ON THE INTERACTION BETWEEN FLUIDS AND ...

ON THE INTERACTION BETWEEN FLUIDS AND EARTHQUAKES IN BOTH NATURAL AND INDUCED SEISMICITY by MATTHEW BENJAMIN WEINGARTEN B.S., University of Wisconsin-Madison, 2009

A thesis submitted to the Faculty of the Graduate School of the University of Colorado in partial fulfillment of the requirement for the degree of Doctor of Philosophy Department of Geologic Sciences 2015

This thesis entitled: On the interaction between fluids and earthquakes in both natural and induced seismicity Written by Matthew Benjamin Weingarten has been approved for the Department of Geological Sciences

Professor Shemin Ge

Professor Anne F. Sheehan

Date

The final copy of this thesis has been examined by the signatories, and we find that both the content and the form meet acceptable presentation standards of scholarly work in the above mentioned discipline.

ii

ABSTRACT

Weingarten, Matthew Benjamin (Ph.D., Geological Sciences) On the interaction between fluids and earthquakes in both natural and induced seismicity Thesis directed by Professor Shemin Ge

Fluid-fault interaction in the subsurface is a critical driver of both natural and induced earthquakes. Fluid-pressure increases inside fault zones lower their frictional resistance to slip and make earthquakes more likely. Inversely, earthquakes also perturb the fluid-pressure field and cause observable changes in groundwater level. This dissertation investigates fluid-fault interaction by studying both groundwater level changes from natural earthquakes and injectioninduced earthquakes from fluid injection wells. Chapter 2 quantifies an extremely sensitive water level to distant earthquakes at Devils Hole in southern Nevada. Examining a 24-year water level record, I find the seismic energy density required to initiate both hydroseismogram and coseismic types of water level response is e ~ 10-6 J/m3, two orders of magnitude smaller than previously documented. This new threshold has implications for the dynamic triggering of earthquakes, as remote earthquakes can lead to pore pressure changes and consequently effective stress changes in fluid-filled fault zones. Chapters 3 through 5 examine the relationship between fluid injection and the unprecedented seismic rate increase in the U.S. mid-continent beginning in 2009. This rate increase occurred in regions where earthquakes were generally uncommon and not predicted by the laws of natural seismicity. Chapter 3 characterizes seismicity and fluid-pressure changes from injection wells in Jones, Oklahoma, showing that high-rate injection wells are likely responsible for the earthquake swarm. The modeled fluid-pressure perturbation propagates iii

throughout the same depth range and tracks earthquakes to distances of 35 km, with a triggering threshold of ~0.07 MPa. Chapter 4 examines the broad-scale relationship between fluid injection and U.S. mid-continent seismicity using a newly assembled injection well database for the central and eastern U.S. Statistical methods find the entire increase in earthquake rate is associated with fluid injection wells. High injection rate wells (>300,000 barrels/month) are statistically much more likely to be associated with earthquakes than lower rate wells. Finally, in Chapter 5, I quantify a novel case of injection-induced seismicity in the Raton Basin of southern Colorado. Hydrogeologic models show injection can induce earthquakes several kilometers below the reservoir injection interval despite a lack of wellhead pressure needed for injection.

iv

ACKNOWLEDGEMENTS This thesis would never have been possible without the determination, hard work and sacrifice of my loving family: Chris, Dan and Sophie Weingarten. I thank them from the bottom of my heart for all they have given me. Of many things, my mother taught me the meaning of hard work and sacrifice; my sister taught me how to laugh and how to treat others; and my father taught me what to value in life, to pursue knowledge, and to love nature. It was our trip to Death Valley with a couple of geology books during high school that got this whole thing rolling. To my advisor, Shemin Ge, who has been an endless pillar of support and patience during my many wanderings over the last 5 years. I have the utmost respect for your clear scientific mind, ability to stay focused, ability to see my potential when I may not have and to create opportunities for me to collaborate with world-class researchers early on in my career. Most of all, I'll remember your sense of humor and the many hilarious conversations that had me doubled over with laughter. To all of the collaborators who have guided me in my studies. The mentorship and wisdom of Justin Rubinstein, Barbara Bekins, Katie Keranen, Anne Sheehan, Jonathan Godt, Paul Hsieh and Jin-Yong Lee has been invaluable. I know this achievement is in large part due to your guidance. To all of my lab mates through the years who provided me support and relief. Listed in order of graduation or arrival since my time here: Miori Yoshino, Lyndsay Ball, Jessica Cochrane, Nadine Reitman, Nora Catolico, Sarah Evans, Steven Henning and Megan Brown. Finally, this research owes a great debt to the United States Geological Survey, the John Wesley Powell Center for Synthesis and Analysis as well as the National Science Foundation for their continued financial support throughout the course of my doctoral research.

v

TABLE OF CONTENTS

ACKNOWLEDGEMENTS ............................................................................................... v LIST OF FIGURES ........................................................................................................... x LIST OF TABLES .......................................................................................................... xiv CHAPTER 1: INTRODUCTION ...................................................................................... 1 1.1 Background ...................................................................................................... 1 1.2 Scope of Work ................................................................................................. 3 1.3 References ........................................................................................................ 5 CHAPTER 2: INSIGHTS INTO WATER LEVEL RESPONSE TO SEISMIC WAVES: A 24 YEAR HIGH-FIDELITY RECORD OF GLOBAL SEISMICITY AT DEVILS HOLE Abstract .................................................................................................................. 7 2.1 Introduction ..................................................................................................... 8 2.2 Background and Dataset .................................................................................. 9 2.3 Identifying Seismically Induced Water Level Fluctuations........................... 13 2.4 Analysis of Water Level Response ................................................................ 15 2.4.1 Characteristic Types of Water Level Response .................................... 15 2.4.2 Water Level Sensitivity......................................................................... 16 2.4.3 Characteristic Frequency of Water Level Response ............................. 19 2.4.4 Water Level Response to Shear Waves ................................................ 19 2.5 Mechanisms of Water Level Response .......................................................... 23 2.6 Conclusions .................................................................................................... 27 2.7 Acknowledgements ........................................................................................ 29 2.8 References ...................................................................................................... 29 vi

2.A Appendix ....................................................................................................... 33 CHAPTER 3: SHARP INCREASE IN CENTRAL OKLAHOMA SEISMICITY SINCE 2008 INDUCED BY MASSIVE WASTEWATER INJECTION Abstract ................................................................................................................ 44 3.1 Introduction .................................................................................................... 45 3.2 Methods and Data .......................................................................................... 45 3.2.1 Earthquake Swarm ................................................................................ 45 3.2.2 Injection Operations in the Jones, Oklahoma Region ........................... 49 3.3 Hydrogeologic Model .................................................................................... 49 3.3.1 Numerical Code .................................................................................... 53 3.3.2 Model Domain, Boundary Conditions & Material Properties .............. 53 3.3.3 Sensitivity Analysis .............................................................................. 56 3.4 Modeling Results ........................................................................................... 61 3.5 Discussion ...................................................................................................... 64 3.6 Conclusion ..................................................................................................... 68 3.7 Acknowledgements ........................................................................................ 68 3.8 References ...................................................................................................... 69 CHAPTER 4: HIGH-RATE INJECTION IS ASSOCIATED WITH THE INCREASE IN U.S. MID-CONTINENT SEISMICITY Abstract ................................................................................................................ 73 4.1 Introduction .................................................................................................... 74 4.2 Methods, Data & Results ............................................................................... 75 4.3 Conclusion ................................................................................................... 107 4.4 Acknowledgements ...................................................................................... 109 vii

4.5 References .................................................................................................... 109 4.A Appendix ..................................................................................................... 115 4.A1 Injection Data Quality Control and Data Sources ............................... 115 4.A2 Calculating Maximum Monthly Injection rate & Cumulative Injected Volume......................................................................................................... 119 4.A3 Statistical Methods to Estimate Confidence Intervals ........................ 120 4.A4 Suspected Mining Blasts Removed..................................................... 127 CHAPTER 5: FLUID-PRESSURE MODELING OF INJECTION IN THE RATON BASIN, COLORADO-NEW MEXICO: A CASE STUDY OF INDUCED SEISMICITY Abstract .............................................................................................................. 129 5.1 Introduction .................................................................................................. 130 5.2 Earthquakes and Injection Wells Overview ................................................. 131 5.2.1 Basin-Scale Injection Overview ......................................................... 131 5.2.2 Earthquake Sequences ........................................................................ 133 5.3 Geologic Background .................................................................................. 138 5.3.1 Geologic Setting.................................................................................. 138 5.3.2 Stratigraphy ......................................................................................... 140 5.4 Regional Hydrogeology ............................................................................... 147 5.4.1 Hydrostratigraphy ............................................................................... 147 5.4.2 Regional Groundwater Flow ............................................................... 151 5.4.3 Regional Underpressure ...................................................................... 151 5.4.4 Geothermal Gradient & Fluid Properties ............................................ 153 5.5 Injection Operations in Detail ...................................................................... 156 5.6 Hydrogeologic Model .................................................................................. 165 viii

5.6.1 Conceptual Model ............................................................................... 166 5.6.2 Model Domain, Boundary Conditions & Grid Discretization ........... 168 5.6.3 Hydrogeologic Scenarios & Modeling Results................................... 171 5.6.3.1 Homogeneous Scenario .......................................................... 171 5.6.3.2 Heterogeneous Scenario1: Sediment-Basement Contrast ....... 175 5.6.3.3 Heterogeneous Scenario2: Depth-Decreasing Basement Conductivity ....................................................................... 178 5.6.3.4 Heterogeneous Scenario3: Permeable Fault ........................... 186 5.7 Conclusions .................................................................................................. 188 5.8 References .................................................................................................... 189

ix

LIST OF FIGURES

CHAPTER 2 FIGURES 2.1

Map of earthquakes producing Devils Hole water level responses ............... 10

2.2

Water level monitoring setup at Devils Hole ................................................. 12

2.3

Devils Hole water level responses recorded by chart recorder ...................... 14

2.4

Seismic energy density of water level responses at Devils Hole ................... 17

2.5

Frequency range of water level responses at Devils Hole ............................. 20

2.6

High-frequency water level response compared with seismometer .............. 22

2.7

Cross-section of Devils Hole with rose diagram of earthquake responses .... 24

2.8

Water level data from 15-minute transducer ................................................. 26


Map of earthquakes in Oklahoma between 1976 and 2014 ........................... 46

3.2

Map of Jones area earthquake catalog and swarm migration ........................ 48

3.3

Fluid injection in four high-rate wells in southeast Oklahoma city ............... 50

3.4

Cross section of the Nemaha Fault ................................................................ 51

3.5

Cumulative monthly injection rate of wells in hydrogeologic model............ 52

3.6

Schematic diagram of model domain ............................................................ 54

3.7

Generalized hydrostratigraphy used in sensitivity analysis ........................... 57

3.8

Model results for 3 sensitivity scenarios including NE wells ........................ 59

3.9

Model results for 3 sensitivity scenarios with only 4 SE OKC wells ............ 60

3.10

Best-fit hydrogeologic model Jones area injection ........................................ 62

x

3.11

Comparison of well contributions to pore pressure change ........................... 63

3.12

Comparison of injected volume and injection rate ........................................ 66


Map of active and associated Class II injection wells in the CEUS .............. 76

4.2

Maps of active SWD and EOR wells & well spatial density ........................ 77

4.3

Map of associated earthquakes in the CEUS from 1973 - 2014 .................... 79

4.4

Number of associated injection wells through time ...................................... 81

4.5

Associated and non-associated earthquakes per year in the CEUS ............... 83

4.6

Sensitivity of number of associated earthquakes to spatial filter radius ........ 84

4.7

Yearly percent of earthquakes associated ...................................................... 85

4.8

Well completions through time ..................................................................... 87

4.9

Maps of cumulative injected volume from 2002-2013............................. 88-89

4.10

Statistical analysis of associated and non-associated injection wells ............ 91

4.11

Sensitivity of injection well statistical analysis to spatial filter radius .......... 92

4.12

Well operational analysis using only earthquakes M3.0+ ............................. 93

4.13

Well operational analysis broken down by state for SWD wells .................. 95

4.14

Well operational analysis broken down by state for EOR wells ................... 96

4.15

Cumulative injected volume within 15 km of associated earthquakes .......... 97

4.16

Statistical analysis of maximum wellhead pressure ...................................... 99

4.17

Statistical analysis of maximum wellhead pressure using only earthquakes M3.0+ .......................................................................................................... 100

4.18

Histogram of injection depths by well type ................................................. 103

xi

4.19

Injection well proximity to crystalline basement and percent association ....................................................................................... 104

4.20

Histogram of injection depths by well type using only earthquakes M3.0+ .............................................................................. 105

4.21

Injection well proximity to crystalline basement and percent association using only earthquakes M3.0+..................................................................... 106

4.22

Map of Class II injection wells in Michigan and injection rates ................. 108

4.A3.1 Spatial ETAS modeling results ..................................................................... 125 4.A3.2 Temporal ETAS modeling results ................................................................ 126


Maps of seismicity and injection rates in the Raton Basin for different time periods ranging from 1966 to 2011. ...................................... 132

5.2

Histograms of earthquakes and injection volume in the Raton Basin from 1973- 2012 .......................................................................................... 134

5.3

Stratigraphic cross section of the 2001 earthquake sequence ...................... 135

5.4

Stratigraphic cross section of the 2011 earthquake sequence ...................... 137

5.5

Map of the boundaries of the Raton Basin................................................... 139

5.6

Elevation of the top of the Pierre Shale formation ...................................... 143

5.7

Elevation of the top of the Dakota Formation ............................................. 145

5.8

Elevation of the top of the Precambrian crystalline basement ..................... 146

5.9

Stratigraphic cross section along strike of earthquake sequences ............... 148

5.10

Hydraulic head map of the Dakota Sandstone ............................................. 152

xii

5.11

Underpressure map of the Dakota Sandstone .............................................. 154

5.12

Geothermal gradient in sedimentary formations.......................................... 155

5.13

Maximum injection rate of wells through time............................................ 161

5.14

Cumulative injected volume of wells through time ..................................... 162

5.15

Stratigraphic cross section with injection well depths ................................. 164

5.16

Stratigraphic cross section of the Dakota potentiometic surface in the injection wells of interest for the 2001 and 2011 earthquake sequences ..... 167

5.17

Boundary conditions of the hydrogeologic model ....................................... 169

5.18

Vertical grid discretization of hydrogeologic model ................................... 170

5.19

Homogeneous scenario model setup ............................................................ 172

5.20

Pressure head changes at wells in homogeneous scenario........................... 173

5.21

Pressure head changes at wells in homogeneous scenario: increased Ss ..... 174

5.22

Heterogeneous scenario 1 model setup ........................................................ 176

5.23

Pressure head changes at wells in heterogeneous scenario 1 ....................... 177

5.24

Heterogeneous scenario 2 model setup ........................................................ 179

5.25

Pressure head changes at wells in heterogeneous scenario 2 ....................... 180

5.26

Cross-section of pore pressure and earthquakes .......................................... 182

5.27

Plan view of pore pressure and earthquakes ................................................ 183

5.28

Cross-section of pore pressure and earthquakes: lower sediment K............ 184

5.29

Plan view of pore pressure and earthquakes: lower sediment K ................. 185

5.30

Permeable fault model setup and results ...................................................... 187

xiii

LIST OF TABLES

CHAPTER 2 TABLES 2.A1.1 Catalog of water level responses to earthquakes at Devils Hole ................. 33

CHAPTER 3 TABLES 3.1

Injection rates from other case studies of induced seismicity..................... 65


Documented induced seismicity identified by spatiotemporal filter .......... 82

4.2

Test of sediment thickness map accuracy ................................................. 102

4.A1.1 State-by-state list of injection well database sources and attributes .......... 116 4.A3.1 Calibrated ETAS parameters for four time periods ................................... 123 4.A4.1 Suspected mining blasts removed from the ANSS earthquake catalog ..... 128


Stratigraphy of interest in the Raton Basin .................................................. 142

5.2

Raton Basin injection wells and their operations ......................................... 157

5.3

Raton Basin injection well depths and perforated depths ............................ 159

xiv

CHAPTER 1 INTRODUCTION 1.1 Background In the last half-century, scientists discovered the critical role fluids play in the nucleation of both natural and induced earthquakes (1). However, several centuries prior to modern scientific research, careful observers had documented fluid-earthquake interactions. Observations of large earthquakes altering streamflow, stream temperature and spring discharge were documented as early as the time of the Roman Empire (2). These novel observations showed earthquakes were capable of perturbing the hydrogeologic system, but it would take several centuries for the community to discover that fluids could also perturb fault systems, producing earthquakes in regions far from tectonic boundaries. Fluid injection-induced earthquakes were one of the key observations of fluid-fault interaction. The first case study of injection-induced earthquakes began in March 1962, when a wastewater disposal well at the Rocky Mountain Arsenal in Denver, Colorado commenced operation. Within several months, earthquakes began to occur in the Denver area, a region not commonly known for earthquake occurrence. By November 1965, a correlation was established between the amount of fluid injected at the Rocky Mountain Arsenal well and the number of earthquakes recorded in the region (3). The well was shut-in a short time later and the earthquakes eventually ceased, becoming known popularly as the Denver Earthquakes. The primary mechanism driving injection-induced earthquakes was first proposed by researchers studying these earthquakes (4). They invoked the theory of effective stress (5), which was first applied to subsurface faults by Hubert and Rubey (1959). The theory states that the total

1

stress (ST) on a rock mass is partially supported by the pore fluid pressure (p) and the effective stress ( ):

(1.1)

=

+

Researchers used this mathematical model to show that increases in pore fluid pressure reduce the effective stress on fault planes and promote slip (6). Due to the importance of fluid-pressure on induced earthquake nucleation, hydrogeologic models became a primary tool used to test and confirm the mechanism. By calculating the timing and magnitude of fluid-pressure changes from injection, researchers could help establish whether or not a given set of earthquakes were induced. Field and hydrogeologic modeling studies of induced earthquakes at the Rocky Mountain Arsenal in Denver, Colorado and the Rangely Oil Field in Rangely, Colorado ultimately confirmed the mechanism proposed by Healy et al. [1968] (7-8). In the decades following their discovery, injection-induced earthquakes were seen as a geologic phenomenon that did not pose a significant hazard. This view was substantiated by the fact that induced earthquakes were rare and had typically been low magnitude (< M 5.0). However, within in the last 5 years, several induced earthquakes have had damaging effects, including the 2011 M 5.6 Prague, Oklahoma earthquake (9), the 2011 M 5.3 Trinidad, Colorado earthquake (10) and the 2012 M 4.8 Timpson, Texas earthquake (11). The state of Oklahoma has been most impacted by the recent surge in induced earthquakes. From 1978-2008, the state experienced about one M 3.0+ earthquake per year. In 2014 alone, the state recorded 584 M 3.0+ earthquakes. By August of 2015, the number of

2

M3.0+ earthquakes had exceeded 2014's total. The U.S. Geological Survey concluded that injection-induced earthquakes have doubled the risk of a damaging M 6.0+ earthquake in Oklahoma (12). This significant seismic rate change demands a better understanding of fluidfault interaction in order to mitigate the hazard. Specifically, one must first quantify the geologic, hydrogeologic and well operational parameters which increase the likelihood of injectioninduced earthquakes.

1.2 Scope of Work This dissertation investigates fluid-fault interaction in both natural and induced earthquakes. Four main chapters are presented in the format of peer-reviewed journal articles. In Chapter 2, I examine a 24-year water level record to characterize water level responses to distant natural earthquakes. Devils Hole, located in Death Valley National Park, Nevada, is a fluid-filled fault cavern which formed in a highly fractured carbonate-rock formation. I document more than 200 water level responses to remote earthquakes at Devils Hole over the 24-year record. A comparison of the sensitivity of Devils Hole to other water levels worldwide shows that Devils Hole is much more sensitive to earthquakes than other water levels documented in the literature. The remaining chapters of this dissertation examine the relationship between fluid injection and the unprecedented seismic rate increase in the U.S. mid-continent beginning in 2009. Chapter 3 examines an uncharacteristic earthquake swarm in central Oklahoma and its relationship to fluid injection in the region. I develop hydrogeologic models of fluid-pressure changes from more than 70 injection wells in central Oklahoma. These models show how four high-rate injection wells, operating close to one another, caused 85% of the reservoir fluidpressure changes at depth. More than 65 low- and medium-rate wells, despite many operating for

3

more than a decade prior to the induced earthquakes, only contribute ~15% of the reservoir fluidpressure changes. These clustered, high-rate wells were able to induce earthquakes tens of kilometers from the injection point because they injected into a highly permeable reservoir near permeable basement faults. In Chapter 4, I study the relationship between injection wells and earthquakes at the scale of the U.S. mid-continent. This broad-scale approach, which has never been attempted before, aims to quantify whether or not injection wells in proximity to earthquakes are operationally distinct from injection wells which are not. To answer this question at the scale of the U.S. midcontinent, I compiled a database of more than 187,000 injection wells in central and eastern U.S. Using a set of injection well-earthquake association criteria, I delineate associated injection wells from non-associated injection wells. These two well populations are then statistically compared across several key operational parameters: injection rate, cumulative injected volume, injection pressure, injection depth and proximity to crystalline basement. I find that high-rate injection wells, wells operating at rates greater than 300,000 barrels per month, are statistically much more likely to be associated with earthquakes than lower rate wells. I also find that wells operating more than 7 km from crystalline basement have nearly no association with earthquakes. None of the other operational parameters tested with our statistical model were show to significantly alter a well's likelihood of earthquake association. The final chapter of this dissertation investigates a novel case of injection-induced seismicity in the Raton Basin of southern Colorado and northern New Mexico. Since 2001, the basin has experienced an 10-fold increase in seismicity since 2001. Earthquake studies of the region have concluded these earthquakes were induced (10), but ambiguities remain in determining the physical mechanism of the quakes. The induced sequences were unique in that

4

few earthquakes occurred at the injection depths, but instead almost all occurred between 0.5-7 km below the injection interval. Geologic, hydrogeologic and injection data were compiled to create a suite of hydrogeologic model which test whether fluid-pressure can migrate to depths several kilometers below the injection interval.

1.3 References [1] W.L. Ellsworth, Injection-induced earthquakes, Science. 341 (2013). [2] M. Manga., I. Beresnev, E.E. Brodsky, J.E. Elkhoury, D. Elsworth, S. Ingebritsen, D.C. Mays, and C.-Y. Wang (2012), Changes in permeability by transient stresses: Field observations, experiments and mechanisms, Reviews of Geophysics, v. 50, RG2004. [3] D. M. Evans, The Denver area earthquakes and the Rocky Mountain disposal well, The Mountain Geologist. 3, 23–26 (1966). [4] J. H. Healy, W.W. Rubey, D. T. Griggs, C.B. Raleigh, The Denver Earthquakes, Science. 161, 1301-1310 (1968). [5] K. Terzaghi, The shear resistance of saturated soils, Proceedings for the 1st. International Conference on Soil Mechanics and Foundation Engineering, Cambridge, MA. 1, 54 56 (1936). [6] M. Hubbert and W. Rubey, Role of fluid pressure in mechanics of overthrust faulting: 1. Mechanics of fluid-filled porous solids and its application to overthrust faulting, Bull. Geol. Soc. Am. 70, 115–166 (1959). [7] C.B. Raleigh, J. H. Healy, J. D. Bredehoeft, An experiment in earthquake control at Rangely, Colorado, Science. 191, 1230-1237 (1976). [8] P.A. Hsieh, J.D. Bredehoeft, A reservoir analysis of the Denver earthquakes: a case of induced seismicity, J. Geophys. Res. 86, 903-920 (1981). 5

[9] K. M. Keranen, H. M. Savage, G. A. Abers, E. S. Cochran, Potentially induced earthquakes in Oklahoma, USA: links between wastewater injection and the 2011 Mw5.7 earthquake sequence, Geology. 41, 699-702 (2013). [10] J. L. Rubinstein, W.L. Ellsworth, A. McGarr, H. Benz, The 2001-present induced earthquake sequence in the Raton Basin of northern New Mexico and southern Colorado, Bull. Seismol. Soc. Am. 104, 1-20 (2014). [11] C. Frohlich et al., The 17 May 2012 M4.8 earthquake near Timpson, East Texas: An event possibly triggered by fluid injection, J. Geophys. Res. 119, 581–593 (2014). [12] M. D. Petersen et al., Incorporating induced seismicity in the 2014 United States National Seismic Hazard Model -- Results of 2014 workshop and sensitivity studies, U.S. Geo. Surve Open-File Report. 2015-1070 (2015).

6

CHAPTER 2 INSIGHTS INTO WATER LEVEL RESPONSE TO SEISMIC WAVES: A 24-YEAR HIGHFIDELITY RECORD OF GLOBAL SEISMICITY AT DEVILS HOLE Abstract We studied the 24-year record of water level responses in Devils Hole, Death Valley National Park, NV to dynamic crustal stresses from earthquakes. The continuous water level record exhibited 219 responses from earthquakes around the world, displaying hydroseismogram and coseismic offset types of response. We found that the water level in Devils Hole is extremely sensitive to earthquakes and the seismic energy density required to initiate both hydroseismogram and coseismic offset responses is e ~ 10-6 J/m3, two orders of magnitude smaller than previously documented. Multiple mechanisms at Devils Hole may be responsible for observed water level responses to distant earthquakes. The hydroseismogram type responses are best explained by poroelastic deformation, while coseismic offset responses are likely the result of localized permeability changes. This study could have implication to studying dynamic triggering of earthquakes, as remote earthquakes can lead to pore pressure changes and consequently effective stress changes in fluid-filled fault zones.

This chapter has been previously published: Weingarten, M. and S. Ge (2014), Insights into water level response to seismic waves: A 24 year high-fidelity record of global seismicity at Devils Hole, Geophysical Research Letters, 41(1), pg. 74-80, doi:10.1002/2013GL058418.

7

2.1 Introduction Groundwater level changes as a result of crustal stress changes from earthquakes have been studied for nearly 50 years (1-2). In the past two decades, more high-frequency water level (1 Hz) sampling has allowed for detailed analysis of water level responses to earthquakes (3-5). Using groundwater level responses to distant earthquakes to study dynamic crustal stress changes has broadly motivated researchers to seek better understanding of how earthquake waves interact with groundwater. Hydrologic responses are interpreted according to the type and magnitude of the stresses imposed on the hydrologic system (6). Earthquakes produce two general types of stress changes in the crust: static and dynamic. Static stress changes are a result of the local crustal deformation surrounding the rupture zone while dynamic stress changes are a result of the deformation caused by passing seismic waves. The relative magnitudes of static and dynamic stresses generally decrease with increasing distance from the earthquake epicenter. Static stress magnitudes fall off far more rapidly with increasing distance than dynamic stress magnitudes. Employing the convention defined by Wang and Manga [2010], hydrologic responses within one rupture length of the fault are termed near-field responses, between 1 and 10 rupture lengths are intermediate field, and greater than 10 rupture lengths are far-field responses. Nearand intermediate field responses typically reflect the effects of static stress, the dominant stress in these distance ranges. Far-field responses typically record the effect of dynamic stress changes as the relative magnitude of dynamic stress change is larger than that of the static stress change (8). Studies of hydrologic responses to earthquakes often analyze responses of several well sites to one earthquake or one site to multiple earthquakes (4,9). One of the complexities of

8

interpreting several well site responses to one earthquake is the site-specific nature of the responses themselves (5). Interpretation of only a few earthquake responses at one site may not capture all potential response types because examining the water level record over a short period of time may capture some earthquakes, but miss others. This study focuses on a unique 24-year continuous water level dataset at Devils Hole, Death Valley National Park, Nevada that responded to global seismicity (Fig. 2.1). The objectives of this study are to: (1) characterize Devils Hole water level response patterns; (2) examine causal mechanisms for the observed water level responses, and (3) explore the potential for using water level responses to study earthquake dynamic triggering.

2.2 Background and Dataset Devils Hole is a fluid-filled fault cavern developed in a highly fractured and faulted carbonate-rock formation. It lies at the intersection of a northeast-trending, high-angle reverse fault and smaller northwest-trending normal faults (10). The inset of Figure 2.1 shows the geologic structural context of Devils Hole in proximity to several faults and fracture zones. The outcrop is representative of a larger, confined carbonate-rock aquifer system with a thickness of 1,500 meters (11). The vertical extent and subsurface geometry of Devils Hole are poorly constrained; however, the cavern forms a generally planar, fault feature and extends to depths of at least 150 meters (12). Previous research has shown that the water level in Devils Hole is sensitive to static stress changes from large earthquakes in the near- and intermediate-fields (13). Coseismic offsets in the water level were shown to recover in periods from days to months. The water level is also sensitive to stresses from earth tides and atmospheric pressure (14-15).

9

Figure 2.1. Worldwide distribution of earthquakes epicenters which produced water level fluctuations in Devils Hole. The number of responses by type is denoted in parenthesis next to the response type. The inset shows a shaded relief map of the location of Devils Hole in southwestern Nevada with solid red lines representing major faults. The seismometer used in the study (TPNV) is 56 km north of Devils Hole. Right panel shows schematic diagrams of the three characteristic types of water level response. The distribution of earthquake epicenters roughly outlines patterns of global seismicity with water level responses observed at epicentral distances of 14,749 km.

10

Since August 1989, the National Park Service has monitored the water level in Devils Hole with multiple instruments. Continuous water level was recorded by a float-pulley device and an accompanying strip chart recorder (Fig. 2.2). The strip chart recorder graphically records real-time water level relative to a fixed elevation. The majority of the water level was recorded on 0.0005 m of strip chart per 30 minutes. The resolution of the float-pulley device is 0.003 m and the maximum possible recorded water level fluctuation recorded is 0.3 m (16). Two pressure transducers and an electronic data logger were also added to record water levels to the nearest 0.0003 m on 15-minute intervals. The strip chart recorder was removed in May 2010. In December 2012 a new pressure transducer was installed and a data logger reprogrammed to record water level at 1-second (1 Hz) intervals when an offset of 0.003 m was detected. The updated monitoring protocol records higher frequency, digital water level data. Seismic data were examined to assess whether any specific earthquake has caused a water level fluctuation. We used the U. S. Geological Survey National Earthquake Information Center (USGS-NEIC) earthquake catalog and ground motion data recorded at seismic station TPNV, located 56 km north of Devils Hole (Fig. 2.1). TPNV is maintained by the U.S. seismometer network, USArray, and is the closest broadband station operating during the period of the Devils Hole water level record. The instrument is a 3-component broadband seismometer with a response range from 0.01 to 10 Hz. We used the seismic data to help determine the type of seismic waves causing the water level fluctuations at Devils Hole in the high-frequency water level record. We also used the TauP toolkit, a seismic wave travel time calculator, to predict arrival times of various seismic waves at Devils Hole from a specific earthquake (17).

11

Figure 2.2. (A) The pool at Devils Hole with human for scale. (B) Water level monitoring setup at Devils Hole. The high-frequency transducer (HF Transducer) was installed on September 20th, 2012. The chart recorder was originally installed in 1989, but has since been upgraded several times.

12

2.3 Identifying Seismically Induced Water Level Fluctuations A protocol was developed to identify and catalog seismically induced water level responses using the 24 years of continuous water level data. We first flag any abrupt changes in water level from 1989 to 2010 from the strip chart record. The digital pressure transducer data were also used to identify and corroborate the timing of abrupt water level changes in the strip chart record. We then checked the USGS-NEIC earthquake database for corresponding seismic events and calculated seismic wave arrival times at Devils Hole. Seismically induced water level responses were identified to be abrupt changes in water level recorded by the strip chart and digital water level data whose timing correlated with seismic wave arrivals (Fig. 2.3). Once these criteria were met, the maximum amplitude and duration of the water level response were taken from the strip chart record. In addition, the corresponding earthquake time, magnitude, depth, azimuth and distance to Devils Hole were cataloged. There are two factors, in addition to earthquakes, that can cause abrupt water level changes. The first is the rare storm precipitation that is readily identifiable by its distinctive character of small (~0.003 m) amplitude oscillations persisting for hours at a time. The second is monthly or bi-weekly instrument checks. Both non-earthquake related changes have been accounted for in the creation of the catalog. The water level is also sensitive to cyclic barometric pressure and earth tides. These cyclic effects on the water level were filtered from the digital water level record using timeseries analysis software (18). The two principal tidal constituents that cause water level to fluctuate are the O1 diurnal and M2 semi-diurnal lunar tides and their periods are 25.82 hours (~1 cpd) and 12.42 hours (~2 cpd), respectively. The dominant frequencies in the unfiltered water level time series are 1 cpd and 2 cpd, reflecting the O1 and M2 lunar tides (19). To remove these

13

Figure 2.3. Devils Hole water level responses to earthquakes as recorded by the Stephens Chart Recorder (16). Each panel represents a water level response to a specific earthquake as recorded in the chart record. (A) Hydroseismogram response to the November 14th, 2001 M 7.8 China earthquake. (B) Hydroseismogram response to the October 24th, 1993 M 6.7 Oaxaca, Mexico earthquake. (C) Hydroseismogram response to the October 4th, 1994 M 8.3 Kuril Islands earthquake. (D) Coseismic + hydroseismogram response to the September 14th, 1995 M 7.4 Oaxaca, Mexico earthquake. All panels exhibit the semi-diurnal response of the Devils Hole water level to earth tides.

14

frequency components from the water level time series data, we implemented a high-pass fast Fourier Transform filtering technique using a cut-off frequency of 5 cycles per day (cpd) such that the tidal and barometric influences are removed and higher frequency seismically induced water level fluctuations remain in the time series.

2.4 Analysis of Water Level Response 2.4.1 Characteristic Types of Water Level Response Using the protocol described above, we identified 219 worldwide earthquakes that produced water level responses in the 24-year Devils Hole water level record (Fig. 2.1; Table 2.A1.1). The water level response exhibited three characteristic types: (1) hydroseismogram, characterized by high-frequency water level oscillation with no persistent change in water level; (2) coseismic offset and a hydroseismogram; and (3) coseismic offset only. The right panel in Figure 2.1 shows the characteristics of these three types of water level responses. Earthquake epicenters shown in yellow in Figure 2.1 produced hydroseismogram response and those shown in red produced response types (2) and (3). For purposes of further discussion we do not make a distinction between the coseismic offset response types (2) and (3), because we believe they are attributable to a similar mechanism of water level response. The hydroseismogram responses comprise the vast majority of water level responses with 183 events. Coseismic offsets were observed for 36 events. The hydroseismogram response is observed at epicentral distances as far as 14,749 km while the coseismic offset responses were limited to distances of less than 4,000 km from Devils Hole. Another notable observation is that the Devils Hole water level does not only respond to large earthquakes but can respond to small

15

earthquakes hundreds of kilometers away. For example, coseismic offset was observed from a M3.3 earthquake at an epicentral distance of 231 km.

2.4.2 Water Level Sensitivity Earthquake magnitude and distance are the two most important parameters controlling the amount of stress at a specific location imposed by an earthquake. Wang [2007] introduced a physical quantity, seismic energy density e, which defines the maximum seismic energy available to do work in a unit volume of rock or sediment. Equation (1) relates the seismic energy density (J/m3) at a distance, r (km), from an earthquake source of magnitude (Mw):

log r = 0.48 Mw – 0.33log e -1.4

(2.1)

Using this general equation of seismic energy, the seismically induced water level fluctuations at Devils Hole can be conveniently compared to other well sites as well as other earthquake-related hydrologic phenomena such as streamflow responses, liquefaction or the eruption of mud volcanoes (7). This metric does not take into account such effects as the focusing of seismic waves around a spherical earth, but these effects do not change the order-ofmagnitude of seismic energy (21). We plotted the cataloged earthquake events using the magnitude versus epicentral distance to Devils Hole in Figure 2.4. Previously observed worldwide water level responses cataloged by Wang and Manga [2010] are also plotted for reference. A threshold of seismic energy density for water level responses at Devils Hole is observed to be e ~ 10-6 J/m3. This threshold is two orders of magnitude below previously documented empirical thresholds of 10-4

16

Figure 2.4. Hydrologic responses to earthquakes at Devils Hole and from a worldwide catalog of well responses compiled by Wang and Manga [2010]. Epicentral distance is the distance between either Devils Hole and the earthquake epicenter or the observed water well and the earthquake epicenter. Magnitude is the moment magnitude of the earthquake. Solid contour lines show equal seismic energy density in unites of J/m3 based on the model of Wang [2007]. The horizontal dashed line denotes the ~4000 km distance threshold observed for coseismic offset responses. The maximum amplitude from the chart record is plotted as the size of the plotted circles.

17

J/m3 suggesting that Devils Hole appears to be the most sensitive to seismic waves among all seismically induced hydrologic responses reported in the literature (6). The amplitude of water level responses at Devils Hole to e ~ 10-6 J/m3 is as much as 0.1 m. The magnitude of the water level variation, shown by the size of the plotted circles in Figure 2.4, generally shows larger magnitudes correspond to larger seismic energy densities. However, the relationship is less clear between the magnitude of the water level variation and the magnitude or distance of an earthquake. Figure 2.4 also shows that the seismic energy density required to trigger both hydroseismogram and coseismic offset responses is similar, i.e., even though the number of hydroseismogram responses is substantially greater than that of the coseismic offset responses, the seismic energy required to initiate the hydroseismogram response is not smaller than that required to initiate coseismic types of responses. The calculated seismic energy density alone does not discern what type of response may occur at Devils Hole. Hydroseismogram responses occur up to epicentral distances of 14,749 km. The coseismic offset responses were only observed up to epicentral distances of 4,000 km (Fig. 2.1; Fig. 2.4). This 4,000-km distance distinguishes hydroseismogram responses from the coseismic offset responses, yet the threshold of seismic energy density is similar for all responses. Attenuation of seismic energy suggests that seismic wave frequencies are lower at increasing distances from the earthquake source, as a result of energy loss to internal friction (22). Thus, the frequency content from earthquakes beyond 4000-km is generally lower. Therefore, we suggest that below a certain frequency content only hydroseismogram type responses occur and no coseismic offset type responses occur at Devils Hole. This observation contrasts with some studies that have found that hydrologic responses such as mud volcanoes are more sensitive to seismic waves with low frequency content (23).

18

2.4.3 Characteristic Frequency of Water Level Response Frequency analysis of the water level response to earthquakes is based upon water level data collected from December 2012 to May 2013 measured at 1 Hz frequency. Devils Hole responded to five earthquakes in the far-field during this time period. All five earthquakes produced hydroseismogram type responses with no coseismic offset. The attributes of these five earthquakes are listed in Figure 2.5. To discern the dominant frequencies of the water level responses, we conducted spectral analysis of the responses to these earthquakes. Figure 2.5 shows the amplitude spectrums of the water level response in the frequency domain. The water level response exhibited a characteristic frequency range between 0.04 Hz and 0.1 Hz. The lower and upper bounds on the frequency range correspond to the farthest and closest earthquakes, respectively. The lower bound, 0.04 Hz, is from the M7.8 Iran earthquake at an epicentral distance of 12,837 km. The upper bound, 0.1 Hz, is from the M6.3 San Diego earthquake at an epicentral distance of 669 km. An analytical model presented by Liu et al. [1989] found that Rayleigh wave components of seismic waves with frequencies of 0.04 Hz to 0.07 Hz produce characteristic water level responses in the same frequency range. The frequency range of the hydroseismogram response at Devils Hole is well predicted by, and provides the field data support for, the analytical models of water level response to passing Rayleigh waves (24-25).

2.4.4 Water Level Response to Shear Waves While spectral analysis of the water level response gives insight into dominant frequencies of water level oscillation, comparison of the high-frequency water level data with seismic wave arrivals at the TPNV broadband seismic station show the water level is sensitive to both shear and Rayleigh waves (Fig. 2.6). The January 5th, 2013 earthquake near British

19

Figure 2.5. Water level amplitude spectrums of the hydroseismogram water level responses to the five earthquakes listed on the table inset. The yellow band highlights the characteristic frequency range of water level response between 0.04 Hz and 0.1 Hz. The table inset lists the attributes of the five earthquakes analyzed with 1 Hz water level data.

20

Columbia, Canada (epicentral distance = 2522 km) caused a water level response at Devils Hole. The predominant water level response occurs to Rayleigh waves about 15 minutes after the earthquake origin time with the maximum amplitude of water level oscillation 0.43 m. However, coincident with the observed shear wave arrival at ~561 seconds after the earthquake origin time, we also observe a water level oscillation of 0.021 m amplitude at Devils Hole (Fig. 2.6). To our knowledge, there has been only one direct observation of seismically induced water level responses to shear waves (26). Shear waves reflect shear particle motion with no associated volume change and therefore would not be expected to induce water level response. For the water level to respond to shear waves, the hydrogeologic system might have a component of anisotropy to convert shear wave energy into volumetric strain (4). Our observation of shear wave triggering water level response possibly indicates a component of anisotropy in the Devils Hole fractured-rock aquifer system. Of the 5 earthquake responses studied in Figure 2.5, the water level responded to both shear and Rayleigh waves for 3 out of the 5 earthquakes listed. The two earthquakes which produced only Rayleigh wave responses were the M6.3 San Diego and M7.8 Iran earthquakes at epicentral distances of 690 km and 12,837 km, respectively. These earthquake responses also exhibited the highest and lowest frequency of water level response, respectively. The earthquakes which generated water level shear wave responses and those that did not exhibited similar values of peak ground velocity. The water level response to shear waves may be dictated by the azimuth of the incident seismic waves at Devils Hole. The two water level responses which did not respond to shear waves were oriented at 002 and 208 degrees, while the three which did respond to shear waves were oriented between 272 and 331 degrees.

21

Figure 2.6. (top) Water level in Devils Hole in response to January 5th, 2013 M 7.5 earthquake near British Columbia, Canada. (bottom) Ground velocity recorded at seismometer TPNV in the Radial, Transverse and Vertical directions. The P-,S- and Rayleigh waves are denoted on each figure respectively. Note the small water level oscillations start to occur at ~561 seconds, which corresponds to the shear wave arrival at TPNV.

22

Figure 2.7 shows the incident seismic wave azimuths at Devils Hole for earthquakes which produced responses and those that did not. The aquifer system connected to Devils Hole is more sensitive to seismic waves incident along a NE-SW strike line. While there are several hundred earthquakes which produce seismic waves incident along a SE-NW strike like, Devils Hole is far less sensitive to earthquakes from those azimuths. This is well-explained by the orientation of the fracture system connected to Devils Hole reproduced from Riggs et al. [2002]. Devils Hole forms a wide cavern along a SW-NE strike line, but a narrow fracture along a strike line of NW-SE. Thus, incident seismic waves from azimuths trending NW-SE are most effective at producing water level responses at Devils Hole.

2.5 Mechanisms of Water Level Response Various mechanisms have been proposed for far-field water level responses to earthquakes. Cooper et al. [1965] suggested that the hydroseismogram is primarily due to passing Rayleigh waves causing aquifer dilatation and vertical ground motion. This mechanism explains the majority of response at Devils Hole. The hydroseismogram responses at Devils Hole exhibit no persistent change in water level as well as a characteristic frequency range that is consistent with theoretical predictions (Fig. 2.5). Near-field coseismic offsets at Devils Hole have already been shown to be the result of static poroelastic deformation (13). However, the observation of coseismic offset responses from far-field earthquakes demands an alternate mechanism to explain these water level responses. Several studies have suggested (1) opening and closing of fractures; (2) clogging and unclogging of fractures by colloidal particles; or (3) mobilization of trapped bubbles through pore throats or fracture asperities (1-2, 28). All three alternative mechanisms imply that permeability change causes far-field coseismic offset. Devils

23

Figure 2.7. (A-B) Cross-section of Devils Hole from (A) SW-NE and (B) NW-SE reproduced from Riggs et al. [2002]. (C-D) Comparison of incident seismic wave azimuths for earthquakes which caused water level responses at Devils Hole (red) versus all earthquakes greater than M 6.0 over the same time period.

24

Hole outcrops in a highly fractured and faulted carbonate-rock aquifer thus being a suitable candidate to entertain one of these alternative mechanisms that involve fracture permeability. Figure 2.8 shows the timing and magnitude of response types as well as the sign of the coseismic offsets over the cataloged record. Red circles indicate the timing of coseismic offset responses from near- and far-field earthquakes while yellow circles indicate the timing of hydroseismogram responses. Hydroseismogram responses occur frequently throughout the record while coseismic offset responses occur less frequently and often separated by several months to a year. The sign of coseismic offsets are shown in the water level record and varied from response to response with no coherent pattern. Two clusters of coseismic offset responses occur during the time periods of the 06/28/1992 M 7.2 Landers Earthquake and the 10/16/1999 M 7.1 Hector Mine earthquake. These large, near field responses have been shown to be the result of deformation-induced fluid flow from the static stresses associated with these earthquakes (13). We interpret these responses as separate from the far-field coseismic offset response. The mean time interval between coseismic offset responses is 210 days with a standard deviation of 194 days. This time interval may be representative of the permeability recovery time. However, the large standard deviation shows time intervals may also differ by several months. Brodsky et al. [2003] suggested that an aquifer system can be interpreted to contain little gaseous phase if its water level shows a great sensitivity to earth tides. This is due to the large compressibility of gas relative to that of water. Thus, being sensitive to earth tides, gas bubble mobilization at Devils Hole is unlikely. Therefore, we favor colloid mobilization as a possible mechanism for the far-field coseismic step-like responses. We further argue that fractures in the aquifer system can be clogged or flushed as colloids accumulate and mobilize over time. The

25

Figure 2.8. Water level data from the 15-minute digital water level record at Devils Hole are plotted through time. Earth tides were filtered out. Plotted above the water level are the timing and maximum amplitude from the chart record of the hydroseismogram (yellow circles) and coseismic offset (red circles) responses in the catalog.

26

coseismic offset responses to far-field earthquakes could be indicative of these local processes with the recovery time for these processes on the order of several months. On the other hand, the hydroseismogram responses do not depend on these processes and they occur much more frequently through time, as illustrated in Figure 2.8. The variety of hydrologic response types at Devils Hole shows that for the same site, multiple causal mechanisms may be responsible for inducing far-field water level responses. At Devils Hole, the mechanism may be dependent on the frequency content of the incident seismic waves. For the same seismic energy density, the lower frequency content of seismic waves from a more distant earthquake produced only the hydroseismogram response as opposed to coseismic offset responses.

2.6 Conclusions Examining the 24-year record of the water level responses in Devils Hole to earthquakes worldwide, we identified 219 responses events of both hydroseismogram and coseismic offset type response. Analysis of the data led to following conclusions: •

The water level is extremely sensitive to earthquakes with seismic energy densities as small as e ~ 10-6 J/m3 from distant earthquakes inducing water level responses at Devils Hole.

•

The hydroseismogram type of water level response exhibited a characteristic frequency range between 0.04 Hz and 0.1 Hz and provides field data support for prior analytical models of water level response to passing Rayleigh waves.

•

High-frequency water level analysis reveals a water level response to shear waves.

27

•

The water level response type to distant earthquakes may be dictated by the frequency content of the passing seismic waves.

•

Different mechanisms, even at a single site, may be responsible for water level response to distant earthquakes. The hydroseismogram type of responses are best explained by poroelastic deformation due to passing waves, while the coseismic offset type responses are likely a result of localized permeability changes.

This study could have implication in studying dynamic triggering of earthquakes. The water level at Devils Hole exhibited a maximum amplitude of 0.1 m oscillation in response to seismic energy densities as small as e ~ 10-6 J/m3. This roughly translates to a pore pressure change of 1 kPa in the aquifer system. Brodsky and Prejean [2005] observed triggered seismicity at Long Valley Caldera from dynamic stresses as small as ~5 kPa. Van der Elst and Brodsky [2010] further dropped the dynamic strain threshold for earthquake triggering by several orders of magnitude with increased seismicity in far-field areas after large earthquakes suspected to be caused by altered pore pressure gradients or changes in permeability (31). This study shows that remote earthquakes can lead to pore pressure changes at distances not previously observed. These pore pressure changes could in turn contribute to dynamic triggering of earthquakes. The hydrogeologic processes seen at Devils Hole could be operating in geologic settings such as subduction zones to cause changes in effective stresses in fluid-filled fault zones in response to far-field earthquakes.

28

2.7 Acknowledgements The authors would like to thank Michael Manga and an anonymous reviewer for their thoughtful comments. This research was generously funded by the National Park Service and the University of Colorado-Boulder. The authors would like to thank P.A. Cutillo, Jennifer Back, and Chris Gable for their knowledge of Devils Hole and help acquiring the dataset. We thank Will Yeck for his discussions on the topic and Ryan Schnell for valuable comments on the frequency techniques applied.

2.8 References [1]

J.D. Bredehoeft, H. H. Cooper, Jr., I. S. Papadopulos, and R. R. Bennet, Seismic fluctuations in an open artesian water well, U.S. Geol. Surv. Prof. Paper, 525-C (1965).

[2]

E. Roeloffs, Persistent water level changes in a well near Parkﬁeld California, due to local and distant earthquakes, J. Geophys. Res. 103, 869–89 (1998).

[3]

M. Ohno, H. Wakita, and K. Kanjo, A water well sensitive to seismic waves, Geophys. Res. Letters. 23, 691-694 (1997).

[4]

E. E. Brodsky, E. Roeloffs, D. Woodcock, I. Gall, and M. Manga, A mechanism for sustained groundwater pressure changes induced by distant earthquakes, J. Geophys. Res. 108, 2390 (2003).

[5]

E. E. Roeloffs, D. L. Nelms, R. A. Sheets, W. L. Cunningham, M. Kozar, J. Williams, C. Dawson, and D. Risser, Groundwater level changes caused by strain and seismic shaking from the August 23, 2011 Mw 5.8 Virginia earthquake, Abstract S14B-07 presented at 2011 Fall Meeting, AGU, San Francisco, Calif., 5–9 Dec (2011).

29

[6]

M. Manga, I. Beresnev, E.E. Brodsky, J.E. Elkhoury, D. Elsworth, S. Ingebritsen, D.C. Mays, and C.-Y. Wang, Changes in permeability by transient stresses: Field observations, experiments and mechanisms, Reviews of Geophysics. 50, (2012).

[7]

C. Y. Wang, M. Manga, Hydrologic responses to earthquakes - a general metric, Geofluids, 10, 206-216 (2010).

[8]

D. Kilb, J. Gomberg, and P. Bodin, Aftershock triggering by complete Coulomb stress changes, J. Geophys. Res. 107, 2060 (2002).

[9]

S. C. Cox, H.J. Rutter, A. Sims, M. Manga, J.J. Weir, T. Ezzy, P.A. White, T.W. Horton and D. Scott, Hydrological effects of the Darfield (Canterbury) Mw 7.1 earthquake, 4 September 2010, New Zealand, New Zealand Journal of Geology and Geophysics. 55, 231-247 (2012).

[10] G. D. Robertson, Ge, S., and P. A. Cutillo, An investigation of regional tectonic strain on water levels in Devils Hole, Death Valley National Park, Nevada, Geophysical Research Letters, 34, (2007). [11] J. Harrill, D. Prudic, Aquifer systems in the great basin region of Nevada, Utah, and adjacent states-summary report, U.S. Geol. Surv. Prof. Paper. 1409-A (1998). [12] A. C. Riggs, W.J. Carr, R.T. Kolesar, and R.J. Hoffman, Tectonic speleogenesis of Devils Hole, Nevada, and implications for hydrogeology and the development of long, continuous paleoenvironmental records, Quaternary Research. 42, 241-254 (1994). [13] P. A. Cutillo, and S. Ge, Analysis of strain-induced ground-water fluctuations at Devils Hole, Nevada, Geofluids. 6, 319–333 (2006). [14] W.W. Dudley, Jr., J.D. Larson, Effect of Irrigation Pumping on Desert Pupfish Habitats in Ash Meadows, Nye County, Nevada, U.S. Geol. Surv. Prof. Paper. 927, 52p (1976).

30

[15] D. Galloway, and W. E. Wilcoxon, Earth-tide induced fluid-pressure changes in Devils Hole, Death Valley National Monument, California-Nevada, 1993 Annual Fall Meeting, EOS, Transactions, AGU. 74, 565 (1993). [16] Stevens Water Monitoring Systems (2012), Type F Chart Recorder Datasheet, [17] H. P. Crotwell, T. J. Owens, and J. Ritsema, The TauP Toolkit: Flexible seismic travel-time and ray-path utilities, Seismological Research Letters. 70, 154–160 (1999). [18] M . Van Camp, and Vauterin, P., Tsoft: graphical and interactive software for the analysis of time series and Earth tides, Computers & Geosciences, 31(5), 631-640 (2005) [19] P. A. Cutillo, and J. D. Bredehoeft, Estimating aquifer properties from the water level response to Earth tides, Groundwater. 49, 600-610 (2010). [20] C. Y. Wang, Liquefaction beyond the near field, Seismological Research Letters. 78, p. 512–517 (2007). [21] T. Lay, and H. Kanamori, Geometric effects of global lateral heterogeneity on long-period surface wave propagation, J. Geophys. Res., 90(B1), 605–621 (1985). [22] P. M. Shearer (2009). Introduction to seismology. Cambridge University Press. [23] M. L. Rudolph, M. Manga, Frequency dependence of mud volcano response to earthquakes, Geophysical Research Letters. 39, L14303 (2012). [24] L. B. Liu, E. Roeloffs, and X. Y. Zheng, Seismically induced water level fluctuations in the Wali well, Beijing, China, J. Geophys. Res. 94, 9453-9462 (1989). [25] H. H. Cooper, Jr., J. D. Bredehoeft, I. S. Papadopulos, and R. R. Bennet, The response of well-aquifer systems to seismic waves, J. Geophys. Res. 70, 3915-3926 (1965).

31

[26] C. Y. Wang, Y. Chia, P. L. Wang, and D. Dreger, Role of S waves and Love waves in coseismic permeability enhancement, Geophysical Research letters, 36 (2009). [27] A. C. Riggs, and J. E. Deacon, Desert aquatic ecosystems: The Devils Hole story, paper presented at Spring-Fed Wetlands: Important Scientific and Cultural Resources of the Intermountain Region, Desert Res. Inst., Las Vegas, Nev. (2002). [28] J. E. Elkhoury, A. Niemeijer, E. E. Brodsky, and C. Marone, Laboratory observations of permeability enhancement by fluid pressure oscillation of in situ fractured rock, J. Geophys. Res. 116 (2011). [29] E. E. Brodsky, S. G. Prejean, New constraints on mechanisms of remotely triggered seismicity at Long Valley Caldera, J. Geophys. Res. 110, B04302 (2005). [30] N. J. Van der Elst, E. E. Brodsky, Connecting near-field and far-field earthquake triggering to dynamic strain, J. Geophys. Res. 115, B07311 (2010). [31] N. J. Van der Elst, H. M. Savage, K. M. Keranen, and G. A. Abers, Enhanced Remote Earthquake Triggering at Fluid-Injection Sites in the Midwestern United States, Science. 341, 164-167 (2013).

32

33

2.A Appendix Table 2.A1. Catalog of water level responses to earthquakes at Devil's Hole. Date

EQ Epicentral Location

EQ Time (hh:mm UTC)

EQ Magnitude

EQ Depth (km)

Epicentral Distance to Devils Hole (km)

Epicenter Latitude

Epicenter Longitude

Epicenter Azimuth to Devils Hole (deg)

Peak-to-Peak Amplitude (ft) Chart Recorder

Peak-to-Peak Amplitude (ft) Transducer

Response Duration (hrs)

Water Level Response Type

5/24/2013

Sea of Okhotsk

5:44

8.3

608.9

3482.920333

54.874

-153.281

318.2472269

Recorder Removed

0.17 (1 Hz data)

unknown

hydroseismogram

5/23/2013

Greenville, CA

3:47

5.7

9.7

604.415789

40.192

-121.059

317.9211867

Recorder Removed

0.17 (1 Hz data)

unknown

hydroseismogram

4/16/2013

Iran-Pakistan Border

0.05 (1 Hz data)

unknown

hydroseismogram

2/6/2013

1/5/2013

11/1/2011 10/21/2011 9/9/2011 9/2/2011

Santa Cruz Islands Off Coast of British Columbia, Canada Off Coast of Manzanillo, Mexico Coral Sea British Columbia Aleutian Islands

10:44

7.8

82

12856.81803

28.107

62.053

1.619048827

Recorder Removed

1:12

8

28.7

8284.361962

10.738

165.138

272.0890263

Recorder Removed

0.07 (1 Hz data)

unknown

hydroseismogram

8:58

7.5

9.8

2539.655715

55.368

-134.621

332.5854094

Recorder Removed

1.5 (1 Hz data)

unknown

hydroseismogram

12:32

6.3

10

1970.111995

19.83

-109.21

157.5981123

Recorder Removed

0.07

0.05

hydroseismogram

17:57

7.4

33

9595.206863

-28.99

-176.24

229.3503764

0.07

0.08

hydroseismogram

19:41

6.4

22

1689.308939

49.53

-126.89

332.9021368

0.074

0.16

hydroseismogram

10:55

6.9

32

4613.298028

52.17

-171.71

310.3384389

0.089

0.08

hydroseismogram

0.043

0.04

hydroseismogram

0.038

0.01

coseismic -- no hydroseismogram

0.125

0.16

hydroseismogram

0.067

0.25

hydroseismogram

0.532

2.25

hydroseismogram

0.036

0.08

hydroseismogram

0.13

0.33

hydroseismogram

0.12

0.2

hydroseismogram

8/20/2011

Coral Sea

18:19

7.1

28

9976.486154

-18.31

168.22

246.7937093

7/26/2011

Gulf of California

17:44

6

12

1414.419006

25.1

-109.53

151.0367495

7/6/2011

New Zealand

19:03

7.6

17

9645.767148

-29.54

-176.34

229.0282872

6/24/2011 3/11/2011 3/9/2011 12/25/2010 12/21/2010

7/18/2010

7/18/2010

Recorder Removed Recorder Removed Recorder Removed Recorder Removed Recorder Removed Recorder Removed Recorder Removed Recorder Removed Recorder Removed Recorder Removed Recorder Removed

Aleutian Islands Tohoku, Japan Tohoku, Japan

3:09

7.3

52

4622.273021

52.05

-171.84

310.1583272

5:46

9

29

8437.9934

38.3

142.37

307.4940312

2:45

7.3

32

8395.582153

38.44

142.84

307.3913107

Coral Sea

13:16

7.3

16

10095.43898

-19.7

167.95

245.8701824

17:19

7.4

14

9087.263193

26.9

143.7

297.4429608

13:34

7.3

35

10676.94683

-5.93

150.59

267.0601987

Recorder Removed

0.082

0.08

hydroseismogram

13:04

6.9

28

10693.86454

-5.97

150.43

267.122005

Recorder Removed

0.044

0.07

hydroseismogram

Marians Trench New Britian Region, Papua New Guinea New Britian Region, Papua New Guinea

Date

5/27/2010 4/4/2010 2/27/2010 1/12/2010 1/10/2010 9/29/2009 9/24/2009 8/5/2009

EQ Time (hh:mm UTC)

EQ Magnitude

EQ Depth (km)


Epicenter Latitude

Epicenter Longitude






Solomon Islands Baja California, Mexico Offshore Maule, Chile

17:14

7.2

31

9788.869213

-13.7

166.64

251.3455993

Recorder Removed

0.03

0.04

hydroseismogram

22:40

7.2

10

486.4688665

32.128

-115.303

168.9526949

Disabled

3.207

2.5

hydroseismogram

6:34

8.8

35

9226.504088

-36.12

-72.9

146.0041533

0.99

0.766

3

hydroseismogram

Haiti

21:53

7

10

4705.097167

18.44

-72.57

103.100307

0.35

0.376

0.45

hydroseismogram

0:27

6.5

29

867.7872218

40.65

-124.69

305.3104955

0.465

0.225

0.18

hydroseismogram

17:48

8.1

18

8218.285335

-15.49

-172.1

236.0594262

0.392

0.37

3.5

hydroseismogram

7:16

6.4

13

2144.271186

18.83

-107.32

153.4517422

N/A

0.065

0.12

hydroseismogram

9:13

5.8

10

792.8998588

29.61

-113.79

162.196915

N/A

0.047

0.03

hydroseismogram

Offshore of Northern California East of Coral Sea Off Coast of Manzanillo, Mexico Gulf of California

8/3/2009

Gulf of California

17:59

6.9

10

880.2581519

29.04

-112.9

157.9393914

Disabled

2.6

1.17

coseismic offset -recorder disabled

7/15/2009

New Zealand

9:22

7.8

12

11952.32082

-45.76

166.56

225.4923068

N/A

0.053

0.08

hydroseismogram

5/28/2009 3/19/2009 2/18/2009 1/15/2009 1/3/2009 1/3/2009 10/16/2008 9/24/2008

5/12/2008

3/15/2008 2/21/2008 1/5/2008

Off Coast of Honduras East of Coral Sea East of Coral Sea

8:24

7.3

19

3679.917665

16.73

-86.22

118.4939847

0.455

0.555

0.75

hydroseismogram

18:17

7.6

31

9009.969577

-23.04

-174.66

232.4875309

N/A

0.058

0.08

hydroseismogram

21:53

7

25

9476.474608

-27.42

-176.33

230.5072637

N/A

0.022

0.2

hydroseismogram

Kuril Islands

17:49

7.4

36

7054.304888

46.86

155.15

310.1631212

0.065

0.146

1.25

hydroseismogram

22:33

7.4

23

11866.79126

-0.69

133.3

281.8916776

0.1

0.156

1.25

hydroseismogram

19:43

7.7

17

11884.30585

-0.41

132.88

282.3926117

0.085

0.134

1.25

hydroseismogram

Irian Jaya Indonesia Irian Jaya Indonesia Off Coast of Guatemala Off Coast of Manzanillo, Mexico Sichuan Province, China Off Coast of Southern Oregon Wells, Nevada Off Coast of British Columbia,

19:41

6.7

24

3409.987538

14.42

-92.36

129.6216708

0.235

0.207

0.5

coseismic + hydro (down)

2:33

6.4

12

2345.477941

17.61

-105.5

150.2708007

0.06

0.061

0.15

coseismic offset (down)

6:28

7.9

19

11456.71772

31

103.32

325.8772975

0.2

0.25

1.5

hydroseismogram

14:44

5.7

10

1122.053433

42.41

-126.83

309.5839341

0.035

0.02

0.1


14:16

6

6

539.5265373

41.15

-114.87

12.76308184

N/A

0.06

0.05

hydroseismogram

11:44

6.4

10

1990.383335

51.16

-130.54

329.8545127

0.245

N/A

0.4

hydroseismogram

34


Date

EQ Magnitude

EQ Depth (km)


Epicenter Latitude

Epicenter Longitude






11:01

6.6

15

2007.475256

51.25

-130.75

329.7176532

0.159

N/A

0.2

hydroseismogram

9:30

7.2

34

5154.440611

51.36

-179.51

309.6155401

0.095

0.1

0.8

hydroseismogram

Malaysia

23:49

7.9

35

14667.67314

-2.62

100.84

305.875151

0.11

N/A

1.75

hydroseismogram

Malaysia

11:10

8.5

34

14780.41013

-4.44

101.37

303.7059263

0.3

N/A

2

hydroseismogram

23:40

8

39

6927.665373

-13.39

-76.6

135.4357918

0.28

N/A

2.6

hydroseismogram

20:39

8.1

24

10270.12451

-8.47

157.04

261.2119697

0.16

N/A

2

hydroseismogram

2:59

6

26

1254.787972

26.26

-110.54

152.6572249

0.032

N/A

0.2

coseismic offset (up)

Canada

1/5/2008

12/19/2007 9/12/2007 9/12/2007 8/15/2007 4/1/2007 3/13/2007

Off Coast of British Columbia, Canada South Bering Sea

Off Coast of Peru Solomon Islands Gulf of California

1/13/2007

Kuril Islands

4:23

8.1

10

7128.00989

46.24

154.52

309.7569583

0.515

0.55

2.5

hydroseismogram

11/15/2006

Kuril Islands

11:14

8.3

10

7195.915619

46.59

153.27

310.5348514

0.21

N/A

2

hydroseismogram

Hawaii

17:07

6.7

38

4260.325121

19.88

-155.93

255.3877686

0.035

N/A

0.5

hydroseismogram

10/15/2006 1/4/2006 10/8/2005 6/15/2005 3/28/2005

12/26/2004 12/23/2004

7/19/2004

6/28/2004

2/7/2004 12/22/2003

Gulf of California Northern Pakistan Off Coast of Eureka, CA Northern Sumatra Indonesia Northern Sumatra Indonesia South of New Zealand Off the Coast of British Columbia, Canada Off the Coast of British Columbia, Canada Irian Jaya Indonesia Cambria, CA

8:32

6.6

14

998.9943954

28.16

-112.12

155.7495961

Disabled

N/A

unknown

coseismic offset -logger disabled

3:50

7.6

26

12057.87107

34.54

73.59

351.4299289

0.045

N/A

0.75

hydroseismogram

2:50

7.2

16

994.9722069

41.29

-125.95

305.8596326

Disabled

N/A

unknown


16:09

8.6

30

14512.39376

2.09

97.11

313.6451644

0.215

N/A

2.25

hydroseismogram

0:58

9.1

30

14473.92649

3.3

95.98

315.7681994

0.58

N/A

2.5

hydroseismogram

14:59

8.1

10

12494.20366

-49.31

161.35

224.3248539

0.048

N/A

1.25

hydroseismogram

8:01

6.4

23

1700.825283

49.62

-126.97

332.9327062

0.105

N/A

0.4

coseismic + hydroseismogram (down)

9:49

6.8

20

2459.186816

54.8

-134.25

331.8313965

0.71

N/A

1.5

hydroseismogram

2:42

7.3

10

11939.43028

-4

135.02

278.0510686

0.025

N/A

0.75

hydroseismogram

19:15

6.6

7

439.3698962

35.71

-121.1

260.9832874

Disabled

N/A

unknown


35

EQ Time (hh:mm UTC)


Date

11/17/2003 9/27/2003 9/25/2003 3/12/2003

2/19/2003

1/22/2003 1/20/2003 1/16/2003

Aleutian Islands Southern Russia Northern Japan Gulf of California Unimak Island Region of Alaska Colima, Mexico Solomon Islands North Pacific Ocean off Oregon Coast

EQ Time (hh:mm UTC)

EQ Magnitude

EQ Depth (km)


Epicenter Latitude

Epicenter Longitude






6:43

7.8

33

5284.459745

51.15

178.65

309.5417619

0.14

N/A

2

hydroseismogram

11:33

7.3

16

10113.23419

50.04

87.81

344.7929523

0.1

N/A

0.75

hydroseismogram

19:50

8.3

27

8107.753858

41.81

143.91

309.7897511

0.4

N/A

2.5

hydroseismogram

23:41

6.4

10

1222.377862

26.56

-110.59

152.2225614

0.34

N/A

0.33

hydroseismogram

3:32

6.6

19

4150.097782

53.76

-164.61

313.2639763

0.045

N/A

0.25

hydroseismogram

2:06

7.6

24

2290.085002

18.84

-104.1

145.3748882

0.87

N/A

1.5

hydroseismogram

8:43

7.3

33

10076.58596

-10.49

160.77

257.3915265

0.35

N/A

1

hydroseismogram

0:53

6.3

10

1384.984011

44.28

-129.02

313.0006243

0.095

N/A

0.33

hydroseismogram

11/3/2002

Denali Fault, Alaska

22:12

7.9

5

3674.394072

63.52

-147.53

334.9091628

Disabled

N/A

unknown

coseismic offsetrecorder disabled

10/23/2002

Central Alaska

11:27

6.7

4

3688.15664

63.51

-147.91

334.69644

0.09

N/A

0.5

hydroseismogram

10/10/2002

Indonesia

10:50

7.6

10

11852.00503

-1.76

134.3

280.3694896

0.163

N/A

1.2

hydroseismogram

10/3/2002

Gulf of California

16:08

6.5

10

1636.711593

23.32

-108.53

150.7824137

0.11

N/A

0.5

hydroseismogram

9/8/2002

Bismarck Sea

18:44

7.6

13

11187.32691

-3.3

142.95

273.7529389

0.37

N/A

1.5

hydroseismogram

4/18/2002

Oaxaca Region, Mexico

5:02

6.8

24

2641.524854

16.99

-100.86

140.8229877

0.25

N/A

0.5

hydroseismogram

2/22/2002

Mexicali

19:32

5.7

7

465.1910471

32.32

-115.32

168.6671649

0.08

N/A

0.083


Mexicali

23:36

5.8

10

503.9588679

32.04

-114.91

165.0055326

0.205

N/A

0.25

hydroseismogram

9:26

7.8

10

11504.42317

35.94

90.54

337.927925

0.3

N/A

1.5

hydroseismogram

9:47

6.1

33

1802.759445

22.37

-106.97

147.5541287

0.095

N/A

0.25

hydroseismogram

20:30

8.4

33

7372.84295

-16.26

-73.64

134.7389115

0.85

N/A

3.5

hydroseismogram

18:54

6.8

51

1305.331148

47.15

-122.73

337.98778

0.11

N/A

0.25

hydroseismogram

20:11

6.2

20

2358.327121

53.92

-133.61

331.0153351

0.13

N/A

0.25

coseismic no offset

14:22

6.6

10

3715.018161

13.67

-88.94

125.8270088

0.18

N/A

0.5

hydroseismogram

12/8/2001 11/14/2001 11/13/2001 6/23/2001 2/28/2001

2/17/2001

2/13/2001

QinghaiXinjiang Border China Gulf of California Peru Coast Olympia, Washington Off Coast of Northern British Columbia El Salvador

36


EQ Time (hh:mm UTC)

EQ Magnitude

EQ Depth (km)


Epicenter Latitude

Epicenter Longitude






1/26/2001

India

3:16

7.7

16

13326.4705

23.41

70.23

353.097718

0.135

N/A

2

hydroseismogram

1/10/2001

Kodiak Island Region, Alaska

16:02

7

33

3558.218668

57.08

-153.21

321.9730916

0.235

N/A

0.5


12/6/2000

Turkmenistan

17:11

7.5

31

11516.63391

39.56

54.79

7.062241836

0.05

N/A

0.25

hydroseismogram

4:43

6.1

33

3262.882347

14.88

-93.94

131.4085643

0.05

N/A

0.16


21:01

7.6

33

10541.77112

-5.49

151.78

266.7153842

0.33

N/A

1

hydroseismogram

7:42

7.6

30

10406.81054

-5.23

153.1

266.146704

0.1

N/A

1

hydroseismogram

4:54

8

33

10408.52837

-3.98

152.16

267.7118816

0.405

N/A

2.25

hydroseismogram

1:32

6.6

43

3626.308478

57.37

-154.21

322.0602979

0.052

N/A

0.16


16:28

7.9

33

14749.48002

-4.72

102.09

302.7270379

0.08

N/A

2

hydroseismogram

11:13

6.2

10

1469.179692

44.51

-130.08

311.9616405

0.085

N/A

0.25


4:21

7.6

26

12736.88263

-1.11

123.57

288.0924201

0.05

N/A

0.75

hydroseismogram

11:00

7.6

126

9390.213954

22.34

143.73

293.7591054

0.15

N/A

1.25

hydroseismogram

15:19

5.9

7

898.008484

40.39

-125.28

302.113818

0.096

N/A

0.16

10:42

6.1

1

2836.567274

58.04

-136.87

334.4052793

0.052

N/A

0.16

23:12

7

66

3643.478108

57.41

-154.49

322.0177782

0.19

N/A

0.25


0:19

6.4

40

3642.397179

57.36

-154.51

321.9249815

0.03

N/A

0.16

hydroseismogram

13:21

7.5

33

9845.557835

-16.42

168.21

248.2723377

0.078

N/A

1.1

hydroseismogram

12/4/2000

11/17/2000

11/16/2000

11/16/2000

7/11/2000

6/4/2000 6/2/2000 5/4/2000

3/28/2000 3/16/2000 1/6/2000

12/6/1999

12/7/1999

11/26/1999

Off Coast of Tapachula, Mexico New Britian Region, Papua New Guinea New Ireland Region, Papua New Guinea New Ireland Region, Papua New Guinea Kodiak Island Region, Alaska Southern Sumatera, Indonesia Off the Coast of Oregon Minahassa Peninsula, Indonesia Volcano Islands, Japan Off Coast of Eureka, CA Petersburg, Alaska Kodiak Island Region, Alaska Kodiak Island Region, Alaska Coral Sea Region,

coseismic no offset coseismic no offset

37


Date

Date

EQ Time (hh:mm UTC)

EQ Magnitude

EQ Depth (km)


Epicenter Latitude

Epicenter Longitude






9:46

7.1

0

204.1635951

34.59

-116.27

179.4347281

Disabled

N/A

unknown


16:31

7.5

60

2963.427052

16.06

-96.93

134.7423227

0.9

N/A

0.8


0:01

7.6

17

10765.23711

40.75

29.86

25.14807398

0.112

N/A

1

hydroseismogram

16:06

5.7

7

128.056598

37.39

-117.08

327.06664

0.46

N/A

0.5


14:14

6.7

10

3590.139811

15.78

-88.33

122.3580389

0.202

N/A

0.5

hydroseismogram

20:42

7

70

2725.163193

18.39

-97.44

132.339119

0.3

N/A

0.5

hydroseismogram

13:22

5.6

5

255.9115842

37.53

-118.82

299.4198248

0.21

N/A

0.16

coseismic no offset

14:13

6.2

20

3513.452232

56.42

-152.94

320.9415558

0.045

N/A

0.3


10:47

6.9

33

5024.995814

51.59

-177.67

309.762418

0.053

N/A

0.5

hydroseismogram

12:25

6.9

56

6535.575199

52.06

159.52

314.3332034

0.035

N/A

0.16

hydroseismogram

21:47

7.3

90

9725.189471

-12.85

166.7

251.9769158

0.06

N/A

0.75

hydroseismogram

18:59

7.2

33

5536.204077

-0.59

-80.39

129.8399412

0.312

N/A

1

hydroseismogram

6:05

6.3

33

3584.907438

13.68

-90.75

128.2496228

0.032

N/A

0.25


23:30

7.5

33

10829.24226

22.31

125.31

304.851845

0.021

N/A

1

hydroseismogram

3:12

8.1

10

13757.70808

-62.88

149.53

213.1370286

0.191

N/A

1.75

hydroseismogram

3:02

6.4

33

3019.782456

15.88

-96.3

133.9083622

0.25

N/A

0.25

hydroseismogram

8:20

6.6

33

3476.106897

14.37

-91.47

128.404597

0.048

N/A

0.25

coseismic no offset

11:48

6

10

2840.44487

16.11

-98.85

138.1018267

0.04

N/A

0.25

coseismic no offset

11:26

7.8

33

6277.52051

54.84

162.04

316.8710989

0.311

N/A

1.75

hydroseismogram

Australia 10/16/1999 9/30/1999

8/17/1999 8/1/1999 7/11/1999 6/15/1999 5/15/1999

5/7/1999

3/20/1999 3/8/1999 2/6/1999 8/4/1998

5/10/1998 5/3/1998 3/25/1998

2/3/1998 1/10/1998 12/16/1997 12/5/1997

Hector Mine Earthquake Oaxaca Region, Mexico East of Istanbul, Turkey North of Beatty, NV Puerto Barrios, Guatemala Tehuacan, Mexico South of Mammoth Lakes, CA Kodiak Island Region, Alaska South Bering Sea Southwest Bering Sea Solomon Islands Bahia de Caraquez, Ecuador Off West Coast of Guatemala Philippine Sea Southern Ocean South of Australia Oaxaca Region, Mexico Rio Bravo, Guatemala Oaxaca Region, Mexico Southwest Bering Sea

38


Date

11/8/1997 10/4/1997 7/19/1997 7/9/1997 5/22/1997 5/1/1997 4/22/1997 4/21/1997

1/11/1997 11/12/1996 10/6/1996

7/15/1996 6/10/1996 6/10/1996 4/29/1996 3/9/1996 3/3/1996

3/3/1996

2/25/1996

2/21/1996 2/17/1996

Eastern China Off Coast of Eureka, CA Santiago, Mexico Carupano, Venezuela Flores, Mexico Off Coast of Manzanillo, Mexico Scarbarough, Tobago Coral Sea Region, Australia Calla de Campos, Mexico Ica, Peru Off Coast of Vancouver, BC, Canada Oaxaca Region, Mexico Aleutian Islands Aleutian Islands Solomon Islands Yucca Valley, CA Off Coast of Managua, Nicaragua Off Coast of Managua, Nicaragua Oaxaca Region, Mexico Off Coast of Chimbote, Peru Southern Indonesia

39


EQ Time (hh:mm UTC)

EQ Magnitude

EQ Depth (km)


Epicenter Latitude

Epicenter Longitude






10:02

7.5

33

11698.01995

35.07

87.32

340.1406897

0.099

N/A

1.25

hydroseismogram

10:57

5.6

7

938.8612627

41.05

-125.36

305.9552953

0.027

N/A

0.16

coseismic no offset

14:22

6.9

33

2858.467962

16.33

-98.22

136.6614262

0.189

N/A

1

hydroseismogram

19:24

7

19

6007.724056

10.6

-63.49

104.6617814

0.126

N/A

0.75

hydroseismogram

7:50

6.5

70

2441.003148

18.68

-101.6

140.0070835

0.041

N/A

0.2

coseismic no offset

11:37

6.9

33

2126.618883

18.99

-107.35

153.3462727

0.182

N/A

0.8

hydroseismogram

9:31

6.7

5

6199.666318

11.11

-60.89

102.2784402

0.02

N/A

0.33

hydroseismogram

12:02

7.7

33

9708.288561

-12.58

166.68

252.2011552

0.29

N/A

2.5

hydroseismogram

20:28

7.2

33

2419.988038

18.22

-102.76

143.1703864

0.145

N/A

0.6

hydroseismogram

16:59

7.7

33

7130.082624

-14.99

-75.68

135.6645202

0.084

N/A

1.5

hydroseismogram

20:13

6.2

10

1688.704528

49.05

-127.88

329.8311794

0.115

N/A

0.33

coseismic no offset

21:23

6.8

18

2578.567696

17.6

-100.96

140.2046635

0.178

N/A

0.5

hydroseismogram

15:24

7.3

26

4969.610693

51.48

-176.85

309.5423093

0.11

N/A

1

hydroseismogram

4:03

7.9

33

5022.558378

51.56

-177.63

309.7170083

0.34

N/A

2.5

hydroseismogram

14:40

7.2

44

10322.39304

-6.52

155

263.9846461

0.049

N/A

0.66

hydroseismogram

23:20

3.3

4

231.4037856

34.35

-116.47

184.0502951

0.063

N/A

0.16


16:37

6.7

33

4018.880758

11.9

-86.77

125.1633177

0.09

N/A

0.4

hydroseismogram

14:55

6.6

33

4032.116071

11.66

-86.86

125.5464012

0.037

N/A

0.3

hydroseismogram

3:08

7.1

21

2899.533176

15.98

-98.07

136.8512362

0.45

N/A

1.2

hydroseismogram

12:51

7.5

10

6394.659398

-9.59

-79.59

135.6803008

0.082

N/A

0.75

hydroseismogram

5:59

8.2

33

11559.48345

-0.89

136.95

279.4109438

0.496

N/A

2.5

hydroseismogram

Date

1/7/1996

1/1/1996

12/11/1995 12/3/1995 10/9/1995 10/3/1995 9/14/1995 8/28/1995 7/30/1995 6/30/1995 6/14/1995

5/31/1995 5/27/1995 5/16/1995

4/7/1995 2/19/1995 2/5/1995 1/19/1995 12/28/1994 10/27/1994 10/4/1994

Ridgecrest, CA Off Coast of Dampal Selatan, Indonesia Off Coast of Manzanillo, Mexico Kuril Islands Manzanillo, Mexico Sucua, Ecuador Oaxaca Region, Mexico Gulf of California Northern Chile Gulf of California Off Coast of Managua, Nicaragua Off Coast of Manzanillo, Mexico Sea of Okhotsk Coral Sea Region, Australia Coral Sea Region, Australia Off Coast of Eureka, CA North of New Zealand Bogota, Columbia Northern Japan Off Coast of Oregon Kuril Islands

EQ Time (hh:mm UTC)

EQ Magnitude

EQ Depth (km)


Epicenter Latitude

Epicenter Longitude






14:32

5.4

5

142.1513842

35.77

-117.65

239.5285892

0.065

N/A

0.1


8:05

7.9

24

12908.39484

0.73

119.93

292.2694894

0.06

N/A

1.75

hydroseismogram

14:09

6.4

20

2214.393704

18.93

-105.47

148.5727026

0.07

N/A

0.45

hydroseismogram

18:01

7.9

33

7569.666109

44.66

149.3

310.1396871

0.465

N/A

2.75

hydroseismogram

15:35

8

33

2263.078484

19.06

-104.21

145.3310183

0.99

N/A

3

hydroseismogram

1:51

7

24

5897.761558

-2.75

-77.88

129.0438529

0.14

N/A

0.5

hydroseismogram

14:04

7.4

23

2794.237265

16.78

-98.6

136.7518727

0.99

N/A

2.7

coseismic + hydroseismogram (up)

10:46

6.6

12

1282.784849

26.09

-110.28

151.9431958

0.151

N/A

0.55

hydroseismogram

5:11

8

45

8213.202597

-23.34

-70.29

136.5604211

0.1

N/A

2.25

hydroseismogram

11:58

6.2

10

1427.37644

24.69

-110.23

154.4135906

0.113

N/A

0.4

hydroseismogram

11:11

6.6

25

3884.737428

12.13

-88.36

126.8977524

0.096

N/A

0.45

hydroseismogram

16:08

6.4

33

2126.905502

18.96

-107.42

153.5679435

0.062

N/A

0.25

coseismic no offset

13:03

7.2

11

7525.981549

52.63

142.83

319.885623

0.267

N/A

1.8

hydroseismogram

20:12

7.7

20

10170.54528

-23.01

169.9

242.1550398

0.147

N/A

2.5

hydroseismogram

22:06

7.4

21

8302.075874

-15.2

-173.53

237.2964524

0.498

N/A

2.5

hydroseismogram

4:03

6.4

10

926.1436981

40.56

-125.54

302.5568491

1

N/A

1.1

hydroseismogram

22:51

7.1

21

10609.52619

-37.76

178.75

226.0111933

0.08

N/A

1.25

hydroseismogram

15:05

6.5

17

5625.170365

5.05

-72.92

117.6988315

0.02

N/A

0.33

hydroseismogram

12:19

7.8

26

8221.747364

40.53

143.42

308.9023042

0.081

N/A

2.1

hydroseismogram

17:45

6.3

20

1231.536823

43.51

-127.43

313.1725196

0.051

N/A

0.16


13:22

8.3

14

7755.877404

43.77

147.32

310.1004626

0.5

N/A

4

hydroseismogram

40


Date

9/1/1994 6/9/1994 3/20/1994 3/14/1994 2/3/1994 1/29/1994 12/4/1993 11/13/1993 11/13/1993

10/24/1993

9/30/1993 9/21/1993 9/21/1993 9/19/1993

9/10/1993

9/3/1993

8/8/1993

8/3/1993

7/12/1993 6/8/1993

Off Coast of Northern California Rio Beni, Bolivia Los Angeles, CA Santa Domingo, Mexico Afton, WY Los Angeles, CA Klamath Falls, OR Kuril Islands Oaxaca Region, Mexico Oaxaca Region, Mexico Oaxaca Region, Mexico Klamath Falls, OR Klamath Falls, OR Oaxaca Region, Mexico Oaxaca Region, Mexico Oaxaca Region, Mexico Northern Mariana Islands Off Coast of British Columbia, Canada Northern Japan Kuril Islands

41


EQ Time (hh:mm UTC)

EQ Magnitude

EQ Depth (km)


Epicenter Latitude

Epicenter Longitude






15:15

7.1

10

928.9525495

40.4

-125.68

301.2435015

Disabled

N/A

unknown


0:33

8.2

631

7571.155039

-13.84

-67.55

128.1168128

0.08

N/A

1.5

hydroseismogram

21:20

5.3

13

314.0899155

34.23

-118.47

219.6136444

0.048

N/A

0.1

coseismic no offset

20:51

6.9

164

3272.598649

15.99

-92.43

127.6821036

0.04

N/A

0.45

hydroseismogram

9:05

5.8

7

838.1642827

42.76

-110.98

31.20874343

0.05

N/A

0.2

11:20

5.4

1

313.667053

34.31

-118.58

222.07151

0.075

N/A

0.1

22:15

5.4

8

817.0323379

42.3

-122.01

324.816129

0.043

N/A

0.16

1:18

7

34

6596.375462

51.93

158.65

314.4146211

0.061

N/A

0.5

hydroseismogram

0:16

6

19

2836.525249

16.29

-98.64

137.4799215

0.041

N/A

0.25

hydroseismogram

7:52

6.7

20

2789.629742

16.75

-98.72

137.0141375

0.33

N/A

0.52

hydroseismogram

18:27

6.5

19

3165.100094

15.42

-94.7

131.9001295

0.322

N/A

0.75

hydroseismogram

5:45

6

5

823.7791441

42.36

-122.04

324.9738788

0.297

N/A

0.25

hydroseismogram

3:28

6

10

817.8990111

42.31

-122.01

324.8650905

0.175

N/A

0.25

hydroseismogram

14:10

6.4

18

3349.713183

14.36

-93.32

131.119721

0.256

N/A

0.45

hydroseismogram

19:12

7.2

34

3365.069888

14.72

-92.64

129.6534382

0.95

N/A

1.5

hydroseismogram

12:35

6.8

26

3377.374714

14.52

-92.71

130.0096972

0.692

N/A

0.55

hydroseismogram

8:34

7.8

59

9931.349403

12.98

144.8

285.6847303

0.23

N/A

2.5

hydroseismogram

7:19

6

10

2004.66652

51.19

-130.8

329.5149046

0.102

N/A

0.25

coseismic no offset

13:17

7.7

16

8357.924865

42.85

139.2

312.7574507

0.145

N/A

2.5

hydroseismogram

13:03

7.5

70

6678.216683

51.22

157.83

313.8687552

0.124

N/A

1.75

hydroseismogram

coseismic no offset coseismic no offset coseismic no offset

Date

5/25/1993 5/19/1993 5/18/1993 5/17/1993

EQ Time (hh:mm UTC)

EQ Magnitude

EQ Depth (km)


Epicenter Latitude

Epicenter Longitude






Aleutian Islands Death Valley, CA Death Valley, CA

23:16

6.2

36

3914.3776

55.02

-160.51

316.088983

0.054

N/A

0.16

coseismic no offset coseismic no offset coseismic no offset coseismic + hydroseismogram (down)

14:13

5.2

0

153.7134036

37.14

-117.77

301.5381261

0.06

N/A

0.16

23:48

5.2

3

150.1603163

37.06

-117.78

298.4437698

0.055

N/A

0.16

Death Valley, CA

23:20

6.1

6

156.1951297

37.17

-117.78

302.4252353

Disabled

N/A

unknown

11:59

6.9

32

3915.400278

55.18

-160.46

316.3704508

0.12

N/A

0.5

hydroseismogram

5:29

7.8

27

13395.17407

-8.48

121.9

282.7954771

0.065

N/A

1.8

hydroseismogram

Aleutian Islands Flores Sea, Indonesia Big Bear City, CA Northern Columbia St. George, UT Off Coast of Nicaragua Gulf of Alaska

16:00

5.6

1

238.4100044

34.34

-116.9

193.5443159

0.058

N/A

0.1


15:11

7.2

10

5157.560622

7.07

-76.86

119.4641463

0.203

N/A

1

hydroseismogram

10:26

5.6

15

262.0063228

37.09

-113.47

72.79181847

0.17

N/A

0.16

hydroseismogram

0:16

7.7

44

3990.651649

11.74

-87.34

126.0509635

0.34

N/A

1

hydroseismogram

18:19

6.9

13

3057.382718

57.59

-142.85

328.7314748

0.088

N/A

0.65

hydroseismogram

7/5/1992

Landers, CA

21:18

5.4

0

205.2814793

34.58

-116.32

180.7156137

0.109

N/A

0.16

6/28/1992

Landers Earthquake

11:57

7.3

1

247.8838086

34.2

-116.44

183.148449

Disabled

N/A

unknown

4/23/1992

Desert Hot Springs, CA

4:50

6.1

12

274.2184889

33.96

-116.32

180.5397442

Disabled

N/A

unknown

coseismic offset recorder disabled

14:09

6

10

1913.865319

50.59

-129.92

329.63428

0.078

N/A

0.16


13:54

6.7

19

1932.642301

50.72

-130.09

329.6367284

0.38

N/A

1.1

hydroseismogram

5/13/1993 12/12/1992 11/27/1992 10/18/1992 9/2/1992 9/2/1992 8/7/1992

4/6/1992

4/6/1992

Off Coast of British Columbia, Canada Off Coast of British Columbia, Canada

coseismic no offset coseismic offset recorder disabled

3/8/1992

Ferndale, CA

3:43

5.5

13

815.3707635

40.23

-124.29

303.6659801

0.07

N/A

0.2


1/2/1992

Off Coast of British Columbia, Canada

16:40

6.1

10

1726.097648

48.74

-129.23

326.5151902

0.067

N/A

0.25


12/22/1991

Kuril Islands

8:43

7.6

24

7404.525212

45.53

151.02

310.3122776

0.086

N/A

1..75

hydroseismogram

15:58

7.2

23

10187.64643

-9.09

158.44

259.8888427

0.13

N/A

1.25

hydroseismogram

22:17

7

13

987.5116551

41.82

-125.4

310.1693501

Disabled

N/A

unknown

coseismic offset recorder disabled

10/14/1991 8/17/1991

Solomon Islands Off Coast of Crescent

42


Date

43


EQ Time (hh:mm UTC)

EQ Magnitude

EQ Depth (km)


Epicenter Latitude

Epicenter Longitude






2:50

6.8

11

1026.694977

42.18

-125.64

311.3954107

>1

N/A

1.45


14:43

5.6

11

286.3565175

34.26

-118

213.2476815

0.2

N/A

0.2

hydroseismogram

0:30

6.1

10

1575.756285

23.92

-108.55

149.801683

0.128

N/A

0.6

hydroseismogram

City, CA

7/13/1991

6/28/1991 6/21/1991

Off Coast of Crescent City, CA Angeles National Forest, CA Gulf of California

4/22/1991

Costa Rica

21:56

7.6

10

4473.425165

9.69

-83.07

123.2595947

0.35

N/A

1.5

hydroseismogram

1/1/1991

Manzanillo, Mexico

0:06

6.3

35

2283.536722

18.07

-105.85

150.582882

0.108

N/A

0.4


20:14

7.1

24

5824.654378

53.45

169.87

313.7706888

0.047

N/A

1.2

hydroseismogram

17:30

5.4

0

93.3034992

37.25

-116.49

349.1734215

0.041

N/A

0.1

coseismic no offset

13:53

5.4

5

483.5898058

36.92

-121.68

278.1262982

0.056

N/A

1.5

hydroseismogram

21:12

7.4

11

9546.945925

15.12

147.6

285.7584285

0.228

N/A

2.5

hydroseismogram

22:57

6.8

52

4091.976906

11.43

-86.3

125.1209724

0.125

N/A

0.5

hydroseismogram

13:22

7.3

22

4325.706571

9.92

-84.81

124.9994741

0.27

N/A

1.3

hydroseismogram

15:52

6.1

10

1456.093358

24.9

-109.04

149.6441018

0.095

N/A

0.35

hydroseismogram

12:16

7.6

33

9695.101682

-22.12

175.16

239.6844159

0.14

N/A

1.6

hydroseismogram

23:43

5.7

5

284.501967

34.14

-117.7

207.1019137

0.218

N/A

0.3

coseismic no offset

18:25

7.4

28

8311.563381

39.84

142.76

308.612463

0.104

N/A

2.2

hydroseismogram

21:04

7.1

24

9974.600328

-11.02

162.35

256.0316808

0.116

N/A

0.6

hydroseismogram

11/6/1990 10/12/1990 4/18/1990 4/5/1990 4/3/1990 3/25/1990 3/16/1990 3/3/1990 2/28/1990 11/1/1989 10/28/1989

Western Aluetian Islands Yucca Flat, NV Watsonville, CA Marianas Trench Managua, Nicaragua Puntarenas, Costa Rica Gulf of California Fiji Islands Claremont, CA Northern Japan Solomon Sea

Table 2.A1. (continued) Catalog of water level responses to earthquakes at Devil's Hole.

CHAPTER 3 SHARP INCREASE IN CENTRAL OKLAHOMA SEISMICITY 2009-2014 INDUCED BY MASSIVE WASTEWATER INJECTION

Abstract Unconventional oil and gas production provides a rapidly growing energy source; however high-producing states in the United States, such as Oklahoma, face sharply rising numbers of earthquakes. Subsurface pressure data required to unequivocally link earthquakes to injection are rarely accessible. Here we use seismicity and hydrogeologic models to show that distant fluid migration from high-rate disposal wells in Oklahoma is likely responsible for the largest swarm. Earthquake hypocenters occur within disposal formations and upper-basement, between 2-5 km depth. The modeled fluid pressure perturbation propagates throughout the same depth range and tracks earthquakes to distances of 35 km, with a triggering threshold of ~0.07 MPa. Although thousands of disposal wells may operate aseismically, four of the highest-rate wells likely induced 20% of 2008-2013 central US seismicity.

A portion of this chapter has been previously published: Keranen, K.M., Weingarten, M., Abers, G.A., Bekins, B.A., and S. Ge (2014), Sharp increase in central Oklahoma seismicity since 2008 induced by massive wastewater injection, Science, 345(6195), pg. 448-451, doi:10.1126/science.1255802. 44

3.1 Introduction Seismicity in the United States midcontinent surged beginning in 2008 (1), predominantly within regions of active unconventional hydrocarbon production (2-6). In Arkansas, Texas, Ohio, and near Prague, Oklahoma, recent earthquakes have been linked to wastewater injection (2-7) although alternative interpretations have been proposed (1, 8). Conclusively distinguishing human-induced earthquakes based solely on seismological data remains challenging. Seismic swarms within Oklahoma dominate the recent seismicity in the central and eastern United States (9), contributing 45% of M3 and larger earthquakes between 2008-2013 (10). No other state contributed more than 11 %. A single swarm, beginning in 2008 near Jones, Oklahoma, accounts for 20% of seismicity in this region (10). East of Jones, the damaging 2011 Mw 5.7 earthquake near Prague, Oklahoma was likely induced by wastewater injection (2,8, 1112), the highest magnitude to date. These earthquakes are part of a 40-fold increase in seismicity within Oklahoma during 2008-2013 as compared to 1976-2007 (Fig. 3.1A) (10). Wastewater disposal volumes have also increased rapidly, nearly doubling in central Oklahoma between 2004-2008. Many studies of seismicity near disposal wells rely upon statistical relationships between the relative timing of seismicity, disposal well location, and injected water volume to evaluate a possible causal relationship (3-7,13). 3.2 Methods & Data 3.2.1 Earthquake Swarm Here we focused on the Jones swarm and compared modeled pore pressure from hydrogeologic models to the best-constrained earthquake hypocenters (14). Using data from

45

Figure 3.1. Earthquakes are magnitude > 1 from the NEIC catalog (10). Black lines are faults (15-17). Small and large dashed gray boxes outline the areas used for analysis of the Jones swarm and of central OK, respectively, in Inset (B). OKC: Oklahoma City. Inset (A): Comparison of M3+ earthquake rate in Oklahoma to California. 2014 earthquakes are through the first four months. Inset (B): Expanding area of the Jones and the broader central Oklahoma swarms. Regions were divided into 5x5 km grid cells and any cell with an earthquake was considered part of the swarm. Swarm area per year is inclusive of all prior years.

46

local U. S. Geological Survey NetQuake accelerometers, the Earthscope Transportable Array and a small local seismic network, we generated a catalog of well-located earthquakes between 20102013. Event-station distances were predominantly less than 10 km and all earthquakes were recorded on at least one seismometer within 20 km of the initial hypocenter. To study pore pressure changes at earthquake hypocenters and the apparent migration in seismicity, we developed a three-dimensional hydrogeologic model of pore pressure diffusion from injection wells. The Jones swarm began within 20 km of the highest-rate wastewater disposal wells in Oklahoma, between two regions of fluid injection (Fig. 3.2A). Earthquakes in our catalog primarily nucleated within either the Arbuckle Group or within the upper 2 km of basement, with 22-33% above basement (Fig. 3.2B). Well-constrained earthquake hypocenters from March to October 2010 migrated northeast from the initial swarm centroid near Jones at 0.1-0.15 km/day (Fig. 3.2C-D), followed by a broad spread in seismicity. Earthquake hypocenters are not diffusely distributed; instead, relocated aftershock sequences of individual earthquakes (18) illuminate narrow faults parallel to one plane of calculated focal mechanisms (19) (Fig. 3.2A, insets). An earthquake on August 2, 2010, ruptured a portion of a 7-km-long mapped fault; if the entire fault had ruptured earthquake scaling laws suggest a maximum magnitude of ~M6.0 (20). Earthquakes later in 2010 ruptured an unmapped east-southeast to west-northwest trending fault, at an oblique angle to the overall northeast-southwest migration direction of the swarm. Although the swarm of seismicity migrates to the northeast parallel to structural dip, the individual faults, as evidenced by earthquake lineations, are not preferentially oriented in this direction.

47

Figure 3.2. (A) Jones earthquake catalog March 2010-March 2013 using local stations. Squares are injection wells over 400,000 barrels/month (21-22), triangles are high-water-production wells. Background color and contours represent depth to the top of the Hunton Group (21). The Hunton Group is higher in section than the Arbuckle Group but has more data on formation depth. (B) Earthquake depth histogram; earthquakes are dominantly in sediment and upper basement. (C) Distance of each March-October 2010 Jones earthquake to the SE OKC disposal wells. The dense region of the swarm increases in distance between days 150 and 250 in 2010. (D) Map view of Jones earthquakes during March-October of 2010, colored by time. Semicircles are equidistant lines from SE OKC disposal wells. Faults at greater distance from the wells become active at later times. Details of two of these fault planes are shown in insets A and B of Fig. 3.2A and are discussed in the text. 48

3.2.2 Injection Operations in the Jones, Oklahoma region The four highest rate wells are southwest of Jones in southeast Oklahoma City (SE OKC) and dispose of over 4 million barrels/month (21) (Fig. 3.3). The target injection depth is 2.2-3.5 km into the Cambrian-Ordovician Arbuckle Group (Fig. 3.4), a dolomitized carbonate; one disposal well reaches Precambrian basement. The large disposal wells are part of dewatering plays. Dewatering production wells produce substantial wastewater volumes with initially up to 200 times greater water per barrel of oil than conventional production wells, and up to 1000:1 water to oil ratio (23-24). The rate of wastewater disposal in central Oklahoma has gradually increased since the mid-1990s (Fig. S5), but disposal rates jumped after 2004 as high-rate injection wells began operating, including the first of the SE OKC wells in 2005 (Fig. 3.5) (21). Seismic moment release escalated in the Jones swarm in 2009, concurrent with the initial application of positive wellhead pressure at the SE OKC wells (Fig. 3.3B).

3.3 Hydrogeologic Model To study pore pressure changes at earthquake hypocenters and the apparent diffusion in seismicity, we developed a three-dimensional hydrogeologic model of pore pressure diffusion from injection wells. Our model aims captures the primary geological features of the region without attempting to include all geological detail. Much of the hydrogeologic detail of the region is unknown, such as the three-dimensional permeability structure. However, we can put constraints on the ranges of the relevant hydrogeologic parameters based upon the region's lithology as well as the combination of the injection rates and pressures of the simulated wells in the region.

49

Figure 3.3. (A) Sum and individual monthly injection volumes and (B) wellhead pressure and cumulative, summed injected volume (21). The Deep Throat, Flower Power, and Sweetheart wells are co-located on one surface pad; the Chambers well is ~3.5 km away. Gray shading denotes injection rates for notable past cases of induced seismicity for reference (Table 3.1). Cumulative seismic moment in (B) is calculated from M3+ earthquakes from 2005 to January 2014 (10) for earthquakes within the box outlining the Jones swarm in Fig.3.1.

50

Figure 3.4. Cross section across the Nemaha fault. Cross section is modified from (17, Fig. 6). The fault crosses approximately north-south through the Oklahoma City metro area (Inset map). Metro area from the University of Oklahoma Center for Spatial Analysis. In the SE OKC wells, fluid is extracted from the Arbuckle Group on the west side of the fault and injected within the Arbuckle Group and basement on the east, downthrown side of the fault. The wavy red line marks the major pre-Pennsylvanian unconformity. Wells noted were used in construction of the cross section in (17). Production and disposal wells are within hundreds of meters of the fault.

51

Figure 3.5. Injection rates at wells included in hydrogeologic model of the Jones swarm. All wells within ~50 km of the center of the Jones swarm that both reached the target depth and injected more than 1,000,000 barrels in any year of operation from 1995-2012 were modeled. The entire lifetime of the well was modeled even if the well only operated for one year above 1,000,000 barrels/year.

52

3.3.1 Numerical Code We use MODFLOW to simulate 3D pore pressure diffusion from injection wells. MODFLOW (25) is a modular finite-difference code developed at the USGS, and solves the groundwater flow equation in three dimensions for a fluid of constant density and dynamic viscosity in a heterogeneous and anisotropic aquifer with sources and sinks (26):

+

(3.1)

where

,

,

+

=

−

( ) ( −

) ( −

) ( − )

are the principal components of the hydraulic conductivity tensor [L/T],

is the specific storage coefficient [L-1], [L3/T], h is the hydraulic head [L] and

is the volumetric injection rate at the i-th point source is the kronecker delta. Hydraulic head change can be

converted to pore pressure change through a simple conversion using the specific weight of water and the known depth of the reservoir.

3.3.2 Model Domain, Boundary Conditions & Material Properties Our model domain is 200 km by 130 km by 6 km deep, including, from top to bottom, the 1-km-thick Arbuckle Group, the 30-m-thick Reagan Sandstone, and the granitic basement (Fig. 3.6). We assumed that all the formations are flat because there is little structural dip across the model domain. The model is bounded on the west by the Nemaha fault, approximated as a no-flow boundary, as the fault juxtaposes the Arbuckle Group on the east with low-permeability granitic basement to the west. The distant north, south, and east boundaries of the model were set to no-flow, as is the top boundary, to represent the overlying low-permeability Oil Creek Formation (predominantly shale). The bottom boundary, within basement at 8 km deep, is also 53

Figure 3.6. Schematic diagram of the model domain with boundary conditions and model domain. The Nemaha Fault no-flow boundary flow does not extend directly north-south as diagrammed here, but instead follows the trend of the mapped fault.

54

assumed to be no-flow. All no-flow boundaries, except the Nemaha fault and the Oil Creek Formation, are sufficiently far from the area of interest such that they will have little influence on the modeled pore pressure distribution. Reported permeabilities in the repeated carbonate and brecciated facies within the Arbuckle Group reach values up to 1.5x10-12 m2 (27). If specific storage is assumed to be 1.0 x 10-5 m-1, this permeability corresponds to a diffusivity range of 1.5 to 4.5 m2/s for temperatures ranging from 20-100°C. The base of the Arbuckle Group has reported temperatures in the vicinity of 100°C. Hydraulic diffusivity is defined as the relationship of the hydraulic conductivity and specific storage of the medium:

(3.2)

=

where

is the hydraulic diffusivity tensor [L2/T]. Higher permeabilities allow faster fluid

!"

movement and, in turn, faster dissipation of fluid pressure. A higher storage capacity allows more fluid to be stored or released under a given pressure change, thus, slows dissipation of fluid pressure. We simulated the fluid-pressure changes of all injection wells within 50 km of the center of the Jones swarm that met two criteria (Fig. 3.5): (1) Injected with a yearly injection volume over 1 million barrels in any year since 1995. (2) Injected at the Simpson-Arbuckle-Reagan layer depth. The models covered the time period of 1995-2012 for models including all 73 wells, and the time period of 2005-2012 for models of only the four southeast Oklahoma City (SE OKC) wells.

55

3.3.3 Sensitivity Analysis Our initial sensitivity analysis started with a wide range of Arbuckle Group hydrogeologic properties to test whether the reported the high-end permeabilities could reproduce wellhead pressures from the four high-rate wells near the Nemaha Fault (Fig. 3.6). The reported monthly wellhead pressures collected by the Oklahoma Corporation Commission did not contain enough temporal resolution to allow a detailed temporal calibration of the hydrogeologic model. Therefore, we used the reported maximum wellhead pressure at the four high-rate wells near the Nemaha fault (Fig. 3.3B) as a constraint. The reported maximum wellhead pressure at the four high-rate wells reached ~4 MPa by the end of the 2012. Figure 3.7 shows the hydrostratigraphic layering of the sensitivity analyses. The initial scenarios tested Arbuckle Group diffusivities over three orders of magnitude from Dxx = Dyy = 0.1 m2/s to Dxx = Dyy = 10 m2/s. In each simulation, the Arbuckle Group’s vertical diffusivity (Dzz) was set one order of magnitude below the horizontal diffusivity to account for layering. The sensitivity of basement diffusivity was tested over two orders of magnitude, ranging from Dxx = Dyy = Dzz = 0.1 m2/s to Dxx = Dyy = Dzz = 0.01 m2/s. In all sensitivity analyses, the specific storage coefficient was held constant while the hydraulic conductivity was varied. The specific storage coefficient used for the Arbuckle Group, Ss = 10-5 m-1, is typical of dolomitic rock. The specific storage coefficient for the basement formation, Ss = 10-6 m-1, is typical of lesscompressible crystalline rock. In the low Arbuckle Group diffusivity case (Dxx = Dyy = 0.1 m2/s), the modeled maximum wellhead pressure at the four high-rate wells all exceed 10 MPa by the end of 2012. In the high Arbuckle Group diffusivity case (Dxx = Dyy = 10 m2/s), the modeled maximum wellhead pressure

56

Figure 3.7. Generalized hydrostratigraphic setup of the sensitivity analysis. Flow model parameters such as K, hydraulic conductivity, and Ss, specific storage, are for the narrowed range of Arbuckle Group diffusivities from Dxx = Dyy = 1 m2/s to Dxx = Dyy = 4 m2/s.

57

at the four high-rate wells does not reach 1 MPa. The basement diffusivity did not have a large effect on the maximum wellhead pressure at the high-rate wells, but did affect the radial distance to which pore pressure propagated in the Arbuckle Group. The higher basement diffusivity case decreased the radial distance to which pore pressure propagates in the Arbuckle Group, allowing more pressure to be accommodated by the basement formation. From these initial runs, it was clear the Arbuckle Group's diffusivity must be in the range of Dxx = Dyy = 1 m2/s. Simulating another set of hydrogeologic scenarios, we narrowed the range of the horizontal hydraulic diffusivity of the Arbuckle Group to between Dxx = Dyy = 1 m2/s to Dxx = Dyy = 4 m2/s (Fig. 3.8). The model was also run using only the four high-rate SE OKC wells to test the relative effect of these four wells as compared to the 69 wells to the northwest of the swarm (Fig. 3.9). We calculated the pore pressure change at each earthquake hypocentral location in space and time to estimate the critical pressure needed to induce failure in the Jones area. Each model result demonstrates how the material properties of the model effect the timing and magnitude of pore pressure at the Jones Swarm. Of all three models, Arbuckle diffusivities of Dxx = Dyy = 1 m2/s produce the highest pore pressure changes, but over the smallest distances from the high-rate wellbores. Conversely, Arbuckle diffusivities of Dxx = Dyy = 4 m2/s produce the smallest pore pressure changes, but over a distance of more than 60 km from the high-rate wells (Fig. 3.8; Fig. 3.9). All of the models predict fluid-pressure changes in vicinity of the Jones swarm near the end of 2009. From 2009 through 2012, modeled perturbations grow and track the expansion of the Jones swarm to its full extent in December 2012. While the pore pressure distribution in each model is unique, the calculated critical pore pressure change needed for failure is quite similar. The critical pore pressure varied for 0.05 0.07 MPa over when the Arbuckle Group diffusivity ranged from 1 to 4 m2/s.

58

Figure 3.8. Hydrogeologic model results using all wells in the Jones Area. Scenarios A-C use Arbuckle Group diffusivities of 1, 2 and 4 m2/s. Histograms of pore pressure change at each earthquake hypocentral location plot the critical pore pressure for failure in each model.

59

Figure 3.9. Hydrogeologic model results using on the four high-rate SE OKC wells. Scenarios A-C use Arbuckle Group diffusivities of 1, 2 and 4 m2/s. Histograms of pore pressure change at each earthquake hypocentral location plot the critical pore pressure for failure in each model.

60

3.4 Modeling Results Using an Arbuckle Group horizontal diffusivity of 2.0 m2/s provided the best fit to the general trend of reported maximum wellhead pressure and matches locally reported values of Arbuckle Group permeability, e.g. in the Mary Unsell #7 well (21). We also used a vertical diffusivity within the Arbuckle Group one order of magnitude smaller to account for layering. Basement is assumed to be isotropic with a depth-decreasing hydraulic diffusivity from 0.01 m2/s at the top to 0.0001 m2/s at the base (28). The final model predicts a region of high fluid-pressure perturbation spreading radially eastward from the SE OKC wells, and a lesser perturbation around the lower-rate wells to the northeast (Fig. 3.10). The high pore pressure increase occurs within the Arbuckle Group and in the upper 1-2 km of the basement in our model; nearly all earthquakes occur within this same depth range (Fig. 3.2B). The migrating front of the Jones earthquake swarm corresponds closely to the expanding pressure perturbation away from the SE OKC wells, which reaches 25 km from the wells by December 2009 and to ~ 35 km by December 2012. The pore pressure change modeled at each hypocenter indicates a critical threshold of ~ 0.07 MPa, above which earthquakes are triggered. This threshold is compatible with prior observations that static stress changes of as little as ~0.01-0.1 MPa are sufficient to trigger earthquakes when faults are near failure in the ambient stress field (29-31). Our results indicate that over 85% of the pore pressure perturbation is contributed by the four high-rate SE OKC wells (Fig 3.11). The 69 wells to the northeast contribute up to ~15% additional pore pressure change at the center of the Jones swarm by the end of 2012, and may contribute to the triggering of earthquakes particularly outside the region affected by the SE OKC wells. The dominance of the SE OKC wells is attributable to their high rate and their

61

Figure 3.10. (A) Modeled pressure perturbation in December 2009 and (B) in December 2012 using a hydraulic diffusivity of 2 m2/s. The model includes the four high-rate SE OKC wells and 85 wells northeast of the Jones swarm near the West Carney field. The modeled pressure perturbation is dominated by fluid injected at the high-rate SE OKC wells. Earthquakes are plotted from 2008-2009 (A) and 2008-2012 (B) (10). (C) Vertical cross-section through model results. Pore pressure rises in the Arbuckle Group and uppermost basement. (D) Pore pressure increase at the hypocenter of each earthquake in our local catalog. A pore pressure increase of ~0.07 MPa is the modeled triggering threshold. Modeled pore pressure rises throughout much of the swarm area for hydraulic diffusivity between 1 m2/s and 4 m2/s.

62

Figure 3.11. A comparison of modeled pore pressure changes from the four high-rate SE OKC wells and from all wells within the model. The comparison shows ~85% of the pore pressure changes come from the four high-rate wells. Pressure changes are plotted through time at the center of the Jones swarm, denoted by the black star in the plan view model plot.

63

proximity to one another. These wells include the largest well in the state and three wells 3.5 km away all on a single surface pad with a combined monthly volume of ~3 million barrels/month. The only other Oklahoma wells of nearly equivalent size, in northern Oklahoma, are on the boundary of a second rapidly growing seismic swarm (Fig. 3.1). The summed rate of this well cluster near SE OKC is higher than previous cases of reported induced seismicity (Table 3.1), including several times higher than the high-rate disposal wells linked to earthquakes near Dallas-Fort Worth, Texas and Cleburne, Texas (5-7). One key aspect of the modeling result is the relative pore pressure contributions of the four high-rate wells in SE OKC and from the 69 wells to the northeast. This distinction highlights the importance of injection rate in driving reservoir pressure perturbation. While the SE OKC wells only began injection in 2005, many of the wells to the northeast began injection as early as 1995 (Fig. 3.12). By 2005, nearly half of all of the volume injected in the Jones swarm area (6 billion barrels) had been injected at relatively low rates over a 25 km2 area. Before the onset of the high-rate wells, the pressure perturbation from the northeast injection wells was less than 0.01 MPa at the Jones swarm center, because it occurred at low rates over a large area. Once the SE OKC wells became operational the reservoir pressures drastically increased (Fig. 3.11). This simple modeling comparison shows how large volumes can be injected in a manner which does not raise reservoir pressure above the critical threshold.

3.5 Discussion We view the expanding Jones earthquake swarm as a response to regionally increased pore pressure from fluids injected at the SE OKC wells. As the pressure perturbation expanded

64

IS Case Study Location (reference number)

Disposal volume (millions of barrels per month)

Dallas-Fort Worth, TX (7) Cleburne, TX (6) Rocky Mountain Arsenal, CO (32) Paradox Valley, UT (33) Guy, Arkansas (4) Youngstown, OH (3) Ashtabula, OH (34)

0.3 0.6 0.214 0.356 0.394 0.077 0.314

Table 3.1. Injection rates at other wastewater disposal sites which have documented induced seismicity. Injection rates is previous case studies of induced seismicity were significantly lower than the rates present at the four high-rate SE OKC wells.

65

Figure 3.12. Cumulative injected volume (top) and cumulative monthly injection rate (bottom) for the four SE OKC wells (red) and the 69 low-rate wells to the northeast of the Jones swarm (blue). By 2005, nearly half of all of the volume injected in the Jones swarm area (6 billion barrels) had been injected at relatively low rates over a large area. 66

and encountered faults at various orientations, critically stressed, optimally oriented faults are expected to rupture first (35). Additional faults at near optimal orientations may rupture following further pressure increase (Fig. 3.10). As fluid pressure continues to propagate away from the wells and disturbs a larger and larger volume, the probability increases that fluid pressure will encounter a larger fault and induce a larger magnitude earthquake. The absence of earthquakes in regions above the critical pressure threshold may result from either a lack of faults or lack of well-oriented, critically-stressed faults. Alternatively, fluid flow may preferentially migrate along bedding structure. Though seven earthquakes were recorded in 2006-2009 near the base of the SE OKC wellbores (10), the main swarm began ~15 km to the northeast, despite the high modeled pressure perturbation near the wells. Earthquakes in 2009 primarily occurred, within location uncertainty, near injection wells or on the nearest known faults to the northeast of the wells. Focal mechanisms near the swarm onset indicate fault planes at orientations favorable to failure (Fig. 3.2B) (19). Faults subparallel to the NNW-SSE-trending Nemaha fault would not be welloriented for failure in the regional ~N70E stress regime (36) and would require substantially larger pressure increase to fail. Recent earthquakes near the fault may be evidence for continued pressure increase. This 50-km-long segment of the Nemaha fault is capable of hosting a M7 earthquake (20) and the fault zone continues for hundreds of kilometers. The increasing proximity of the earthquake swarm to the Nemaha fault presents a potential hazard for the Oklahoma City metropolitan area.

67

3.6 Conclusion Our earthquake relocations and pore pressure models indicate that four high-rate disposal wells are capable of increasing pore pressure above the reported triggering threshold (29-31) throughout the Jones swarm. The Jones swarm represents ~20% of 2008-2013 central and eastern US seismicity. Nearly 45% of this region’s seismicity, and currently nearly 15 M>3 earthquakes per week, may be linked to disposal of fluids generated during Oklahoma dewatering and following hydraulic-fracturing. Recent Oklahoma seismicity dominantly occurs within seismic swarms in the Arbuckle Group, Hunton Group, and Mississippi Lime dewatering plays. The injection-linked seismicity near Jones occurs up to 35 km away from the source wells, much further than previously considered in existing criteria for induced seismicity (13). Modern, very high-rate injection wells can therefore impact regional seismicity and increase seismic hazard. Regular measurements of reservoir pressure at a range of distances and azimuths from high-rate disposal wells could verify our model and potentially provide early indication of seismic vulnerability.

3.7 Acknowledgments This research benefited from discussion with E. Cochran, W. Ellsworth, and participants in a US Geological Survey (USGS) Powell Center Working Group on Understanding Fluid Injection Induced Seismicity (M.W., B.B., and S.G. are part of this group). C. Hogan identified many P- and S-phases. K.M.K was partially supported by USGS National Earthquake Hazards Reduction Program (NEHRP) grant G13AP00025, M.W. was partially supported by the USGS Powell Center grant G13AC00023, and G.A.A. was partially supported by NEHRP grant G13AP00024. This project used seismic data from EarthScope’s Transportable Array, a facility

68

funded by the National Science Foundation. Seismic waveforms are from the Incorporated Research Institutions for Seismology Data Management Center and the USGS CWB Query. Well data are from the Oklahoma Corporation Commission and the Oklahoma Geological Survey.

3.8 References [1]

W.L. Ellsworth, Injection-induced earthquakes, Science, 341 (2013).

[2]

K. M. Keranen, H. M. Savage, G. A. Abers, E. S. Cochran, Potentially induced earthquakes in Oklahoma, USA: links between wastewater injection and the 2011 Mw5.7 earthquake sequence, Geology. 41, p. 699-702 (2013).

[3]

W.-Y., Kim, Induced seismicity associated with fluid injection into a deep well in Youngstown, Ohio, J. Geophys. Res. 118, p. 3506–3518 (2013).

[4]

S. Horton, Disposal of hydrofracking waste fluid by injection into subsurface aquifers triggers earthquake swarm in central Arkansas with potential for damaging earthquake, Seism. Res. Lett. 83, p. 250–260 (2012).

[5]

C. Frohlich, C. Hayward, B. Stump, E. Potter, The Dallas-Fort Worth earthquake sequence: October 2008 through May 2009, Seism. Soc. Am. Bull. 101, p. 327–340 (2011).

[6]

A.H. Justinic, B. Stump, C. Hayward, C. Frohlich, Analysis of the Cleburne, Texas earthquake, sequence from June 2009 to June 2010, Bull. Seism. Soc. Am. 103, p. 3083-3093 (2013).

[7]

C. Frohlich, Two-year survey comparing earthquake activity and injection-well locations in the Barnett Shale, Texas, PNAS 109, p. 13934–13938, (2012).

69

[8]

A. McGarr, Maximum magnitude earthquakes induced by fluid injection, J. Geophys. Res. 119, 1008–1019 (2014).

[9]

The Central and Eastern United States is considered the portion of the contiguous United States east of 109° W.

[10] ANSS catalog, United States Geological Survey, http://earthquake.usgs.gov/earthquakes/search/, accessed 4/1/2014. [11] N.J. van der Elst, H.M. Savage, K.M. Keranen, G.A. Abers, Enhanced Remote Earthquake Triggering at Fluid-Injection Sites in the Midwestern United States, Science, 341, p. 164-167 (2013). [12] D.F. Sumy, E. S. Cochran, K. M. Keranen, M. Wei, G. A. Abers, Observations of Static Coulomb Stress Triggering of the November 2011 M5.7 Oklahoma Earthquake Sequence, J. Geophys. Res. 119 (2014). [13] S. D. Davis, C. Frohlich, Did (or will) fluid injection cause earthquakes? - criteria for a rational assessment, Seis. Res. Letters, 64, p. 207–224 (1993). [14] K. M. Keranen, M. Weingarten, G.A. Abers, B.A. Bekins, and S. Ge, Sharp increase in central Oklahoma seismicity since 2008 induced by massive wastewater injection, Science. 345(6195), 448-451 (2014). [15] K.V. Luza, J.E. Lawson, Seismicity and tectonic relationships of the Nemaha uplift in Oklahoma—part III, Oklahoma Geological Survey Spec. Pub. 81-3, p. 1-67 (1981). [16] S.P. Gay, The Nemaha trend--a system of compressional thrust-fold, strike-slip structural features in Kansas and Oklahoma: part 2, Shale Shaker 54, p. 39-49 (2003). [17] L.E. Gatewood, in Geology of giant petroleum fields, AAPG Memoir 14, Halbouty, M.T. editor, Tulsa: American Association of Petroleum Geologists (1970).

70

[18] F. Waldhauser, W.L. Ellsworth, A Double-Difference Earthquake Location Algorithm: Method and Application to the Northern Hayward Fault, California, Bull. Seism. Soc. Am. 90, p. 1353-1368, (2000). [19] A.A. Holland, Optimal fault orientations within Oklahoma, Seism. Res. Lett., 84(5), p. 876890 (2013). [20] D.L. Wells, K.J. Coppersmith, New Empirical Relationships among Magnitude, Rupture Length, Rupture Width, Rupture Area, and Surface Displacement, Bull. Seism. Soc. Am., 84(4), p. 974-1002 (1994). [21] Oklahoma Corporation Commission Imaging Web Application, http://imaging.occeweb.com/ [22] Monthly average volume was calculated using reported volumes for any month with nonzero volume (15). Injection rates over 90% larger than the median monthly value in a given year for each well were removed from calculations to remove data entry errors. [23] D. Chernicky, World Oil (2000); www.worldoil.com/September- 2000-Major-reserveincrease-obtained-by-dewatering-highwater-saturation-reservoirs.html. [24] K. Murray, State-scale perspective on water use and production associated with oil and gas operations, Oklahoma, US, Env. Sci. and Tech., 47, 4918-4925 (2013). [25] A.W. Harbaugh, E.R. Banta, M.C. Hill, M.G. McDonald, MODFLOW-2000, The U.S. Geological Survey Modular Ground-Water Model - User guide to modularization concepts and the ground-water flow process, U.S. Geological Survey Open-File Report 00-92, 121 p. (2000). [26] M.G. McDonald, A.W. Harbaugh, A modular three-dimensional finite-difference groundwater flow model U.S. Geol. Survey Open-File Report 83-875 (1984).

71

[27] E.K. Franseen, A.P. Byrnes, in J. R. Derby, R. D. Fritz, S. A. Longacre, W. A. Morgan, and C. A. Sternbach, Arbuckle Group platform strata in Kansas: A synthesis, eds., AAPG Memoir 98, p. 1031–1047 (2012). [28] B.C. Haimson and T.W. Doe, State of stress, permeability, and fractures in the Precambrian granite of northern Illinois, J. Geophys. Res., 88(B9), p. 7355–7371 (1983). [29] P. Reasenberg, R.W. Simpson, Response of Regional Seismicity to the Static Stress Change Produced by the Loma Prieta Earthquake, Science 255, p. 1687−1690 (1992). [30] Seeber, L., and Armbruster, J., Earthquakes as beacons of stress change, Nature 407, p. 6972 (2000). [31] R. Stein, The role of stress transfer in earthquake occurrence, Nature 402 (1999). [32] J. H. Healy, W.W. Rubey, D. T. Griggs, C.B. Raleigh, The Denver Earthquakes, Science 161, p. 1301-1310 (1968). [33] J. Ake, K. Mahrer, D. O’Connell, L. Block, Deep injection and closely monitored induced seismicity at Paradox Valley, Colorado, Bull. Seism. Soc. Am. 95, p. 664-683 (2005). [34] L. Seeber, J. G. Armbruster, W.-Y. Kim, A fluid-injection-triggered earthquake sequence in Ashtabula, Ohio: Implications for seismogenesis in stable continental regions, Bull. Seism. Soc. Am. 94, 76-87 (2004). [35] M.D. Zoback, J. Townend, B. Grollimund, Steady state failure equilibrium and deformation of intraplate lithosphere, Intl. Geology Rev. 44 (2002). [36] M.L. Zoback, First- and second-order patterns of stress in the lithosphere: The World Stress Map Project, J. Geophys. Res. 97 (B8), p. 11703-11728 (1992).

72

CHAPTER 4 HIGH-RATE INJECTION IS ASSOCIATED WITH THE INCREASE IN U.S. MID-CONTINENT SEISMICITY Abstract An unprecedented increase in earthquakes in the U.S. mid-continent began in 2009. Many of these earthquakes have been documented as induced by wastewater injection. We examine the relationship between wastewater injection and U.S. mid-continent seismicity using a newly assembled injection well database for the central and eastern U.S. We find the entire increase in earthquake rate is associated with fluid injection wells. High injection rate wells (>300,000 barrels/month) are much more likely to be associated with earthquakes than lower rate wells. At the scale of our study, a well's cumulative injected volume, monthly wellhead pressure, depth, and proximity to crystalline basement do not strongly correlate with earthquake association. Managing injection rates may be a useful tool to minimize the likelihood of induced earthquakes.

This chapter has been previously published: Weingarten, M., Ge, S., Godt, J.W., Bekins, B.A. and J.L. Rubinstein (2015), High-rate injection is associated with the increase in U.S. mid-continent seismicity, Science, 348 (6241), pg. 1336-1340, doi:10.1126/science.1255802.

73

4.1 Introduction The injection of fluids into the subsurface has been known to induce earthquakes since the mid-1960s (1-3). However, few additional cases of earthquakes induced by wastewater injection have been documented until 2009 (4). The hazard from these earthquakes was considered small because they were infrequent and not expected to be large (largest observed prior to 2011 was the M 4.9 Rocky Mountain Arsenal earthquake in 1967) (4-6). The central and eastern United States (CEUS) has seen an unprecedented increase in earthquake rate since 2009, and many of these earthquakes are believed to be induced. Along with the increased rate, several damaging earthquakes have occurred such as the 2011 M 5.6 Prague, Oklahoma earthquake (89), the 2011 M 5.3 Trinidad, Colorado earthquake (10), the 2012 M 4.8 Timpson, Texas earthquake (11), and the 2011 M 4.7 Guy, Arkansas earthquake (12). The increased earthquake rate and occurrence of multiple damaging earthquakes has prompted the scientific community to refocus efforts to understand the hazard posed by injection-induced earthquakes (13). The sudden appearance of several large, potentially induced earthquakes led to many sitespecific case studies (4). These case studies examined the operation of injection wells in close proximity to the earthquakes showing a link between the timing and location of injection and seismicity (12, 14-18). While useful to understand the individual systems in which these earthquakes occurred, broader scale studies are needed to understand the phenomenon as a whole. One previous study examined earthquakes in Texas' Barnett Shale region and found that earthquakes are commonly located near wells injecting more than 150,000 barrels/month (19). However, to fully understand the possibility of induced seismicity associated with a given injection well, we must analyze a range of geologic, hydrogeologic and operational differences between injection wells that are potentially associated with earthquakes and those that are not.

74

4.2 Methods, Data & Results We examined the location and timing of earthquakes and their relationship to the location and operation of injection wells across the CEUS (Fig. 4.1). We have compiled from publicly available sources, an injection well database that documents the location and operational parameters of Underground Injection Control Class II injection wells in the CEUS (Fig. 4.1; Appendix 4.A1). Class II injection wells inject fluids associated with oil and gas production and are distinct from hydraulically fractured production wells (20). The database contained 187,570 wells as of December 2014, with 56% actively injecting fluid (Fig. 4.1) while the remaining 44% are inactive or abandoned. About 75% of the active Class II injection wells operated for the purposes of enhanced oil recovery (EOR), while nearly all of the remaining wells were designated as salt water disposal wells (SWD) (Fig. 4.2). EOR wells inject fluid into depleted oil reservoirs to increase oil production. SWD wells inject to dispose of waste fluids produced by oil and gas production, which would otherwise be hazardous to surface waters or underground sources of drinking water. Injection wells are geographically clustered in the basins and regions of major oil and gas operations. Texas, Oklahoma, Kansas, and Wyoming contain ~85% of all Class II injection wells in the CEUS. The spatial density of active SWD wells is highest (≥ 5 wells / 5 km2) in the Forth Worth Basin of north-central Texas and the Mississippi Lime Play extending from north-central Oklahoma northward into central Kansas. The spatial density of active EOR wells is highest (≥ 5 wells / 5 km2) in the Permian Basin of West Texas, Fort Worth Basin, south-central Oklahoma, and southeastern Kansas (Fig. 4.2).

75

Figure 4.1. (A) Map showing the location of active Class II injection wells in the central and eastern United States (CEUS). Active injection wells from the database are shown as blue circles. Spatiotemporally associated injection wells, defined as those within a 15 km radius and active at the time of an earthquake, are shown as yellow circles. The CEUS region comprises all states intersected by 109° W longitude and eastward. The total number of wells, including inactive or abandoned wells in the CEUS is 188,570. Of the 18,757 associated injection wells, 14,490 (>77%) are currently active. (B) The inset pie diagram shows spatiotemporally associated injection wells by state. Only 8% of all injection wells are located in Oklahoma, but 40% of the associated injection wells in the CEUS are located in Oklahoma. 76

Figure 4.2. Maps of (A) active salt water disposal (SWD) wells, (B) active enhanced oil recovery wells (EOR) wells, (C) the number of active SWD wells per five square kilometers and (D) the number of EOR wells per five square kilometers. EOR wells are spatially dense in the Permian Basin and Fort Worth Basin in Texas, the Ardmore Basin of south-central Oklahoma, and the southern Fort City Basin of southeastern Kansas. SWD wells are spatially dense in the Fort Worth Basin in Texas and from central Oklahoma northwards into central Kansas.

77

We obtained earthquake location and magnitude data from the Advanced National Seismic System’s (ANSS) comprehensive earthquake catalog (21). During the study period (1973 - 2014), we identified 7,175 M ≥ 0.0 events in the ComCat catalog in the CEUS region (Fig. 4.3). Although the ComCat catalog is not complete down to M 0.0 during the study period, we treated all earthquakes as potentially induced events to capture the most comprehensive dataset of associated earthquakes and injection wells. We used a magnitude of completeness of 3.0 when comparing associated versus non-associated earthquakes through time (7). We used spatial and temporal filtering methods to discriminate injection wells that may be associated with earthquakes from those that are probably not. We considered any earthquake within 15 km of an active injection well to be associated with that well. This distance of association is based upon the sum of a 5 km radius within which earthquakes are traditionally considered as potentially induced (22) and a 10 km estimate of the spatial uncertainty in earthquake epicenter location in the CEUS (23). We designed the temporal filter to include only injection wells active at the time of the spatially associated earthquake. Both filters could be considered conservative as induced seismicity has been found tens of kilometers from injection wells (24) and also after a well is shut-in (25) due to the injection prior to the well becoming inactive (4-5). To analyze the sensitivity of our results to these filtering parameters, we also tested our analysis using spatial association distances of 5 km and 10 km. This first-order analysis attempts to understand which basic well properties affect the likelihood of earthquake association.

78

Figure 4.3. Map showing the locations of M0.0 to M5.0+ earthquakes in the ANSS ComCat earthquake catalog from January 1st, 1973 through December 31st, 2014. White dots denote earthquakes that are not spatiotemporally associated with injection wells. Red dots denote earthquakes that are spatiotemporally associated with injection wells. Following Ellsworth (2013), the U.S. mid-continent is defined by the dashed lines inside of the greater CEUS.

79

We find 18,757 injection wells (~10% of all wells) associated with earthquakes in the CEUS after filtering, mostly in the states of Oklahoma and Texas (Fig. 4.1). The number of associated injection wells has tripled since the year 2000 (Fig. 4.4). The spatiotemporal filter identifies every case of induced seismicity from Class II injection wells documented in the literature for the CEUS region (Table 4.1). We identify far more injection wells that are potentially related to earthquakes than those indicated by published cases. Of the wells that are associated with earthquakes, 66% are EOR wells. However, active SWD wells are more than 1.5 times as likely as active EOR wells to be associated with an earthquake, which accounts for their respective well totals (Fig. 4.1). The finding that SWD wells are preferentially associated with earthquakes likely resides with difference in well operation. SWD injection causes a net-positive reservoir pressure change, whereas EOR injection and extraction wells are typically operated in tandem with injection rates managed carefully to balance reservoir pressures (7). Over the last four decades, more than 60% of all CEUS seismicity (M 3.0+) is associated with injection wells using our filtering criteria (Fig. 4.3, Fig. 4.5). Varying the spatial distance of association by several kilometers only changes this percentage by +/-5% (Fig. 4.6). Prior to the year 2000, an average of 22% of all CEUS seismicity was associated with injection wells. The yearly percent of associated earthquakes has risen sharply to ~89% from 2011 to 2014 (Fig. 4.5, Fig. 4.7). The percentage increase of associated earthquakes, combined with the rising CEUS earthquake rate, implies that recent seismicity in the CEUS is preferentially occurring near injection wells. The number of non-associated earthquakes during the same period has also remained stable (Fig. 4.5).

80

Figure 4.4. The number of associated injection wells through time identified using a spatial filter of 15 km combined with a temporal filter for active wells only. For each associated well, we find the date of the first associated earthquake. We then count the cumulative number of associated wells prior to the given year to produce the curve. We observe an increase in the number of associated wells through time; 6,129 wells were associated with earthquakes in the year 2000. By the end of 2014, the number of associated wells has tripled to more than 18,757.

81

Site

Date

Maximum Earthquake Magnitude

Reference Number

Spatiotemporal Filter

Painesville, OH Youngstown, OH Raton Basin, CO Raton Basin, CO Jones, OK Prague, OK Cogdell, TX Cogdell, TX Dagger Draw, NM Timpson, TX El Dorado, AR Guy, AR Southern Alabama Kermit Field, TX Hunt Field, MS Rangely, CO Sleepy Hollow Field, NE

1/31/1986 12/31/2011 9/5/2001 8/23/2011 4/16/2013 11/6/2011 6/16/1978 9/11/2011 12/19/2005 5/17/2012 12/9/1983 2/28/2011 10/24/1997 1/19/1976 5/4/1977 3/19/1979 7/16/1979

5 4 4.5 5.3 4.4 5.6 4.4 4.3 4.1 4.8 3 4.7 4.9 3.5 3.6 3.3 3.2

15 18 40 10 24 8 41 42 43 11 44 12 45 41 41 41 41

Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes

Table 4.1. Cases of documented and suspected induced seismicity from EPA Class II wells used to test the spatiotemporal filter.

82

Figure 4.5. The grey bars represent the number of >M3.0 earthquakes per year in the U.S. midcontinent (Fig. 4.3) located by the networks of the ANSS ComCat earthquake catalog from 1/1/1973 - 12/31/2014. The red bars represent the number of earthquakes that are spatiotemporally associated with injection wells. The black line denotes the number of nonassociated earthquakes per year. Over the time period of the catalog, the number of nonassociated earthquakes per year has stayed roughly constant at 10 - 25 per year. Meanwhile, the number of associated earthquakes per year has risen from ~1 -7 /year in the 1970s to 75 190/year between 2011-2013 and >650 earthquakes in 2014.

83

Figure 4.6. The number of associated earthquakes as a function of spatial filter radius using the Advanced National Seismic System's (ANSS) earthquake catalog from 1973 - March 24, 2014. The number of associated earthquakes is not linearly related to spatial filter radius. Similar numbers of earthquakes are associated using spatial filters of 15 km or 30 km. This is an unexpected result, as one might expect the rate of associated earthquakes to increase with increasing spatial distance. The number of earthquakes associated with a 15 km filter is 2,219 whereas a 30 km filter associates 2,510 earthquakes. Thus, the lower sensitivity of earthquake association to spatial filter radius past 15 km indicates this radius is appropriate for spatial filtering.

84

Figure 4.7. The curve shows the yearly percent of central and eastern U.S. (CEUS) earthquakes (M ≥ 3.0) spatiotemporally associated with injection wells through time. Since 2000, the yearly percent of earthquakes associated with injection wells has risen from an average of 22% from 1973-2000 to over ~98% in 2014. During the same time period, the cumulative percent of all CEUS earthquakes associated with injection wells has increased from 20% from 1973-2000 to 70% in the last 14 years.

85

This increase in associated earthquake rate does not correspond to an increase in the rate of wells completed; the well completion rate has remained constant over this period (Fig. 4.8) (26). A portion of the increase in associated earthquakes may be due to increasing well area or density, but we suggest this effect is minimal considering the relative increase in the spatial coverage of wells was much more rapid between 1960 and 1980 than in recent years (Fig. 4.8). The main contributors to the dramatic increase in seismicity associated with injection wells are in central and north-central Oklahoma. New production methods in these regions are generating large volumes of produced water, which are injected at high rates (Fig. 4.9A, Fig. 4.9B) (27). Regions such as west Texas, southern Colorado, central Arkansas and southern Illinois also show concentrations of seismicity associated with injection wells (Fig. 4.3). However, several regions with large numbers of injection wells appear to be aseismic during the study period, including the Williston Basin of North Dakota (28), the Michigan Basin, and extensive areas of the Texas and Louisiana Gulf Coast (Fig. 4.1, Fig. 4.3). Several operational parameters are hypothesized to influence the likelihood of an induced seismic event: injection rate (19,24), cumulative injected volume (29-30), wellhead injection pressure (31), and injection in proximity to crystalline basement (18, 32). Four states, Oklahoma, Arkansas, Colorado, and New Mexico, with both natural and induced earthquakes all have over 15 years of injection data that include readily accessible information on monthly injection rate and pressure for a large proportion of operational wells. Using these data, we explore injection operational parameter control on the likelihood that SWD and EOR wells will be associated with earthquakes.

86

Figure 4.8. (A) A histogram of the number of wells completed in a given year through time. The number of completed Class II injection wells each year varies according to fluctuations in domestic oil and gas operations. (B) A cumulative curve of the number of Class II injection wells completed by a given year. The overall rate of wells completed each year has not markedly changed since the early 1980s.

87

Figure 4.9A.Maps of cumulative injected volume per five square kilometer boxes in Oklahoma (OK), Colorado (CO), New Mexico (NM) and Arkansas (AR) for three time periods: 2002-2005, 2006-2009 and 2010-2013. Each map sums the volume injected over a given time period. Areas of highest cumulative injected volume during these time periods are in southern, central and north-central OK as well as southern AR. Prior to the onset of seismic rate changes in the midlate 2000s, cumulative injected volumes in several regions in these four states were similar to that of today. However, some regions, such as north-central Oklahoma, have experienced distinct injection increases in the past decade. 88

Figure 4.9B. Maps showing the change in cumulative injected volume per five square kilometer boxes relative to four time periods. The upper map differences volumes summed from 20022005 from volumes summed from 1998-2001. The middle map differences volumes summed from 2006-2009 from volumes summed from 2002-2005. The lower map differences volumes summed from 2010-2013 from volumes summed from 2006-2009. The middle map shows the activation of injection operations in central Oklahoma between 2006-2009. The bottom map shows the activation of injection operations in north-central Oklahoma between 2010-2013. 89

The maximum monthly injection rate of wells across these four states vary several orders of magnitude, ranging from 100 barrels (~15.9 m3) per month (bbl/month) up to 2 million bbl/month (~318,000 m3) (Fig. 4.10A) (33). The average SWD well operates at a monthly rate of ~13,000 bbl/month. For each histogram bin in Figure 4.10A and 4.10B, we calculate the percent of wells associated with earthquakes. The likelihood that an SWD well is associated with earthquakes increases as the maximum injection rate increases (Fig. 4.10C). To discern whether the association is random, we estimated upper (95%) and lower (5%) confidence bounds using a bootstrapped resampling method with 10,000 resamples (Fig. 4.10C) (34-35). Wells operating at maximum injection rates greater than 300,000 bbl/month fall outside the bootstrap resampling confidence bounds, suggesting a greater-than-expected likelihood of association with an earthquake at a statistical significance near 99%. This is contrasted with wells operating at maximum injection rates less than 100,000 bbl/month, which mostly fall within the bounds of random association. We confirmed this result using spatial distances of association of 5 km and 10 km (Fig. 4.11) as well as restricting our well associations to earthquakes greater than M 3.0 (Fig. 4.12). Of the 413 wells operating at injection rates greater than 300,000 bbl/month, 253 (61%) of them are spatiotemporally associated compared with only 40% of wells operating at injection rates less than 10,000 bbl/month. Additionally, 34 (76%) of the 45 highest-rate SWD wells (injecting over a million barrels per month) are associated with an earthquake. When SWD operations are examined state-by-state, the overall percent associated varies but the trend of increased earthquake association at higher rates is generally preserved (Fig. 4.13A). Fewer data are available for EOR wells, but we do not observe a clear trend of increasing earthquake association with increasing injection rate for EOR wells (Fig. 4.14A). Without considering

90

Figure 4.10. (A) Histogram showing the maximum monthly injection rate of salt water disposal (SWD) wells in Oklahoma, Arkansas, Colorado, and New Mexico. The blue bars show the maximum monthly injection rate for all SWD wells, whereas the yellow bars show the maximum monthly injection rate for SWD wells spatiotemporally associated with an earthquake. (B) Histogram showing the cumulative injected volume at all wells in the same states as (A). The blue distribution represents the cumulative injection volume for SWD wells, while the yellow distribution is for spatiotemporally associated wells. Injection data for Oklahoma was available from 1995-2013, for Arkansas from 1999-2013, for Colorado from 1999-2014, and for New Mexico from 1994-2014. (C-D) The percentage of all wells that are associated with an earthquake in each histogram bin is plotted as a function of (C) maximum monthly injection rate and (D) cumulative injected volume. The two dashed red lines represent the upper (95%) and lower (5%) confidence bounds in each bin generated by 10,000 bootstrap resamples and following the assumption that the rate of association is random. The shaded grey region of (D) indicates a lack of associated wells at the given volume. These data are also broken down state by state and for enhanced oil recovery (EOR) wells in Fig. 4.13 and Fig. 4.14.

91

Figure 4.11. Injection parameter analysis for three radii of spatial association: 5 km, 10 km and 15 km. The number of associated wells decreases as a function of decreasing spatial radius as shown in the histograms of associated versus all SWD wells. However, the statistics show high injection rate wells are still preferentially associated with earthquakes using both a 5 km and 10 km radius. In addition, cumulative injected volume does not show a clear trend of increasing association with increasing cumulative injected volume.

92

Figure 4.12. We re-ran the same operational parameter analysis from Figure 4 (maximum injection rate and cumulative injected volume) associating wells with earthquakes only greater than M3.0. (A) Histogram showing the maximum monthly injection rate of salt water disposal (SWD) wells in Oklahoma, Arkansas, Colorado, and New Mexico. The blue bars show the maximum monthly injection rate for all SWD wells, whereas the yellow bars show the maximum monthly injection rate for SWD wells spatiotemporally associated with an earthquake. (B) Histogram showing the cumulative injected volume at all wells in the same states as (A). The blue distribution represents the cumulative injection volume for SWD wells, while the yellow distribution is for spatiotemporally associated wells. Injection data for Oklahoma was available from 1995-2013, for Arkansas from 1999-2013, for Colorado from 1999-2014, and for New Mexico from 1994-2014. (C-D) The percentage of all wells that are associated with an earthquake in each histogram bin is plotted as a function of (C) maximum monthly injection rate and (D) cumulative injected volume. The two dashed red lines represent the upper (95%) and lower (5%) confidence bounds in each bin generated by 10,000 bootstrap resamples and following the assumption that the rate of association is random. The shaded grey region of (D) indicates non-associated wells at the given volume.

93

geologic or hydrogeologic setting, the highest-rate SWD wells are nearly twice as likely to be near an earthquake as low-rate SWD wells. We next examine whether cumulative volume injected affects the likelihood of well association with earthquakes. For the four states examined during the period from 1973 to 2014, cumulative volume injected ranged from 1,000 bbl to nearly 100 million bbl (Fig. 4.10B). Many large cumulative volume wells inject at moderate rates for decades, providing a contrasting dataset from maximum injection rate. We do not observe a strong trend of increasing SWD well association as a function of increasing cumulative injected volume (Fig. 4.10D). The difference between the association rate of wells, which have injected more than 1,000,000 bbl cumulatively (45%) and those which have injected less than 10,000 bbl cumulatively (38%) is not statistically significant using a bootstrap resampling method (35). The percent of wells associated with earthquakes at high cumulative injected volumes can be mostly explained by random variation given the total number of associated and non-associated wells. EOR wells exhibit a similar trend of earthquake association to SWD wells as a function of cumulative injected volume (Fig. 4.14B). If we instead calculate cumulative injected volume not for individual wells, but for all wells within 15 km of an associated earthquake, we observe a log-normal distribution of volumes without a clear threshold of increased earthquake association (Fig. 4.15). We do not observe cumulative injected volume as a significant parameter affecting the likelihood of an injection well's association with an earthquake. The majority of Class II injection wells operate at monthly wellhead injection pressures less than 500 pounds per square inch (psi). Reported wellhead pressures for both SWD and EOR wells ranged from 0 - 3,000 psi (Fig. 4.16A; Fig. 4.16B). In the same four states studied, the proportion of SWD and EOR wells associated with earthquakes show no strong correlation

94

Figure 4.13. (A) Upper row shows histograms of the maximum monthly injection rate of all SWD wells in Oklahoma (OK), Arkansas (AR), Colorado (CO), and New Mexico (NM). The blue bars indicate the maximum monthly injection rate for all salt-water disposal (SWD) wells; yellow bars show the maximum monthly injection rate for wells spatiotemporally associated with an earthquake. The lower row of charts shows the percentage of all wells associated with an earthquake in each histogram bin. The two dashed red lines represent the upper (95%) and lower (5%) confidence bounds in each bin generated by 10,000 bootstrap resamples and following the assumption that the rate of association is random. The shaded regions of the plots denote that no associated maximum monthly injection rate is observed at those values. (B) Histograms of cumulative injected volume for the same wells and states depicted in (A) as well as their respective percent associated plots.

95

Figure 4.14. (A) Upper row shows histograms of the maximum monthly injection rate of all EOR wells in Oklahoma (OK), Arkansas (AR) and Colorado (CO). No EOR injection data was available for New Mexico (NM). The blue bars indicate the maximum monthly injection rate for all enhanced oil recovery (EOR) wells; yellow bars show the maximum monthly injection rate for wells spatiotemporally associated with an earthquake. The lower row of charts shows the percentage of all wells associated with an earthquake in each histogram bin. The two dashed red lines represent the upper (95%) and lower (5%) confidence bounds in each bin generated by 10,000 bootstrap resamples and following the assumption that the rate of association is random. The shaded regions of the plots denote that no associated maximum monthly injection rate is observed at those values. (B) Histograms of cumulative injected volume for the same wells and states depicted in (A) as well as their respective percent associated plots.

96

Figure 4.15. Histogram of cumulative injected volume within 15 km of each associated earthquake. While all associated earthquakes have some amount of volume injected within 15 km, we observe a log-normal distribution of the number of associated earthquakes at a given injected volume. Our interpretation of the data suggests the level of injected volume within 15 km of associated earthquakes does not correlate to the number of associated earthquakes observed.

97

towards increased monthly wellhead pressures (Fig. 4.16; Fig. 4.17). However, we note that reported monthly wellhead pressure may not always be reliable because many wells report constant wellhead pressures despite changing injection rates. In addition, wellhead injection pressure data may not reflect the pore-pressure conditions in the injection formation due to friction in the wellbore and other factors. Wells reporting zero wellhead pressure also still create bottomhole pressure from the hydrostatic fluid column in the well which could be large enough induce an earthquake. There are several hundred wells with zero wellhead pressures that are associated with earthquakes (Fig. 4.16A; Fig. 4.16B). This is consistent with field observations of earthquakes induced by wells with zero wellhead pressure (10, 36). We do not consider the reported maximum wellhead pressure to be a controlling factor on injection well and earthquake association. This finding, together with the indication SWD wells are preferentially associated with earthquakes, underscores the need to collect reservoir pressure data. Ideally, pre-injection reservoir pore pressure and bottomhole formation pressure measurements during injection would prove more useful in determining whether a link exists between net injection, injection pressure and earthquakes. Injection depth and proximity to crystalline basement have been hypothesized to affect the likelihood that wells are associated with earthquakes (32). Comparison of injection depths for most states in CEUS, excluding Mississippi, Indiana, West Virginia and Alabama, is possible as these data are more readily available than injection rates and pressures (Appendix 4.A1). Class II injection wells are permitted over a wide range of injection depths from 300 m to 4,000 m (Fig. 4.18). The majority of both SWD and EOR wells inject between 300 and 1,500 m (Fig 4.18B; Fig. 4.18E). Wells associated with earthquakes also inject over a similarly wide range of injection depths (Fig. 4.18A; Fig. 4.18D). We find no clear evidence that increasing injection

98

Figure 4.16. (A) Histogram of maximum wellhead pressure for salt water disposal wells (SWD) and associated SWD wells (yellow) for Oklahoma, New Mexico, Arkansas and Colorado combined. (B) Histogram of maximum wellhead pressure for enhanced oil recovery wells (EOR) and associated EOR wells in Oklahoma, Arkansas and Colorado combined. (C-D) The percent of wells associated with an earthquake in each histogram bin at a given maximum wellhead pressure for (C) SWD wells and (D) EOR wells. The dashed red lines are the upper (95%) and lower (5%) confidence intervals in each bin generated by 10,000 bootstrap resamples and following the assumption that the rate of association is random. No clear trend of increasing wellhead pressure for associated wells is evident for either SWD wells or EOR wells.

99

Figure 4.17. Injection pressure analysis for wells associated with earthquakes greater than M3.0 (A) Histogram of maximum wellhead pressure for salt water disposal wells (SWD) and associated SWD wells (yellow) for Oklahoma, New Mexico, Arkansas and Colorado combined. (B) Histogram of maximum wellhead pressure for enhanced oil recovery wells (EOR) and associated EOR wells in Oklahoma, Arkansas and Colorado combined. (C-D) The percent of wells associated with an earthquake in each histogram bin at a given maximum wellhead pressure for (C) SWD wells and (D) EOR wells. The dashed red lines are the upper (95%) and lower (5%) confidence intervals in each bin generated by 10,000 bootstrap resamples and following the assumption that the rate of association is random. No clear trend of increasing wellhead pressure for associated wells is evident for either SWD wells or EOR wells.

100

depth increases the likelihood a well will be associated with seismicity; the proportion of both SWD and EOR wells associated with earthquakes does not increase with increasing injection depth (Fig. 4.18C; Fig. 4.18F). However, comparison of injection depths neglects the large variations in sediment thickness across the CEUS. Using a map of sediment thickness across the CEUS (37), we estimate injection proximity to basement for all wells by subtracting sediment thickness at the closest sediment thickness data point from the injection well depth. The sediment thickness map was tested against known reference depths of crystalline basement and found to have errors in thickness up to +/- 15% (Table 4.2). Thousands of wells in the CEUS inject fluid within 500 meters of crystalline basement rock, but only a small proportion are associated with seismicity. When taking into account the error in basement depth over the CEUS region, we did not observe a significant correlation between well's injecting near basement and earthquakes using a bootstrap resampling method (Fig. 4.19). However, injection wells operating very far from basement, between 7 and 12 km vertically, exhibited an association rate near zero. We found similar results for both depth parameters using only well associations with earthquakes greater than M3.0 (Fig. 4.20; Fig. 4.21). This finding supports the notion that detailed stratigraphic knowledge surrounding the injection interval is necessary to quantify the mechanistic linkage between injection and seismicity (32). The lack of spatiotemporal association between injection and seismicity in several regions highlights the apparent influence of factors other than injection well operation. The San Juan Basin of New Mexico, the Williston Basin of North Dakota, the Michigan Basin, and extensive areas of the Texas and Louisiana Gulf Coast contain thousands of SWD and EOR wells that are not associated with seismicity (Fig. 4.1). In some of these regions, wells inject at

101

Test Site

Latitude Longitude

Reference Depth (km)

MooneyKaban Depth (km)

Residual Percent (km) Error

Reference Number

Youngstown, OH

41.12

-80.68

2.74

3.15

0.41

14.96

18

Denver, CO

39.85

-104.86

3.64

3.56

-0.08

-2.20

1

Jones, OK

35.40

-97.44

3.51

3.08

-0.43

-12.13

24

38.21

-108.88

4.71

5.16

0.45

9.62

38

48.00

-104.00

4.58

4.76

0.18

3.93

39

48.00

-99.00

1.15

1.10

-0.05

-4.35

39

31.87

-94.45

5.00

5.41

0.41

8.20

11

43.45

84.38

4.97

4.93

-0.04

-0.80

46

Paradox Valley Williston Basin West Williston Basin East Timpson, TX Michigan Basin

Table 4.2. Test of Mooney and Kaban (2010) sediment thickness depths against known reference depths in several regions across the CEUS where the proximity to basement calculation is made.

102

Figure 4.18. Histograms of injection depths for (A) associated salt-water disposal (SWD) wells and (B) all SWD wells. (C) The percent of SWD wells associated with seismicity for each injection depth bin. Histograms of injection depths for (D) associated enhanced oil recovery (EOR) wells and (E) all EOR wells. (F) The percent of EOR wells associated with seismicity for each injection depth bin. The dashed red lines in C and F represent the 5% and 95% confidence intervals generated by 10,000 bootstrap resamples. No clear trend of increasing depth correlating to increased numbers of associated wells is evident.

103

Figure 4.19. Injection well depths plotted by each well's proximity to crystalline basement rock. (A) Proximity to basement for associated wells (yellow bars) versus all wells (blue bars) and (B) percent of all wells associated with earthquakes at a given proximity to basement (shaded grey region) with a plus or minus 15 percent error in the basement depth included. The dashed red lines represent the 5% and 95% confidence intervals generated by 10,000 bootstrap resamples. Increasing positive values indicate increasing distance between a given injection well's depth and the depth of crystalline basement rock. We do not identify any correlation between the proximity of injection to basement and association rate. However, well's injecting between 7 - 12 kilometers from crystalline basement exhibiting a near-zero association rate.

104

Figure 4.20. Histograms of injection depths for (A) associated salt-water disposal (SWD) wells and (B) all SWD wells. (C) The percent of SWD wells associated with seismicity for each injection depth bin. Histograms of injection depths for (D) associated enhanced oil recovery (EOR) wells and (E) all EOR wells. (F) The percent of EOR wells associated with seismicity for each injection depth bin. Spatiotemporal filtering analysis was re-run using only earthquakes greater than M3.0 to see the effect of injection depth on large magnitude earthquakes. The dashed red lines in C and F represent the 5% and 95% confidence intervals generated by 10,000 bootstrap resamples. No clear trend of increasing depth correlating to increased numbers of associated wells is evident.

105

Figure 4.21. Injection well depths plotted by each well's proximity to crystalline basement rock. In this analysis, we used a spatiotemporal filter which only associates earthquakes greater than M3.0 to see the effect of proximity to basement on larger magnitude earthquakes. (A) Proximity to basement for associated wells (yellow bars) versus all wells (blue bars) and (B) percent of all wells associated with earthquakes at a given proximity to basement (shaded grey region) with a plus or minus 15 percent error in the basement depth included. The dashed red lines represent the 5% and 95% confidence intervals generated by 10,000 bootstrap resamples. Increasing positive values indicate increasing distance between a given injection well's depth and the depth of crystalline basement rock. We do not identify a strong correlation between the proximity of injection to basement and association rate. However, well's injecting between 7 and 12 kilometers from basement exhibit a near-zero association rate.

106

similar rates to those in areas such as central Oklahoma where large numbers of wells are associated with earthquakes. In the aseismic Michigan Basin, 30 wells operate at maximum injection rates greater than 200,000 bbl/month (Fig. 4.22). Obviously, other factors in addition to high injection rate must play a role; the regional state of stress, fault size, fault orientation, the presence of fluid pathways between the injection point and faults, as well as other geologic factors must be examined to assess the potential for injection-induced seismicity (4).

4.3 Conclusion Our analysis shows injection rate is the most important well operational parameter affecting the likelihood of an induced seismic event, in regions and basins potentially prone to induced seismicity. High-rate SWD wells are nearly twice as likely as low-rate wells to be near an earthquake. These high-rate wells perturb the ambient reservoir pressure by a larger magnitude and over a larger area than low-rate wells, thus increasing the likelihood pressure changes will reach an optimally oriented, critically stressed fault. Previous studies have shown high-rate wells exert greater influence on the extent and magnitude of reservoir and fault pressure perturbation (24). At the scale of our study, no other operational parameter was found to have a strong influence on the likelihood of association with an earthquake. The important distinction between operational parameters such as injection rate and cumulative injected volume shows the effect of the recent rise of new production methods and high-rate SWD wells. Thus, the oil and gas industry and regulatory bodies can use this operational parameter to lower the likelihood of earthquakes associated with injection wells.

107

Figure 4.22. The spatial distribution of Class II injection wells in the Michigan Basin. There are no injection wells associated with seismicity in the entire Michigan Basin. The lower figure shows a histogram of injection rates for Class II wells in the Michigan Basin. The solid red line shows the Gaussian fit to the distribution and the dashed red line denotes the threshold of injection rates greater than 200,000 bbl/month. The mean injection rate of wells in Michigan is ~4,300 barrels/month, which is lower than that of other states with associated wells. However, more than 30 injection wells have maximum injection rates greater than 200,000 bbl/month.

108

4.4 Acknowledgements This work was conducted as a part of the Understanding Fluid Injection Induced Seismicity Working Group supported by the John Wesley Powell Center for Analysis and Synthesis, funded by the U.S. Geological Survey (Grant G13AC00023). The authors thank J. Hardebeck and W. Ellsworth for their thoughtful comments. This project was aided by injection data contributed by A. Holland (OK), C. Eisenger (CO), T. Kropatsch (WY), J. Amrheim (IN), T. Tomastik (OH), S. Platt (PA), I. Allred (UT), M. Berry (UT), A. Wickert (TX) and I. VanVloten (CEUS). This project used earthquake data from the ANSS Comprehensive Catalog. The well data used in this study are available as supplementary materials on corresponding authors ResearchGate profile.

4.5 References [13]

D. M. Evans, The Denver area earthquakes and the Rocky Mountain disposal well, The Mountain Geologist. 3, 23–26 (1966).

[14]

J. H. Healy, W.W. Rubey, D. T. Griggs, C.B. Raleigh, The Denver Earthquakes, Science. 161, 1301-1310 (1968).

[15]

C.B. Raleigh, J. H. Healy, J. D. Bredehoeft, An experiment in earthquake control at Rangely, Colorado, Science. 191, 1230-1237 (1976).

[16]

National Research Council, Induced Seismicity Potential in Energy Technologies (National Academies Press, Washington, DC, 2013).

[17]

R. B. Herrmann, S. K. Park, C. Y. Wang, The Denver earthquakes of 1967-1968, Bull. Seismol. Soc. Am. 71, 731-745 (1981).

109

[18]

M.D. Petersen et al., Documentation for the 2008 update of the United States National Seismic Hazard Maps, U.S. Geo. Survey Open-File Report 08-1128 (2008).

[19]

W.L. Ellsworth, Injection-induced earthquakes, Science. 341 (2013).

[20]

K. M. Keranen, H. M. Savage, G. A. Abers, E. S. Cochran, Potentially induced earthquakes in Oklahoma, USA: links between wastewater injection and the 2011 Mw5.7 earthquake sequence, Geology. 41, 699-702 (2013).

[21]

A. L. Llenos, A. J. Michael, Modeling earthquake rate changes in Oklahoma and Arkansas: possible signatures of induced seismicity, Bull. Seismol. Soc. Am. 103, 2850-2861 (2013).

[22]

J. L. Rubinstein, W.L. Ellsworth, A. McGarr, H. Benz, The 2001-present induced earthquake sequence in the Raton Basin of northern New Mexico and southern Colorado, Bull. Seismol. Soc. Am. 104, 1-20 (2014).

[23]

C. Frohlich et al., The 17 May 2012 M4.8 earthquake near Timpson, East Texas: An event possibly triggered by fluid injection, J. Geophys. Res. 119, 581–593 (2014).

[24]

S. Horton, Disposal of hydrofracking waste fluid by injection into subsurface aquifers triggers earthquake swarm in central Arkansas with potential for damaging earthquake, Seism. Res. Lett. 83, 250–260 (2012).

[25]

M. D. Petersen et al., Incorporating induced seismicity in the 2014 United States National Seismic Hazard Model -- Results of 2014 workshop and sensitivity studies, U.S. Geo. Surve Open-File Report 2015-1070 (2015).

[26]

P.A. Hsieh, J.D. Bredehoeft, A reservoir analysis of the Denver earthquakes: a case of induced seismicity, J. Geophys. Res. 86, 903-920 (1981).

110

[27]

C. Nicholson, E. Roeloffs, R. L. Wesson, The northeastern Ohio earthquake of 31 January 1986: was it induced?, Bull. Seismol. Soc. Am. 78, 188-217 (1988).

[28]

J. Ake, K. Mahrer, D. O'Connell, L. Block, Deep injection and closely monitored induced seismicity at Paradox Valley, Colorado, Bull. Seismol. Soc. Am. 95, 664-683 (2005).

[29]

C. Frohlich, C. Hayward, B. Stump, E. Potter, The Dallas-Fort Worth earthquake sequence: October 2008 through May 2009, Seism. Soc. Am. Bull. 101, 327–340 (2011).

[30]

W.-Y., Kim, Induced seismicity associated with fluid injection into a deep well in Youngstown, Ohio, J. Geophys. Res. 118, 3506–3518 (2013).

[31]

C. Frohlich, Two-year survey comparing earthquake activity and injection-well locations in the Barnett Shale, Texas, PNAS 109, 13934–13938, (2012).

[32]

U.S Environmental Protection Agency, Evaluation of impacts to underground sources of drinking water by hydraulic fracturing of coalbed methane reservoirs, EPA Publication 816-F-04-017 (2004).

[33]

ANSS Comprehensive Catalog (ComCat), U.S. Geological Survey, http://earthquake.usgs.gov/earthquakes/search, accessed 02/15/15.

[34]

S. D. Davis, C. Frohlich, Did (or will) fluid injection cause earthquakes? - criteria for a rational assessment, Seis. Res. Letters, 64, p. 207–224 (1993).

[35]

A. Frankel, Mapping seismic hazard in the central and eastern United States, Seis. Res. Letters, 66, p. 8-21 (1995).

[36]

K. M. Keranen, M. Weingarten, G. A. Abers, B. A. Bekins, S. Ge, Sharp increase in central Oklahoma seismicity since 2008 induced by massive wastewater injection, Science, 345, 448-451 (2014).

111

[37]

Well shut-in occurs when injection operations are either temporarily or permanently ceased, but the tubing and casing of the well remain in place for possible future reactivation of injection operations.

[38]

Well completion is the process to ready for injection. This process includes installing tubing used to inject fluid, perforating the portion of the well in the injection zone and casing the well to ensure no injection fluids leakage.

[39]

K. E. Murray, A. A. Holland, Inventory of Class II Underground Injection Control Volumes in the Midcontinent, Shale Shaker, 65, 98-106 (2014).

[40]

C. Frohlich, J. I. Walter, J. F. W. Gale, Analysis of Transportable Array (USArray) data shows earthquakes are scarce near injection wells in the Williston Basin, 20082011, Seis. Res. Letters, 86, 1-8 (2015).

[41]

A. McGarr, Seismic moments and volume changes, J. Geophys. Res. 81, 1487-1495 (1976).

[42]

A. McGarr, Maximum magnitude earthquakes induced by fluid injection, J. Geophys. Res. 119, 1008–1019 (2014).

[43]

L. V Block, C. K. Wood, W. L. Yeck, V. M. King, The 24 January 2013 ML4.4 earthquake near Paradox, Colorado, and its relation to deep well injection, Seis. Res. Letters 85, 609–624 (2014).

[44]

Y. Zhang et al., Hydrogeologic controls on induced seismicity in crystalline basement rocks due to fluid injection into basal reservoirs, Groundwater, 51, 525–538 (2013).

[45]

The volume conversion from the oil industry standard of barrels to the metric standard of meters cubed is ~6.29 barrels per meter cubed assuming a 42-gallon oil barrel.

[46]

B. Efron, R. J. Tibshirani, An introduction to the bootstrap (CRC press, 1994).

112

[47]

Information on materials and methods is available on Science Online.

[48]

J. Rutqvist et al., The northwest Geysers EGS demonstration project, California: prestimulation modeling and interpretation of the stimulation, Mathematical Geosciences. 47, 3-26 (2015).

[49]

W. D. Mooney, M. K. Kaban, The North American upper mantle: density, composition, and evolution, J. Geophys. Res. 115, B12424 (2010).

[50]

V. M. King, L. Block, W. L. Yeck, C. Wood, and S. A. Derouin, Geological structure of the Paradox Valley Region, Colorado, and relationship to seismicity induced by deep well injection, J. Geophys. Res. 119, 4955-4978 (2014).

[51]

F. J. Anderson, Depth to precambrian basement rock in North Dakota, North Dakota Geological Survey, 85, (2009).

[52]

M. E. Meremonte et al., Investigation of an earthquake swarm near Trinidad, Colorado, August-October 2001, U.S. Geological Survey Open-File Report 02-0073 (2002).

[53]

C. Nicholson, R. L. Wesson, Triggered earthquakes and deep well activities, PAGEOPH, 139, 561-578 (1992).

[54]

W. Gan and C. Frohlich, Gas injection may have triggered earthquakes in the Cogdell oil field, Texas, PNAS, 110, 18786-18791 (2013).

[55]

J. Pursley, S. L. Bilek, and C. J. Ruhl, Earthquake catalogs for New Mexico and bordering areas: 2005-2009, New Mexico Geology, 35, 3-12 (2013).

[56]

R. T. Cox, Possible triggering of earthquakes by underground waste disposal in the El Dorado, Arkansas area, Seis. Res. Letters, 62, 113-122 (1991).

113

[57]

J. Gomberg, L. Wolf, Possible cause for an improbable earthquake: the 1997 Mw4.9 southern Alabama earthquake and hydrocarbon recovery, Geology, 27, 367-370 (1999).

[58]

L. Brown, L. Jensen, J. Oliver, S. Kaufman, D. Steiner, Rift structure beneath the Michigan Basin from COCORP profiling, Geology, 10, 645-649 (1982).

[59]

Utsu, T., A statistical study on the occurrence of aftershocks, Geophys. Mag., 30, 521– 605 (1961).

[60]

Utsu, T., A method for determining the value of b in a formula log n = a – bM showing the magnitude frequency for earthquakes: Geophysical Bulletin of Hokkaido University, 13, 99–103 (1965).

[61]

Felzer, K. R., T. W. Becker, R. E. Abercrombie, G. Ekstršm, and J. R. Rice, Triggering of the 1999 Mw 7.1 Hector Mine earthquake by aftershocks of the 1992 Mw 7.3 Landers earthquake, J. Geophys. Res., 107, 2190 (2002).

[62]

Llenos, A. L., Michael, A. J., Modeling earthquake rate changes in Oklahoma and Arkansas: Possible signatures of induced seismicity, Bull. Seismol. Soc. Am. 103, 2850-2861 (2013).

114

4.A Appendix 4.A1 Injection Data Quality Control and Data Sources Injection data quality control was applied to the monthly injection rate and monthly injection pressure data for Oklahoma, New Mexico, Colorado and Arkansas. The corrected data were used in the statistical comparison of associated wells by maximum monthly injection rate, cumulative injected volume and injection pressure (Figs. 4.10-4.12; Table 4.A1.1). The most common errors noted in the recorded data were simple keying errors in the entry of the monthly injection rate. Class II disposal wells are typically drilled by the operator and permitted by the state regulatory body with a priori knowledge of the expected monthly injection rate in a given area. Therefore, a given well will typically operate at similar rates over its lifetime. If there are changes in injection rate they typically occur gradually over several months to years. To account for keying errors, we expect that when a well is operational (i.e. injection rate > 0 bbl/month) each monthly injection rate should fall within one standard deviation of the expected monthly injection rate. If an observed rate is several standard deviations from the expected monthly rate, the specific monthly injection rate is almost always a keying error in which digits were accidentally added to the value or digits were not entered in the correct order. These errors were present in each state studied and were most often evident in the highest monthly injection rates (i.e. millions of bbl/month). To correct these errors, we calculated a simple percent error for each monthly injection rate in a given well's record: #$%&$' (%%)% = where

*+

*+

−

, -.

*+

is the observed monthly injection rate at month i and year j and

, -.

is the

observed mean monthly injection rate over the well's lifetime. Errors greater than +/- 71% were manually reviewed to ensure the given well was not experiencing an uncharacteristic increase or 115

Table 4.A1.1 State-by-state list of injection well database sources and attributes. State

Database Source Location

How were data acquired?

What data acquired?

Alabama

http://www.gsa.state.al.us/o gb/db_main.html

Web Download Available

API, Well Type, Authorization Date, Latitude, Longitude, Well Status

Arkansas

http://www.aogc.state.ar.us/ JDesignerPro/JDPArkansas/ AR_Welcome.html


Colorado

http://cogcc.state.co.us/


Florida

http://www.dep.state.fl.us/ Water/mines/oil_gas/permit _data.htm


Illinois

http://maps.isgs.illinois.edu/ iloil/

Manual web download using GIS server No online download available; state contact FOIA Request Needed. Only 7 wells.

Indiana

Closed Database

Iowa

EPA has Primacy. Data not available on web.

Kansas

http://www.kgs.ku.edu/PRS/ petroDB.html


Kentucky

http://kgs.uky.edu/kgsweb/ DataSearching/oilsearch.asp


Louisiana

http://sonriswww.dnr.state.la.us/gis/ags web/IE/JSViewer/index.htm l?TemplateID=181

Manual web download using GIS server.

Michigan

http://www.michigan.gov/de q/0,4561,7-1353311_4111_4231-214727-,00.html

Mississippi

http://gis.ogb.state.ms.us/M SOGBOnline/

Web Download Available.


API, Water Injection Permit, County Name, Well Name, Well Type, Latitude, Longitude, Depth To Be Drilled, Perforations Measured Depth, Well Status, Monthly Injection Volume, Monthly Injection Pressure API, Formation, Monthly Injection Volume, Monthly Injection Pressure, Well Status Date, Spud Date, Total Depth, Measured Depth, Perforations Top and Bottom, UTM Northing, UTM Easting, Latitude, Longitude Permit Number, Well Type, Well Status, Completion Date, Plug Date, Total Depth, Latitude, Longitude API, Permit Number, Well Status, Latitude, Longitude, Injection Formation, Company Name, Elevation, Completion Date, Total Depth Well Type, Completion Date, Max Pressure, Average Monthly Injection, Latitude, Longitude None API, Kansas ID, Lease, Well, Field, Latitude, Longitude, Operator, Elevation, Depth, Permit Date, Spud Date, Completion Date, Plug Date, Well Type, Well Status EPA_ID, Kentucky_ID, Company Name, Well Name, Well Class, Well Type, Well Status, Total Depth, Elevation, Injection Formation, Injection Top, Injection Bottom, Latitude, Longitude API, Well Serial, Well Name, Well Status, Well Description, Latitude, Longitude, Perforation Depths, Measured Depths API, Well Name, Company Name, Permit Number, Well Type, Well Status, Latitude, Longitude, Monthly Injection Volume, Monthly Injection Pressure, Formation Tops. API, Well Type, Well Status, Well Name, Operator, Latitude, Longitude.

116

How were data acquired?

State


Missouri

http://www.dnr.mo.gov/geol ogy/geosrv/ogc/index.html# permitted


Montana

http://bogc.dnrc.mt.gov/We bApps/DataMiner/


Nebraska

http://www.nogcc.ne.gov/N OGCCPublications.aspx


New Mexico

https://wwwapps.emnrd.stat e.nm.us/ocd/ocdpermitting// Data/Wells.aspx


New York

http://www.dec.ny.gov/cfm x/extapps/GasOil/search/wel ls/index.cfm

North Dakota

https://www.dmr.nd.gov/oil gas/

Ohio

http://oilandgas.ohiodnr.gov /industry/undergroundinjection-control http://www.occpermit.com/ WellBrowse/

Oklahoma



State contact

What data acquired? API, Well Type, Well Status, Company Name, Approval Date, Spud Date, Completion Date, Latitude, Longitude, Elevation, Total Depth, Perforation Depths, Formations. API, Company Name, Well Number, Well Status, Well Type, UIC Permit, Application Date, Effective Date, Max Pressure, Latitude, Longitude, Completion Date, Depth to be Drilled, Monthly Injection Volumes, Monthly Injection Pressure. API, County, Well Type, Well Status, Formation, Top Depth, Spud Date, Completion Date, Plug Date, Depth to be Drilled, Latitude, Longitude, Elevation API, Well Name, Well Number, Well Type, Well Status, Latitude, Longitude, Total Depth, Spud Date, Plug Date, Monthly Injection Volume, Monthly Injection Pressure API, Formation Tops, Well Name, Well Type, Well Status, Status Date, Permit App Date, Permit Issue Date, Spud Date, Total Depth Date, Completion Date, Plug Date, Latitude, Longitude, True Vertical Depth, Measured Depth File Number, Well Name, Well Type, Well Status, Well Status Date, UIC Status, Date of First Injection, Last Report Date, Last PPSI, Last Monthly Injection, Cumulative Water Barrels, Max Allowed Pressure, Latitude, Longitude API, Well Type, Latitude, Longitude, Deepest Formation, Total Depth, First Injection, Maximum Pressure, Permit Date


API, Lease Name, Packer Depth, Latitude, Longitude, Monthly Injection Volume, Monthly Injection Pressure API, Facility ID, Facility Name, Injection Formation, Surface Injection Pressure Average, Average Monthly Injection Rate, Latitude, Longitude API, Facility ID, Well Name, Well Type, Well Status, Average Daily Injection rate, Injection Pressure Average, Injection Pressure Max, Latitude, Longitude, Permit Date, Completion Date, Plug Date

http://imaging.occeweb.com

Pennsylvania

Closed Database

Only SWD data available through EPA Contact

South Dakota

http://www.sdgs.usd.edu/sd oil/oilgas_databases.html


Tennessee

EPA Authority. Called but never received a response from district head.

FOIA Request Needed. Only 16 wells.

None

117

State


How were data acquired? Web Download Available.

Texas

http://gis2.rrc.state.tx.us/pub lic/startit.htm

Virginia

Only about a dozen wells. All in western Virginia.

EPA contact

West Virginia

https://apps.dep.wv.gov/oog /wellsearch_new.cfm


Wyoming

http://wogcc.state.wy.us/

State contact

What data acquired? API, Latitude, Longitude, Original Authority Date, Permit Cancel Date, Plug Date, Injection Type, Max Liquid Pressure, Max Liquid Volume, Top Injection Depth, Bottom Injection Depth API, Facility ID, Facility Name, Injection Formation, Surface Injection Pressure Average, Average Monthly Injection Rate, Latitude, Longitude API, Permit Type, Permit Issue Date, Operator, Formation, Well Status, Well Type, Latitude, Longitude API, Latitude, Longitude, First Injection Date, Monthly Injection Volumes, Monthly Injection Rate, Total Depth

Table 4.A1.1. (continued). State-by-state list of injection well database sources and attributes.

118

decrease in injection rate for a period of time greater than the month in question. If the reviewed monthly injection rate was considered a keying error, the rate was changed to the mean monthly injection rate of the well's record. Fewer than 1% of all injection rates were changed according to this protocol. Another source of potential bias are that injection rates less than 500 bbl/month were typically reported in 100 barrel increments (i.e. 100, 200 or 300 bbl/month). This is potentially an operational reporting bias and was not corrected in the dataset.

4.A2. Calculating maximum monthly injection rate & cumulative injected volume Maximum monthly injection rate and cumulative injected volume were calculated differently depending on whether a given injection well was associated or non-associated with an earthquake. If an injection well was not associated with an earthquake, the maximum monthly injection rate was determined from the maximum rate from the well's entire injection history. Similarly, the cumulative injected volume for non-associated wells was calculated as the sum of the well's entire injection history. However, if the well was associated with an earthquake, the maximum monthly injection rate was determined from only the portion of the injection history prior to the date of the associated earthquake. The cumulative injected volume of associated wells was similarly calculated as the sum of the injection history prior to the date of the associated earthquake. If a well was associated with multiple earthquakes, the most recent earthquake was used as the prior date for calculation of injection metrics. Calculating the two injection parameters differently depending on a well's association with an earthquake is a function of the potential causal relationship between the given well and earthquake. The spatiotemporal filter identifies wells that are associated with a given earthquake.

119

However, a causal relationship between well operation and earthquake occurrence can be due only to the portion of the injection history prior to the associated earthquake date. Injection after the associated earthquake date can only be causally related to subsequent associated earthquakes. Therefore, the maximum monthly injection rate and cumulative injected volume for associated well's are only calculated for the portion of well's history prior to the associated earthquake (Fig. 4.10).

4.A3.Statistical methods to estimate confidence intervals 4.A3.1 Bootstrap resampling to estimate confidence intervals To assess whether the association between operational parameters and earthquake occurrence is a random process, we estimate confidence intervals for the expected percent of wells associated in any given bin of the respective distributions using a bootstrap resampling method (34). For each of the bins in the distribution we perform the bootstrap resampling protocol: 1.) Create a set of wells equal to all of the wells in the entire distribution. 2.) Randomly sample from that set, with replacement, a number of wells equal to the number of wells in the particular bin. 3.) Count how many of the random samples in step 2 are associated with earthquakes. 4.) Repeat steps 2 & 3 10,000 times. 5.) Sort the number of associated counts from smallest to largest. 6.) The 500th and 9,500th largest number of counts are the 5% and 95% confidence for that bin.

120

After the protocol is run for each bin in the histogram distribution we generate a 5% and 95% confidence bound from the 5% and 95% confidence in each bin (Fig. 4.10-4.14).

4.A3.2 Statistical significance test using Epidemic-Type Aftershock Sequence model The Epidemic-Type Aftershock Sequence (ETAS) model can be used to reproduce an observed seismicity catalog with the appropriate levels of spatial-temporal clustering. We aim to use multiple realizations of ETAS model to account for spatial-temporal clustering of earthquakes and thus generate more realistic confidence intervals for the injection parameter analysis of maximum injection rate, cumulative injected volume and maximum injection pressure. Ultimately, we did not achieve a statistical model which significantly improved the error estimation provided by the bootstrap resampling model. The ETAS model describes the seismicity rate in a given area as the sum of the background rate of independent events plus the number of aftershocks triggered by each event (47). Equation 1 describes the ETAS model formulation in the time domain:

/( ) = 0 + 1

8: : ;

10-(4 546) ( − + &)7

where / is the earthquake occurrence rate, 0 is the background seismicity rate,

(1)

is the

occurrence time of event i with magnitude < , K is the aftershock productivity, a ensures that the largest aftershock is on average 1.2-magnitude units less than the mainshock, -? ~10+4 , where >-? is the number of aftershocks, b is the Gutenberg-Richter parameter, and M is mainshock magnitude. The Gutenberg-Richter parameter was estimated for the observed catalog using the maximum likelihood following Utsu (1965): A=

ln(10) ∑HIJ(
− 300,000 bbl/month). Here we use MODFLOW to calculate injection-induced fluid-pressure changes relative to ambient reservoir conditions. The modeled reservoir includes the Dakota Sandstone, a fluvial and conglomeratic sandstone which is the primary injection target; the overlying shale confining units; the underlying sequence of Mid-Cretaceous to Pennsylvanian sediments; and crystalline basement. We compare the spatial and temporal evolution of fluid pressure to the migration and timing of earthquake hypocenters. The results show several plausible hydrogeologic models fit the available injection, geologic and hydrogeologic data, some of which produce fluid-pressure changes sufficient to trigger the two distinct sequences of seismicity in 2001 and 2011.

129

5.1 Introduction In this chapter, I focus on a remarkable seismic rate increase in the Raton Basin of southern Colorado and northern New Mexico (1-3). This seismic rate increase was punctuated by the M 5.3 August 23rd, 2011 earthquake, which caused structural damage in the nearby town of Trinidad, Colorado. Recent studies concluded these earthquakes, including the M 5.3 mainshock, were induced and attributed their cause to deep wastewater injection wells in the region (2-3). The studies concluded the seismicity was induced by relying on three lines of evidence: (1) the seismic rate change was very unlikely to be caused by random variations in the natural rate of earthquakes; (2) the vast majority of the seismicity lies within 5 km spatially of wastewater injection wells; and (3) proximal wastewater injection wells to the seismicity were high-volume and high-rate, exceeding the rates observed in previous case studies of induced seismicity. These facts provide strong evidence for a connection between wastewater injection and earthquakes in the Raton Basin. However, the Raton Basin case of induced seismicity also differs from classical cases of induced seismicity. While spatially near injection wells in plan view, the vast majority of recent seismicity occurred at depths between 0.5 - 7 km below the dominant injection interval. In classic case studies of induced seismicity, at least a portion of the seismic activity occurs at injection depths, because this is where fluid-pressure changes are expected to be largest from injection. In addition, the sedimentary basin stratigraphy and compaction often provides hydrogeologic barriers to vertical migration of fluid pressure. The goal of the present study is to determine whether or not realistic models of fluidpressure change from injection wells support the induced hypothesis. The vertical migration of fluid-pressure is of specific interest in this case study. Thus, I focus on the vertical component of heterogeneity between the sedimentary and basement formations. I test a range of hydrogeologic models to find the best fit to the available injection, geologic and hydrogeologic data. The results 130

show several plausible models which fit the available data, some of which produce fluid-pressure changes from injection wells large enough to induce earthquakes at the observed depths. Fluidpressure changes are broadly coincident in time with the dominant seismic sequences in the region. These models add to the existing body of the evidence supporting an induced origin of the recent seismicity in the Raton Basin.

5.2 Earthquakes and Injection Wells Overview Historical seismicity in the Raton Basin region is limited (Fig. 5.1). However, there were earthquakes in the Raton Basin prior to the onset of injection activities. Nine earthquakes, only one above M4.0, occurred within the study region from 1970 to 1993. These earthquakes do not delineate a defined fault feature, but instead are spread throughout the study region. The largest earthquake magnitude during this time period was a M4.3. While the seismic instrument coverage varied over this time period, Rubinstein et al. [2014] documented the magnitude of completeness for the region to be M3.8.

5.2.1 Basin-Scale Injection Overview The onset of wastewater injection activities coincided with the development of coal-bed methane production by several energy companies in 1994. Coal-bed methane production in the Raton Basin is accomplished through a dewatering procedure, where producing formations are pumped and the methane gas comes out of solution during the pumping associated depressurization. In addition to methane, this process yields large volumes of wastewater, or water which cannot be disposed of on the surface. While coal-bed methane production and

131

Figure 5.1. Maps of seismicity and injection rates in the Raton Basin for different time periods ranging from 1966 to 2011. (1966-1993) Historical seismicity in the Raton Basin. (1994-2001) Seismicity since the onset of injection in 1994, including the 2001 earthquake sequence. (20022006) Seismicity and injection since the onset of injection in the New Mexico portion of the basin. (2007-2011) Seismicity and injection prior to and during the 2011 earthquake sequence. Relocated earthquake data are from Rubinstein et al. [2014]. 132

wastewater injection began in Colorado in 1994, production and injection volumes did not substantially increase until the year 2001 (Fig. 5.2). In the year 2001, the annual volume of injected fluid doubled from ~6 million barrels in year 2000 to ~12 million barrels in year 2001 (Fig. 5.2). Between 2001 and 2011 injection volumes increased in most years to a yearly total of over 30 million barrels in 2011. As of March 2012, 28 injection wells were operating in the basin and the cumulative injected volume into the basin totaled more than 262 million barrels. For some perspective, the 262 million barrels injected into the Raton Basin would cover an area of more than 33,000 acres, roughly double the size of the city of Boulder, Colorado, with water a foot deep.

5.2.2 Earthquake Sequences The first portion of the seismic rate increase in the Raton Basin began in 2001 (1). Earthquake relocations performed by Rubinstein et al. [2014] show the 2001 earthquake sequence occurred within a few kilometers of newly active, high- rate injection well operating at rates greater than 400,000 barrels per month (Fig. 5.1; Fig.5.3). The Wild Boar well, operating at depths of ~600 masl, was the closest well to the 2001 sequence and became operational in April 2000. The sequence highlighted a previously unknown, steeply dipping normal fault trending NE-SW, which produced a mainshock earthquake magnitude of M4.5. Most of the earthquakes occurred 1- 1.5 km below the injection interval. Some earthquakes were observed in the sedimentary strata, but few occurred at the injection depth. Meremonte et al. [2002] studied this earthquake sequence to determine whether or not the recent installation of a high-rate well nearby indicated the sequence was induced. Ultimately, the authors were equivocal in regards to a possible induced origin. Their reasoning stemmed from a

133

Injection Volume (million barrels) Number of Earthquakes (M3.0+)

1973 1974 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012

A

1973 1974 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012

B 25

20

15

10

5

0

35

30

25

20

15

10

5

0

Figure 5.2. Histograms from 1973- 2012 of (A) yearly Raton Basin seismicity greater than M3.0 and (B) yearly waste fluid injected volume summed over all injection wells. Earthquake are from the ANSS catalog (4). Well data are from the COGCC and NMOCD (5-6).

134

Figure 5.3. Stratigraphic cross section from southwest to northeast along a transect which follows the strike of the 2001 earthquake sequence. Earthquake hypocenters are denoted by red and hollow dots. The filled red dots are earthquakes which have magnitudes calculated. Hollow dots are earthquakes with no magnitude calculated. The largest filled red dot is the hypocenter of the September 5th, 2001 M4.5 earthquake. Earthquake data are relocated hypocenters from Rubinstein et al. [2014]. Well data are from the COGCC and NMOCD (5-6). Stratigraphic contacts were interpolated from well stratigraphic data. The dashed lines in the stratigraphy indicate an interpreted stratigraphic contact where well data were lacking.

135

lack of seismometer coverage in the basin prior to the sequence and thus their inability to clearly distinguish this sequence from the historical seismicity in the basin. However, they added a caveat to their conclusion that if earthquakes continued at a similar or greater rate to the 2001 sequence, then the likelihood of induced seismicity was high. This sequence was followed by nearly a decade of elevated seismic activity with 41 M3.0+ earthquakes, seven of which were M3.8+ earthquakes, from 2002 through 2010 (Fig. 5.1). This period included the second-largest earthquake in the Raton Basin over the 45-year study period: a M5.0 earthquake occurring on August 10th, 2005. Most of the earthquakes from 2002-2010, including the M5.0 earthquake, occurred in the New Mexico portion of the basin. This seismic activity was broadly spatially co-located with the activation of several injection wells in this portion of the basin. Due to a lack of seismometer spatial coverage, earthquake locations during this time period may have had location uncertainties of greater than 10 km (2). However, a new seismic dataset indicates the M5.0 earthquake and many subsequent quakes align on a steeply dipping, narrow fault zone (7). In August 2011, another earthquake sequence ruptured along the same strike as the previous NE-SW trend of the 2001 earthquake sequence (Fig. 5.4). The focal mechanism of the M5.3 mainshock showed a steep, eastwardly dipping normal fault structure which was consistent with focal mechanisms from the 2001 sequence. The 2011 sequence most likely nucleated near the southwest end of the structure and subsequent aftershocks propagated along a trend striking to the northeast. Seismicity ranged in depth between 1.5 - 6 km below the injection interval with very little seismicity occurring in the sedimentary strata. Figure 5.1 shows the 2011 sequence is bounded to the NE and SW by some of the highest-rate injection wells operating in the basin. The Wild Boar well bounds the sequence to

136

Figure 5.4. Stratigraphic cross section from southwest to northeast along a transect which follows the strike of the 2011 earthquake sequence. Earthquake hypocenters are denoted by red and hollow dots. The filled red dots are earthquakes which have magnitudes calculated. Hollow dots are earthquakes with no magnitude calculated. The largest filled red dot is the hypocenter of the August 23rd, 2015 earthquake. Earthquake data are relocated hypocenters from Rubinstein et al. [2014]. Well data are from the COGCC and NMOCD (5-6). Stratigraphic contacts were interpolated from well stratigraphic data. The dashed lines in the stratigraphy indicate an interpreted stratigraphic contact where well data were lacking.

137

the northeast while wells VPR C 14, VPR C 39 and VPR A 182 bound the sequence to the southwest. All of these wells injected at maximum rates greater than 300,000 barrels per month from 2006-2011. While only one M4.0+ earthquake occurred over the 30-year period from 1970 to 2000, ten M4.0+ earthquakes occurred from 2001 through 2011. The number of earthquakes in any given year may differ due to natural variations in seismicity. Using techniques in the field of statistical seismology, Rubinstein et al. [2014] calculated the chance of this seismic rate increase being natural at less than 3%.

5.3 Geologic Background 5.3.1 Geologic Setting The Raton Basin is an asymmetric sedimentary basin which straddles the Colorado-New Mexico border and lies just to the east of the Sangre de Cristo Mountains (Fig. 5.5). The basin, as defined by outcrop of the Vermejo and Raton formations, is roughly 130 km long and 60 km wide at its maximum points. The Vermejo and Raton formations range from 200-800 m in depth in the basin proper. The deeper strata of the Raton Basin outcrop tens of kilometers to the east of the outline defined in Figure 5.5. The deepest stratigraphy in the basin indicates deposition began during the Pennsylvanian, but the major structural features of the basin were developed during the Laramide orogeny. During the Laramide, roughly east-west compression resulted in several of the crystalline basement uplifts which define the current boundaries of the basin (8). Along the basin's western edge are the Sangre De Cristo Mountains, which are composed

138

Figure 5.5. A map of the Raton Basin boundaries defined by uplifts in the Precambrian crystalline basement. The bolded dark polygon represents the outcrop of the Vermejo Formation and broadly defines the surface expression of the basin, while deeper strata of the basin outcrop to the east. Blue squares depict 29 injection wells which operated at one time or another in the Raton Basin during the period from 1994 - 2011.

139

of several east-verging thrust faults which deform and completely overturn many of the sedimentary units of the deep Raton Basin. The basin's northern and eastern boundaries are defined by Laramide-aged deformation known as the Apishapa Uplift and Sierra Grande Uplift, respectively. The southern boundary is defined by the Cimarron Arch, which separates the Raton Basin from the shallower Las Vegas Basin in north-central New Mexico (9). The Raton Basin's west margin is more than 11,000 feet above sea level. From west to east, over a distance of roughly 100 km, the basin's topography drops to 4000 feet above sea level. In addition to the topographic relief in the basin, most of the sedimentary section has also experienced significant deformation. Lorenz et al. [2004] documented the extensive and pervasive fracturing of most of the significant strata in the deeper stratigraphic section of the Raton Basin (10). These fractures most likely provide the majority of the permeability in the deep sedimentary formations of the basin. Pervasive fracturing of the sedimentary strata has produced enough permeability for economic quantities of oil and gas to be extracted from the Raton Basin (11). While many of the sedimentary formations are highly fractured, little geologic evidence exists at the surface for faults which can produce felt seismic events. The only published study documenting faults lying within the Raton Basin does not document any seismogenesis associated with these faults (12). These faults are also not documented in the U.S. Geological Survey's Quaternary Fault and Fold Database, indicating they are likely inactive (2).

5.3.2 Stratigraphy Geologic units in the Raton Basin range from Precambrian to the Holocene age (13). Only a subset of the geologic units in the Raton Basin are of interest in this study as injection wells operate at several kilometers depth within the basin. The units of interest for our study lie 140

hundreds of meters to several kilometers below the surface and are Mid-Cretaceous to Precambrian in age (Table 5.1). It is important to note that most of the sedimentary formations in the basin actually lie above sea level despite their depth due to the fact the average elevation within the basin is 2200 meters above sea level. I compiled well stratigraphic data from the New Mexico Oil Conservation Division (NMOCD), the Colorado Oil and Gas Corporation Commission (COGCC), the New Mexico Geological Survey (NMGS) and the Colorado Geological Survey (CGS). These data were used to create maps of the stratigraphic tops of each formation included in the model domain. The surfaces of each stratigraphic unit were interpolated using a natural neighborhood algorithm (14). The sedimentary units used by injection wells in the Raton Basin are highlighted in yellow in Table 5.1. The shallowest unit of interest in this study consists of the Mid-Cretaceous Pierre Shale formation (Fig. 5.6). The Pierre Shale consists of a dark-gray, fine-grained mudstone with small interbedded layers of limestone and sandstone. The average thickness in the basin is 600 meters. The Niobrara and Benton formations are also composed of mostly light- to dark-gray shales with one sandstone member, the Codell, in the Benton formation. These formations combined make up another 400 meters of stratigraphy on average in the basin. The variation in depth of the Pierre and underlying Niobrara and Benton formations is negligible at the basin scale with a structural dip less of less than 1% over the entire model domain. The primary stratigraphic horizon for injection wells in the Raton Basin is the Early Cretaceous-age Dakota Formation, but underlying sedimentary strata are also used for disposal. Comprised of two members, the Dakota Sandstone and Purgatoire Sandstone, the formation contains fluvial and conglomeratic sandstone facies. While only 50-100 meters thick throughout

141

Raton Basin Members

Raton Basin Formations

Predominant Rock Type

Deposition Age

Pierre

Pierre

Shale

Mid-Cretaceous

Formation Top Average Depth (meters below surface) 500 1100

300

1400

100

1500

68

Niobrara

Niobrara

Shale

Mid-Cretaceous

Fort Hays

Niobrara

Shale

Mid-Cretaceous

Carlile

Benton

Shale

Early Cretaceous

Codell

Benton

Sandstone

Early Cretaceous

Greenhorn

Benton

Shale

Early Cretaceous

Graneros

Benton

Shale

Early Cretaceous

Dakota

Dakota

Sandstone

Early Cretaceous

Puragtoire

Dakota

Early Cretaceous

Morrison

Morrison

Entrada

Entrada

Sandstone Shaley Sandstone Sandstone

Dockum

Dockum

Glorieta Sangre De Cristo Basement

Formation Average Thickness (m) 600

Late Jurassic

109

Early Jurassic

36

Shale

Triassic

125

Dockum

Sandstone

Sangre De Cristo

Sandstone

Permian PermianPennsylvanian

Crystalline Basement

Mafic Gneiss, Granite, Granite Gneiss

Precambrian

1860

1100

2750

N/A

Table 5.1. Stratigraphy of interest for the present study includes formations dating from Precambrian to Mid-Cretaceous age. Injection wells operate in the sedimentary formations highlighted in yellow. Well stratigraphic data from the COGCC and NMOCD (5-6).

142

Figure 5.6. Elevation of the top of the Pierre Shale formation in the Raton Basin in meters above sea level. The formation is deepest in the central and north-central portions of the basin. Control points of known formation elevation are noted by the red dots. Data come from the COGCC and NMOCD (5-6).

143

the majority of the basin, this formation is the key reservoir target for operators in the region. The formation rises from ~400 meters above sea level in the north-central portion of the basin to ~700 meters above sea level to the east (Fig. 5.7). The Jurassic-age sediments of the Raton Basin are the Morrison and Entrada formations. These formations are predominantly composed of limestone, shaley sandstones and sandstones. Their combined thicknesses averages roughly ~130 meters. The formations are laterally extensive as they were intersected by most wells at depth over a large portion of the basin. The Permian and Triassic-age sediments of the Dockum Formation are composed of one shale member and one sandstone member (11). The formation thickness averages 125 meters in the basin. The formation was not encountered in the wells in the northeast portion of the basin, which should have penetrated the formation given its depth in other regions of the basin. The formation is roughly flat in the central portion of the basin at an elevation of 300-400 meters above sea level. The Pennsylvanian strata intersected in the Raton Basin consist of the Sangre De Cristo Formation. This formation is mainly comprised of conglomeratic sandstones and arkosic sands (15). The protolith of this formation is the weathering of the ancestral Rocky Mountains in which great alluvial fans piled up along the mountain range's front. In the basin deep, the formation has an average thickness of 1,100 m and directly overlies the crystalline basement rock. Precambrian crystalline basement in the Raton Basin is composed of mafic gneiss, metaquartzite, granite and granitic gneiss (11). Using NMGS and CGS basement maps, I estimate the top elevation of the crystalline basement in the Raton Basin (Fig. 5.8; 16-17). The deepest portion of the basin is in the central and north-central portions of the basin in Colorado. The basement rock shallows rapidly in the New Mexico portion of the basin to

144

Figure 5.7. Elevation of the top of the Dakota Formation in the Raton Basin in meters above sea level. The formation is deepest in the central and north-central portions of the basin. Control points of known formation elevation are noted by the red dots. Data come from the COGCC and NMOCD (5-6).

145

Figure 5.8. Precambrian crystalline basement elevation in the Raton Basin interpolated from well data and structural contours in Suleiman and Keller [1985] and Hemborg [1996] (16-17). The deepest portion of the basin occurs in Colorado portion of the basin and extends out to the northeast of the surface of the expression of the basin proper. The dashed black line delineates the transect used to create the cross section profile of stratigraphy in Figure 5.7.

146

elevations of 3000 feet above sea level. The reconciliation of the two basement datasets, which did not agree along the Colorado-New Mexico border is a source of error in the model. Using data from all stratigraphic tops from well data, I constructed a geologically constrained cross section of sedimentary and basement depths in the Raton Basin (Fig. 5.9). This cross section runs along the zone of seismicity active during the 2001 and 2011 earthquake sequences. The structural dip of the sedimentary formations is very small at the basin-scale and they are essentially flat-lying.

5.4 Regional Hydrogeology 5.4.1 Hydrostratigraphy A hydrostratigraphic unit is a formation or group of formations which exhibit similar hydrogeologic properties. Rock formations often help delineate hydrostratigraphic units due their heterogeneity in rock properties. However, at the scale of a regional groundwater flow model, hydrostratigraphic units may combine several formations of similar lithology as their hydrogeologic properties are similar at that scale. In the Raton Basin, many of the rock strata are of similar lithology, yielding a less complex hydrostratigraphy than geologic data would suggest. Geldon [1989] defined much of the hydrostratigraphy of the Raton Basin. One of the important facets of the hydrostratigraphy in the Raton Basin is the thick sequence of shale formations in the middle depths of the basin. These shale formations, described above, include the Pierre Shale, the Niobrara Formation and the individual members which make up the larger Benton Formation (Table 5.1). These formations combine to provide more than 1 kilometer of low-permeability hydrogeologic media.

147

Figure 5.9. Stratigraphic cross section from southwest to northeast along a transect which follows the strike of the 2001 and 2011 earthquake sequences. The cross section shows the Dakota Formation overlain by roughly ~1 km of shale formations and underlain by ~1 km of sedimentary formations mostly comprised of sandstone lithology. The dashed lines indicate an interpreted stratigraphic contact where well data were lacking.

148

They also serve to confine the underlying half kilometer of sandstone and limestone formations which serve as the primary target for injection wells in the basin. The permeability of the Pierre Shale has been documented at field scale to be as large as 1.4 x 10-19 m2 with specific storage values of 1.7 x 10-5 m-1 (18). The permeability of this formation is typically on the order of 10-20 m2 to 10-21 m2. This range in permeability yields hydraulic conductivity values for water at 100 degrees C on the order of 10-13 m/s. These fluid flow properties indicate the shale formations in the Raton Basin are extremely low permeability. All of the hydrogeologic studies of the upper Cretaceous aquifers above the Pierre Shale use the formation as the deepest formation for flow in the upper portion of the Raton Basin (19-20). The underlying shale formations of the Niobrara and Benton are considered to be confining units with calibrated hydraulic conductivities of 1.5 x 10-11 m/s (21). The Dakota-Purgatoire aquifer has long been known to be a quality water supply formation outside of the Raton Basin's boundary (13). Studies of regional groundwater flow in the Dakota aquifer from nearby basins have calibrated hydraulic conductivities as large as 2.0 x 10-5 m/s with specific storage values of 1.0 x 10-7 m-1 (21). Most oil-field core data suggest the hydraulic conductivity of the formation to be on the order of 10-6 m/s (22). Calibrated hydraulic conductivities as small as 1.0 x 10-8 m/s have been observed in the Denver Basin in isolated regions (23). These values will be used as the range of hydraulic conductivities for the Dakota Formation in the Raton Basin. The Jurassic aquifer is comprised of the Morrison and Entrada sandstone formations. These formations have shown calibrated hydraulic conductivity values as high as 5.0 x 10-7 m/s in the Denver Basin (23). Oil-field core data show hydraulic conductivities ranging from 10-6 to 10-8 m/s (24). We use this range of hydraulic conductivities in our modeling study.

149

The Dockum Formation, comprised of a shale and sandstone member, does not produce economically viable quantities of water. In the Geldon (1989) model, this formation is not considered a regional aquifer. However, many other studies have shown the permeability of this formation to be consistent with the overlying Jurassic formations (23). In this framework, we consider the formation to be an aquifer on the order of its up-section counterparts. The Sangre De Cristo Formation is another regional aquifer in the Geldon (1989) hydrostratigraphic framework. Harbaugh and Davie (1964) found permeability of the Pennsylvanian-aged sediments in Kansas to be between 10-13 to 10-15 m2 in laboratory testing (25). These permeabilities would yield hydraulic conductivities between 5.0 x 10-7 m/s to 5.0 x 10-8 m/s at the reservoir temperatures found in the Raton Basin. The mid-point of this range in hydraulic conductivity falls within the calibrated range of conductivities for this formation in the Denver Basin (23). There is no hydrogeologic data available for the Precambrian crystalline basement in the Raton Basin. Thus, we employ the use of permeability-depth relationships for crystalline basement rocks used in the literature. Manning and Ingebritsen (1999) derived a permeabilitydepth relationship constrained by geothermal data and calculated fluid fluxes during metamorphism: (5.1)

log S = −14 − 3.2 log

where z is depth in kilometers and k is permeability in meters squared (26). Given an average depth of 2.75 km across the basin, the top of the crystalline basement has a permeability on the order of 4 x 10-16 m2 and a hydraulic conductivity of 1.5 x 10-8 m/s, assuming a reservoir temperature of 120 degrees C.

150

5.4.2 Regional Groundwater Flow Regional groundwater flow in the Raton Basin is topographically driven, but also greatly affected by the heterogeneity of the hydrostratigraphy. On human timescales, groundwater flow can essentially be thought of as occurring in two distinct aquifer systems, separated by the thick sequence of low-permeability shale in the middle depths of the basin (27). Here, I use drillstem test data from Nelson et al. (2013) and a natural neighbor algorithm to interpolate hydraulic head in the Dakota Formation of the Raton Basin (Fig. 5.10). The resulting potentiometric surface indicates groundwater is flowing from southeast to northwest across the basin. Hydraulic head values reach as high as 6900 feet on the western side of the basin, but these values are still well below land surface in this region. Measurements of hydraulic head in the eastern portion of the basin indicate the potentiometric surface is essentially flat over much of the basin's area at roughly 5100 feet. The east-west hydraulic continuity of the Dakota Formation has been shown to be on the order of 80-90 km (27).

5.4.3 Regional Underpressure Subnormal fluid pressure, or underpressure, was first recognized in the Raton Basin in the late 1970s (28). Underpressure is defined as ambient formation fluid pressure which is less than hydrostatic, resulting in fluid levels in the confined aquifers of the basin which do not reach the land surface. Early oil and gas exploration wells noted lost circulation zones while drilling in the Raton Basin. Underpressure in the intermountain west region is not uncommon. Both the San Juan and Denver Basins also exhibit subnormal fluid pressures. Belitz and Bredehoeft (1988) concluded that underpressured formations in the Denver Basin are a consequence of regional groundwater flow. Recharge areas are prevented from recharging the formation due to the basin structure 151

Figure 5.10. Interpolated map of hydraulic head in the Dakota Sandstone as measured by several drillstem tests throughout the basin. Control points of known hydraulic head are noted by the red dots. Data for drillstem tests are from Nelson et al. [2013].

152

reducing hydrogeologic connectivity in the basin deep. Meanwhile, shallower hydrogeologic strata to the east are still hydrogeologically connected to discharge areas. The imbalance in recharge versus discharge rates eventually results in a decrease of hydrogeologic storage. Nelson et al. (2013) interpreted the geologic structure and several drillstem tests to conclude the underpressure in the Raton Basin occurred as a result of a similar mechanism to that of the Denver Basin. We use the hydraulic head values calculated in the previous section and elevation data to estimate underpressure in the Dakota Formation as a depth-to-water calculation (Fig. 5.11): (5.2)

WX = Y − Z

where DTW is depth-to-water, Z is elevation and H is hydraulic head in feet. Underpressure is greatest along the western boundary of the basin. Elevations in this portion of the basin are high and fall moving eastward. Within the central portion of the basin, underpressure in the Dakota Sandstone ranges from 2,100 to 3,000 feet (640 - 915 meters). This level of underpressure is consistent throughout most of the central portion of the basin.

5.4.4 Geothermal Gradient & Fluid Properties The northwest portion of the Raton Basin has experienced volcanism in the geologic past dating to the mid- to late-Tertiary (29). These igneous rocks formed the current day Spanish Peaks in the northwest Raton Basin as well as many sills and dikes in that area. The highest concentration of these features is in the northwest portion of the basin. This geologic history has created an elevated geothermal gradient which persists to the present day throughout the basin proper (11). Temperature measurements in the Dakota Formation generally range between 110130 degrees C at depths below surface of ~1.5-1.7 km. (Fig. 5.12). The geothermal gradient in the stratigraphy of interest in this study is in excess of 55 degrees C per km. 153

Figure 5.11. Interpolated map of underpressure in the Dakota Sandstone as calculated from hydraulic head and elevation data in Equation 5.3. Thousands of feet of underpressure in the Dakota Sandstone is observed throughout the central portion of the basin.

154

Figure 5.12. Geothermal gradient from temperature data collected during the drilling of Rocky Mountain #7 well (modified from 11, Fig. 14). The well was drilled about 15 miles south of the CO-NM border in the central portion of the basin.

155

The fluid properties of water are affected by the elevated temperatures of the injection reservoir. At standard room temperatures, water has density of 997 kg/m3 and dynamic viscosity of 9.31x10-4 Pa-s. At reservoir temperatures in the Raton Basin of ~120 C, the density of water is reduced to 941 kg/m3 and the dynamic viscosity decreases to 2.25x10-4 Pa-s. Using the conversion from the fluid-independent material property of permeability to fluid-dependent property of hydraulic conductivity, I calculate the elevated temperatures effect on hydraulic conductivity: S=

(5.3)

where S is the permeability of the medium [L2],

[ \

is the hydraulic conductivity [L/T], 0 is the

dynamic viscosity of the fluid [M/(LT)] and ] is the specific weight of water [M L-2T-2]. Elevated temperatures in the Dakota Formation enhance the hydraulic conductivity of the fluid in the reservoir by a factor of 3.9, allowing fluid pressure to migrate more easily in the reservoir.

5.5. Injection Operations in Detail Injection operations commenced in November 1994 with the first well, Cottontail Pass, operating 15 miles north of the Colorado-New Mexico border in the north-central portion of the basin (Table 5.2). The Colorado portion of basin contained all of the injection wells in the basin from 1994 through 1999. In January 1995, two more wells had come online in the western portion of the basin, Apache Canyon 10-3 and Apache Canyon 19-10, respectively. Only one injection well was drilled, PCW well, over the next 4 years from January 1995 to January 1999. A rapid increase in the field's injection capacity began in January 1999. In the Colorado portion of the basin, 8 new wells came online over a 2-year period, tripling the injection capacity in this portion of the basin. Injection well operations also began in the New Mexico portion of

156

Average Injection Rate (bbl/month)

Maximum Injection Rate (bbl/month)

Cumulative Injected Volume (barrels)

Maximum Injection Pressure (psi) 0

Well Name

Lat.

Long.

Injection Start Date

Cottontail Pass Apache Canyon 10-3 Apache Canyon 19-10 PCW

37.22

-104.78

1994-Nov

106,389

448,374

23,405,471

37.10

-104.99

1995-Jan

23,043

96,690

4,558,989

37.07

-104.93

1995-Jan

56,497

265,405

12,316,262

37.12

-104.68

1997-Jul

124,932

327,572

23,487,303

0

0 0

VPR C 14

37.02

-104.78

1999-Sep

147,491

378,554

17,293,848

1243; 0

Sawtooth

37.20

-104.67

2000-Apr

52,016

131,967

8,062,451

0

VPR C 39

37.02

-104.78

2000-May

118,649

356,826

18,451,302

0

Wild Boar

37.13

-104.70

2000-Aug

129,776

491,058

19,596,216

0

Beardon

37.25

-104.66

2001-Jan

35,071

525,949

11,687,123

0

Cimarron

37.26

-104.93

2001-Jan

15,659

58,671

1,503,295

0

Long Canyon

37.09

-104.62

2001-Apr

78,780

166,039

11,265,533

0

La Garita

37.16

-104.80

2001-Aug

30,925

79,306

4,298,601

0

Weston

37.15

-104.86

2004-Jan

89,560

211,903

9,851,548

0

Del Agua

37.28

-104.74

2005-Jul

99,568

196,239

9,160,226

0

Hill Ranch Deep

37.09

-104.74

2005-Jul

22,394

67,081

2,060,254

0

VPR A 007

36.96

-104.83

2006-Jan*

105,639

237,672

8,605,028

0

VPR A 042

36.96

-104.83

2006-Jan*

161,161

387,704

13,125,002

0

VPR A 182

36.98

-104.80

2006-Jan*

368,267

485,190

30,126,570

0

VPR B 027

36.80

-104.94

2006-Jan*

68,937

115,768

5,633,810

0

VPR D 025

36.86

-105.02

2006-Jan*

198,863

283,307

16,276,620

0

VPR E 099

36.96

-104.90

2006-Jan*

272,187

367,588

22,336,411

0

Jarosa

37.30

-104.78

2007-May

138,444

334,640

9,691,063

0

Ferminia

37.29

-104.83

2007-Sep

61,125

146,789

4,034,277

0

VPR A 500

36.89

-104.71

2008-Jun

134,595

228,404

7,799,816

0

Southpaw

37.30

-104.73

2009-Apr

121,376

282,880

5,704,682

0

Polly

37.23

-104.70

2009-Jul

2,682

23,440

118,028

0

Lopez Canyon

37.15

-104.89

2010-Sep

11,958

33,871

358,733

64

VPR C 204

37.02

-104.83

2012-Mar

155,236

222,363

2,069,674

0

Table 5.2. Injection well operations in the Raton Basin. Wells highlighted in gray are of particular interest for in the 2001 and 2011 earthquake sequences. The maximum injection pressure column is highlighted to note that virtually all wells injected at zero wellhead pressure (gravity feed) for their entire lifetimes. Well VPR C 14 initially injected with pressure at the wellhead, but since 2005 has operated under zero wellhead pressure. Data from the NMOCD and COGCC (5-6). 157

the basin during this time period. However, injection data from New Mexico are unavailable for the period from 1999 - 2006. In January 2006, 6 wells began reporting injection rates in New Mexico meaning either a portion or all of these wells were operating in the basin prior to that date. After the tripling of the injection well count from 1999-2001, injection wells were drilled sporadically throughout the basin on both sides of the border. Excluding the New Mexico wells, most of which probably came online prior to 2006, 8 more injection wells were drilled from 2002-2011. This brought the total number of wells operating at the time of the 2011 earthquake sequence to 27 wells, all of which were operating within a 10-15 mile radius of the ColoradoNew Mexico Border. One of the important features of injection well operations in the Raton Basin is the lack of wellhead pressure required to inject into the deep basin. Zero wellhead pressure means the injection occurs with gravity providing the necessary pressure at the reservoir depth to allow flow. Part of the explanation for this phenomenon is the large amount of underpressure (~600 meters) in most areas of injection reservoir. In addition, the injection reservoir must be sufficiently permeable over large areas to allow for two decades worth of injection with reservoir pressures never reaching the surface. The lack of injection pressure needed in the Raton Basin is one of the main constraints on the hydrogeologic modeling. The deepest injection wells operate in the New Mexico portion of the basin (Table 5.3). However, injection depths are relatively uniform when compared relative to sea level as the dip of formations is relatively small at the basin scale. All of the injection wells inject fluid into the aquifer units below the Pierre-Niobrara-Benton shale groups. Some wells operate in deeper stratigraphy than the Dakota-Purgatoire aquifer, but most wells inject a portion of fluid in

158

Well Name Long Canyon PCW Wild Boar Sawtooth Hill Ranch Deep Beardon Polly La Garita Weston Jarosa Del Agua Southpaw Ferminia VPR C 204 Apache Canyon 19-10 Lopez Canyon Cottontail Pass VPR A 007 VPR C 39 VPR C 14 VPR A 182 Cimarron VPR A 500 Apache Canyon 10-3 VPR B 027 VPR D 025 VPR A 042 VPR E 099

37.09 37.12 37.13 37.2 37.09 37.25 37.23 37.16 37.15 37.3 37.28 37.3 37.29 37.02

-104.62 -104.68 -104.7 -104.67 -104.74 -104.66 -104.7 -104.8 -104.86 -104.78 -104.74 -104.73 -104.83 -104.83

1950 1970 1986 2074 2151 2296 2127 2182 2148 2159 2292 2156 2237 2244

Top Injection Elevation (masl) 822 762 729 730 697 837 696 558 468 492 559 445 403 596

37.07

-104.93

2241

362

303

1879

1938

37.15 37.22 36.96 37.02 37.02 36.98 37.26 36.89

-104.89 -104.78 -104.83 -104.78 -104.78 -104.8 -104.93 -104.71

2205 2269 2522 2380 2353 2468 2421 2429

368 370 571 407 671 626 305 351

220 283 521 373 333 320 272 258

1837 1899 1951 1973 1682 1842 2116 2078

1985 1986 2001 2007 2020 2148 2149 2171

37.1

-104.99

2637

490

458

2147

2179

36.8 36.86 36.96 36.96

-104.94 -105.02 -104.83 -104.9

2460 2606 2527 2620

510 671 464 457

220 366 254 319

1950 1935 2063 2163

2240 2240 2273 2301

Latitude Longitude

Elevation (m)

Bottom Injection Elevation (masl) 778 745 694 691 662 806 515 550 438 449 555 383 338 315

Top Injection Depth (m) 1128 1208 1257 1344 1454 1459 1431 1624 1680 1667 1733 1711 1834 1648

Bottom Injection Depth (m) 1172 1225 1292 1383 1489 1490 1612 1632 1710 1710 1737 1773 1899 1929

Table 5.3. Injection well operational depths in the Raton Basin. Wells highlighted in gray are of particular interest for in the 2001 and 2011 earthquake sequences. Data from the NMOCD and COGCC (5-6).

159

this reservoir. No wells inject fluid into the Sangre De Cristo Formation, indicating the Dockum Formation is the deepest formation for injection. The highest rate injection wells operate within a relatively small area in the central portion of the basin near the Colorado-New Mexico Border (Fig. 5.13). In 2001, the highest rate wells were the Cottontail Pass, Wild Boar, VPR C 14 and VPR C 39 wells. Both the Wild Boar and Cottontail Pass wells had operated at injection rates greater than 400,000 barrels per month by the end of 2001. Of the VPR C wells, which operate adjacent to one another, VPR C 39 was the dominant injector during this time period. By the end of 2011, wells in the New Mexico portion of the basin became the dominant injectors in the basin. Specifically, the VPR A 182 well stood out as the dominant injection well in this portion of the basin. The VPR A 182 well had an average injection rate of 368, 267 barrels per month, or more than double all but one of the other wells in the basin. Wells VPR A 042 and VPR E 099 also began injecting at high rates during the time period from 2007-2011. Cumulative injected volume in the basin is difficult to distinguish over the decade from 2001 to 2011 (Fig. 5.14). The VPR A 182 well dominated cumulative injection with more than 30,000,000 barrels injected over a 5 year time period. This was 7,000,000 barrels more than any other well in the basin, many of which operated for double and triple the operating history of the VPR A 182 well. The pattern of large cumulative volume wells looks similar to that of the pattern of high-rate injection wells in 2011. The main wells of interest in this study are the 8 injection wells along the transect highlighted in Figure 5.14 and in cross section in Figure 5.15. The transect follows the strike of the fault zone involved in the 2001 and 2011 earthquake sequences and also includes most of the

160

Figure 5.13. Maximum injection rate of wells in the Raton Basin for three time periods: 2001, 2006 and 2011. The size of the blue square scales with injection rate. The dashed line shows the transect of interest which corresponds to the 2001 and 2011 earthquake sequences.

161

Figure 5.14. Cumulative injected volume of wells in the Raton Basin for three time periods: 2001, 2006 and 2011. The size of the blue square scales with injected volume. The dashed line shows the transect of interest which corresponds to the 2001 and 2011 earthquake sequences.

162

high-rate, high-volume injection wells in the basin. The cumulative injected volume of the six high-volume wells along the transect account for more than 40% of all of the injected volume in the basin. These same six wells injected at maximum rates greater than 300,000 barrels/month during their lifetimes. Pore pressure changes are expected to be largest from the wells along this transect due to their injection rate, volume and proximity to one another.

163

Figure 5.15. Stratigraphic cross section from southwest to northeast along a transect which follows the strike of the 2001 and 2011 earthquake sequences. Injection wells of specific interest are labeled by name above the topographic surface. Injection well depths are noted by the black and red lines. The red lines denote the interval of injection for each well, respectively.

164

5.6. Hydrogeologic Model Here we calculate the fluid-pressure changes from injection in the Raton Basin over the 17-year period from November 1994 through December 2011. We utilize the groundwater modeling software, MODFLOW, a modular finite-difference code developed at the USGS. This numerical code will calculate the build-up of pore pressure associated with fluid injection in three dimensions. MODFLOW is considered one of the industry standard codes in numerical groundwater modeling (30). MODFLOW solves the groundwater flow equation in three dimensions for a fluid of constant density and dynamic viscosity in a heterogeneous and anisotropic aquifer with sources and sinks:

+

(5.4)

where

,

,

+

=

−

( ) ( −

) ( −

) ( − )

are the principal components of the hydraulic conductivity tensor [L/T] and

is the specific storage coefficient [L-1]. Pressure head can be converted to pressure by:

= ]^

(5.5)

where ^ is the pressure head [m], ] is the specific weight of water [N/m3] and

is the pressure

change [N/m2]. The construction of a finite-difference groundwater flow model first involves defining the question the model wishes to answer. In this case, the numerical model will attempt to put bounds on the extent, timing and magnitude of fluid-pressure changes from injection wells in the Raton Basin. We specifically aim to capture fluid-pressure perturbations in the vertical direction 165

to address the extent to which injection perturbs the ambient pressures in the crystalline basement.

5.6.1 Conceptual Model We aim to keep our hydrogeologic model as simple as possible in its initial stages and build hydrogeologic scenarios one step at a time. Ultimately, we aim to capture the threedimensional pressure perturbation from all of the injection in the Raton Basin. Geologic and hydrogeologic data suggest the injection reservoir and underlying sedimentary units are permeable and hydraulically connected over a large lateral extent of the basin. Therefore, the key heterogeneity explored in our model is the contrast in hydraulic conductivity between the sedimentary formations used for injection and the underlying crystalline basement. We have three main constraints on the hydraulic conductivity of the sedimentary formations in the Raton Basin:

(1) The established ranges of K for each hydrogeologic unit from laboratory, field and calibrated modeling studies. (2) The known quantity of initial underpressure at each injection well (Fig. 5.16). (3) The fact that almost all injection wells in the basin have operated at zero wellhead pressure over their entire operational lifetimes.

Any model which predicts reservoir fluid-pressure increases in excess of the initial underpressure present is considered unrealistic. Using data from Figure 5.16, models must not calculate

166

Figure 5.16. Estimated depth-to-water in each of the 8 wells along strike of the 2001 and 2011 earthquake sequences. Model-predicted fluid-pressure increases cannot exceed these values because of the lack of wellhead pressure required to inject fluid into the Dakota Formation. Potentiometric surface is interpolated from drillstem test data of Nelson et al. [2013].

167

pressure head increases greater than 732 m, 699 m, or 388 m at the VPR A 182, VPR C 14 and Wild Boar wells, respectively. Given this constraint, our modeling effort should be able to constrain the lower bound of hydraulic conductivity for the sedimentary formations using the underpressured condition and known injection rates. The difficulty in our modeling effort will be whether or not we can differentiate between mid-range and upper-range models of hydraulic conductivity.

5.6.2 Model Domain, Boundary Conditions & Grid Discretization Our model domain comprises the entire Raton Basin as defined by the crystalline basement uplifts discussed in section 5.3.1 (Fig. 5.17). The model domain is 110 km from north to south and 100 km east to west at its largest extent. No-flow boundary conditions are implemented along all of the model edges. The eastern, northern and southern boundaries are sufficiently far from injection wells that the vast majority of model scenarios will not interact with these boundaries. The western boundary, defined by the complex thrust faulting of the Sangre De Cristo mountain front, will interact with most model scenarios. The model simulates fluid-pressure changes from 500 m to 15 km depth in order to capture the vertical migration of fluid-pressure. The hydrostratigraphy described in section 5.4 is used to delineate the different depths at which units are simulated in the model domain. All model layers were simulated as flat due to the lack of structural dip at the scale of the model. We assume each well's injection is uniformly distributed over the perforated interval defined in Table 5.3. The primary input for each injection well is the reported monthly injection rate from the COGCC or NMOCD data (5-6). Our model discretization aims to accurately capture the vertical propagation of fluidpressure from injection (Fig. 18). We focused most of our computational resources on 168

Figure 5.17. No-flow boundary conditions of the hydrogeologic model are defined by the thick black dashed line. The Raton Basin boundaries are defined by uplifts in the Precambrian crystalline basement. Blue squares depict 28 injection wells which operated at one time or another in the Raton Basin during the period from 1994 - 2011.

169

Figure 5.18. Vertical discretization of the model domain. The model simulates 43 layers over the entire domain with an average thickness in the sedimentary section of ~68 meters. The highlight region is the zone of injection in the model domain. Overall, the model simulates nearly ~1.9 million cells.

170

discretizing a fine mesh in the vertical direction. The vertical discretization of the model contains 43 layers. Average layer spacing over the ~2.2 km thick sedimentary sequence is 68 meters (Figure. 5.18). The model domain solves the groundwater flow equation over ~1.9 million cells with a discretization in the x- and y- directions of 500 m.

5.6.3 Hydrogeologic Scenarios, Modeling Results & Discussion 5.6.3.1 Homogeneous Scenario We first aim to determine the lower boundary of hydraulic conductivity for the entire model domain. Our first hydrogeologic scenario consists of a purely hypothetical homogeneous, isotropic reservoir of equivalent hydraulic conductivity in the sedimentary and basement formations (Fig 5.19). The shale formations which confine the Dakota Formation are simulated as no-flow for this initial hydrogeologic scenario. We simulated hydraulic conductivity over four orders of magnitude. The effect of specific storage was also tested. The upper bound on hydraulic conductivity of 10-5 m/s is on the order of the highest reported value for the Dakota Formation from the literature. The lower bound of 10-8 m/s is set by calibrated modeling studies of the same formation in the Denver Basin. A specific storage value of 10-7 m-1 comes from calibrated values for the Dakota Formation (see section 5.4). Simulated pressure head values in two of the high-rate injection wells, VPR A 182 and Wild Boar, show a wide range of responses depending on the material properties (Fig. 5.20). At the low-end of hydraulic conductivity (10-8 m/s), pressure rises thousands of meter above land surface at both wells. At the high-end of hydraulic conductivity (10-5 m/s), pressure barely reaches 10 m above ambient reservoir conditions. One caveat to the high-end result, however, is

171

Figure 5.19. Homogeneous hydrogeologic scenario in which hydraulic conductivity of sedimentary and basement units are similar.

172

Figure 5.20. Logarithm of pressure head change in meters versus time in 4 homogeneous model scenarios: K =10-5 m/s (dashed line), 10-6 m/s (triangle), 10-7 m/s (dot), 10-8 m/s (line) and Ss = 10-7 m-1. Plots represent pressure head change at the (A) Wild Boar well and (B) VPR A 182 well.

173

Figure 5.21. Logarithm of pressure head change in meters versus time in 4 homogeneous model scenarios: K =10-5 m/s (dashed line), 10-6 m/s (triangle), 10-7 m/s (dot), 10-8 m/s (line) and Ss = 10-6 m-1. Plots represent pressure head change at the (A) Wild Boar well and (B) VPR A 182 well.

174

the effect of model boundaries in both of the well responses. This scenario is most likely unrealistic because pressure changes extend to all boundaries in the model by the end of 2011. The mid-range K model of 10-7 m/s shows a realistic pressure curve, but just exceeds the underpressure threshold at both wells. Increasing specific storage from 10-7 m-1 to 10-6 m-1 had the effect of delaying the reservoir pressure changes at both the VPR A 182 and Wild Boar wells (Fig. 21). However, the maximum pressure change for both wells is similar to the lower specific storage models. From these initial scenarios, the sedimentary formation K is likely in the range from 10-6 m/s to 2.5x107

m/s. These K values produce realistic pressure curves at both wells and are a range supported

by previous studies. We rely on the literature value of specific storage, Ss = 10-7 m-1, for the Dakota Formation as the model scenarios do not appear to further distinguish this value.

5.6.3.2 Heterogeneous Scenario 1: Sediment-Basement Contrast While the homogeneous scenarios constrain the lower bound of the hydraulic conductivity of the reservoir, a more realistic portrayal of the subsurface explores sedimentarybasement heterogeneity (Fig. 22). This scenario uses a range of K in the sedimentary formation from 10-6 m/s to 5x10-7 m/s. Meanwhile, we vary the hydraulic conductivity of the basement from 10-7 to 10-10 m/s. At the depths and temperatures of interest, the hydraulic conductivity of the top of the crystalline basement is predicted to be ~10-8 m/s (Equation 5.1). In this scenario, we also introduce the shale confining units, using the material properties reported in Garravito et al. [2006]. The maximum pressures head changes calculated at VPR A 182 range from ~280 m to ~520 m by varying the basement K over 4 orders of magnitude (Fig. 23A). The pressure head changes for all model scenarios stayed below the critical threshold at both VPR 175

Figure 5.22. Heterogeneous scenario where the basement formation is between 2.5 and 10,000 times less permeable than the sedimentary formation.

176

Figure 5.23. Pressure head changes versus time for heterogenous scenarios varying basement from K =10-7 m/s (dashed line), 10-8 m/s (triangle), 10-9 m/s (dot), 10-10 m/s (line). (A) VPR A 182 well with sediment hydraulic conductivity of 10-7 m/s. (B) Wild Boar well with sediment hydraulic conductivity of 5x10-7 m/s. (C) VPR A 182 well with sediment hydraulic conductivity of 10-6 m/s. (D) Wild Boar well with sediment hydraulic conductivity of 10-7 m/s.

177

A182 and Wild Boar well. One interesting result of this scenario is the relative insensitivity of the pressure at the wellbores to the basement K. Lowering the basement K over 4 orders of magnitude only increased the maximum pressure from ~170 m to ~340 m (Fig. 5.23C). Conversely, one model predicted the effect of lowering the sediment K by half an order of magnitude raised the maximum pressure change from ~170 m to ~280 m (Fig. 5.23A-C). Thus, the pressures calculated at the wellbores are far more sensitive to changes in sediment K than basement K.

5.6.3.3 Heterogeneous Scenario 2: Depth-Decreasing Basement Conductivity I add another layer of complexity to the hydrogeologic model with a component of depthdecreasing basement K. In this model scenario, I explore both wellbore pressure changes and how the model simulates 3D pressure changes in space and time. This scenario explores three models for depth-decreasing hydraulic conductivity in the crystalline basement (Fig. 5.24). Following the guide of Manning and Ingebritsen [1999], I vary the top of the basement K from 10-7 to 10-9 m/s. For each of the runs, the basement K decays two orders of magnitude below the top basement K. Due to the sensitivity of the model to sediment K, as found in the previous section, I also vary the sediment K in this scenario from 2.5x10-7 m/s to 10-6 m/s. Pressure head at the VPR A 182 and Wild Boar wells show a strong sensitivity to the sediment K and some sensitivity to the depth-decaying basement K (Fig. 5.25). The low-end scenario, using a sediment K of 2.5x10-7 m/s and basement K of 10-9 to 10-11 m/s, shows pressure in excess of land surface at VPR A 182. All other runs in this model scenario do not produce pressure changes in excess of land surface and therefore cannot be ruled out as possible fits to the hydrogeologic data. 178

Figure 5.24. Heterogeneity scenario with a depth-decreasing basement hydraulic conductivity.

179

Figure 5.25. Pressure head change versus time for nine models of sediment-basement heterogeneity with depthdecaying conductivity in the crystalline basement. Results are plotted for the VPR A 182 well and the Wild Boar well.

180

The threshold of underpressure alone does not distinguish between several of the model scenarios performed up to this point. There are multiple heterogeneous models, ranging orders of magnitude in hydraulic conductivity, which do not produce any wellhead pressure at the injection wells in Raton Basin. Data provided by the literature on the hydrostratigraphic units do not help further constrain a preferred model result. Therefore, I analyze the 3D pressure perturbation through time to help discern a preferred model result. I analyze the 3D pore pressure perturbation for two model outputs, both of which have depth-decreasing basement hydraulic conductivity and a sediment K of 5x10-7 m/s. Neither model produces wellhead pressure at any of the injection wells in the basin. The first model result uses a depth-decreasing basement K from 10-8 m/s at the top of basement to 10-10 m/s at the bottom of basement (Fig. 5.26; Fig. 5.27). The 3D pressure perturbation tracks earthquake hypocenters well in space and time, reaching depths of 6 to 7 km below the injection reservoir by September 2011. A histogram of calculated pore pressures at each hypocentral location finds most failures occur at pressure increases of 0.5 and 1 MPa (Fig. 27C). In plan view, the model predicts a reasonable lateral extent of pressure change, without noticeable excursions of pressure to north, east and southern boundaries. The second observed model run uses a depth-decreasing basement K from top to bottom of 10-9 m/s to 10-11 m/s (Fig. 5.28; Fig. 5.29). Previously, I showed pressure head at the wellbore to be relatively insensitive to basement hydraulic conductivity. Analyzing the data in three dimensions shows the vertical pressure perturbation is highly sensitive to basement K (Fig. 5.28). Lowering basement K one order of magnitude drives pressure much further laterally, while drastically reducing the amount of pressure diffusing to depth (Fig. 5.29). Qualitatively, the plan view prediction of pressure change in this scenario seems to move pressure further than expected

181

Figure 5.26. Cross-sections of pore pressure distribution and earthquake hypocenters for two time periods: October 2001 and September 2011. The plotted pore pressures come from a model with sediment K = 5 x10-7 m/s and a depth-decreasing basement K from 10-8 m/s to 10-10 m/s. The seismicity plotted represents the relocated earthquake hypocenters between August 2001- January 2002 and August-December 2011(2). Larger circles denote earthquakes of larger magnitudes. Model results plotted using USGS Model Viewer (31).

182

Figure 5.27. Plan view of pore pressure distribution and earthquake hypocenters for two time periods: (A) October 2001 and (B) September 2011. The plotted pore pressures come from the same model as Fig. 5.25. (C) Histogram of simulated pore pressure changes at the location and time of each earthquake hypocenter in the 2001 and 2011 earthquake sequences denoted by the box with the dashed line in (A) and (B). The seismicity plotted represents the relocated earthquake hypocenters between August 2001- January 2002 and August-December 2011(2). Larger circles denote earthquakes of larger magnitudes. Model results plotted using USGS Model Viewer (31).

183

Figure 5.28. Cross-sections of pore pressure distribution and earthquake hypocenters for two time periods: October 2001 and September 2011. The plotted pore pressures come from a model with sediment K = 5 x10-7 m/s and a depth-decreasing basement K from 10-9 m/s to 10-11 m/s. The seismicity plotted represents relocated earthquake hypocenters between August 2001- January 2002 and August-December 2011 (2). Larger circles denote earthquakes of larger magnitudes. Model results plotted using USGS Model Viewer (31).

184

Figure 5.29. Plan view of modeled pore pressure distribution and earthquake hypocenters for two time periods: (A) October 2001 and (B) September 2011. The plotted pore pressures come from the same model as Fig. 5.27. (C) Histogram of simulated pore pressure changes at the location and time of each earthquake hypocenter in the 2001 and 2011 earthquake sequences denoted by the box with the dashed line in (A) and (B). The seismicity plotted represents the relocated earthquake hypocenters between August 2001- January 2002 and August-December 2011(2). Larger circles denote earthquakes of larger magnitudes. Model results plotted using USGS Model Viewer (31).

185

laterally. Pressure changes interact heavily with the northern, southern and even eastern boundaries of the model. In addition, the 3D pressure perturbation does not track hypocenters in depth through time. By 2011, the pressure perturbation has only reached a maximum depth of 3 km below the injection interval (Fig. 5.28). Pore pressure changes at hypocentral locations do not show clear evidence of a pore pressure range at which failure occurs (Fig. 5.29C). These contrasting model results both fit the current hydrogeologic data and show that the vertical propagation of pressure is highly sensitive to basement hydraulic conductivity. By lowering basement K, fluid pressure migrates further in the x- and y- directions. The current hydrogeologic data lacks good constraints on the extent of pressure migration in the x- and ydirections. Lateral pressure barriers in the sedimentary formation could restrict pressure perturbations in the x- and y- directions. Knowledge of lateral pressure barriers could allow for a better understanding of the basement conductivity structure, because wellbore pressures would be more sensitive to changes in vertical heterogeneity. A better understanding of heterogeneity in the x- and y- directions may help further constrain future model runs. Most of the other scenarios of varying sediment K and depth-decreasing basement K predict pressure perturbations at earthquake depths. Only the model runs which used conductivities of 10-9 m/s or smaller at the top of the crystalline basement did not allow pressure to migrate to earthquake depths by September 2011. From the available data, one cannot rule out models which use basement conductivities of 10-9 m/s or lower.

5.6.3.4 Heterogeneous Scenario 3: Permeable Fault The final modeling scenario includes the addition of a permeable fault in an otherwise low-permeability crystalline basement. The 2001 and 2011 earthquakes sequences both highlight

186

Figure 5.30. (top) Heterogeneity scenario with a permeable fault represented as a zone of high vertical hydraulic conductivity in an otherwise low-K basement. The fault zone is 15 km long and 2 km wide, following the combined trend of the 2001 and 2011 earthquake sequences. (bottom) Pressure head results at the VPR A 182 and Wild Boar well versus time.

187

a series of faults which I approximate as a fault zone trending SW-NE along the transect plotted in Figure 5.28. Fitting the relocated earthquake sequences as a continuous SW-NE trending zone produces a 15 km long and 2 km wide fault zone. The fault is set to be permeable from the top of the crystalline basement to the base of the model as an anisotropic zone of high vertical hydraulic conductivity (Fig. 5.30). Vertical hydraulic conductivity of the fault zone is varied from 10-6 m/s to 10-7 m/s in an otherwise depth-decaying, low-K basement of 10-9 m/s to 10-11 m/s. Model results show the conductivity of the fault zone is almost indistinguishable in the calculated pressure head at both the VPR A 182 and Wild Boar wells (Fig. 5.30). When compared with results lacking a fault zone altogether, the presence of the fault zone decreases pressure measurements at the VPR A 182 well by a maximum of ~50 m. For most of the runs, however, the pressure effect of the fault is unperceivable at the wells. The fault zone is located directly beneath both the VPR A 182 and Wild Boar wells. These wells should be the most sensitive wells to pressure changes caused by the fault zone. A permeable fault zone is a possible conduit for fluid-pressure to migrate relatively quickly to depth, but the results show that this type of model may not be distinguishable from an otherwise low-permeability basement scenario.

5.7 Conclusions This study provides the first quantitative evaluation of the fluid-pressure changes associated with injection in the Raton Basin of southern Colorado and northern New Mexico. Previous studies had shown a correlation between onset of fluid injection in the basin and seismic rate increases (2-3). I synthesized the available geologic, hydrogeologic and injection data to build 3D hydrogeologic models of more than 25 injection wells operating in the deep

188

basin. Modeled pressure changes were compared against the known ambient underpressure in the injection reservoir prior to the start of injection. I find the reservoir's hydraulic conductivity to be in the range between 10-6 m/s to 10-7 m/s. The hydraulic conductivity of the crystalline basement is difficult to constrain as modeled pressures at the injection wells were far more sensitive to sedimentary formation hydraulic conductivity. A range of plausible models of sedimentbasement formation heterogeneity fit the available data. Many of the models within this range show 3D pressure perturbations track the migration of earthquake hypocenters to depths of more than 6 km from the injection interval. The calculated threshold for models that track hypocenters suggest earthquake nucleation occurs at pressure changes from 0.5 to 1 MPa. Models which use basement hydraulic conductivities of 10-9 m/s or less still fit the available hydrogeologic data and do not predict perturbations reaching the depths of recent earthquakes in the Raton Basin. Ultimately, the range of plausible models developed in this study add to the existing body of the evidence supporting an induced origin of the recent seismicity in the Raton Basin.

5.8 References [1]

M.E. Meremonte, J.C. Lahr, A.D. Frankel, J.W. Dewey, A.J. Crone, D.E. Overturf, D.L. Carver, and W. T. Bice, Investigation of an Earthquake Swarm near Trinidad, Colorado, August-October 2001, US Geological Survey Open-File Report. 02-0073 (2002).

[2]

J. L. Rubinstein, W.L. Ellsworth, A. McGarr, H. Benz, The 2001-present induced earthquake sequence in the Raton Basin of northern New Mexico and southern Colorado, Bull. Seismol. Soc. Am. 104, 1-20 (2014).

[3]

W. D. Barnhart, H. M. Benz, G. P. Hayes, J. L. Rubinstein, and E. Bergman, 189

Seismological and geodetic constraints on the 2011 Mw5. 3 Trinidad, Colorado earthquake and induced deformation in the Raton Basin. Journal of Geophysical Research: Solid Earth. 119(10), 7923-7933 (2014). [4]

ANSS catalog, United States Geological Survey, http://earthquake.usgs.gov/earthquakes/search/, accessed 04/01/2012.

[5]

Colorado Oil and Gas Corporation Commission Website, http://cogcc.state.co.us/#/home, accessed 01/01/2013.

[6]

New Mexico Oil Conservation Division Permitting Website, https://wwwapps.emnrd.state.nm.us/ocd/ocdpermitting//Data/Wells.aspx, last accessed 01/01/2013.

[7]

J. Nakai, A. F. Sheehan, M. Weingarten and S. L. Bilek, Potential induced seismicity in the Raton Basin, Colorado and New Mexico, 2008-2009, Seismological Research Letters, 86(2B), 691 (2015).

[8]

D.A. Lindsey, Laramide structures of the central Sangre de Cristo Mountains and adjacent Raton Basin, southern Colorado, The Mountain Geologist,.35 (2), 55–70 (1998).

[9]

E. H. Baltz, Stratigraphy and history of Raton Basin and notes on San Luis Basin, Colorado–New Mexico, American Association of Petroleum Geologists Bulletin. 49, 2041–2075 (1965).

[10] J.C. Lorenz, S. P. Cooper, B. W. Arnold, P. M. Basinski, J. M. Herrin, J. F. Holland, R. G. Keefe, R. Larson, W. A. Olsson and C. A. Rautman, Natural Gas Production Problems: Solutions, Methodologies and Modeling, Sandia National Laboratories Report, 2004-4859 (2004). [11] D. K. Higley, Petroleum systems and assessment of undiscovered oil and gas in the Raton

190

Basin–Sierra Grande Uplift Province, Colorado and New Mexico, USGS Province 41: U.S. Geological Survey Digital Data Series DDS–69–N. ch. 2, 124p. (2004). [12] S. G. Robson, E.R. Banta, Geology and hydrology of the deep bedrock aquifers of eastern Colorado, USGS Water Investigations Report. 85–4240, 6 sheets (1987). [13] A. L. Geldon, Ground-water hydrology of the central Raton Basin, Colorado and New Mexico, USGS Water-Supply Paper. 2288, 81p (1989). [14] R. Sibson, A Brief Description of Natural Neighbor Interpolation, Interpolating Multivariate Data, New York: John Wiley & Sons. ch. 2, 21–36 (1981). [15] G. L. Shaw, Pennsylvanian History and Stratigraphy of the Raton Basin, Rocky Mountain Association of Geologists, (1958). [16] A. S. Suleiman, G. R. Keller, A geophysical study of basement structure in northeastern New Mexico, New Mexico Geological Society 36th Annual Fall Field Conference Guidebook. p.153-159, (1985). [17] H. T. Hemberg, Basement Structure Map of Colorado with Major Oil and Gas Fields, Colorado Geological Survey Maps, (1996). [18] A. M. Garavito, H. Kooi, and C. E. Neuzil, Numerical modeling of a long-term in situ chemical osmosis experiment in the Pierre Shale, South Dakota, Advances in Water Resources. 29 (3), 481-492, (2006). [19] R. Topper, K. Scott, and N. Watterson, Geologic model of the Purgatoire River watershed within the Raton Basin, Colorado, Colorado Geological Survey, (2011). [20] K. R. Watts, Hydrostratigraphic framework of the Raton, Vermejo, and Trinidad aquifers in the Raton Basin, Las Animas County, Colorado. USGS Scientific Investigations Report. 2006–5129, p.37 (2006).

191

[21] J. D. Bredehoeft, C. E. Neuzil, and P. C. D. Milly, Regional Flow in the Dakota Aquifer: A study in the Role of Confining Layers, USGS Water-Supply Paper 2237. (1983). [22] R. H. Miller, P. H. Rahn, Recharge to the Dakota sandstone from outcrops in the Black Hills, South Dakota, Association of Engineering Geologists Bulletin, 11 (3), p 221234 (1974). [23] K. Belitz, J. D. Bredehoeft, Hydrodynamics of Denver basin: explanation of subnormal fluid pressures, American Association of Petroleum Geologists Bulletin. 72, 1334 – 1359 (1988). [24] F. S. Jensen, H. H. R. Sharkey, and D. S. Turner, The oil and gas fields of Colorado, Rocky Mountain Association of Geologists, p.302 (1954). [25] J. W. Harbaugh, W. Davie, Jr., Upper Pennsylvanian calcareous rocks cored in two wells in Rawlins and Stafford Counties, Kansas, Kansas Geological Survey Bulletin. 170 (6), p18 (1964). [26] C. E. Manning, and S. E. Ingebritsen, Permeability of the continental crust: Implications of geothermal data and metamorphic systems. Reviews of Geophysics. 37(1), 127150, (1999). [27] P. H. Nelson, N. J. Gianoutsos, N. J., and L. O. Anna, Outcrop control of basin-scale underpressure in the Raton Basin, Colorado and New Mexico, The Mountain Geologist, 50 (2), 37-63 (2013). [28] E. D. Dolly, F.F. Meissner, Geology and gas exploration potential, upper Cretaceous and lower Tertiary strata, northern Raton Basin, Colorado, in Veal, H.K., ed., Exploration frontiers of the central and southern Rockies, Rocky Mountain Association of Geologists, 247–270 (1977).

192

[29] R. B. Johnson, Geology of the Huerfano Park area, Huerfano and Custer Counties, Colorado, USGS Bulletin. 1071-D, 119 p. (1959). [30] A.W. Harbaugh, E.R. Banta, M.C. Hill, M.G. McDonald, MODFLOW-2000, The U.S. Geological Survey Modular Ground-Water Model - User guide to modularization concepts and the ground-water flow process, U.S. Geological Survey Open-File Report 00-92, 121 p. (2000). [31] P. A. Hsieh, R. B. Winston, User’s Guide To Model Viewer, A Program For ThreeDimensional Visualization of Ground-water Model Results. U.S. Geological Survey Open-File Report. 02-106, 18 p. (2002).

193

ON THE INTERACTION BETWEEN FLUIDS AND ...

ON THE INTERACTION BETWEEN FLUIDS AND ...

Suggest Documents

Interaction between SPH Fluids and Dynamic ... - Semantic Scholar

Investigation on the Interaction between Cyclophosphamide and

ON THE INTERACTION BETWEEN CONVECTION AND MAGNETIC ...

on the interaction between azide and manganese

On the interaction between sharing and linearity

The Interaction between Nidovirales and

THE INTERACTION BETWEEN CULTURE AND ...

Interaction Between Daidzein and Hesperetin on Antispasmodic ...

Spectroscopic studies on the interaction between erlotinib ...

Investigation on the Interaction between Cyclophosphamide - MDPI

Investigation on the interaction between natural

On the functional interaction between nicotinic ...

Interaction between exopolysaccharide and

The Interaction between Saccharomyces

Study on the interaction between the food and beverage servicescape

Interaction between fructose and glucose on the regulation of the ...

On the interaction between bathymetry and climate in the ... - CiteSeerX

Isotopic exchange between thermal fluids and

Exploring the interaction between siRNA and the

Finite Element Investigations on the Interaction between a Pile and ...

Geochemical and isotopic constraints on the interaction between ...

On the interaction between knowledge and social ... - Semantic Scholar

Economic Crisis and its Influences on the Interaction between ... - CORE

On the Interaction between Mobile Processes and Objects - CiteSeerX