University of Southern California Dissertations and Theses

Integrated workflow for characterizing and modeling fracture network in unconventional reservoirs using microseismic data

(USC Thesis Other)

Integrated workflow for characterizing and modeling fracture network in unconventional reservoirs using microseismic data

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content INTEGRATED WORKFLOW FOR CHARACTERIZING AND MODELING FRACTURE NETWORK IN UNCONVENTIONAL RESERVOIRS USING MICROSEISMIC DATA by Tayeb Ayatollahy Tafti A Dissertation Presented to the FACULTY OF THE USC GRADUATE SCHOOL UNIVERSITY OF SOUTHERN CALIFORNIA In Partial Fulfillment of the Requirements for the Degree DOCTOR OF PHILOSOPHY (PETROLEUM ENGINEERING) December 2013 Copyright 2013 Tayeb Ayatollahy Tafti Dedication DedicatedtomybelovedwifeMinooforherpartnershipinthesuccess ofmylife ii Acknowledgments I would like to specially thank my supportive advisor, Professor Fred Aminzadeh for his great advices and kind support of my academic life. I am also grateful to Pro- fessor Iraj Ershaghi for his continued support and for providing me with illuminating advice throughout my education. It is my honor to thank Professor Muhammad Sahimi, Professor Charles Sammis, and Professor Kristian Jessen of USC for their continued collaboration and valuable inputs at various stages of this thesis. I would like to thank the Department of Energy for supporting this work at the Uni- versity of Southern California under ARRA GRANT10379400. I owe my gratitude to Chevron Corporations for providing me an internship opportunity in which I have gained invaluable industry experience. I am also indebted to all of my mentors, managers, and colleagues, who helped me with my projects at both the Information Technology Com- pany and the Energy Technology Company. I would like to acknowledge the developers of Zmap, Matlab, SGeMS, CMG, Eclipse, FracGen, Petrel, and Opendtect for providing us academic license. I also would like to thank Ernie Majer, Stephen Jarpe, and Katie Boyle for their expert advice and support in providing the initial velocity models and catalogs. I also benefited from interactions and many useful discussions with Joseph Beall, Mark Walter, and Craig Hartline from Calpine Corporation. iii My gratitude goes to anyone who has helped and motivated me in my progress through doctoral studies and to all of the good friends whom I have been fortunate to have in my life. I would like to thank all the staff members and graduate students at the Petroleum Engineering program for their support and friendliness; in particular, Hamid Reza Jahangiri, Reza Rastegar, Parham Ghods, Debotyam Maity, Mohammad Javaheri, Tao Yu, Mohammad Evazi, Shahram Farhadnia, Hasan Shojaie, Asal Rahimi Zeynal, Arman Khodabakhshnejad, Nima Jabbari, Amin Rezapour, Azarang Golmo- hammadi, Atefeh Jahandideh, Juli Legat, Xiaoyan Zhang, Sorin marghitoiu, Aimee Barnard, Tracey Charles, Margery Berti, Jennifer Gerson, and Idania Takimoto. In the end, my deepest gratitude goes to my parents, Mahvash and Javad, and my sister and brother, Taher and Afsar, for their constant support and encouragement. My special gratitude and appreciation goes to Minoo Malek for her unconditional love and care and for staying by my side whenever I needed encouragement and support. iv Publications produced from this work 1. Tayeb A. Tafti, Muhammad Sahimi, Fred Aminzadeh, and Charles G. Sammis, 2013, Use of microseismicity for determining the structure of the fracture net- work of large-scale porous media, Physical Review Letter E, V olume 87, Issue 3, 032152 2. Fred Aminzadeh, Tayeb A. Tafti, and Debotyam Maity, 2013, An Integrated Method- ology for Sub-surface Fracture Characterization using Microseismic data: A Case Study at the NW Geysers, Computer and Geosciences journal, V olume 54, April, Pages 39-49 3. Tayeb A. Tafti and Fred Aminzadeh, 2012, Characterizing fracture network in shale reservoir using microseismic data, SPE-153814-PP, Western Regional meet- ing, Bakersfield, CA 4. Tayeb A. Tafti and Fred Aminzadeh, 2012, Fracture Network Interpretation Through High Resolution Velocity Models: Application to The Geysers Geothermal Field, Geothermal Resources Council Transactions, V ol. 36, pp. 550-560. 5. Tayeb A. Tafti and Fred Aminzadeh, 2012, Time Lapse Stress and Rock Property Profiling using Microseismic Data: A Case Study at NW Geysers, AGU annual Meeting, San Francisco, CA v 6. Tayeb A. Tafti and Fred Aminzadeh, 2011, Application of high-resolution passive seismic tomographic inversion and estimating reservoir properties, AGU annual Meeting, San Francisco, CA 7. Tayeb A. Tafti and Fred Aminzadeh, 2011, Fracture characterization at The Gey- sers geothermal field using time lapse velocity modeling, fractal analysis and microseismic monitoring, Geothermal Resources Council Transactions, V ol. 35, pp. 547-551. 8. Fred Aminzadeh, Debotyam Maity, Tayeb A. Tafti, and Friso Brouwer, 2011, Artificial neural network based autopicker for micro-earthquake data, In: SEG Annual Meeting. pp. 1623-1626. 9. Fred Aminzadeh, Tayeb A. Tafti, Debotyam Maity, Katie Boyle, Muhammad Sahimi, and Charles G. Sammis, 2010, Analysis of microseismicity using fuzzy logic and fractals for fracture network characterization, AGU annual meeting, San Francisco, CA. 10. Fred Aminzadeh, Tayeb A. Tafti, and Debotyam Maity, 2010, Characterizing Fractures in Geysers Geothermal Field Using Soft Computing, Geothermal Resources Council Transactions, V ol. 34, pp. 1193-1198. 11. Fred Aminzadeh and Tayeb A. Tafti, 2009, Characterizing Fractures in Geysers Geothermal Field by Micro-seismic Data Using Soft Computing, Fractals, and Shear Wave Anisotropy, Geothermal Resource Council Annual meeting, Reno, NV . vi Table of Contents Dedication ii Acknowledgments iii Publications produced from this work v List of Figures x List of Tables xvii Abstract xviii Chapter 1: Introduction 1 1.1 Data and Study Area . . . . . . . . . . . . . . . . . . . . . . 4 1.2 Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6 1.3 Research Objectives . . . . . . . . . . . . . . . . . . . . . . . 6 1.4 Organization of the Manuscript . . . . . . . . . . . . . . . . . 8 Chapter 2: Primary Steps of the Workflow 15 2.1 Recording the Microseismic Events . . . . . . . . . . . . . . 15 2.2 First Arrival Picking for Shear and Compressional Waves . . . 16 2.3 Hypocentral Location and Tomographic Inversion . . . . . . . 18 2.4 Multiplet Analysis . . . . . . . . . . . . . . . . . . . . . . . . 19 2.5 Shear Wave Splitting . . . . . . . . . . . . . . . . . . . . . . 20 vii Chapter 3: Fuzzy Clustering 27 3.1 Fuzzy Clustering Algorithm . . . . . . . . . . . . . . . . . . . 27 3.2 Result and Discussion . . . . . . . . . . . . . . . . . . . . . . 28 3.3 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . 30 Chapter 4: Fractal Analysis 34 4.1 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . 35 4.2 True Fractal Dimension Analysis . . . . . . . . . . . . . . . . 37 4.3 Physical Interpretation of the Results . . . . . . . . . . . . . . 42 4.4 Concluding Remarks . . . . . . . . . . . . . . . . . . . . . . 48 Chapter 5: b-value Analysis 50 5.1 Background and Application of b-value Analysis . . . . . . . 50 5.2 b-value Analysis . . . . . . . . . . . . . . . . . . . . . . . . . 52 5.3 Result and Discussion . . . . . . . . . . . . . . . . . . . . . . 54 5.4 Fractal Dimension Versus b-value . . . . . . . . . . . . . . . 59 5.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . 64 Chapter 6: Application of Seismic Velocity Tomography in Frac- ture Characterization 68 6.1 Joint Interpretation of Seismic Wave Velocities . . . . . . . . 70 6.2 Stress and Rock Property Profiling . . . . . . . . . . . . . . . 74 6.2.1 Definitions: Stress and Elasticity . . . . . . . . . . . . 74 6.2.2 Stress and Rock Properties Derivation Using Seismic Velocity V olumes . . . . . . . . . . . . . . . . . . . . 76 6.3 Effect of Geologic Parameters and Uncertainties . . . . . . . . 80 6.4 Characterizing the Fracture Network Using Changes in Stress and Rock Properties . . . . . . . . . . . . . . . . . . . . . . . 85 6.4.1 Stress and Fracture . . . . . . . . . . . . . . . . . . . 85 6.4.2 Elastic Moduli and Fracture . . . . . . . . . . . . . . 86 6.5 Enhancing the Resolution of Seismic Velocity Models . . . . . 87 6.6 Fracture Network Interpretation through High Resolution Veloc- ity Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89 6.7 Tomographic Inversion Versus Fuzzy Clustering . . . . . . . . 91 viii 6.8 Tomographic Inversion Versus Production/Injection Data . . . 92 6.9 Time-lapse Fracture Characterization Using Tomographic Inver- sion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92 6.10 Concluding Remarks . . . . . . . . . . . . . . . . . . . . . . 115 Chapter 7: Modeling and Simulation of Hydraulic Fracturing using MDFN model 118 7.1 Background and Objective . . . . . . . . . . . . . . . . . . . 118 7.2 Microseismic based Discrete Fracture Network (MDFN) . . . 122 7.2.1 Estimating Fracture Aperture . . . . . . . . . . . . . . 123 7.2.2 Degree of Complexity and Fracture Pattern . . . . . . 124 7.2.3 Estimating Fracture Size . . . . . . . . . . . . . . . . 127 7.3 MDFN Case Studies . . . . . . . . . . . . . . . . . . . . . . . 129 7.3.1 Real field Case Study 1: Simple pattern . . . . . . . . 130 7.3.2 Real field Case Study 2: Complex Pattern . . . . . . . 132 7.4 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133 Chapter 8: Summary and Concluding Remarks 144 Chapter 9: Future Work 147 Bibliography 149 ix List of Figures 1.1 Application of microseismicity cloud to characterize the fracture network.(a) Associated microseismic cloud with production data (Phillips et al., 1998), (b)Original, absolute-determined hypocenters(Left) and relative mapping technique (right) (Rutledge et al., 1998), (c) Frac- ture growth through four stimulation times in Barnett shale (Maxwell et al., 2002) (d) Examples of microseismic cloud maps in tight sand- stone reservoirs (Warpinski et al., 2010). . . . . . . . . . . . . . . . 9 1.2 Novel microseismicity analaysis to characterize the fracture net- work, (a) Comparison of fault plane solutions and microseismic doublet distribution planes (Tezuka, 2000), (b) Most positive princi- pal curvature range of microseismic occurrence (top), P-impedance range of microseismic occurrence (Refunjol et al., 2010), (c) Micro- seismic events interpreted as related to fracture stimulation (red, b- value 2) and fault related events (blue, b-value 1), Barnett shale (Wessels et al., 2011), (d) Evolution of the P-wave seismic velocity at one horizon during the stimulation test (Charlety et al., 2006). . . 10 1.3 Unconventional reservoirs in which sufficient microseismic activity is present.(a) Shale and tight sand reservoirs, (b) Enhanced geother- mal systems(EGS) . . . . . . . . . . . . . . . . . . . . . . . . . . . 11 1.4 Principle of microseismic fracture mapping (Cipolla and Wright, 2000) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11 1.5 Map of The Geysers geothermal field (GGF), the area under study (in the rectangular area), and the locations of the injection wells and seismic activities (after Beall,J., Calpine). . . . . . . . . . . . . . . 12 1.6 Integrated workflow architecture for characterizing a fracture net- work in unconventional reservoirs using microseismic data. . . . . . 13 1.7 The schematic view for the integration of microseismic data with other sources of data. . . . . . . . . . . . . . . . . . . . . . . . . . 14 x 2.1 (a) Downhole monitoring for microseismic data (ASC, 2011)(b) Loca- tion of the surface receivers (white) and treatment well location (red) (Diller and Gardner, 2011). . . . . . . . . . . . . . . . . . . . . . . 16 2.2 (a) ANN Autopicking workflow, (b) sample pick with ANN outputs and actual traces compared. . . . . . . . . . . . . . . . . . . . . . . 22 2.3 Schematic view of passive seismic tomography (Duncan, 2005). This can be modified horizontally for Downhole monitoring. . . . . 23 2.4 Microseismic event cloud in 3D at The Geysers . . . . . . . . . . . 23 2.5 The initial velocity model provided by LBNL . . . . . . . . . . . . 23 2.6 A decision tree for the real-time application of the multiplet-identification algorithm in the analysis of microseismic events. (Arrowsmith and Eisner, 2006) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24 2.7 (a) Plane of fractures related to multiplet analysis, (b) the microseis- mic locations colored with respect to the multiplet groups. Red lines represent two dominated fracture system interpreted by the princi- ple direction of the clusters and by connecting the clusters.(Tezuka et al., 2008) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24 2.8 Shear wave splitting in anistropic medida (Garnero, 2011) . . . . . . 25 2.9 The representative waveforms and observed splitting for the events inside the shear-wave window (Vlahovic et al., 2003) . . . . . . . . 25 2.10 Rose diagrams of the fast shear-wave polarization directions observed, (a) at Icleland stations (Tang et al., 2008), (b)at The NW Geysers (Elkibbi, 2005) . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26 2.11 Equal-area projection plots of t measurements for the stations in the NW Geysers (Elkibbi, 2005). . . . . . . . . . . . . . . . . . . . 26 3.1 Selected zones for fuzzy cluster center evaluation . . . . . . . . . . 29 3.2 Workflow to find the fuzzy cluster centers of microseismic Data . . . 30 3.3 Zones boundary for applying the fuzzy clustering technique and the production wells locations. . . . . . . . . . . . . . . . . . . . . . . 31 3.4 Fuzzy cluster centers for all the years at the HTZ zone . . . . . . . . 31 xi 3.5 Fuzzy cluster centers for all the years at the zone (a)one, (b)two, (c)three, (d)four, (e)five. . . . . . . . . . . . . . . . . . . . . . . . . 32 3.6 Movement of microseismic cluster centers at The Geysers; start of arrow is cluster center in 2006; end of arrow is cluster center in 2010 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33 4.1 Clusters of the earthquakes hypocenters and the locations of active injection wells from 2006 to 2011. Each point represents one event. 42 4.2 The four regions studied, as well as the four subregions. Each point represents the location of an event. . . . . . . . . . . . . . . . . . . 43 4.3 The logarithmic plot of the correlation function C(r) vs distance in region 2 of The Geysers geothermal field. Also shown are the variations of the local slopes around a constant value, indicating the accuracy of the data and the overall slope of the plot. . . . . . . . . 44 4.4 Dependence of the fractal dimensionD f on the density of the micro- seismic events in region 1 from 2006 to 2010. . . . . . . . . . . . . 45 4.5 Same as in Figure 4.4, but for the four subregions carved out of region 2 from 20062010. . . . . . . . . . . . . . . . . . . . . . . . 46 4.6 Same as in Figure 4.4, but for the northwest region, from 2006 to 2010. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47 5.1 b-values of normal (green), strike-slip (red) and thrust events (blue); average b-value (grey line); standard error (vertical bars)(Schorlemmer et al., 2005; Zoback, 2007) . . . . . . . . . . . . . . . . . . . . . . 51 5.2 Microseismic histogram of fault (blue) and fracture related events (red)(Wessels et al., 2011) . . . . . . . . . . . . . . . . . . . . . . . 53 5.3 b-value map within the fracture treatment area. Lower values are associated with fault activity whereas higher b values are indicative of more fracture event creation.(Wessels et al., 2011) . . . . . . . . 54 5.4 Comparison of b-value analysis for different stages of stimulation in horizontal Barnett Shale well (Downie et al., 2010) . . . . . . . . . 55 5.5 Temporal evolution of b-values for a heavy-oil dataset to character- ize opening or closing phase of fractures.), (Grob and van der Baan, 2011) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55 xii 5.6 Magnitude distributions of The Geysers seismicity in 2006 . . . . . 56 5.7 Map for west of Kern County, the area under study. . . . . . . . . . 59 5.8 Distribution of earthquakes magnitudes in The Geysers. . . . . . . . 60 5.9 Distribution of earthquakes magnitudes in west Kern County( 0-10 km) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61 5.10 The frequency-magnitude plot for extracting the b value for the entire seismicity catalog of west Kern County from 1990 to 2013 . . 62 5.11 Time dependence of the b values in the four regions of The Gey- sers. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63 5.12 Annual probability and recurrence time for seismicity in west Kern County . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64 5.13 Annual probability and recurrence time for seismicity at The GGF 65 5.14 Schematic diagram showing two different types of damage localiza- tion: (a) Model A, (b) Model B. The left diagrams show the ini- tial seismicity distribution; the middle diagrams show the evolution toward more concentrated activity; and the right diagrams show the correlation plotP r (C r ) associated with this change.(Main, 1992) . . 66 6.1 V P section showing structural details and V P =V S section showing lithological details.(Martakis et al., 2006) . . . . . . . . . . . . . . 70 6.2 Factors affecting seismic compresional velocity - Qualitative overview, www.seg.org . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71 6.3 Seismic velocities for various rock samples versus (a) depth for dry samples, (b) porosity for dry samples, (c) effective confining pres- sure for dry and brine saturated rocks (Boitnott, 2003). . . . . . . . 72 6.4 Compressional, extensional, and shear stress in rock (Image cour- tesy of Michael Kimberly, North Carolina State Univ.) . . . . . . . . 75 6.5 Cartoons describing (a) Bulk modulus, (b) Young modulus, (c) Shear modulus (Nave, 2010), (d) Poisson’s ratio (Christopher et al., 2006), (e) confining stress(Cramez, 2006)) . . . . . . . . . . . . . . . . . . 76 6.6 Elastic deformations and particle motions associated with the prop- agation of seismic waves: (a) P-wave, (b) S-wave (Barton, 2006) . . 79 xiii 6.7 Created density volume based on available lithology logs at the NW Geysers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82 6.8 Density and Bulk Modulus for 4 density realizations at the NTR . . 83 6.9 Normalized absolute standard deviations along every evaluation point on selected horizons for all density realizations indicating small variability of target properties . . . . . . . . . . . . . . . . . . . . 84 6.10 Difference maps (percentage difference) for various Bulk modulus realizations. The baseline density used for computation is a con- stant value from the horizon of interest. We observe the percentage difference to be< 5% . . . . . . . . . . . . . . . . . . . . . . . . 94 6.11 (a) Diagram illustrating the fracture subjected to dimensional stress condition (Tezuka, 2000), (b) Associated fracture mode to different states of stress (Rountree et al., 2002) . . . . . . . . . . . . . . . . 95 6.12 Fracture opening for different normal stresses in a hydraulic fractur- ing job (Dershowitz and Doe, 2011) . . . . . . . . . . . . . . . . . 95 6.13 (a) 3-dimensional Mohr-Coulomb diagram showing shear and nor- mal stress on a fracture surface as a function of insitu stress field. The dark triangles represent fractures that are experiencing a state of stress sufficient to inducing shear failure whereas the dots represent fractures in a more stable state of stress (Ameen, 2003) (b) Change of differential stress during hydraulic fracturing job to reach failure mode and create fracture . . . . . . . . . . . . . . . . . . . . . . . 96 6.14 Coefficient of friction as a function of orientation, displayed stereo- graphically as a lower hemisphere polar projection (Ameen, 2003). 96 6.15 Bulk and Shear modulus versus pressure and brine saturation for two different core samples (Boitnott, 2003). . . . . . . . . . . . . . 97 6.16 Laboratory measurement of crack effect on bulk moduli(Berge et al., 2001) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97 6.17 Enhancing the resolution of seismic velocity models, (a) the ini- tial velocity model, (b) variogram model based on initial velocity model,(c) actual output of tomographic inversion (black dots) and new kriged data point (orange dots), (d) the final kriged velocity model. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98 xiv 6.18 Velocity anomalies around active injection wells, Inline view . . . . 99 6.19 Velocity anomalies below injection wells; the lateral extension of these anomalies increases up to the middle of a normal temperature reservoir–where most of wells are completed –and then decreases by depth . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100 6.20 NTR horizon at The NW Geysers (a)V P , (b)V S , (c)V P =V S , HTZ horizon (d)V P , (e)V S , (f)V P =V S . . . . . . . . . . . . . . . . . . . 101 6.21 Laboratory measurement for The Geysers (Berge et al., 2001) . . . . 102 6.22 NTR horizon at The Geyser (a) Poisson’s ratio, (b) Normal stress, (c) Confining stress, HTZ horizon at The Geyser (d) Poisson’s ratio, (e) Normal stress, (f) Confining stress . . . . . . . . . . . . . . . . 103 6.23 Correlation between microseismic cluster movement and velocity anomaly direction; the red circle is the microseismic cluster center in 2006 and the yellow one is for 2009 . . . . . . . . . . . . . . . . 104 6.24 NTR horizon at The NW Geysers, (a)V P , (b)V P =V S , (c) Poisson’s ratio (d) Normal stress, with production/injection data superimposed (Production: White, Injection: Black) . . . . . . . . . . . . . . . . 105 6.25 Compressional wave seismic velocity distribution in NTR horizon at The NW Geysers (a) 2005, (b) 2006, (c) 2007, (d) 2008, (e) 2009, (f) 2010 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108 6.26 Shear wave seismic velocity distribution in NTR horizon at The NW Geysers (a) 2005, (b) 2006, (c) 2007,(d) 2008, (e) 2009, (f) 2010 . . 111 6.27 Bulk modulus distribution in NTR horizon at The NW Geysers (a)2005, (b)2006, (c)2007, (d)2008, (e)2009, (f)2010 . . . . . . . . . . . . . 112 6.28 Shear modulus distribution in NTR horizon at The NW Geysers (a)2005, (b)2006, (c)2007, (d)2008, (e)2009, (f)2010 . . . . . . . . 113 6.29 Young modulus distribution in NTR horizon at The NW Geysers (a)2005, (b)2006, (c)2007,(d)2008, (e)2009, (f)2010 . . . . . . . . . 114 6.30 Normal stress distribution in NTR horizon at The NW Geysers (a)2005, (b)2006, (c)2007, (d)2008, (e)2009, (f)2010 . . . . . . . . . . . . . 117 7.1 Authors’ visualization of complex fracture network from Warpinski and Teufel (1987); Fisher and Wright (2002) . . . . . . . . . . . . 135 xv 7.2 (a) Discrete fracture model workflow (b) Source of data for DFN (b) Defining property distribution for fracture properties (d) Sample created network (Dershowitz and Doe, 2011) . . . . . . . . . . . . . 136 7.3 Microseismic based discrete fracture network (MDFN) model work- flow . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137 7.4 Conceptual model for aperture - normal stress relationship in frac- tured rocks Rutqvist (2002). . . . . . . . . . . . . . . . . . . . . . . 137 7.5 Aperture distribution in GGF core samples from CT scan result (He, 1998) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 138 7.6 Aperture distribution in NTR horizon at The NW GGF (a)2005, (b)2010 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 138 7.7 frequency-magnitude distributions for different stages of stimula- tion in horizontal Barnett Shale well, (after Downie et al. (2010)) . . 139 7.8 Calculated frequency-fracture length distributions for different stages of stimulation in horizontal Barnett Shale well, . . . . . . . . . . . 139 7.9 Microseismic distributions for hydraulic fracture treatment where simple seismicity pattern is observed at Barnett Shale,D f = 1, with its associated production (after Maxwell et al. (2002)) . . . . . . . . 140 7.10 Fracture model and history match when microseismic distribution hasD f ' 1:2 0:1 in 2D orD f ' 2 0:1 in 3D . . . . . . . . . . 141 7.11 Gas production mechanisms from fractured reservoir in tight forma- tion such as Shale(Ghods, 2012) . . . . . . . . . . . . . . . . . . . 141 7.12 Microseismic distributions for hydraulic fracture treatment, where a complex seismicity pattern is observed at Barnett Shale,Df' 1:82, with its associated production (after Maxwell et al. (2002)) . . . . . 142 7.13 Fracture model to match the fractal dimension of seismicity distri- bution (D f ' 1:82 0:16 and its fractal dimension calculated using the box count method . . . . . . . . . . . . . . . . . . . . . . . . . 143 7.14 Fracture model and history match when microseismic distribution hasD f ' 1:89 0:05 in 2D orD f ' 2:5 0:1 in 3D . . . . . . . . 143 xvi List of Tables 4.1 Estimates of the fractal dimensions for the individual regions. . . . . 41 5.1 b-values versus stress regime and dominant faulting mechanism, based on work by Schorlemmer et al. (2005) . . . . . . . . . . . . 52 5.2 Estimates of theb values for the individual regions. Estimates ofb are for the 2006 to 2010 period. . . . . . . . . . . . . . . . . . . . 58 5.3 Observation of large earthquakes in seismicity history of west Kern County, 1990 to 2013 between 0 to 10 km depth, from SCSN cata- log. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59 5.4 Observation of large earthquakes in seismicity history of The GGF, 1990 to 2013 between 0 to 7 km depth, from NCEDC catalog. . . . 60 5.5 Estimates of the fractal dimensions and the b values for the individ- ual regions. Estimates of b are for the 20062010 period. . . . . . . . 67 6.1 Effect of saturation on velocities (Boitnott, 2003). . . . . . . . . . . 73 6.2 Major Production= injection wells within the area of interest, with their associated rates from DOGGR database. . . . . . . . . . . . . 106 6.3 Production= injection well clustering . . . . . . . . . . . . . . . . 107 7.1 Obtaining fracture properties using microseismic data for MDFN modeling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123 7.2 Empirical relationship among seismic moment magnitude, rupture length, width, area, and average displacement (Wells and Copper- smith, 1994) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128 7.3 Set of reservoir parameters . . . . . . . . . . . . . . . . . . . . . . 131 xvii Abstract We develop a new method for integrating information and data from different sources. We also construct a comprehensive workflow for characterizing and modeling a frac- ture network in unconventional reservoirs, using microseismic data. The methodology is based on combination of several mathematical and artificial intelligent techniques, including geostatistics, fractal analysis, fuzzy logic, and neural networks. The study contributes to scholarly knowledge base on the characterization and modeling fractured reservoirs in several ways; including a versatile workflow with a novel objective func- tions. Some the characteristics of the methods are listed below: 1. The new method is an effective fracture characterization procedure estimates dif- ferent fracture properties. Unlike the existing methods, the new approach is not dependent on the location of events. It is able to integrate all multi-scaled and diverse fracture information from different methodologies. 2. It offers an improved procedure to create compressional and shear velocity models as a preamble for delineating anomalies and map structures of interest and to correlate velocity anomalies with fracture swarms and other reservoir properties of interest. xviii 3. It offers an effective way to obtain the fractal dimension of microseismic events and identify the pattern complexity, connectivity, and mechanism of the created fracture network. 4. It offers an innovative method for monitoring the fracture movement in different stages of stimulation that can be used to optimize the process. 5. Our newly developed MDFN approach allows to create a discrete fracture network model using only microseismic data with potential cost reduction. It also imposes fractal dimension as a constraint on other fracture modeling approaches, which increases the visual similarity between the modeled networks and the real network over the simulated volume. xix Chapter 1 Introduction Unconventional reservoirs have low permeability and porosity; thus, they are difficult to produce, and often stimulation techniques, such as hydraulic fracturing, must be per- formed. Examples of unconventional reservoirs are tight gas, shale reservoirs, and enhanced geothermal systems (EGS) (Figure 1.3). Fractures provide permeability for fluid movement and play an important role in production from this kind of reservoir. Hence, the characterization of these fractures has become vitally important considera- tion for every aspect of exploration, development, and production in order to find new energy resources. Fractures are mechanical breaks or discontinuities in rocks that naturally form in response to high fluid pressures, lithostatic, thermal, and tectonic stresses. During the injection or production of fluid, microseismic events can be induced on created fractures or triggered on natural fractures or faults (Figure 1.4). In many geothermal reservoirs, induced seismicity (micro-seismic events) can be caused by reactivated shears created fractures or the natural fractures in shear zones and faults. Although many character- ization methods exist, microseismic monitoring probably offers the best resolution to image fracture complexity. Fractures can grow asymmetrically, have variable confine- ment across geologic interfaces, and change orientation. Fracture growth in a naturally fractured reservoir has additional complexities associated with interaction between the hydraulic fracture and the pre-existing fracture network (Maxwell et al., 2002). Several earlier studies reported that the total stimulated reservoir volume from microseismic active volume relates to well production. Also, this volume has been related to surface 1 contact area with the reservoir and, ultimately, to production rates and reservoir drainage (Maxwell, 2011). With the aim of monitoring microseismic data, we can finally estimate the properties of this kind of complexity, modify reservoir models, predict drainage pat- terns, assess changes to hydraulic fracturing design, and possibly optimize the stimula- tion process. Today, microseismic monitoring is a proven technology for characterizing a frac- ture network created from fluid injection and hydraulic fracturing. In the last decade, many authors have published their research into using the spatial/temporal progression of microseismic activity to interpret the fracture height, length, azimuth, zonal coverage, and fracture complexity in terms of a simple, planar fracture, or a complex fracture net- work. They have also correlated production data with the dimensions of the microseis- mic clouds and volume estimates based on the density of microseismic events (Albright and Pearson, 1982; Brady et al., 1994; Rutledge et al., 1998; Phillips et al., 1998; Fisher et al., 2004; Downie et al., 2009; Barree et al., 2002; Xu and Calvez, 2009; Warpin- ski et al., 2005; Tezuka et al., 2008). Among these researchers, Albright and Pearson (1982) reported that high pore pressure correlates with microseismicity and that perme- ability may be estimated from this relationship. He also used microseismic hypocentral location as an indicator for fracture location. Brady et al. (1994) demonstrated that comparing different monitoring wells to find the cloud of microseismicity can elimi- nate the errors in determining associated fracture network height and width. Rutledge et al. (1998) used relative mapping of microseismic events to obtain high precision frac- ture images (Figure 1.1b). Phillips et al. (1998) applied Hodogram-inclination to locate microseismic events and, consequently, fracture zones. Then, they used the cloud of microseismicity patterns to indicate different stimulation-zone widths; the wider zone yielded higher post-stimulation production. They concluded that microseismic cloud 2 represents the high-porosity, low-angle, reverse-slip fracture zones (Figure 1.1a). Fig- ure 1.1 shows how a cloud of microseismicity may help accomplish this tedious task. However, other content of microseismic events can potentially provide additional insight into the fracturing process. With recent advances in computing techniques, some authors have tried to use methods other than hypocentral location analysis to character- ize the fracture network. For instance, Tezuka (2000); Moriya et al. (2000); Rowe et al. (2002); Rutledge and Phillips (2003); Baig and Urbancic (2010) used focal mechanism, moment tensors, and microseismic multiplet analysis to delineate the reservoir structure and to examine the associated permeability enhancements later. The microseismic dou- blets tend to align on the plane (fracture) that has been activated by hydraulic stimulation (Figure 1.2a). Refunjol et al. (2010) implemented joint interpretation of microseismic cloud and seismic attributes such as curvature, and the P- and S- impedance (Figure 1.2b). Wessels et al. (2011); Grob and van der Baan (2011); Downie et al. (2010) ana- lyzed both b-value and fractal dimension of microseismic events to identify open and closed fractures or to distinguish the microseismic events from associated near-fault effects or created fracture network (Figure 1.2c). Grechka and Mazumdar (2010); Hum- mel and Shapiro (2011); Rozhko (2010) predicted the permeability and production of hydraulically fractured hydrocarbon reservoirs from microseismic data using diffusiv- ity equations, microseismic event cloud geometry, and their temporal changes. Finally, Charlety et al. (2006) used 4D tomographic inversion at the Soultz Enhanced Geother- mal System (EGS) site to evaluate the fracture stimulation process. As mentioned, many scientists use microseismic data with different methods to understand fracture properties in unconventional reservoirs, but none of them has intro- duced a robust workflow to utilize most information content from microseismic events. In this thesis, we introduce a workflow that has the advantage of being more comprehen- sive and flexible than current applied technology. Here, we develop innovative methods 3 to utilize different types of information inherent to 3 Component (3C) microseismic data. We demonstrate how such data can be integrated to characterize the fracture net- work, thus providing a better understanding of the fracture network geometry, connec- tivity, and density. We go beyond existing methods that use the microseismic cloud to locate the fracture network. Our technical analysis of microseismic data involves an integrated workflow, which allows us to utilize other information content about the events, such as their size, relationship to other events, attributes, and relationship to other data (conventional seismic, well data, etc.). Based on the theory of geostatistics, fractal geometry, fuzzy logic, and neural network, we have developed a methodology for the comprehensive characterization of fractured reservoirs in which sufficient microseismic activity is present. 1.1 Data and Study Area We apply most techniques discussed in this thesis to the data from The Geysers geother- mal field (GGF), which provides a large acquisition array and high levels of microseis- micity, to test the proposed methods and workflows. We used the dataset provided by Lawrence Berkeley National Laboratory (LBNL) and the online data set from Northern California Earthquake Data Center (NCEDC) 1 from 2006 to 2011 earthquake catalogs. We analyzed initial velocity models provided by the LBNL extracted from 2004-2010 events (Boyle et al., 2011). The area covered by our study was the northwest (NW) region of the GGF, indicated by the rectangle in Figure 1.5. The GGF is located about 150 km north of San Francisco, California; see Figure 1.5. The field contains a large number of wells, some of which are used for injecting cold water into the metamorphic and crystalline rock. When water comes into contact with 1 http://www.ncedc.org/SeismiQuery/events_f.html 4 the hot matrix, it evaporates, generating steam produced from a network of fractures in the crystalline rock. Most of the induced fractures at The Geysers are created through the thermal contraction process (Stark, 1990; Rutqvist et al., 2006). At The Geysers, water is not pumped into the reservoir under pressure, but rather ”free falls” down the well bore to the reservoir level. Limited reservoir pressure results from the development of a standing water column near the base of the well, but is generally insufficient to initiate fracturing. The cold fluid injectate interacts with hot rock causing contraction at and near fracture surfaces. Rutqvist et al. (2006) reported that evaporation cooling with contraction reduces both effective stress and static friction, triggers slip along planes of weakness, and results in the slight opening of fracture. Due to the very low permeability of the GGF’s formation matrix, steam production depends on the presence of a fracture network. Cost-effective production of the steam requires that the trajectories of the wells intersect with densely fractured regions. Therefore, locating the fracture network and identifying regions with higher fracture density is important to the economics of steam production at the GGF. Various approaches have been already used to characterize the fracture network of the GGF, including geologic mapping (Hebein, 1986; Sternfeld, 1989), outcrop analysis (Sammis et al., 1991), core analysis (Nielson et al., 1991), shear wave splitting (Lou et al., 1997; Malin and Shalev, 1999; Erten et al., 2001; Elkibbi et al., 2004), and tomographic inversion (Gritto et al., 2013). We also elaborate more on fracture characterization at GGF using microseismic data in our recent publications; see Tafti et al. (2013); Aminzadeh et al. (2013). In The Geysers, the producing reservoir rock is a Jurassic-Cretaceous Franciscan graywacke. Pleistocene Felsite is the basement rock, which is intrusive in some parts of the reservoir 2 . Finally, the cap rock is a melange of metamorphic Franciscan rocks (Berge et al., 2001). 2 The zone for developing EGS 5 1.2 Motivation Large amounts of oil and gas are currently being produced from unconventional reser- voirs, especially shale ones. At the same time, there is considerable interest in develop- ing enhanced geothermal systems. These reservoirs have such low porosity and perme- ability that they must be effectively stimulated by hydraulic fracturing jobs. In addition, economic production from these reservoirs requires a well trajectory to intersect with densely fractured volume of either created or natural one. Once the fractures are delin- eated, operators can create reservoir models to predict production, calculate reserves, or design future stimulation jobs. Due to the high level of interest in unconventional reservoirs, a surge in the micro- seismic monitoring of hydraulic fracturing has occurred in recent years. To avoid mak- ing assumptions and to better understand the created fracture geometry, many efforts have aimed at improving the methodology of using these data and eliminating any ambiguity within them. But, important unanswered questions remain. Many questions answered by seismologists, but their findings or theories have not been implemented completely in microseismic monitoring. The motivation of this study is to introduce an innovative workflow for characterizing fracture networks using microseismic data to fill the gap between the highly studied subject of earthquake seismology and engineering concerns related to developing these reservoirs. 1.3 Research Objectives Microseismic evaluation that considers the location of microseismic activity, microseis- mic dimensions, and microseismic volume cannot always provide an explanation for changes in observed fracture behavior or determine if those changes are being caused by the presence of faults near (tectonic) or in contact with the fracturing treatment(induced). 6 In addition, they may not completely delineate the significant variability and complexity in fracture growth during well stimulation operations or fluid injection which is observ- able especially in naturally fractured reservoirs such as Barnett shale or The Geysers. We believe that monitoring microseismicity data can define the location of fractures, fracture size, shape, density, and orientation if a robust workflow is applied in analyzing them. This workflow should allow the creation of improved models based on introduced concepts and can be supported by real-time data. The objective of this study is to introduce an integrated workflow to apply all useful information from microseismicity. This critical insight allows the stimulation treatment plan to be refined and optimized. It also provides useful tools for long-term improve- ments to the well spacing plan, the well design, and the completion design. In this methodology, microseismic data in the form of seismic response are processed and then go through various schemes to obtain useful results for fracture characterization. The workflow is initiated by calculating the hypocentral location of events and by detecting first arrival times for both compressional and shear waves. Workflow then proceeds with fuzzy clustering, fractal analysis, b-value analysis, tomographic inversion, stress analy- sis, and shear wave splitting. Figure 1.6 shows the workflow architecture, and how each step may help us map the fracture network and estimate the fracture properties. In addi- tion, the extracted results can be integrated with other sources of data, such as well data, seismic, and geology, for both validation and achieving higher accuracy in our estimates (Figure 1.7). We believe that the advent of cheap geophone sensors and a rising use of micro- seismic data analysis, has made it increasingly important to have a fast, efficient, and accurate workflow to monitor this kind of data. 7 1.4 Organization of the Manuscript The remainder of this manuscript is organized as follows to describe each step of our innovative workflow. The next chapter presents a brief overview of our workflow’s primary steps, which are currently being implemented in some microseismic monitoring projects worldwide. Chapter 3 discusses a new fuzzy clustering technique for finding fracture movement direction and zones in which fractures penetrate from stimulation. Chapter 4 describes the significance of the impact of fractal analysis on the evaluation of microseismic events spatial/temporal distribution. The impact of considering b-value analysis in this distribution is presented in chapter 5. In chapter 6, we demonstrate how velocity models and tomographic inversion can help us characterize the fracture network with various approaches. Finally, chapter 7 concludes the manuscript by presenting a new approach to modeling the fracture network in unconventional reservoirs integrating information obtained from previous chapters. 8 The storage capacity computed for the most seismically active fracture, based on mapped surface area and well-log porosity estimates, implies that total oil production represents about 20% of the pore volume of the fracture. 40 The presence of low-angle, oil-bearing fractures has implications for field development. Drilling horizontal or deviated wells should not increase the probability of intersecting productive fractures. Dipmeter and formation microscanning logs may be very useful in determining the orientations of low-angle, productive fractures and thereby may aid in a more effective placement of offset wells. Interwell correlation and mapping of the conductive fractures will allow better planning in plug-and-abandonment operations so as to avoid premature contamination of pay zones with water. Pressure- maintenance operations also could be attempted once the conduc- tive fracture zones between the wells have been mapped. Conclusions 1. More than 480 and 770 stimulation-induced microearthquakes were recorded at two sites in the Austin chalk, Giddings field, Texas, and more than 3,200 production-induced microearthquakes Fig. 11—Pre- and post-stimulation production rates, Wells CPU2-2 and Matcek 1, Giddings field, Texas. Fig. 10—Map and cross section (C-D) views showing event locations associated with production in Well HT1, Clinton County, Kentucky. 119 SPE Reservoir Evaluation & Engineering, April 1998 (a) 250 250 300 350 400 450 200 150 100 50 Distance North (ft) 8950 9000 9050 9100 9150 Depth (ft) 8950 9000 9050 9100 9150 Depth (ft) Distance East (ft) Map View View Along Strike View Across Strike -100 -50 0 50 100 Distance (ft) -100 -50 0 50 100 Distance (ft) 3Oi i l b l d i dh Thi l i 250 250 300 350 400 450 200 150 100 50 Distance North (ft) Distance East (ft) 8950 9000 9050 9100 9150 Depth (ft) 8950 9000 9050 9100 9150 Depth (ft) -100 -50 0 50 100 Distance (ft) -100 -50 0 50 100 Distance (ft) Map View View Along Str ke View Across Str ke Fi 4 H f i l i l (b) 2 S.C. MAXWELL, T.I. URBANCIC, N. STEINSBERGER & R. ZINNO SPE 77440 wireline array is mechanically clamped in place ensuring mechanical coupling with the rock. The imaging used ESG’s hydraulic fracture microseismic diagnostic commercial service, FRACMAP 7 . Continuous acoustic signals from the geophone array are input into ESG’s microearthquake detection systems, which archive the detected events and automatically process for 3D event locations, location errors, magnitudes and other seismic source attributes 8 . The microseismic results are then transmitted using radio telemetry to the frac van, and available to the engineer during the frac for real-time diagnostics of the fracture characteristics. Since the images are available to the engineer during the treatment, they can be used to diagnose the effectiveness of both planned and reactive frac operations. For example, the technology can be used “on the fly” to image the frac and change the injection parameters to gain the desired fracture characteristics. In this report, the real-time images are used to assess the effectiveness of controlling the fracture growth in the presence of complex fracture geometries controlled by the fracture network. Microseismic images can also be useful in a post-mortem analysis of the fracture 9 , to: 1. calibrate numerical simulations, 2. predict drainage patterns, 3. assess changes to frac design, and 4. optimize frac design. Fracture Complexity in the Barnett Shale Microseismic imaging of hydraulic fracturing in the Barnett Shale has shown dramatic interaction of the stimulations with the pre-existing fracture networks, generating complex fracture geometry. Figure 1 shows the image of one well (described as Well ‘A’ here), resulting from a 56 bpm - 6 hour water frac. The image shows events distributed over three main parallel lineations, trending NE-SW (parallel to the anticipated frac orientation) at a spacing of roughly 300’. In this example, the average location accuracy of the individual microseismic events is approximately 45’. The lineations are interpreted to be the interaction of the frac slurry with pre- Figure 1. Plan view of stimulation of well ‘A’ showing the fracture growth through four periods, a) 80 minutes, b) 130 minutes, c) 190 minutes, d) end of the treatment. Microseismic event locations scaled by magnitude. (c) 6 SPE 131776 natural fractures. These fractures are probably the most important for production, are potentially damageable by the stimulation fluids, and are possibly the reason for the wider microseismic envelopes and shorter fracture lengths (enhanced leakoff). -1200 -800 -400 0 400 800 1200 1600 -2000 -1600 -1200 -800 -400 0 400 Easting (ft) Northing (ft) -1000 -600 -200 200 600 1000 1400 -1200 -800 -400 0 400 800 1200 Easting (ft) Northing (ft) -400 0 400 800 1200 1600 -1200 -800 -400 0 400 Easting (ft) Northing (ft) Stage 1 Stage 2 Stage 3 Stage 4 Stage 5 Stage 6 Stage 7 Stage 8 Stage 9 -1200 -800 -400 0 400 800 1200 -1000 -600 -200 200 600 1000 Easting (ft) Northing (ft) Stage 1 Stage 2 Stage 3 Stage 4 Stage 1 Stage 2 Stage 3 Stage 4 -1400 -1000 -600 -200 200 600 1000 1400 1800 -2200 -1800 -1400 -1000 -600 -200 200 600 1000 1400 Easting (ft) Northing (ft) -400 0 400 800 -1600 -1200 -800 -400 0 400 800 1200 Easting (ft) Northing (ft) -1200 -800 -400 0 400 800 -1200 -800 -400 0 400 800 Easting (ft) Northing (ft) Piceance Basin – Grand Valley field Piceance Basin – Rulison field Piceance Basin – Mamm Creek field Uintah Basin West Tavaputs field Greater Green River Basin – Jonah field Greater Green River Basin – Jonah field San Juan Basin – Rio Arriba County SPE 124673 SPE 95637 SPE 95637 SPE 108103 Malone et al., 2009 Malone et al. 2009 Weijers et al. 2009 Monitor Well Monitor Well Monitor Well Monitor Well Monitor Well Monitor Well Monitor Well Monitor Well -1200 -800 -400 0 400 800 1200 1600 -2000 -1600 -1200 -800 -400 0 400 Easting (ft) Northing (ft) -1000 -600 -200 200 600 1000 1400 -1200 -800 -400 0 400 800 1200 Easting (ft) Northing (ft) -400 0 400 800 1200 1600 -1200 -800 -400 0 400 Easting (ft) Northing (ft) Stage 1 Stage 2 Stage 3 Stage 4 Stage 5 Stage 6 Stage 7 Stage 8 Stage 9 -1200 -800 -400 0 400 800 1200 -1000 -600 -200 200 600 1000 Easting (ft) Northing (ft) Stage 1 Stage 2 Stage 3 Stage 4 Stage 1 Stage 2 Stage 3 Stage 4 -1400 -1000 -600 -200 200 600 1000 1400 1800 -2200 -1800 -1400 -1000 -600 -200 200 600 1000 1400 Easting (ft) Northing (ft) -400 0 400 800 -1600 -1200 -800 -400 0 400 800 1200 Easting (ft) Northing (ft) -1200 -800 -400 0 400 800 -1200 -800 -400 0 400 800 Easting (ft) Northing (ft) Piceance Basin – Grand Valley field Piceance Basin – Rulison field Piceance Basin – Mamm Creek field Uintah Basin West Tavaputs field Greater Green River Basin – Jonah field Greater Green River Basin – Jonah field San Juan Basin – Rio Arriba County SPE 124673 SPE 95637 SPE 95637 SPE 108103 Malone et al., 2009 Malone et al. 2009 Weijers et al. 2009 Monitor Well Monitor Well Monitor Well Monitor Well Monitor Well Monitor Well Monitor Well Monitor Well Fig. 4—Examples of microseismic maps in lenticular tight sandstone reservoirs, with same scale for direct comparison. The third example in the Piceance basin is from Mamm Creek field (Weijers et al. 2009), which is east to southeast of the previous two projects. Fracture treatments in Mamm Creek by the operator of these projects have more stages than those from the previous example and larger total volumes. Resultant fracture lengths are greater than those in Grand Valley, ranging from about 750 ft to as much as 1,600 ft. Some interesting features are immediately observable in the results. First, both wings of the fracture can be resolved because there are two monitoring wells. Fracture wings appear to be symmetric, although the southeast wings of the fractures are probably better resolved because the monitor well is farther out on that side than it is on the northwest wings. Second, there is a general fan-shaped appearance to the microseismic envelopes, which is a result of a steady rotation of the fracture azimuths as treatments progress uphole. This rotation is consistent with a similar rotation observed in the MWX test in the Piceance basin (Warpinski and Teufel 1990). Because of this rotation, it is difficult to assess the width of the microseismic envelope for each individual stage on a composite map such as this, but a breakdown will be given in a later section. Third, the effect of viewing bias is very evident in the maps. The relatively small number of (d) Figure 1.1: Application of microseismicity cloud to characterize the fracture network.(a) Associated microseismic cloud with production data (Phillips et al., 1998), (b)Original, absolute-determined hypocenters(Left) and relative mapping technique (right) (Rut- ledge et al., 1998), (c) Fracture growth through four stimulation times in Barnett shale (Maxwell et al., 2002) (d) Examples of microseismic cloud maps in tight sandstone reservoirs (Warpinski et al., 2010). 9 53 K. Tezuka, H. Niitsuma / Engineering Geology 56 (2000) 47–62 one is the true failure plane. Multiplet analysis to be that of the fault. Once the fault plane is identified, the slip direction is also identified from often provides a way of doing this since members of cluster often align along a plane which can be the fault plane solution. Hayashi and Masoka (1995) proposed a interpreted as a fracture (Augliea et al., 1994). If indeed the events of the multiplet are located on method for estimating principle stress directions and their magnitude ratios by using slip data the same plane that has slipped in each of the events, as illustrated in Fig. 5, that plane must be recorded on fracture surfaces in core samples, assuming that the slip striations are created by consistent with one of the nodal planes of the fault plane solutions of the events. Fig. 6 shows a com- frictional slip between two fracture surfaces. The slip directions and the unit normal vectors of the parison of a fault plane solution (left figure) with the associated multiplet distribution plane (right fracture surfaces are used to estimate the stress field. Moriya and Niitsuma (1994) combined this figure) for one multiplet group of the Hijiori data. In the latter, the relative locations of each member method with a microseismic doublet analysis and used doublet distribution planes as fracture planes. of the multiplet group have been projected onto the lower hemisphere of a Wulff net. The arc We adapt this technique to our purpose and use the slip directions and the unit normal vectors of represents the most likely aligned plane as esti- mated by a principle component analysis. The the fault planes obtained from microseismic clus- ters instead of those from core samples. Fig. 7 nodal plane that dips to the northeast in Fig. 6(a) is almost identical to the multiplet distribution summarizes the flow chart of the analysis pro- cedure including the technique for microseismic plane of Fig. 6(b). Thus, this plane is considered Fig. 8. Comparison of fault plane solutions and microseismic doublet distribution planes for seven major doublet groups. Six of the seven groups (except for group E1) resulted in consistent solutions. This result indicates that the microseismic doublets tend to align on the plane (fracture) which has been activated by hydraulic stimulation. (a) (b) Identifying fault activation during hydraulic stimulation in the Barnett shale: source mechanisms, b values, and energy release analyses of microseismicity Scott Wessels*, Michael Kratz, Alejandro De La Pena, MicroSeismic Inc. Summary Identification of fault planes that intersect horizontal wellbores is critical to optimizing formation stimulation, preventing waste of valuable time and materials, and avoiding the establishment of fluid flow pathways into non- target formations, such as aquifers. We can detect and locate microseismic events accurately over a broad area using a large near surface seismic monitoring array. In addition, source mechanism inversion techniques can be used to determine the method of failure experienced by the rock formation, expanding our understanding of the dynamics involved in hydraulic fracturing. Events in this analysis are segregated into two populations based upon the distinct source mechanisms present. Spatial and temporal analysis of frequency magnitude distributions (FMD) allows us to characterize trends useful in assessing the hydraulic treatment efficiency. This information can assist in interpretation of faults in a 3D seismic volume to delineate faults in reservoirs, or when used alone, identify faults of subseismic displacement to further optimize future well placement. Also b values and source mechanisms can help to better define stimulated reservoir volumes (SRV) by indicating the effective level of stimulation. Introduction Microseismic monitoring of hydraulic fracture stimulation in low permeability reservoirs has grown to be a critical source of information for operators working to optimize well treatment plans. Information obtained from microseismic monitoring provides us with a better understanding of rock mechanics within the reservoir. One aspect of utmost importance is identifying pre-existing tectonic faults that have the capacity to capture one or more stages of valuable frac fluids and proppant without adding significantly to the productivity of the well, or possibly harming productivity by accessing a nearby aquifer. Source mechanism inversion of microseismic events can be used to identify different failure mechanisms induced by fluid pressure in the reservoir. By comparing the energy distributions of event populations segregated by source mechanism and integrating other information, such as spatial and temporal distribution of events, we begin to see clear differences between these populations that can be used to further identify what type of activity is taking place within the reservoir. By examining the FMD of event populations we can also determine if a particular type of mechanism is associated with fault activity or reactivation of natural fractures (De La Pena et al., 2011). In this study we examine the observed microseismicity associated with four horizontal wells completed in the Barnett Shale Formation in the Ft. Worth Basin, Midcontinent USA (Figure 1). The well horizontals are oriented perpendicular to maximum horizontal stress (Heidbach et al., 2009). Figure 1. Map of wellbores and microseismic events colored by mechanism and sized by magnitude. Events interpreted to be related to natural fracture stimulation are in red, while fault related events are shown in blue. There are a total of 12,094 events shown here, of which 86% are natural fracture events. Grid squares are 500 ft by 500 ft. Methodology The dataset used for this study was acquired using BuriedArray TM technology and microseismic events were located using Passive Seismic Emission Tomography (PSET ® ). Two different source mechanisms were identified by inversion of several individual events following the same method explained by Williams-Stroud et al. (2010). The failure mechanisms are used to differentiate between events generated by reactivation of natural fractures and events generated by stimulation of a pre-existing fault plane intersecting the wellbore (De La Pena et al., 2011), herein referred to as fracture and fault events, respectively. © 2011 SEG SEG San Antonio 2011 Annual Meeting 1463 1463 Downloaded 27 Sep 2011 to 128.125.153.126. Redistribution subject to SEG license or copyright; see Terms of Use at http://segdl.org/ (c) J. Charl´ ety et al. / Geothermics 35 (2006) 532–543 539 Fig. 3. Evolution of the P-wave seismic velocity at 4.6 km depth during the 2000 stimulation test. Images are in chronolog- ical order from Subsets 1 to 14 (labelled “set” in this figure). Yellow dots represent the 500 events used in the computation for each panel. The green line corresponds to the open-hole section of injection well GPK2. Grey areas (zones of poor resolution) were not considered in the analysis. Distances given in kilometers. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of the article.) (d) Figure 1.2: Novel microseismicity analaysis to characterize the fracture network, (a) Comparison of fault plane solutions and microseismic doublet distribution planes (Tezuka, 2000), (b) Most positive principal curvature range of microseismic occurrence (top), P-impedance range of microseismic occurrence (Refunjol et al., 2010), (c) Micro- seismic events interpreted as related to fracture stimulation (red, b-value 2) and fault related events (blue, b-value 1), Barnett shale (Wessels et al., 2011), (d) Evolution of the P-wave seismic velocity at one horizon during the stimulation test (Charlety et al., 2006). 10 (a) _________________ dependence on naturally occurring hydrothermal reservoirs involves human intervention to engineer hydrothermal reservoirs in hot rocks for commercial use. This alternative is known as “Enhanced Geothermal Systems,” or EGS. EGS reservoirs are made by drilling wells into hot rock and fracturing the rock sufficiently to enable a fluid (water) to flow between the wells. The fluid flows along permeable pathways, picking up in situ heat, and exits the reservoir via production wells. At the surface, the fluid passes through a power plant where electricity is generated. Upon leaving the power plant, the fluid is returned to the reservoir through injection wells to complete the circulation loop (Figure 2). If the plant uses a closed-loop binary cycle to generate electricity, none of the fluids vent to the atmosphere. The plant will have no greenhouse gas emissions 6 other than vapor from water that may be used for cooling. Figure 2. EGS Cutaway Diagram Energy Conversion Plant Injection Well Hot Rock Production Well Engineered Fracture System A complete geothermal system includes both surface and underground components, and the MIT study analyzed elements of both components. DOE has focused this technology evaluation on the underground component (i.e., the EGS reservoir), rather than the energy conversion surface component. The surface component represents a significant fraction of the overall cost of a commercial EGS and will be a major factor in ultimately determining economic viability. However, by far the greater knowledge gaps and technology uncertainties involve the reservoir. Enex Binary Plants, “Benefits of the Enex Binary Plant,” 6 Apr. 2008. http://www.enex.is/?PageID=191. An Evaluation of Enhanced Geothermal Systems Technology 3 6 (b) Figure 1.3: Unconventional reservoirs in which sufficient microseismic activity is present.(a) Shale and tight sand reservoirs, (b) Enhanced geothermal systems(EGS) SPE 59735 DIAGNOSTIC TECHNIQUES TO UNDERSTAND HYDRAULIC FRACTURING: WHAT? WHY? AND HOW? 3 these techniques is that they map the total extent of hydraulic fracture growth, but provide no information about the effective propped fracture length or conductivity. The resolution of these techniques decreases with increasing distance away from the fracture (see Table 1 “Main Limitations” for details). -Well Testing -Production Analysis -Net Pressure Matching -Pressure Decline Analysis -Microseismic Fracture Mapping - Downhole Tiltmeter Mapping Surface Tiltmeters -Radioactive Tracers -Production Logging -Borehole Image Logging Figure 1 – Fracture diagnostic techniques Surface and Downhole Tilt Fracture Mapping. 4-7 The principle of tiltmeter fracture mapping is quite simple (see Figure 2). A created hydraulic fracture results in a characteristic deformation pattern of the rock surrounding the fracture. By measuring the hydraulic fracture induced tilt (deformation) of the earth at several locations (surface and/or downhole) using extremely accurate “carpenter’s levels”, the fracture orientation (when using surface tiltmeters) and geometry (when using downhole tiltmeters) can be obtained. Figure 2 – Principle of tiltmeter fracturing mapping Surface tiltmeters are deployed in shallow holes (20 to 40 feet deep) at radial distances from as close as a few hundred feet to as far as one mile around the injection well depending on the depth of the treatment zone. The array of surface tiltmeters measures the gradient of the displacement and provides a “map” of the deformation of the earth’s surface above the fracture. Analysis of this “tilt” field provides a measurement of the fracture azimuth, dip, depth to fracture center and total fracture volume. Because surface tiltmeters are typically very far away from the created fracture, they can not precisely resolve fracture length and height. Downhole tiltmeter mapping is based on the same concept as surface tiltmeter mapping but instead of being at the surface, the tiltmeters are positioned via wireline in one or more offset wellbores at the depth of the hydraulic fracture. Downhole tiltmeters provide a map of the deformation of the earth “adjacent” to the hydraulic fracture. In most applications, downhole tiltmeters can be placed much closer to the fracture than surface tiltmeters and are therefore significantly more sensitive to fracture dimensions. The measured tilt is used to determine fracture height, length, and width versus time. Microseismic Fracture Mapping. 8,9 Microseismic fracture mapping provides an image of the fracture by detecting microseisms or micro-earthquakes that are triggered by shear slippage on bedding planes or natural fractures adjacent to the hydraulic fracture ( see Figure 3). The location of these microseismic events is obtained using a downhole receiver array of accelerometers or geophones that are positioned at the depth of the hydraulic fracture in one or more offset wellbores. TIP REGION LEAKOFF REGION X Y RECEIVER DETECTS GROUND MOTION FROM MICROSEISM ELASTIC WAVES EMITTED ADDED SHEAR NATURAL FRACTURE INCREASED PORE PRESSURE Figure 3 – Principle of Microseismic Fracture Mapping Both wireline-conveyed and cemented-in receiver arrays have been used in field applications. The data is gathered and processed with a surface data acquisition system, and the microseismic events are located using techniques based on P (compressional) and S (shear) – wave arrivals to provide time- dependent images of fracture growth and geometry. Figure 1.4: Principle of microseismic fracture mapping (Cipolla and Wright, 2000) 11 Mercuryville Fault Zone APPR NORT LIMIT STEAM OXIMATE HERN Cobb Mountain Fault Big Sulphur Creek H i g h V a l l e y Ot t obon i R i dge Collayomi Fault Zone Aidlin Bottlerock 2000 1000 FEET 3000 0.5 1.0 MILE Active power plant Inactive or retired power plant Injection Well Communities Strong Motion Station Steam Chloride >= 1.0 ppmw Current Production Area Study Area PRODUCED BY AN AUTODESK EDUCATIONAL PRODUCT PRODUCED BY AN AUTODESK EDUCATIONAL PRODUCT PRODUCED BY AN AUTODESK EDUCATIONAL PRODUCT PRODUCED BY AN AUTODESK EDUCATIONAL PRODUCT Figure 1.5: Map of The Geysers geothermal field (GGF), the area under study (in the rectangular area), and the locations of the injection wells and seismic activities (after Beall,J., Calpine). 12 Figure 1.6: Integrated workflow architecture for characterizing a fracture network in unconventional reservoirs using microseismic data. 13 Workflow For Characterizing Fracture Network Using Microseismic Data Inverse Modeling Microseismic Conventional ANN Joint inversion Fracture model seismic Well logs Integrati Integrati on using on using novel novel Reservoir properties Well logs Geology schemes schemes Geo-statistics Improved characterization Production Hybrid ANN –FL - GA Improved reservoir reservoir management Figure 1.7: The schematic view for the integration of microseismic data with other sources of data. 14 Chapter 2 Primary Steps of the Workflow This chapter presents a brief overview of the primary steps of our workflow. Some of which are fundamental for processing any microseismic data; the rest are widely accepted methods according to many authors both seismologists and engineers. 2.1 Recording the Microseismic Events Microseismic events are micro-earthquakes mainly induced by changes in stress or pore pressure associated with hydraulic fracturing or fluid injection interaction. They are caused by slippage or tensile deformation along pre-existing planes of flaws (e.g., nat- ural fractures, joints, etc.) and emit seismic energy that can be detected using a down- hole or surface array of three-component receivers (Le Calvez et al., 2007; Cipolla and Wright, 2000). Downhole monitoring (Figure 2.1a) is more sensitive than surface mon- itoring (Figure 2.1b) near the observation well; the former can detect 4 to 5 times more events, but may lose this advantage at a farther distance (Diller and Gardner, 2011). In addition, to acquiring better tomographic inversion and shear wave splitting results, sur- face monitoring appears to be essential, as we will discuss later in this chapter. To detect more reliable and accurate events, we propose installing permanent surface arrays near treatment wells in addition to having downhole monitoring arrays in an offset well when stimulation is in progress. 15 (a) Comparison of simultaneous downhole and surface microseismic monitoring in the Williston Basin David E. Diller*, Nanoseis, and Stephen P. Gardner, Whiting Oil and Gas Summary Microseismic events reported from simultaneous downhole and surface monitoring of a hydraulic fracture well stimulation were matched on an event-by-event basis and compared. Downhole monitoring was much more sensitive than surface monitoring near the observation well, detecting 4-5 times the number of events, but the downhole monitoring appeared to be lose much of it sensitivity advantage at distances greater than about 3,000 feet. The picks reported from surface monitoring varied dramatically depending on the picking criteria that were used, which emphasizes the need for a reliable pick confidence factor. In the strict assessment of the surface data the existence of the majority of the events were corroborated by the downhole data, including many sub-visible surface events. Considerable differences exist in the reported spatial location of the events picked by the two methods despite correct positioning of perforation shots for each method. The events picked by the surface method are clustered much more closely to the wellbore, which may represent a considerably lower estimate of the stimulated rock volume. Introduction From Dec. 9-12, 2010 a 30-stage hydraulic fracture well stimulation in the Bakken-Three Forks Petroleum System of the Williston Basin (~10,500’ depth) was simultaneously monitored using a downhole array of 45 three-component geophones located in an observation well approximately 530 feet offset, and 1,246 single-component geophones located in a star shaped array at the surface over the area of interest (Fig. 1). The purpose of conducting simultaneous monitoring was to compare the effectiveness of the two methods in detecting and locating microseismic events. Method The downhole and surface data were processed independently using commercial microseismic processing services. Events in the downhole data were located using a combination of traveltime-based moveout analysis of the compressional waves, and multi-component rotation analysis (Fuller et al., 2007). Events in the surface data were located using the passive seismic emission tomography method (Duncan and Eisner, 2010). Initial comparison of the reported hypocenters from the two methods showed poor agreement. It was determined that the threshold(s) that discriminates between valid and invalid picks was set too low for the surface data, and the Figure 1: Location of the surface receivers (white points) relative to the borehole of the treatment well (red points). Figure 2: Map view of the hypocenters reported from the downhole data. The symbol sizes correspond to event amplitudes, and the colors correspond to stages. From Engelbrecht and Fuller, 2011. © 2011 SEG SEG San Antonio 2011 Annual Meeting 1504 1504 Downloaded 27 Sep 2011 to 128.125.153.126. Redistribution subject to SEG license or copyright; see Terms of Use at http://segdl.org/ (b) Figure 2.1: (a) Downhole monitoring for microseismic data (ASC, 2011)(b) Location of the surface receivers (white) and treatment well location (red) (Diller and Gardner, 2011). 2.2 First Arrival Picking for Shear and Compressional Waves The number and reliability of events reported through microseismic monitoring from either surface or downhole monitoring varies dramatically, depending on the first arrival picking criteria being used. The detection of phase arrivals is vitally important to obtain- ing accurate time differentials to be used for tomographic inversion schemes so as to obtain estimates of velocity and microseismic hypocentral locations. Improved accuracy in the arrival times should lead to better inversion results. These form an indispensable link in microseismic data analysis and are an important element in determining the effi- cacy of the eventual property prediction workflow presented in this thesis. Traditionally, human analysts have carried out this task in a visual way. But visual analysis is a very time-consuming and subjective task that cannot manage the huge volume of digital and 16 real-time microseismic data recorded today by different arrays of receivers. The liter- ature discusses various techniques for automatic detection and for picking first arrivals of different seismic waves. Over the last two decades, numerous algorithms have been developed for identifying arrival times based on energy analysis, polarization analysis, artificial neural networks, maximum likelihood methods, fuzzy logic theory, autoregres- sive techniques, higher order statistics, or the wavelet transform (Leonard, 2000; Dai and MacBeth, 1995, 1997; Zhao and Takano, 1999; Al-Ghamdi, 2007; Sabbione, 2010; McCormack et al., 1993; Veezhinathan et al., 1993; Aminzadeh et al., 1994). With large arrays and a significant quantity of data being collected, traditional autopickers face issues with run times and low SNR (signal to noise ratio) environments. One major challenge is to working in an environments in which external contingencies pose serious restrictions on effectively applying various inversion schemes. Some such scenarios include a limited number of recording stations (coverage issues) or noise- related (low SNR) issues. It is important to have a minimum number of ”acceptable” arrival detections for both compressional (P-wave) and shear (S-wave) data for a reli- able velocity inversion and satisfactory levels of accuracy in the final microseismic event locations. This necessitates the development of an advanced autopicker to improve accu- racy, speed, and range of applicability over any dataset. For instance, Figure 2.2 shows the overall workflow for extracting phase arrivals using an artificial neural network we have developed at University of Southern California and a sample pick obtained through this method. 17 2.3 Hypocentral Location and Tomographic Inversion The ’simulps14’ code 1 is available for 3D velocity modeling using phase arrivals. The original work was conducted by Thurber (1983) and has been subsequently improved (Um and Thurber, 1987; Eberhart-Phillips, 1986, 1990; Eberhart-Phillips and Michael, 1998). The algorithm uses an approximate ray tracing and pseudo ray bending algo- rithm to invert for velocity values (V P and V P =V S ). This computer code utilizes an iterative, damped, least-squares method to invert arrival times, simultaneously estimat- ing earthquake locations and the 3DV P andV P =V S fields. Hypoinverse (Klein, 2002) is a freely available software available from USGS 2 and allows for earthquake location and magnitude estimation. The hypoinverse algorithm provides necessary information to generate event pairs, which are used as inputs for the hypoDD algorithm in order to relocate the hypocenters using a double difference algorithm (Waldhauser, 2001). The method correlates hypocenters based on P- and S-wave arrival times, and/or waveforms. Then, relocates them with respect to other related earthquakes. This process eliminates the uncertainties associated with the velocity model and picking errors. We used a workflow that involves generating ’phase files’ including a minimum of 4 P-phase picks and 2 S-phase picks from any event which can be used as input to SimulPS as well as Hypoinverse/ hypoDD algorithms to generate velocity models as well as microseismic event locations. Figure 2.4 shows the 3D distribution of hypocen- tral location of microseismic events from 2006 to 2009 at The Geysers. Figure 2.5 shows the SimulPS output structure for a compressional velocity model that has 4950 points and spacing of about 600m covering the study area 3 . In the following chapters, we will 1 http://faldersons.net/Software/Simulps/Simulps.html 2 http://earthquake.usgs.gov/research/software/#hypoDD 3 It is shown in Figure 1.5 18 discuss how these fundamental data can help us characterize the fracture network in unconventional reservoirs. 2.4 Multiplet Analysis Multiplet analysis has been successfully implemented for microseismic monitoring projects at unconventional reservoirs worldwide; the utility of waveform correlation and clustering of multiplets has been investigated by many researchers (Rutledge and Phillips, 2003; Rowe et al., 2002; Moriya et al., 2000). Microseismic events that occur on the same fracture plane or along adjacent, similarly-oriented ones produce similar waveforms at a given receiver with the same focal mechanism. Grouping such similar waveforms with the objective of illuminating an active fracture structure is called a relative location technique. We could identify the relative arrival times by picking identified peaks or troughs repeatedly. Then, frac- ture can be imaged accurately by locating these microseismic events relative to a master event location (Rutledge et al., 1998; Tezuka, 2000). Two such similar events are called doublets, and a group of more than two is called a multiplet (Arrowsmith and Eisner, 2006). The slip and fracture direction can be identified from a cluster of multiplets often aligned along a fracture/fault planes (Tezuka, 2000). Multiplet clustering within a large number of microseismic events (Figure 1.1 and 2.4) allows us to identify events located on geometrically and/or geophysically related structures (especially fractures), because the focal mechanism and the travel path of waveforms are nearly identical. The hypocen- tral locations of a multiplet cluster usually illuminate the planar structure defining a frac- ture or a fault where the events occur (Kumano et al., 2007; Moriya et al., 1994). They are also useful for estimating fracture density (Lees, 1998). 19 We may apply different approaches to identifying the multiplet in automated pro- cesses, such as the cross correlation coefficient. Figure 2.6 shows a decision tree for the real-time application of the multiplet-identification algorithm in an analysis of micro- seismic events (Arrowsmith and Eisner, 2006). Figure 2.7a shows the ellipsoids of the multiplet clusters. Red lines represent two dominated fracture systems interpreted by the principle direction of the connected clusters (Tezuka et al., 2008). 2.5 Shear Wave Splitting One effective way of detecting the fracture orientation and density of the subsurface is through shear wave splitting of microseismic events. When shear wave hits the fracture, it splits into two components which have fast and slow arrivals on the seismogram. The polarization angle() of a fast component that is parallel to the fracture indicates the fracture orientation and the time delays(t) 4 observed between the slow and fast shear waves give information about the density of fracture (Figure 2.8). Normalized time difference, t divided by total travel time (or to the length of the ray path), is proportional to fracture density along the seismic ray path (Vlahovic et al., 2003). Shear wave splitting has been used successfully to identify the orientation of primary and secondary stress-aligned fractures (Crampin, 1981; Lou et al., 1997; Tang et al., 2008; Crampin, 2005). Figure 2.9 shows the representative waveforms and observed splitting for the events inside the shear-wave window. We can use rose diagrams (polar histograms) to determine the dominant direction of fast shear-wave motion. Then, they can be plotted at each receiver of the array. Figure 2.10 shows fracture orientation result for two different fields. Equal-area projection 4 Typically a few tens of milliseconds 20 plots of normalized time delays can be used to display the crack density. Figure 2.11 demonstrates such a plot at The Geysers. 21 (a) (b) Figure 2.2: (a) ANN Autopicking workflow, (b) sample pick with ANN outputs and actual traces compared. 22 Figure 2.3: Schematic view of passive seismic tomography (Duncan, 2005). This can be modified horizontally for Downhole monitoring. Figure 2.4: Microseismic event cloud in 3D at The Geysers 4950 Points (15*33)*10 Spacing ~ 600m Figure 2.5: The initial velocity model provided by LBNL 23 V32 Arrowsmith and Eisner Figure 1 is a proposed scheme for the real-time location of microseismic events. Crosscorrelating every event with every other event in a data set would scale with the square of the number of events and would quickly cause the algorithm to become very expensive (J. Rickett, 2005, personal communi- cation). The scheme in Figure 1 avoids this problem. When a microseismic event is detected, the waveform is compared with a library of master events to determine if the event is a doublet of a master trace. If the event is not the doublet of a master trace, it is located using a conventional location algo- rithm and added to the library of master events. If the event is the doublet of a master trace, it is located relative to a master event for that group. Next, if the event has a higher S/N ratio than the master event, it replaces the old master. For tectonic earthquakes, multiplets are especially common in the creeping zones of faults (e.g., Nadeau et al., 1995; Wald- hauser and Ellsworth, 2000; Stich et al., 2001). Multiplets also are observed commonly during hydraulic fracture monitor- ing in hot, dry rocks (e.g., Lees, 1998; Li et al., 1998; Fr´ echet et al., 1989; Moriya et al., 2003) or oil and gas reservoirs (e.g., Moriya et al., 1994; Rutledge and Philips, 2003). Multiplet analyses from large industrial data sets require the identifica- tion of multiplets in a data set. The practical real-time applica- tion of multiplet identification in large data sets, such as those recorded in hydraulic fracturing, requires a fully automatic, fast technique. Previous techniques for the automatic detec- tion of multiplets (Maurer and Deichmann, 1995; Cattaneo et al., 1999; Stich et al., 2001; Schaff and Richards, 2004) fo- cus on earthquake seismology and use receivers distributed on the free surface of the earth. These studies use data recorded at networks of seismic receivers. Only the vertical compo- nents of the recorded wavefields are used on account of gen- erally worse S/N ratios on horizontal components. In every Figure 1. A decision tree for the real-time application of the multiplet-identification algorithm in analyses of microseismic events. case, they use preprocessed data with prepicked P- and S-wave phases. We present a novel identification technique for identify- ing multiplets which is suitable for borehole data sets such as those recorded during hydraulic fracturing or passive seismic monitoring, and it uses all three components. Using all three components helps to compensate for the lack of receiver dis- tribution (compared to surface seismic data sets) and is a natu- ral generalization to borehole data where there is no preferred component. The technique does not require previously picked P- or S-wave arrivals. We then apply the technique to a passive seismic data set recorded in the Valhall field, North Sea. TECHNIQUE FOR IDENTIFYING DOUBLETS We define a doublet as two events that have highly corre- lated waveforms. Thus, the time and polarization of the P- and S-waves are very similar, and we use the crosscorrelation coef- ficient to help identify doublets. The normalized crosscorrela- tion coefficient is evaluated in the frequency domain by anal- ogy with the convolution theorem (Ifeachor and Jervis, 1993). For two traces, x 1 (t)and x 2 (t), the normalized crosscorrelation function may be evaluated as C x (τ i ) = F −1 D (X ∗ 1 (f )X 2 (f )) x 2 1 (t) x 2 2 (t) , (1) where F −1 D denotes the inverse discrete Fourier transform, X ∗ 1 (f ) is the complex conjugate of the Fourier transform of x 1 (t), and X 2 (f ) is the Fourier transform of x 2 (t). The information provided by three-component receivers is exploited in such a way that the effect of noise on the calcula- tions is down weighted. This is particularly important when we apply this algorithm in real-time processing to randomly ori- ented borehole receivers. A simple crosscorrelation of com- ponents with low S/N ratios may distort the averaged coeffi- cient. Assuming the maximum amplitude of each component reflects the S/N ratio, we average the normalized crosscor- relation functions from all components with a weight of the maximum amplitude on each component. Subsequently, for receiveri, the signal weighted average of the crosscorrelation functions from the different components is given by C Ri (τ i ) = A x C x (τ i ) + A y C y (τ i ) + A z C z (τ i ) A x + A y + A z , (2) where C x ,C y ,C z are the normalized crosscorrelation func- tions for each component; A x ,A y ,A z are the maximum ampli- tudes for each component; and τ i is the lag time of the cross- correlation for theith receiver. Here, all times are defined as when the events are actually recorded. Amplitude weighting significantly reduces the effect of noise in our computations by down-weighting components where the S/N ratio is poor. When multiple receivers exist, the peak crosscorrelation co- efficient for two events in a doublet occurs at the same time lag at each receiver within a time window of length t. The time window t is less than one-fourth of the dominant period in the signal, because we define a doublet as two similar events with a maximum event separation of one-fourth of the dom- inant wavelength. We find the peak crosscorrelation coeffi- cient C for all receivers by averaging the peak crosscorrelation functions for each separate receiver for τ i within one-fourth Downloaded 15 Feb 2011 to 128.125.153.49. Redistribution subject to SEG license or copyright; see Terms of Use at http://segdl.org/ Figure 2.6: A decision tree for the real-time application of the multiplet-identification algorithm in the analysis of microseismic events. (Arrowsmith and Eisner, 2006) IPTC 12391 5 (a) (b) Figure 7: (a) The PCA ellipsoids and (b) the microseismic locations coloured with respect to the multiplet groups. Red lines represent two dominated fracture system interpreted by the principle direction of the clusters and by connecting the clusters. Dip Strike Slip Fracture plane Normal ｆ Figure 8: Definition of the dip, strike, and slip angles for the grid search method. Figure 9: Lower hemisphere pole projection of the fault planes for four groups. The intensity of the color indicates the number of the solutions. N-S strike: red (128 events), ENE-WSW strike: blue (95 events), NE-SW strike: green (40 events), NW-SE strike: cyan (26 events). (a) (b) Figure 10: (a) Microseismic location coloured with respect to the group of the strike direction. (b) Fracture system estimated by microseismic strike direction. Although the solutions for 35 multiplet representatives vary widely, they can be classified into four groups if we put focus on the strike direction as shown in Fig. 9. Those are N- S, ENE-SWS, NE-SW, and NW-SE directions where N-S and ENE-SWS are two dominant groups. Fig. 10(a) shows the microseismic locations coloured with respect to the group of the strike direction. Considering the sort of colouring and their strike direction, syetmatic linear structures can be interpreted. Fig. 10(b) shows the linearment system estimated by the mechanism analysis. Dominant Orientation of Fracture System According to the results of the multiplet clusters analysis and the mechanism analysis an existing of two dominant orientation of the fracture system are highlighted by the two independent analyses. The linear structures are commonly inferred from the two analyses in N-S trending and also in NE- SW to ENE-SWS trending despite that a slight discrepancy can be seen between the two analyses. The fracture orientations derived by the multiplet analysis include some degree of error, because the PCA results are quite sensitive to the location errors and distributions of the multiplet groups are not always in plane. If the orientation estimated by the mutiplet analysis include 20 degrees of uncertainty, the orientation of 23 out of 35 multiplet clusters are within the range of orientation constrained by the mechanism analysis. The state of stress around the injection well is estimated as the strike-slip stress regime whose maximum horizontal stress axis is in NNE-SSW direction. 7 In this condition, shear slippages are most likely induced along N-S and NE-SW directions. These directions are consistent with the dominant orientation of fracture system estimated by the multiplet analysis and the mechanism analysis. Figure 2.7: (a) Plane of fractures related to multiplet analysis, (b) the microseismic loca- tions colored with respect to the multiplet groups. Red lines represent two dominated fracture system interpreted by the principle direction of the clusters and by connecting the clusters.(Tezuka et al., 2008) 24 Figure 2.8: Shear wave splitting in anistropic medida (Garnero, 2011) Fig. 1. VOLGEO 2524 4-11-02 G. Vlahovic et al./Journal of Volcanology and Geothermal Research 120 (2002) 123^140 125 Figure 2.9: The representative waveforms and observed splitting for the events inside the shear-wave window (Vlahovic et al., 2003) 25 etal.,1988).Thedifferentialtimedelaybetweenthearrivalofthefast and the slow shear-waves (typically a few tens of milliseconds) is proportional to crack density, or number of cracks per unit volume within the rock body traversed by the seismic wave (Hudson, 1981; Crampin,1987; Crampin and Lovell,1991). Measuring the fast shear- wave polarization and time delay from local microearthquakes has thus become a valuable technique to detect the orientation and intensity of fracturing in the subsurface of fracture-controlled geothermal field (e.g. Lou and Rial, 1997; Vlahovic et al., 2002a,b; Elkibbi and Rial, 2003, 2005; Elkibbi et al., 2004, 2005; Yang et al., 2003; Rial et al., 2005; Tang et al., 2005). In cracked geothermal reservoirs such as Krafla the anisotropy is likely to have been caused by aligned systems of open, fluid-filled microfractures. Fortunately, the anisotropy effects on seismic waves induced by small, aligned open cracks in an otherwise isotropic rock are indistinguishable from those produced by an unfractured, but transversely isotropic medium. Seismic anisotropy characterizes the Neovolcanic zones of Iceland where shear-wave splitting of 0.1–0.3 s have been observed (Menke et al., 1994). Shear-wave splitting was clearly recorded at most of the Krafla stations. In fact, we have re- cordedunusually welldevelopedsplitting,inwhichthefastandslow shear-waves are naturally separated in time, that strongly points to theprevalenceofatleasttwofracturesystemsorientedapproximately inN–SandE–W.Fig.2(a)showsevidenceforastrong,nearlyE–Wfast shear-wave polarization that suggests the presence of E–W oriented, probably vertical cracks in the neighborhood of the station. Fig. 2(b) shows evidence of a N8°E fast shear-wave polarization, close to the overallN15°EstrikeofthenormalfaultsoftheKraflariftzone.Indeed, SWShas detected notonly the predominant N–S fabric related to the riftzone,butalsoprovidesstrongevidenceforanequallypervasiveE– Woriented lineament of subsurface fractures. 2.2. Measuring polarization and time delay The polarization direction of the fast split shear-wave is usually paralleltothestrikeofthepredominantcracks,regardlessofitsinitial polarizationatthesource,andthetimedelaybetweenthefastandthe slow waves is proportional to the crack density, assuming constant crustalvelocities.Thesesplitshear-waveparameters(fastshear-wave polarization direction φ and differential time delay δt) constitute a valuabledatasettoinvertforthesubsurfacefracturegeometryandto estimate the crack density and permeability within fractured geothermal reservoirs. An important limitation to shear-wave split- ting analysis is that seismic rays must be within the shear-wave window of the seismic stations. This window can be visualized as a rightcircularconewithvertexatthestationandvertexangleic=sin −1 (β/α), where α and β are the P- and S-wave surface velocities, res- pectively.Foranglesofincidencegreaterthanic,shear-wavesinteract strongly with the free surface, distorting the incoming waveform (Crampin, 1981; Booth and Crampin, 1985). For a half space with a typicalPoisson'sratioof0.25,thewindow'svertexangle,asmeasured from thevertical, is equal to35.2°. All earthquakesused for thestudy in this paper are restricted within this window. Forthepurposeofthisstudy,weusethoseφandδtmeasurements from the Krafla array that correspond to high signal-to-noise ratio seismograms displaying linear horizontal particle motion and a clear well-defined shear-wave splitting event. Polarization diagrams (also known as particle motion plots) are used to accurately detect the switch in polarity of the two orthogonally polarized fast and slow shear-wavesandtomeasurethesplitparametersφandδt.Fastshear- wave polarization angle φ is measured by interactive rotation of the seismogram until the horizontal particle motion plot shows that fast and slow shear-waves are oriented along the instrument's horizontal Fig.3.Equal-arearosediagrams(polarhistograms)of thefastshear-wavepolarizationdirectionsobservedattenseismicstations.Thegreencurverepresentstheroadandtheblue squaresindicatethelocationoftheKraflapowerplant.NVitisacraternearby.Seedetailsinthetext.(Forinterpretationofthereferencestocolorinthisfigurelegend,thereaderis referred to the web version of this article.) 319 C. Tang et al. / Journal of Volcanology and Geothermal Research 176 (2008) 315–324 (a) Shear-wave splitting in The Geysers 1027 Table 1. The NW Geysers station orientation corrections based on P-wave analysis. Predicted and observed P-wave polarizations from an average of 50 events per station were compared. The value of each correction (in degrees) indicates the amount of rotation of the geophone horizontal compo- nents. With the exception of stations S9 and S14, standard deviation values are mostly within 10 degrees. Station Name Station Correction Standard deviation (NW Geysers) (Degrees) S1 56 8.2 S2 −25 7.8 S3 179 11 S4 −165 7.4 S5 −168 10.1 S6 −65 10.4 S7 −148 13 S8 15 6.3 S9 130 14.7 S10 −105 8.8 S11 −132 4.2 S12 170 9.8 S13 133 9.8 S14 107 22.8 S15 41 11.8 S16 4 7.8 coincide with true geographic North and East. Hence, geophone ori- entation corrections had to be performed on all 16 stations in the NW Geysers by comparing predicted and observed P-wave polarizations from an average of 50 microearthquakes surrounding each seismic station. Table 1 shows station orientation correction estimates and the corresponding standard deviations. In the SE Geysers, stations were on the ground surface, so no correction was necessary. In the last few years, we have analyzed shear-wave splitting events from several thousand seismograms in The Geysers. As a result, we accumulated a set of 1757 high-quality φ and δt pairs. Our experi- ence with the data suggests that automatic picking of polarization directions and time delays, in particular, without human interven- tion is unreliable. This is due to the great variability and diversity of wave patterns, which often result from interaction of shear-waves with complexly cracked rock bodies. The trained human eye is in- dispensable to carefully investigate seismic records, especially those in which the arrival of the slow shear-wave is less clearly defined. 4SHEAR-WAVE SPLITTING RESULTS 4.1 The NW Geysers 4.1.1 Fast shear-wave polarization directions 4.1.1.1 Rose diagrams Fig. 3 shows rose diagrams (polar his- tograms) of fast shear-wave polarization directions observed within the shear-wave window of each station in the NW Geysers. The bin size in the rose diagrams is 10 ◦ and the length of each bin is pro- portional to the number of polarizations within it. Typically, stations with more than 35 polarization observations show one predominant polarization orientation. Stations exhibiting generally uniform po- larization directions are stations S1, S3, S4, S5, S7, S8, S9, S11, S15 and S16. Table 2 lists mean polarization orientation and stan- dard deviation for stations showing a main polarization orientation, excluding those with less than 35 observations to avoid possible scat- 38 o 48' 38 o 50' 122 o 48' 122 o 50' 1 km Squaw Creek Fault Zone Caldwell Pines Fault CCSF S1(89) S2(64) S3(72) S7(34) S8(66) S9(37) S10(35) S12(35) S4(92) S5(57) S6(120) S11(72) N S13(11) S14(65) S16(33) S15(26) NW Geysers Figure 3. Rose diagrams (polar histograms) showing fast shear-wave po- larization directions as recorded by each station in the NW Geysers. Most stations display a predominant φ direction ranging between N and N60E. The station names along with the number of events (between parentheses) used to generate the rose diagrams are indicated. The approximate extent of the Coldwater Creek Steam Field (CCSF) is delimited by the shaded area. Table 2. Mean polarization directions (in degrees, measured clockwise from North) and standard deviations for stations in the NW Geysers showing generally uniform polarization orientations. Station Mean polarization direction Standard deviation S1 −14 18 S3 59 11 S4 47.5 16 S5 9 11 S8 39 12 S9 −18 16 S11 15 10 ter in polarization directions induced by a relatively limited number of data. We restrict the analysis of mean polarization direction to sta- tions showing only one prominent polarization orientation because, as we will discuss later, sets of secondary polarizations oriented at an angle to the main polarization direction often provide impor- tant clues on fracture geometry (e.g. fracture dip amount and direc- tion, presence of intersecting fractures, etc.) and should be analyzed separately. The predominant observations of φ vary mainly between N and N60E for all stations, with the exception of station S9 which displays mainly NNW polarizations (note however that station S9 has a rela- tively high standard deviation for orientation correction (Table 1)). Stations S2 and S6 show, in addition to the main polarization orien- tation, a distinct subset of NW-striking polarizations nearly perpen- dicular to the main set. These secondary polarizations strike parallel to nearby NW-trending faults. As numerous readings were made for both stations S2 and S6 (64 and 120 readings respectively; Fig. 3), the secondary sets of polarization orientations are likely real clues to local fracturing patterns and not artifacts generated by a limited dataset. 4.1.1.2 Equal-area projection plots To study the azimuthal distri- bution of observed fast shear-wave polarizations, φ measurements C 2005 RAS, GJI, 162, 1024–1035 (b) Figure 2.10: Rose diagrams of the fast shear-wave polarization directions observed, (a) at Icleland stations (Tang et al., 2008), (b)at The NW Geysers (Elkibbi, 2005) Shear-wave splitting in The Geysers 1029 S8 S7 S3 S1 S2 S14 S16 S10 S11 S13 S9 S6 S5 S4 S12 S15 10 ms km −1 40 ms km −1 20 ms km −1 SCALE S8 S7 N N N N N N N N N N N N N N N Figure 5. Equal-area projection plots of δt measurements for 16 stations in the NW Geysers (homogeneous velocity model). Normalized time delays generally vary from 8 to 40 ms km −1 .For clarity purposes, equal-area plots are located in the approximate relative locations of the stations. The plot radius is unity and corresponds to 45 degrees. window as a function of ray azimuth and incident angle (e.g. Crampin 1993). Highest standard deviations, however, are probably related to measurement errors and geological inhomogeneities along the raypath. Fig. 5 shows equal-area projection plots of normalized time delays in the NW Geysers. Patterns in observed time delays as a function of ray azimuth and incident angle are rather subtle and not easily quantifiable. A higher sampling rate may be required to detect systematic changes in time delays within the shear-wave window. Spatial variations in observed time delays show some correlation with the location of seismic stations. For instance, stations S4, S5, S6, and S11 are located along the Squaw Creek Fault Zone and show the highest values of mean time delays among all stations in the NW Geysers. These four stations cover the southwestern edge of the steam field and show mean normalized δt values between 12.8 ms km −1 and 16.47 ms km −1 . Stations S1, S2, and S8 show medium-to-high average time delays and are located in the east- central part of the steam field (mean normalized δt values range between 11.13 ms km −1 and 12.13 ms km −1 ). In general, spatial variations in time delays, as recorded by different seismic stations, may be related to variations in the geometry of crack systems, or variations in crack density, crack aspect-ratio (defined as the ratio of crack thickness to crack diameter), degree of fluid saturation, fluid type, and pore-fluid pressure. 4.2 The SE Geysers 4.2.1 Fast shear-wave polarization directions 4.2.1.1 Rose diagrams Although seismic stations in the SE Geysers are on the ground surface, and thus likely to be noisier than downhole ones, numerous high-quality seismograms were recorded with high signal-to-noise ratios as well as robust and impulsive shear-wave arrivals. Compared to the NW Geysers, observed fast shear-wave polarizations in the SE Geysers show more variation in orientation (Fig. 6). The major polarization directions are generally N5E, N35E-to-N60E, N75E-to-N85E, and N20W-to-N55W . Signif- icant variations in observed polarization angles can be traced even in neighboring stations, such as stations S10 and S11, which are less than 1 km apart and show orthogonally polarized fast shear-waves. This may indicate that orientations of main fracture systems beneath stations S10 and S11 are orthogonal, if we assume the simple case of vertical fractures. Stations with more than 38 SWS observations in the SE Geysers typically show one prevalent polarization orientation, with the ex- ception of stations S6, S8, and S14. Table 4 lists mean polarization directions for these stations with one predominant polarization ori- entation. Standard deviations are comparable to those calculated with the NW Geysers dataset. As in the NW Geysers, some stations C 2005 RAS, GJI, 162, 1024–1035 Figure 2.11: Equal-area projection plots oft measurements for the stations in the NW Geysers (Elkibbi, 2005). 26 Chapter 3 Fuzzy Clustering While observing the induced microseismicity cloud around geothermal injection wells, such as GGF (Figure 2.4), and in the hydraulic fracturing process in tight hydrocar- bon reservoirs (Figure 1.1), we have found that the identification of microseismic event clusters and their relation to the injection process or stimulation stages is a difficult task. Figure 2.4 shows microseismic event clouds as observed at The Geysers. This figure clearly indicates a high degree of overlap between different identifiable microseismic event clusters. Therefore, to find a movement of microseismicity cloud or the possible fracture network propagation with time, we look at the potential use of a fuzzy logic based clustering tool.(Aminzadeh and de Groot, 2006; Aminzadeh et al., 2010). 3.1 Fuzzy Clustering Algorithm The fuzzy clustering algorithm involves dividing the events into zones based on the loca- tion of injection wells in geothermal reservoirs or on stimulation stages in the hydraulic fracturing of tight hydrocarbon reservoirs where sufficient aggregations of seismicity exist. A radial distance is selected for clustering in order to understand how the cluster centers move with time. The window time for a hydraulic fracturing process is normally a matter of hours 1 . Nonetheless, the time periods for geothermal stimulation, may vary from daily and monthly periods to biannual or annual periods based on the observed seismicity in the area. 1 Depends on time required for each stimulation stage 27 Figure 3.1 shows the classification of zones for evaluation as well as the associated overlap of events falling within different zones. Given the extent of overlap between different regions, a fuzzy logic approach could prove advantageous. Figure 3.2 shows the workflow to find the fuzzy cluster centers. In this workflow, Eq. 3.1 defines each membership valueu ij and Eq. 3.2 defines the location of the centers whereas is the fuzzifier coefficient and can be defined based on the level of overlap of microseismic events. Interpreting the movement of these cluster centers can help us locate the front of the propagating fracture network and allows for a more accurate interpretation of fracture patterns and a broader understanding of the underlying physics. u ij = 1 P k=fuzzy cluster dis(center(i);event(j)) dis(center(i);event(j)) 2=(1) (3.1) V i (x) = P j u ij x P j u ij (3.2) 3.2 Result and Discussion In our dataset, we are only interested in the high temperature zone (HTZ) for developing EGS reservoirs. Thus,-cut of 3.2 km in depth is used for limiting the top of the desired locations and-cut of 6 km in depth is used for limiting the bottom of the desired area. This -cut representation is extracted from the geologic map of The Geysers. This subset of microseismic events is used for further analysis and processing. To implement the fuzzy clustering technique, the set of microseismic events is divided into five different zones. These zones are demonstrated in Figure 3.3 along with the production well locations. They are chosen based on both the locations of the production wells and the locations showing aggregations of seismic events. They 28 Figure 3.1: Selected zones for fuzzy cluster center evaluation should be defined based on the location of each stimulation stage or perforation areas in hydraulic fracturing monitoring. In each zone, three clusters were considered to prop- erly define the three dimensional direction of the fracture network movement. Annual time windows are used to analyze microseismic events. Figure 3.4 depicts cluster cen- ters for 2006–2009 at the HTZ zone. Figure 3.5 shows these cluster centers in each zone separately. These cluster center locations and their movement through time may help determining a new methodology for locating the appropriate locations of fracture network fronts. Considering the tail of microseismic events within the high temperature zone (below 3 km depth) in the Northwest Geysers, and after clustering the microseismic events, cluster centers show significant movement within the high temperature zone from 2006 to 2009 (indicating possible evolution of fracture network). The arrows in Figure 3.6 indicate the direction of movement of these clusters over time. We believe that these 29 Figure 3.2: Workflow to find the fuzzy cluster centers of microseismic Data directions may also represent the fracture network propagation direction (in the case of small movements.) Large movements may indicate fault movements or may have other reasons. The start of arrows indicates the 2006 cluster centers; the end of them indicates 2009 cluster centers or centers of fracture network. 3.3 Conclusion One application of fuzzy logic is the use of fuzzy clustering techniques to find the frac- ture network areas. Fuzzy cluster centers may represent centers of the connected frac- ture network which are ideal for reservoir creation and for enhancing production. The movement of these centers through time shows successful stimulation jobs, may help operators optimize and design future stimulation jobs, and provides useful information for long-term improvements to well spacing plan. 30 Figure 3.3: Zones boundary for applying the fuzzy clustering technique and the pro- duction wells locations. Figure 3.4: Fuzzy cluster centers for all the years at the HTZ zone 31 (a) (b) (c) (d) (e) Figure 3.5: Fuzzy cluster centers for all the years at the zone (a)one, (b)two, (c)three, (d)four, (e)five. 32 Figure 3.6: Movement of microseismic cluster centers at The Geysers; start of arrow is cluster center in 2006; end of arrow is cluster center in 2010 33 Chapter 4 Fractal Analysis We show that microseismic events - earthquakes with small magnitudes - can be fruit- fully used to gain insight into the properties of the fracture network of large-scale porous media, such as oil, gas, and geothermal reservoirs. As an example, we analyze exten- sive data derived from the GGF in northeast California. Injection of cold water into the reservoir to produce steam leads to microseismic events. We also demonstrate that anal- ysis can provide insight into whether the fractures are of tectonic type, or are induced by injection of cold water.As such, using the catalogue of the microseismic events, we estimate the fractal dimensionD f of the spatial distribution of hypocenters of the events in three seismic clusters associated with the injection of cold water into the field. The fractal dimensions are all in a narrow range centered around,D f ' 2:57 0:06, com- parable to the measured fractal dimension of fracture sets in the graywacke reservoir rock. Our results imply that the stress regime in the reservoir allows the activation of less favorably-oriented fractures that produce an increase inD f . The estimateDf 2 for tectonic seismicity has been interpreted as indicating that most tectonic events occur either on the subset of near-vertical faults, because they have lower normal stress, or on the backbone of the fracture and fault network, the multiply-connected part of the net- work that enables finite shear strain. Our results lend support to the latter. The results that the entire fracture network, and not just its backbone, is active at The Geysers indicate that the seismicity is not a result of the triggered release of tectonic stress, but 34 rather is induced by the release of local stress concentrations, driven by thermal contrac- tion that is not constrained by friction. The possible implication for hydraulic fracturing ,so-called fracking, is also briefly discussed. 4.1 Background Fractal geometry is expected for the tectonic fractures and many tectonic fault networks have been shown to have a fractal structure. Hirata et al. (1987) mapped the fault pat- terns, demonstrating that fractal structure could be anticipated from the fracture process generated by small or large earthquakes and that this type of rock fracturing from the macroscopic to the microscopic level is a scale-dependent process. Sammis et al. (1992) used fractal geometry to analyze the fracture pattern at The Geysers over a wide range of scales including regional maps, outcrops, and drill-cores. They concluded that the fracture network in the graywacke reservoir rock is fractal, with a dimension between 1.6 and 1.9 in 2-D planar section. Sahimi et al. (1993) reported fractal dimensions of 1.9 and 2.5 in 2D and 3D, respectively, for the fracture patterns in heterogeneous rocks. Studies by several groups have suggested that the fracture network of rock forma- tions may be self-similar and scale-invariant (for a comprehensive recent review, see Sahimi (2011) and Bonnet et al. (2001), implying that, statistically, the fracture network appears the same over a range of length scales, and that long-range correlations, which are a fundamental feature of fractal structures, affect any phenomenon that may occur in the network. Such studies began in 1985, when the geologic and hydrologic framework at Yucca Mountain in Nevada was being studied. Barton, C. C. (1985); Barton et al. (1987); Barton and Hsieh (1989) developed the so-called pavement method, whereby one clears a subplanar surface and maps the fracture surface, in order to measure its connectivity, trace length, density, and scaling, in addition to its orientation, surface 35 roughness, and aperture. An important finding from the Yucca Mountain study was that the fractured pavements had a scale-invariant structure, characterized by a fractal dimensionfractaldimensionD f , defined by n(`)/` D f ; (4.1) where n(`) is the number of fractures of length `, and D f is the fractal dimension of the network, which is less than the Euclidean dimension of space in which the network is embedded. The Yucca Mountain study indicated that it is possible to represent the distribution of fractures ranging from 20 cm to 20 m by a single parameter, D f . For the fracture surfaces analyzed by Barton and co-workers, D f ' 1:6 1:7. La Pointe (1988) carried out a careful reanalysis of three fracture-trace maps of Barton, C. C. (1985) estimating that the corresponding three-dimensional (3D) fracture networks are also fractal withD f ' 2:37, 2.52, and 2.68. Velde et al. (1991) analyzed the structure of fracture patterns in granites, while Vignes-Adler et al. (1991) carried out the same type analysis for fracturing in two African regions, reporting strong evidence for the fractality of the fracture patterns, while 2D maps of fracture traces spanning nearly ten orders of magnitude, ranging from microfractures in Archean Albites to large fractures in South Atlantic seafloors, were analyzed by Barton and LaPointe (1992), who reported thatD f ' 1:3 1:7. Sammis et al. (1992) analyzed the fracture pattern in the GGF over a wide range of scales, including regional maps, outcrops, and drill-cores, concluding that the fracture network in the graywacke reservoir rock has a fractal structure with a fractal dimension, 1:6 D f 1:9, in 2D (planar) sections. Sahimi et al. (1993) suggested that the fractal dimensions of the fracture patterns in heterogeneous rocks should be around 1.9 and 2.53 in two and three dimensions, respectively (see below). See also Hatton et al. (1993) for further discussion of the issue of 2D and 3D sampling 36 in laboratory tests. The results of such tests may depend on the heterogeneity and the anisotropy of the fracture set. On the other hand, Hirata et al. (1987) mapped the fault patterns in a certain rock formation demonstrating that a fractal pattern should be expected from the fracturing process, generated by earthquakes of various sizes, and that the fractures generated are scale-invariant over multiple length scales, ranging from the macroscopic to the field scale. Computer simulations (Sahimi and Goddard, 1986; Sahimi and Arbabi, 1992, 1996) as well as the simulation of hydraulic fracturing in which water is injected into a heterogeneous solid to generate fracture (Herrmann et al., 1993), indicated that the resulting fracture networks are self-similar fractals. Since earthquakes usually occur on existing faults, the spatial pattern of their hypocenters is often used to reveal the structure of their underlying fault network. Hirata (1989) estimated the fractal dimension of the spatial distribution of seismic events hypocenters in the Tohoku region, based on a correlation function (Equation 4.2). He reported fractal dimensions between 1.34 and 1.79 in 2D sections. Robertson et al. (1995) estimated the fractal dimension of the spatial distribution of the hypocenters of several aftershock sequences in south and central California, reportingD f to be between 1.82 and 2.07 in three dimensions, with an average of about 1.95. 4.2 True Fractal Dimension Analysis Microseismic events can be characterized by the fractal dimensionD f of hypocenters. The fractal dimensions are estimated based on a correlation function defined by (Hirata et al., 1987; Wiemer, 2001). C(r) = 2 N t (N t 1) N r (R<r) (4.2) 37 whereN r (R<r) is the number of pairs of events that have a spacingR less thanr, andN t is the total number of events within the region of interest. As pointed out earlier, injecting of cold water into a geothermal reservoir or hydraulic fracturing job induces microseismic events; thus, if the spatial distribution of such microseismic events has a fractal structure,C(r) should follow a power law, C(r)/r D f (4.3) The fractal dimensionD f 1 defined by Equation 4.3 is also called the correlation dimen- sion, and denoted sometimes byD c . Throughout this thesis, whenever we refer to the fractal dimension of our own data, we meanD f , as defined by Equation 4.3. We used catalogs provided by the Lawrence Berkeley National Laboratory and the online data set from Northern California Earthquake Data Center 2 . The area covered by our study was the northwest (NW) region of the GGF, indicated by rectangle in Fig- ure 1.5. As mentioned earlier, injecting cold water into the GGF induces microseismic events-earthquakes of small magnitudes. Their hypocenters and the locations of injec- tion wells are shown in Figure 4.1. Beall et al. (2010) reported a strong correlation between seismic activity and the rate of injection of cold water into the NW region of the GGF. Based on the area’s seismic activity and the location of injection wells, we initially defined three clusters consisting of the spatial distributions of the hypocenters of the microseismic events; see Figure 4.2. Cluster number 2, shown in Figure 4.2, was then divided into four subclusters, each of which was also analyzed to delineate possi- ble size effects. As Figures 4.1 and 4.2 indicate, some clusters and subclusters are more 1 D is equal to 0 for a point, 1 for a line, 2 for a plane, and 3 for a sphere. Non-integer values reveal a clustering of the events closest to the shape described by the nearest integer value 2 http://www.ncedc.org/SeismiQuery/events\_f.html 38 dense than others. We deliberately selected such clusters in order to also understand the effect of the events density on the properties computed. The spatial distribution of the hypocenters of the seismic events in the clusters that are shown Figure 4.2 was characterized by the fractal dimensionD f . The fractal dimen- sions were computed using the Zmap program (Wiemer, 2001) which determines D f using Equations 4.2 and 4.3. As an example, Figure 4.3 presents a plot of logC(r) for the entire spatial distribution of the hypocenters in region 2 of Figure 4.2. The linear portion of the curve yields a fractal dimension,D f ' 2:59. The interpretation of such values ofD f will be given shortly. We should point out that the fractal dimension estimated from the data presented in Figure 4.3 is for a bit less than two orders of magnitude variations in the distancer. In principle, the distancer over which the correlation function C(r) is varied and used to estimateD f must vary by about four orders of magnitude (Grassberger and Procaccia, 1983). If the range of variations of r is not broad enough, then one must consider an alternative interpretation of the data (Sahimi, 2011; Bonnet et al., 2001). Unfortunately, however, the range of length scales that can be explored in seismicity distributions is severely limited by the accuracy with which the individual events can be located, which itself is limited by the heterogeneity of the crust. At the same time, however, our data are not indicative of other interpretations and spatial distributions of the events. Thus, although one must, in principle, be cautious about attributing fractal characteristics to the data set and consider other possibilities, our results are completely consistent with such characteristics. Moreover, when the variations of local slopes are well-behaved and indicate a plateau, we may obtain a reliable estimate of the fractal dimensionD f . In Figure 4.3, we also show the variations of the local slope; it is zero over the tail region of C(r) and rises where the power law region begins. The scaling region where the power law is observed begins at r 0:04 km. The local slope at that point is about 39 2.3. The maximum of the local slope is about 2.7, only 17% larger. Atr 2 km where the power-law region ends andC(r) reaches a plateau, the local slope is about 2:1. The average of all the local slopes is 2.59. Hence, we conclude that the estimate ofD f from Figure 4.3 is reliable. Care must be applied in using seismicity to estimate the fractal dimension of a frac- ture network. Smith (1988) and Robertson et al. (1995) illustrated that a minimum number of data points exists for estimating the true fractal dimension of the underly- ing fracture network. Eneva (1996) also illustrated that the number of data points, the size of the region under study, and the measurements’ errors can significantly affect the estimate of the fractal dimension, and that assigning a specific physical meaning to the fractal dimension associated with a limited data set might be problematic. Thus, to ensure that we sampled a large enough number of data points to compute the true frac- tal dimension of the underlying fracture-fault network, the effective values ofD f were plotted as a function of the density of the events in a given cluster.Figure 4.5 indicates the same trends for the fractal structure of the spatial distributions of the hypocenters in the four subregions carved out of region 2. For all of the subregions, the fractal dimen- sions associated with the spatial distribution of the hypocenters converge to values that vary in a very narrow range. As a further test, the northwest region of the GGF was ana- lyzed separately, with the results shown in Figure 4.6, indicating again that the spatial distribution of the hypocenters in this region forms a fractal cluster. All of the estimated fractal dimensions are listed in Table 4.1. When earthquakes are not induced, for example, by the injection of cold water into a rock formation, and are of tectonic type, the value ofD f is always close to 2 (Sahimi et al., 1993). Such a value of D f has been interpreted in two different ways, One, it indicates that most tectonic events occur on a subset of near-vertical faults, because they have lower normal stress; the second interpretation (Sahimi et al., 1993) is that the 40 events occur on the fault networks backboneits multiply connected partwhich enables finite shear strain. The latter proposal is supported by the recent work of Past´ en et al. (2011), who analyzed the spatial distributions of hypo- and epicenters of earthquakes in central Chile and reported estimates ofD f , which are consistent with this hypothesis. We shall come back to this point shortly. In any case, estimates ofD f reported here are significantly larger than 2, thus con- firming that seismic activity in the GGF is more likely to have been induced by the injection of cold water into the formation, rather than being of tectonic type. Therefore, the estimates ofD f provide significant insight into the structure of the fracture networks, as well as into their origin. It is, therefore, possible to directly use the spatial distribution of microseismic events to map out the fracture network of a large-scale porous medium, such as a geothermal reservoir. Also, the pattern of microseismic events is the same as the nucleation and growth of fractures in random media Sahimi et al. (1993). In addi- tion, the fact that computedD f is significantly smaller than 3– the spatial dimension of the region in which the hypocenters are embedded– implies that only a small part of the overall structure contributes to distributing the strains. Region Measured Fractal Dimension D C NW Geysers 2:58 0:03 Region 1 2:50 0:03 Region 2 2:63 0:06 Region 3 2:58 0:03 Region 2-1 2:60 0:04 Region 2-2 2:60 0:04 Region 2-3 2:62 0:06 Region 2-4 2:51 0:03 Table 4.1: Estimates of the fractal dimensions for the individual regions. 41 ZMAP 25−Jan−2011 −122.84 −122.83 −122.82 −122.81 −122.8 −122.79 −122.78 −122.77 38.785 38.79 38.795 38.8 38.805 38.81 38.815 38.82 38.825 38.83 38.835 Longitude (degrees) Latitude (degrees) Active Injection well Figure 4.1: Clusters of the earthquakes hypocenters and the locations of active injection wells from 2006 to 2011. Each point represents one event. 4.3 Physical Interpretation of the Results If fractures nucleate and grow more or less at random in a highly heterogeneous medium, such as large-scale porous formations, then they should form a network of interconnected fractures that resembles what is called apercolationcluster (Stauffer and Aharony, 1994; Sahimi, 1994), i.e. a cluster of (more or less) randomly distributed inter- connected fractures that percolates between two widely-separated planes. To describe this phenomenon in more intuitive but physically understandable basis, we appeal to the critical path analysis (CPA) first developed by Ambegaokar et al. (1971) and con- firmed by many sets of simulations. They argued that transport processes in a highly heterogeneous medium can be reduced to one in a percolation system at or very near 42 ZMAP 24−Oct−2012 −122.84 −122.83 −122.82 −122.81 −122.8 −122.79 −122.78 −122.77 38.785 38.79 38.795 38.8 38.805 38.81 38.815 38.82 38.825 38.83 38.835 Longitude (degrees) Latitude (degrees) z<2.1 km z<4.2 km z<7.0 km 3 2−1 2−2 2−3 2−4 1 Figure 4.2: The four regions studied, as well as the four subregions. Each point repre- sents the location of an event. the percolation threshold. The idea is that in a medium with broadly distributed het- erogeneities, a finite portion of the system possesses a very small conductivity, hence making a negligible contribution to the overall conductivity or other effective flow or transport properties. Therefore, zones of low conductivity may be eliminated from the medium, which would then reduce it to a percolation system. Ambegaokar et al. (1971) described a procedure by which the equivalent percolation network, called the critical path, is built up. They showed that the resulting percolation system is at or very near its percolation threshold. When applied to heterogeneous fractured rock (Sahimi, 2011), CPA suggests that the fracture network must have the connectivity of a percolation clus- ter because, for example, the fractures are the main conduits for fluid flow in rock as their permeabilities or hydraulic conductances are much larger than those of the matrix in which they are embedded. Using the procedure of Ambegaokar et al. (1971), one finds that the fracture network of rock must be at, or very near, its percolation threshold. 43 10 −4 10 −3 10 −2 10 −1 10 0 10 1 10 −4 10 −3 10 −2 10 −1 10 0 D f ' 2.59±0.02 Distance r [km] Correlation Integral C(r) −5 −4 −3 −2 −1 0 1 2 3 4 5 Figure 4.3: The logarithmic plot of the correlation functionC(r) vs distance in region 2 of The Geysers geothermal field. Also shown are the variations of the local slopes around a constant value, indicating the accuracy of the data and the overall slope of the plot. The relationship among percolation theory, the spatial distribution of earthquakes hypo- and epicenters, and fault-fracture networks was first explored by Stark and Stark (1991); Trifu and Radulian (1989) in a qualitative manner, but was configured in a quan- titative foundation by Sahimi et al. (1993). The utility of identifying the fracture net- work of large-scale porous media with the sample-spanning percolation cluster is that the latter has been studied extensively (Stauffer and Aharony, 1994; Sahimi, 1994). 44 2.2 2.3 2.4 2.5 2.6 2.7 Region 1 ractal Dimension 1.9 2 2.1 0 25 50 75 100 125 150 175 200 225 2006 2007 2008 2009 2010 2009-2010 2006-2010 2008-2010 g   3 km Events Density Fr Figure 4.4: Dependence of the fractal dimensionD f on the density of the microseismic events in region 1 from 2006 to 2010. In particular, it is well known that the sample-spanning percolation cluster at or very near the percolation threshold is a self-similar fractal object with a fractal dimension, D f ' 1:9 and 2:53, in two and three dimensions, respectively. Moreover, the multiply connected part of the cluster, which allows various phenomena such as fluid flow and stress transport to occur in the network, is the aforementioned backbone, which is also a fractal object with fractal dimensions of 1.64 and 1.9 in two and three dimensions, respectively. Indeed, as mentioned earlier, the recent work of Past´ en et al. (2011), who analyzed the spatial distributions of hypo- and epicenters of earthquakes in central Chile, yielded D f ' 2:02 0:05 and 1:73 0:02, respectively, which are within 5% of the fractal dimensions of 3D and 2D percolation backbones. We must point out that a network of interconnected fractures and/or faults with irregular shapes and sizes resembles what is usually referred to as continuum perco- lation (Balberg, 2009), which differs from the better known and more studied lattice 45 2.7 2.7 2.5 2.5 23 23 n n 2.3 2.3 nsio nsion 2.1 2.1 Dimen imen tal D tal D 1.9 Region 2-1 1.9 Region 2-2 Frac Fract 17 2006 2007 2008 2009 17 2006 2007 2008 2009 F F 1.7 2010 2009-2010 2008-2010 2007-2010 1.7 2010 2009-2010 2008-2010 2007-2010 1.5 0 50 100 150 200 250 300 350 400 450 1.5 0 50 100 150 200 250 300 350 400 450 0 50 100 150 200 250 300 350 400 450 0 50 100 150 200 250 300 350 400 450   3 Events Density   3 k Events Density 2.7 2.7   3 km Density   3 km Density 2.5 2.5 23 23 on on 2.3 2.3 nsio ensio 2.1 2.1 Dime Dime Region 2-4 ctal D ctal D 1.9 2006 2007 Region 2-3 1.9 2006 2007 Region 2-4 Frac Frac 17 2006 2007 2008 2009 2010 2009 2010 17 2008 2009 2010 2006-2010 1.7 2010 2009-2010 2008-2010 1.7 2008-2010 2009-2010 2007-2010 1.5 0 50 100 150 200 250 300 350 400 450 1.5 0 50 100 150 200 250 300 350 400 450   Events Dit 0 25 50 75 100 125 150 175 200 225 250 275 300   Events Density   3 km Events Density   3 km Events Density Figure 4.5: Same as in Figure 4.4, but for the four subregions carved out of region 2 from 20062010. percolation, which deals with networks of bonds and sites. All numerical and analyt- ical works have indicated (Balberg, 2009), however, that the fractal dimensions of the sample-spanning clusters and their backbones are the same for lattice and continuum percolation. The estimates of the fractal dimensions listed in Table 4.1 deviate from that of the sample-spanning percolation cluster by, at most, 4%, well within the estimated errors, but not close to that of the percolation backbone. Thus, the seismicity induced by the injection of cold water happens on a fracture network that is similar to the 3D sample- spanning percolation cluster, whereas the tectonic events occur on the backbone of the fault network. The reason is that when cold water is injected into the GGF, the path the fluid takes within the porous formation and the fractures that it generates within the rock are, due to the heterogeneity of the formation, random. Even if the path is not random but contains extended correlations, the structure of the cluster at the largest 46 2.2 2.3 2.4 2.5 2.6 2.7 tal Dimension 1.9 2 2.1 0 25 50 75 100 125 2010 2009 2008 2007 2006 2009-2010 2008-2010 Northwest Region   3 km Events Density Fract Figure 4.6: Same as in Figure 4.4, but for the northwest region, from 2006 to 2010. length scale should still resemble that of a percolation cluster. The high-pressure cold water generates some fractures that are dead-ends, because the growth of such fractures stops only when the pressure of the water cannot overcome the resistance offered by the rock. As a result, the network generated by the injection contains both dead-end as well as multiply connected fractures, i.e., the sample-spanning percolation cluster. On the other hand, for earthquakes of tectonic origin to occur, finite strains and deformations must occur on the fault or fracture network. But that is possible only on the multiply connected part of the cluster, as the singly connected faults or fractures are dead-ends and cannot contribute to strain release. Therefore, such earthquakes should occur on the backbone of the fault-fracture network, which has a much lower fractal dimension close to 2. 47 The significance of the link between the structure of a fracture network and those of percolation clusters and their backbones is that the latter have been studied exten- sively, and deep insights into their structural properties have been gained (Stauffer and Aharony, 1994; Sahimi, 1994). This knowledge can, therefore, be used for realistically modeling a fracture network of the GGF or that of any other rock formation, for that matter. We will elaborate on this modeling approach in chapter 7 of this manuscript. 4.4 Concluding Remarks In this chapter, we analyzed the structure of the spatial distribution of hypocenters of microseismic events in the GGF. The results indicate that the distribution forms a fractal cluster with a fractal dimension very close to that of a 3D sample-spanning percolation cluster. The results also indicate that the spatial locations of microearthquakes hypocen- ters provide deeper insight into the structure of the fracture network of large-scale porous media. The dimensionD = 2 for tectonic seismicity has been interpreted as an indication that most tectonic events occur on the subset of near-vertical faults (because they have lower normal stress), or occur on the percolation 3 backbone of the fracture network which enables finite shear strain. The fractal dimension of about 2.6 is identical to the fractal dimension of nucleation and growth of fractures in random media. Hence, microseismic locations with a fractal dimension of 2.6 may reveal the connected fracture network and the reservoir heterogeneity. In addition, the results indicate that by calculating the fractal dimension of a micro- seismic cloud we may identify whether stimulated microseismic data are triggered (tec- tonic) or induced. Hence, we may find an explanation for changes in observed fracture 3 Percolation theory is a mathematical theory that examines the likelihood of connectivity, through a generated fracture network. 48 behavior or determine if those changes might be caused by the presence of nearby faults (tectonic) or by contact with the fracturing treatment (induced). Finally, determining the fractal behavior of microseismic event clouds in different stages of stimulation and their dimensions, allows us to assume that the fracture network at the underlying unconventional reservoir is self-similar (scale independent), and thus that its structure, mechanical, and transport properties are best described by using fractal geometry. We will discuss the latter in the chapter 7. 49 Chapter 5 b-value Analysis This chapter discusses the impact of considering b-value analysis in microseismic dis- tribution. We will discuss how this critical insight can lead to characterize a fracture net- work created from either hydraulic fracturing or fluid injection. The following method- ology is not specific to any kind of unconventional reservoir, and the only requirement is to have sufficient microseismicity with reliable locations. As an example, we analyze extensive data for the GGF in northeast California and limited induced seismicity in west Kern County in southern California. This type of analysis can also lead to insight into whether the fractures are of tectonic type, or induced by the injection of cold water. To demonstrate this phenomenon, we estimate the b values in the Gutenberg-Richter frequency-magnitude distribution using the catalogue of the microseismic events. For most cases, the b values are about b' 1:3 0:1. The b are significantly higher than those commonly observed for regional tectonic seismicity or aftershock sequences for whichb 1 are typical. Our results indicates that the seismicity is not a result of the triggered release of tectonic stress, but is induced by the release of local stress concen- trations, driven by thermal contraction that is not constrained by friction in the GGF or pore pressure increase, in the case of hydraulic fracturing. 5.1 Background and Application of b-value Analysis Some authors reported b = 1 as a universal constant for earthquakes in general (Frohlich, 1993; Kagan, 1999), but Schorlemmer et al. (2005); Gulia et al. (2010) showed how the 50 type of fault /fracture mechanism can affect b values. They concluded that a normal fault has a greater b-value than a strike-slip one, and that a strike-slip has a greater b-value than a thrust one. Wessels et al. (2011) said that this behavior is inversely related to the stress regime, which means that where we have higher b-value, we expect a lower stress regime (Table 5.1). The b-value may be proportional to vertical stress minus horizontal stress, because normal faults tend to happen under lower horizontal stresses than thrust faults do (Grob and van der Baan, 2011). systematically from the regional average and reach their extreme for pure event classes with range g# 58. Strike-slip events show only minor deviation from the regional averageb value, regardless of the rangeg. Only forg¼ 58 is theb value significantly higher than the average. The differences in the frequency–magnitude distribution between pure thrust and pure normal events are statistically highly significant (Fig. 1c and Table 2). The b-value variations of the Harvard catalogue are less pro- nounced;however,theyshowthesamedependenceonfaultingstyle (Fig.1d,e).Theseparationisclearerwhenusingonly thefirstnodal plane for classifying events (Fig. 1e, inset) rather than both nodal planes (Fig. 1e). Although each catalogue covers a different range of magnitudes, the frequency–magnitude distributions emphasize the similarity of the b values for the two displayed classes of events (Fig. 1c). The data in all three additional catalogues show essentially the same systematic dependence of b on faulting style (Fig. 2). The absolute level of b does not concern us here, because it can vary in different regions and for different networks. The important obser- vation is that, in all cases, normal events have the highest b values, followed by strike-slip and thrust events. In some data sets the separation isno longerclearfor thenarrowestrangesgoflbecause the error bars increase as a result of small sample sizes. The data sets analysed are very different in depth distribution, magnitude range and tectonic environment. Nevertheless, all data sets show the same dependence ofb on focal mechanisms. The uniformityof theseparation ofbvaluesaccording tofaulting type in these tectonically different data sets requires a universal interpretation. Weproposethatthedifferentialstress,and indirectly the confining pressure (towhich the differential stressis tied), is the parameter most stronglycontrollingfaultingtypes,thus influencing differences in b. This inverse relationship follows because we show thatb TH ,b SS ,b NR ,anditisknownthatforagivenverticalstress j v the mean stresses obey the relationship j TH . j SS . j NR . Ourresultsimplythattheinverserelationshipbetweendifferential stress Dj and b is universally valid, spanning the range from submillimetres to hundreds of kilometres of rupture length. The b value can therefore be interpreted as a ‘stressmeter’ in the Earth’s crust. However, locally, secondary effects can come from material properties 10,11 and from modifications of the stress tensor by pore pressure 12 . The idea that depths might be the controlling factor, in general, can be ruled out because the average depths of the three classes of eventsarenotsystematicallyorderedaccordingtothebvalues.Depth is also not a good candidate to be a universal factor controlling b, because the trends reported are opposite in California 13,14 and Japan 15 . The degree of material heterogeneity 10 as a fundamental causeoftheseuniversaldifferencescanalsoberuledout,becausethe observationisvalidforsmallaswellaslargerearthquakesandforall existing tectonic settings. Previous studies 4,16 based on fewer data and a less quantitative analysis of focal mechanism reported that the size distributionb of moments in the Harvard CMT catalogue (0–70km depth) obey the relationship b NR .b TH .b SS . However, the most comparable computation of Kagan, using the entire catalogue in the period 1987–1999withacompletenessofM c ¼ 5.4reproducestherelation that we found: (b NR ¼ 0.731). (b SS ¼ 0.643). (b TH ¼ 0.642). The inverse relationship between stress andb is consistent with a range of other observations: laboratory rock specimens 11,17 , mines 18 and increased pore pressure 12,19 (which decreases the differential stress). For the Los Angeles Basin, it was found 20 that b SS ¼ 1.1 and b TH ¼ 0.7 for strike-slip and thrust events, respectively. Although this result was based on only 144 and 78 earthquakes, respectively,itsupportsourfindings.Finally,thestrongdifferencesin bbetweenlockedpartsoffaults,wherebvaluesarelow,andcreeping sections with much higher b (refs 21–23), are consistent with the model.Theshearstrengthisassumedtobehigherinalockedsection thanincreepingsections.ThisresultsinhighDjandlowbvaluesfor Figure1|Plotsofbagainstrakeangleandrangeofrakeangle. a,b l plotof events in southern California, from the SCSN catalogue. In all frames the green,redandbluelines(solid,firstplane;outlined,secondplane)markthe b values of mainly normal, strike-slip and thrust events, respectively; the grey line marks the averageb value and the vertical bars indicate the standarderror 7 .(SolidbarsareusedforsampleswithN$ 200,anddashed bars for samples with 200.N$ 100). The circles at the top of the frame showtherakelusedforcomputingthebvaluesl¼2908^g,l¼ 08^g andl¼ 908^g,g¼ 208. b,b g plot of events in southern California. The circlesat the topof the frame show the rangeofrakel usedforcomputing theb values (g¼ 158,458 and 758). Inset, circle explaining the rake values and the corresponding colours of the classes of events. c, Frequency– magnitude distributions for pure normal (green) and pure thrust (blue) eventsoftheSCSNandHarvardcatalogues(g¼ 58).d,AsafortheHarvard catalogue.e,Asb for the Harvard catalogue. Inset, as main panel but considering only the rake of the first nodal plane. LETTERS NATURE|Vol 437|22 September 2005 540 © 2005 Nature Publishing Group (a) 9 The tectonic stress field T P B Normal S Hmax S hmin S v S v > S Hmax > S hmin P T B Reverse S hmin S Hmax S v S Hmax > S hmin > S v P B T Strike-slip S Hmax S hmin S v S Hmax > S v > S hmin Figure 1.2. E. M. Anderson’s classification scheme for relative stress magnitudes in normal, strike-slip and reverse faulting regions. Earthquake focal mechanisms, the beach balls on the right, are explained in Chapter 5. where ρ(z)is the density as a function of depth, g is gravitational acceleration and ρ is the mean overburden density (Jaeger and Cook 1971). In offshore areas, we correct for water depth S v = ρ w gz w + z z w ρ(z)gdz ≈ ρ w gz w + ρg(z − z w ) (1.6) where ρ w is the density of water and z w is the water depth. As ρ w ∼ 1 g/cm 3 (1.0 SG), water pressure (hydrostatic pressure) increases at a rate of 10 MPa/km (0.44 psi/ft). Most (b) Figure 5.1: b-values of normal (green), strike-slip (red) and thrust events (blue); average b-value (grey line); standard error (vertical bars)(Schorlemmer et al., 2005; Zoback, 2007) In recent years, b-value analysis has found application in monitoring and character- izing the fracturing process. For instance, Downie et al. (2010); Maxwell et al. (2010); Wessels et al. (2011) distinguished the fault movement from fracture stimulation by comparing the b-values for various distributions of microseismic events. Figure 5.2 demonstrates b-value of 2 for fracture related events and 1 for associated fault 51 Table 5.1: b-values versus stress regime and dominant faulting mechanism, based on work by Schorlemmer et al. (2005) b-value Stressregime Faulttype b< 1 S H >S h >S v 1 Reverse(compressive) b 1 S H >S v >S h Strike-slip b> 1 S v >S H >S h Normal(extensional) events. Figure 5.3 shows the application of b-value map on the same dataset to dif- ferentiate fault (darkblue) from fracture areas (green). Downie et al. (2010) also used the same concept to evaluate the efficiency of different stimulation stages. Figure 5.4 demonstrates that microseismic events in stage 7 of stimulation are fault related which can affect the oil production by water invasion or other factors. Furthermore, Grob and van der Baan (2011) demonstrate how the temporal evolution of b-values in steam flooding a heavy-oil reservoir may indicate the opening or closing fractures. Figure 5.5 demonstrates three regimes in their observation with relation to Schorlemmer et al. (2005) work: b-values larger than 1.1 (extensional faulting [normal] or opening of fractures), b-values around 1.0 (a strike-slip regime) and a final regime with values around 0.65 (closing of fractures or compressive faulting [reverse]). 5.2 b-value Analysis Microseismic events can be characterized by theb-value of their frequency-magnitude distribution in the Gutenberg-Richter relationship logN(m>M) =abM (5.1) 1 S H : maximum horizontal stress,S h : minimum horizontal stress,S v : vertical stress 52 Identifying fault activation in the Barnett We used Zmap software (Weimer, 2001) to analyze spatial variation of b values over the areal extent of the microseismic volume. Grid cells are set at 0.001 degrees latitude and longitude. The magnitude of completeness, M c , which is the lowest magnitude at which all events of that size are detected, was calculated using the maximum curvature method and b values were calculated using the maximum likelihood method (Woessner and Weimer, 2005). Frequency magnitude distributions are determined by plotting moment magnitude bins of 0.1 against the log of the bin count. The b value was determined by measuring the slope of the histogram for magnitudes above a common M c of -1.6 for consistency. Cumulative energy release plots for each mechanism were created by first converting moment magnitude into Joules released per event using the following formula: logJ = 1.5M w + 4.8 This formula is derived from the equation used by Kanamori and Anderson (1975) to convert moment magnitude to ergs and follows the Gutenberg-Richter energy relationship (Gutenberg and Richter, 1956). Cumulative energy released for a 24 hour period was then summed and normalized to 1 for each population to correct for the disparity in event count and total energy released by events of each mechanism. Slopes were then calculated within chosen time intervals to illustrate the variability in energy release rates of each mechanism. Statistical Analysis Examining the FMD of events can give us more information about how the rock formation breaks and what external factors may be involved. Maxwell et al. (2009) and Downie et al. (2010) differentiate fault movement from fracture stimulation by comparing the slope (b value) on FMD plots for different populations of events. Figure 2 shows two different b value populations —slopes of ~2 for fracture events and ~1 for fault events (De La Pena et al., 2011). Schorlemmer et al. (2005) and Gulia et al. (2010) show empirically that b values vary systematically with the type of fault motion, such that the b value of a normal fault will be greater than that of a strike-slip fault, and a strike- slip fault will have a greater b value than a thrust fault. Because thrust faults occur in higher stress regimes than normal faults, it was inferred that b values can be used as a stress state indicator that is inversely related to the stress regime. Fault events on average release a greater amount of energy and have a greater maximum magnitude when compared to fracture events; however, there are significantly fewer fault events than fracture events within this dataset. This is a result of the relatively small but linear area affected by the fault relative to the broad and well distributed fracture development. The spatial variation in b values (Figure 3) indicates the relative presence of fault activity vs. fracture activity, and largely mirrors the event distribution seen in Figure 1. Lower b values are concentrated where the fault crosses the wellbores whereas higher b values indicate typical fracture event generation. The spatial distribution of b values in Figure 3 show the variation between areas of effectively stimulated reservoir relative to areas that may not have been so effectively stimulated. The highest b values located near the outer edges of the treated area are anomalously high due to an insufficient event population per cell. FMD plots for individual cells (Figure 4) within Figure 3 are in agreement with our b values for the total population of fault and fracture events. Figure 2. A non-cumulative FMD histogram of fault (blue) and fracture events (red) shows the log of the number of events (y-axis) per 0.1 magnitude bin (x-axis). The b value of the fracture events is ~2 whereas the b value for fault events is ~1. © 2011 SEG SEG San Antonio 2011 Annual Meeting 1464 1464 Downloaded 27 Sep 2011 to 128.125.153.126. Redistribution subject to SEG license or copyright; see Terms of Use at http://segdl.org/ Figure 5.2: Microseismic histogram of fault (blue) and fracture related events (red)(Wessels et al., 2011) According to Eq. 5.1,b is the slope of the linear portion of the plot of logN versus M. The plot has negative curvature for small earthquakes, due to undersampling caused by the detection threshold. The break from negative curvature represents the so-called minimum magnitude of completeness, M c . There is also deviation from linearity for large values ofM, due to the limited observation times for properly sampling much-less- frequent larger events. In most cases,M c may be estimated by the maximum curvature method (Wyss, 2000). But, when we used it, the method did not yield physical estimates ofb in some cases, in which case manual curve fitting or the least-squares method was utilized for estimating the b values. If M c is determined by the maximum curvature method, then the b is estimated using the maximum likelihood method, according to which b = 0:433 hMiM c ; (5.2) 53 Identifying fault activation in the Barnett Figure 3. Cellular map view of b values within the treatment area. Lower values are associated with fault activity whereas higher b values are indicative of more natural fracture event generation. The highest values on the outer edges are due to insufficient event populations for FMD calculations. The location individual cells used for Figures 4a and 4b are indicated. Figure 4. Cumulative FMD plots for individual cells in Figure 3. A) A b value of 1.3 indicates fault related activity. B) A higher b value of 2.05 suggests a lower local strain rate and therefore a typical fracture failure mechanism. Timing and Rates of Energy Release Microseismicity generated by hydraulic pumping within a formation with low differential stress should theoretically be coincident with the time of pumping, i.e. when fluid and proppant are forcibly added to the formation. Similarly, we would expect high stress areas of the formation to generate microseismicity during pumping and also between pumping intervals when accumulated stress is released. Using microseismic events generated during one 24 hour period (Figure 5), we show that fracture generated microseismicity is largely isolated to a restricted period of pumping time whereas fault events are distributed over a much longer period of time (Figure 6). Figure 5. Map view of events used for time energy release comparison. All events shown occurred within a single 24 hour time period, which includes one fault affected stage (blue) and two fracture dominated stages (red). Figure 6. Normalized cumulative energy release for a 24 hour time period, separated by mechanism. Three complete hydraulic fracture stages took place during this time. Stage A has significant fault interaction while stage B and C generate very little fault related microseismicity. Looking at Table 1 and Figure 6, fault energy release during pumping occurs at a moderate rate (slope 1a) followed by a lower rate (slope 2a) that approximates an exponential decline curve. In contrast, we see that hydraulically stimulated natural fractures release microseismic energy at a very high rate initially (slope 1b), © 2011 SEG SEG San Antonio 2011 Annual Meeting 1465 1465 Downloaded 27 Sep 2011 to 128.125.153.126. Redistribution subject to SEG license or copyright; see Terms of Use at http://segdl.org/ Figure 5.3: b-value map within the fracture treatment area. Lower values are associ- ated with fault activity whereas higher b values are indicative of more fracture event creation.(Wessels et al., 2011) wherehMi is the average magnitude of the earthquakes. A typical plot is shown in Fig- ure 5.6 for the northwestern region of the GGF during 2006, illustrating the application of the maximum likelihood method for estimating b. 5.3 Result and Discussion We analyzed seismicity for the NW GGF using the NCEDC catalogues and seismicity in the west Kern County –which can be related to oil and gas activities 2 – using the SCEC database 3 ; see Figure 5.7. We determined thebvalue by measuring the slope frequency-magnitude distribu- tion for magnitudes above a commonM c where most seismic activities greater than this magnitude have been recorded. This value is also near to the mode of magnitudes. The 2 Mostly from hydraulic fracturing jobs or water waste disposal 3 http://www.data.scec.org/eq-catalogs/date_mag_loc.php 54 Figure 5.4: Comparison of b-value analysis for different stages of stimulation in hori- zontal Barnett Shale well (Downie et al., 2010) Statisticsofmicroseismiceventsandgeomechanicalimplications CASESTUDY We compute the average b and D-values for a microseismic dataset acquired above a heavy-oil field. The heavy-oil reser- voir is drained using cyclic steam stimulation. A total of 3211 events are recorded from August 2009 to April 2010. A change in injection/production strategy occurred in December 2009. Figure 2 shows the frequency-size distribution obtained using all events. This distribution follows a power law with an ex- ponent b equal to 1.35. The plateau for magnitudes less than -2.5 indicates that many smaller events are not successfully recorded, so the b-value is calculated only over the reliable magnitude-distribution part of the catalogue. The postulated link between magnitude distribution b, stress regime and domi- nant faulting mechanism in Table 1 implies that the local stress regime is most likely to lead to normal faulting (extension) and thus S v >S H >S h . The occurrence of extensional faulting is plausible in the areas above the steam cloud. Figure 2: Global event-size distribution for a heavy-oil dataset. The size distribution follows a power-law with a b-value of 1.35. The correlation integral to estimate the spatial fractal dimen- sion D is plotted in Figure 3. The D-value is estimated over the linear part of the curve and equals 2.72, meaning the events are distributed quite uniformly in space. The change in the slope of the curve after a distance r of 10 m is a sign of depopulation which could lead to a bias in the statistics, so the value D is only evaluated over the first part of the distribution. The large number of recorded events allows for an analysis of temporal variations in the b and D-values. Figure 4 repre- sents the variations of the b-value in time from August 2009 to March 2010. The b-values are computed over 300 events with an overlap of 30 events. Ranges are defined on the number of events and not on time to reduce potential bias in the estimated statistical value. Three different stages could be seen in Figure 4. High b-values are found at the very beginning (b > 1.1); then there is a de- crease to intermediate values (b∼1); and finally after January 2010, the b-values are quite low (b∼0.65). This could indi- cate a change in the stress state of the reservoir from extension (i.e., opening of fractures) to compression (closing of frac- tures) with an intermediate stage of mostly strike-slip events (Schorlemmer et al., 2005,and table 1). It would also indicate that horizontal stresses were originally smaller than the vertical stress (S v >S H >S h ) but dominated in the end (S H >S h >S v ). This is a plausible scenario from the initial phase of steam in- jection until the start of production. Figure 3: Global correlation integral for a heavy-oil dataset. The average spatial fractal dimension D equals 2.72 indicating a predominantly spherical event cloud. Figure 4: Temporal evolution of b-values for a heavy-oil dataset. Three regimes are visible with initial b-values larger than 1.1 (extensional faulting or opening of fractures) until November 2009, followed by b-values around 1.0 and finally a last stage with values around 0.65 (closing of fractures or compressive faulting), starting end of January 2010. Figure 5 represents the temporal evolution of fractal dimen- sion D for the same period and with the same parameters for the moving window as employed for computing the b-values. The variations in dimension D are more chaotic than those of the b-values. Dimension D varies mostly between 2 and 3, in- dicating no real clustering of events can be observed, except during December 2009 when the D value dropped to 1. This period corresponds to a change in injection/production strat- egy in the reservoir. Such processes can change the stress state in the reservoir, resulting in a different clustering of events at places where the variations are the highest. This would then be followed by a relaxation phase and events happen again in a more uniform or spherical distribution. The only D-values less than 2 also happen to occur in the strike-slip regime when more linear to planar microseismic event clouds are observed in hydraulic fracturing experiments in areas where strike-slip © 2011 SEG SEG San Antonio 2011 Annual Meeting 1572 1572 Downloaded 27 Sep 2011 to 128.125.153.126. Redistribution subject to SEG license or copyright; see Terms of Use at http://segdl.org/ Figure 5.5: Temporal evolution of b-values for a heavy-oil dataset to characterize open- ing or closing phase of fractures.), (Grob and van der Baan, 2011) distribution of magnitudes for microseismic events are interesting and important. For example, Figure 5.8 presents the distribution for the northwestern region of the GGF from 2006 to 2011. It peaks at M ' 1 but is not Gaussian (symmetric), as it has a relatively long tail for larger earthquakes. The same type of distribution was obtained for other regions of the GGF as well as for west Kern County; see Figure 5.9. This peak (mode) can be used asM c for estimatingbvalue. We usedM c ' 1:1 at The Geysers 55 0 1 2 3 4 10 0 10 1 10 2 10 3 10 4 Magnitude Cumulative Number M c = 1.1 a'5.05 b'1.32±0.05 Figure 5.6: Magnitude distributions of The Geysers seismicity in 2006 for events in the year 2006, which can also be obtained through the Maximum Curvature Method (Figure 5.6) and 1.67 for the seismicity distribution in west of Kern County 4 . Figure 5.11 presents the variations of b over the time period in which we studied the microseismic events at the GGF for the four subregions. Except for region 3, the b values are all larger than 1.3, indicating a very large number of very small seismic events, as larger values of b correspond to smaller earthquakes, which explains why the b values we obtained are all larger than 1, the typical value for large earthquakes with a tectonic origin. More interestingly, the b values for the same three regions approach 1.2 4 This value varies depending on the location of analysis and the relative sensitivity of geophones. For example, for the microseismicity of hydraulic fracturing where monitoring geophones are located so close to the events, there will be high sensitivity, with a value of about -1 to -2 56 and appear not to change for the next year or so between 2009 and 2010, and the nature of the events was still more likely to be of the induced type (b > 1) rather than the tectonic type (b 1). These findings are all consistent with the catalog of events that we studied. The same behavior has been seen inb-value estimation of seismicity in west Kern County; see Figure 5.10. Wyss (2000) emphasized the significance of studying the time variations of the b values. Estimates of the b values at the GGF vary from 1.11 to 1.32, and are all listed in Table 5.2. They represent estimates for the entire 2006 to 2010 period. b-values from both GGF and Kern County are more than 1.2 consistent for an induced event. These values are significantly higher than those commonly observed for regional tectonic seismicity or aftershock sequences, in whichb = 1 are typical. There- fore, seismicity is probably not the triggered release of tectonic stress, but is induced seismicity releasing local stress concentrations most likely driven by thermal contrac- tion at the GGF and by pore pressure increase in Kern County. Moreover, the frequency-magnitude distribution can be used to make probabilistic hazard forecasts for the discussed areas, by simply rewriting the equation 5.1 in terms of the annual probability of a target magnitudeM P (M >M target ) = 10 abM T (5.3) WhereT is the observation period, recurrence time is inverse of this annual probability. We have used this analysis to show how an accurateb value is critical to hazard analysis and physical understanding. We show how small changes in b value result in large changes in projected numbers of larger seismicity. Calculated Probability of more than 1 is considered 1 in the final results which means that an earthquake of this magnitude will probably happen each year. 57 Figure 5.12 demonstrates how an inaccurate estimation ofb value, instead of using the universal value of 1, can change the forecast for Kern County. Whereas usingb = 1:25 shows that an earthquake greater than 4 is not probable at all in this area, ab value of 1 will result in a recurrence of about 5 years for an earthquake of this magnitude. Our observation and forecast can be validated by looking at the history of seismicity in this area, where a magnitude of 4 has not occurred during 23 years of data gathered in the SCSN catalog; see Table 5.3. We implement the same analysis for The GGF data. Figure shows that an earthquake greater than 4.7 is not probable at The GGF, a conclusion that can be validated by both the history and the nature of seismicity there. But, using b-value of 1 results in a misleading forecast for the probability of an earthquake at The GGF. A through investigation of seismicity from 1990 to 2013 validated such a probabilistic model using b value of 1.32 at The GGF. During 33 years of production at The GGF, no earthquake with a magnitude greater than 5 was reported, and only 12 events with magnitude greater than 4 were reported–findings that are compatible with the forecast we have made in Figure 5.13. Hence, large earthquake can not be triggered from both of case studies and we are in a safe production zone, away from causing any possible hazard in nearby urban areas. Table 5.2: Estimates of theb values for the individual regions. Estimates ofb are for the 2006 to 2010 period. Region b-value NW Geysers 1:27 0:02 Region 1 1:33 0:02 Region 2 1:36 0:02 Region 3 1:28 0:02 Region 2-1 1:20 0:05 Region 2-2 1:10 0:02 Region 2-3 1:20 0:03 Region 2-4 1:17 0:03 58 S O N O M A B A S I N V E N T U R A B A S I N I M P E R I A L V A L L E Y B A S I N P A C I F I C O C E A N DISTRICT 3 DISTRICT 6 DISTRICT 5 DISTRICT 4 DISTRICT 2 DISTRICT 1 GEOTHERMAL DISTRICT 1 GEOTHERMAL DISTRICT 2 38 118 117 119 39 120 33 40 41 122 121 120 121 34 37 115 116 o o o S A N B E R N A R D I N O o o o o o 115 36 o 116 117 o o o o O R A N G E o o o 118 119 o o o T U L A R E S A N T A B A R B A R A L A S S E N o M O N T E R E Y S A N L U I S O B I S P O o o o o G L E N N o o o o o F R E S N O o o o o D E L o o o o o o M A D E R A M O N O M A R I P O S A A M A D O R E L D O R A D O N E V A D A A L P I N E T E H A M A T U O L U M N E M A R I N Y O L O C O N T R A C O S T A S O L A N O H U M B O L D T S I E R R A P L U M A S M O D O C S I S K I Y O U T R I N I T Y B U T T E N A P A SUTTER Y U B A S H A S T A 42 41 40 S O N O M A S A N M A T E O 35 34 33 R I V E R S I D E K E R N S A N D I E G O V E N T U R A S A N T A C R U Z L O S A N G E L E S I M P E R I A L K I N G S S T A N I S L A U S M E R C E D S A N B E N I T O 122 36 35 123 37 123 39 124 38 124 The Geysers Calistoga Susanville Litchfield Wendel Lake City Amedee Coso Desert Hot Springs Salton Sea Heber Brawley (Abd.) Mesquite (Abd.) East Mesa . STATE OF CALIFORNIA GRAY DAVIS, Governor Crescent City . Eureka . Ukiah . Lakeport . . Santa Rosa Calistoga Red Bluff . Redding . . San Jose . Monterey . Salinas . Department of Conservation Division of Oil, Gas, and Geothermal Resources 801 K Street, MS 20-20 Sacramento, CA 95814-3530 (916) 445-9686 www.consrv.ca.gov Alfred J. Zucca, Cartographer . King City . Paso Robles . . . Coalinga . Santa Barbara . Kettleman City . . . Susanville . . Alturas RESOURCES AGENCY MARY D. NICHOLS, Secretary Stockton Modesto . . Fresno . Ventura . Bakersfield . . Long Beach Los Angeles . Cypress . . Lone Pine . Bridgeport DEPARTMENT OF CONSERVATION DARRYL YOUNG, Director . Barstow . Palm Springs . . Desert Hot Springs . El Centro . Needles DIVISION OF OIL, GAS, AND GEOTHERMAL RESOURCES WILLIAM F. GUERARD, JR., State Oil and Gas Supervisor 44 43 21 20 22 23 24 25 46 2 11 13 12 4 5 7 6 8 9 10 6 3 2 9 10 12 11 13 15 33 34 42 41 40 38 39 37 36 8 7 19 20 21 2 3 4 5 35 16 1S 1N 2 9 8 7 6 5 4 3 10 12 16 32 30 31 15 14 13 1S 11 26 23 2 17 18 19 20 22 21 24 25 27 28 29 30 31 32 18 17 7 6 5 4 3 2 1N 8 27 25 26 28 29 11 10 9 19 18 22 19 11 12 13 14 16 15 17 18 8 10 10 11 12 14 13 15 16 8 10 9 1S 2 3 4 6 5 7 8 5 1E 2 3 4 6 7 9 27 18 17 24 23 22 12 13 14 15 1N 16 11 1W 1E 9 7 6 5 3 4 2 25 26 17 14 15 13 30 29 28 16 18 27 25 24 23 22 21 20 19 26 45 46 47 32 31 30 28 29 45 7 6 19 20 21 1W 2 3 4 5 42 44 43 35 36 37 38 39 41 40 17 16 9 8 10 11 12 14 13 15 34 32 26 28 29 30 31 33 31 25 26 28 27 29 30 32 4 3 17 18 19 21 20 1N 1S 11 14 13 5 6 7 8 9 10 12 10 8 15 2 4 3 5 6 7 9 2 14 15 16 7 8 9 10 11 12 13 3 2 11 12 14 13 15 16 17 6 5 4 14 15 16 17 18 1E 3 1W 4 S A N B E R N A R D I N O B A S E L I N E S A N B E R N A R D I N O M E R I D I A N 20 19 21 22 23 24 15 17 8 9 10 11 12 13 7 1E 16 2 4 3 5 6 30 27 23 24 25 26 28 29 31 33 32 34 35 23 24 23 24 21 21 M O U N T D I A B L O B A S E L I N E 22 20 13 14 16 15 17 18 19 22 20 13 14 15 16 17 18 19 2 1W 7 8 6 5 4 2 3 1N 12 11 4 3 5 6 7 9 8 10 22 13 14 15 16 17 19 18 20 32 31 23 24 26 25 27 28 29 30 9 10 9 10 11 15 14 12 13 12 5 4 6 7 8 9 10 11 8 H U M B O L D T M E R I D I A N M O U N T D I A B L O M E R I D I A N 3 4 5 6 7 2 3 4 2 3 4 5 8 7 5 29 19 20 21 15 14 16 17 16 17 18 11 1S 30 22 23 24 26 27 28 12 5 4 6 7 8 9 3 2 9 7 6 5 4 3 2 1E 1N 10 8 28 29 30 15 14 13 12 27 26 18 19 21 22 24 23 25 3 2 2 1W 1E 2 3 4 7 17 16 7 5 6 3 4 2 1W 6 10 20 8 9 17 16 14 13 11 12 10 18 19 26 27 25 24 23 22 21 8 9 17 16 15 14 12 13 11 28 19 18 17 16 15 14 13 29 6 8 7 1W 1E 5 3 4 2 3 2 39 38 12 11 10 9 8 35 36 37 32 34 33 1N 30 2 3 4 5 31 6 48 40 41 44 43 42 45 46 47 42 45 43 35 36 37 38 39 40 41 32 34 33 12 13 14 15 16 17 31 11 12 44 46 47 17 16 14 15 13 9 11 7 2 3 4 5 6 10 8 1E 2 10 9 8 6 7 5 4 3 44 42 32 2 5 45 1W 2 3 48 4 47 46 1E 1W 17 14 7 6 5 4 3 1S 43 35 36 37 38 39 41 40 33 34 5 4 3 2 1E 31 6 7 6 7 9 8 10 11 12 16 15 7 8 9 12 11 10 13 18 8 6 19 Petrolia (Abd.) Petaluma Half Moon Bay Brentwood Livermore Oil Creek Moody Gulch (Abd.) Sargent Bitterwater Hollister Quinado Canyon (Abd.) Monroe Swell King City Paris Valley Vallecitos McCool Ranch Lynch Canyon (Abd.) San Ardo - Main Area Cantua Nueva (Abd.) Cantua Creek (Abd.) Coalinga Jacalitos Point Pedernales Offshore (Fed.) Point Arguello Offshore (Fed.) Alegria (Abd.) Santa Maria Valley Casmalia Guadalupe Harris Canyon, NW. (Abd.) Orcutt Jesus Maria Point Conception Lompoc Huasna Alegria Offshore (Abd.) Arroyo Grande Lopez Canyon (Abd.) Cuarta Offshore (Abd.) Conception Offshore (Abd.) Four Deer (Abd.) Careaga Canyon Los Alamos Zaca Barham Ranch Cat Canyon Capitan Kreyenhagen Kettleman North Dome Turk Anticline Pleasant Valley San Joaquin Coalinga, E., Extension Raisin City Guijarral Hills Helm Five Points (Abd.) Burrel Burrel, SE. Kettleman City Russell Ranch Cuyama, Central (Abd.) Cuyama, S. Cienaga Canyon Pioneer Los Lobos San Emigdio (Abd.) Eagle Rest San Emigdio Creek (Abd.) Pleito White Wolf Landslide Pyramid Hills Devils Den Beer Nose Antelope Hills, N. Morales Canyon Taylor Canyon (Abd.) Antelope Hills Temblor, E. (Abd.) McDonald Anticline Carneros Creek Temblor Hills Chico-Martinez Cymric Temblor Ranch Belgian Anticline Lost Hills, NW. Lost Hills Belridge, N. Pescado Offshore (Fed.) Sisquoc Ranch (Abd.) Hondo Offshore (Fed.) Elwood,S. Offshore Las Varas Canyon (Abd.) Goleta (Abd.) Elwood Elwood, Offshore Area Gonyer Anticline (Abd.) Midway-Sunset Monument Junction Cal Canal Semitropic McKittrick Asphalto Belridge,S. Jerry Slough (Abd.) Railroad Gap Wasco Shafter (Abd.) Bowerbank Rio Bravo Shafter, SE. (Abd.) Goosloo Canal Elk Hills Buena Vista Greeley Westhaven (Abd.) Camden Riverdale Tulare Lake Van Ness Slough Hanford (Abd.) Dos Cuadras Offshore (Fed.) Summerland Summerland Offshore (Abd.) Mesa (Abd.) Carpinteria Offshore Santa Clara Offshore (Fed.) Sockeye Offshore (Fed.) Hueneme Offshore (Fed.) Capitola Park Yowlumne Paloma San Emidio Nose Rio Viejo Coles Levee, S. Coles Levee, N. Valpredo Tejon, N. Bellevue Edison Strand Lakeside, S. (Abd.) Lakeside (Abd.) Ten Section Shafter, N. English Colony Rosedale Rosedale Ranch Fruitvale Kern Bluff Jasmin, W. (Abd.) Jasmin Mount Poso Poso Creek Dyer Creek (Abd.) Terra Bella (Abd.) Deer Creek Deer Creek, N. Kern River Kern Front ~ Rincon Canada Larga Oakview (Abd.) Rincon Creek (Abd.) Ventura Ojai San Miguelito Saticoy West Montalvo El Rio (Abd.) Long Canyon (Abd.) Oxnard Santa Paula Bardsdale West Mountain South Mountain Las Posas Somis (Abd.) Moorpark, W. Moorpark Oak Park Sespe Timber Canyon Hopper Canyon Piru Creek (Abd.) Canton Creek (Abd.) Fillmore Chaffee Canyon Eureka Canyon Santa Clara Avenue Conejo (Abd.) Shiells Canyon Del Valle Temescal Holser Ramona, N. Ramona Oak Canyon Ant Hill Round Mountain Edison, NE. Mountain View Comanche Point Tejon Hills Tejon Flats (Abd.) Castaic Hills Hasley Canyon Tapia Tapo Canyon, S. Venice Beach (Abd.) Hyperion Playa Del Rey Sawtelle Salt Lake, S. Beverly Hills San Vicente Sherman (Abd.) Cheviot Hills Inglewood Lawndale Alondra Saugus (Abd.) Honor Rancho Wayside Canyon Torrance El Segundo Gaffey (Abd.) Wilmington Oakridge Big Mountain Torrey Canyon Simi Tapo Ridge Piru (Abd.) Tapo, N. Santa Susana Oat Mountain Newhall Las Llajas Newhall-Portrero Lyon Canyon (Abd.) Placerita Aliso Canyon Cascade Pacoima Mission (Abd.) Charlie Canyon (Abd.) Elizabeth Canyon (Abd.) Castaic Junction (Abd.) Bouquet Canyon (Abd.) Potrero Las Cienegas Bandini Rosecrans, S. Rosecrans, E. Beta Offshore (Fed.) Montebello Los Angeles, E. Dominguez Huntington Beach Belmont Offshore Rosecrans Howard Townsite Los Angeles City Long Beach Los Angeles Downtown Union Station Newgate Santa Fe Springs Long Beach Seal Beach Sunset Beach (Abd.) Walnut Lapworth (Abd.) Sansinena Brea-Olinda Olive Newport, W. Leffingwell (Abd.) La Mirada (Abd.) Talbert (Abd.) Coyote, E. Coyote, W. Newport Anaheim (Abd.) Rowland (Abd.) Yorba Buena Park, W. (Abd.) Buena Park, E. (Abd.) Esperanza Richfield Mahala Chino-Soquel Kraemer (Abd.) Kraemer, NE. (Abd.) Kraemer, W. (Abd.) Prado-Corona San Clemente (Abd.) Cristianitos Creek (Abd.) OIL, GAS, AND GEOTHERMAL FIELDS IN CALIFORNIA 2001 MAP S-1 Table Bluff (Abd.) Tompkins Hill Bunker Cotati (Abd.) Ryer Island Suisun Bay Kirby Hill Tremont (Abd.) Honker (Abd.) Winters Liberty Cut (Abd.) Liberty Island (Abd.) Dixon, E. (Abd.) Van Sickle Island Rio Vista Sherman Island Lindsey Slough Williams Bounde Creek Compton Landing, S. (Abd.) Princeton Stegeman Malton-Black Butte Kirkwood Red Bank Creek (Abd.) Artois (Abd.) Greenwood, S. (Abd.) Wilson Creek (Abd.) Larkin, W. Greenwood Orland (Abd.) Ord Bend Corning, S. (Abd.) Corning (Abd.) Rice Creek Perkins Lake Afton Llano Seco Afton, S. (Abd.) Durham Lone Star Dry Slough (Abd.) Moon Bend Sycamore Concord (Abd.) Willow Pass (Abd.) Los Medanos Mulligan Hill (Abd.) River Break Oakley, S. Pleasant Creek Dunnigan Hills Arbuckle Buckeye Grimes, W. Fairfield Knolls Eldorado Bend (Abd.) Madison (Abd.) Harlan Ranch (Abd.) Kirk Grimes Dufour Woodland Howells Point Sycamore Slough Zamora, N. (Abd.) Merritt Zamora Sutter City Hospital Nose (Abd.) Compton Landing Butte Sink West Butte Angel Slough (Abd.) Butte Slough Peace Valley (Abd.) Wild Goose (Abd.) Schohr Ranch (Abd.) Maine Prairie Cache Slough Cache Creek Crossroads (Abd.) Karnak Robbins Pierce Road Catlett (Abd.) Rio Jesus (Abd.) Nicolaus Sacramento Airport Brentwood, E. Oakley Dutch Slough Knightsen (Abd.) Roberts Island Tracy Union Island Vernalis Davis, SE. (Abd.) Florin (Abd.) Poppy Ridge (Abd.) Stone Lake (Abd.) Galt (Abd.) Lodi, SE. Lodi (Abd.) Lone Tree Creek Collegeville, E. Ash Slough Cheney Ranch (Abd.) Merrill Ave Mint Road Chowchilla Gill Ranch Moffat Ranch Caliente Offshore (Abd.) Molino Offshore (Abd.) Gaviota Offshore Naples Offshore (Abd.) Refugio Cove (Abd.) Glen Annie (Abd.) Pitas Point Offshore (Fed.) La Goleta Shale Point Shale Flats (Abd.) Antelope Plains (Abd.) Dudley Ridge (Abd.) Buttonwillow Harvester (Abd.) Trico, NW. (Abd.) Trico Semitropic, NW. (Abd.) (Abd.) 0 30 Miles 60 EEL RIVER BASIN Salton Sea Owens Lake (dry) Santa Catalina Is. San Clemente Is. San Nicolas Is. Santa Barbara Is. Anacapa Is. Santa Cruz Is. Santa Rosa Is. San Miguel Is. Mono Lake San Francisco Bay San Pablo Bay Goose Lake Eagle Lake Lake Tahoe Lake Almanor Lake Oroville Lake Berryessa Shasta Lake Clear Lake S A C R A M E N T O L A K E C O L U S A P L A C E R Bullock Bend Sacramento S A N San Francisco La Honda B A S I N H A L F M O O N 1W 18 B A S I N San Joaquin, NW. (Abd.) S A N J O A Q U I N Ave. 21 C U Y A M A B A S I N S A N T A M A R I A B A S I N San Luis Obispo Airport L O S A N G E L E S B A S I N Linda Headquarters 801 K Street, 20th Floor, MS 20, Sacramento, CA 95814-3530 Phone: (916) 445-9686, TDD (916) 324-2555 Telefax: (916) 323-0424 District No. 1 5816 Corporate Avenue, Suite 200, Cypress, CA 90630-4731 Phone: (714) 816-6847 Telefax: (714) 816-6853 District No. 2 1000 S. Hill Rd., Suite 116, Ventura, CA 93003-4458 Phone: (805) 654-4761 Telefax: (805) 654-4765 District No. 3 5075 S. Bradley Rd., Suite 221, Santa Maria, CA 93455 Phone: (805) 937-7246 Telefax: (805) 937-0673 District No. 4 4800 Stockdale Hwy., Suite 417, Bakersfield, CA 93309 Phone: (661) 322-4031 Telefax: (661) 861-0279 District No. 5 466 N. Fifth St., Coalinga, CA 93210 Phone: (559) 935-2941 Telefax: (559) 935-5154 District No. 6 801 K Street, 20th Floor, MS 22, Sacramento, CA 95814-3530 Phone: (916) 322-1110 Telefax: (916) 323-0424 OIL AND GAS DISTRICT BOUNDARIES AND OFFICES Headquarters & 801 K Street, 20th Floor, MS 21, Sacramento, CA 95814-3530 District No. G1 Phone: (916) 323-1788 Telefax: (916) 323-0424 District No. G2 1699 West Main Street, Suite E, El Centro, CA 92243-2235 Phone: (760) 353-9900 Telefax: (760) 353-9594 District No. G3 50 D Street, Room 300, Santa Rosa, CA 95404 Phone: (707) 576-2385 Telefax: (707) 576-2611 GEOTHERMAL DISTRICT BOUNDARIES AND OFFICES The State of California and the Department of Conservation/Division of Oil, Gas, and Geothermal Resources make no representation or warranties regarding the accuracy of the data from which this map was derived. Neither the State nor the Department shall be liable under any circumstances for any direct, indirect, special, incidental, or consequential damages with respect to any claim by any user or any third party on account of or arising from the use of this map. Scale 1:1,500,000 LEGEND CALIFORNIA COUNTIES WITH OIL, GAS, OR GEOTHERMAL PRODUCTION Alameda Contra Costa Fresno Kern Kings Los Angeles Monterey Orange San Benito San Bernardino San Luis Obispo San Mateo Santa Barbara Santa Clara Tulare Ventura Butte Colusa Glenn Humboldt Madera Merced Sacramento San Joaquin Solano Stanislaus Sutter Tehama Yolo Imperial Inyo Lake Lassen Mono Sonoma Alpine Colusa Fresno Imperial Inyo Kern Lake Lassen Mendocino Modoc Mono Monterey Napa Plumas Riverside San Bernardino San Diego San Luis Obispo Santa Barbara Shasta Sierra Sonoma Ventura Commercial Low-temperature Geothermal Use Electrical Generation from Geothermal Energy Gas Production Only Oil and Gas Production \ 101 299 5 299 299 395 395 101 5 80 5 80 A L A M E D A 580 680 101 101 101 101 395 46 5 5 5 15 10 8 10 15 40 15 99 58 15 58 395 99 99 99 J O A Q U I N 50 1W H U M B O L D T B A S E L I N E GEOTHERMAL DISTRICT 3 M E N D O C I N O 15 S A N F R A N C I S C O S A N T A C L A R A 14 Boyle Heights (Abd.) Whittier 99 Rice Creek, E. Rancho Capay Chico (Abd.) Willows-Beehive Bend Knights Landing (Abd.) Conway Ranch Willow Slough Sacramento By-Pass (Abd.) S A C R A M E N T O Todhunters Lake Greens Lake (Abd.) Putah Sink Winchester Lake Freeport (Abd.) Clarksburg Saxon Grand Island (Abd.) Merritt Island Elkhorn Slough W. Thornton-Walnut Grove Snodgrass Slough Thornton (Abd.) River Island King Island McDonald Island Harte (Abd.) East Islands Lodi Airport (Abd.) Sand Mound Slough (Abd.) Stockton (Abd.) French Camp Lathrop Lathrop, SE. (Abd.) McMullin Ranch Vernalis, SW. (Abd.) 14 Kettleman Middle Dome 11 Fremont Landing (Abd.) Verona (Abd.) B A S I N Tisdale Sutter Buttes Little Butte Creek 25 13 10 Dixon (Abd.) Millar Garrison City (Abd.) Calders Corner 27 20 (Abd.) Union Shafter, SE. S A L I N A S B A S I N Santa Maria Salt Lake San Diego Bishop I N Y O Casa Diablo C A L A V E R A S 11 South Lake Tahoe N O R T E Sacate Offshore (Fed.) 5 . Middletown . Oakland 42 o Horse Meadows (Abd.) Whittier Heights, N. (Abd.) Turnbull (Abd.) Wheeler Ridge Tejon Canfield Ranch McClung (Abd.) Bellevue, W. Stockdale Kernsumner Seventh Standard Welcome Valley Blackwells Corner Kirby Hill, N. (Abd.) Potrero Hills (Abd.) Denverton (Abd.) Denverton Creek Pinole Point (Abd.) Oil field Gas field Geothermal field Sedimentary basin with oil, gas, or geothermal production Division oil and gas district boundaries Division geothermal district boundaries Figure 5.7: Map for west of Kern County, the area under study. Table 5.3: Observation of large earthquakes in seismicity history of west Kern County, 1990 to 2013 between 0 to 10 km depth, from SCSN catalog. 1990-2013 (23 years of seismicity) Magnitude > 3 > 4 Number 5 none 5.4 Fractal Dimension Versus b-value In chapter 5 and 4, we reported that the fractal dimensions are all in a narrow range centered around, D f ' 2:57 0:06, and that in most cases the b values are about b' 1:3 0:1, consistent with the Aki relation, D f = 2b. Both D f and b are signif- icantly higher than those commonly observed for regional tectonic seismicity or after- shock sequences for which Df 2 and b 1 are typical. Our results indicate that the activation of less favorably-oriented fractures produce an increase in bothb andD f . This result further validate that the seismicity is not a result of the triggered release of 59 0 0.5 1 1.5 2 2.5 0 1000 2000 3000 4000 5000 6000 7000 8000 Magnitude M Number of Earthquakes Figure 5.8: Distribution of earthquakes magnitudes in The Geysers. Table 5.4: Observation of large earthquakes in seismicity history of The GGF, 1990 to 2013 between 0 to 7 km depth, from NCEDC catalog. 1980-2013 (33 years of seismicity) Magnitude > 4 > 5 Number 12 none tectonic stress, but is induced by the release of local stress concentrations, driven by thermal contraction unconstrained by friction. Aki (1981) proposed an important relation between the fractal dimension D f of a fault network and the bvalue in the Gutenberg-Richter (GR) law (Equation 5.1)– magnitude-frequency distribution of seismicity that develops on that network. If during an earthquake slip scales with the area of the active fault plane, then the Aki relation is 60 −1 0 1 2 3 4 5 0 5 10 15 20 25 30 Magnitude Histogram Magnitude Number Figure 5.9: Distribution of earthquakes magnitudes in west Kern County( 0-10 km) given by,D f = 3b=c, wherec is a scaling constant between moment and magnitude rela- tionship, which has a world-wide average of about 1.5 Kanamori and Anderson (1975). But, whereas Hirata et al. (1987) did not observe a correlation between the two for acoustic emissions in laboratory experiments, Hirata (1989) reported the approximate relation,D f ' 2:3 0:73b, for seismicity in the Tohoku region in Japan. Comprehen- sive discussions of the relation between D f and the b values are offered by Wyss and Sammis (2004); Chen et al. (2006). Main (1992) conducted a very notable work in an effort to find the origin of the positive/negative correlation between the fractal dimensions and b-values. Figure 5.14 61 1 1.5 2 2.5 3 3.5 10 0 10 1 10 2 10 3 Magnitude Cumulative Number b' 1.25±0.02 a' 4.65±0.02 M c =1.7 Figure 5.10: The frequency-magnitude plot for extracting the b value for the entire seismicity catalog of west Kern County from 1990 to 2013 demonstrates his results. He assumed two different models; model A with the align- ment of uniformly distributed events along a plane, and model B where earthquakes around potential nucleation points clustered on an existing failure plane such as a fault or fracture. He reported that model A has interaction and weakening potential, with more concentrated deformation of cracks where energy release potential is high, and that positive correlation exists between the two values. On the other hand, model B has a negative correlation between b and D, whereby mechanical hardening of the system forces a reduction in the potential energy release rate associated with distributed damage to decrease the local stresses on a crack. 62 1.3 1.4 1.3 1.4 1.2 1.3 b value 1.2 1.3 b value 1.1 2006 2007 2008 2009 2010 year NW Region year 1.1 1.2 2006 2007 2008 2009 2010 year year Region 1 1.4 1.4 year year year year 1.3 1.4 ue 1.3 1.4 ue 1.2 1.3 b value Region 2 1.2 1.3 b value Region 3 1.1 2006 2007 2008 2009 2010 year Region 2 1.1 2006 2007 2008 2009 2010 year Region 3 Figure 5.11: Time dependence of the b values in the four regions of The Geysers. In this section, we address whether the Aki relation betweenD f andb holds for the GGF. The b-value and stable fractal dimension in the 3 regions and 4 sub-regions in Figure 4.2 are given in Table 5.5. As Table 5.5 indicates, values of b are more scat- tered than those of the fractal dimension D f . This finding is, in fact, not surprising because although each hypocenter is on a fault, it is not obvious that each earthquake fully activates a fracture in the network, hence more uncertainty results in theb values. Note that, according to Table 5.5, values ofD f andb for the three regions roughly follow the Aki relation,D f 2b, whereas those for the subregions do not. This find- ing is presumably due to the higher sensitivity of theb values to the size of the area in which seismic activity and microearthquakes occur. Note also that the estimatesb > 1 confirm that the seismicity at the GGF is induced and does not have tectonic origin because, as pointed out earlier, the fractal dimension of the spatial distribution of earth- quakes’ hypocenters with a tectonic origin is usually close to 2 with ab value of about 63 AP=0.03% AP=1.1% RT= 3861 Yrs RT= 92 Yrs 0.0001 0.0010 0.0100 0.1000 1.0000 10.0000 100.0000 1000.0000 10000.0000 1.5 2 2.5 3 3.5 4 4.5 5 5.5 Annual Probility Recurrence Time(Yrs) Magnitude Not Probable Figure 5.12: Annual probability and recurrence time for seismicity in west Kern County 1 (Frohlich, 1993). Some researchers have suggested (Wiemer et al., 1998) that high b values are a necessary, but not sufficient, condition for earthquakes to occur near an active magmatic body– but that hypothesis is not applicable to the GGF. We would like to emphasize that, at this point, we are only documenting our findings for the GGF and providing a plausible explanation. Clearly, much work needs to be done to check the generality of the proposal. See Refs. (Main, 1992; Henderson and Main, 1994; ¨ Oncel et al., 1996) for alternative interpretations and discussions. [ht] 5.5 Conclusion It is possible to use microseismic moment-magnitude values from selected time periods to determine the b-values of those events, and to ascertain if the microseismic events that are stimulated have been triggered or induced. In other words, we can differentiate 64 AP=0.1% AP=7% RT= 810 Yrs RT= 14 Yrs 0.001 0.01 0.1 1 10 100 1000 1.5 2 2.5 3 3.5 4 4.5 5 5.5 Annual Probility Recurrence Time(Yrs) Magnitude Not Probable Figure 5.13: Annual probability and recurrence time for seismicity at The GGF fracture-related events from fault-related ones in real-time; a microseismic cloud with b- value larger than 1.2 is induced and not tectonic, and higher b-values mean lower stress. Furthermore, b-value evaluation helps us identify areas in which we have a fracture opening process or closing one. Whereas increase in b-value means an opening fracture, a decrease means a closing fracture. With this kind of calculation in real-time microseismic monitoring, we will be able to avoid triggering larger earthquakes, distinguish triggered microseismic events (not connected to fracture network) from induced ones (has permeability and is connected to the network), and optimize the stimulation cost and time by screening the nearby fault. Hitting the nearby fault through stimulation can create a channel for excess water production 5 , results in extra time and expense for completion, and cause a deviation of fracturing materials and fluids from their designed path. 5 Water production increases costs, and necessitates septation, storage, transportation, and disposal facilities for the produced water. 65 Damage mechanics with long-range interactions 533 where local stress concentrations lead to strong crack-crack interactions and a much more heterogeneous stress field. Specifically, the modified Griffith theory of Main (1991) is extended to consider the effect of long-range elastic interactions on a potential energy release rate G’ (modified for the case of an ensemble of isolated cracks), and hence a physical connection is determined between material weakening and the concentration of deformation on dominant fault zones. Although primarily aimed at explaining observations of stress corrosion cracking, the theory is completely general, and applies to any mechanism of quasi-static subcritical crack growth which produces a power-law distribution of fracture size and spacing. Two processes are described: (A) localization of two-dimensional damage onto an eventual one-dimensional rupture plane; and (B) fracture clustering in two dimensions around potential rupture nucleation points on a pre-existing fault. These two end-member models are schematically illustrated in Fig. 1. For case A a weak negative correlation between b and Dc is shown to be consistent with distributed deformation and the idea that crack growth and associated acoustic emission in intact specimens in the early stages of damage is associated with a stabilization of the material against macroscopic, dynamic failure. This establishes a direct connection (in an analytic form) between the distributions of crack size and spacing. The mechanical hardening effect is consistent with the usual tenets of damage mechanics, in which sample failure does not occur immediately on formation of the first crack, for example due to the stability of tensile crack growth in a compressive stress field (Ashby when taken over several cycles, could be explained by self-organized criticality, where fractally distributed ‘aval- anches’ of large events result from the local interaction of smaller elements. Usually this has been illustrated by spring-block-slider models for earthquakes on a fault of the type first proposed by Burridge & Knopoff (1967). The most important property of self-organized criticality in terms of the present work is that it represents a stationary or average state, far from thermodynamic equilibrium, to which the complex dynamical system has already evolved, perhaps over millenia (P. Bak, personal communication). Therefore it cannot by definition be applied to the evolution of damage within a single earthquake cycle or even a few cycles. If we are interested in determining this evolution, say for predictive purposes, it may be more fruitful to consider a mainshock as a kind of critical phase transition which the system is driven through (e.g. Bruce & Wallace 1989). This approach was applied by Henderson & Main (1992) to a one-dimensional fault model, based on fracture mechanics, which showed systematic variations in the capacity dimension and b-value, depending on the applied stress and the type of local interaction in the model. In the present paper the problem of how damage evolves in terms of the ordering of crack size and position is addressed using a mean field theory based on weakly interacting Griffith cracks. The main advantage of a mean field theory is the degree of analytic tractability, thereby allowing the correlations between the spacing and length distributions to be explicitly included. The main disadvan- tage is the lack of applicability near the point of critical, dynamic rupture on the scale of the sample in question, (a) (b) r------i I \* I \d . . r -------I I I I p7 I I I I I I I I Figure 1. Schematic diagram showing two different types of damage localization: (a) alignment of epicentres on an incipient fault plane (model A); (b) concentration of epicentres on jogs and asperities on a pre-existing fault plane (Model B). The left-hand diagrams show the initial background seismicity, the central diagrams the evolution towards more concentrated activity, and the right-hand diagrams the correlation plot P(r) associated with this change. P(r) is defined in equation (6) of the main text, and the correlation plot is shown with log-log axes. The fractal ranges (ro, r,) and (ri, r;) are confined to the linear part of the correlation plot P(r). Model (A) has fixed boundary conditions determined by the size of the laboratory sample. Model (B) has more arbitrary boundaries for the epicentre distribution which are chosen by the investigator, and are hence shown as dashed lines. Figure 5.14: Schematic diagram showing two different types of damage localization: (a) Model A, (b) Model B. The left diagrams show the initial seismicity distribution; the middle diagrams show the evolution toward more concentrated activity; and the right diagrams show the correlation plotP r (C r ) associated with this change.(Main, 1992) In addition, both the probabilistic forecast and physical understanding of the seismic- ity at the GGF and west Kern County indicates that large earthquake cannot be triggered from them, and that we are in a safe production zone, away from geohazards. Finally, the relationship between D f and the b values opens up another path for the characterization of a fracture network of highly heterogeneous rock with higher confidence compare to analyzing them individually. 66 Table 5.5: Estimates of the fractal dimensions and the b values for the individual regions. Estimates of b are for the 20062010 period. Region Measured Fractal Dimension b-value D-C NW Geysers 2:58 0:03 1:27 0:02 Region 1 2:50 0:03 1:33 0:02 Region 2 2:63 0:06 1:36 0:02 Region 3 2:58 0:03 1:28 0:02 Region 2-1 2:60 0:04 1:20 0:05 Region 2-2 2:60 0:04 1:10 0:02 Region 2-3 2:62 0:06 1:20 0:03 Region 2-4 2:51 0:03 1:17 0:03 67 Chapter 6 Application of Seismic Velocity Tomography in Fracture Characterization In chapter 2, we discussed how to obtain seismic velocity models from microseismic data. In this chapter, we will describe how we use that information to characterize the fracture network in unconventional reservoirs through a fundamental understanding of the relationship between seismic velocity models (both shear and compressional wave) and reservoir properties. Effective, reliable, and accurate characterization of unconventional reservoirs– especially their complex fracture system–requires a fundamental understanding of the geophysical and geomechanical properties of the reservoir rocks and fracture systems. Geophysical and geomechanical anomalies in the reservoirs can be associated with var- ious features of the reservoir such as porosity, fracture density, salinity, saturation, tec- tonic stress, fluid pressures, and lithology. Using tomographic inversion results is one method of accomplishing this task. For instance, Charlety et al. (2006) used 4D tomographic inversion for the Soultz EGS site to evaluate the stimulation process and identify the fracture network evolution. (Berge et al., 2001) used effective medium theories to estimate the effects of fractures on seis- mic velocity volumes at The Geysers. Finally, (Gritto et al., 2013) and (Julian et al., 68 1996) carried out analysis using tomographic inversion results at The Geysers to hypoth- esize reasons for observed velocity anomalies. Our work involves improved models and a more comprehensive interpretation of workflow using derived properties, as discussed hereafter. In this chapter, we use lithology logs, laboratory measurement of rock properties, effective medium theories, and production/injection rates to better understand the frac- ture network evolution and to locate regions with potentially higher fracture density. Our study utilizes P-wave and S-wave seismic velocity volumes provided by the Lawrence Berkeley National Laboratory (LBNL), which were derived from microseismic events during 2005 to 2010 (Boyle et al., 2011; Hutchings et al., 2011). Then, we apply krig- ing on these velocity volumes to enhance the resolution as described later and in our recent publication (Aminzadeh et al., 2013). We next integrate P-wave and S-wave velocity volumes with each other and with density volumes to profile the stress and rock properties at the reservoir (effective normal stress [ n ], confining stress[ h ], Poisson’s ratio, Bulk modulus, Young’s modulus, and Shear modulus). We correlate the derived property maps with associated injection/production data–obtained from the DOGGR database 1 –, and our fuzzy clustering method to investigate their effect on the fracture network at The Geysers. Figure 1.5 shows the area of interest with respect to The GGF. This kind of research has applications for both unconventional oil and gas reservoirs and geothermal ones. For example, lots of microseismic events are measured during hydraulic fracturing job in a shale oil or gas reservoir. The velocity models could be generated based on those events, and then a new porosity and fracture density map could be generated after the job. The same process can take place in an enhanced geothermal system. 1 http://maps.conservation.ca.gov/doms/doms-app.html 69 6.1 Joint Interpretation of Seismic Wave Velocities In this section, we review the existing rich literature about seismic wave veloci- ties to define a framework to relate seismic velocity models (both shear(V S ) and compressional(V P ) waves) and fracture properties. Martakis et al. (2006) generalized thatV P shows the structural details of the reservoir andV P =V S illuminates lithological details; see Figure 6.1. However, we and many other authors believe that many factors affect these velocities in all scales–such as porosity (Wyllie et al., 1956; Wyllie and Gre- gory, 1958; Berryman et al., 2002), fractures (Berge et al., 2001; Berryman and Wang, 2000), pore pressure, and saturation (Nur and Simmons, 1969; Berryman et al., 2002); see Figure 6.2. 836 N. Martakis, S. Kapotas and G.-A. Tselentis Figure 6 (a) P-wave velocity section showing structural details of the crust. Green is related to evaporite and blue to carbonate characteristic velocities. (b) V P /V S section showing lithological details. Brown indicates the existence of evaporite along thrust faults. Figure 7 (a) Plane sections of P-wave velocity and (b) V P /V S ratio at 4 km depth. C 2006 European Association of Geoscientists & Engineers, Geophysical Prospecting, 54, 829–847 Figure 6.1: V P section showing structural details andV P =V S section showing lithologi- cal details.(Martakis et al., 2006) White (1975) reported an increase in V P with saturation, while V S remained unchanged. Although Figure 6.3c clearly shows that saturation 2 has small or no effect on theV P , we can observe a modest reduction in theV S at The Geysers’ core measure- ment. Table 6.1 explains the effect of saturation from three different scenarios: fluid density, fluid compressibility, and shear weakening. Saturation raises the V P =V S and 2 Brine saturation 70 Courtesy SEG Figure 6.2: Factors affecting seismic compresional velocity - Qualitative overview, www.seg.org reducesV S in two of these scenarios, with little resulting effect onV P . Hence, we can conclude that high V P =V S anomalies associated with low V S anomalies are saturation anomalies. On the other hand, lowV P =V S anomalies associated with lowV P are caused from other phenomena, such as lithology or fracture density variations (Boitnott, 2003). Figure 6.3b shows that V P negatively correlates to porosity. Notably, different lithologies demonstrate dissimilar trends. Some with a wide range of velocities are confined to a narrow range of porosities, whereas some have a wide range of porosi- ties are confined to a narrow range of velocities, and some show a strong correlation of velocities to porosities. Charlety et al. (2006) also reported a decrease in V P with porosity using velocity and flowrate correlation. In conclusion, velocities are sensitive to porosity and rock type, but other factors such as textural and mineralogic variations 71 can also impact this behavior. Figure 6.3c illustrates that seismic velocities are insen- sitive to pressure, persisting to high effective pressures. Charlety et al. (2006) reported that reservoir pore pressure can reduce theV P in field scale measurement. Final Report: DE-FG07-99ID13761 New England Research, Inc. 11 Vp (m/s) Vs (m/s) 7000 4000 3000 2000 Lahar Andesite Microdiorite Dacite_Autobreccia Unspecified Tuff Vp (m/s) Depth (m) 7000 700 3000 2000 Vp (m/s) Vs (m/s) 7000 4000 3000 2000 Vp (m/s) Depth (m) 7000 700 3000 2000 Vp (m/s) Vs (m/s) 7000 4000 3000 2000 Vp (m/s) Depth (m) 7000 700 3000 2000 Vp (m/s) Vs (m/s) 7000 4000 3000 2000 Vp (m/s) Depth (m) 7000 700 3000 2000 Vp (m/s) Vs (m/s) 7000 4000 3000 2000 Vp (m/s) Depth (m) 7000 700 3000 2000 Vp (m/s) Vs (m/s) 7000 4000 3000 2000 Vp (m/s) Depth (m) 7000 700 3000 2000 Vp (m/s) Vs (m/s) 7000 4000 3000 2000 Vp (m/s) Depth (m) 7000 700 3000 2000 Vp (m/s) Vs (m/s) 7000 4000 3000 2000 Vp (m/s) Depth (m) 7000 700 3000 2000 Vp (m/s) Vs (m/s) 7000 4000 3000 2000 Vp (m/s) Depth (m) 7000 700 3000 2000 Vp (m/s) Vs (m/s) 7000 4000 3000 2000 Vp (m/s) Depth (m) 7000 700 3000 2000 Vp (m/s) Vs (m/s) 7000 4000 3000 2000 Vp (m/s) Depth (m) 7000 700 3000 2000 Vp (m/s) Vs (m/s) 7000 4000 3000 2000 Vp (m/s) Depth (m) 7000 700 3000 2000 Vp (m/s) Vs (m/s) 7000 4000 3000 2000 Vp (m/s) Depth (m) 7000 700 3000 2000 Vp (m/s) Vs (m/s) 7000 4000 3000 2000 Vp (m/s) Depth (m) 7000 700 3000 2000 Vp (m/s) Vs (m/s) 7000 4000 3000 2000 Vp (m/s) Depth (m) 7000 700 3000 2000 Vp (m/s) Vs (m/s) 7000 4000 3000 2000 Vp (m/s) Depth (m) 7000 700 3000 2000 Vp (m/s) Vs (m/s) 7000 4000 3000 2000 Vp (m/s) Depth (m) 7000 700 3000 2000 Vp (m/s) Vs (m/s) 7000 4000 3000 2000 Vp (m/s) Depth (m) 7000 700 3000 2000 Vp (m/s) Vs (m/s) 7000 4000 3000 2000 Vp (m/s) Depth (m) 7000 700 3000 2000 Vp (m/s) Vs (m/s) 7000 4000 3000 2000 Vp (m/s) Depth (m) 7000 700 3000 2000 Vp (m/s) Vs (m/s) 7000 4000 3000 2000 Vp (m/s) Depth (m) 7000 700 3000 2000 Vp (m/s) Vs (m/s) 7000 4000 3000 2000 Vp (m/s) Depth (m) 7000 700 3000 2000 Vp (m/s) Vs (m/s) 7000 4000 3000 2000 Vp (m/s) Depth (m) 7000 700 3000 2000 Vp (m/s) Vs (m/s) 7000 4000 3000 2000 Vp (m/s) Depth (m) 7000 700 3000 2000 Vp (m/s) Vs (m/s) 7000 4000 3000 2000 Vp (m/s) Depth (m) 7000 700 3000 2000 Vp (m/s) Vs (m/s) 7000 4000 3000 2000 Vp (m/s) Depth (m) 7000 700 3000 2000 Vp (m/s) Vs (m/s) 7000 4000 3000 2000 Vp (m/s) Depth (m) 7000 700 3000 2000 Vp (m/s) Vs (m/s) 7000 4000 3000 2000 Vp (m/s) Depth (m) 7000 700 3000 2000 Vp (m/s) Vs (m/s) 7000 4000 3000 2000 Vp (m/s) Depth (m) 7000 700 3000 2000 Vp (m/s) Vs (m/s) 7000 4000 3000 2000 Vp (m/s) Depth (m) 7000 700 3000 2000 Figure 1: Summary of dry velocities for all samples tested. Note that Vp correlates well with Vs . There is no clear separation between lithologies and no apparent signature with depth. Vp/Vs Vp (m/s) 2 7000 1.3 3000 Lahar Andesite Microdiorite Dacite_Autobreccia Unspecified Tuff porosity Vp (m/s) 0.2 7000 0 3000 Vp/Vs Vp (m/s) 2 7000 1.3 3000 porosity Vp (m/s) 0.2 7000 0 3000 Vp/Vs Vp (m/s) 2 7000 1.3 3000 porosity Vp (m/s) 0.2 7000 0 3000 Vp/Vs Vp (m/s) 2 7000 1.3 3000 porosity Vp (m/s) 0.2 7000 0 3000 Vp/Vs Vp (m/s) 2 7000 1.3 3000 porosity Vp (m/s) 0.2 7000 0 3000 Vp/Vs Vp (m/s) 2 7000 1.3 3000 porosity Vp (m/s) 0.2 7000 0 3000 Vp/Vs Vp (m/s) 2 7000 1.3 3000 porosity Vp (m/s) 0.2 7000 0 3000 Vp/Vs Vp (m/s) 2 7000 1.3 3000 porosity Vp (m/s) 0.2 7000 0 3000 Vp/Vs Vp (m/s) 2 7000 1.3 3000 porosity Vp (m/s) 0.2 7000 0 3000 Vp/Vs Vp (m/s) 2 7000 1.3 3000 porosity Vp (m/s) 0.2 7000 0 3000 Vp/Vs Vp (m/s) 2 7000 1.3 3000 porosity Vp (m/s) 0.2 7000 0 3000 Vp/Vs Vp (m/s) 2 7000 1.3 3000 porosity Vp (m/s) 0.2 7000 0 3000 Vp/Vs Vp (m/s) 2 7000 1.3 3000 porosity Vp (m/s) 0.2 7000 0 3000 Vp/Vs Vp (m/s) 2 7000 1.3 3000 porosity Vp (m/s) 0.2 7000 0 3000 Vp/Vs Vp (m/s) 2 7000 1.3 3000 porosity Vp (m/s) 0.2 7000 0 3000 Vp/Vs Vp (m/s) 2 7000 1.3 3000 porosity Vp (m/s) 0.2 7000 0 3000 Vp/Vs Vp (m/s) 2 7000 1.3 3000 porosity Vp (m/s) 0.2 7000 0 3000 Vp/Vs Vp (m/s) 2 7000 1.3 3000 porosity Vp (m/s) 0.2 7000 0 3000 Vp/Vs Vp (m/s) 2 7000 1.3 3000 porosity Vp (m/s) 0.2 7000 0 3000 Vp/Vs Vp (m/s) 2 7000 1.3 3000 porosity Vp (m/s) 0.2 7000 0 3000 Vp/Vs Vp (m/s) 2 7000 1.3 3000 porosity Vp (m/s) 0.2 7000 0 3000 Vp/Vs Vp (m/s) 2 7000 1.3 3000 porosity Vp (m/s) 0.2 7000 0 3000 Vp/Vs Vp (m/s) 2 7000 1.3 3000 porosity Vp (m/s) 0.2 7000 0 3000 Vp/Vs Vp (m/s) 2 7000 1.3 3000 porosity Vp (m/s) 0.2 7000 0 3000 Vp/Vs Vp (m/s) 2 7000 1.3 3000 porosity Vp (m/s) 0.2 7000 0 3000 Vp/Vs Vp (m/s) 2 7000 1.3 3000 porosity Vp (m/s) 0.2 7000 0 3000 Vp/Vs Vp (m/s) 2 7000 1.3 3000 porosity Vp (m/s) 0.2 7000 0 3000 Vp/Vs Vp (m/s) 2 7000 1.3 3000 porosity Vp (m/s) 0.2 7000 0 3000 Vp/Vs Vp (m/s) 2 7000 1.3 3000 porosity Vp (m/s) 0.2 7000 0 3000 Vp/Vs Vp (m/s) 2 7000 1.3 3000 porosity Vp (m/s) 0.2 7000 0 3000 Figure 2: Summary of dry velocities for all samples tested. Note that Vp exhibits a positive correlation with Vp /Vs and a negative correlation with porosity. (a) Final Report: DE-FG07-99ID13761 New England Research, Inc. 11 Vp (m/s) Vs (m/s) 7000 4000 3000 2000 Lahar Andesite Microdiorite Dacite_Autobreccia Unspecified Tuff Vp (m/s) Depth (m) 7000 700 3000 2000 Vp (m/s) Vs (m/s) 7000 4000 3000 2000 Vp (m/s) Depth (m) 7000 700 3000 2000 Vp (m/s) Vs (m/s) 7000 4000 3000 2000 Vp (m/s) Depth (m) 7000 700 3000 2000 Vp (m/s) Vs (m/s) 7000 4000 3000 2000 Vp (m/s) Depth (m) 7000 700 3000 2000 Vp (m/s) Vs (m/s) 7000 4000 3000 2000 Vp (m/s) Depth (m) 7000 700 3000 2000 Vp (m/s) Vs (m/s) 7000 4000 3000 2000 Vp (m/s) Depth (m) 7000 700 3000 2000 Vp (m/s) Vs (m/s) 7000 4000 3000 2000 Vp (m/s) Depth (m) 7000 700 3000 2000 Vp (m/s) Vs (m/s) 7000 4000 3000 2000 Vp (m/s) Depth (m) 7000 700 3000 2000 Vp (m/s) Vs (m/s) 7000 4000 3000 2000 Vp (m/s) Depth (m) 7000 700 3000 2000 Vp (m/s) Vs (m/s) 7000 4000 3000 2000 Vp (m/s) Depth (m) 7000 700 3000 2000 Vp (m/s) Vs (m/s) 7000 4000 3000 2000 Vp (m/s) Depth (m) 7000 700 3000 2000 Vp (m/s) Vs (m/s) 7000 4000 3000 2000 Vp (m/s) Depth (m) 7000 700 3000 2000 Vp (m/s) Vs (m/s) 7000 4000 3000 2000 Vp (m/s) Depth (m) 7000 700 3000 2000 Vp (m/s) Vs (m/s) 7000 4000 3000 2000 Vp (m/s) Depth (m) 7000 700 3000 2000 Vp (m/s) Vs (m/s) 7000 4000 3000 2000 Vp (m/s) Depth (m) 7000 700 3000 2000 Vp (m/s) Vs (m/s) 7000 4000 3000 2000 Vp (m/s) Depth (m) 7000 700 3000 2000 Vp (m/s) Vs (m/s) 7000 4000 3000 2000 Vp (m/s) Depth (m) 7000 700 3000 2000 Vp (m/s) Vs (m/s) 7000 4000 3000 2000 Vp (m/s) Depth (m) 7000 700 3000 2000 Vp (m/s) Vs (m/s) 7000 4000 3000 2000 Vp (m/s) Depth (m) 7000 700 3000 2000 Vp (m/s) Vs (m/s) 7000 4000 3000 2000 Vp (m/s) Depth (m) 7000 700 3000 2000 Vp (m/s) Vs (m/s) 7000 4000 3000 2000 Vp (m/s) Depth (m) 7000 700 3000 2000 Vp (m/s) Vs (m/s) 7000 4000 3000 2000 Vp (m/s) Depth (m) 7000 700 3000 2000 Vp (m/s) Vs (m/s) 7000 4000 3000 2000 Vp (m/s) Depth (m) 7000 700 3000 2000 Vp (m/s) Vs (m/s) 7000 4000 3000 2000 Vp (m/s) Depth (m) 7000 700 3000 2000 Vp (m/s) Vs (m/s) 7000 4000 3000 2000 Vp (m/s) Depth (m) 7000 700 3000 2000 Vp (m/s) Vs (m/s) 7000 4000 3000 2000 Vp (m/s) Depth (m) 7000 700 3000 2000 Vp (m/s) Vs (m/s) 7000 4000 3000 2000 Vp (m/s) Depth (m) 7000 700 3000 2000 Vp (m/s) Vs (m/s) 7000 4000 3000 2000 Vp (m/s) Depth (m) 7000 700 3000 2000 Vp (m/s) Vs (m/s) 7000 4000 3000 2000 Vp (m/s) Depth (m) 7000 700 3000 2000 Vp (m/s) Vs (m/s) 7000 4000 3000 2000 Vp (m/s) Depth (m) 7000 700 3000 2000 Figure 1: Summary of dry velocities for all samples tested. Note that Vp correlates well with Vs . There is no clear separation between lithologies and no apparent signature with depth. Vp/Vs Vp (m/s) 2 7000 1.3 3000 Lahar Andesite Microdiorite Dacite_Autobreccia Unspecified Tuff porosity Vp (m/s) 0.2 7000 0 3000 Vp/Vs Vp (m/s) 2 7000 1.3 3000 porosity Vp (m/s) 0.2 7000 0 3000 Vp/Vs Vp (m/s) 2 7000 1.3 3000 porosity Vp (m/s) 0.2 7000 0 3000 Vp/Vs Vp (m/s) 2 7000 1.3 3000 porosity Vp (m/s) 0.2 7000 0 3000 Vp/Vs Vp (m/s) 2 7000 1.3 3000 porosity Vp (m/s) 0.2 7000 0 3000 Vp/Vs Vp (m/s) 2 7000 1.3 3000 porosity Vp (m/s) 0.2 7000 0 3000 Vp/Vs Vp (m/s) 2 7000 1.3 3000 porosity Vp (m/s) 0.2 7000 0 3000 Vp/Vs Vp (m/s) 2 7000 1.3 3000 porosity Vp (m/s) 0.2 7000 0 3000 Vp/Vs Vp (m/s) 2 7000 1.3 3000 porosity Vp (m/s) 0.2 7000 0 3000 Vp/Vs Vp (m/s) 2 7000 1.3 3000 porosity Vp (m/s) 0.2 7000 0 3000 Vp/Vs Vp (m/s) 2 7000 1.3 3000 porosity Vp (m/s) 0.2 7000 0 3000 Vp/Vs Vp (m/s) 2 7000 1.3 3000 porosity Vp (m/s) 0.2 7000 0 3000 Vp/Vs Vp (m/s) 2 7000 1.3 3000 porosity Vp (m/s) 0.2 7000 0 3000 Vp/Vs Vp (m/s) 2 7000 1.3 3000 porosity Vp (m/s) 0.2 7000 0 3000 Vp/Vs Vp (m/s) 2 7000 1.3 3000 porosity Vp (m/s) 0.2 7000 0 3000 Vp/Vs Vp (m/s) 2 7000 1.3 3000 porosity Vp (m/s) 0.2 7000 0 3000 Vp/Vs Vp (m/s) 2 7000 1.3 3000 porosity Vp (m/s) 0.2 7000 0 3000 Vp/Vs Vp (m/s) 2 7000 1.3 3000 porosity Vp (m/s) 0.2 7000 0 3000 Vp/Vs Vp (m/s) 2 7000 1.3 3000 porosity Vp (m/s) 0.2 7000 0 3000 Vp/Vs Vp (m/s) 2 7000 1.3 3000 porosity Vp (m/s) 0.2 7000 0 3000 Vp/Vs Vp (m/s) 2 7000 1.3 3000 porosity Vp (m/s) 0.2 7000 0 3000 Vp/Vs Vp (m/s) 2 7000 1.3 3000 porosity Vp (m/s) 0.2 7000 0 3000 Vp/Vs Vp (m/s) 2 7000 1.3 3000 porosity Vp (m/s) 0.2 7000 0 3000 Vp/Vs Vp (m/s) 2 7000 1.3 3000 porosity Vp (m/s) 0.2 7000 0 3000 Vp/Vs Vp (m/s) 2 7000 1.3 3000 porosity Vp (m/s) 0.2 7000 0 3000 Vp/Vs Vp (m/s) 2 7000 1.3 3000 porosity Vp (m/s) 0.2 7000 0 3000 Vp/Vs Vp (m/s) 2 7000 1.3 3000 porosity Vp (m/s) 0.2 7000 0 3000 Vp/Vs Vp (m/s) 2 7000 1.3 3000 porosity Vp (m/s) 0.2 7000 0 3000 Vp/Vs Vp (m/s) 2 7000 1.3 3000 porosity Vp (m/s) 0.2 7000 0 3000 Vp/Vs Vp (m/s) 2 7000 1.3 3000 porosity Vp (m/s) 0.2 7000 0 3000 Figure 2: Summary of dry velocities for all samples tested. Note that Vp exhibits a positive correlation with Vp /Vs and a negative correlation with porosity. (b) Final Report: DE-FG07-99ID13761 New England Research, Inc. 12 3100_1a Effective Pressure (MPa) Velocity (m/s) 60 7000 0 0 dry 3100_1a Effective Pressure (MPa) Velocity (m/s) 60 7000 0 0 saturated 3100_1a Effective Pressure (MPa) Velocity (m/s) 60 7000 0 0 3100_1a Effective Pressure (MPa) Velocity (m/s) 60 7000 0 0 Shear Compressional 4072_7a Effective Pressure (MPa) Velocity (m/s) 60 7000 0 0 dry 60 7000 0 0 brine 60 7000 0 0 60 7000 0 0 60 7000 0 0 60 7000 0 0 Shear Compressional 5091_1a Effective_Pressure (MPa) Velocity (m/s) 60 7000 0 0 dry 5091_1a Effective_Pressure (MPa) Velocity (m/s) 60 7000 0 0 saturated 5091_1a Effective_Pressure (MPa) Velocity (m/s) 60 7000 0 0 5091_1a Effective_Pressure (MPa) Velocity (m/s) 60 7000 0 0 Shear Compressional Figure 3: Ultrasonic velocity data plotted as a function of effective confining pressure for dry and brine saturated samples from AWI 1-2. (c) Figure 6.3: Seismic velocities for various rock samples versus (a) depth for dry samples, (b) porosity for dry samples, (c) effective confining pressure for dry and brine saturated rocks (Boitnott, 2003). Furthermore, effective medium theories 3 would suggest that lowV P andV S indicate highly fractured regions, whereas high V P and V S may indicate unfractured regions (Berge et al., 2001). Increasing depth may close fractures and cracks which can increase seismic velocities–Figure 6.3a indicates no clear separation between lithologies and no 3 Physical models that describe the macroscopic properties of a medium based on the properties and relative fractions of its components. 72 Table 6.1: Effect of saturation on velocities (Boitnott, 2003). Mechanism Description V P V S V P =V S FluidDensity Saturation causes a increase in bulk density, resulting in a decrease in velocities decreases decreases No Change FluidCompressibility Adding an incompressible fluid stiffens the pores to compression, increasing the Bulk modulus of the rock increases No Change increases ShearWeakening The presence of liquid water acts to weaken the Shear modulus of the rock. decreases decreases increases apparent signature with depth in core analysis where there is no overburden pressure to close the fractures. In addition, Charlety et al. (2006) reported that the porosity created from microcracks increases fluid saturation and raises the pore pressure. They also reported a slow increase in seismic velocity with time, which related to the cooling of the rock by the injected colder fluid and the increase in water saturation; but, they observed an overall reduction in the P-wave velocity that resulted from creating microcracks in the reservoir. Although the correlation to other properties is not general, Figure 6.3a indicates that the compressional and shear velocities strongly correlate to each another. Figure 6.3b also demonstrates thatV P andVp=Vs has a positive correlation. In summary 4 , the closing of small cracks due to pressure increase with depth, increasing overburden pressure, and cementation are the main causes of seismic veloc- ity increases. Fracturing, chemical alteration, extreme temperature gradient with depth, pore pressure, and increased porosity are the main causes of decreases in seismic veloc- ity. Fluid saturation has different results, but overall has little effect on V P , reduces V S , and enhances V P =V S . Effective medium theories suggest that low V P and V S are 4 based on the literature, laboratory measurement of The Geysers’ core samples performed by Boitnott (2003), and effective medium theories 73 consistent with highly fractured volumes, whereas highV P andV S are consistent with unfractured regions (Berge et al., 2001). 6.2 Stress and Rock Property Profiling Having both compressional and shear wave velocity models makes it possible to define most of the elastic rock properties uniquely (Tokosoz and Johnson, 1981). In particular, we calculate the effective normal stress , confining stress, Poisson’s ratio, Bulk modulus, Young’s modulus, and Shear modulus volume to characterize the fracture network in unconventional reservoirs in this step of workflow. 6.2.1 Definitions: Stress and Elasticity Elasticity is the property of a substance causing it to resist and recover from deformation produced by forces. The earth is considered almost elastic for small displacements. In this section, we use the theory of elastic waveforms in elastic media. Stress is a force applied to the unit area of material. It can be compressional, exten- sional, or shear (Figure .6.4). Stress, strain, Youngs modulus (Figure 6.5a), Bulk mod- ulus (Figure 6.5b), and Shear modulus (Figure 6.5c) are defined through the following equations: S = F A =E" (6.1) " = Change in dimension Original dimension (6.2) B = volume stress volume strain (6.3) 74 E = tensile or compressive stress tensile or compressive strain (6.4) = shear stress shear strain (6.5) Where S is stress, F is force, A is area," is strain, E is Young’s modulus, B is Bulk modulus and is Shear modulus. Measurement of the geometric change in shape due to normal stress could be considered Poissons ratio() (Figure 6.5d). It is defined as the ratio of relative change in diameter (" y = D=D) to relative compression or elongation (" x = L=L). = D=D L=L (6.6) Figure 6.4: Compressional, extensional, and shear stress in rock (Image courtesy of Michael Kimberly, North Carolina State Univ.) 75 (a) (b) (c) (d) (e) Figure 6.5: Cartoons describing (a) Bulk modulus, (b) Young modulus, (c) Shear modulus (Nave, 2010), (d) Poisson’s ratio (Christopher et al., 2006), (e) confining stress(Cramez, 2006)) 6.2.2 Stress and Rock Properties Derivation Using Seismic Velocity Volumes Compressional waves (primary or P-waves) propagate by alternating compression and dilation (Figure 6.6a) in the direction of the waves. Shear waves (secondary or S-waves) propagate through a sinusoidal pure shear strain (Figure 6.6b) in a direction perpendicu- lar to the direction of the waves. These velocities are approximately related to the square root of rock elastic properties and inversely related to its inertial properties. V (approx:) = s elastic property inertial property (6.7) For instance, in rock materials,V P andV S are defined in equation 6.8 and 6.9. 76 V P = s B + 4 3 (6.8) V S = r (6.9) where is density (an inertial property), B is Bulk modulus, and is Shear modulus. Then, effective normal stress ( n ), confining stress( h ), Poisson’s ratio(), Bulk modu- lus, Young’s modulus(E), and Shear modulus can theoretically be described in term of these seismic wave velocities 5 (Tokosoz and Johnson, 1981) 2 n = V 2 S (3V 2 P 4V 2 S ) (V 2 P V 2 S ) (6.10) 2 h =V 2 P 4 3 V 2 S (6.11) = V 2 P 2V 2 S 2 (V 2 P V 2 S ) (6.12) B =V 2 P 4 3 (6.13) =V 2 S (6.14) E = 2V 2 S ( + 1) (6.15) 5 Modulus are inGPa = 1:4510 5 psi and velocities are in km s . 77 It should be noted that the effects of wave propagation and attenuation on esti- mated properties are a function of the medium and vary from sandstones to carbon- ates to metamorphic rocks. Although (Tokosoz and Johnson, 1981) developed these relations primarily based on sedimentary depositional environments, the findings may be applicable to fractured metamorphic and crystalline rock formations such as The Geysers field. Variation of acoustic properties, such as velocities in fractured and hard crystalline rocks, has been studied before (Simmons, 1964; Nur and Simmons, 1969; Toks¨ oz et al., 1976). One possibility for explaining elastic wave propagation from the presence of particular propagation media is to look at the combination of Biot and Squirt flow simultaneously (BISQ theory). Studies have shown the potential applicability of the same in dual porosity medium (Qi-Zhen et al., 2009) and fractured crystalline rocks (Miyata et al., 2006). Elastic property estimates observed within this manuscript relate to relationships expected of any elastic medium of wave propagation. Several studies have been carried out which the same elastic moduli relationships used in this manuscript have been used to characterize changes in the properties of different rock types. Bina (1993) demon- strated how the elastic property variations can be used to characterize contrasts observed in upper mantle ( 400km depth). Sahimi and Arbabi (1992) used the same type of esti- mate to interpret the behavior of fractures in any disordered solids and granular media. Just as the velocities that have been predicted at any discrete location within the study area are an averaged estimate of a predefined volume (based on inversion param- eters), the elastic properties that are estimated can also be considered averaged values of said grid volumes. Fractures and variations in microstructures that might be expected within sandstone reservoirs and=or fractured metamorphic or crystalline deposits have an impact on the computed elastic properties due to variations in said structure and to the presence of different fluid types within the pore space. Theoretical models on the effect 78 of fractures–including their anisotropy–have been studied over the years and are avail- able, including many studies on effective medium theories (EMT), which deal fractures in non-porous media. However, due to the absence of good estimates of fracture prop- erties and anisotropy within the study area, using complicated theoretical models that map such affects is not possible. For a detailed discussion of elastic wave propagation in cracked media, refer to a recent monograph titled Mechanicsof CrustalRocks; 2012 (Leroy and Lehner, 2012). Therefore, within the confines of inaccuracies due to inversion schemes, data quality, effect of rock fabric anisotropy, fractures, and so forth, the elastic properties should be usable for qualitative analysis and interpretations made in this manuscript. Future work should look into more complex models based on data availability and some work needs to be conducted to find the optimal model and the specific properties required to apply said EMT models. The oscillating uniaxial strain involved in the case of a confined body, means that the axial modulus () con- trols the velocity of propagation, thus: (1.1) Shear bodywave waves, termed secondary, transverse or S-waves propagate by a sinusoidal pure shear strain (Figure 1.1 b) in a direction perpendicular to the direc- tion of the waves. The shear modulus ( ), which is given by the ratio of shear stress ( ) divided by the shear strain (tan ), will therefore control the (lower) velocity of propagation, thus: (1.2) The third important elastic modulus influencing the conversion between dynamic properties is the bulk modu- lus (K), defined as the ratio of the volumetric stress (P) and the volumetric strain (v/v). Since the three mod- uli are linked by the equation K 4/3 , it follows that V p can also be expressed as: (1.3) This equation therefore demonstrates the fundamen- tally faster nature of V p in relation to V s . The ratio of these two dynamic properties are also linked by the dynamic Poisson’s ratio for the material, as will be shown in the next section, which contains some standard equations. 1.2.1 Some sources of reduced elastic moduli In the case of micro-cracked, fractured, or jointed rock masses, there is a correspondingly reduced set of moduli in relation to the undisputed elastic nature of the intact matrix, because of micro (and probably elastic) displace- ments in normal and/or shear directions across and/ or along the micro-cracks, fractures or joints. These repre- sent an important part of the source of attenuation of the seismic waves in the dry state, due both to various scales of wave scattering and due to the intrinsic micro- deformations. Added losses are incurred if these micro- or macro-discontinuities are partly saturated, since there is communication with the pores and eventual pore fluid, and minute flows may be initiated to equilibrate pressures. These micro-imbalances will only be equili- brated when the frequency is sufficiently low. The above mechanisms mean that dynamic proper- ties, such as the velocities, Poisson’s ratio and attenuation tend in practice to be dispersive, or frequency depend- ent. They are also of course rock quality and environment- dependent, in the broadest possible meanings of these words. As rock quality declines, or the surface is approached, there develops a serious discrepancy between the dynamic or elastic properties of the intact matrix and the dynamic properties of the (partly discontinu- ous) medium. The ratio between the dynamic proper- ties of the (partly discontinuous) medium and the static deformation properties, such as the (rock mechanics) deformation moduli and joint stiffnesses (the inverse of compliances), may rise into double figures in this V K43 p 1 2 /         V s 1 2            V p 1 2            Shallow seismic refraction, some basic theory, and the importance of rock type 5 Figure 1.1 Elastic deformations and particle motions associated with the propagation of body waves: a) P-wave, b) S-wave. Based on Bott, 1982. Figure 6.6: Elastic deformations and particle motions associated with the propagation of seismic waves: (a) P-wave, (b) S-wave (Barton, 2006) 79 6.3 Effect of Geologic Parameters and Uncertainties To define the elastic moduli in the medium, we integrate P-wave and S-wave velocity volumes with each other and with density volumes created from lithology logs (Equa- tions 6.13 to 6.15). Figure 6.7 shows the created 3D density volume at the NW GGF. We identify five major rock types in this area which are color coded, then assign constant density to each type of rock. Notably, There are two major uncertainties in this assump- tion: First, each rock type has a range of recorded density with respect to its location and condition; for instance, the grain densities of the graywacke have a median value of 2.71 and a weighted average of 2.72, with a standard deviation of 2.5 percent. Sec- ond, geologic features–such as intrusion of bottom formation into the top– create higher degree of uncertainty in locating different rock types. The latter enables the presence of other rock types in the known formation that was detected from limited lithology logs. Because of these uncertainties, we carry out statistical simulation runs to generate an adequately large number of density realizations to test the effect of the variability of density in the computation of elastic properties that make use of density values as per defined relations in Equations 6.13 to 6.15. For this purpose, we test our methodology on the horizon located at 1.2 km below sea level 6 . This depth corresponds to the location of the normal temperature reservoir (NTR) where most of injection and production wells have steam entries. There are two rock types in the NTR: crystalline ”felsite” with a density of about 2.63 g/cc, and metamorphic graywacke, with a density around 2.7 g/cc. According to the available lithology logs, felsite density is only in the southern portion of our study area. The uncertainties in density estimates made by using lithology logs are deemed insignificant in the horizon of interest due to the high presence of graywacke formation, which is known to have a nearly constant density of around 2.7 gr cm 3 (Satik 6 Approximately 2.2 km depth from surface 80 et al., 1996; Gunderson, 1992). This horizon is affected by felsite intrusion; hence, lithology variations, including crystalline felsite of lower density with the metamorphic graywacke is the main reason for density uncertainty in this horizon. We use Sequential Gaussian Simulation to generate 1000 density realizations for our study area. Figure 6.8 shows 4 sample density realizations for a specific horizon of interest as well as Bulk modulus at the same horizon based on the accompanying density realization. From this figure–and similar results from other density-dependent properties–we observe that substantial changes within the range of interest in density values over the spread of the horizons do not create major changes in the actual elastic moduli values. Moreover, even with substantial variations in density, qualitative inter- pretations from the results should be reasonably accurate. Figure 6.9 shows the nor- malized absolute standard deviations of three properties (density, Poisson’s ratio, and Bulk modulus) at different evaluation points. We submit that the variations observed with the target properties are in the same order of magnitude as those observed with density variability. This finding indicates that errors due to density variations should not have a substantial impact on the final property maps for qualitative–and some degree of quantitative–evaluation due to the lower impact of density (degree of 1) compared to phase velocities (degree of 2) as well as relatively small density variability (2:1%). To further validate the relatively small property variation as a result of varying density maps, we have computed difference maps for the mapped properties; the results for Bulk modulus are shown in Figure 6.10. We can observe from this figure that as expected, the difference maps follows the same trend as the density realization (because the second density used in computation is a constant baseline map). Also, the maximum difference observed for the all of properties stays below5%. Thus using constant density for a major portion of the horizon of interest should allow us to make a qualitative interpreta- tion. 81 Figure 6.7: Created density volume based on available lithology logs at the NW Geysers 82 −2699 −2698 −2697 −2696 −2695 −2694 −2693 −4185 −4184 −4183 −4182 −4181 X(km) Y(km) DENSITY 2.67 2.68 2.69 2.7 2.71 2.72 2.73 2.74 −2699 −2698 −2697 −2696 −2695 −2694 −2693 −4185 −4184 −4183 −4182 −4181 X(km) Y(km) BULK MODULUS 10 15 20 25 30 35 40 45 50 −2699 −2698 −2697 −2696 −2695 −2694 −2693 −4185 −4184 −4183 −4182 −4181 X(km) Y(km) DENSITY 2.67 2.68 2.69 2.7 2.71 2.72 2.73 2.74 −2699 −2698 −2697 −2696 −2695 −2694 −2693 −4185 −4184 −4183 −4182 −4181 X(km) Y(km) BULK MODULUS 10 15 20 25 30 35 40 45 50 −2699 −2698 −2697 −2696 −2695 −2694 −2693 −4185 −4184 −4183 −4182 −4181 X(km) Y(km) DENSITY 2.67 2.68 2.69 2.7 2.71 2.72 2.73 2.74 −2699 −2698 −2697 −2696 −2695 −2694 −2693 −4185 −4184 −4183 −4182 −4181 X(km) Y(km) BULK MODULUS 10 15 20 25 30 35 40 45 50 −2699 −2698 −2697 −2696 −2695 −2694 −2693 −4185 −4184 −4183 −4182 −4181 X(km) Y(km) DENSITY 2.67 2.68 2.69 2.7 2.71 2.72 2.73 2.74 −2699 −2698 −2697 −2696 −2695 −2694 −2693 −4185 −4184 −4183 −4182 −4181 X(km) Y(km) BULK MODULUS 10 15 20 25 30 35 40 45 50 Figure 6.8: Density and Bulk Modulus for 4 density realizations at the NTR 83 0 500 1000 1500 2000 2500 3000 3500 0.002 0.004 0.006 0.008 0.01 0.012 0.014 0.016 0.018 Evaluation Point Normalized Absolute Standard Deviation Z = 25 POISSONS RATIO BULK MODULUS DENSITY Figure 6.9: Normalized absolute standard deviations along every evaluation point on selected horizons for all density realizations indicating small variability of target prop- erties 84 6.4 Characterizing the Fracture Network Using Changes in Stress and Rock Properties Microseismic events release seismic energy from shear or mixed mode failure of rocks along pre-existing planes of weakness. Therefore, understanding this failure process can help us characterize the created fracture network from stimulation stages. 6.4.1 Stress and Fracture Figure 6.11a illustrates the behavior of a fracture under a two-dimensional stress con- dition. Three types of stresses typically act on the fracture surface; they are referred to as normal stress ( n ), shear stress, and static frictional stress. Each of these states of stress can cause an associated fracture mode, as shown in Figure 6.11b. Figure 6.11a also shows the state before pore pressure (p) increase caused by an injection from stim- ulation in hydraulic fracturing or initial contact of water with new planes of weakness. Most induced fractures in tight reservoirs are either created through an hydraulic fracturing process, such as shale reservoir, or through a thermal contraction process, such as The Geysers. In the hydraulic fracturing process, pressurized fluid injection reduces the effective normal stress and raises the pore pressure, which reduces the peak stress that the interface can support. Hence, at some level of pore pressure increase, shear slip may occur once the peak frictional stress becomes smaller than the shear stress. In addition, reduced effective stress may cause fracture dilation. Tezuka (2000); Ameen (2003) also reported that shear slippage along pre-existing fractures or flaws caused by pore pressure increase can create fluid flow pathways for production. At The Geysers, water is not pumped into the reservoir under pressure, but ”free falls” down the well bore to the reservoir level. Limited reservoir pressure results from the development of a standing water column near the base of the well, but is generally insufficient to 85 initiate fracturing. Stark (1990) reported that most seismicity in The Geysers geothermal field is related to cold water injection due to a thermal contraction process. The cold fluid injectate interacts with hot rock causing contraction at and near fracture surfaces. Rutqvist et al. (2006) reported that evaporation cooling with contraction reduces both effective stress, and static friction, triggers slip along planes of weakness, and results in the slight opening of fracture. In both scenarios, we can use the effective normal stress as an index for fracture opening 7 , which is essential to fluid movement and production. Figure 6.12 illustrates this phenomenon in a hydraulic fracturing job. Sammonds et al. (1989) reported that fracture patterns are controlled by confining stress that impacts on closure of pre-existing fractures and the growth of new microc- racks. We might also use the confining stress as an indicator of pore pressure. Tezuka (2000); Ameen (2003) also reported that shear slippage along pre-existing fractures or flaws caused by pore pressure increase can create fluid flow pathways for produc- tion. They hypothesized that because shear failure fractures have the associated condi- tion whereby Coulomb failure (Figure 6.13) is reached with the smallest pore pressure increase, these types of fractures mainly contribute to fluid production and to the distri- bution of favorably oriented pre-existing fractures and their interaction with the regional stress field which strongly controls the growth direction of the reservoir (Figure 6.14). In summary, we can use confining stress–and especially normal stress– distribution to identify and delineate fractured areas from unfractured ones. 6.4.2 Elastic Moduli and Fracture Furthermore, Figure 6.15 demonstrates the apparent signature of saturation on bulk and Shear modulus. Saturation effects include an increase in Bulk modulus and a reduc- tion in Shear modulus. Shear modulus reduces due to rock-water interaction. The low 7 the fracture walls become completely separated when the effective normal stress becomes zero 86 frequency Biot-Gassmann theory (Biot, 1956; Boitnott, 2003) explains the changes in Shear modulus as due to void geometry with resolution control of the stress field at the void space scale 8 (Berryman et al., 2002; Berryman and Wang, 2000; Berryman, 1999). Berge et al. (2001) used the same theory to predict the dry Bulk modulus for rocks with different void ratios. As shown in Figure 6.16, increasing the void ratio significantly reduces the Bulk modulus especially for thin cracks or fractures. Moreover, Barton (2006) stated that a set of moduli decreased because of micro displacements of the frac- tures. In partly saturated fractures, this reduction is more significant. He also reported that Young’s modulus in fractured rock decreased exponentially with fracture density. 6.5 Enhancing the Resolution of Seismic Velocity Mod- els For estimating various properties from the velocity models extracted from tomographic inversion (section 2.3), integrating them with other data types, or finding the possible temporal changes in the velocity models, finer grid size is desirable. Figure 2.5 shows the tomographic inversion output from microseismic data for the compressional veloc- ity model, which has 4950 points and spacing of about 600m covering the Northwest Geysers study area; hence, we are obligated to find a way to enhance the resolution of these models. In general, the resolution of the 3-D tomography can increases with the number of receiver arrays or seismic stations. Because the number of raypaths between the hypocenter and the seismic receivers that cross through each volume of a designated 3-D grid increases with the number of receivers, the volumes within the 3-D grid reduce and enhance the resolution (Kenedi et al., 2010). But, with a fixed number of receivers 8 Poisson’s effect 87 and created velocity models, we recommend using kriging 9 to generate high resolution velocity models. Kriging allows us to have the velocity models and locations that are reliable for characterizing the unconventional reservoirs where sufficient seismicity is available. We begin with the initial velocity model provided by the LBNL (Boyle et al., 2011) 10 . Figure 6.17a shows the initial compressional velocity model with 4950 points and spacing of about 600m. We perform the kriging analysis in SGeMS and Gslib environ- ment. To create the high resolution velocity models, we calculate experimental vari- ogram values for different lag separations from 500m to 10km from the output of tomo- graphic inversion software such as SimulPS 11 . Then, we use the Gaussian model with a practical range of 57 and sill of 0.4 to fit the data (equation 6.16) for kriging analysis; see Figure 6.17b. We implement the ordinary kriging on 1395360 points in Cartesian grid. Figure 6.17d shows the final compressional velocity model on a fine grid size volume with spacing of less than 100m. In other words, we introduce 212 new points of data for any 4 actual tomographic inversion outputs; see Figure 6.17c. We apply the same procedure for the shear wave velocity model. Although the created velocity models have inherent uncertainty, they could be a powerful tool in characterizing the unconventional reservoir. (h) = 0:4[1exp( 3h 2 a 2 )] (6.16) 9 A geostatistical tool for estimating the value of missing points (velocity values in fine grid mesh) from the known values (the initial velocity field). 10 This model is output of SimulPS and preliminary one based on 2004 events 11 http://faldersons.net/Software/Simulps/Simulps.html 88 6.6 Fracture Network Interpretation through High Res- olution Velocity Models Based on a smoothed velocity model, Figure 6.18 shows the injection wells and corre- sponding velocity anomalies. Figure 6.19 also shows that the lateral extension of veloc- ity anomalies below injection wells increases up to the middle of a normal temperature reservoir, where most wells are completed, and then decreases by depth. The same kind of anomalies could be seen in the high temperature zone. According to section 6.1, an increase in porosity created from fractures is the main cause for depreciation in the com- pressional velocity of the zone of interest. Hence, low velocity anomalies may correlate with fracture network densities and spacing in the system. Thus, changes in velocities may serve as other indicators of the evolution of fracture network in high temperature zone. We observed a reduction in velocity and a growth of the region, which are indicative of this decrease in the reservoir depth (Figure 6.19). In the deeper regions, velocity anomalies tend to diminish slightly. They can be related to closing fractures with depth or to reduction in number of cracks or void ratio with depth 12 . These observations lead to an expected distribution of the velocity with depth at The Geysers. The area where we have both lowV P andV S indicates highly fractured regions, whereas highV P and V S may delineate unfractured regions. We consider two different horizons in The GGF to deeper investigate our methodol- ogy. The first horizon is located in the normal temperature reservoir (NTR), where the injection and production wells have been completed ;see Figure 2.4. The existing frac- ture network within this zone has the main role in the production of steam. Our aim in this horizon is locating and characterizing the fracture network. The second horizon is 12 This findings is consistent with observation made by Berge et al. (2001) at The Geysers 89 the area 1500 ft below the depth of the deepest wells and within high temperature zone (HTZ), where there is little or no production 13 . Our goal here is to identify zones in which the fracture network propagates in order to create an enhanced geothermal system (EGS). The same method might be applied to an unconventional reservoir in which both seismic velocity models are available or created from microseismic data using tomo- graphic inversion. Figure 6.20 shows V P , V S , and V P =V S on these two horizons in the NW Geysers. HighV P =V S anomalies associated with lowV S anomalies may be saturation anomalies. On the other hand, highV P =V S anomalies associated with highV P are caused by other phenomena such as lithology effects. Finally, it is reasonable to assume that lowV P =V S anomalies associated with lowV P anomalies are fracture related anomalies. Based on the identified framework, we can identify highly fractured zones of interest by interpret- ing the observed velocity anomalies. As per our discussion, we interpret the area below and around the SB27 and DX23 wells as having the highest fracture density. In contrast, the area below and around LF2 and CMHC2 is interpreted to have the lowest fracture density within this horizon. Moreover, we can successfully locate the propagated frac- ture network in the high temperature zone below the SB27 and DX77 wells. Figure 6.22 clearly shows that Poisson’s ratio anomalies follow a similar trend when compared withV P =V S anomalies shown in Figure 6.20. This is a good indicator of fluid saturation. Reduction in normal stress indicates the areas in which fractures are open to provide sufficient permeability along with low velocity anomalies, which acts as further validation that the identified anomalies are fracture-related and not from lithology or other phenomena. Boitnott (2003) also showed laboratory measurement for theV P andV S results from The Geysers which are plotted against void ratio in Figure 6.21. For The Geysers,V P 13 A reasonable assumption is that we have dominantly graywacke in NTR and Felsite in HTZ 90 increased, whereasV S remained unchanged or decreased slightly with saturation (Berge et al., 2001). Hence, although we have enough saturation below injection wells, the only factor that explains the low velocity anomalies is fractures. Furthermore, Berge et al. (2001) used medium theories to investigate the effects of fractures on velocities for The Geysers rocks. They also clearly reported that the regions with lowV p are significantly fractured. In summary, after joint interpretation of velocity, Poissons ratio, normal stress, and confining stress, we could identify the NW trend of the fracture network and the zones in which a denser fracture network exists in the NTR. In the HTZ, it is also possible to locate the penetrating fracture network–the area in which change inV P , confining stress, and normal stress are remarkable. 6.7 Tomographic Inversion Versus Fuzzy Clustering We also sought to accurately locate the boundaries of the connected fracture network; thus, hypocenters were analyzed to examine possible correlation between microseismic events and fracture network. Fuzzy clustering was used to investigate the movement of microseismic events in the high temperature zone, and velocity models were created to find the fracture related anomalies. The cluster centers were overlaid on the kriged compressional velocity models to validate the results obtained from the two methods. We have explored the relationship between microseismic cluster movement and tem- poral change in the compressional velocity model at The GGF. Figure 6.23 shows that the identified anomalies from the velocity models correlate with microseismic event clusters and their movement. Fracture propagation or fluid movement within the frac- ture network may be identified by considering the contribution of the fractured area to velocity model anomalies and the movement of microseismic fuzzy cluster centers. 91 The successful integration of fuzzy clustering and velocity modeling can help us clarify our hypothesis about velocity anomalies or fuzzy cluster movement. This elimi- nates the errors that may arise in locating the fractured areas and may help us target the stimulated area for development plans. 6.8 Tomographic Inversion Versus Production/Injection Data Table 6.2 and 6.3 shows production / injection rate for the wells in this area and their clustering definition, respectively. Figure 6.24 validates that regions with higher frac- ture density anomalies (such as low V P and low normal stress as discussed) correlate with higher steam production with relatively low water injection levels. This finding is consistent with our hypothesis and theory from the preceding discussions. On the other hand, lower Poisson ratio or lowerV P =V S are not self-indicative of higher fracture intensity, and these anomalies could be associated with other phenomena such as degree of fluid saturation or fluid type. 6.9 Time-lapse Fracture Characterization Using Tomo- graphic Inversion Monitoring the change in velocity models for a series of equivalent time periods may help determine the evolution of a fracture network growth. The 2005 to 2010 compres- sional (P) and shear (S) wave seismic velocity volumes, provided by Lawrence Berkeley National Laboratory at The GGF, were fed into the workflow presented in this chapter to monitor changes in the reservoir. As discussed, we derived various properties of the 92 reservoir by integrating these volumes and available lithology logs. This study points out the changes in the reservoir properties during steam production. We discuss the stress variation and rock properties within the NTR. For each year, the associated aver- age daily production and injection rates were plotted to assist in the correlation of fluid movement with derived volumes. This analysis has allowed us to determine regions with potentially higher fracture densities and may provide better understanding of the effects of water injection on reservoir properties. Results from studies within the overall framework of the workflow may help operators optimize injection schemes, improve the production profiles, and locate potential zones of interest for future development. Figure 6.25 presents the temporal evolution of the compressional wave seismic velocity distribution at the horizon of interest. An interesting feature is observed between these models; the low velocity anomalies (represented as blue and green areas), expands through the middle of the horizon of interest ((x,y)=[-2696.36 km -4183 km]) through time, indicating an evolution of the fracture network in this area. In addition, the tails of low velocity anomalies in the NW part of this horizon expand through this period of time before injection starts at CA-87A-2, and a relatively unfractured zone connects with the fracture network. The figures also indicate that regions perceived to have higher fracture density anomalies (low V P ) correlate with higher steam produc- tion and relatively low water injection rates. Other phenomena observed in 2010 are increasing velocities in the area between DX-77 and DX-38, possibly due to the closing of fractures. The shear wave seismic velocity distribution displayed in Figure 6.26 may also clarify the latter behavior. Moreover, areas with both low V P and low V S have a higher possibility of being a result of fractures rather than of other possible effects. 93 −2699 −2698 −2697 −2696 −2695 −2694 −2693 −4185 −4184 −4183 −4182 −4181 X(km) Y(km) DENSITY 2.67 2.68 2.69 2.7 2.71 2.72 2.73 2.74 −2699 −2698 −2697 −2696 −2695 −2694 −2693 −4185 −4184 −4183 −4182 −4181 X(km) Y(km) DELTA PLOT − BULK MODULUS 0.5 1 1.5 2 2.5 3 −2699 −2698 −2697 −2696 −2695 −2694 −2693 −4185 −4184 −4183 −4182 −4181 X(km) Y(km) DENSITY 2.64 2.66 2.68 2.7 2.72 −2699 −2698 −2697 −2696 −2695 −2694 −2693 −4185 −4184 −4183 −4182 −4181 X(km) Y(km) DELTA PLOT − BULK MODULUS 0.5 1 1.5 2 2.5 3 3.5 4 −2699 −2698 −2697 −2696 −2695 −2694 −2693 −4185 −4184 −4183 −4182 −4181 X(km) Y(km) DENSITY 2.69 2.7 2.71 2.72 2.73 2.74 −2699 −2698 −2697 −2696 −2695 −2694 −2693 −4185 −4184 −4183 −4182 −4181 X(km) Y(km) DELTA PLOT − BULK MODULUS 0.5 1 1.5 2 2.5 −2699 −2698 −2697 −2696 −2695 −2694 −2693 −4185 −4184 −4183 −4182 −4181 X(km) Y(km) DENSITY 2.67 2.68 2.69 2.7 2.71 2.72 2.73 2.74 −2699 −2698 −2697 −2696 −2695 −2694 −2693 −4185 −4184 −4183 −4182 −4181 X(km) Y(km) DELTA PLOT − BULK MODULUS 0.5 1 1.5 2 2.5 3 Figure 6.10: Difference maps (percentage difference) for various Bulk modulus realiza- tions. The baseline density used for computation is a constant value from the horizon of interest. We observe the percentage difference to be< 5% 94 (a) Normal Shear Frictional (b) Figure 6.11: (a) Diagram illustrating the fracture subjected to dimensional stress condi- tion (Tezuka, 2000), (b) Associated fracture mode to different states of stress (Rountree et al., 2002) Natural Fracture Stimulation Figure 6.12: Fracture opening for different normal stresses in a hydraulic fracturing job (Dershowitz and Doe, 2011) 95 CRITICAL STRESS-RELATED PERMEABILITY Fig. 1. Three-dimensional Mohr diagram showing shear and normal stress on a fracture surface as a function of the in situ stress field. Adapted from Jaeger & Cook (1984). The dark triangles represent fractures that are experiencing a state of stress sufficient to induce shear failure whereas the dots represent fractures in a more stable state of stress. of 0.6 is used after the work of Byerlee (1978), who compiled a wealth of shear test data and found a remarkably consistent pattern. Fractures with an intermediate and above confining stress (greater than 200 MPa) are best described by a coefficient of friction of 0.6. However, the applic- ability of this value to most fractured reservoirs is limited given that the 200 MPa threshold is in the range of 5-6 km depth. Byerlee showed that for fractures with a normal stress less than 200 MPa, a value of m=0.85 better defined the failure state although there was a much greater scatter in the data. Intuitively the higher coefficient should be used as this better fits the typical stress state for most reservoirs. However, there is considerable empirical evidence to indicate that the lower value should be used. It should be noted that the failure criterion lacks a value for cohesion or intrinsic shear strength. This is justified in that the magnitude of this value is very small compared to the stresses experienced by fracture surfaces. Barton et al. (1995) used the term 'critically stressed' fractures to refer to those fractures whose shear and normal stresses fall in excess of the 0.6 failure criterion and therefore exist in states likely to result in enhanced permeability. However, geology is seldom binary with the onset of a phenomenon happening when a threshold is crossed. It is considered that the coefficient of friction calculated for each fracture should be used as an indicator of the likelihood of failure and hence flow rather than simply applying a test for failure or non-failure. For a typical UK onshore stress field (strike slip stress regime), the ratio of shear to normal stress can be computed for all orientations. These data can be plotted stereographically as a polar plot, with the data contoured with respect to the stress ratio (see Fig. 2). This will then indicate those orientations most likely to experience enhanced permeability. Given a maximum hori- zontal stress direction of approximately 340°, it is clear from Figure 2 that there are two main strike orientations that cause the highest ratio of shear to normal stress. These correspond to fracture and fault surfaces striking in two distinct directions: NW–SW and also almost north-south. So in contrast to the empirical evidence of Heffer & Lean (1993), which suggests that flow is greatest in the direction of ohmax, given this particular stress tensor the optimum fracture orientations for flow would be approximately o - hmax±30°. Case study: Sellafield RCF3 borehole To test the hypothesis of shear-stress-dependent flow being a significant controlling mechanism 9 τ σ τ σ Φ, friction angle Φ Increase of overall stress Increase of differential stress (a) (b) Figure 2. Two kinds of stress increase used in this study. (a) Increase of stress magnitudes while keeping the stress ratio of horizontal to vertical stress constant. (b) Increase of differential stresses while keeping the vertical, i.e. the lowest stress component, stress constant. The Mohr-Coulomb failure criterion is used with the zero cohesion. 40 Figure 6.13: (a) 3-dimensional Mohr-Coulomb diagram showing shear and normal stress on a fracture surface as a function of insitu stress field. The dark triangles rep- resent fractures that are experiencing a state of stress sufficient to inducing shear failure whereas the dots represent fractures in a more stable state of stress (Ameen, 2003) (b) Change of differential stress during hydraulic fracturing job to reach failure mode and create fracture 10 STEPHEN F. ROGERS Fig. 2. Coefficient of friction as a function of orientation, displayed stereographically as a lower hemisphere polar projection. for flow through fractured reservoirs, detailed fracture, stress and flow data are required. Borehole RCF3 was drilled as part of the UK radioactive waste disposal programme through 525 m of Permo-Triassic sandstone and breccia and approximately 450 m of Borrowdale Volcanic Group basement rocks. These rocks are a thick pile of mixed intrusive and extrusive volcanics and volcaniclastics with generally little primary porosity. The vast majority of flow in these rocks is through the complex fracture and vuggy vein system. In addition to the almost complete coring and full wireline logging of this borehole, 100 short interval tests (SITs) were carried out over zones 1.56 m long, using Schlumberger's wireline conveyed testing string, the Modular Dynamics Tester™ (Gutmanis et al. 1998). This provides a detailed continuous record of borehole transmissivity over a total distance of approximately 150m of fractured basement rocks. This hydraulic testing identified seven zones where the transmissivity was several orders of magnitude higher than the background level. If the critical stress theory is to account for the locations of flow within the borehole, then there needs to be a state of stress within these seven zones that is different from the borehole as a whole. Orientated borehole discontinuity information was available from the interpretation of the Fullbore Microlmager FMI™ and Ultrasonic Borehole Imager UBI™. Careful picking of all discontinuities in the test zone from 640 to 790 mbRT provided a continuous record of frac- turing in the borehole. All discontinuities were picked rather than just the larger ones, as it was the importance of orientation, rather than apparent aperture that was being tested. Having produced a data set of all discon- tinuity information from the test interval, the shear and normal stresses on every discontinuity were calculated. These results are shown in Figure 3a, with a Mohr diagram showing the ratio of shear to normal stress calculated for all discon- tinuities from this section. The distribution of these values is also displayed as a histogram alongside. To test whether there was a stress dependency to the origin of the zones where the transmissivity was significantly higher than the background, discontinuity data from just these zones were also analysed. Figure 3b shows a Mohr diagram with the shear and normal stresses for all discontinuities from the seven transmissive zones. Again, a histogram of the distribution of coefficients of friction for all of these fractures is displayed alongside. Comparison of the data for Figure 6.14: Coefficient of friction as a function of orientation, displayed stereographi- cally as a lower hemisphere polar projection (Ameen, 2003). 96 Final Report: DE-FG07-99ID13761 New England Research, Inc. 13 4072_7a Effective_Pressure (MPa) Modulus (GPa) 60 40 0 0 Dry Bulk 4072_7a Effective_Pressure (MPa) Modulus (GPa) 60 40 0 0 Saturated Bulk 4072_7a Effective_Pressure (MPa) Modulus (GPa) 60 40 0 0 Dry Shear 4072_7a Effective_Pressure (MPa) Modulus (GPa) 60 40 0 0 Saturated Shear 3100_1a Effective_Pressure (MPa) Modulus (GPa) 60 40 0 0 3100_1a Effective_Pressure (MPa) Modulus (GPa) 60 40 0 0 3100_1a Effective_Pressure (MPa) Modulus (GPa) 60 40 0 0 3100_1a Effective_Pressure (MPa) Modulus (GPa) 60 40 0 0 Figure 4: The effect of pressure and brine saturation on the inferred dynamic bulk and shear moduli. Saturation causes an in crease in the dynamic bulk modulus and a decrease in the dynamic shear modulus. Following previous work on core from The Geysers, the effect of saturation on the velocities of these and related cores is modeled using a modified low frequency Biot-Gassmann poroelastic theory. The required modification from traditional theory is the addition of a shear modulus weakening term which is found to be independent of stress. The model can be described by the following equations, which describe the effect of saturation on the elastic moduli of the matrix material. (1a) Κ sat = Κ dry + ΔΚ ; (1b) G sat = G dry + ΔG (1c) ΔK = (K soild –K dry ) 2 . [K solid (1-ϕ-(K dry /K solid ) + ϕ(K solid /K f ))] ΔK represents the increase in bulk modulus due to saturation as described by Biot's low frequency poroelastic theory, where ϕ is the porosity, K soild is the bulk modulus of the solid grains, K dry is the bulk modulus of the dry frame (i.e. the rock in its dry state), and K f is the bulk modulus of the saturating fluid. In this modified model, the shear modulus G is assumed to weaken with saturation by an amount ΔG. Using this model, we are lead to a physical understanding of the effect of saturation resulting from three basic mechanisms: fluid density, fluid compressibility, and shear weakening (see Table 1). Other mechanisms such as inertial coupling and local flow which are commonly observed in similar data on crystalline rocks, sandstones, and carbonates, cannot be ruled out, but do not appear necessary to explain the results. This likely results in part from the fine grained nature of the pore space in hydrothermally altered rock. Figure 6.15: Bulk and Shear modulus versus pressure and brine saturation for two dif- ferent core samples (Boitnott, 2003). Figure 3 Draft: Joint Inversion at the Geysers 3/26/2007 2:12:00 PM Page 15 o Figure 6.16: Laboratory measurement of crack effect on bulk moduli(Berge et al., 2001) 97 4950 Points (15*33)*10 Spacing ~ 600m (a) 0.00 0.00 37.98 0.25 75.97 0.50 113.95 0.76 151.93 1.01 189.92 1.26 227.90 1.51 distance (b) (c) 1395360 points (152*153)*60 Spacing~100m (d) Figure 6.17: Enhancing the resolution of seismic velocity models, (a) the initial velocity model, (b) variogram model based on initial velocity model,(c) actual output of tomo- graphic inversion (black dots) and new kriged data point (orange dots), (d) the final kriged velocity model. 98 1 0 -1 -2 MSL -3 -4 -5 1 0 -1 -2 MSL -3 -4 -5 Figure 6.18: Velocity anomalies around active injection wells, Inline view 99 1 0 -1 MSL -2 -3 -4 -5 Figure 6.19: Velocity anomalies below injection wells; the lateral extension of these anomalies increases up to the middle of a normal temperature reservoir–where most of wells are completed –and then decreases by depth 100 −4184000 −4182000 −4180000 −4184000 −4182000 −4180000 −2698000 −2696000 −2694000 −2698000 −2696000 −2694000 Compressional Velocity 2011 Dec 13 16:11:41 Compressional Velocity Scale 1:40000 CA−45A−12 CMHC−2 CMHC−6 DX−19 DX−24 DX−47 DX−5 GDC−26 GDC−53−13 GDC−88−12 LF−2 LF−23 OS−11 OS−12 OS−16 SB−15 CA32−12 CA45−12 CA48A−2 CA52−11 Curry3 DX23 DX25 DX28 DX3 DX38 DX42 DX59 DX77 GDC2 GDC6 OS15 OS19 OS23 SB20 SB27 4.50 4.75 5.00 5.25 H−1.2kmNw VP2004OK Injection Wells Production Wells (a) −4184000 −4182000 −4180000 −4184000 −4182000 −4180000 −2698000 −2696000 −2694000 −2698000 −2696000 −2694000 Shear Velocity 2011 Dec 13 16:50:59 Shear Velocity Scale 1:40000 CA−45A−12 CMHC−2 CMHC−6 DX−19 DX−24 DX−47 DX−5 GDC−26 GDC−53−13 GDC−88−12 LF−2 LF−23 OS−11 OS−12 OS−16 SB−15 CA32−12 CA45−12 CA48A−2 CA52−11 Curry3 DX23 DX25 DX28 DX3 DX38 DX42 DX59 DX77 GDC2 GDC6 OS15 OS19 OS23 SB20 SB27 2.7 2.8 2.9 3.0 H−1.2kmNw VS2004OK Injection Wells Production Wells (b) −4184000 −4182000 −4180000 −4184000 −4182000 −4180000 −2698000 −2696000 −2694000 −2698000 −2696000 −2694000 VP/VS 2011 Dec 13 16:58:54 VP/VS Scale 1:40000 CA−45A−12 CMHC−2 CMHC−6 DX−19 DX−24 DX−47 DX−5 GDC−26 GDC−53−13 GDC−88−12 LF−2 LF−23 OS−11 OS−12 OS−16 SB−15 CA32−12 CA45−12 CA48A−2 CA52−11 Curry3 DX23 DX25 DX28 DX3 DX38 DX42 DX59 DX77 GDC2 GDC6 OS15 OS19 OS23 SB20 SB27 1.6 1.7 1.8 H−1.2kmNw VP/VS Injection Wells Production Wells (c) −4184000 −4182000 −4180000 −4184000 −4182000 −4180000 −2698000 −2696000 −2694000 −2698000 −2696000 −2694000 Compressional Velocity 2011 Dec 15 09:22:32 Compressional Velocity Scale 1:40000 CA−45A−12 CMHC−2 CMHC−6 DX−19 DX−24 DX−47 DX−5 GDC−26 GDC−53−13 GDC−88−12 LF−2 LF−23 OS−11 OS−12 OS−16 SB−15 CA32−12 CA45−12 CA48A−2 CA52−11 Curry3 DX23 DX25 DX28 DX3 DX38 DX42 DX59 DX77 GDC2 GDC6 OS15 OS19 OS23 SB20 SB27 4.50 4.75 5.00 5.25 H−2.4kmNw VP2004OK Injection Wells Production Wells (d) −4184000 −4182000 −4180000 −4184000 −4182000 −4180000 −2698000 −2696000 −2694000 −2698000 −2696000 −2694000 Shear Velocity 2011 Dec 15 09:41:56 Shear Velocity Scale 1:40000 CA−45A−12 CMHC−2 CMHC−6 DX−19 DX−24 DX−47 DX−5 GDC−26 GDC−53−13 GDC−88−12 LF−2 LF−23 OS−11 OS−12 OS−16 SB−15 CA32−12 CA45−12 CA48A−2 CA52−11 Curry3 DX23 DX25 DX28 DX3 DX38 DX42 DX59 DX77 GDC2 GDC6 OS15 OS19 OS23 SB20 SB27 2.9 3.0 3.1 3.2 H−2.4kmNw VS2004OK Injection Wells Production Wells (e) −4184000 −4182000 −4180000 −4184000 −4182000 −4180000 −2698000 −2696000 −2694000 −2698000 −2696000 −2694000 VP/VS 2011 Dec 15 09:41:24 VP/VS Scale 1:40000 CA−45A−12 CMHC−2 CMHC−6 DX−19 DX−24 DX−47 DX−5 GDC−26 GDC−53−13 GDC−88−12 LF−2 LF−23 OS−11 OS−12 OS−16 SB−15 CA32−12 CA45−12 CA48A−2 CA52−11 Curry3 DX23 DX25 DX28 DX3 DX38 DX42 DX59 DX77 GDC2 GDC6 OS15 OS19 OS23 SB20 SB27 1.6 1.7 1.8 H−2.4kmNw VP/VS Injection Wells Production Wells (f) Figure 6.20: NTR horizon at The NW Geysers (a)V P , (b)V S , (c)V P =V S , HTZ horizon (d)V P , (e)V S , (f)V P =V S 101 Figure 2 Draft: Joint Inversion at the Geysers 3/26/2007 2:12:00 PM Page 14 o Figure 6.21: Laboratory measurement for The Geysers (Berge et al., 2001) 102 −4184000 −4182000 −4180000 −4184000 −4182000 −4180000 −2698000 −2696000 −2694000 −2698000 −2696000 −2694000 Poisson Ratio 2011 Dec 13 17:11:32 Poisson Ratio Scale 1:40000 CA−45A−12 CMHC−2 CMHC−6 DX−19 DX−24 DX−47 DX−5 GDC−26 GDC−53−13 GDC−88−12 LF−2 LF−23 OS−11 OS−12 OS−16 SB−15 CA32−12 CA45−12 CA48A−2 CA52−11 Curry3 DX23 DX25 DX28 DX3 DX38 DX42 DX59 DX77 GDC2 GDC6 OS15 OS19 OS23 SB20 SB27 0.175 0.200 0.225 0.250 0.275 H−1.2kmNw Poisson2004 Injection Wells Production Wells (a) −4184000 −4182000 −4180000 −4184000 −4182000 −4180000 −2698000 −2696000 −2694000 −2698000 −2696000 −2694000 Extensional Stress 2011 Dec 13 16:57:49 Extensional Stress Scale 1:40000 CA−45A−12 CMHC−2 CMHC−6 DX−19 DX−24 DX−47 DX−5 GDC−26 GDC−53−13 GDC−88−12 LF−2 LF−23 OS−11 OS−12 OS−16 SB−15 CA32−12 CA45−12 CA48A−2 CA52−11 Curry3 DX23 DX25 DX28 DX3 DX38 DX42 DX59 DX77 GDC2 GDC6 OS15 OS19 OS23 SB20 SB27 4.2 4.4 4.6 H−1.2kmNw VE2004 Injection Wells Production Wells (b) −4184000 −4182000 −4180000 −4184000 −4182000 −4180000 −2698000 −2696000 −2694000 −2698000 −2696000 −2694000 Hydrostatic Stress 2011 Dec 13 17:00:30 Hydrostatic Stress Scale 1:40000 CA−45A−12 CMHC−2 CMHC−6 DX−19 DX−24 DX−47 DX−5 GDC−26 GDC−53−13 GDC−88−12 LF−2 LF−23 OS−11 OS−12 OS−16 SB−15 CA32−12 CA45−12 CA48A−2 CA52−11 Curry3 DX23 DX25 DX28 DX3 DX38 DX42 DX59 DX77 GDC2 GDC6 OS15 OS19 OS23 SB20 SB27 3.00 3.25 3.50 3.75 4.00 H−1.2kmNw VK2004 Injection Wells Production Wells (c) −4184000 −4182000 −4180000 −4184000 −4182000 −4180000 −2698000 −2696000 −2694000 −2698000 −2696000 −2694000 Poisson Ratio 2011 Dec 15 09:29:25 Poisson Ratio Scale 1:40000 CA−45A−12 CMHC−2 CMHC−6 DX−19 DX−24 DX−47 DX−5 GDC−26 GDC−53−13 GDC−88−12 LF−2 LF−23 OS−11 OS−12 OS−16 SB−15 CA32−12 CA45−12 CA48A−2 CA52−11 Curry3 DX23 DX25 DX28 DX3 DX38 DX42 DX59 DX77 GDC2 GDC6 OS15 OS19 OS23 SB20 SB27 0.15 0.20 0.25 H−2.4kmNw Poisson2004 Injection Wells Production Wells (d) −4184000 −4182000 −4180000 −4184000 −4182000 −4180000 −2698000 −2696000 −2694000 −2698000 −2696000 −2694000 Extensional Stress 2011 Dec 15 09:31:08 Extensional Stress Scale 1:40000 CA−45A−12 CMHC−2 CMHC−6 DX−19 DX−24 DX−47 DX−5 GDC−26 GDC−53−13 GDC−88−12 LF−2 LF−23 OS−11 OS−12 OS−16 SB−15 CA32−12 CA45−12 CA48A−2 CA52−11 Curry3 DX23 DX25 DX28 DX3 DX38 DX42 DX59 DX77 GDC2 GDC6 OS15 OS19 OS23 SB20 SB27 4.4 4.6 4.8 H−2.4kmNw VE2004 Injection Wells Production Wells (e) −4184000 −4182000 −4180000 −4184000 −4182000 −4180000 −2698000 −2696000 −2694000 −2698000 −2696000 −2694000 Hydrostatic Stress 2011 Dec 15 09:32:33 Hydrostatic Stress Scale 1:40000 CA−45A−12 CMHC−2 CMHC−6 DX−19 DX−24 DX−47 DX−5 GDC−26 GDC−53−13 GDC−88−12 LF−2 LF−23 OS−11 OS−12 OS−16 SB−15 CA32−12 CA45−12 CA48A−2 CA52−11 Curry3 DX23 DX25 DX28 DX3 DX38 DX42 DX59 DX77 GDC2 GDC6 OS15 OS19 OS23 SB20 SB27 3.00 3.25 3.50 3.75 4.00 H−2.4kmNw VK2004 Injection Wells Production Wells (f) Figure 6.22: NTR horizon at The Geyser (a) Poisson’s ratio, (b) Normal stress, (c) Confining stress, HTZ horizon at The Geyser (d) Poisson’s ratio, (e) Normal stress, (f) Confining stress 103 1 0 -1 -2 -3 -4 -5 Depth (km) Figure 6.23: Correlation between microseismic cluster movement and velocity anomaly direction; the red circle is the microseismic cluster center in 2006 and the yellow one is for 2009 104 -2698.32 -2697.33 -2696.35 -2695.37 -2694.38 -2693.4 -4185 -4184 -4183 -4182 -4181 X (m) Y (m) LF-4 LF-37 GDC-6 GDC-2 OS-19 OS-15 DX-38 OS-23 DX-29 DX-28 DX-77 DX-23 DX-59 CA-48A-2 CA-52-11 PG1 PG2 PG3 PG4 PG5 PG6 GDC-18 CA-117A-19 LF-2 DX-5 GDC-88-12 SB-15 OS-11 OS-12 OS-16 DX-24 DX-19 LF-23 DX-47 CA-45A-12 CA-87A-2 IG1 IG2 IG3 54254 T/m 28100 T/m 1947 T/m 296422 T/m 150521 T/m 4621 T/m X10 3 4.5 4.6 4.7 4.8 4.9 5 5.1 5.2 5.3 (a) -2698.32 -2697.33 -2696.35 -2695.37 -2694.38 -2693.4 -4185 -4184 -4183 -4182 -4181 X (m) Y (m) LF-4 LF-37 GDC-6 GDC-2 OS-19 OS-15 DX-38 OS-23 DX-29 DX-28 DX-77 DX-23 DX-59 CA-48A-2 CA-52-11 PG1 PG2 PG3 PG4 PG5 PG6 GDC-18 CA-117A-19 LF-2 DX-5 GDC-88-12 SB-15 OS-11 OS-12 OS-16 DX-24 DX-19 LF-23 DX-47 CA-45A-12 CA-87A-2 IG1 IG2 IG3 54254 T/m 28100 T/m 1947 T/m 296422 T/m 150521 T/m 4621 T/m X10 3 1.6 1.65 1.7 1.75 1.8 1.85 (b) -2698.32 -2697.33 -2696.35 -2695.37 -2694.38 -2693.4 -4185 -4184 -4183 -4182 -4181 X (m) Y (m) LF-4 LF-37 GDC-6 GDC-2 OS-19 OS-15 DX-38 OS-23 DX-29 DX-28 DX-77 DX-23 DX-59 CA-48A-2 CA-52-11 PG1 PG2 PG3 PG4 PG5 PG6 GDC-18 CA-117A-19 LF-2 DX-5 GDC-88-12 SB-15 OS-11 OS-12 OS-16 DX-24 DX-19 LF-23 DX-47 CA-45A-12 CA-87A-2 IG1 IG2 IG3 54254 T/m 28100 T/m 1947 T/m 296422 T/m 150521 T/m 4621 T/m X10 3 0.15 0.2 0.25 0.3 (c) -2698.32 -2697.33 -2696.35 -2695.37 -2694.38 -2693.4 -4185 -4184 -4183 -4182 -4181 X (m) Y (m) LF-4 LF-37 GDC-6 GDC-2 OS-19 OS-15 DX-38 OS-23 DX-29 DX-28 DX-77 DX-23 DX-59 CA-48A-2 CA-52-11 PG1 PG2 PG3 PG4 PG5 PG6 GDC-18 CA-117A-19 LF-2 DX-5 GDC-88-12 SB-15 OS-11 OS-12 OS-16 DX-24 DX-19 LF-23 DX-47 CA-45A-12 CA-87A-2 IG1 IG2 IG3 54254 T/m 28100 T/m 1947 T/m 296422 T/m 150521 T/m 4621 T/m X10 3 4.1 4.2 4.3 4.4 4.5 4.6 4.7 4.8 (d) Figure 6.24: NTR horizon at The NW Geysers, (a)V P , (b)V P =V S , (c) Poisson’s ratio (d) Normal stress, with production/injection data superimposed (Production: White, Injection: Black) 105 Table 6.2: Major Production = injection wells within the area of interest, with their associated rates from DOGGR database. Well Name Type 2005 2006 2007 2008 2009 2010 Production (Injection) Average daily rate (1000 kg per day) CA-117A-19 I 17624.66 15914.47 14851.29 11040.52 15143.39 18250.07 CA-48A-2 P 115.27 0 0 0 0 97.86 CA-52-11 P 937.8 952.81 949.95 968.55 892.85 913.72 CA-87A-2 I 3878.28 3476.07 3592.98 3532.73 2809.31 3575.65 DX-29 P 486.3 480.3 479.39 446.18 407.39 400.67 DX-38 P 275.45 270.22 265.72 282.18 260.05 258.83 DX-47 I 4961.38 3451.8 2397.08 1557.14 796.13 677.5 DX-77 P 248.29 216.93 224.19 216.26 233.2 215.44 GDC-18 I 3218.41 3432.39 4724.39 2832.3 0 0 GDC-2 P 292.49 305.51 259.89 271.3 254.15 275.54 GDC-6 P 662.66 681.31 728.7 681.95 635.75 678.18 LF-2 I 401.96 1351.71 0 1230 0 1562 LF-37 P 211.8 332.13 354.41 296.82 284.77 301.19 LF-4 P 462.48 476.82 471.46 444.63 412.3 452.62 OS-11 I 4664.91 4301.19 3757.21 2953.66 2544.54 2558.64 OS-12 I 2725.47 2052.19 2188.11 937.92 0 0 OS-15 P 224.66 219.21 177.27 160.38 163.39 119.85 OS-19 P 511.45 500.55 414.35 396.25 387.6 378.4 OS-23 P 704.15 758.48 767.78 767.89 693.58 639.95 SB-15 I 463 1574.98 1663.33 502 450.54 221 IG1 I 11397.36 10675.89 4661.88 6538.75 0 0 IG2 I 7146.01 7080.48 8782.13 6836.39 7142.13 7053.58 IG3 I 924.34 1618.18 1451.33 1075.79 791.91 606.2 PG1 P 1481.92 1572.19 1585.76 1616.17 1568.05 1618.25 PG2 P 1371 1379.8 1424.19 1700.79 1608.34 1612.38 PG3 P 1856.33 2033.19 2061.68 2278.12 2050.65 2033.79 PG4 P 831.432 895.32 1185.83 1205.6 1140.18 1126.85 PG5 P 861.42 833.715 820.51 782.15 704.27 630.39 PG6 P 1520.19 1436.34 1468.14 1417.67 1313.88 1298.91 106 Table 6.3: Production= injection well clustering Group Name Well Names IG1 GDC-53-13, GDC-53A-13 IG2 CMHC-2, DX-26, GDC-36-18 IG3 CA-96-18, CMHC-6 PG1 CA-32-12, CA-32A-12, CA-32B-12 PG2 SB-20, SB-30, SB-21 PG3 SB-27, SB-24, SB-28 PG4 Curry-1, Curry-3 PG5 DX-25, DX-12 PG6 DX-41, DX-42 107 −2699 −2698 −2697 −2696 −2695 −2694 −2693 −4185 −4184 −4183 −4182 −4181 X (km) Y (km) LF−4 LF−37 GDC−6 GDC−2 OS−19 OS−15 DX−38 OS−23 DX−29 DX−28 DX−77 DX−23 DX−59 CA−48A−2 CA−52−11 PG1 PG2 PG3 PG4 PG5 PG6 GDC−18 CA−117A−19 LF−2 GDC−53−13 SB−15 OS−11 OS−12 DX−26 CMHC−6 DX−47 GDC−53A−13 CA−87A−2 IG1 IG2 IG3 2268 1152 36 18250 9130 10 3.5 4 4.5 5 5.5 (a) −2699 −2698 −2697 −2696 −2695 −2694 −2693 −4185 −4184 −4183 −4182 −4181 X (km) Y (km) LF−4 LF−37 GDC−6 GDC−2 OS−19 OS−15 DX−38 OS−23 DX−29 DX−28 DX−77 DX−23 DX−59 CA−52−11 PG1 PG2 PG3 PG4 PG5 PG6 GDC−88−12 GDC−88−12 LF−23 LF−23 GDC−18 CA−117A−19 LF−2 GDC−53−13 GDC−88−12 GDC−88−12 SB−15 OS−11 OS−3 OS−12 DX−26 LF−23 LF−23 CMHC−6 DX−47 GDC−53A−13 CA−87A−2 IG1 IG2 IG3 2268 1152 36 18250 9130 10 3.5 4 4.5 5 5.5 (b) −2699 −2698 −2697 −2696 −2695 −2694 −2693 −4185 −4184 −4183 −4182 −4181 X (km) Y (km) LF−4 LF−37 GDC−6 GDC−2 OS−19 OS−15 DX−38 OS−23 DX−29 DX−28 DX−77 DX−23 DX−59 CA−52−11 PG1 PG2 PG3 PG4 PG5 PG6 GDC−88−12 GDC−88−12 LF−23 LF−23 GDC−18 CA−117A−19 GDC−53−13 GDC−88−12 GDC−88−12 SB−15 OS−11 OS−3 OS−12 DX−26 GDC−36−18 LF−23 LF−23 SB−25 CMHC−6 DX−47 CA−73B−12 GDC−53A−13 CA−87A−2 IG1 IG2 IG3 2268 1152 36 18250 9130 10 3.5 4 4.5 5 5.5 (c) −2699 −2698 −2697 −2696 −2695 −2694 −2693 −4185 −4184 −4183 −4182 −4181 X (km) Y (km) LF−4 LF−37 GDC−6 GDC−2 OS−19 OS−15 DX−38 OS−23 DX−29 DX−28 DX−77 DX−23 DX−59 CA−52−11 PG1 PG2 PG3 PG4 PG5 PG6 GDC−88−12 GDC−88−12 LF−23 LF−23 GDC−18 CA−117A−19 LF−2 Curry−85−13 GDC−32A−13 GDC−53−13 GDC−88−12 GDC−88−12 SB−15 OS−11 OS−12 DX−26 GDC−36−18 LF−23 LF−23 SB−25 CMHC−6 DX−47 CA−73B−12 GDC−53A−13 CA−87A−2 CA−51A−11 IG1 IG2 IG3 2268 1152 36 18250 9130 10 3.5 4 4.5 5 5.5 (d) −2699 −2698 −2697 −2696 −2695 −2694 −2693 −4185 −4184 −4183 −4182 −4181 X (km) Y (km) LF−4 LF−37 GDC−6 GDC−2 OS−19 OS−15 DX−38 OS−23 DX−29 DX−28 DX−77 DX−23 DX−59 CA−52−11 PG1 PG2 PG3 PG4 PG5 PG6 GDC−88−12 GDC−88−12 LF−23 LF−23 CA−117A−19 Curry−85−13 GDC−32A−13 GDC−88−12 GDC−88−12 SB−15 OS−11 DX−26 GDC−36−18 LF−23 LF−23 SB−25 CMHC−6 DX−47 CA−73B−12 CA−87A−2 CA−51A−11 IG1 IG2 IG3 2268 1152 36 18250 9130 10 3.5 4 4.5 5 5.5 (e) −2699 −2698 −2697 −2696 −2695 −2694 −2693 −4185 −4184 −4183 −4182 −4181 X (km) Y (km) LF−4 LF−37 GDC−6 GDC−2 OS−19 OS−15 DX−38 OS−23 DX−29 DX−28 DX−77 DX−23 DX−59 CA−48A−2 CA−52−11 PG1 PG2 PG3 PG4 PG5 PG6 LF−23 LF−23 CA−117A−19 LF−2 Curry−85−13 GDC−32A−13 SB−15 OS−11 DX−26 GDC−36−18 LF−23 LF−23 SB−25 CMHC−6 DX−47 CA−73B−12 CA−87A−2 CA−51A−11 IG1 IG2 IG3 2268 1152 36 18250 9130 10 3.5 4 4.5 5 5.5 (f) −2699 −2698 −2697 −2696 −2695 −2694 −2693 −4185 −4184 −4183 −4182 −4181 X (m) Y (m) LF−4 LF−37 GDC−6 GDC−2 OS−19 OS−15 DX−38 OS−23 DX−29 DX−28 DX−77 DX−23 DX−59 CA−48A−2 CA−52−11 PG1 PG2 PG3 PG4 PG5 PG6 GDC−18 CA−117A−19 LF−2 GDC−53−13 SB−15 OS−11 OS−12 DX−26 CMHC−6 DX−47 GDC−53A−13 CA−87A−2 IG1 IG2 IG3 2268 1152 36 18250 9130 10 X10 3 3.5 4 4.5 5 5.5 km s Figure 6.25: Compressional wave seismic velocity distribution in NTR horizon at The NW Geysers (a) 2005, (b) 2006, (c) 2007, (d) 2008, (e) 2009, (f) 2010 108 Figure 6.27 shows a reduction of the Bulk modulus in the period from 2005 to 2010 for The NW Geysers, which seems to indicate that the void ratio or fracture density increased and the fractured network expanded within the reservoir; see section 6.4.2. It is also interesting to note that the Bulk modulus value in the area of low velocity anomaly in Figure 6.25 is less than 25Gpa, strongly suggesting an area of heavily fractured rock. There are two mechanisms of solid/fluid interaction in the rocks: the Biot mech- anism and the squirt-flow mechanism. Squirt flow, localized between compliant and noncompliant pores, is considered the cause of predominant loss in the Bulk modulus, and a small loss in the Shear modulus (Barton, 2006). The areas corresponding to OS- 12, DX-77, and GDC-18 shows this kind of behavior, as observed in Figures 6.27 and 6.28. Therefore, we can hypothesize that these may be fracture induced anomalies, as saturation alone would raise the Bulk modulus and reduce the Shear modulus. Biot flow is considered the cause for a predominant loss in the Shear modulus and a compara- tively smaller loss in Bulk modulus. This kind of flow arises from fluid movement in open pores and macro-permeability where fracture density is very high. The areas asso- ciated with PG1 and SB25 show this kind of behavior where–with relatively low water injection rates–higher steam production occurs than in any other areas. The behavior of both moduli also suggests a fracture closure of the area between DX-77 and DX-38 from 2005 to 2010. Barton (2006) reported that Youngs modulus in fractured rock decreased exponentially with fracture density. Figure 6.29 shows a northwest trend of Young’s modulus reduction at The Geysers, which the authors interpret as a spatial distribution of the fracture network. The higher degree of reduction shows higher fracture density in the area. As mentioned in section 6.4.1, in the hydraulic fracturing process–such as in the shale reservoirs–, or thermal contraction process–such as in The Geysers–we can use the effective normal stress as an index for fracture opening which is essential for fluid 109 movement and production. Figure 6.30 shows the distribution of normal stress at the horizon of interest. A low normal stress anomaly (represented by the blue and green areas) indicates where fractures may be open to providing enough permeability for pro- duction. 110 −2699 −2698 −2697 −2696 −2695 −2694 −2693 −4185 −4184 −4183 −4182 −4181 X (km) Y (km) LF−4 LF−37 GDC−6 GDC−2 OS−19 OS−15 DX−38 OS−23 DX−29 DX−28 DX−77 DX−23 DX−59 CA−48A−2 CA−52−11 PG1 PG2 PG3 PG4 PG5 PG6 GDC−18 CA−117A−19 LF−2 GDC−53−13 SB−15 OS−11 OS−12 DX−26 CMHC−6 DX−47 GDC−53A−13 CA−87A−2 IG1 IG2 IG3 2268 1152 36 18250 9130 10 2.3 2.4 2.5 2.6 2.7 2.8 2.9 3 3.1 3.2 (a) −2699 −2698 −2697 −2696 −2695 −2694 −2693 −4185 −4184 −4183 −4182 −4181 X (km) Y (km) LF−4 LF−37 GDC−6 GDC−2 OS−19 OS−15 DX−38 OS−23 DX−29 DX−28 DX−77 DX−23 DX−59 CA−52−11 PG1 PG2 PG3 PG4 PG5 PG6 GDC−88−12 GDC−88−12 LF−23 LF−23 GDC−18 CA−117A−19 LF−2 GDC−53−13 GDC−88−12 GDC−88−12 SB−15 OS−11 OS−3 OS−12 DX−26 LF−23 LF−23 CMHC−6 DX−47 GDC−53A−13 CA−87A−2 IG1 IG2 IG3 2268 1152 36 18250 9130 10 2.3 2.4 2.5 2.6 2.7 2.8 2.9 3 3.1 3.2 (b) −2699 −2698 −2697 −2696 −2695 −2694 −2693 −4185 −4184 −4183 −4182 −4181 X (km) Y (km) LF−4 LF−37 GDC−6 GDC−2 OS−19 OS−15 DX−38 OS−23 DX−29 DX−28 DX−77 DX−23 DX−59 CA−52−11 PG1 PG2 PG3 PG4 PG5 PG6 GDC−88−12 GDC−88−12 LF−23 LF−23 GDC−18 CA−117A−19 GDC−53−13 GDC−88−12 GDC−88−12 SB−15 OS−11 OS−3 OS−12 DX−26 GDC−36−18 LF−23 LF−23 SB−25 CMHC−6 DX−47 CA−73B−12 GDC−53A−13 CA−87A−2 IG1 IG2 IG3 2268 1152 36 18250 9130 10 2.3 2.4 2.5 2.6 2.7 2.8 2.9 3 3.1 3.2 (c) −2699 −2698 −2697 −2696 −2695 −2694 −2693 −4185 −4184 −4183 −4182 −4181 X (km) Y (km) LF−4 LF−37 GDC−6 GDC−2 OS−19 OS−15 DX−38 OS−23 DX−29 DX−28 DX−77 DX−23 DX−59 CA−52−11 PG1 PG2 PG3 PG4 PG5 PG6 GDC−88−12 GDC−88−12 LF−23 LF−23 GDC−18 CA−117A−19 LF−2 Curry−85−13 GDC−32A−13 GDC−53−13 GDC−88−12 GDC−88−12 SB−15 OS−11 OS−12 DX−26 GDC−36−18 LF−23 LF−23 SB−25 CMHC−6 DX−47 CA−73B−12 GDC−53A−13 CA−87A−2 CA−51A−11 IG1 IG2 IG3 2268 1152 36 18250 9130 10 2.3 2.4 2.5 2.6 2.7 2.8 2.9 3 3.1 3.2 (d) −2699 −2698 −2697 −2696 −2695 −2694 −2693 −4185 −4184 −4183 −4182 −4181 X (km) Y (km) LF−4 LF−37 GDC−6 GDC−2 OS−19 OS−15 DX−38 OS−23 DX−29 DX−28 DX−77 DX−23 DX−59 CA−52−11 PG1 PG2 PG3 PG4 PG5 PG6 GDC−88−12 GDC−88−12 LF−23 LF−23 CA−117A−19 Curry−85−13 GDC−32A−13 GDC−88−12 GDC−88−12 SB−15 OS−11 DX−26 GDC−36−18 LF−23 LF−23 SB−25 CMHC−6 DX−47 CA−73B−12 CA−87A−2 CA−51A−11 IG1 IG2 IG3 2268 1152 36 18250 9130 10 2.3 2.4 2.5 2.6 2.7 2.8 2.9 3 3.1 3.2 (e) −2699 −2698 −2697 −2696 −2695 −2694 −2693 −4185 −4184 −4183 −4182 −4181 X (km) Y (km) LF−4 LF−37 GDC−6 GDC−2 OS−19 OS−15 DX−38 OS−23 DX−29 DX−28 DX−77 DX−23 DX−59 CA−48A−2 CA−52−11 PG1 PG2 PG3 PG4 PG5 PG6 LF−23 LF−23 CA−117A−19 LF−2 Curry−85−13 GDC−32A−13 SB−15 OS−11 DX−26 GDC−36−18 LF−23 LF−23 SB−25 CMHC−6 DX−47 CA−73B−12 CA−87A−2 CA−51A−11 IG1 IG2 IG3 2268 1152 36 18250 9130 10 2.3 2.4 2.5 2.6 2.7 2.8 2.9 3 3.1 3.2 (f) −2699 −2698 −2697 −2696 −2695 −2694 −2693 −4185 −4184 −4183 −4182 −4181 X (m) Y (m) LF−4 LF−37 GDC−6 GDC−2 OS−19 OS−15 DX−38 OS−23 DX−29 DX−28 DX−77 DX−23 DX−59 CA−48A−2 CA−52−11 PG1 PG2 PG3 PG4 PG5 PG6 GDC−18 CA−117A−19 LF−2 GDC−53−13 SB−15 OS−11 OS−12 DX−26 CMHC−6 DX−47 GDC−53A−13 CA−87A−2 IG1 IG2 IG3 2268 1152 36 18250 9130 10 X10 3 2.3 2.4 2.5 2.6 2.7 2.8 2.9 3 3.1 3.2 km s Figure 6.26: Shear wave seismic velocity distribution in NTR horizon at The NW Gey- sers (a) 2005, (b) 2006, (c) 2007,(d) 2008, (e) 2009, (f) 2010 111 −2699 −2698 −2697 −2696 −2695 −2694 −2693 −4185 −4184 −4183 −4182 −4181 X (km) Y (km) LF−4 LF−37 GDC−6 GDC−2 OS−19 OS−15 DX−38 OS−23 DX−29 DX−28 DX−77 DX−23 DX−59 CA−48A−2 CA−52−11 PG1 PG2 PG3 PG4 PG5 PG6 GDC−18 CA−117A−19 LF−2 GDC−53−13 SB−15 OS−11 OS−12 DX−26 CMHC−6 DX−47 GDC−53A−13 CA−87A−2 IG1 IG2 IG3 2268 1152 36 18250 9130 10 10 15 20 25 30 35 40 (a) −2699 −2698 −2697 −2696 −2695 −2694 −2693 −4185 −4184 −4183 −4182 −4181 X (km) Y (km) LF−4 LF−37 GDC−6 GDC−2 OS−19 OS−15 DX−38 OS−23 DX−29 DX−28 DX−77 DX−23 DX−59 CA−52−11 PG1 PG2 PG3 PG4 PG5 PG6 GDC−88−12 GDC−88−12 LF−23 LF−23 GDC−18 CA−117A−19 LF−2 GDC−53−13 GDC−88−12 GDC−88−12 SB−15 OS−11 OS−3 OS−12 DX−26 LF−23 LF−23 CMHC−6 DX−47 GDC−53A−13 CA−87A−2 IG1 IG2 IG3 2268 1152 36 18250 9130 10 10 15 20 25 30 35 40 (b) −2699 −2698 −2697 −2696 −2695 −2694 −2693 −4185 −4184 −4183 −4182 −4181 X (km) Y (km) LF−4 LF−37 GDC−6 GDC−2 OS−19 OS−15 DX−38 OS−23 DX−29 DX−28 DX−77 DX−23 DX−59 CA−52−11 PG1 PG2 PG3 PG4 PG5 PG6 GDC−88−12 GDC−88−12 LF−23 LF−23 GDC−18 CA−117A−19 GDC−53−13 GDC−88−12 GDC−88−12 SB−15 OS−11 OS−3 OS−12 DX−26 GDC−36−18 LF−23 LF−23 SB−25 CMHC−6 DX−47 CA−73B−12 GDC−53A−13 CA−87A−2 IG1 IG2 IG3 2268 1152 36 18250 9130 10 10 15 20 25 30 35 40 (c) −2699 −2698 −2697 −2696 −2695 −2694 −2693 −4185 −4184 −4183 −4182 −4181 X (km) Y (km) LF−4 LF−37 GDC−6 GDC−2 OS−19 OS−15 DX−38 OS−23 DX−29 DX−28 DX−77 DX−23 DX−59 CA−52−11 PG1 PG2 PG3 PG4 PG5 PG6 GDC−88−12 GDC−88−12 LF−23 LF−23 GDC−18 CA−117A−19 LF−2 Curry−85−13 GDC−32A−13 GDC−53−13 GDC−88−12 GDC−88−12 SB−15 OS−11 OS−12 DX−26 GDC−36−18 LF−23 LF−23 SB−25 CMHC−6 DX−47 CA−73B−12 GDC−53A−13 CA−87A−2 CA−51A−11 IG1 IG2 IG3 2268 1152 36 18250 9130 10 10 15 20 25 30 35 40 (d) −2699 −2698 −2697 −2696 −2695 −2694 −2693 −4185 −4184 −4183 −4182 −4181 X (km) Y (km) LF−4 LF−37 GDC−6 GDC−2 OS−19 OS−15 DX−38 OS−23 DX−29 DX−28 DX−77 DX−23 DX−59 CA−52−11 PG1 PG2 PG3 PG4 PG5 PG6 GDC−88−12 GDC−88−12 LF−23 LF−23 CA−117A−19 Curry−85−13 GDC−32A−13 GDC−88−12 GDC−88−12 SB−15 OS−11 DX−26 GDC−36−18 LF−23 LF−23 SB−25 CMHC−6 DX−47 CA−73B−12 CA−87A−2 CA−51A−11 IG1 IG2 IG3 2268 1152 36 18250 9130 10 10 15 20 25 30 35 40 (e) −2699 −2698 −2697 −2696 −2695 −2694 −2693 −4185 −4184 −4183 −4182 −4181 X (km) Y (km) LF−4 LF−37 GDC−6 GDC−2 OS−19 OS−15 DX−38 OS−23 DX−29 DX−28 DX−77 DX−23 DX−59 CA−48A−2 CA−52−11 PG1 PG2 PG3 PG4 PG5 PG6 LF−23 LF−23 CA−117A−19 LF−2 Curry−85−13 GDC−32A−13 SB−15 OS−11 DX−26 GDC−36−18 LF−23 LF−23 SB−25 CMHC−6 DX−47 CA−73B−12 CA−87A−2 CA−51A−11 IG1 IG2 IG3 2268 1152 36 18250 9130 10 10 15 20 25 30 35 40 (f) −2699 −2698 −2697 −2696 −2695 −2694 −2693 −4185 −4184 −4183 −4182 −4181 X (m) Y (m) LF−4 LF−37 GDC−6 GDC−2 OS−19 OS−15 DX−38 OS−23 DX−29 DX−28 DX−77 DX−23 DX−59 CA−48A−2 CA−52−11 PG1 PG2 PG3 PG4 PG5 PG6 GDC−18 CA−117A−19 LF−2 GDC−53−13 SB−15 OS−11 OS−12 DX−26 CMHC−6 DX−47 GDC−53A−13 CA−87A−2 IG1 IG2 IG3 2268 1152 36 18250 9130 10 X10 3 10 15 20 25 30 35 40 Gpa Figure 6.27: Bulk modulus distribution in NTR horizon at The NW Geysers (a)2005, (b)2006, (c)2007, (d)2008, (e)2009, (f)2010 112 −2699 −2698 −2697 −2696 −2695 −2694 −2693 −4185 −4184 −4183 −4182 −4181 X (km) Y (km) LF−4 LF−37 GDC−6 GDC−2 OS−19 OS−15 DX−38 OS−23 DX−29 DX−28 DX−77 DX−23 DX−59 CA−48A−2 CA−52−11 PG1 PG2 PG3 PG4 PG5 PG6 GDC−18 CA−117A−19 LF−2 GDC−53−13 SB−15 OS−11 OS−12 DX−26 CMHC−6 DX−47 GDC−53A−13 CA−87A−2 IG1 IG2 IG3 2268 1152 36 18250 9130 10 15 20 25 30 (a) −2699 −2698 −2697 −2696 −2695 −2694 −2693 −4185 −4184 −4183 −4182 −4181 X (km) Y (km) LF−4 LF−37 GDC−6 GDC−2 OS−19 OS−15 DX−38 OS−23 DX−29 DX−28 DX−77 DX−23 DX−59 CA−52−11 PG1 PG2 PG3 PG4 PG5 PG6 GDC−88−12 GDC−88−12 LF−23 LF−23 GDC−18 CA−117A−19 LF−2 GDC−53−13 GDC−88−12 GDC−88−12 SB−15 OS−11 OS−3 OS−12 DX−26 LF−23 LF−23 CMHC−6 DX−47 GDC−53A−13 CA−87A−2 IG1 IG2 IG3 2268 1152 36 18250 9130 10 15 20 25 30 (b) −2699 −2698 −2697 −2696 −2695 −2694 −2693 −4185 −4184 −4183 −4182 −4181 X (km) Y (km) LF−4 LF−37 GDC−6 GDC−2 OS−19 OS−15 DX−38 OS−23 DX−29 DX−28 DX−77 DX−23 DX−59 CA−52−11 PG1 PG2 PG3 PG4 PG5 PG6 GDC−88−12 GDC−88−12 LF−23 LF−23 GDC−18 CA−117A−19 GDC−53−13 GDC−88−12 GDC−88−12 SB−15 OS−11 OS−3 OS−12 DX−26 GDC−36−18 LF−23 LF−23 SB−25 CMHC−6 DX−47 CA−73B−12 GDC−53A−13 CA−87A−2 IG1 IG2 IG3 2268 1152 36 18250 9130 10 15 20 25 30 (c) −2699 −2698 −2697 −2696 −2695 −2694 −2693 −4185 −4184 −4183 −4182 −4181 X (km) Y (km) LF−4 LF−37 GDC−6 GDC−2 OS−19 OS−15 DX−38 OS−23 DX−29 DX−28 DX−77 DX−23 DX−59 CA−52−11 PG1 PG2 PG3 PG4 PG5 PG6 GDC−88−12 GDC−88−12 LF−23 LF−23 GDC−18 CA−117A−19 LF−2 Curry−85−13 GDC−32A−13 GDC−53−13 GDC−88−12 GDC−88−12 SB−15 OS−11 OS−12 DX−26 GDC−36−18 LF−23 LF−23 SB−25 CMHC−6 DX−47 CA−73B−12 GDC−53A−13 CA−87A−2 CA−51A−11 IG1 IG2 IG3 2268 1152 36 18250 9130 10 15 20 25 30 (d) −2699 −2698 −2697 −2696 −2695 −2694 −2693 −4185 −4184 −4183 −4182 −4181 X (km) Y (km) LF−4 LF−37 GDC−6 GDC−2 OS−19 OS−15 DX−38 OS−23 DX−29 DX−28 DX−77 DX−23 DX−59 CA−52−11 PG1 PG2 PG3 PG4 PG5 PG6 GDC−88−12 GDC−88−12 LF−23 LF−23 CA−117A−19 Curry−85−13 GDC−32A−13 GDC−88−12 GDC−88−12 SB−15 OS−11 DX−26 GDC−36−18 LF−23 LF−23 SB−25 CMHC−6 DX−47 CA−73B−12 CA−87A−2 CA−51A−11 IG1 IG2 IG3 2268 1152 36 18250 9130 10 15 20 25 30 (e) −2699 −2698 −2697 −2696 −2695 −2694 −2693 −4185 −4184 −4183 −4182 −4181 X (km) Y (km) LF−4 LF−37 GDC−6 GDC−2 OS−19 OS−15 DX−38 OS−23 DX−29 DX−28 DX−77 DX−23 DX−59 CA−48A−2 CA−52−11 PG1 PG2 PG3 PG4 PG5 PG6 LF−23 LF−23 CA−117A−19 LF−2 Curry−85−13 GDC−32A−13 SB−15 OS−11 DX−26 GDC−36−18 LF−23 LF−23 SB−25 CMHC−6 DX−47 CA−73B−12 CA−87A−2 CA−51A−11 IG1 IG2 IG3 2268 1152 36 18250 9130 10 15 20 25 30 (f) −2699 −2698 −2697 −2696 −2695 −2694 −2693 −4185 −4184 −4183 −4182 −4181 X (m) Y (m) LF−4 LF−37 GDC−6 GDC−2 OS−19 OS−15 DX−38 OS−23 DX−29 DX−28 DX−77 DX−23 DX−59 CA−48A−2 CA−52−11 PG1 PG2 PG3 PG4 PG5 PG6 GDC−18 CA−117A−19 LF−2 GDC−53−13 SB−15 OS−11 OS−12 DX−26 CMHC−6 DX−47 GDC−53A−13 CA−87A−2 IG1 IG2 IG3 2268 1152 36 18250 9130 10 X10 3 15 20 25 30 Gpa Figure 6.28: Shear modulus distribution in NTR horizon at The NW Geysers (a)2005, (b)2006, (c)2007, (d)2008, (e)2009, (f)2010 113 −2699 −2698 −2697 −2696 −2695 −2694 −2693 −4185 −4184 −4183 −4182 −4181 X (km) Y (km) LF−4 LF−37 GDC−6 GDC−2 OS−19 OS−15 DX−38 OS−23 DX−29 DX−28 DX−77 DX−23 DX−59 CA−48A−2 CA−52−11 PG1 PG2 PG3 PG4 PG5 PG6 GDC−18 CA−117A−19 LF−2 GDC−53−13 SB−15 OS−11 OS−12 DX−26 CMHC−6 DX−47 GDC−53A−13 CA−87A−2 IG1 IG2 IG3 2268 1152 36 18250 9130 10 35 40 45 50 55 60 65 70 (a) −2699 −2698 −2697 −2696 −2695 −2694 −2693 −4185 −4184 −4183 −4182 −4181 X (km) Y (km) LF−4 LF−37 GDC−6 GDC−2 OS−19 OS−15 DX−38 OS−23 DX−29 DX−28 DX−77 DX−23 DX−59 CA−52−11 PG1 PG2 PG3 PG4 PG5 PG6 GDC−88−12 GDC−88−12 LF−23 LF−23 GDC−18 CA−117A−19 LF−2 GDC−53−13 GDC−88−12 GDC−88−12 SB−15 OS−11 OS−3 OS−12 DX−26 LF−23 LF−23 CMHC−6 DX−47 GDC−53A−13 CA−87A−2 IG1 IG2 IG3 2268 1152 36 18250 9130 10 35 40 45 50 55 60 65 70 (b) −2699 −2698 −2697 −2696 −2695 −2694 −2693 −4185 −4184 −4183 −4182 −4181 X (km) Y (km) LF−4 LF−37 GDC−6 GDC−2 OS−19 OS−15 DX−38 OS−23 DX−29 DX−28 DX−77 DX−23 DX−59 CA−52−11 PG1 PG2 PG3 PG4 PG5 PG6 GDC−88−12 GDC−88−12 LF−23 LF−23 GDC−18 CA−117A−19 GDC−53−13 GDC−88−12 GDC−88−12 SB−15 OS−11 OS−3 OS−12 DX−26 GDC−36−18 LF−23 LF−23 SB−25 CMHC−6 DX−47 CA−73B−12 GDC−53A−13 CA−87A−2 IG1 IG2 IG3 2268 1152 36 18250 9130 10 35 40 45 50 55 60 65 70 (c) −2699 −2698 −2697 −2696 −2695 −2694 −2693 −4185 −4184 −4183 −4182 −4181 X (km) Y (km) LF−4 LF−37 GDC−6 GDC−2 OS−19 OS−15 DX−38 OS−23 DX−29 DX−28 DX−77 DX−23 DX−59 CA−52−11 PG1 PG2 PG3 PG4 PG5 PG6 GDC−88−12 GDC−88−12 LF−23 LF−23 GDC−18 CA−117A−19 LF−2 Curry−85−13 GDC−32A−13 GDC−53−13 GDC−88−12 GDC−88−12 SB−15 OS−11 OS−12 DX−26 GDC−36−18 LF−23 LF−23 SB−25 CMHC−6 DX−47 CA−73B−12 GDC−53A−13 CA−87A−2 CA−51A−11 IG1 IG2 IG3 2268 1152 36 18250 9130 10 35 40 45 50 55 60 65 70 (d) −2699 −2698 −2697 −2696 −2695 −2694 −2693 −4185 −4184 −4183 −4182 −4181 X (km) Y (km) LF−4 LF−37 GDC−6 GDC−2 OS−19 OS−15 DX−38 OS−23 DX−29 DX−28 DX−77 DX−23 DX−59 CA−52−11 PG1 PG2 PG3 PG4 PG5 PG6 GDC−88−12 GDC−88−12 LF−23 LF−23 CA−117A−19 Curry−85−13 GDC−32A−13 GDC−88−12 GDC−88−12 SB−15 OS−11 DX−26 GDC−36−18 LF−23 LF−23 SB−25 CMHC−6 DX−47 CA−73B−12 CA−87A−2 CA−51A−11 IG1 IG2 IG3 2268 1152 36 18250 9130 10 35 40 45 50 55 60 65 70 (e) −2699 −2698 −2697 −2696 −2695 −2694 −2693 −4185 −4184 −4183 −4182 −4181 X (km) Y (km) LF−4 LF−37 GDC−6 GDC−2 OS−19 OS−15 DX−38 OS−23 DX−29 DX−28 DX−77 DX−23 DX−59 CA−48A−2 CA−52−11 PG1 PG2 PG3 PG4 PG5 PG6 LF−23 LF−23 CA−117A−19 LF−2 Curry−85−13 GDC−32A−13 SB−15 OS−11 DX−26 GDC−36−18 LF−23 LF−23 SB−25 CMHC−6 DX−47 CA−73B−12 CA−87A−2 CA−51A−11 IG1 IG2 IG3 2268 1152 36 18250 9130 10 35 40 45 50 55 60 65 70 (f) −2699 −2698 −2697 −2696 −2695 −2694 −2693 −4185 −4184 −4183 −4182 −4181 X (m) Y (m) LF−4 LF−37 GDC−6 GDC−2 OS−19 OS−15 DX−38 OS−23 DX−29 DX−28 DX−77 DX−23 DX−59 CA−48A−2 CA−52−11 PG1 PG2 PG3 PG4 PG5 PG6 GDC−18 CA−117A−19 LF−2 GDC−53−13 SB−15 OS−11 OS−12 DX−26 CMHC−6 DX−47 GDC−53A−13 CA−87A−2 IG1 IG2 IG3 2268 1152 36 18250 9130 10 X10 3 35 40 45 50 55 60 65 70 Gpa Figure 6.29: Young modulus distribution in NTR horizon at The NW Geysers (a)2005, (b)2006, (c)2007,(d)2008, (e)2009, (f)2010 114 6.10 Concluding Remarks To estimate the fracture network properties from the velocity models, we initially enhanced the resolution of velocity models extracted from tomographic inversion. Then, we simultaneously interpret the seismic velocity volumes with other estimated volume to successfully locate and characterize the fractured area with higher fracture density and the propagation of a fracture network in targeted zones–with the aim of laboratory measurement of rock properties. We were able to identify the NW trend of the regional fracture network and the zones having higher fracture density within the NTR. Within the HTZ, it is possible to identify fracture networks that penetrate the NTR and move into the HTZ. We have also demonstrated how integrating with production= injection data and fuzzy clustering can help us clarify our hypothesis, which relates the velocity and stress anomalies to the propagating fracture network. Moreover, we demonstrated that integrating the passive seismic tomography with density information allows us to detect the space-time dependency of elastic properties in response to local variations of fluid pressure or fracture creation. It was then pos- sible to identify areas where fractures are proposed to have propagated over time. In addition, the use of multiple properties for characterization can improve the overall pro- cess, providing a valuable tool to cross-validate observations and improve on the initial interpretations, as shown in this chapter. The time lapse stress and rock property deter- minations presented in this article may also be applicable to other geothermal reservoirs with sufficient microseismicity and time-variant high resolution velocity models. Finally, we believe that the methodology proposed in this chapter provides a use- ful tool for long-term improvements to well spacing plan, well design, and comple- tion design at The GGF. This technique eliminates errors in locating the fractured areas and can help us target the stimulated area for future development plans. Furthermore, 115 expanding this technique to other stimulation projects, or analyzing them in real-time microseismic monitoring, can help us optimize the stimulation stages in cost and time, and better identify targets for hydraulic fracturing jobs. Conducting this technique also helps companies plan field development and increase their success rate from drilling and fracturing. 116 −2699 −2698 −2697 −2696 −2695 −2694 −2693 −4185 −4184 −4183 −4182 −4181 X (km) Y (km) LF−4 LF−37 GDC−6 GDC−2 OS−19 OS−15 DX−38 OS−23 DX−29 DX−28 DX−77 DX−23 DX−59 CA−48A−2 CA−52−11 PG1 PG2 PG3 PG4 PG5 PG6 GDC−18 CA−117A−19 LF−2 GDC−53−13 SB−15 OS−11 OS−12 DX−26 CMHC−6 DX−47 GDC−53A−13 CA−87A−2 IG1 IG2 IG3 2268 1152 36 18250 9130 10 3.5 4 4.5 5 (a) −2699 −2698 −2697 −2696 −2695 −2694 −2693 −4185 −4184 −4183 −4182 −4181 X (km) Y (km) LF−4 LF−37 GDC−6 GDC−2 OS−19 OS−15 DX−38 OS−23 DX−29 DX−28 DX−77 DX−23 DX−59 CA−52−11 PG1 PG2 PG3 PG4 PG5 PG6 GDC−88−12 GDC−88−12 LF−23 LF−23 GDC−18 CA−117A−19 LF−2 GDC−53−13 GDC−88−12 GDC−88−12 SB−15 OS−11 OS−3 OS−12 DX−26 LF−23 LF−23 CMHC−6 DX−47 GDC−53A−13 CA−87A−2 IG1 IG2 IG3 2268 1152 36 18250 9130 10 3.5 4 4.5 5 (b) −2699 −2698 −2697 −2696 −2695 −2694 −2693 −4185 −4184 −4183 −4182 −4181 X (km) Y (km) LF−4 LF−37 GDC−6 GDC−2 OS−19 OS−15 DX−38 OS−23 DX−29 DX−28 DX−77 DX−23 DX−59 CA−52−11 PG1 PG2 PG3 PG4 PG5 PG6 GDC−88−12 GDC−88−12 LF−23 LF−23 GDC−18 CA−117A−19 GDC−53−13 GDC−88−12 GDC−88−12 SB−15 OS−11 OS−3 OS−12 DX−26 GDC−36−18 LF−23 LF−23 SB−25 CMHC−6 DX−47 CA−73B−12 GDC−53A−13 CA−87A−2 IG1 IG2 IG3 2268 1152 36 18250 9130 10 3.5 4 4.5 5 (c) −2699 −2698 −2697 −2696 −2695 −2694 −2693 −4185 −4184 −4183 −4182 −4181 X (km) Y (km) LF−4 LF−37 GDC−6 GDC−2 OS−19 OS−15 DX−38 OS−23 DX−29 DX−28 DX−77 DX−23 DX−59 CA−52−11 PG1 PG2 PG3 PG4 PG5 PG6 GDC−88−12 GDC−88−12 LF−23 LF−23 GDC−18 CA−117A−19 LF−2 Curry−85−13 GDC−32A−13 GDC−53−13 GDC−88−12 GDC−88−12 SB−15 OS−11 OS−12 DX−26 GDC−36−18 LF−23 LF−23 SB−25 CMHC−6 DX−47 CA−73B−12 GDC−53A−13 CA−87A−2 CA−51A−11 IG1 IG2 IG3 2268 1152 36 18250 9130 10 3.5 4 4.5 5 (d) −2699 −2698 −2697 −2696 −2695 −2694 −2693 −4185 −4184 −4183 −4182 −4181 X (km) Y (km) LF−4 LF−37 GDC−6 GDC−2 OS−19 OS−15 DX−38 OS−23 DX−29 DX−28 DX−77 DX−23 DX−59 CA−52−11 PG1 PG2 PG3 PG4 PG5 PG6 GDC−88−12 GDC−88−12 LF−23 LF−23 CA−117A−19 Curry−85−13 GDC−32A−13 GDC−88−12 GDC−88−12 SB−15 OS−11 DX−26 GDC−36−18 LF−23 LF−23 SB−25 CMHC−6 DX−47 CA−73B−12 CA−87A−2 CA−51A−11 IG1 IG2 IG3 2268 1152 36 18250 9130 10 3.5 4 4.5 5 (e) −2699 −2698 −2697 −2696 −2695 −2694 −2693 −4185 −4184 −4183 −4182 −4181 X (km) Y (km) LF−4 LF−37 GDC−6 GDC−2 OS−19 OS−15 DX−38 OS−23 DX−29 DX−28 DX−77 DX−23 DX−59 CA−48A−2 CA−52−11 PG1 PG2 PG3 PG4 PG5 PG6 LF−23 LF−23 CA−117A−19 LF−2 Curry−85−13 GDC−32A−13 SB−15 OS−11 DX−26 GDC−36−18 LF−23 LF−23 SB−25 CMHC−6 DX−47 CA−73B−12 CA−87A−2 CA−51A−11 IG1 IG2 IG3 2268 1152 36 18250 9130 10 3.5 4 4.5 5 (f) −2699 −2698 −2697 −2696 −2695 −2694 −2693 −4185 −4184 −4183 −4182 −4181 X (m) Y (m) LF−4 LF−37 GDC−6 GDC−2 OS−19 OS−15 DX−38 OS−23 DX−29 DX−28 DX−77 DX−23 DX−59 CA−48A−2 CA−52−11 PG1 PG2 PG3 PG4 PG5 PG6 GDC−18 CA−117A−19 LF−2 GDC−53−13 SB−15 OS−11 OS−12 DX−26 CMHC−6 DX−47 GDC−53A−13 CA−87A−2 IG1 IG2 IG3 2268 1152 36 18250 9130 10 X10 3 3.5 4 4.5 5 Mpa Figure 6.30: Normal stress distribution in NTR horizon at The NW Geysers (a)2005, (b)2006, (c)2007, (d)2008, (e)2009, (f)2010 117 Chapter 7 Modeling and Simulation of Hydraulic Fracturing using MDFN model We have demonstrated the importance of including different methods in microseismic monitoring of the fracture characterization procedure in unconventional reservoirs. We have presented, up to now, a robust workflow that utilizes most of the information con- tent of microseismic events, which has advantage over current applied technology of being both comprehensive and flexible. In this chapter, we use that information to model and simulate the complex fracture network using an innovative approach called theMDFN method. 7.1 Background and Objective Advances in hydraulic fracturing techniques in the past few decades have significantly improved unconventional resources development. Accurate descriptions and modeling of hydraulic fractured reservoirs have been a hindrance in these years. Various models have been presented to demonstrate the stimulated reservoir based on the degree of complexity of the created fracture network and available information at the site. The final objective of these models is to create a reliable estimation of the associated porosity and permeability of these reservoirs in order to accurately match the history and to reliably predict the future performance. 118 Most of these models involve simplistic assumptions regarding geometry and trans- port properties, which result in a failure to show the predicted behavior of a complex fracture network. They assumed an ideal, planar fracture model to represent the stimu- lated area. However, geologic discontinuities such as natural fractures, joints, bedding planes, and stress contrast make this assumption unrealistic. For instance, Wright et al. (1970); Fisher and Wright (2002); Maxwell et al. (2002) reported highly complex frac- ture growth behavior in McClure Shale 1 in North Shafter field, Strawn Limestone in West Texas, and Barnett Shale after hydraulic fracturing job. Moreover, Warpinski and Teufel (1987); Warpinski et al. (1993) revealed this complex fracture growth and their causes through comprehensive mine-back experiments, core analysis, and laboratory measurements for various geologic discontinuities. They concluded that hydraulic frac- turing has much more complex behavior than is generally accepted. In a laboratory experiment, Beugelsdijk et al. (2000) also investigated these interactions on hydraulic fracture propagation. They observed hydraulic fracture complexity reduction by apply- ing higher flow rates and more viscous fluid as the horizontal stress difference was increased. They also claimed that complexity may result in the arrest of hydraulic frac- ture propagation, fluid flow into discontinuities, and the creation of multiple fractures and fracture offsets, which can reduce fracture length and width. On the other hand, Fisher and Wright (2002) claimed that complex behavior can create larger surface areas that potentially contribute to higher production. Figure 7.1 shows visualizations of a complex fracture network by Fisher and Wright (2002); Warpinski and Teufel (1987) Although it has been used to interpret the field results, this complexity is seldom con- sidered in modeling the hydraulic fracturing process. In this chapter, we are going to use microseismic data to identify the degree of complexity of the created fracture network; 1 McClure Shale in naturally fractured quartz phase siliceous shale. 119 however, we mainly focus on modeling this complex growth with higher reliability and less cost. As of now, various methods have been proposed for describing and modeling a com- plex fracture network, mostly in naturally fractured reservoirs. For instance, a dual porosity dual permeability (DPDP) model is widely used. This method is only based on smoothing the heterogeneities with assuming averages fracture porosities and perme- abilities over tens to hundreds of feet. This model is fast and does not require knowledge of the exact location and properties of each individual fracture(Ghods, 2012). However, it ignores the complexity of the flow behavior, which may control the recovery, and requires considerably long history to reliably adjust reservoir properties to forecast the future. Therefore, it cannot be used as an initial model to predict the reservoir behav- ior or to optimize the stimulation jobs real-time. In addition, DPDP cannot model the conductivity of individual fractures to match the strongly anisotropic fracturing. Discrete fracture model (DFN) is more rigorous and robust in capturing the physical properties of the fractured rock and can adapt the complexity of the flow behavior by modeling the flow in each individual fractures. This method is mainly based on local grid refinements or unstructured griding. This model is slower than DPDP and requires vast amounts of data with time and space resolution to populate a reservoir model with a reasonable number of fractures and their measured properties. Figure 7.2a and 7.2b shows the workflow for creating a DFN model and its com- mon data requirement. This figure shows that various fracture properties, such as length, aperture, conductivity, orientation, and porosity, should be collected for creating a DFN model. Currently, most of these properties are obtained from different source of data types and different experts, which makes this simulation extremely costly (Figure 7.2b). For instance, storage aperture can be obtained from image and conventional logs; 120 hydraulic aperture can be measured with respect to well tests and interference obser- vations; size and orientation can be estimated using outcrops; and permeability can be obtained from well test analysis and mud loss data. Although values of these properties are measured near accuracy, they may have considerable uncertainty because of censor- ing, length bias, truncation, and reservoir condition. For instance, outcrops may reveal the pattern and structure of an inclusive natural fracture within the area of stimulation, but due to stress and history, they may not be identical to the ones in the reservoir. More- over, fracture azimuth and dip varies spatially and vertically throughout the hydraulic fracture stimulation jobs or in natural fracture patterns with respect to the local stress field. The local stress field is also impacted by production, injection, or stimulation of reservoir which result in re-orientation of fracture (Wright et al., 1970). Therefore real monitoring of orientation or other fracture properties by using microseismic data or tilt meter is preferable to using only outcrops, logs, or cores. In addition, acquiring most fracture properties using only microseismic data significantly reduce the cost and time of analysis. Next, we need to define the distribution for each of these individual properties (Fig- ure 7.2c). For instance, Gale et al. (1991) reported that fracture orientation, length, and spacing follow the log-normal distribution on the outcrop analysis of Monterey formation. However, the average and range values differ in different clusters of frac- ture network in the same rock type. Then, we used these fracture statistics collected from outcrop or other resources to create the 2D and 3D stochastic fracture network model (DFN) as an input for simulation (Figure 7.2d). One of the main disadvantages of this type population is extrapolating the fracture properties collected from outcrops, well logs, and cores to the stimulated reservoir volume, which adds another degree of uncertainty in computing the average flow properties in the large-scale reservoir mod- els. However, estimating properties from microseismic data and populating data with 121 the inherent property of microseismic data can prove advantageous, because microseis- mic monitoring is able to measure these properties at distances of several hundred feet away from the wellbore over entire stimulated reservoir volume (SRV). Constraining the fracture model to microseismic derivative information such as frac- tures location, size, shape, and orientation implies significant improvement in model performance and can be supported by the real time data (Williams-Stroud, 2008; Warpinski et al., 1996; Barree et al., 2002; Mayerhofer et al., 2005). Here we intro- duce a modified DFN model not only to constraint the model but also to create it just by utilizing microseismic data. We have developed a MDFN (Microseismic based Discrete Fracture Network) method with the benefits of a DFN model, but eliminates most of discussed disadvan- tages. An MDFN model is only dependent on accurate and extensive microseismic data analysis, as discussed in this manuscript. Therefore, it enables us to simulate the stimulated reservoir in the absence of other information or for the purpose of reducing cost and time. Furthermore, it is able to validate other collected data upon availability. Finally, it imposes another important constraints on the traditional DFN model, which increases the visual similarity between the model and the real network over stimulated reservoir volume (SRV). 7.2 Microseismic based Discrete Fracture Network (MDFN) Figure 7.3 shows the workflow for creating an MDFN model. A discrete fracture net- work model can be created using just microseismic data to define fracture set parameters such as size, orientation, density, and fracture pattern. In section 2.4 (multiplet analy- sis) and 2.5 (shear wave splitting), we have shown methods for estimating the fracture 122 density and orientation from microseismic data. We have also defined Young’s modulus volume in chapter 6, which is another indicator for fracture density; see section 6.2.2 and 6.4.2. In this section, we will describe how to obtain fracture size, aperture, and population method from microseismic data. Table 7.1 lists the fracture properties and their associated methods for measuring them from microseismic data. Table 7.1: Obtaining fracture properties using microseismic data for MDFN modeling Size Density Orientation Aperture Population SRV Shear Wave Splitting X X Multiplet Analysis X X b-value X X Fractal Analysis X X Tomographic inversion X a X X Cloud of Microseismicity X b X b X a Qualitative b For the entire fracture network 7.2.1 Estimating Fracture Aperture We are able to quantitatively map the fracture aperture from microseismic data. Using the methodology presented earlier 2 , we define normal stress distribution from micro- seismic data. As discussed in section 6.4.1 and 6.9, the change of normal stress can directly indicate the fracture opening which affects the fluid flow path and regime. Rutqvist (2002) estimated the fracture aperture based on a cubic-block conceptual model and a parallel plate fracture flow model (Equation 7.1; Figure 7.4). b =b r +b m =b r +b max (exp( n )) (7.1) 2 Chapter 6 123 Where n is normal stress,b r is residual aperture,b m is the mechanical aperture,b max is maximum mechanical aperture in the medium, is a coefficient of curvature with value of 1:1 10 6 when aperture is in m and stress is in MPa. b r and b max may be determined based on priori knowledge about the formation characteristics whether from previously reported data 3 or expert experience. For instance, in naturally fractured reservoirs, these value are 0.001 and 0.1 inch, successively and in hydraulic fracturing b max may increase to 0.4 inch, according to the reported data from Monterey formation, Barnett shale, and Lost Hills field (Gale et al., 1991; Fast et al., 1994; Gale et al., 2007). In this study, we used the core CT scanner measurements by He (1998) at the GGF to estimate the set of parameters. Figure 7.5 shows the distribution of aperture in their laboratory measurements. Therefore, we imply b r = 50m( 0:002in) and b max = 2500m( 0:1in) into our normal stress model to derive the aperture volume at the GGF. Figure 7.6 shows the distribution of fracture aperture in NTR horizon at NW GGF for 2005 and 2010. 7.2.2 Degree of Complexity and Fracture Pattern In this section, we identify the degree of complexity of a created fracture network using microseismic data. We impose microseismic data fractal dimension as a new constraint on our fracture network to make it more consistent with underlying physics and field evidence and to increase its visual similarity to the real network. There are two major differences between modeling natural fractures and hydraulic fracturing and their complexities. Hydraulic fracturing occurs when the effective stress is reduced sufficiently by compressive forces, which are an increase in the pressure of fluids within the rock. Therefore, the minimum principal stress becomes tensile and exceeds the tensile strength of the material. This process usually does not create a large 3 image logs, cores, or outcrops 124 number of fractures in heterogeneous rock Sahimi (1995). While natural fracturing can be created everywhere in the rock, hydraulic fracturing can only occur at the weakest points that contact the injection fluid. Therefore, we might model the hydraulic fracturing as single planar. But, according to our discussion in section 7.1, we observed an interaction of hydraulic fractures and geologic discontinuities in many case studies. Therefore, we might be dealing with a complex fracture network here as well. The method presented in this section is able to identify the degree of complexity using microseismic data to accurately estimate the fluid flow within the fracture network. Considering the fractal geometry and the dimen- sion of a fracture network seems to have a significant impact on the simulation results, which may provide further information for creating DFN in a traditional way. It has been shown by several authors that faults and fracture networks of rock are fractal 4 . Therefore, they used fractal geometry to describe and interpret the true physics and behavior of the fractured medium. In particular, Acuna and Yortsos (1995) created a fractal fracture network based on their observation from field outcrops. Their numerical simulation behaves according to the analytical prediction of pressure transient analysis. Ferer et al. (1994) imposed fractal behavior constraint to model the natural fracture net- work with the dimension ofD f = 1:4 0:1 in a two-dimensional pattern. He reported that fractal based simulation increases the visual similarity of models to the real net- work. (Gale et al., 1991) also claimed that anisotropy to flow, interconnectivity,and relative flow rate are highly dependent on fracture geometry and its statistical informa- tion. Finally, Fomin et al. (2003) carried out a series of computations, which indicate that the connectivity of the fracture network is greatly affected by the fractal dimension of the fracture network. 4 see section 4.1 125 According to our discussion in chapter 4 of this manuscript, we may define the guideline for using microseismic data fractal dimension (D f ) to reveal the structure of the associated fracture network. Therefore, we use a fractal interpretation of the microseismic data to impose another constraint to MDFN model. Microseismic data with a fractal dimension ofD f ' 1:20:1 in 2D orD f ' 20:1 in 3D, are consistent with data for the surface of a single major fault, and represent the single large fracture that spans the stimulated volume. Therefore, a single planar hydraulic model will be only appropriate in this case. Microseismic data with a fractal dimension ofD f ' 1:89 0:05 in 2D orD f ' 2:5 0:1 in 3D , represent a fracture network similar to a percolation cluster. Therefore, we have the highest complexity, and full interaction with geologic discontinuities where both natural fractures and Hydraulic fractures contribute to the flow. Microseismic data with a fractal dimension of D f ' 1:5 0:1 in 2D or D f ' 2:3 0:1 in 3D , reveal a fracture network pattern similar to the observed natural fracture pattern on outcrops 5 . Therefore, hydraulic fracture follows the natural fracture network pattern, and hydraulic fracture stimulation mainly inflates the natural fractures or creates most fractures on existing planes of weakness. It is interesting to note that most of the current complex fracture models have a fractal dimension equal to the space dimension of a box or cube 6 which overestimates the volume of fracture network. We recommend estimating the fractal dimension from microseismic clusters near wellbore to the farther one with an appropriate density of events. Any change in fractal dimension in scale may represent discontinuity in the fracture network which may result in reducing the estimation of stimulated reservoir volume; see chapter 4 for more details. 5 For more information see Ferer et al. (1994); Wilson (1996); Babadagli (2001); Sahimi (1995) 6 2 in 2D and 3 in 3D 126 The less reliable the information determining the probability distributions of fracture properties, the less reliable the fracture network. However, it is impossible to determine the reliable information about all aspects of the real fracture network. Here, we assume the self-similar fractal behavior (if valid) with dimensionD f to populate the fractures properties into the stimulated reservoir volume in a manner more consistent with under- lying physics and field evidence 7 . For instance, equation 4.3 for microseismic location can be modified to define the length distribution in a fracture network N(l) =Cl D f (7.2) WhereN(l) is the number of fractures longer than a given fracture length ,l is fractal length , andC is fracture density; see Sahimi and Yortsos (1990); Sammis et al. (1991, 1992); Acuna and Yortsos (1995); Fomin et al. (2003) for more detailed information. We will discuss finding the appropriate range of fracture length in the next section. 7.2.3 Estimating Fracture Size We are able to use microseismic event magnitudes to control the fracture size. Kanamori and Anderson (1975) reported the following relationship among moment magnitude (M o ), rigidity or Shear modulus (), rupture areaS, and displacementd for earthquakes. M 0 =Sd (7.3) Williams-Stroud (2008) also calculated the size of fracture plane from seismic moment of microseismic events using the same relationship. Wells and Coppersmith (1994) theorized an empirical relationship among seismic moment magnitude, rupture length, 7 Microseismic distribution Tafti et al. (2013) and laboratory measurements(Sahimi and Yortsos, 1990) 127 width, and area through historical studies of earthquakes worldwide. Table 7.2 summa- rizes their remarkable work. It should be noted thata andb coefficients vary for different slip types and failure mechanisms. Table 7.2: Empirical relationship among seismic moment magnitude, rupture length, width, area, and average displacement (Wells and Coppersmith, 1994) Relationship with Moment Magnitude log(P ) =a +bM a a b Rupture length [-2.57, -1.51] [0.4 , 0.62] Rupture width [-1.81, -0.64] [0.25 , 0.44] Rupture area [-4.35, -2.37] [0.74 , 1.04] a P can be defined as L-rupture length (km); W-rupture width (km); A-rupture area (km 2 ). Although, there is a notable difference between tectonic earthquake and induced microseismicity, as we discussed earlier, we might still use these empirical relationships for microseismic events to make an estimate of the associated fracture size. To adapt them in modeling the hydraulic fracturing job, we imply the rupture length relationship to Barnett Shale microseismic data to find the correspondinga andb value. Fisher et al. (2004); Mayerhofer et al. (2006); Bruner and Smosna (2011) reported the total length of 25000 ft in any treatment at this area, and the hydraulic fracture length varies between 4ft to 2000 ft. Gale et al. (2007) also reported a natural length between 0.05 to 100 ft. We apply two extreme values ofa andb for calculating the length of the Wells and Coppersmith (1994) equation to match the Barnett shale data. Therefore we can define equation 7.4 to match the reported length at Barnett shale and other hydraulic fracture jobs using the lower end values of an empirical relationship of Wells and Coppersmith (1994). L = 3280:8 10 1:51+0:4M (7.4) 128 Where rupture length L is in ft and M is the size of microseismic event. Figure 7.7 demonstrates the frequency magnitude distribution of microseismic events for differ- ent stages of hydraulic fracturing stimulation at the Barnett shale reservoir. Figure 7.8 applies this equation to this data to estimate the frequency–fracture length distribution during different stage of stimulation. The range of fracture length is 7 to 40 ft, which represents the high segmentation of observed large fractures in this data set. The rup- ture length of 1600 ft corresponds to a magnitude 3 event, which is considered as large microseismic event. This magnitude has been reported in a few hydraulic fracturing jobs associated with long fracture creation. It should be noted that rupture length is always less than or equal to the fracture= fault length. However, the rupture length is assumed to be equal to the fracture length when we are creating new fractures, not triggering the old ones. The former is the case for hydraulic fracturing. There is another uncertainty in this approach, whereby micro- seismic events happen more than once on the same fracture, which results in an overes- timation of the number of fractures within the stimulated reservoir volume. Therefore, the number fractures should be reduced to match the fractal pattern (Equation 7.2). To create a discrete fracture network, we use the range of fracture length estimated in this section with the population obtained in section 7.2.2 7.3 MDFN Case Studies Heretofore, we have presented how to use only microseismic data to create a discrete fracture network to model fractured reservoirs including shale oil and gas formations. In this section, we illustrate how to apply the MDFN method for history matching a shale gas reservoir in a Barnett shale formation to real field case studies. 129 The Mississippian-age Barnett Shale at Fort Worth Basin has a thickness between 200 to 800 feet. The gas is produced from organic-rich shale, non-siliciclastic rocks with permeability between 10 to 100 nanodarcies (nd). Due to this extremely low permeabil- ity, production is highly dependent on hydraulic fracturing job.(Fisher and Wright, 2002; Maxwell and Urbancic, 2002; Mayerhofer et al., 2006). We use unstructured griding and nodal material balance techniques to apply our MDFN method to these case studies. We use FracGen and NFlow 8 to simulate the fluid flow in our MDFN model consistent with only observed microseismic data. Because we have not been able to find any tomographic inversion or shear wave splitting data for hydraulic fracture patterns 9 , thus we cannot follow all the steps proposed in our method to fully characterize the fracture network. Here, we only build our network from public data provided by Maxwell et al. (2002); Mayerhofer et al. (2006). Table 7.3 shows the reservoir constant input parameters for the NFlow simulator. We identify the range of fracture length from b-value analysis (section 7.2.3); next we identify the degree of complexity with fractal analysis of microseismic data (section 7.2.2). We only change fracture aperture to match the history of production with bottom hole pressure constraint. We will show how these two steps can help us to history match and reliably predict the future behavior of an hydraulic fractured reservoir. 7.3.1 Real field Case Study 1: Simple pattern This case is a simple fracture pattern observed at Barnett shale during a hydraulic frac- turing job. Figures 7.9 shows its microseismic distribution, in which nearly linear geom- etry is observed withDf' 1. It also shows the associated normal daily production rate 8 National Energy Technology Laboratory (2013) 9 These technique are primarily designed for surface monitoring. Although, surface monitoring has been implemented in some hydraulic fracturing job, downhole monitoring is a common practice for mon- itoring them due to their higher accuracy; for more detail see section 2.1 130 Table 7.3: Set of reservoir parameters Parameters Value Depth,ft 7000 Initial Pressure,psi 6319 Porosity, % 6 Permeability,nd 23 Temperature, F 200 Gas Gravity 0.572 after treatment. According to our guidelines presented in section 7.2.2 of this chapter, this fractal dimension may represent a single large fracture that spans the stimulated volume. Therefore, a single planar hydraulic model will be sufficient to model the reser- voir. Figure 7.10 shows our model in FracGen and simulated production data, which nearly match the actual production 10 . Therefore, the traditional approach for modeling hydraulic fracturing works where we observe a simple pattern. We are also able to use analytical models 11 for improving the speed of simulation if desired. In this case, we observe high initial gas production rates from the storage in hydraulic fracture. Then, after first shut in, the fracture was filled again with gas from diffusion and flow of gas molecules through the shale matrix into the fractures. Opening the well again resulted in flowing the gas stored in the fractures flows toward the well at high velocities–results in a rapid decline of production. Next, we observe the pressure depletion with a lower rate, which is from the production of diffused gas from the shale rock matrix into the fracture. Finally, steady and low gas production rates caused by the diffusion of gas in the matrix and its desorption from organic content can be observed. Figure 7.11 demonstrates these steps in simple cartoons. 10 We do not have access to actual bottom hole pressure; therefore we assume common sense values, as shown at the bottom of the figure 11 see Shojaei et al. (2013) for more details on an analytical solution to single planar fracture modeling 131 Although other steps presented through this chapter may results in desired fractures properties to characterize and model the fracture network, exact knowledge of individ- ual fractures may not be required to match the history of production. Mayerhofer et al. (2005) also demonstrated that a wide range of hydraulic fracture and reservoir variables result in a nearly accurate production match for a single planar fracture model and that the answer is not unique. However, hydraulic fracture optimization and stage planning may not be possible without the integration of all information gathered from microseis- mic data, as discussed in this manuscript. 7.3.2 Real field Case Study 2: Complex Pattern In this section, we study the case of a complex fracture pattern observed at Barnett shale during a hydraulic fracturing job. Figures 7.9 shows its microseismic distribution and its associated unusual daily production rate. This complex behavior, represents the high degree of interaction between hydraulic fracture and natural fractures. The hydraulic fracturing job creates new fractures and inflates the natural fractures in a complex pat- tern. Therefore, a fractal dimension of 1:82 may represent its complexity, as discussed in section 7.2.2. This guideline indicates that a single planar hydraulic model and tra- ditional way to model the hydraulic fracture will not be sufficient to model this fracture network. Thus, we create the fracture network, shown in Figure 7.13, to match the frac- tal dimension of induced seismicity. The range of fracture length is also extracted from b-value analysis, as discussed in section 7.2.3. Figure 7.14 shows our model in FracGen and simulated production data, which nearly match the actual production 12 . We are not able to match the 70-day production period after the third shut-in of the well with good 12 We do not have access to actual bottom hole pressure; therefore we assume common sense values, as shown at the bottom of the figure 132 accuracy because of limited access to other reservoir parameters. However, a model is able to match the rest of the production rate with reasonable error after the forth shut-in. In this case, we observe high initial gas production rates from the storage in hydraulic fractures, followed by d the iffusion and flow of gas molecules through the shale matrix into the fractures in each stage of production. The depletion rate of production is much lower than case 1 (simple pattern), in which production is supported by a cluster of fracture network. Therefore, a percolation cluster is reasonable interpretation for this fracture network which has fractal dimension of 1.89 (2D)–close to our designed net- work. This type of complex behavior confirms the importance of applying the various steps presented throughout this manuscript to fracture characterization and modeling using microseismic data. This result also matches the Warpinski et al. (1993) recommendation for modeling the complex fracture network. After extensive laboratory measurements, they recom- mended applying shorter fracture lengths, and having tip zone effects and multiple frac- ture zones at the offset for the modeling and analysis of an hydraulic fracturing job. 7.4 Summary In this chapter, we have introduced a comprehensive methodology for creating a discrete fracture network using only microseismic data, called MDFN. This method enables us to create the model using only microseismic data in the absence of other information or for the purpose of reducing the cost. It can also validate other available data from cores, logs, or outcrops, then populate them within the stimulated reservoir volume using the fractal dimension of microseismic data to generate a more realistic appearance and to reduce uncertainties. 133 Moreover, our proposal for including fractal behavior in the hydraulic fracture model, based on the observed microseismicity, is a good approximation of the real pat- tern of complex networks. Our proposed technique was tested on two case studies at Barnett Shale formation with both simple and complex fracture networks. At the very least, we have imposed another constraint on other fracture modeling approaches in a manner more consistent with underlying physics and field evidence, which increases the visual similarity between our networks and the real network over the simulated volume. 134 CORE HOLES o 10 ft ~ o 3 m o FRACTURE INTERCEPT CORE HOLES SIDE VIEW I. WIDTH OF .1 ZONE OF FRACTURING PLAN VIEW MINE BACK DRIFT Fig. 3-Results of exploratory coring showing large num ber of fracture intercepts. Fig. 1 is a photograph and illustration of fracture be havior at one location where many of these elements could be observed together. In naturally fractured regions, mul tiple stranding was predominant; a single fracture was sel dom observed. These often originated from natural fractures that were filled (and probably opened) by frac ture fluid. Fractures were usually offset, as seen at the top of the photograph, and sometimes two or three frac tures initiated at these offsets. The volcanic rubble zone, which is probably similar to a conglomerate with respect to property variations, also was a site of multiple strand ing and considerable fracture meandering. Fig. 2 is a photograph and sketch of a fracture initiat ing from the borehole. Many of the cross joints are filled with the cement fracturing fluid, and nearly all of the cross joints offset the hydraulic fracture to some degree. The effect on leakoff, pressure drop in the hydraulic fracture, and proppant transport would certainly be severe in such a case. Even these figures do not adequately describe the com plexity of the fracture system. Information obtained from exploratory coring of the induced fracture(s) in the welded tuff provide a better picture. Fig. 3, a plan view and side view (parallel to the fracture) of the coring results, shows the locations where fractures filled with cement were found. Several fractures are obviously propagating side by side in this region. Thus it is clear that the fracture Journal of Petroleum Technology, February 1987 PROPAGATION DIRECTION ~ Fig. 4-Authors' visualization of results of fracture treat ment in jointed rock mass. might be better termed a "zone of fracturing" that spans 20 to 30 ft [6 to 9m]. Our best representation of the zone of fracturing from mineback and coring is shown in Fig. 4. Generally, we see that fractures are offset when they cross the joints, and often two or more fractures initiate from the joints. Some of these fracture strands die out in a short distance, while others persist for long distances. These fractures divide and coalesce many times along the fracture length. We should stress that the evolution, scale, and total ef fect of this zone of fracturing depends to a large degree on the ancillary parameters. In this case, the orientation of the joints with respect to the maximum horizontal stress direction (or fracture orientation) varied from about 30 to 90° [0.5 to 1.6 rad] (there was more than one joint set). The minimum stress was about 100 psi [690 kPa] in this jointed region, apparently still showing the residual of the stress system that caused the fracturing. The maxi mum horizontal stress in the welded tuffs is not as cer tain. In this area, maximum horizontal stresses are typically 300 to 400 psi [2 to 2.8 MPa] above the mini mum stress over a large range of stresses in the ash-fall tuffs. An overcoring measurement in the welded tuff, however, indicated that the maximum stress may be as great as 800 psi [5.5 MPa], 17 although we think this is much too high. Permeability of the fractures may be as great as 1,000 darcies, 17 but some cross joints show no 211 PHOTO --- , o 12FT. L.....-L....... VITRle ZONE Fig. 1-Photograph and sketch of complex fracture be havior. Containment. One of the most striking results of the mineback experiments, and one that has been previously documented,6.13,14,16 is the overriding influence of the in-situ stress distribution on hydraulic fracture contain ment. Many different fracture experiments in different lo cations and geologic conditions have shown that the confinement resulting from a high-stress region is a first order effect, whereas interfaces, modulus or strength changes, fluid pressure gradients, and most bedding planes have only second-order effects on height growth. Stress contrasts on the order of 600 to 900 psi [4 to 6 MPa] have been shown to confine small test fractures completely, 6 210 PHOTO MAIN FRACTURE I ' o 12FT. ~ \ \ MAIN FRACTURE "'---t- NATURAL FRACTURES FILLED WITH GROUT Fig. 2-Photograph and sketch showing fracture offsets at joints. while smaller stress contrasts have been shown to restrict fracture height growth compared with growth in other directions. While most other factors have much smaller or negligible effects, we will show later that some dis continuities, particularly joints and unbonded bedding planes, may provide good containment by making frac ture growth inefficient. Effects of Joints. Even in the most homogeneous of the ash-fall tuff formations, the hydraulic fractures are ob served to diverge considerably from the usual picture of a planar feature; multiple stranding, fracture meander ing, and large-scale surface roughness are common oc currences. In the vicinity of geological discontinuities, the situation is often worse. Frequently there is nothing that can be called a planar fracture; instead, the hydraulic frac ture could be better described as a zone of multiple frac turing, sometimes 15 to 30 ft [5 to 10 m] wide. The most illustrative example of this complexity was observed in a fracture created with 15,000 gal [57 m 3 ] of colored Class A cement. 13 This experiment was de signed to test the effect of a material property interface on hydraulic fracture growth by initiating the fracture in the low-modulus ash-fall tuff below the interface with the much-higher-modulus welded tuffs. These welded tuffs are highly jointed, and when the hydraulic fracture penetrated through the interface into this jointed region, the effects were severe. A mineback of part of the frac ture and exploratory coring to characterize the rest of the fracture revealed important elements of the fracture be havior with respect to the joint system. Journal of Petroleum Technology, February 1987 PHOTO --- , o 12FT. L.....-L....... VITRle ZONE Fig. 1-Photograph and sketch of complex fracture be havior. Containment. One of the most striking results of the mineback experiments, and one that has been previously documented,6.13,14,16 is the overriding influence of the in-situ stress distribution on hydraulic fracture contain ment. Many different fracture experiments in different lo cations and geologic conditions have shown that the confinement resulting from a high-stress region is a first order effect, whereas interfaces, modulus or strength changes, fluid pressure gradients, and most bedding planes have only second-order effects on height growth. Stress contrasts on the order of 600 to 900 psi [4 to 6 MPa] have been shown to confine small test fractures completely, 6 210 PHOTO MAIN FRACTURE I ' o 12FT. ~ \ \ MAIN FRACTURE "'---t- NATURAL FRACTURES FILLED WITH GROUT Fig. 2-Photograph and sketch showing fracture offsets at joints. while smaller stress contrasts have been shown to restrict fracture height growth compared with growth in other directions. While most other factors have much smaller or negligible effects, we will show later that some dis continuities, particularly joints and unbonded bedding planes, may provide good containment by making frac ture growth inefficient. Effects of Joints. Even in the most homogeneous of the ash-fall tuff formations, the hydraulic fractures are ob served to diverge considerably from the usual picture of a planar feature; multiple stranding, fracture meander ing, and large-scale surface roughness are common oc currences. In the vicinity of geological discontinuities, the situation is often worse. Frequently there is nothing that can be called a planar fracture; instead, the hydraulic frac ture could be better described as a zone of multiple frac turing, sometimes 15 to 30 ft [5 to 10 m] wide. The most illustrative example of this complexity was observed in a fracture created with 15,000 gal [57 m 3 ] of colored Class A cement. 13 This experiment was de signed to test the effect of a material property interface on hydraulic fracture growth by initiating the fracture in the low-modulus ash-fall tuff below the interface with the much-higher-modulus welded tuffs. These welded tuffs are highly jointed, and when the hydraulic fracture penetrated through the interface into this jointed region, the effects were severe. A mineback of part of the frac ture and exploratory coring to characterize the rest of the fracture revealed important elements of the fracture be havior with respect to the joint system. Journal of Petroleum Technology, February 1987 2 C.A. WRIGHT, M.K. FISHER, B.M. DAVIDSON, A.K. GOODWIN, E.O. FIELDER, W.S. BUCKLER & N.P. STEINSBERGER SPE 77441 conductivity in these treatments. Due to the low permeability nature of the reservoir, it is imperative that extremely large fracture surface areas are created by the fracture treatments. The use of “light sand” or waterfrac treatments has considerably improved both the production performance and the economics in this reservoir. Because of its extremely low permeability, the drainage distance from the fracture face is very small. Introduction The classical description of a hydraulic fracture is a single bi- wing planar crack with the wellbore at the center of the two wings. However, almost all physical fracture verifications performed to date, from core-throughs to minebacks have proven this description to be oversimplified. Therefore, fracture mapping technologies that directly measure actual fracture geometry can provide insight into reservoir depletion dynamics and significantly help optimize reservoir management. Fractures can be categorized as simple (classical description), complex, or very complex and an illustration of how these fractures may look is pictured below. Simple Fracture Com plex Fracture Very Complex Fracture Figure 1 – Fracture Complexity Due to several factors including the presence of natural fractures, a fracture treatment in the Barnett is more likely to look like the “very complex” fracture description than the “simple” case – this allows a fracture “fairway” to be created during a treatment with many fractures in multiple orientations resulting in large surface areas potentially contributing to production. Numerous treatments have been mapped in the Barnett to gain a better understanding about how these fractures propagate. The primary (hydraulic) fracture orientation is NE to SW in this area and this has been verified from both surface tiltmeters and microseismic fracture mapping. Additionally, the natural fractures identified from borehole imaging surveys in this area are oriented orthogonal to the primary fracturing (NW to SE) and these natural fractures may be activated (opened) during a hydraulic fracture treatment. The length and width of the resulting fracture “fairway” is important in determining the area contacted by the fracture so that well location and spacing can be optimized. Because of the aforementioned small drainage distance from a fracture, the density of fractures within this fairway is very important. There may be opportunities for additional wells to be drilled in less densely fractured areas within a fracture fairway or for refracs to be performed that may extend the fairway or more densely populate it with new fractures. Fracture Mapping Technologies A combination of fracture mapping technologies was used to develop an accurate image of fracture growth in the Barnett. These complementary technologies allowed for a better understanding of the different aspects of fracture growth. Production results strongly correlated to the orthogonal fracture growth, visual in the Surface Tiltmeters as the NW component and in the Microseismic data as a wide band of seismic activity. Utilizing two independent mapping technologies allowed for a more complete image of the fracture “fairway”. Table 1 shows the component of fracture geometry determined from each of the mapping technologies. Table 1 – Barnett Fracture Geometry Measurements determined by various technologies • Surface Tiltmapping is a direct fracture diagnostic tool utilized on more than one thousand treatments per year to map the deformation caused at the surface of the earth by hydraulic fractures or dislocations in the subsurface. The tiltmeter is a very sensitive device, similar to an electronic carpenter’s level, that can sense changes in the gradient of displacement (or tilt) of as little as one part per billion. Surface deformation, measured by tiltmeter arrays, is used to directly determine the azimuth and dip of a hydraulic fracture and also the percent of treatment volume placed in each plane or orientation when fracture growth occurs in multiple planes 2 . About 30 surface tiltmeter sites were required in each of the mapping areas based on the depth of the fracture treatments, the expected fracture complexity and the large area encompassing the number of wells stimulated in each field. • Downhole Tiltmapping is a separate application of the same technology 3,4 . By ruggedizing the surface tiltmeter instruments and placing them in offsetting wellbores to the treatment wells, hydraulic fracture dimensions can be determined. Most of the treatments were monitored with Surface Downhole Tilt Tilt Microseismic Azimuth X X Height X (Tip) X (Wellbore) Length X Asymmetry X NW Volume Component X Figure 7.1: Authors’ visualization of complex fracture network from Warpinski and Teufel (1987); Fisher and Wright (2002) 135 DFN MDFN Collection of Fracture Geometric Data Size, Orientation, Density Image logs, Seismic facies maps, Outcrops, VSP, Well test data, Core, Seismic Collection of Fracture Data Size, Orientation, Density, Fractal Dimension Microseismic Data ,, y ,, sections, Tilt meter, Microseismic Create Fracture Property Values Permeability Pattern Define distribution for each property Mean, Standard deviation Type of Distribution Generate fractures sets using the Permeability Pattern Generate fractures sets using stochastic methods Generate fractures sets using the underlying Physics Reservoir Simulator Reservoir Simulator (a) Discrete Fracture Network Model Courtesy FracMan Technology Group (b) Discrete Fracture Generation Oi t ti  Stochastic Inputs & Deterministic Features Orientation Si e Size Intensity kh (c) Discrete Fracture Generation Oi t ti  Stochastic Inputs & Deterministic Features Orientation Si e Size Intensity kh (d) Figure 7.2: (a) Discrete fracture model workflow (b) Source of data for DFN (b) Defin- ing property distribution for fracture properties (d) Sample created network (Dershowitz and Doe, 2011) 136 DFN MDFN Collection of Fracture Geometric Data Size, Orientation, Density Image logs, Seismic facies maps, Outcrops, VSP, Well test data, Core, Seismic Collection of Fracture Data Size, Orientation, Density, Fractal Dimension Microseismic Data ,, y ,, sections, Tilt meter, Microseismic Create Fracture Property Values Permeability Pattern Define distribution for each property Mean, Standard deviation Type of Distribution Generate fractures sets using the Permeability Pattern Generate fractures sets using stochastic methods Generate fractures sets using the underlying Physics Reservoir Simulator Reservoir Simulator Figure 7.3: Microseismic based discrete fracture network (MDFN) model workflow k 2 /k i ≥ 0.1 Niches Drift Scale Test 200 ºC Permeability Stress σ 1 = σ i σ 3 σ 2 k 2 k 3 k i k 3 /k i ≈10-100 Fracture aperture Stress b max b r b i Drift Scale Test Niches Figure 4. Schematic of normal stress versus aperture relation (a) and calibration of stress- permeability function (b). 29 Figure 7.4: Conceptual model for aperture - normal stress relationship in fractured rocks Rutqvist (2002). 137 SECTION 4. RESUL TS 35 0 0.5 1 1.5 2 2.5 3 0 50 100 150 200 250 300 350 aperture (mm) number of fractures Figure 4.8: Ap erture distribution in Geysers sample Figure 7.5: Aperture distribution in GGF core samples from CT scan result (He, 1998) −4184000 −4182000 −4180000 −4184000 −4182000 −4180000 −2698000 −2696000 −2694000 −2698000 −2696000 −2694000 Aperture 2005 2013 Mar 07 15:27:10 Aperture 2005 Scale 1:40000 CA−45A−12 CMHC−2 CMHC−6 DX−19 DX−24 DX−47 DX−5 GDC−26 GDC−53−13 GDC−88−12 LF−2 LF−23 OS−11 OS−12 OS−16 SB−15 CA32−12 CA45−12 CA48A−2 CA52−11 Curry3 DX23 DX25 DX28 DX3 DX38 DX42 DX59 DX77 GDC2 GDC6 OS15 OS19 OS23 SB20 SB27 70 80 90 100 110 120 130 140 150 160 170 180 190 200 210 220 230 240 250 H−1.2kmNw Aperture2005 Injection Wells Production Wells (a) −4184000 −4182000 −4180000 −4184000 −4182000 −4180000 −2698000 −2696000 −2694000 −2698000 −2696000 −2694000 Aperture 2010 2013 Mar 07 15:32:55 Aperture 2010 Scale 1:40000 CA−45A−12 CMHC−2 CMHC−6 DX−19 DX−24 DX−47 DX−5 GDC−26 GDC−53−13 GDC−88−12 LF−2 LF−23 OS−11 OS−12 OS−16 SB−15 CA32−12 CA45−12 CA48A−2 CA52−11 Curry3 DX23 DX25 DX28 DX3 DX38 DX42 DX59 DX77 GDC2 GDC6 OS15 OS19 OS23 SB20 SB27 70 80 90 100 110 120 130 140 150 160 170 180 190 200 210 220 230 240 250 H−1.2kmNw Aperture2010 Injection Wells Production Wells (b) −4184000 −4182000 −4180000 −4184000 −4182000 −4180000 −2698000 −2696000 −2694000 −2698000 −2696000 −2694000 Aperture 2005 2013 Mar 07 15:27:10 Aperture 2005 Scale 1:40000 CA−45A−12 CMHC−2 CMHC−6 DX−19 DX−24 DX−47 DX−5 GDC−26 GDC−53−13 GDC−88−12 LF−2 LF−23 OS−11 OS−12 OS−16 SB−15 CA32−12 CA45−12 CA48A−2 CA52−11 Curry3 DX23 DX25 DX28 DX3 DX38 DX42 DX59 DX77 GDC2 GDC6 OS15 OS19 OS23 SB20 SB27 70 80 90 100 110 120 130 140 150 160 170 180 190 200 210 220 230 240 250 H−1.2kmNw Aperture2005 Injection Wells Production Wells m Figure 7.6: Aperture distribution in NTR horizon at The NW GGF (a)2005, (b)2010 138 Figure 7.7: frequency-magnitude distributions for different stages of stimulation in horizontal Barnett Shale well, (after Downie et al. (2010)) Figure 7.8: Calculated frequency-fracture length distributions for different stages of stimulation in horizontal Barnett Shale well, 139 SPE 77440 MICROSEISMIC IMAGING OF HYDRAULIC FRACTURE COMPLEXITY IN THE BARNETT SHALE 7 EW orientation. During the final stage, relatively fewer events were recorded, with most of the events in the lineation offset from the well. Therefore, the sequences of stages resulted in an extension of the previous fracture network, without of significant generation of new fractures. From this set of figures, it is clear that the fracturing is offset to the NW of the treatment well, which could be related to interaction with previously hydraulic fractures in neighboring wells. However, this is only speculative since images were not recorded for those treatments. Production Activity In order to compare the fracture effectiveness of various wells documented in this paper, plots of daily production rates are shown in Figures 15, 16, 17, and 18. Figure 15 shows the production for well ‘A’, showing relatively high initial production that is maintained at a more uniform rate relative to the other wells which have a more significant hyperbolic decline. This high production rate could be related to an extensive contact between the reservoir and the fracture network, as can be seen by the complexity of the fracture network depicted in Figure 1. Figure 16 shows the production rate for well ‘B’ (shown in Figure 3. In relative terms, the production rate is moderate. As shown in Figure 3, there is evidence that the treatment stimulated parallel fractures in a fracture network, at least close to the treatment well. However, the fractures are much less extensive and more closely spaced compared to the wider fracture network geometry found with well ‘A’, perhaps explaining the differences in production between the two wells. Figure 17 shows the production of well ‘C’, where the frac grew into the monitoring well. In this case, the production is very poor, indicative of all the examples of fracs that encountered a depleted zone around a neighboring monitoring Figure 14. Final stage of well ‘F’ stimulation. Figure 13. Second stage of well ‘F’ stimulation. Figure 15. Daily production for well ‘A’. Figure 16. Daily production for well ‘B’. Figure 7.9: Microseismic distributions for hydraulic fracture treatment where simple seismicity pattern is observed at Barnett Shale,D f = 1, with its associated production (after Maxwell et al. (2002)) 140 Figure 7.10: Fracture model and history match when microseismic distribution has D f ' 1:2 0:1 in 2D orD f ' 2 0:1 in 3D (a) Desorption of gas from organic content (b) Gas diffusion in ultra-tight matrix (c) Darcy flow in fractures Figure 1.7: Production mechanisms in shale gas reservoirs. 1989). High initial gas production rates from the storage in fractures, rapid decline of production rates when micro fractures and small pore volumes are depleted, and steady and low gas production rates caused by diffusion of gas in the matrix and its desorption from organic content can be observed in Figure 1.8. In this study, we propose a combined methodology aimed for assisted history match- ing of the numerical models of fractured reservoirs including shale gas formations. We show that our proposed methodology is fast, reliable and robust. It does not depend on discrete fracture networks that are based on local grid refinements or unstructured gridding, which make simulations painfully slow because of the large number of the grid blocks in the model and the complexity of the flow behavior. Furthermore, in assisted history matching processes, reservoir models are automatically adjusted and updated based on the errors observed in the prediction of measurements. If locally refined or 8 Figure 7.11: Gas production mechanisms from fractured reservoir in tight formation such as Shale(Ghods, 2012) 141 SPE 77440 MICROSEISMIC IMAGING OF HYDRAULIC FRACTURE COMPLEXITY IN THE BARNETT SHALE 7 EW orientation. During the final stage, relatively fewer events were recorded, with most of the events in the lineation offset from the well. Therefore, the sequences of stages resulted in an extension of the previous fracture network, without of significant generation of new fractures. From this set of figures, it is clear that the fracturing is offset to the NW of the treatment well, which could be related to interaction with previously hydraulic fractures in neighboring wells. However, this is only speculative since images were not recorded for those treatments. Production Activity In order to compare the fracture effectiveness of various wells documented in this paper, plots of daily production rates are shown in Figures 15, 16, 17, and 18. Figure 15 shows the production for well ‘A’, showing relatively high initial production that is maintained at a more uniform rate relative to the other wells which have a more significant hyperbolic decline. This high production rate could be related to an extensive contact between the reservoir and the fracture network, as can be seen by the complexity of the fracture network depicted in Figure 1. Figure 16 shows the production rate for well ‘B’ (shown in Figure 3. In relative terms, the production rate is moderate. As shown in Figure 3, there is evidence that the treatment stimulated parallel fractures in a fracture network, at least close to the treatment well. However, the fractures are much less extensive and more closely spaced compared to the wider fracture network geometry found with well ‘A’, perhaps explaining the differences in production between the two wells. Figure 17 shows the production of well ‘C’, where the frac grew into the monitoring well. In this case, the production is very poor, indicative of all the examples of fracs that encountered a depleted zone around a neighboring monitoring Figure 14. Final stage of well ‘F’ stimulation. Figure 13. Second stage of well ‘F’ stimulation. Figure 15. Daily production for well ‘A’. Figure 16. Daily production for well ‘B’. Figure 7.12: Microseismic distributions for hydraulic fracture treatment, where a com- plex seismicity pattern is observed at Barnett Shale, Df ' 1:82, with its associated production (after Maxwell et al. (2002)) 142 10 0 10 1 10 2 10 3 10 0 10 1 10 2 10 3 10 4 10 5 Df' 1.82±0.16 Box size (l) Number of boxes N(l) 0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2 Figure 7.13: Fracture model to match the fractal dimension of seismicity distribution (D f ' 1:82 0:16 and its fractal dimension calculated using the box count method Figure 7.14: Fracture model and history match when microseismic distribution has D f ' 1:89 0:05 in 2D orD f ' 2:5 0:1 in 3D 143 Chapter 8 Summary and Concluding Remarks In this dissertation, we introduced a comprehensive workflow for characterizing and modeling a fracture network in unconventional reservoirs, using microseismic data. This new workflow is an effective fracture characterization procedure estimates different frac- ture properties. Unlike the existing methods, the new approach is not dependent on the location of events. It is able to integrate all multi-scaled and diverse fracture information from different methodologies. In chapter 3, we offered an innovative method for monitoring the fracture movement in different stages of stimulation that can be used to optimize the process. We applied fuzzy logic concept in clustering the microseismic data to find the fracture network areas. Fuzzy cluster centers may represent centers of the connected fracture network which are ideal for reservoir creation and for enhancing production. In chapter 4, we offered an effective way to obtain the fractal dimension of micro- seismic events and identify the pattern complexity, connectivity, and mechanism of the created fracture network. This results indicate that the spatial distribution of hypocenters of microseismic events provide deeper insight into the structure of the fracture network of large-scale porous media. In addition, the results indicate that by calculating the fractal dimension of a microseismic cloud we may identify whether stimulated micro- seismic data are triggered (tectonic) or induced. Hence, we may find an explanation for changes in observed fracture behavior or determine if those changes might be caused by the presence of nearby faults (tectonic) or by contact with the fracturing treatment (induced). 144 Determining the fractal behavior of microseismic event clouds in different stages of stimulation and their dimensions, allows us to assume that the fracture network at the underlying unconventional reservoir is self-similar (scale independent), and thus that its structure, mechanical, and transport properties are best described by using fractal geometry. Microseismic data with a fractal dimension of D f ' 1:2 0:1 in 2D or D f ' 2 0:1 in 3D, are consistent with data for the surface of a single major fault, and represent the single large fracture that spans the stimulated volume. Therefore, a single planar hydraulic model will be only appropriate in this case. Microseismic data with a fractal dimension of D f ' 1:89 0:05 in 2D or D f ' 2:5 0:1 in 3D , represent a fracture network similar to a percolation cluster. Therefore, we have the highest complexity, and full interaction with geologic discontinuities where both natural fractures and Hydraulic fractures contribute to the flow. Microseismic data with a fractal dimension of D f ' 1:5 0:1 in 2D or D f ' 2:3 0:1 in 3D , reveal a fracture network pattern similar to the observed natural fracture pattern on outcrops. Therefore, hydraulic fracture follows the natural fracture network pattern, and hydraulic fracture stimulation mainly inflates the natural fractures or creates most fractures on existing planes of weakness. In chapter 5, we claimed that b-values of microseismic events can ascertain if the those that are stimulated have been triggered or induced. In other words, we can differ- entiate fracture-related events from fault-related ones in real-time; a microseismic cloud with b-value larger than 1.2 is induced and not tectonic, and higher b-values mean lower stress. In chapter 6, we offered an improved procedure to create compressional and shear velocity models as a preamble for delineating anomalies and map structures of interest and to correlate velocity anomalies with fracture swarms and other reservoir properties of interest. 145 We also demonstrated how integrating with production = injection data and fuzzy clustering can help us clarify our hypothesis, which relates the velocity and stress anomalies to the propagating fracture network. Moreover, we demonstrated that inte- grating the passive seismic tomography with density information allows us to detect the space-time dependency of elastic properties and stress in response to local variations of fluid pressure or fracture creation. Finally in chapter 7, we developed an innovative MDFN approach to create a discrete fracture network model by implementing the technique discussed in the previous steps of the workflow only using microseismic data with potential cost reduction. It also imposed fractal dimension as a constraint on other fracture modeling approaches, which increased the visual similarity between the modeled networks and the real network over the simulated volume. 146 Chapter 9 Future Work We presented a robust workflow to utilize most information content of the microseis- mic events which has advantage of being both comprehensive and flexible over current applied technology. In the study, we used geothermal reservoir data to validate our complete workflow. It was noticed that there is a small difference between geothermal data and hydraulic fracturing data. Therefore, testing our proposed workflow and methodology on micro- seismic data emitted from hydraulic fracturing stimulation would be useful. In this dissertation, we were not able to find any tomographic inversion or shear wave splitting data for hydraulic fracture patterns 1 , thus we could not follow all the steps proposed in our method to fully characterize the fracture network in hydraulic fracturing job. How- ever, we successfully tested our methodology for fractal dimension and b-value analysis on two case studies at Barnett Shale formation with both simple and complex fracture networks. We studied only microseismic data application in our innovative workflow. It was noticed that other data such as 3D seismic and well logs, have potential to be integrated in this workflow. Implementing advanced seismic attribute analysis and neural network based pseudo well logs can be great additions to the workflow. We studied b-values and compared them with true fractal dimension to characterize the associated fracture network. Extending this investigation on the b-values by creating 1 These technique are primarily designed for surface monitoring. Although, surface monitoring has been implemented in some hydraulic fracturing job, downhole monitoring is a common practice for mon- itoring them due to their higher accuracy; for more detail see section 2.1 147 b-value maps in different horizons and compare them with velocity and stress models to gain better understanding of stimulated areas, is another potential for future work. In this study, we used isotropic seismic velocity models to characterize the fractured areas. Building anisotropic velocity models to consider inherent anisotropy within the stimulated fracture network in the unconventional reservoirs would be another tool for improvement of this work. 148 Bibliography Acuna, J. a., Yortsos, Y . C., Mar. 1995. Application of Fractal Geometry to the Study of Networks of Fractures and Their Pressure Transient. Water Resources Research 31 (3), 527–540. Aki, K., 1981. A probabilistic synthesis of precursory phenomena, in earthquake pre- diction: an international review. In: Simpson, D., Richards, P. (Eds.), AGU, Maurice Ewing ser. V ol. 4. Washingtonn D.C., pp. 566–574. Al-Ghamdi, A. S., 2007. Automatic first arrival picking using energy ratios. Master of science, King Fahad University of Petroleum and Minerals. Albright, J., Pearson, C., 1982. Acoustic emissions as a tool for hydraulic fracture loca- tion: Experience at the Fenton Hill Hot Dry Rock site. Old SPE Journal (August). Ambegaokar, V ., Halperin, B., Langer, J., 1971. Hopping conductivity in disordered systems. Physical review B 4 (8), 2612–2620. Ameen, M., 2003. Fracture and in situ stress characterization of hydrocarbon reservoirs. The Geological Society, London. Aminzadeh, F., de Groot, P., 2006. Neural Networks and Soft Computing Techniques, with applications in the oil industry. EAGE Publishing. Aminzadeh, F., Katz, S., Aki, K., 1994. Adaptive neural nets for generation of artificial earthquake precursors. Geoscience and Remote Sensing, IEEE Transactions on 32 (6), 1139–1143. Aminzadeh, F., Tafti, T. A., Maity, D., 2010. Characterizing fractures in the geysers geothermal field using soft computing. GRC Transactions 34, 1193–1198. Aminzadeh, F., Tafti, T. a., Maity, D., Apr. 2013. An integrated methodology for sub- surface fracture characterization using microseismic data: A case study at the NW Geysers. Computers & Geosciences 54, 39–49. Arrowsmith, S. J., Eisner, L., 2006. A technique for identifying microseismic multiplets and application to the Valhall field, North Sea. Geophysics 71 (2), 31–40. 149 ASC, 2011. Microseismics for Carbon Storage. Babadagli, T., Dec. 2001. Fractal analysis of 2-d fracture networks of geothermal reservoirs in south-western turkey. Journal of V olcanology and Geothermal Research 112 (1-4), 83–103. Baig, A., Urbancic, T., 2010. Microseismic moment tensors A path to understanding frac growth. The Leading Edge (March). Balberg, I., 2009. Continuum Percolation. In: Sahimi, M., Myres, R. (Eds.), Encyclo- pedia of Complexity and System Science. Springer, New York, New York, USA, pp. 1443–1475. Barree, R., Fisher, M., Woodroof, R., 2002. A Practical Guide to Hydraulic Fracture Diagnostic Technologies. In: SPE Annual Technical Conference and Exhibition. SPE, p. SPE 77442. Barton, C. C., Hsieh, P. A., 1989. Physical and Hydrologic-Flow Properties of Fractures. In: Field Trip Guidebook T of AGU. Washington, D. C, p. 385. Barton, C. C., LaPointe, P., 1992. Fractals and Their Use in the Earth Sciences. Geolog- ical Society of America, Washington, DC. Barton, C. C., Schutter, T., Page, W., Samuel, J., 1987. Computer generation of frac- ture networks for hydrologic-flow modeling. In: Transactions American Geophysical Union 68. p. 1295. Barton, N., 2006. Rock quality, seismic velocity, attenuation and anisotropy. Taylor & Francis Group. Barton, C. C., E. L., 1985. Fractal geometry of two-dimensional fracture networks at Yucca Mountain, southwestern Nevada. In: International Symposium on Fundamen- tals of Rock Joints. Utah Geological Association, Salt Lake City, pp. 77–84. Beall, J. J., Wright, M. C., Pingol, A. S., Atkinson, P., 2010. Effect of high rate injection on seismicity in the geysers. GRC. Transactions 34, 1203–1208. Berge, P., Hutchings, L., Wagoner, J., 2001. Rock Physics Interpretation of P-Wave Q and Velocity Structure, Geology, Fluids and Fractures at the Southeast Portion of The Geysers Geothermal Reservoir. Geothermal Res. Council, Transactions 25, 383–389. Berryman, J. G., 1999. Origin of Gassmanns equations. Geophysics 64 (5), 1627. Berryman, J. G., Berge, P. a., Bonner, B. P., 2002. Estimating rock porosity and fluid saturation using only seismic velocities. Geophysics 67 (2), 391. 150 Berryman, J. G., Wang, H. F., 2000. Elastic wave propagation and attenuation in a double-porosity dual-permeability medium. Science 37, 63–78. Beugelsdijk, L., de Pater, C., Sato, K., Apr. 2000. Experimental Hydraulic Fracture Propagation in a Multi-Fractured Medium. In: SPE Asia Pacific Conference on Inte- grated Modelling for Asset Management. Society of Petroleum Engineers, p. SPE 59419. Bina, C., 1993. Mutually consistent estimates of upper mantle composition from seismic velocity contrasts at 400 km depth. pure and applied geophysics 141 (1), 101–109. Biot, M., 1956. Theory of propagation of elastic waves in a fluid-saturated porous solid. I. Low-frequency range. J. Acoust. Soc. Am 28 (2), 168–178. Boitnott, G., 2003. Core analysis for the development and constraint of physical models of geothermal reservoirs. Tech. rep., New England Research. Bonnet, E., Bour, O., Odling, N., Davy, P., 2001. Scaling of fracture systems in geolog- ical media. Reviews of Geophysics 39 (3), 347–383. Boyle, K., Jarpe, S., Hutchings, L., Saltiel, S., Peterson, J., 2011. Preliminary investiga- tion of an aseismic ’Doughnut Hole’ region in the northwest Geysers,California. In: Thirty-Sixth Workshop on Geothermal Reservoir Engineering. V ol. 1995. Brady, J., Withers, R., Fairbanks, T., Dressen, D., Sep. 1994. Microseismic monitoring of hydraulic fractures in prudhoe bay. Proceedings of SPE Annual Technical Confer- ence and Exhibition, 387–398. Bruner, K. R., Smosna, R., 2011. A Comparative Study of the Mississippian Barnett Shale , Fort Worth Basin , and Devonian Marcellus Shale , Appalachian Basin. Tech. rep., URS Corporation. Charlety, J., Cuenot, N., Dorbath, C., Dorbath, L., Oct. 2006. Tomographic study of the seismic velocity at the Soultz-sous-Forˆ ets EGS/HDR site. Geothermics 35 (5-6), 532–543. Chen, C.-C., Wang, W.-C., Chang, Y .-F., Wu, Y .-M., Lee, Y .-H., Dec. 2006. A corre- lation between the b -value and the fractal dimension from the aftershock sequence of the 1999 Chi-Chi, Taiwan, earthquake. Geophysical Journal International 167 (3), 1215–1219. Christopher, B., Schwartz, C., Boudreau, R., 2006. Geotechnical Aspects of Pavements. Tech. rep., U.S. Department of Transportation-Federal Highway Administration. Cipolla, C., Wright, C., 2000. Diagnostic techniques to understand hydraulic fracturing: What? why? and how? In: SPE/CERI Gas Technology Symposium. pp. 1–13. 151 Cramez, C., 2006. Basic Principles in Tectonics. Crampin, S., Oct. 1981. A review of wave motion in anisotropic and cracked elastic- media. Wave Motion 3 (4), 343–391. Crampin, S., Jan. 2005. A review of shear-wave splitting in the compliant crack-critical anisotropic Earth. Wave Motion 41 (1), 59–77. Dai, H., MacBeth, C., 1995. Automatic picking of seismic arrivals in local earthquake data using an artificial neural network. Geophysical journal international 120 (3), 758–774. Dai, H., MacBeth, C., 1997. The application of back-propagation neural network to automatic picking seismic arrivals from single-component recordings. Journal of Geophysical Research 102 (B7), 15105–15113. Dershowitz, W., Doe, T., 2011. Discrete Fracture Network Models of Natural and Hydraulic Fractures. In: AIPG Marcellus Shale Hydraulic fracturing Conference. p. Golder Assocaites. Diller, D., Gardner, S., 2011. Comparison of simultaneous downhole and surface micro- seismic monitoring in the Williston Basin. In: SEG Annual Meeting. SEG, pp. 1504– 1508. Downie, R., Kronenberger, E., Maxwell, S., 2010. Using Microseismic Source Param- eters To Evaluate the Influence of Faults on Fracture Treatments: A Geophysical Approach to Interpretation. In: SPE Annual Technical Conference and Exhibition. p. SPE 134772. Downie, R., Le Calvez, J., Kerrihard, K., 2009. Real-Time Microseismic Monitoring of Simultaneous Hydraulic Fracturing Treatments in Adjacent Horizontal Wells in the Woodford Shale. In: cspg CSEG CWLS Convention. Frontiers + Innovation, Calgary, Canada, pp. 484–492. Duncan, P., 2005. Passive Seismic: Something Old, Something New. AAPG Explorer (May 2005) 40153, 10–13. Eberhart-Phillips, D., 1986. hree-dimensional velocity structure in Northern California Coast Ranges from inversion of local earthquake arrival times. Bulletin of the Seis- mological Society of America 76 (4), 1025–1052. Eberhart-Phillips, D., 1990. Three-dimensional p and s velocity structure in the coalinga region, california. Journal of Geophysical Research 95 (B10), 15343–15. 152 Eberhart-Phillips, D., Michael, A. J., 1998. Seismotectonics of the Loma Prieta, Cali- fornia, region determined from three-dimensionalV p ;V p =V s , and seismicity. Journal of Geophysical Research 103 (B9), 21099–21120. Elkibbi, M., 2005. The Geysers geothermal field: results from shear wave splitting anal- ysis in a fractured reservoir. Geophysical Journal International, 1024–1035. Elkibbi, M., Yang, M., Rial, J., 2004. Imaging crack systems in the geysers with shear- wave splitting. Geothermal Resources Council (GRC) Transactions 28, 789–800. Eneva, M., Mar. 1996. Effect of limited data sets in evaluating the scaling properties of spatially distributed data: an example from mining-induced seismic activity. Geo- physical Journal International 124 (3), 773–786. Erten, D., Elkibbi, M., Rial, J., 2001. Shear wave splitting and fracture patterns at the geysers (california) geothermal field. In: Proceedings 26th Workshop on Geothermal Reservoir Engineering, Stanford University, Stanford. pp. 139–147. Fast, R., Murer, A., Timmer, R., 1994. Description and analysis of cored hydraulic fractures, Lost Hills Field, Kern County, California. SPE Production & Facilities SPE 24853 (May), 107–114. Ferer, M. V ., Dean, B. H., Mick, C., 1994. WGS-Fracture Modeling of Fractures in Tight Gas Reservoirs. Tech. rep. Fisher, M., Heinze, J., Harris, C., Wright, C., Dunn, K., 2004. Optimizing horizontal completion techniques in the Barnett shale using microseismic fracture mapping. In: SPE Annual Technical Conference and Exhibition. p. SPE 90051. Fisher, M., Wright, C., 2002. Integrating fracture mapping technologies to optimize stimulations in the barnett shale. SPE Annual Technical Conference and Exhibition, SPE 77441. Fomin, S., Hashida, T., Shimizu, a., Matsuki, K., Sakaguchi, K., Oct. 2003. Fractal concept in numerical simulation of hydraulic fracturing of the hot dry rock geothermal reservoir. Hydrological Processes 17 (14), 2975–2989. Frohlich, C., 1993. Teleseismic b Values; Or, Much Ado About 1.0. Journal of Geo- physical Research 98, 631–644. Gale, J., Schaefer, R., Carpenter, A., 1991. Collection, Analysis, and Integration of Discrete Fracture Data From the Monterey Formation for Fractured Reservoir Simu- lations. SPE Annual Technical Conference and Exhibition (SPE 22741), 823–834. 153 Gale, J. F. W., Reed, R. M., Holder, J., Apr. 2007. Natural fractures in the Barnett Shale and their importance for hydraulic fracture treatments. AAPG Bulletin 91 (4), 603– 622. Garnero, E., 2011. Earth interior research overview. Ghods, P., 2012. Data Assimilation for fractured shale gas reservoirs using ensemble kalman filter. Phd dissertation, University of Southern California. Grassberger, P., Procaccia, I., 1983. Measuring the strangeness of strange attractors. Physica D: Nonlinear Phenomena 9, 189–208. Grechka, V ., Mazumdar, P., 2010. Predicting permeability and gas production of hydraulically fractured tight sands from microseismic data. Geophysics 75 (1). Gritto, R., Yoo, S., Jarpe, S., 2013. Three dimensional seismic tomography at The Gey- sers geothermal field, CA, USA. In: Thirty-Eighth Workshop on Geothermal Reser- voir Engineering. Stanford University, Stanford, California, pp. 1241–1252. Grob, M., van der Baan, M., 2011. Statistical Analysis of Microseismic Event Locations and Magnitudes, and their Geomechanical Implications. In: SEG Annual Meeting. pp. 1570–1575. Gulia, L., Wiemer, S., Schorlemmer, D., 2010. Asperity-based earthquake likelihood models for Italy. Annals of Geophysics 53 (3), 63–75. Gunderson, R., 1992. Porosity of reservoir greywacke at The Geysers, in: Monograph on The Geysers Geothermal Field, special report. Tech. rep., GRC. Hatton, C., Main, I., Meredith, P., Dec. 1993. A comparison of seismic and structural measurements of scaling exponents during tensile subcritical crack growth. Journal of Structural Geology 15 (12), 1485–1495. He, M., 1998. Application of X-ray tomography to measurement of fractures in rocks. Master of science thesis, Stanford University. Hebein, J., 1986. Reservoir fracturing in the geysers hydrothermal system: Fact or fal- lacy? In: Proceedings, 11th Stanford Workshop on Geothermal Reservoir Engineer- ing. pp. 43–50. Henderson, J., Main, I., 1994. Seismicity in north-eastern Brazil: fractal clustering and the evolution of the b value. Geophysical Journal International 116, 217–226. Herrmann, H. J., Sahimi, M., Tzschichholz, F., 1993. Examples of fractals in soil mechanics. Fractals 1 (4), 795. 154 Hirata, T., 1989. A correlation between the b value and the fractal dimension of earth- quakes. Journal of Geophysical Research 94 (B6), 7507–7514. Hirata, T., Satoh, T., Ito, K., 1987. Fractal structure of spatial distribution of microfrac- turing in rock. Geophysics 90, 369–374. Hummel, N., Shapiro, S., 2011. Nonlinear diffusion estimates from hydraulic fracturing of shales. SEG Annual Meeting, 1544–1549. Hutchings, L., Jarpe, S., Boyle, K., Viegas, G., Majer, E., 2011. Inexpensive , Auto- mated Micro-Earthquake Data Collection and Processing System for Rapid , High- Resolution Reservoir Analysis. GRC Transactions 35, 1679–1685. Julian, B., Ross, A., Foulger, G., Evans, J., 1996. Three-dimensional seismic image of a geothermal reservoir: The Geysers, California. Geophysical Research Letters 23 (6), 685–688. Kagan, Y ., 1999. Universality of the seismic moment-frequency relation. Pure and Applied Geophysics 155, 537–573. Kanamori, H., Anderson, D., 1975. Theoretical basis of some empirical relations in seismology. Bulletin of the Seismological Society of America 65 (5), 1073–1095. Kenedi, C., Shalev, E., Lucas, A., Malin, P., 2010. Microseismicity and 3-d mapping of an active geothermal field, kilauea lower east rift zone, puna, hawaii. In: Proceedings World Geothermal Congress 2010. No. apr. Bali, Indonesia,. Klein, F. W., 2002. User’s guide to hypoinverse-2000, a fortran program to solve for earthquake locations and magnitudes. Kumano, Y ., Asanuma, H., Hotta, A., 2007. Reservoir structure delineation by micro- seismic multiplet analysis at Basel, Switzerland, 2006. 77th SEG. La Pointe, P., 1988. A method to characterize fracture density and connectivity through fractal geometry. International Journal of Rock Mechanics and Mining 25 (6), 421 – 429. Le Calvez, J., Craven, M., Klem, R., Baihly, J., Bennett, L., Brook, K., 2007. Real- Time Microseismic Monitoring of Hydraulic Fracture Treatment: A Tool to Improve Completion and Reservoir Management. In: SPE Hydraulic Fracturing Technology Conference. p. SPE 106159. Lees, J., 1998. Multiplet analysis at Coso geothermal. Bulletin of the Seismological Society of America 88 (5), 1127–1143. 155 Leonard, M., 2000. Comparison of manual and automatic onset time picking. Bulletin of the Seismological Society of America 90 (6), 1384–1390. Leroy, Y . M., Lehner, F. K., 2012. Mechanics of Crustal Rocks. Springer Vienna. Lou, M., Shalev, E., Malin, P., 1997. Shear-wave splitting and fracture alignments at the northwest geysers, california. Geophysical Research Letters 24 (15), 1895–1898. Main, I. G., 1992. Damage mechanics with long-range interactions : correlation between the seismic b-value and the fractal two-point correlation dimension. Main, 531–541. Malin, P., Shalev, E., 1999. Shear-wave splitting crack density maps for the geysers and mammoth. In: Proceedings, 24th Stanford Workshop on Geothermal Reservoir Engineering. Martakis, N., Kapotas, S., Tselentis, G., Others, 2006. Integrated passive seismic acqui- sition and methodology. Case studies. Geophysical prospecting 54 (6), 829–847. Maxwell, S., Jones, M., Parker, R., Leaney, W., Mack, M., Dorvall, D., D’Amico, D., Logel, J., Anderson, E., Hammermaster, K., 2010. Fault Activation During Hydraulic Fracturing. EAGE (June 2010), 14–17. Maxwell, S., Urbancic, T., 2002. Real-Time 4D Passive Seismic Imaging of Hydraulic Fracturing. SPE/ISRM Rock Mechanics Conference (October), SPE / ISRM 78191. Maxwell, S., Urbancic, T., Steinsberger, N., Zinno, R., 2002. Microseismic imaging of hydraulic fracture complexity in the barnett shale. In: SPE Annual Technical Confer- ence and Exhibition. p. SPE 77440. Maxwell, S. C., 2011. What Does Microseismicity Tells Us About Hydraulic Fractures ? In: SEG Annual Meeting. pp. 1565–1569. Mayerhofer, M., Lolon, E., Youngblood, J., Heinze, J., 2006. Integration of Microseismic-Fracture-Mapping Results With Numerical Fracture Network Produc- tion Modeling in the Barnett Shale. In: SPE Annual Technical Conference and Exhi- bition. p. SPE 102103. Mayerhofer, M. J., Technologies, P., Bolander, J. L., Williams, L. I., Energy, S., Sands, T., 2005. Integration of Microseismic Fracture Mapping , Fracture and Production Analysis With Well Interference Data To Optimize Fracture Treatments in the Overton Field , East Texas. In: SPE Annual Technical Conference and Exhibition. p. SPE 95508. McCormack, M., Zaucha, D., Dushek, D., Jan. 1993. First-break refraction event pick- ing and seismic data trace editing using neural networks. Geophysics 58 (1), 67–78. 156 Miyata, A., Yuzo, O., Satoshi, N., Takao, Y ., Takahashi, M., 2006. A study on estimat- ing hydraulic characteristics of rock specimens using elastic wave. 4th Asian Rock Mechanics Symposium. Moriya, H., Nagano, K., Niitsuma, H., 1994. Precise source location of AE doublets by spectral matrix analysis of triaxial hodogram. Geophysics 59 (1), 36–45. Moriya, H., Nakazato, K., Niitsuma, H., Baria, R., 2000. Study of microseismic dou- blet/multiplet for evaluation of subsurface fracture system in Soultz HDR field. In: Proc. World Geothermal Congress 2000. pp. 3807–3812. National Energy Technology Laboratory, 2013. FracGen & NFlow: Fractured Reservoir Simulator. Nave, C., 2010. HyperPhysics. Nielson, D., Walters, M., Hulen, J., 1991. Fracturing in the northwest geysers, sonoma county, california. Transactions, Geothermal Resources Council 15, 27–33. Nur, A., Simmons, G., 1969. The effect of saturation on velocity in low porosity rocks. Earth and Planetary Science Letters 7 (2), 183–193. ¨ Oncel, A. O., Main, I., Alptekin, O., Cowie, P., Jun. 1996. Spatial variations of the fractal properties of seismicity in the Anatolian fault zones. Tectonophysics 257 (2- 4), 189–202. Past´ en, D., Mu˜ noz, V ., Cisternas, A., Rogan, J., Valdivia, J. A., Dec. 2011. Monofractal and multifractal analysis of the spatial distribution of earthquakes in the central zone of Chile. Physical Review E 84 (6), 066123. Phillips, W., Rutledge, J., Fairbanks, T., Gardner, T., Miller, M., Maher, B., 1998. Reser- voir fracture mapping using microearthquakes: Two oilfield case studies. SPE Reser- voir Evaluation and Engineering 1 (2), 114–121. Qi-Zhen, D., Li-Yun, K., Shi-Chun, H., 2009. Wavefield propagation characteristics in the fracture-induced anisotropic double-porosity medium. Chinese Journal of Geo- physics 52 (2), 490–500. Refunjol, X., Keranen, K., Marfurt, K., 2010. Inversion and attribute-assisted hydrauli- cally induced microseismic fracture prediction: A North Texas Barnett Shale case study: 80th Annual International Meeting, SEG. In: SEG Annual Meeting. pp. 2161– 2165. Robertson, M. C., Sammis, C. G., Sahimi, M., Martin, A. J., 1995. Fractal analysis of three-dimensional spatial distributions of earthquakes with a percolation interpreta- tion. Journal of Geophysical Research 100 (B1), 609–621. 157 Rountree, C. L., Kalia, R. K., Lidorikis, E., Nakano, A., Van Brutzel, L., Vashishta, P., Aug. 2002. Atomistic aspects of crack propagation in brittle materials: Multimillion atom molecular dynamics simulations. Annual Review of Materials Research 32 (1), 377–400. Rowe, C., Aster, R., Phillips, W., Jones, R., 2002. Using automated, high-precision repicking to improve delineation of microseismic structures at the Soultz geothermal reservoir. Pure and Applied 159, 563–596. Rozhko, A. Y ., Nov. 2010. Role of seepage forces on seismicity triggering. Journal of Geophysical Research 115 (B11), 1–12. Rutledge, J., Phillips, W., 2003. Hydraulic stimulation of natural fractures as revealed by induced microearthquakes, Carthage Cotton Valley gas field, east Texas. Geophysics 68 (2), 441–452. Rutledge, J., Phillips, W., House, L., Zinno, R., 1998. Microseismic mapping of a cotton valley hydraulic fracture using decimated downhole arrays. Proc. 68th Ann. Mtg., SEG, 338–341. Rutqvist, J., Jun. 2002. A modeling approach for analysis of coupled multiphase fluid flow, heat transfer, and deformation in fractured porous rock. International Journal of Rock Mechanics and Mining Sciences 39 (4), 429–442. Rutqvist, J., Majer, E., Oldenburg, C., Peterson, J., Vasco, D., 2006. Integrated mod- eling and field study of potential mechanisms for induced seismicity at The Geysers Goethermal Field, California. Lawrence Berkeley National Laboratory (June). Sabbione, J., 2010. Automatic first-breaks picking: New strategies and algorithms. Geo- physics 75 (4), V67–V76. Sahimi, M., 1994. Application of Percolation Theory. Taylor and Francis, London. Sahimi, M., 1995. New Models for Natural and Hydraulic Fracturing of Heterogeneous Rock. SPE Western Regional Meeting (March), SPE 29648. Sahimi, M., 2011. Flow and Transport in Porous Media and Fractured Rock. In: Flow and Transport in Porous Media and Fractured Rock, 2nd Edition. Wiley-VCH, Wein- heim, Ch. 6. Sahimi, M., Arbabi, S., 1992. Percolation and fracture in disordered solids and granular media: Approach to a fixed point. Physical review letters 68 (5), 608–611. Sahimi, M., Arbabi, S., Oct. 1996. Scaling Laws for Fracture of Heterogeneous Materi- als and Rock. Physical review letters 77 (17), 3689–3692. 158 Sahimi, M., Goddard, J., Jun. 1986. Elastic percolation models for cohesive mechani- cal failure in heterogeneous systems. Physical review. B, Condensed matter 33 (11), 7848–7851. Sahimi, M., Robertson, M., Sammis, C., 1993. Fractal distribution of earthquake hypocenters and its relation to fault patterns and percolation. Physical review letters 70 (14), 2186–2189. Sahimi, M., Yortsos, Y ., 1990. Applications of fractal geometry to porous media: A review. In: Annual Fall Meeting of the Society of Petroleum Engineers. V ol. 3. SPE, New Orleans, LA. Sammis, C., An, L., Ershaghi, I., 1992. Determining the 3-d fracture structure in the geysers geothermal reservoir. In: Proceedings, 17th Stanford Workshop on Geother- mal Reservoir Engineering. University of Southern California, Los Angeles, CA, pp. 79–85. Sammis, C. G., An, L. J., Ershaghi, I., 1991. Fracture patterns in graywacke outcrops at the geysers geothemral field. In: Sixteenth Workshop on Geothermal Reservoir Engineering. Stanford. Sammonds, P. G., Meredith, M. R., Ayling, C. J., Murell, S. A. F., 1989. Fracture of Concrete and Rock. Elsevier, London. Satik, C., Walters, M., Horne, R., 1996. Adsorption characteristics of rocks from vapor- dominated geothermal reservoir at the Geysers, CA. In: Twenty-First Workshop on Geothennrll Reservoir Engineering. No. 1 973. Stanford University, Stanford, pp. 469–479. Schorlemmer, D., Wiemer, S., Wyss, M., Sep. 2005. Variations in earthquake-size dis- tribution across different stress regimes. Nature 437 (7058), 539–42. Shojaei, H., Tajer, E. S., California, S., 2013. Analytical Solution of Transient Multi- phase Flow to a Horizontal Well with Multiple Hydraulic Fractures. In: SPE Eatern Regional Meeting. No. 1993. Pittsburg, p. SPE 165703. Simmons, G., Mar. 1964. Velocity of shear waves in rocks to 10 kilobars, 1. Journal of Geophysical Research 69 (6), 1123–1130. Smith, L., 1988. Intrinsic limits on dimension calculations. Physics Letters A 133 (6), 283–288. Stark, C. P., Stark, J. A., 1991. Seismic fluids and percolation theory. Journal of Geo- physical Research 96 (B5), 8417. 159 Stark, M., 1990. Imaging injected water in The Geysers reservoir using microearthquake data. Geothermal Res. Council, Transactions 14 (2), 1697–1704. Stauffer, D., Aharony, A., 1994. Introduction to Percolation Theory, 2nd Edition. Taylor and Francis, London. Sternfeld, J., 1989. Lithologic influences on fracture permeability and the distribution of steam in the northwest geysers steam field, sonoma county, california. GRC Trans 13, 473–480. Tafti, T. a., Sahimi, M., Aminzadeh, F., Sammis, C. G., Mar. 2013. Use of microseismic- ity for determining the structure of the fracture network of large-scale porous media. Physical Review E 87 (3), 032152. Tang, C., Rial, J. A., Lees, J. M., 2008. Seismic imaging of the geothermal field at Krafla , Iceland using shear-wave splitting. Journal of V olcanology and Geothermal Research 176, 315–324. Tezuka, K., Apr. 2000. Stress estimated using microseismic clusters and its relationship to the fracture system of the Hijiori hot dry rock reservoir. Engineering Geology 56 (1- 2), 47–62. Tezuka, K., Kamitsuji, R., Tamagawa, T., 2008. Fractured Reservoir Characteriza- tion Incorporating Microseismic Monitoring and Pressure Analysis During Massive Hydraulic Injection. In: International Petroleum Technology Conference. Thurber, C., 1983. Earthquake locations and three-dimensional crustal structure in the coyote lake area, central california. Journal of Geophysical Research 88 (B10), 8226– 8236. Tokosoz, M. N., Johnson, D. H., 1981. Seismic wave attenuation. Society of Exploration Geophysicists. Toks¨ oz, M., Cheng, C., Timur, A., 1976. Velocities of seismic waves in porous rocks. Geophysics 41 (4), 621–645. Trifu, C., Radulian, M., 1989. Asperity distribution and percolation as fundamentals of an earthquake cycle. Physics of the earth and planetary interiors 58, 277–288. Um, J., Thurber, C., 1987. A fast algorithm for two-point seismic ray tracing. Bulletin of the Seismological Society of America 77 (3), 972–986. Veezhinathan, J., Kemp, F., Threet, J., 1993. A hybrid of neural net and branch and bound techniques for seismic horizon tracking. Proceedings of the 1993 ACM/SIGAPP symposium on Applied computing states of the art and practice - SAC ’93, 173–178. 160 Velde, B., Dubois, J., Moore, D., Touchard, G., May 1991. Fractal patterns of fractures in granites. Earth and Planetary Science Letters 104 (1), 25–35. Vignes-Adler, M., Le Page, A., Adler, P. M., Sep. 1991. Fractal analysis of fracturing in two African regions, from satellite imagery to ground scale. Tectonophysics 196 (1- 2), 69–86. Vlahovic, G., Elkibbi, M., Rial, J., Jan. 2003. Shear-wave splitting and reservoir crack characterization: the Coso geothermal field. Journal of V olcanology and Geothermal Research 120 (1-2), 123–140. Waldhauser, F., 2001. hypodd – a program to compute double-difference hypocenter locations. Warpinski, N., Kramm, R., Heinze, J., Waltman, C., 2005. Comparison of single-and dual-array microseismic mapping techniques in the Barnett shale. In: SPE Annual Technical Conference and Exhibition. October, p. SPE 95568. Warpinski, N., Lorenz, J., Branagan, P., Myal, F., Gall, B., 1993. Examination of a Cored Hydraulic Fracture in a Deep Gas Well. SPE Production & Facilities SPE26302 (August), 150–164. Warpinski, N., Teufel, L., 1987. Influence of geologic discontinuities on hydraulic frac- ture propagation. Journal of Petroleum Technology SPE 13224 (February), 209–220. Warpinski, N., Waltman, C., Weijers, L., 2010. An Evaluation of Microseismic Monitor- ing of Lenticular Tight-Sandstone Stimulations. SPE Production & Operations 25 (4), 498–508. Warpinski, N., Wright, T., Uhl, J., 1996. Microseismic monitoring of the B-sand hydraulic fracture experiment at the DOE/GRI multi-site project. SPE Annual Tech- nical Conference and Exhibition SPE 36450 (October), 327–335. Wells, D., Coppersmith, K., 1994. New empirical relationships among magnitude, rup- ture length, rupture width, rupture area, and surface displacement. Bulletin of the Seismological . . . 84 (4), 974–1002. Wessels, S., Kratz, M., Pena, A. D. L., 2011. Identifying fault activation during hydraulic stimulation in the Barnett shale : source mechanisms , b values , and energy release analyses of microseismicity. In: SEG Annual Meeting. pp. 1463–1467. White, J., 1975. Computed seismic speeds and attenuation in rocks with partial gas saturation. Geophysics 40 (2), 224–232. Wiemer, S., 2001. Zmap. 161 Wiemer, S., McNutt, S. R., Wyss, M., Aug. 1998. Temporal and three-dimensional spa- tial analyses of the frequency-magnitude distribution near Long Valley Caldera, Cal- ifornia. Geophysical Journal International 134 (2), 409–421. Williams-Stroud, S., 2008. Using Microseismic Events to Constrain Fracture Network Models and Implications for Generating Fracture Flow Properties for Reservoir Sim- ulation. In: SPE Shale Gas Production Conference. p. SPE 119895. Wilson, T., 1996. Fractal Analysis of fracture systems. Tech. rep., West Virgina Univer- sity, Deaprtment of Geology and Geography. Wright, C. A., Davis, E. J., Minner, W. A., Ward, J. F., Weijers, L., Technologies, P., Schell, E. J., Pacific, U., Hunter, S. P., Livermore, L., Laboratories, N., 1970. Surface Tiltmeter Fracture Mapping Reaches New Depths -10,000, and Beyond? In: SPE Rocky Mountain Regional Meeting. V ol. April. p. SPE 39919. Wyllie, M., Gregory, A., 1958. An experimental investigation of factors affecting elastic wave velocities in porous media. Geophysics 23 (3), 459–493. Wyllie, M., Gregory, A., Gardner, L., 1956. Elastic wave velocities in heterogeneous and porous media. Geophysics 21 (1), 41–70. Wyss, M., Nov. 2000. Change in the probability for earthquakes in southern california due to the landers magnitude 7.3 earthquake. Science 290 (5495), 1334–1338. Wyss, M., Sammis, C., 2004. Fractal dimension and b-value on creeping and locked patches of the San Andreas fault near Parkfield, California. Bulletin of the Seismo- logical Society of America 94 (2), 410–421. Xu, W., Calvez, J. L., 2009. Characterization of Hydraulically-Induced Fracture Net- work Using Treatment and Microseismic Data in a Tight-Gas Sand Formation: A Geomechanical Approach. SPE Tight Gas Completions. Zhao, Y ., Takano, K., 1999. An artificial neural network approach for broadband seismic phase picking. Bulletin of the Seismological Society of America 89 (3), 670–680. Zoback, M., 2007. Reservoir geomechanics. Cambridge Univ Pr. 162

Abstract (if available)

Abstract We develop a new method for integrating information and data from different sources. We also construct a comprehensive workflow for characterizing and modeling a fracture network in unconventional reservoirs, using microseismic data. The methodology is based on combination of several mathematical and artificial intelligent techniques, including geostatistics, fractal analysis, fuzzy logic, and neural networks. The study contributes to scholarly knowledge base on the characterization and modeling fractured reservoirs in several ways

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Integrated reservoir characterization for unconventional reservoirs using seismic, microseismic and well log data

PDF

Integration of multi-physics data into dynamic reservoir model with ensemble Kalman filter

PDF

A method for characterizing reservoir heterogeneity using continuous measurement of tracer data

$History matching for fractured reservoirs with Hough-transform-based parameterization$

PDF

History matching for fractured reservoirs with Hough-transform-based parameterization

PDF

Assessing induced seismicity rate increase based on deterministic and probabilistic modeling

PDF

Modeling and simulation of multicomponent mass transfer in tight dual-porosity systems (unconventional)

PDF

Modeling and simulation of complex recovery processes

PDF

Deep learning architectures for characterization and forecasting of fluid flow in subsurface systems

$Coarse-scale simulation of heterogeneous reservoirs and multi-fractured horizontal wells$

PDF

Coarse-scale simulation of heterogeneous reservoirs and multi-fractured horizontal wells

PDF

Continuum modeling of reservoir permeability enhancement and rock degradation during pressurized injection

$An extended finite element method based modeling of hydraulic fracturing$

PDF

An extended finite element method based modeling of hydraulic fracturing

PDF

A coupled multiphase flow-geomechanics framework for modeling material, kinematic and contact nonlinearities in faulted reservoirs

PDF

Deep learning for subsurface characterization and forecasting

PDF

Application of data-driven modeling in basin-wide analysis of unconventional resources, including domain expertise

$Aspects of fracture heterogeneity and fault transmissivity in pressure transient modeling of naturally fractured reservoirs$

PDF

Aspects of fracture heterogeneity and fault transmissivity in pressure transient modeling of naturally fractured reservoirs

PDF

Mass transfer during enhanced hydrocarbon recovery by gas injection processes

PDF

Aggregation and modeling using computational intelligence techniques

PDF

Investigating statistical modeling approaches for reservoir characterization in waterfloods from rates fluctuations

PDF

A study of diffusive mass transfer in tight dual-porosity systems (unconventional)

PDF

Optimization of coupled CO₂ sequestration and enhanced oil recovery

Asset Metadata

Creator Ayatollahy Tafti, Tayeb (author)

Core Title Integrated workflow for characterizing and modeling fracture network in unconventional reservoirs using microseismic data

School Viterbi School of Engineering

Degree Doctor of Philosophy

Degree Program Petroleum Engineering

Publication Date 08/20/2013

Defense Date 06/13/2013

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag b-value,discrete fracture network,fractal,fracture,fuzzy logic,geothermal,microseismic data,Modeling,OAI-PMH Harvest,reservoir characterization,reservoir simulation,seismic velocity,shale reservoirs,tomographic inversion,unconventional reservoirs

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Aminzadeh, Fred (committee chair), Ershaghi, Iraj (committee member), Sammis, Charles G. (committee member)

Creator Email ayatolla@usc.edu,tayeb.ayat@gmail.com

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c3-322486

Unique identifier UC11287942

Identifier etd-Ayatollahy-2015.pdf (filename),usctheses-c3-322486 (legacy record id)

Legacy Identifier etd-Ayatollahy-2015.pdf

Dmrecord 322486

Document Type Dissertation

Rights Ayatollahy Tafti, Tayeb

Type texts

Source University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Access Conditions The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the a...

Repository Name University of Southern California Digital Library

Repository Location USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA