SI: Wastewater disposal and earthquake swarm activity at the southern end of the Central Valley, California

Transcription

1 SI: Wastewater disposal and earthquake swarm activity at the southern end of the Central Valley, California T. H. W. Goebel 1*, S. M. Hosseini 2, F. Cappa 3,4, E. Hauksson 1, J. P. Ampuero 1, F. Aminzadeh 2, and J. B. Saleeby 1 1 Division of Geological and Planetary Sciences, CA Institute of Technology, Pasadena, CA, USA. 2 Department of Chemical Engineering and Material Sciences, University of Southern CA, Los Angeles, CA, USA. 3 Geoazur Laboratory, University of Nice Sophia-Antipolis, Nice, France. 4 Institut Universitaire de France, Paris, France * now at University of California, Santa Cruz Contents 1 Tectonic setting at the southern end of the Central Valley 4 2 Geologic structures involved in rupture and fluid migration processes 6 3 FMD, b-value and confidence interval 10 4 Earthquake cluster relocation 11 5 Additional earthquake detection using the template matching method 13 6 Moment tensor inversion and moment magnitudes 20 1

2 7 Hydrogeological model of fluid injection induced fluid pressure perturbations Intro: Crustal permeability structure close to injection well WD Theoretical background Sensitivity analysis of permeability and fault zone width

3 Additional Supporting Information (Files uploaded separately) File Name data.zip WF_new.zip modelresults.zip WWF_2D_v2.gif WWF_3D_v1.gif Description Seismicity catalogs and injection rates in well WD01, WD04 & WD05 Waveform plots of newly detected events and templates Results and initial conditions of pore-pressure modeling 2D animation of WWF swarm epicenter locations 3D animation of WWF swarm hypocenter locations 3

4 1 2 1 Tectonic setting at the southern end of the Central Valley In order to elucidate the complex structure in the area of the injection well and to discuss the plausible fluid pathways that could induce seismicity, we present a simplified fault map of the Tejon area (Figure 1), and generalized cross sections that show fault depth relations and fault inter-relationships (Figure 2). Figure 2 shows the position of the injection well above the approximate surface traces of buried, Early-Miocene, normal faults. A comparison of Figure 2 with the seismicity map in the main text (Figure 1a) shows a band of background seismicity along one of the principal buried normal faults, which along with geomorphic observations along the Tehachapi foothills where the fault is observed at basement levels, indicates Holocene fault-remobilization 20 (J.B. Saleeby, unpub. mapping). Recognizing the dominate north-south directed thrust or reverse sense to the M w 4.6 mainshock, we show in Figure 2 a south-north cross section that crosses the mainshock focal point (synthesized from 10,11,15 ; and J.B. Saleeby, unpub. mapping). The focal point of the M w 4.6 mainshock is deep within the footwall domains of the Pleito and Wheeler Ridge north-directed thrust faults, but lies within the depth projected White Wolf fault zone. On the south-north cross section we also denote the focal point of the north-directed reverse dominated 1952, M w 7.3 Kern County earthquake, which also lies along the depth projected White Wolf fault, at 19 km depth. These relations lead us to hypothesize that the M w 4.6 mainshock is associated with the White Wolf fault. 4

5 119.2 W W W W W W W 35.2 N 35.1 N 35.0 N 34.9 N 34.8 N Figure 1: Generalized fault map for Tejon area showing approximate surface traces of an active blind thrust and several reverse faults, buried early Miocene normal faults, and areas where faults are exposed in adjacent exhumed basement (sources: 6,14,20,21, J.B. Saleeby, unpub. mapping. Final drafting by Z.A. Saleeby.) 23 The principal basement structure of the White Wolf fault formed in the Late 24 Cretaceous in continuity with the early-phases of the Breckenridge-Kern Canyon 25 system to the north 6. Tens of kilometers of spatially varying oblique slip on the in- 26 tegrated system led to severe basement damage zones, numerous subsidiary fault 27 breaks, and rapid differential basement exhumation of up to 15 km 31. The typi- 28 cal width of the ductile to brittle basement damage zone, where exposed, is shown 29 on the south-north cross section in Figure 2. From Early-Miocene to Early-Pliocene 30 the White Wolf segment of the system functioned primarily as a north-down nor- 31 mal fault 10,11,33, resulting in its profound north wall sediment accommodation 32 space (Figure 2, south-north cross section). With the late Pliocene-Quaternary 5

6 33 34 growth of the Tehachapi-San Emigdio fold thrust belt the White Wolf zone has inverted along much of its length to a north-directed reverse/thrust zone 8, Geologic structures involved in rupture and fluid mi- gration processes Several plausible tectonic structures exist that could be associated with the M w 4.6 mainshock such as: 1) Rupture processes along a Riedel shear within the principal basement damage zone; 2) slip near the intersection of a hypothetical blind thrust fault zone, suggested by Davis and Lagoe, (1988) and Gordon & Gerke (2009), and approximated in depth as the hypothetical blind thrust on the Figure 2 south-north cross section; or 3) slip along a zone of stress concentration at the basement surface footwall cut-off of the White Wolf fault (Figure 2 south-north cross section). Potential fluid migration paths from the injection wells into the White Wolf fault zone can be visualized by projecting the structure of the Figure 2 1 west-east cross section northwestward onto the fault zone. Most notable are remobilized Early-Miocene normal faults. Fault damage zone permeability and hanging wall brecciation provide potential direct fluid pathways. Our geologic model is supplemented by detailed reservoir models based on standard petroleum industry data such as seismic profiles and well log data, available from the CA Division of Oil, Gas and Geothermal Resources (DOGGR) 3,4. The 1 In Figure 2, we included an uncertainty ellipse for the injection position shown on the west-east cross section. The uncertainty is derived from: 1) Geological uncertainty in subsurface stratigraphic position resulting from having only a mud log, and no supporting electric log nor core data for the injection well; 2) Rapid south to north facies changes in the Neogene section from the basin margin to interior 10,14,20 ; 3) Northwards projection of the structure across relatively small offset NE SW trending normal faults that are transverse to the principal NW SE trending normal faults that the projection follows; and 4) The effect of the 1500 m long NNW trending horizontal well segment for which fluids are incrementally injected over the last 520 m. 6

7 S 1 Tehachapi - San Emigdio foothills Pleito flt. Wheeler Ridge W-E section San Joaquin Basin N 1 km MSL -1? MSL W 1 Wheeler Ridge flt. Quaternary alluvium Base of Pleistocene Tulare Fm. Base of up. Pliocene San Joaquin Fm. Base of up. Miocene Santa Margarita Fm. Base of up. Miocene Monterey Fm. Base of lw. Miocene Tunis-Tecuya volcanics (Tv)? Basement Pre-Cenozoic White Wolf fault damage zone? 1952 M 7.3 reverse event at 16 km depth Hypothetical Blind Thrust M 4.7 White Wolf fault S-N section E injection well projected ~6 km 1 km northward along normal fault strike normal growth fault sedimentation? approx. depth of Tv isochronous surface km mi km MSL MSL -1-2 uncertainty on injection position N - plunging turbiditic Holocene sand lenses in Monterey remobilization Lw. Miocene normal growth faults Tunis volcanics km mi -6-6 km Figure 2: Simplified south to north, and west to east geologic cross sections for area of injection well and hypothesized induced seismicity. Locations given in Figure 1. Selected stratigraphic horizons are shown as structural form lines. Sources given in Supplementary text. Final drafting by Z.A. Saleeby. 7

8 industry data highlights many fracture and fault zones located between injection site and the White Wolf fault (Figure 3). The WNW orientation of these faults is in agreement with a linear feature in the seismicity catalog highlighting that at least one of these faults is seismically active. This fault is referred to as Tejon fault in the main text. The seismicity catalog indicates that the Tejon fault deepens toward the White Wolf fault which is also seen within some of the interpreted well-logs that extent to 3.7 km depth 4. The closest fault zones identified in the DOGGR data is located less than 0.5 km from injection site WD05. Thus, these faults likely influenced the nearby permeability structure and fluid flow path ways of injected wastewater. 8

9 M4.7 o W e hit W au lf F lt Tejon, North Pleito Fault Wh ult lf Fa o ite W Fault and Fracture Zones 0 1 WD05 Figure 3: Mapped faults and fractures within the Tejon, North oilfield. The top plot shows the approximate boundaries of the oil field (blue region) and earthquake locations. Seismicity likely associated with the Tejon fault is highlighted in black and background seismicity in gray. The bottom plot shows the more detailed oilfield boundaries (gray shaded area) based on a DOGGR technical report 3. Faults and fracture zones were identified based on industry data such as seismic profiles, geological mapping and well logs (Modified after California Division of Oil, Gas & Geothermal Resources, Technical Report 3 ). The displayed seismicity was predominantly recorded before the 2005 swarm and highlights the approximately linear trend of the Tejon fault which connects the injection site and largest magnitude epicenter. 9

10 62 3 FMD, b-value and confidence interval Frequency-magnitude distributions (FMD) of earthquakes within the study area generally followed a power-law of the form: 65 log N = a b M (1) where b is the power-law exponent, N and M are the event number and magnitude and a is a constant related to the overall seismic productivity within the area 17. To compute b-values, we used the maximum likelihood approach 1, b = 1 M M c log(e) (2) where M c is the magnitude of completeness, corrected for bin-size 39 and M is the mean magnitude of events above M c. The lower power-law bound, M c, was determined by minimizing the Kolmogorov-Smirnoff distance between the observed and modeled power-law distributions 7. We compared the observed FMD with expectations for FMDs with a b-value of unity using Monte Carlo simulations 13. The corresponding confidence limits as function of magnitude are reported as the 2.5 and 97.5 percentiles of 1,000 randomly sampled FMDs with the same parameters (a, b, M c ) as the observed distribution. In the main manuscript, we show that b-values show significant variations that coincide with peak injection rates in well WD05. b-values are close to unity before and after peak injection but reach values as low as b=0.6 during the highest injection rates in In addition, we investigate long-term temporal variations in b-value within a 10 km radius of well WD05 between 1995 and 2011 (Figure 4). b-values show strong variations between 0.6 and 1.4. We observed a systematic 10

11 decrease in b starting in 2004 which is when one of the high-rate injection wells became active. This first order correlation between injection rates and temporal b-value variations may indicate a connection between pressure changes and the tendency of earthquakes to grow to larger sizes resulting in lower b-values. However, variations in b-value should be interpreted carefully because of the inherent uncertainties in estimating power-law exponents over a limited number of decades and from seismicity records with changing magnitude of completeness over time. Moreover, a precise correlation between b-value variations and other time-series, e.g. injection rates, is complicated by different temporal resolutions. Thus, the most robust features within our data are generally low b-values during the time of peak injection rates Earthquake cluster relocation In addition to the standard single event locations within the Southern California Seismic Network (SCSN) earthquake catalog, we use a state-of-the-art waveformrelocated catalog 19. The relocated catalog is based on a detailed, 3D velocity model and precise, travel-time differences determined by waveform cross-correlations of earthquakes within clusters 26,36. The relocated seismicity catalog highlights that most of the earthquakes that occurred during the time of peak injection in WD05 occurred at depths between the injection site and the M w 4.6 hypocenter (Figure 5). The seismic activity that deepened toward the White-Wolf fault was likely a key component in maintaining a high-permeability pathway that transmitted pressure changes to large depths. 11

12 WD05 Figure 4: Temporal b-value variations within a 10 km radius around well WD05 and injection rates in wells WD01, WD04 and WD05. 12

13 0 5 Depth [km] Number of Earthquakes Figure 5: Focal depth distributions of events within a 10 km radius of well WD05 between 04/2005 to 09/2005. The red star highlights the focal depth of the M w 4.6 hypocenter Additional earthquake detection using the template matching method We extend the available earthquake catalog by cross-correlating event template waveforms with the continuous waveform record. The waveform templates are created from the earthquake record within the SCSN catalog. We use event waveforms as templates that were recorded at the closest station with high signal-tonoise ratio, i.e. station MPI. For this study, we selected 14 template event waveforms with magnitudes between 0.6 and 1.8, which occurred within a ten kilometer radius of well WD05 and were recorded and produced high-quality template waveforms on all three components at station MPI. This station was installed in Our implementation of the template-matching method largely follows Meng et al., (2013) 30 and consists of 4 main steps: Template selection and processing: We down-sample all waveforms to 20 Hz to match the continuous, broadband waveform record which is available af- ter June 2005 and bandpass-filter all waveforms between 5 to 10 Hz. This fre- 13

14 quency band is generally above the largest, coherent noise sources, providing the most stable results for event detection. After bandpass-filtering, the template waveforms are cut from 1 s before to 3 s after P-arrival on the vertical component. Waveforms on the radial components are selected between 2 s before and after the S-phase arrivals accounting for larger uncertainties in phase picks Cross-correlation: The template waveforms are then cross-correlated with the continuous waveform record between June 30th 2005 until the mainshock occurrence on September 22nd Cross-correlation coefficients (CC) between templates and continuous waveforms are computed for a 4 sec sliding window that is advanced by one sample steps for all three traces Event detection: An event detection is declared if the individual and stacked cross-correlation coefficients exceed a certain threshold. For the present application, we chose a conservative CC threshold to minimize false detection. We tested different CC thresholds and find that a factor of 7 times the standard deviation provides the best results Earthquake-catalog creation: For the newly-detected events, we use the times of highest correlation coefficients as origin times and adopt the locations and magnitudes of the template events for the new event We test our implementation of the template matching method by first evaluating its ability to self-detect events within the earthquake catalog that were recorded between June 2001 and October Expectedly, the corresponding correlationcoefficients (CC) across all traces are close to unity for these events (Figure 6). We minimize false-detection by requiring relatively high CC values that are at least 7 times larger than the standard deviation of noise correlations. In addition, 14

15 Figure 6: Example of waveform cross-correlations for a test-run of the template matching algorithm to self-detect a catalog event. Top: Template (blue) and continuous seismograms (gray) for the vertical, and the two horizontal components. CC values are shown in upper left corner of each subplot. Bottom Left: Time series of different CC values (black), detection threshold (red line) and event detection (red dot). Bottom Right: Histogram of CC values (black), detection threshold (red line) and event detection (red dot) we test if S-P times and event durations are in approximate agreement between template and newly detected events for each of the 21 new detection. Lastly, we check if the first motion polarities agree between templates and new detection. Figure 8 to 10 show three examples of successfully detected events including waveform records and CC values. The waveform plots for the other detected events are part of the Online Supplement. Seven of the 14 template waveforms were connected to additional seismic event detection resulting in a total of 21 additional event detection between the start of injection in well WD05 and the occurrence the mainshock (Figure 7). 15

16 Newly Detected Events Template Loca,on MPI Figure 7: Map-view of template (blue dots and waveforms) and newly-detected (red dots and gray waveforms) events based on cross-correlating waveforms at station MPI (red triangle). The White-Wolf swarm is highlighted by red, orange, yellow and green dots analogous to Figure 1a in the main manuscript.the injection site is highlighted by a blue triangle and the largest magnitude swarm event by a red star. 16

17 Figure 8: Same as Figure 6 but now showing an example of waveforms and CC values of a newly detected event Template matching is generally superior to simple phase picking algorithms, which rely on a notable amplitude or frequency difference between signal and noise, allowing for event detections even in high-noise environments 29,30,37. However, the limited availability of template events due to only one nearby station with continuous waveforms records that was recording over several years before and after the White Wolf swarm in 2005, prevents a systematic lowering of the completeness threshold within the area. Nevertheless, the template matching and catalog extension allow us to test if part of the Tejon fault was active during the high-rate injection period in WD05 and provides information about possibly systematic migration patterns (see main manuscript). 17

18 Figure 9: Same as Figure 6 but now showing an example of waveforms and CC values of a newly detected event. 18

19 Figure 10: Same as Figure 6 but now showing an example of waveforms and CC values of a newly detected event. 19

20 165 6 Moment tensor inversion and moment magnitudes We test the reliability of magnitude estimates in the upper tail of the FMD through determining moment magnitudes of events with M L 4.0. We find that the local magnitudes of M L = 4.5, 4.7 and 4.3 within the SCSN earthquake catalog exceed moment magnitude estimates of M w = 4.1, 4.6 and 4.0. However, three events above M w 4 are in good agreement with the expected number of M 4 events for the determined power-law parameters (b=0.6, a=2.8) and further support a significant deviation from a b-value of unity. Moment magnitudes were computed by fitting long-period waveforms in a 10 to 20 sec window immediately after P and S-phase arrivals using the Cut-And- Paste method for pure, double-couple source tensors 40,41. The synthetic far-field displacement for a double-couple source can be described as follows: 177 s(t) = M 0 3 i=1 A i (φ θ, δ, λ)g i (t), (3) here, i = 1,2,3 correspond to three fundamental faults, i.e. vertical strike slip, vertical dip slip and 45 dip slip, G i s are the Green s functions, φ is the station azimuth, θ, δ, λ are the strike, dip and rake of the source, and M 0 is the seismic moment. The parameters describing the source tensor can be determined by minimizing the misfit between observed and synthetic seismograms using a grid-search method within the limited parameter space. The CAP method fits both body and surface wave phases allowing for small time shifts (here 2 to 8 sec for P and S phases) between different phases, which account for unmodeled velocity structures and resulting uncertainties in Green s function estimates. We use a maximum fitting bandwidth of 0.02 to 0.2 Hz with a slightly narrower band for the smaller magnitude events between 0.08 to 0.2 Hz. We perform another 20

21 grid-search to solve to the focal depth of the synthetic source that minimizes the waveform misfits. For the focal depth inversion, we require reduction in waveform misfits of at least 10% and use the depth within the relocated catalog otherwise. For the largest magnitude event with M w 4.6, we find a focal depth of 9 km which is in good agreement with the relocated catalog depth. The following Figures 11 to 13 show inversion results and waveform misfits for the three events with M w

22 M4.4 M w 4.6 Mainshock, September 22 nd, 2015 M4.5 M4.5 M4.5 M4.5 M4.5 M4.5 M4.5 M4.6 M4.6 M4.6 strike = 89 dip = 30 rake = 72 sec Distance Sta@on Azimuth Time Shi= Correla@on Coefficient Figure 11: Focal mechanism (upper left), results of focal depth grid search including moment magnitude (upper right) and waveforms misfits (bottom) for the largest magnitude event that was part of the 2005 White Wolf swarm. The best-fitting focal mechanism is highlighted in red, observed waveforms are shown in black and synthetic waveforms in red. 22

23 M w 4.1 Forehock, September 22 nd, 2015 M4.1 M4.1 M4.1 M4.1 M4.1 M4.1 M4.1 M4.2 M4.2 M4.0M4.0 M4.0 M4.1 M4.0 M4.1 strike = 110 dip = 32 rake = 90 sec Distance Sta?on Azimuth Time Shi; Correla?on Coefficient Figure 12: Same as Figure 11 for an earthquake above magnitude 4 that occurred 44 sec before the M w 4.6 event. 23

24 M w 4.0 A)ershock, September 22 nd, 2015 M3.9 M3.9 M3.9 M4.0 M4.0 M4.0 M4.0 M4.0 M3.9 M3.9 M3.9 M4.0 M4.0 M4.0 M4.0 strike = 89 dip = 40 rake = 72 sec Distance StaEon Azimuth Time Shi) CorrelaEon Coefficient Figure 13: Same as Figure 11 for an earthquake above magnitude 4 that occurred 6 min after the M w 4.6 event. 24

25 Hydrogeological model of fluid injection induced fluid pressure perturbations Intro: Crustal permeability structure close to injection well WD05 We assess injection induced fluid pressure changes as a function of distance and time from injection operations in well WD05 by modeling pressure diffusion in three dimensions. The 3-D numerical diffusion model is based on the most complete available datasets within the upper 2 3 km of sedimentary basins and includes seismicity records, geologic mapping results, and industry data (i.e. welllogs, stratigraphic columns and interpreted reservoir structure). Below 3 km depth little data is available except for seismicity catalogs. The model includes three principal stratigraphic zones, in addition to the high-permeability portion of the Tejon fault which is implemented as a vertical zone of elevated permeability. These three zones are: 1) The injection zone, i.e. a m thick turbiditic sand lens in the Monterey formation with a lateral extent of up to 1.5 km, labeled as "Transition zone" in the industry data 4 ; 2) the crystalline (gneissic) basement complex, and 3) the Monterey formation (see 10,11,15 ). Permeability is generally high within the sand lenses (i.e. 1 D) and very low ( 10 4 md) above injection depth which is one of the requirements for the selection of an injection site. Similarly, permeability is low ( 10 4 md) within the basement and Monterey formations outside of the injection zone (permeabilities are reported in milli-darcy, 1 md m 2 ). High permeability within the injection zone is supported by publicly available well-log data, i.e., low Gamma-ray and high resistivity values 4. 25

26 Theoretical background In our model, injection induced pore-pressure diffusion is described by the uncoupled diffusivity equation for slightly compressible fluids for which changes in pressures do not affect the elastic deformation of the rock matrix 28,35 : p k x x 2 + k 2 p y y 2 + k 2 p z z 2 = φµc p t, (4) where k x, k y and k z are permeability in x, y and z directions, p is pore-pressure, φ is porosity, µ is fluid viscosity and c is rock matrix compressibility. To present the diffusivity equation in the presence of sources and sinks, we derive the mathematical description of single-phase fluid flow in porous media. We start with the continuity equation, Darcy s law for flow in 3D and the equation of state. The continuity equation in Cartesian coordinates with added source terms can be written as 25,28 : 231 x (ρq x) + y (ρq y) + z (ρq z) +... N i=1 [ρq wδ(x x w )δ(y y w )δ(z z w )] (5) = t (φρ), where q x, q y and q z are volumetric rates of flow per unit cross-sectional area, ρ is fluid density, x w, y w and z w are the injection locations, Q w is the flow rate per unit volume, N is the total number of injection wells in the domain, δ(x) is the Dirac delta function and φ is porosity. Darcy s law for flow in the x, y and z directions 26

27 236 can be written as 28 : q x = k xρ µ p x q y = k yρ p µ y q z = k ( ) zρ p µ z + ρg (6) (7) (8) where k x, k y and k z are permeability in x, y and z directions, p is pore-pressure, µ is fluid viscosity and ρ is fluid density. Using Darcy s flow law and Eq. 5, we can determine diffusion in the presence of sources and gravitational effects: ( ) ( ) ( ) kx ρ p x µ x + y ky ρ p µ y + z kz ρ µ ( p z + ρg)... (9) i=1 N [ρq wδ(x x w )δ(y y w )δ(z z w )] = t (φρ), In order to study isothermal single-phase flow of fluids with small and constant permeability, we use the following expression for fluid compressibility, c: c = 1 ρ ρ p (10) After integration of Eq. 10, we obtain the equation of state for slightly com- pressible fluids: ρ = ρ 0 e c(p p0). (11) as: The combined compressibility of fluid (c) and rock matrix (c r ) can be expressed c c = c + c r, (12) 254 The rock or formation compressibility can be described as change in porosity as a 27

28 255 function of change in pore pressure: c r = 1 φ φ p, (13) where φ again is porosity. Using the equation of state (Eq. 11) and Eq. 9, 12 & 13, we get: 259 k x 2 p x 2 + k y 2 p y 2 + k z 2 p z 2 + 2k z cρg p z... N i=1 [ρq wδ(x x w )δ(y y w )δ(z z w )] = φµc t p t, (14) which is the diffusivity equation used in our model. This description of the diffusive process assumes that pressure gradients and compressibility are small, and porosity does not change with time. This diffusivity equation is solved numerically using the finite-difference reservoir simulator ECLIPSE 32, Numerical model set-up For the present study, we used the single-phase reservoir simulator module of ECLIPSE ( black-oil simulator ) 34. The initial conditions of the numerical model are constrained by the available industry and geologic data described in Section 7.1 and include the high-permeability part at the top of the Tejon fault, identified through a combination of geologic mapping, seismicity locations and industry data. We assume a pressure gradient of 12 MPa/km. We use an average value for the porosity, φ = 0.2, rock compressibility, c r = Pa 1 and fluid compressibility, c = Pa 1. 3,18 The water formation volume factor at a reference pressure of 30 MPa is 1.13 (reservoir m 3 / standard condition m 3 ). The water viscosity at reference pressure is 0.5 cp. The size of the injection layer of 2.1 km 2 is determined from the publicly available data-sets on the California Division of Oil, 28

29 Gas and Geothermal Resources (DOGGR) website 3. For the present study, we assign no-flux boundary conditions. We chose a large modeling domain (x=30 km, y=5 km, z=18 km) and placed the injection well at the center of this domain to minimize potential effects of boundary conditions on the pressure solutions. The model domain is discretized in the following grid sizes: dx = m, dy = 500 m and dz = 100 m. At the vicinity of injection point and high permeability pathway, we use local grid refinement down to dx = 5.78, dy = 9.55 and dz = 1.58 m. The injection well is modeled as a vertical injector with a perforation zone of 30 m that spans the entire extent of the injection zone and exact monthly injection schedule is obtained from DOGGR 4, plotted in Figure Sensitivity analysis of permeability and fault zone width Crust and fault zone permeability vary non-linearly over several orders of magnitude both laterally and with depth so that permeability is one of the modeling parameters most difficult to constrain. The permeability within fault and fracture zones is generally higher than in the fault core and protolith and can be even higher in the presence of rupture induced damage, for example, toward the up-dip end of blind faults 12,23,27,38. The interplay between fault motion and permeability increase has repeatedly been observed in controlled injection experiments and is essential in maintaining relatively high fault permeability at seismogenic depth 2,5,9,16. This is also in agreement with observations at the continental deep drilling project (KTB), Germany where zones of elevated permeability were associated with fracture zones preferably oriented to slip within the local stress field 38,42,43. These zones of elevated permeability persisted down to depth of 9.1 km 42. Similarly, we expect regions with systematically higher seismicity density toward the upper limit of the Tejon fault to exhibit higher permeability. 29

30 In our model, we vary both permeability (k) and width (w) of the upper portion of the Tejon fault over a wide range of values (i.e. k=10 6 to 1 D and w=100 to 1000 m) to determine a plausible range of pore-pressure changes at the M w 4.6 hypocenter (Table 1 & 2). The sensitivity analysis showed that differences in permeability have a stronger influence on pore-pressure variations; however both higher fault zone permeability and narrower fault zone width can lead to larger pore-pressure changes at the M w 4.6 hypocenter. For fault zone permeability above 200 md and fault width below 1000 m, we observe a pressure increase at the M w 4.6 hypocenter of more than 0.1 MPa which is sufficient to trigger earthquakes on faults close to failure 22,24. The modeled pore-pressure increase at the M w 4.6 hypocenter is lower for a 312 fault zone with homogeneous permeability, i.e. without the high-permeability pathway toward the top of the fault. For fault zones with a width between w=300 and 600 m and permeability k >400 md, we expect an increase in pore-pressure at M w 4.6 hypocenter of less than 0.01 MPa (Figure 14). This further highlights that a narrow, high permeability pressure channel is expected to have the largest seismogenic consequences for the White Wolf fault. 30

31 Table 1: Overview of model runs, initial conditions and pore pressure increase at the M w 4.6 hypocenter for a wide range of permeabilities. Model run k i k f w IZL P p [md] [md] [m] [km] [MPa] < < < < < k i - injection zone permeability; k f - fault zone permeability; w - fault zone width; IZL - injection zone lateral extent; P p - pore pressure change at the M w 4.6 hypocenter after 150 days of high-rate injection 31

32 Table 2: Overview of model runs, initial conditions and pore pressure increase at the M w 4.6 hypocenter for a narrow range of permeabilities. Model run k i k f w IZL P p [md] [md] [m] [km] [MPa] k i - injection zone permeability; k f - fault zone permeability; w - fault zone width; IZL - injection zone lateral extent; P p - pore pressure change at the M w 4.6 hypocenter after 150 days of injection 32

33 P [MPa] p Figure 14: Expected pore-pressure increase at distances similar to M w 4.6 hypocentral distance for pressure diffusion along a fault with homogeneous permeability structure. The resulting pressure perturbations are significantly smaller than in the case of a highpermeability pathway at the top of the Tejon fault. 33

34 318 References [1] Aki, K. (1965), Maximum likelihood estimate of b in the formula log N = a - bm and its confidence limits, Bull. Earthquake Res. Inst., Tokyo Univ., 43, [2] Bourouis, S., and P. Bernard (2007), Evidence for coupled seismic and aseismic fault slip during water injection in the geothermal site of Soultz (France), and implications for seismogenic transients, Geophys. J. Int., 169(2), [3] CA Department of Conservation (1998), California oil and gas fields - Central California, DOGGR technical reports, 1, [4] CA Department of Conservation (2012), Division of Oil, Gas and Geothermal resources: Production history in California, (ftp://ftp.consrv.ca.gov/pub/oil/annual_reports, accessed Sept. 26, 2014) [5] Chabora, E., et al. (2012), Hydraulic stimulation of well 27 15, Desert Peak geothermal field, Nevada, USA, in Proceedings of thirty seventh workshop on geothermal reservoir engineering, Stanford University, Stanford, vol [6] Chapman, A., J. Saleeby, D. Wood, A. Piasecki, S. Kidder, M. Ducea, and K. Farley (2012), Regional displacement analysis and palinspastic restoration of dispersed crustal fragments in the southern sierra nevada, california, Geosphere, 8(2) [7] Clauset, A., C. R. Shalizi, and M. E. J. Newmann (2009), Power-law distribu- tions in empirical data, SIAM review, 51(4), [8] Clinton, J. F., E. Hauksson, and K. Solanki (2006), An evaluation of the SCSN 34

35 moment tensor solutions: Robustness of the m w magnitude scale, style of faulting, and automation of the method, Bull. Seis. Soc. Am., 96(5), [9] Cornet, F., J. Helm, H. Poitrenaud, and A. Etchecopar (1997), Seismic and aseismic slips induced by large-scale fluid injections, Pure appl. geophys., 150, [10] Davis, T. (1983), Late Cenozoic Structure and Tectonic History of the Western Big Bend of the San Andreas Fault and Adjacent San Emigdio Mountains, unpublished Ph.D. Dissertation, University of California, Santa Barbara, p. 578 pp [11] Davis, T. L., and M. Lagoe (1988), A structural interpretation of major tectonic events affecting the western and southern margins of the San Joaquin Valley, California, in Studies of the Geology of the San Joaquin Basin, Pacific Section, Society of Economic Paleontologists and Mineralogists, Pacific Section, Los Angeles, California, 60, 60 65, ed. S.A. Graham [12] Faulkner, D. R., C. A. L. Jackson, R. J. Lunn, R. W. Schlische, Z. K. Shipton, C. A. J. Wibberley, and M. O. Withjack (2010), A review of recent developments concerning the structure, mechanics and fluid flow properties of fault zones, J. Struct. Geol., 32(11), , doi: /j.jsg [13] Felzer, K. R., T. W. Becker, R. E. Abercrombie, G. Ekström, and J. R. Rice (2002), Triggering of the 1999 Mw 7.1 Hector Mine earthquake by aftershocks of the 1992 Mw 7.3 Landers earthquake, J. Geophys. Res., 107(B9), ESE [14] Goodman, E. D., and P. E. Malin (1992), Evolution of the Southern San Joaquin Basin and mid-tertiary transitional tectonics, central California, Tectonics, 11(3),

36 [15] Gordon, S., and H. Gerke (2009), Controls on petroleum occurrence and exploration prospectiveness in the southern San Joaquin Basin, California, Contributions to the Geology of the San Joaquin Basin, California, Pacific Section, American Association of Petroleum Geologists, Miscellaneous Publication, 48, [16] Guglielmi, Y., F. Cappa, J.-P. Avouac, P. Henry, and D. Elsworth (2015), Seis- micity triggered by fluid injection induced aseismic slip, Science, 348(6240), [17] Gutenberg, B., and C. F. Richter (1944), Frequency of earthquakes in Califor- nia, Bull. Seismol. Soc. Am., 34, [18] Hall, H. N. (1953), Compressibility of reservoir rocks, Journal of Petroleum Tech- nology, 5(01), [19] Hauksson, E., W. Yang, and P. M. Shearer (2012), Waveform relocated earth- quake catalog for Southern California (1981 to june 2011), Bull. Seism. Soc. Am., 102(5), [20] Hirst, B. (1986), Tectonic development of the Tejon and adjacent areas, Kern County, California, Pacific Section of AAPG, Guidebook, pp [21] Hirst, B. (1986), Early Miocene tectonism and associated turbidite deposys- tems of the Tejon area, Kern County, California, Pacific Section, Society of Eco- nomic Paleontologists and Mineralogists, 60, [22] Hornbach, M. J., et al. (2015), Causal factors for seismicity near Azle, Texas, Nature communications, [23] Ingebritsen, S., and C. Manning (2010), Permeability of the continental crust: 36

37 dynamic variations inferred from seismicity and metamorphism, Geofluids, 10(1-2), [24] Keranen, K., M. Weingarten, G. Abers, B. Bekins, and S. Ge (2014), Sharp in- crease in central Oklahoma seismicity since 2008 induced by massive wastew- ater injection, Science, 345(6195), [25] Lee, J., J. B. Rollins, and J. P. Spivey (2003), Pressure transient testing, vol. 9, Henry L. Doherty Memorial Fund of Aime Society of Petroleum [26] Lin, G., P. M. Shearer, and E. Hauksson (2007), Applying a three-dimensional velocity model, waveform cross correlation, and cluster analysis to locate southern California seismicity from 1981 to 2005, J. Geophys. Res., 112(B12), doi: /2007jb [27] Manga, M., I. Beresnev, E. E. Brodsky, J. E. Elkhoury, D. Elsworth, S. Ingebritsen, D. C. Mays, and C.-Y. Wang (2012), Changes in permeability caused by transient stresses: Field observations, experiments, and mechanisms, Rev. Geophys., 50(2) [28] Matthews, C., and D. G. Russel (1967), Pressure buildup and flow tests in wells, Society of Petroleum Engineers Monograph Series, 1, 163 pp [29] Meng, X., X. Yu, Z. Peng, and B. Hong (2012), Detecting earthquakes around Salton Sea following the 2010 m w 7. 2 El Mayor-Cucapah earthquake using GPU parallel computing, Procedia Computer Science, 9, [30] Meng, X., Z. Peng, and J. L. Hardebeck (2013), Seismicity around Park- field correlates with static shear stress changes following the 2003 m w 6.5 San Simeon earthquake, J. Geophys. Res., 118(7),

38 [31] Nadin, E. S., and J. B. Saleeby (2008), Disruption of regional primary structure of the Sierra Nevada batholith by the Kern Canyon fault system, California, Special Papers - Geol.L Soc. Am., 438, [32] Pettersen, Ø. (2006), Basics of reservoir simulation with the eclipse reservoir simulator, Bergen, Norway: University of Bergen, Department of Mathematics, Lec- ture Notes [33] Saleeby, J., Z. Saleeby, and L. Le Pourhiet (2013), Epeirogenic transients related to mantle lithosphere removal in the southern Sierra Nevada region, California: Part II: Implications of rock uplift and basin subsidence relations, Geosphere, 9(3), [34] Schlumberger (2012), Eclipse reservoir simulation software manual, [35] Shapiro, S. A., E. Huenges, and G. Borm (1997), Estimating the crust perme- ability from fluid-injection-induced seismic emission at the KTB site, Geophys. J. Int., 131, F15 F [36] Shearer, P., E. Hauksson, and G. Lin (2005), Southern California hypocenter relocation with waveform cross-correlation, part 2: Results using sourcespecific station terms and cluster analysis, Bull. Seismol. Soc. Am., 95(3), [37] Skoumal, R. J., M. R. Brudzinski, B. S. Currie, and J. Levy (2014), Optimizing multi-station earthquake template matching through reexamination of the Youngstown, Ohio, sequence, Earth Plant. Sc. Lett., doi: /j.epsl

39 [38] Townend, J., and M. D. Zoback (2000), How faulting keeps the crust strong, Geology, 28(5), [39] Utsu, T., Y. Ogata, and M. Ritsuko (1965), The centenary of Omori formula for a decay law of afterhock activity, Journal of Physics of the Earth, 43, [40] Zhao, L.-S., and D. V. Helmberger (1994), Source estimation from broadband regional seismograms, Bullet. Seismol. Soc. Am., 84(1), [41] Zhu, L., and D. V. Helmberger (1996), Advancement in source estimation tech- niques using broadband regional seismograms, Bullet. Seismol. Soc. Am., 86(5), [42] Zoback, M. D., and H.-P. Harjes (1997), Injection-induced earthquakes and crustal stress at 9 km depth at the KTB deep drilling site, Germany, J. Geophys. Res., 102(B8), [43] Zoback, M. D., and J. Townend (2001), Implications of hydrostatic pore pres- sures and high crustal strength for the deformation of intraplate lithosphere, Tectonophysics, 336(1),