due the Laramide orogeny (Late -Paleogene) forming the

8 Canadian (Steinhoff, 2013).

1.2.3 Petroleum system

Pouce Coupe Field is considered a tight gas field because of the characteristics of the reservoir (low permeability and porosity). Hydrocarbon production began in 2002 and con- tinues to the present day. Initially, the wells were vertical or deviated and fracture stimu- lation techniques were used to increase recovery volumes. In 2008, the first horizontal well was completed in the Montney unit C. Figure 1.7 shows a schematic section line with the conventional and unconventional fields found in the Montney Formation.

Figure 1.7: The stratigraphic framework at Farrell Creek and Pouce Coupe. Maximum regressive surfaces are defined by red lines while maximum flooding surfaces are defined by green lines (courtesy of Lindsay Dunn, Talisman Energy Inc.). Modified from Davey (2012)

The entire Montney section in the area of application is gas charged, with hydrocarbon production occurring at depth of roughly 1700-2000 meters. Gas is trapped as a continuous phase in the fine-grained low permeability silts and shales, where capillary forces restrict migration either up-dip or into coarser grained rock. The Montney is also organic rich,

9 with TOC ranging from 0.51 to 4.18 wt.% and contains Type II/III kerogen, suggesting that this unit generates a significant amount of hydrocarbons where it is thermally mature, which creates a self-sourcing hydrocarbon system which is generally overpressured (Jones, 2008). This combination of local sourcing, low permeability and overpressure culminates in a widespread unconventional gas resource.

1.3 Time-lapse multicomponent seismic, microseismic data

Three time-lapse (4D), multicomponent (3C) seismic surveys were acquired by Talisman Energy in the Pouce Coupe Field in December 2008 (Figure 1.8). These seismic data were acquired to characterize the Montney reservoir and to monitor the hydraulic fracturing of two horizontal wells. Figure 1.8 describes the dates of the multicomponent surface seismic and field operations timeline in the horizontal wells. The baseline was acquired from December 8-10 of 2008, to characterize the in-situ reser- voir conditions before hydraulic fracturing of the two horizontal wells shown in Figure 1.8. Monitor 1 was acquired from December 13-14 of 2008, after 24 hours of the fracture treat- ment of the Montney unit C in the 02/02-07 horizontal well. The purpose of this survey is to determine if the extent of the induced fractures can be detected using surface seismic meth- ods. Monitor 2 was acquired from December 18-19 of 2008. This monitor was also acquired to monitor the fractured extent of a treated horizontal well. In this case the well 02/07-07 in Montney unit D, that was stimulated in December 17 of 2008 has similar parameters of completion as the first one.

1.3.1 Surface seismic acquisition

The seismic was acquired by CGG Veritas and covers a typical patch of about 5 km (1600m by 3000 m). The survey grid consists of 144 permanently buried 3C receivers (3.5 m depth) and 1241 cased shot holes (5.5 m depth). The configuration is shown in the Figure 1.9 and the acquisition parameters the Table 1.2. The field layout resulted in 41 inlines and 101 crosslines. The natural bin size of the acquisition is 100 m by 50 m.

10 Figure 1.8: Pouce Coupe time-lapse, multicomponent surface seismic and field operations timeline. The map shows the horizontal wells hydraulically stimulated (blue and green color). Red circles are the vertical wells. Orange shows wells with microseismicity receivers. Modified from Atkinson (2010)

1.3.2 Seismic processing

Pouce Coupe seismic survey was originally processed in 2011 by CGG Veritas, and re- processed by Sensor Geophysical in 2012 and 2013. The latest reprocessing effort by Sensor Geophysical improved the preservation of amplitudes. In 2013, the PP and PS seismic was reprocessing and the key differences in this new seismic processing (2013) include:

• Use of 100m by 50m natural bin size. Previous processing used 5D interpolation to reduce the bin size to 50 m by 50 m, which increased the risk of amplitude smearing across adjacent bins.

11 Table 1.2: Pouce Coupe time-lapse, multicomponent seismic acquisition parameters.

Recorded By CGG Survey Geometry Megabin Source Line Spacing 100 m Receiver Line Spacing 200 m Typical Patch 9 lines x 31 stations 1600m x 3000 m Charge Size (Dynamite) Baseline: 0.5 Kg Monitor 1: 0.2 Kg Monitor 2: 0.2 Kg Geophones 144OYO Geospace 3C Nails Sample Interval 2 ms

• Application of radon multiple suppression, which helped improve the signal to noise ratio.

• Removal of two passes of spectral whitening and AGC. Both of these processes have a significant impact on relative seismic amplitudes and needed to be removed for AVO/in- version.

• Converted wave layer stripping process. Previously, three iterations were performed between 700-900ms, 1000-1600ms, and 1700-2000ms to remove anisotropy above the reservoir. It was concluded that the best results were obtained by using a fixed PS1 orientation in the direction of the maximum horizontal stress N40E.

• Sorting and interpolating in the COV domain prior to migration. The objective of this step is to address the impact of irregular and sparse acquisition geometry on migration. 5D interpolation was also tested and superior results were produced using COVs for the PS data, but the best results for the PP data was after 5D interpolation. Results can be seen in Figure 1.11.

Workflows of the latest seismic reprocessing can be seen in the Figure 1.10. Figure 1.11 shows an inline number 4 of the baseline survey of PP data in the seismic versions of 2012 and 2013. The red box highlights the area of interest (Montney). Improved frequency content and balanced amplitudes are apparent in the new version. Input data for the seismic

12 Figure 1.9: Multicomponent seismic survey acquisition layout of Pouce Coupe time-lapse. 5km2 patch centered over horizontal wells 02/02-07 and 02/07-07. Modified from Atkinson (2010). inversion is the PP baseline survey that was reprocessed in 2013.

1.3.3 Microseismicity

Three microseismic methods were used to monitor the hydraulic fracture treatments of the 02/02-07 and 02/07-07 horizontal wells: surface microseismic were acquired by Microseismic Inc., downhole microseismic data were acquired by Pinnacle Technologies and water-well microseismic were acquired by Apex-HiPoint. For this study, only the downhole microseismic data were used in the integration part (Chapter 4). Figure 1.12 shows microseismic events for the two horizontal wells. (a) 02/07-07 well recorded by the 02/08-07 horizontal array (b) 02/07-07 well recorded by the 09-07 vertical array (c) 02/02-07 well recorded by the 02/08-07 horizontal array (d) 02/02-07 well recorded by the 09-07 vertical array. For each of the figures stage 1 is in blue, stage 2 red, stage 3 light blue, stage 4 orange and stage 5

13 Figure 1.10: Workflow of seismic processing (2013) PP data of Pouce Coupe. Processed by Sensor Geophysical. green. In general the quality of the microseismic data is good and the recorded amplitudes are consistent. There are some issues with event locations in (c) which leads to only the stage

14 Figure 1.11: Pouce Coupe iline 4 baseline survey - PSTM of PP data. Comparison of version 2012 and 2013 PSTM in PP data. Red box show the zone of Montney Formation.

5 events being used in the amplitude ratio analysis. Stages are occasionally co-located such as stages 1 and 2 in (a) and stages 2 and 3 in (d). In such cases similar failure mechanisms are found and only one solution is given. Stages 3, 4 and 5 in (b) have no amplitude data available so no ratios are calculated in these cases.

1.4 Available thesis data

The study area is located within Township 78, Ranges 11W6 and Alberta-Canada. The 3D seismic covers 10 km2 of area and the six vertical wells used in this study penetrated the base of Montney Formation. Figure 1.13 shows a map of the seismic grid outline and the

15 Figure 1.12: Results of the downhole microseismic events on both treatment wells, showing the total combined event location from the individual 5 stage of each treatment. wells used in this study. The seismic used in the pre-stack and post-stack inversion is the pre-stack time migration (PSTM) of the PP-waves of the baseline of the multicomponent data reprocessed by Sensor Geophysical in 2013. The Pouce Coupe well database is shown in Figure 1.14 and Table 1.3. These wells and microseismicity were used at different stages of this project.

16 Figure 1.13: Map of the Pouce Coupe 3D seismic outline (orange), vertical wells used in the cluster analysis (black points) and the important horizontal wells (blue and red).

Figure 1.14: Pouce Coupe wells used in this project. Some wells are used in different stages of the project.

1.4.1 Production and completion data

The first horizontal well was drilled and completed in the Montney unit C in February of 2008 (07-07). This well started to produce in March of 2008, but its production is not

17 Table 1.3: Abbreviation legend of Figure 1.14

Microseismic MS Rock Quality Index RQI Static and Dynamic Triax- TRP ial Rock Properties Mohr-Coulomb Failure MC Spinner Tracer log ST Fracture gradient FG Porosity Por. Permeability Perm Fullsuite GR, RHOB, NEUTRON, PEF, Resistivity, DT. Cluster Analysis CA Seismic Inversion SeismicInv Integration Microseismicity Integration Rock physics RP affecting the amplitudes of the baseline of Pouce Coupe data. The wells of interest in this thesis are the two horizontal wells within the Pouce Coupe 4D-3C survey area targeting the Lower Montney unit C (2/2-7-78-10W1 referred to as the 02/02-07 well) and the Lower Montney unit D (2/7-7-78-10W6 referred to as the 02/07-07 well). Hydraulic fracturing in the 02/02-07 well was done individually with five 200 m- spaced perforation (stage) locations using an openhole packer system. The parameters used in the second horizontal well (02/07- 07) were identical to the first well (02/02-07), except at a constant interval spacing of 250 m. Each well was fractured and then allowed to flow back just enough to retrieve the treatment balls at surface, and then shut in to maintain full pressure at reservoir level. The parameters are shown in Table 1.4. There are spinner log data from the two horizontal completed wells that were collected in January 13 of 2009 (02/02-07 well) and January 15 of 2009 (02/07-07), a month after hydraulic stimulation of each well (Table 1.5). Stages one and two on the 02/07-07 well could not be separated so they are combined to a total of 32%. The 02/02-07 well shows a

18 Table 1.4: Pouce Coupe Field horizontal well hydraulic stimulation parameters.

Date (m/d/y) Well Fluid Type # of Proppant H2O Closure stages/ size Load Pressure proppant (m3) (Mpa) 12-12-08 02/02-07 Clear 5/100T 20/40 1328 30 12-12-08 02/07-07 Frac 5/100T 20/40 1330 28 more uniform distribution of production across the five stages.

Table 1.5: Spinner log data from the two horizontal completed wells

Stage 02/02-07 02/07-07 1 0.2 0.16 (combined) 1 0.13 0.16 (combined) 1 0.25 0.43 1 0.17 0.1 1 0.24 0.14

The production data for three horizontal wells: 02/07-07, 02/02-07 and 00/07-07 are shown in Figure 1.15. The 02/02-07 in Montney unit C shows significantly better IP and better production on an average m3 of gas per day.

1.5 Previous work

Atkinson (2010) initiated RCP research on the Pouce Coupe dataset. This work demon- strated the ability of time- lapse multicomponent seismic to detect changes in stress sensi- tive attributes induced by hydraulic fracturing. Components of Atkinson’s work (2010) are strongly related to the ongoing project and can be divided into two broad categories; seis- mic response modeling, and shear wave splitting analysis. Through geomechanical analysis and P-wave modeling, hydraulic fracturing was demonstrated to have a major effect on the local stress regime. The stacked P-wave data were unable to monitor time-lapse changes in the reservoir. The acoustic response to reservoir pressurization could not be detected thus shifting the study to the anisotropic response quantification using converted waves. Inter-

19 Figure 1.15: Production data decline of the horizontal wells. The wells drilled in Montney unit C have higher initial production. pretation of the converted wave data utilized time delays caused by shear wave splitting. The methodology summed negative time-variant time-shifts over the reservoir, which is in- dicative of seismic anisotropy induced by fractures (Figure 1.16). A reasonable correlation was found between shear wave splitting maps and the microseismic dataset showing that this methodology can be used as a seismic anisotropy indicator. The 4D seismic across the field was reprocessed for Steinhoff (2013) to allow for post- stack Shear Wave Splitting (SWS) analysis. The reprocessing involved the use of Receiver Azimuth Detection and Rotation (RADAR), layer stripping and simultaneous processing of the three surveys. Through this simultaneous processing the NRMS error between the surveys was reduced significantly and comparison could be drawn between them through the new high repeatability. This summary will focus on the interpretation drawn from the

20 Figure 1.16: Difference in sum of monitor 2 negative shifts and monitor 1 negative shifts from 2000-2200 ms, showing areas of increased azimuthal anisotropy. Indicated in red is the outline of the anisotropy present in the baseline. Calculated by Atkinson (2010). seismic data and its new processing. Fault mapping across the survey was performed and identified a number of strike-slip/wrench faults (Figure 1.17). These fault lineaments could not be identifed as a displacement in either the PS1 or PS2 images so were picked using the offsets on deeper reflectors as guides, which were tracked up to the Montney interval where they produced a dimming effect. The two main fault trends seen (NE-SW and NW-SE) are believed to correlate with areas of high fracture density. An important aspect of this work was to define a method to calculate SWS amplitudes from converted (P to S) data. Using PS1 and PS2 stack seismic volumes (Steinhoff, 2013) the time delays between the PS1 and PS2 arrivals from the base of the Montney reservoir

21 Figure 1.17: Shear wave splitting map from the baseline survey of Pouce Coupe. (Steinhoff, 2013).

were calculated. The base reflector was used as there was a lack of coherent reflectors within the reservoir and using the basal reflector allowed an estimate of SWS across the whole of the reservoir. It is acceptable to estimate SWS levels across the whole of the reservoir interval with the assumption that the dominant fracture orientation does not vary vertically within the formation. A measure of SWS was then calculated (1.1) using these time-shifts for the baseline survey and each of the two monitors.

(t − t ) SWS = PS2 PS1 , (1.1) tPS1 SWS analysis of the baseline survey (Figure 1.17) shows two dominant natural fracture sets, one parallel to Shmax at N40E near the 02/07-07 horizontal well and the other ap- proximately perpendicular to this oriented towards Shmin close to the 02/02-07 well. This

22 pre-stimulation survey shows SWS magnitudes of 2-3% from the natural fractures in the reservoir. Small arrows indicate the direction of PS1 with the size of the arrows scaled to the amount of SWS defined by 1.1. The large red arrows indicate the two predominant di- rections of natural fractures, with one set at N40E and the other roughly orthogonal to this. Shadowed black lines show the interpreted strike-slip/wrench faults within the survey region in the Montney. Fracture stage locations are numbered from 1 at the toe of the horizontal wells to 5 near the heel of the wells. After hydraulic stimulation of the 02/02-07 well the monitor 1 survey was acquired. SWS magnitudes increase around the well on the monitor 1 survey (Figure 1.18) indicating increased fracture dominance in the direction of Shmax provided by the newly created hy- draulic fractures. Magnitudes reach up to 8% from this survey with particular concentration between stages 3 and 4. The anomaly between these stages is attributed to a stress concen- tration and preferential fracture conditions and is also considered an indicator of the growth of hydraulic fractures in the direction of regional Shmax at N40E. A similar pattern but of less intensity is seen towards the toe of the horizontal which also extends in the direction of Shmax but may also interact with the identified fault in that zone. The final anomaly near stage 5 at the heel of the well seems to grow in both the direction of Shmax and Shmin but due to the natural fracture orientation seen in the baseline map (Figure 1.17) that seems to orientate more towards Shmin (Figure 1.17). Following the hydraulic stimulation of the 02/07-07 horizontal well the second survey was acquired. The SWS results from this monitor show the effects from the treatments of both wells (Figure 1.19). This treatment seems to create a lesser SWS response with no higher magnitude anomalies occurring around the wellbore as seen in the monitor 1 results. The anomaly seen at the toe of the 02/02-07 well seems to have grown in magnitude; this correlates to a concentration of microseismic activity in this area interpreted as a minimal offset fault lineament. The other two anomalies identified appear to have decreased in SWS magnitude and diffused out into the reservoir. Two possible interpretations are that this

23 Figure 1.18: Shear wave splitting map from the monitor 1 survey after the 02/02-07 well had been stimulated. (Steinhoff, 2013). result is representing an equibrilating of stimulation pressure within the reservoir or that fractures are closing on the proppant. The dominant fracture orientations did not change in any significant manner between the baseline and monitor surveys which may be an indication of the hydraulic treatment mainly stimulating the pre-existing natural fractures and the induced SWS is a measure of the propping of this original network. Steinhoff theorized that SWS anomalies may be a good indicator of effective fracture permeability created during the hydraulic fracturing. The induced SWS anomalies were seen as highlighting areas where propped fractures pro- vided a conduit to the well bore. To further emphasize the SWS anomalies difference maps were created between the baseline and monitor 1 and baseline and monitor 2 surveys SWS results (Figure 1.20). From these difference maps the overall stimulation of each well can be

24 Figure 1.19: Shear wave splitting map from the monitor 2 survey after the 02/07-07 well had been stimulated. (Steinhoff, 2013). determined. The lower level of SWS splitting seen close to the 02/07-07 well is seen as an explanation as to why the 02/02-07 well produced 39% more gas than the 02/07-07 well. Figure 1.20 also shows spinner production data for each of the stages in each well. From the monitor 1 minus baseline map a correlation between SWS anomalies and high production is interpreted. Low production in stage 2 is attributed to proximity to the interpreted fault around that stage with energy leak off into the fault along the path of least resistance. The monitor 2 minus baseline section shows a greater correlation between SWS anomalies and production and was seen as a good view of the hydraulic stimulation success and resultant effective stimulated volume. The leak off of injection fluid further into the reservoir and associated pressure equilibration is also visible on the monitor 2 minus baseline map as SWS magnitudes spread across the seismic zone. The second monitor was therefore considered to

25 Figure 1.20: On the left is monitor 1 minus baseline of the SWS magnitudes. On the right is monitor 2 minus baseline of the SWS magnitudes. Also plotted is the production from each stage for each well indicated by percentage. (Steinhoff, 2013). be a better image of the effective stimulated volume compared to the monitor one survey, which showed a more instantaneous response to the stimulation of the 2-07 horizontal well.

1.5.1 Geomechanical characterization

Davey (2012) created a geomechanical model and analized the mechanical stratigraphy framework of the Montney Shale in the Farrell Creek and Pouce Coupe areas based on an integrated multi-scale approach including well logs, engineering, completion data , time- lapse seismic and production analysis. The rock variability in the tight gas siltstones within the Pouce Coupe area is controlled by two main factors: composition (petrography and TOC) and texture (laminations, microfractures, large-scale fractures). Davey (2012) defined a parameter named RQI (Rock Quality Index) in order to understand both the ideal stress conditions and rock property conditions for effective hydraulic stimulations. The mechanical stratigraphy was defined using the rock properties and standard log-based measurements such as density, sonic velocity and gamma ray, which indirectly examine both compositional

26 and fabric-based brittleness, but exclude stress. The main conclusions of the geomechanical model are:

• The behavior of the Montney as a reservoir includes: Hydraulic fractures are easily initiated and grow in zones of homogeneity. When these fractures reach the interface of a heterogeneous/brittle zone, the hydraulic energy is dissipated into the highly stressed zone. This hydraulic fracture should create complex fracturing in homogeneous zones of natural fractures and may fail laminated zones to create a complex fracture network.

• RQI is an accurate indicator of heterogeneity, which incorporates both stress and brit- tleness elements. This heterogeneity has a strong formation influence on the progression of hydraulic fractures based on the observations of microseismic data and Diagnostic Fracture Injection Tests (DFIT).

• RQI can be used as an indicator to design differential spacing and fluid/proppant volumes in a potential re-fracturing of the wells that have been previously stimulated. The best options are located in the interfaces between homogeneous and heterogeneous RQI.

• Stress shadowing amplifies the effects of energy dissipation in brittle zones. The RQI was correlated with microseismicity as shown in Figure 1.21.

Figure 1.21 shows in the first panel the RQI log, the second panel shows the Gamma Ray, the next panel is mechanical stratigraphy and microseismic event abundance is the last panel. These microseismic events are the number of events less than 500 m away from the observation well with signal/noise ratio greater than 5. This correlation shows that consistent RQI of the lower facies reflects areas of reservoir homogeneity. In the lower facies we see a greater number of higher intensity events occurring close to the wellbore; interpreted as a manifestation of the large pressure and stress changes introduced by the hydraulic fractures. Any bedding-plane interfaces and proximal fractures will likely shear, causing microseismic

27 Figure 1.21: Correlation of microseismicity with RQI. First panel is RQI, second panel is Gamma Ray and third panel is mechanical stratigraphy. (Davey, 2012) signature to be recorded. Another comparison was made with the multicomponent shear- wave splitting results shown in Figure 1.22. Homogeneous zones along the 02/07-07wellbore were determined using the Rock Quality Index line (black line). Once a hydraulic fracture commences in a homogenous area, shear failure on natural fractures is likely to occur proximal to the induced pressure site. Stress shadowing theory states that as stimulation occurs in a reservoir, reservoir pressurization is an additive effect. Depending on the lithology and initial stresses, there is a threshold pressure at which the stress contrast generated by the propped-open fracture exceeds the in-situ stress contrast, thereby creating a localized zone of stress reversal. Shmin-parallel fractures occurred. The presence of stronger azimuthal anisotropy here also indicates that proppant must be present at some level in these fractures, as they are remaining open for some time following stimulation.

28 Figure 1.22: Correlation of RQI log with the SWVA in the baseline. Hydraulic energy will preferentially propagate to more homogeneous areas of the reservoir. (Davey, 2012)

1.6 Methodology of this work

This project consists of analysis of various data sources. Figure 1.23 presents a workflow to integrate core, log and seismic measurements across multiple scales for the assessment of reservoir and completion quality in a reservoir. The first step is defining the variability at log scale of the heterogeneity of composition using cluster analysis, subsample this variability for core measurements, then upscale the integrated core and log data to seismic scale using the results of seismic inversion. Numbers after heading denote the chapter in which that methodology is elaborated upon.

29 Figure 1.23: Workflow followed in this thesis research. Numbers after heading denote the chapter.

30 CHAPTER 2 CLUSTER ANALYSIS

This chapter discusses cluster analysis of the Pouce Coupe log data for rock class iden- tification. The main purpose of the cluster analysis is to quantify and describe the lateral and vertical variability of composition of the Montney at well log scale. The cluster anal- ysis allows us to correlate rock properties and provide more representative upscale input parameters for characterization that can be compared with seismic results. The first part of this chapter is the theory of cluster analysis of well logs , some examples in the literature and the workflow were performed using the HRA (Hetereogeneous Rock Analysis) module of Schlumberger Techlog software version 2013.1.0. The second part of the chapter is the cluster analysis of the Pouce Coupe wells, the description of the six rock classes chosen and the integration with core data from well 13-12. Finally, there is a correlation of the cluster analysis in all the wells and the interpretation of the variability of units C and D.

2.1 Theory and workflow

Cluster analysis is a name given to a wide variety of mathematical techniques designed for classification (Doveton, 1994). Those techniques are grouping objects that are similar and distinguishing them from dissimilar objects on the basis of their measured characteristics. This clustering is known as unsupervised classification, because the operation is neither dictated by an external model nor determined by a reference training set of objects previously known. We used cluster analysis to define rock classes based on their fundamental attributes of composition. In this study we used well logs that are affected by composition (NPOR, RHOB, GR and PEF). Heterogeneous Rock Analysis (HRA) is the module of Techlog software that was used to perform the cluster analysis. Cluster analysis finds the uniqueness based on a

31 similar set of measurements and it becomes easy to discriminate and identify where these rock classes exist elsewhere. It also helps to understand the spatial variability in tight shale plays, which is fundamental for hydrocarbon exploration and production. There are two main steps in the process of cluster analysis. The first one is to define the master model and the second is the application of the master model to logs in other zones or other wells in the field. This second step is called tagging, which uses the unsupervised master model as supervised training data for logs in other wells. The first step in the master model analyzes the input data using Principal Component Analysis (PCA) and ensures that the data used in the clustering is independent. The Prin- cipal component analysis solution is simply a geometric rotation in a multidimensional space that locks onto the orthogonal axes of relative elongation in a cloud of data points (Dove- ton, 1994). Figure 2.1 shows a multivariate cross plot of log data, plotted in the principal components space that captures the majority of the variability between the logs and not the logs themselves in their original form. The spread of data along each axis is associated with a degree of correlation between each input log curve. The colors in Figure 2.1 represent the rock classes that were assigned from an unsu- pervised classification of the principal components. The class centroids are shown as black stars, so each data point is applied to the rock class whose centroid is the closest in principal component space. The logs within each class are statistically similar, so it is possible to summarize and discretize the bulk log data into zones that are unique and different. After the PCA analysis, the principal components are used in the k-means-clustering algorithm to create the master model. The goal of the k-means-clustering is to minimize the distance between each data point and the center of each defined cluster and to maximize the distance between the centroid of each cluster. There are some diagnostic plots that are used to determine if the number of cluster specified is statically significant and has enough variability. One plot that is very important is the box and whisker plot (Figure 2.2), which shows the variability of the input logs for each class. Another plot is the fall-off plot

32 Figure 2.1: Scatterplot of the first three principal components form a principal components analysis of multivariate log data. Schlumberger (Handwerger et al., 2011)

(Figure 2.3), where the horizontal axis represents the number of clustering runs made, while the vertical axis represents the sum of the distance of each data point to the centroid of its assigned cluster. The cumulative distance (fall-off) will decrease until the minimum distance has been determined. When the fall-off is relatively flat for the last 10% of the runs, the clustering converges to the optimum number of clusters.

Figure 2.2: Box and whisker plot by class of 10 clusters showing the variability of GR log using Techlog software

33 Figure 2.3: Fall-off plot of HRA of 6 clusters showing 200 runs in horizontal axes and distance in vertical axes using Techlog software.

There are several problems when applying cluster analysis to the raw well-log data. The well logs were acquired by different instrumentation. The normalization of the well logs is important before running the PCA. We will show the example of normalization in the analysis part of this chapter. The final step is tagging the clusters in the other wells, which is essentially assigning a cluster to all the well logs. The distance between the data point and the centroid of the cluster is termed the compliance distance of that data point. The compliance distance is a Mahalanobis distance used to measure the similarity of the data. The degree of compliance (error) is the similarity between the master model and the new data to predict, which allows us to recognize existing rock classes or identify new rock classes. According to Schlum- berger, a compliance distance of less than 20 is considered valid (Bratton, 2014)(personal communication). The cluster analysis steps are summarized in Figure 2.4.

34 Figure 2.4: Cluster analysis workflow using a 2-D crossplot of neutron and density porosity. A) Determine how many cluster are represented by the data. B) Example of 4 clusters, where the centroids are the black points. The compliance distance would be the distance from the square data point to the centroid of the red cluster. Modified from Silvis (2011).

2.1.1 Example in literature

Cluster analysis in geology has been in use since 1980 by Romesburg (Romesburg & Marshall, 1980) with the hierarchical analysis. It is a clustering of hypothetical zones based on their log responses. Some authors have given different applications. Robinson et al. (1989) correlated cluster groups with potentially oil-productive stratigraphic zones in four wells of the North Riley Unit, Gaines County, Texas. Moline (1992) provides a good practical example of the application of cluster analysis to logs as a means to identify pressure seals. Those examples were applied to certain logs perceived to respond to a desired property in an effort to define a set of rock classes that correlate with facies defined by a geologist using core or outcrop descriptions or stratigraphy. There are some examples of HRA analysis in conventional fields (Klepacki, 2012) and (Silvis, 2011). We are interested in the studies of unconventional fields (Tight gas shales), and there are few publications of this HRA analysis. Handwerger et al. (2011), Suarez-Rivera (2011) applied HRA to identifying rock units with similar texture and composition based

35 on consistent data structures, which is defined by pattern recognition of the measured data channels. Willis (2013) used cluster analysis to combine geological characteristics as well as the elastic properties of shale formations to help in better aiding fracturing design and placement in VacaMuerta Formation-Argentina.

2.2 Cluster Analysis of composition in four wells of Pouce Coupe

The rock variability of the tight-gas siltstones in Pouce Coupe Field is controlled by two main factors: composition (petrography and TOC) and fabric (laminations, microfractures, large-scale fractures)(Davey, 2012). The main purpose of the cluster analysis in this thesis is to quantify and describe the lateral and vertical variability of the composition of Montney at well log scale. We used the well logs that are most affected by composition: neutron-porosity, density, photoelectric and gamma ray. Those logs are present in all of the wells that we are using in the analysis (05-14, 13-12, 02-07 and 09-07). Figure 2.5 shows a map of the Pouce Coupe 3D seismic with the four vertical wells of this analysis. The master model in the cluster analysis was running in the master well 13-12 that has a complete dataset and a core from the Montney unit C.

Figure 2.5: Map of the Pouce Coupe 3D seismic outline (green), vertical wells used in the HRA analysis (black points) and important horizontal wells (blue, red).

36 2.2.1 Correction and edition of well logs

Before running the cluster analysis, it is necessary to edit and normalize the well logs. As mentioned before, the wells were logged with different instrumentation, so each suite of well logs may have different environmental corrections. The neutron tool measures the hydrogen index in a reservoir, which is directly related to porosity. The Hydrogen Index (HI) of a material is defined as the ratio of the concentration of hydrogen atoms per cm3 in the material of pure water at 75oF . The measurement allows estimation of the amount of liquid-filled porosity due the hydrogen atoms are present in both water and oil filled reservoirs (Wikipedia, 2014). A thermal neutron tool detects only the thermal neutrons and relates the count rates. The neutron log required a correction depending on matrix and this case the Limestone matrix was chosen . Figure 2.6 shows the correction of the four wells and the mean value of the neutron-porosity in the reference well is 0.076 m3/m3.

Figure 2.6: Histogram of NPHI of four wells before and after enviromental and matrix correction in the interval Doig to Belloy.

The gamma ray tool statically measures the natural gamma ray radiation of a forma- tion. Different vendors acquired this Gamma Ray (GR) log, so normalization is required before running the cluster analysis. Normalization generally consists in compensating the log measurements for one or more conditions, such as difference in tool responses, differences in rock and fluid properties, the relative angle between borehole and formation, anisotropy

37 and environmental effects.(Techlog software help) . The GR logs were normalized. The process consist that the mean value of a log over a selected zone is calculated. The maximum and minimum log values that are used to remove outliers were defined before computing the mean or percentile values. All log values of the new wells are then transformed so that the mean corresponds to a desired calibration value. The reference well log is 13-12 and the mean value of the GR is 116 API. Figure 2.7 shows the GR log histogram of the four wells in the interval Montney before and after the normalization. The red curve is the well 13-12. The same process of normalization was done in the Photoelectric log. The mean value of the well log of reference is 3.47. The histogram can be seen in the Figure 2.8.

Figure 2.7: Histogram of GR log of four wells before and after normalization in the interval Doig to Belloy.

The histogram of the density log is showed in Figure 2.9. The red curve is the reference well 13-12. This log does not require a normalization. The density mean of Montney is 2634 kg/m3.

2.2.2 Definition of clusters in the master well log

The cluster analysis was done using a master well (13-12) that has a complete dataset and a core from the Montney unit C . The interval of this analysis is between the Top of Belloy Formation and the Top of Montney E (253 m of thickness). This interval covers the field’s zone of interest which is the Montney C and D units. The k-means clustering algorithm was

38 Figure 2.8: Histogram of photoelectric log of four wells before and after normalization in the interval Doig to Belloy.

Figure 2.9: Histogram of density log of four wells in the interval Doig to Belloy. This log did not required normalization. run on the master log (well 13-12) and was tested at cluster ranging from 5 to 10. Based on the analysis of box and whisker plots, silhouette plots, and distance fall-off plots, six clusters were chosen. Figure 2.10 shows the box and whisker plot for the six clusters. The black line in the box represents the mean values of the logs. The other plots are shown in Appendix. Figure 2.11 shows the resulting master log or HRA reference model. The unit C exhibits a strong blocky response, the dark blue (Cluster 1), and some interbedded layers of green,

39 Figure 2.10: Box and whisker plot for 6 clusters in the logs Neutron-Porosity, GR, RHOZ and PEFZ. yellow, and light blue towards the bottom. The unit D shows interbedded layers of yellow (Cluster 4), blue (Cluster 1), and green (Cluster 3). The unit A and B shows a blocky response of red (Cluster 5). Unit E shows a two blocks of light blue (Cluster 2) and coffee (Cluster 6). The intervals in black color in the first panel are not reservoir.

2.2.3 Integration of cluster analysis and core data

The well 13-12 was cored in the unit C of the Montney Formation (MD: 2196-2214 m, red box Figure 2.11). There are some mineralogy analyses using XRD and microprobe by the University of Calgary (2009-2011) that are incorporated in this study (Derder, 2012) and (Freeman, 2011). Bulk X-Ray Diffraction (XRD) was performed by Derder (2012) to determine mineral composition of the siltstone and shale of the core well 13-12. This analysis shows that the sandstone, siltstone and shale consist in order of abundance of silica, dolomite, orthoclase, and muscovite (Table 2.1). Microprobe analysis was done by Freeman (2011) to create chemical maps on the same samples. The mineral compositions of samples were demonstrated

40 to be constant across the coarser lamination. The resulting mineralogy for each sample is shown in the Table 2.2.

Table 2.1: XRD analysis for samples of the well 13-12. Mineral by constitution is by weight %

Core Sample Qz % Dolomite Orthoclase Muscovite Calcite Illite Pyrite Albite Depth % % % % % % % (m) 2197.71 1 41.36 20.30 18.40 9.80 0.02 1.64 1.64 3.26 2198.57 2 29.77 21.40 18.64 15.28 0 1.92 1.81 5.08 2207.60 3 30.80 29.84 12.57 9.72 8.50 0.82 2.15 2.69

Table 2.2: Microprobe analysis for sample of the well 13-12 (Freeman, 2011)

Core Sample Qz % Dolomite Orthoclase Muscovite Calcite Illite Pyrite Albite Depth % % % % % % % (m) 2197.71 1 41.36 20.44 7.92 14.60 0.86 9.72 0.87 2.95 2198.57 2 35.88 19.96 10.09 14.61 0.87 13.96 0.95 2.85 2207.60 3 34.87 24.43 7.01 11.90 1.42 9.07 1.11 1.84

The interpretation of the six clusters of the master model is showing in the neutron vs. density plot. This plot helps to understand qualitatively which cluster has better petro- physical rock properties (high porosity, low clay content and low saturation of water (Sw)). Figure 2.12 shows the Neutron-density porosity plot of the six clusters of the well 13-12 in the Montney interval. The porosity trend is increasing to the right (black arrow) and the direction of the blue arrow is correlated with the increment of content of quartz and decrement of clay content. Clusters 1 and 2 are located in the best position of these two conditions. The blocky dark blue Cluster 1 represents homogeneity in composition, which can be correlated with the XRD analysis of the core (red box in Figure 11). The presence of high percentage of quartz and the reduced clay content is indicative of good rock quality properties. I conclude that clusters 1 and 2 have the best petrophysical rock properties.

41 2.2.4 Cluster tagging

After this master classification was established, other wells in the area were tagged in- cluding 05-14, 09-07, and 02-07 (Figure 2.13). Figure 2.14 shows a zoom of the units C and D. The highlighted formations are the Montney D and two Montney C subsets. The quality control of this tagging is based on the compliance log that is located in the panel No. 6 of every well in Figure 2.13. Confidence can be expressed in the results because the compliance is generally lower than 20, which is deemed to be the limit of acceptability.

42 Figure 2.11: Master well log 13-12 and HRA reference model of cluster analysis Six clusters. First column is the intervals of Montney Formation. Panels from left to right: Neutron- Porosity, GR, RHOZ, PEFZ and cluster composition. The legend is in the right part, blue to red color. The core interval is located in the red box. The units C and D are highlighted in the dashed black box.

43 Figure 2.12: Neutron porosity vs. density plot of the results of master well log 13-12 in the Montney interval. The legend is the color of the six clusters identified.

Figure 2.13: Log correlation of the four wells of Pouce Coupe. Master well is 13-12 and three wells were tagged. The last panel is the compliance.

44 Figure 2.14: Cluster analysis results in the units C and D of Montney of the four wells. Cluster colors are showed in the legend - lower part of the figure.

45 2.2.5 Interpretation

The pattern of Montney unit C is consistent in all of the wells except in 09-07. It looks more laminated than the others but it has the same clusters as the other wells. Unit C can be divide in two levels, where the upper is more homogeneous except in the east side of the survey and the lower is more heterogeneous and laminated. Montney unit D has the same pattern in three of the wells but well 02-07 looks more homogeneous in the upper part. The horizontal wells of this study are located between the wells 09-07 and 02-07, one is located in Montney unit C and the other in Montney unit D. We can interpret the composition of Montney unit C is more heterogeneous near to well 09-07 and more homogeneous near the wells 2-7, 13-12 and 5-14. Montney unit D is more heterogeneous in general but looks more homogeneous close to the well 2-7.

46 2.3 Summary

Cluster analysis was performed in four wells in the Pouce Coupe Field using the well logs that are most affected by composition: NPOR, GR, PEF and RHOB logs. The main purpose of the cluster analysis in this thesis is to quantify and describe the lateral and vertical variability of composition of the Montney at well log scale. The cluster analysis provides more representative upscale input parameters for reservoir characterization that can be compared with seismic results. The master well log is the well 13-12 that has a complete set of logs and a core in unit C of Montney. Six clusters were chosen. The result of this cluster analysis has indicated a lateral variation of composition of the unit C to the east side of the area (Figure 2.14). Unit C can be divided in upper and lower level: the upper level shows a blocky response of cluster 1 in all of the wells except in 09-07, the lower level is more laminated with a high percent of cluster 3. Unit D is more heterogeneous in general but looks more homogeneous close to the well 2-7. There is a poor presence of clusters 1 and 2 in the interval of unit D. This cluster analysis can be used in the rock physics analysis to find the elastic parameters that will highlight areas of better quality reservoir.

47 CHAPTER 3 INVERSION

The transformation of seismic data into a quantitative rock property description such as acoustic impedance, shear impedance and density is commonly referred to as seismic inversion. The two main advantages of seismic inversion are: diminishing of tuning in the data and an increase in bandwidth. The tuning is diminished through removal of the wavelet, and bandwidth is increased because of adding a low frequency component from the well log that is missing in the seismic data. Latimer et al. (2000) shows different examples of tuning and the increment in bandwidth. The seismic inversion in this study was computed using the Jason software (CGG Com- pany). This package performs a Constrain Sparse Spike Inversion (CSSI) algorithm. CSSI is a recursive inversion method based on sparse impulse deconvolution (Russell, 1988). This chapter will introduce the inversion method, the post-stack and pre-stack inversion theory, workflow, and assumptions in CSSI. The second part involves the data used in the seismic inversion, the detailed workflow of the post-stack and pre-stack seismic inversions of the baseline of PP wave data of Pouce Coupe survey and its comparison. The objective to do these two seismic inversions (post and pre-stack) is to answer the question: what is the robustness of the pre-stack seismic inversion compared with post-stack in the prediction of the elastic properties? In the end, we want to use those elastic properties to predict good quality rock in the 3D survey (Chapter 4).

3.1 Method and theory

This part will explain the convolutional model, the constrained sparse spike algorithm used in the seismic inversion, the workflow and parameters of post and pre-stack inversions and the importance of the wavelet in the inversion.

48 3.1.1 Convolutional model

The basic problem that we want to solve in seismic inversion is to find an impedance series corresponding to a series of reflectivity series r(t) that satisfies the convolutional model:

s(t) = w(t) ∗ r(t) + noise, (3.1)

where s(t) is the recorded seismic trace and w(t) is the wavelet (Russell, 1988). The seismic trace is simply the convolution of the earth’s reflectivity with the seismic source or wavelet function with the addition of noise. The wavelet is time-varying and complex in shape. It is dependent on the depth of the formation of interest and its frequency content decreases with depth because the high-frequencies are absorbed. When there is normal incidence reflectivity of the P wave into two homogeneous, isotropic and elastic layers, the reflection coefficient of this normal incidence of P-wave (Rp) can be estimated:

V2ρ2 − V1ρ1 RP = (3.2) V2ρ2 + V1ρ1 where V is compressional velocity and ρ is density. The subscripts 1 and 2 refer to the upper and lower media respectively. Seismic impedance is the product of velocity and density (V ρ). When there is an incident angle in the reflectivity of the P-wave into two layers homogeneous, isotropic and elastic, the P-wave reflection coefficient is estimated from the elastic parameters using Knott-Zoeppritz equations or the first-order approximation to reflectivity as given by (Aki & Richards, 1980).

"  2 #  2   1 VS 2 ∆ρ 1  2  ∆VP VS 2 ∆VS RP (θ) = 1 − 4 sin θ + 1 − tan θ − 4 sin θ , (3.3) 2 VP ρ 2 VP VP VS

where VP , VS are the average P-wave and S-wave velocities, ρ is density and θ is the average of the P-wave incidence and transmission angles.

3.1.2 Constrained Sparse Spike Inversion (CSSI)

The categories of seismic inversion can be seen in Figure 3.1. Constrained sparse spike is a deterministic or trace-based seismic inversion. The reflection coefficient series underlying

49 the acoustic impedance is assumed sparse.

Figure 3.1: Categories of seismic inversion. Constrained sparse spike is a trace-based seismic inversion. (Mesdag, 2013).

The advantage of using sparse-spike inversion is that the inversion itself is not directly dependent on the impedance log information; it is only to set the correct absolute impedance level. Well log information is important for accurate wavelet estimation (velocity and density logs). This method is referred to as constrained sparse-spike inversion, because it has simple constraints that are usually provided, in order to obtain a more accurate solution to the inversion. One constraint is a range of values in which the impedance can vary laterally away from the well locations. A second constraint is the location of the major horizons on the input data to provide solutions consistent with the conventionally picked horizon interpretation, which form the priori model.

3.1.3 Post-stack and pre-stack inversion

Post-stack inversion transforms the seismic amplitudes into a band-limited estimated of P-wave acoustic impedance or P-impedance. The algorithm used to perform this post-stack

50 inversion is Constrained Sparse Spike Inversion (CSSI). This inversion method assumes a limited number of reflection coefficients with larger amplitude and transforms seismic data to a pseudo-acoustic impedance log at every trace. There are a lot of advantages of acoustic impedance (AI) (Latimer et al., 2000). AI is a rock property that can be measured in seismic and well logs, it is also closely related to lithology, porosity, pore fill and other factors that can be used to produce more accurate and detailed structural and stratigraphic interpretations that can be obtained from seismic interpretation. The workflow of post-stack seismic inversion CSSI is showing in Figure 3.2.

Figure 3.2: P-impedance seismic inversion workflow using CSSI. Colors represent input, process and product. (Mesdag, 2013).

Acoustic-impedance inversion using CSSI involves the following steps:

• An impedance model is constructed using geologic modeling based on input horizons, structural information, and well control. This model is known as the earth model and it is an input data. Interpolated data from each horizon interval define the constraints in CSSI and the inversion trend model. The impedance model may be used to construct

51 low-frequency information that cannot be estimated from seismic data.

• Wavelets are estimated, preferably using well control. The goal is to minimize the misfit between the seismic trace and the synthetic seismic trace.

• A band-limited impedance model is obtained from the convolutional model using sparse spike inversion. Inversion is carried out with constraints from the trend model.

• The final impedance model is created by trace merging of the bandlimited impedance model and the low-frequency impedance model. This final impedance model contains frequencies from close to zero hertz to the highest frequencies present in the seismic data.

In pre-stack seismic inversion, both the low and the high frequency components of the P- wave acoustic impedance are extracted from the seismic data (Mallick, 1995). Low-frequency components of interval velocity are implicit in the moveout curves. When the full elastic earth model is included, in addition to the P-wave acoustic impedance, S-wave information, and Poisson’s ratio can also be estimated from pre-stack data. Pre-stack elastic inversion has a definite advantage over post-stack inversion methods. The workflow of pre-stack seismic inversion using CSSI is shown in Figure 3.3: The simultaneous inversion CSSI creates a set of elastic models from multiple seismic partial (angle or offsets) stacks. At each CMP the seismic data are modeled as the convolu- tion of a set of reflection coefficients with one or more wavelets. The reflection coefficients are estimated from elastic properties using the Aki-Richards approximation (only for PP reflectivity) (equation 3.3). The three principal steps of this process are:

1) Elastic parameter contrasts are estimated which create synthetics and simultaneously honor all of the input seismic stacks.

2) The elastic parameter contrasts are integrated