JOURNAL OF GEOPHYSICAL RESEARCH, VOL. ???, XXXX, DOI:10.1029/, An objective method for the assessment of fluid injection induced seismicity and application to tectonically active regions in central California T. H. W. Goebel, 1 E. Hauksson, 1 F. Aminzadeh, 2 J.-P. Ampuero 1 Seismological Laboratory, California Institute of Technology, 1200 E. California Blvd., Pasadena California 91125, USA. ([email protected]) 1 Seismological Laboratory, California Institute of Technology, Pasadena, California, USA. 2 Department of Petroleum Engineering, University of Southern California, Los Angeles, California, USA. DRAFT April 12, 2015, 11:17am DRAFT
54
Embed
An objective method for the assessment of uid injection ...tgoebel/preprints/indSeism_ceCA.pdf · X - 4 GOEBEL ET AL.: INDUCED SEISMICITY CENTRAL CA signi cantly in uence the potential
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
JOURNAL OF GEOPHYSICAL RESEARCH, VOL. ???, XXXX, DOI:10.1029/,
An objective method for the assessment of fluid
injection induced seismicity and application to
tectonically active regions in central California
T. H. W. Goebel,1
E. Hauksson,1
F. Aminzadeh,2
J.-P. Ampuero1
Seismological Laboratory, California Institute of Technology,
1200 E. California Blvd., Pasadena California 91125, USA. ([email protected])
1Seismological Laboratory, California
Institute of Technology, Pasadena,
California, USA.
2Department of Petroleum Engineering,
University of Southern California, Los
Angeles, California, USA.
D R A F T April 12, 2015, 11:17am D R A F T
X - 2 GOEBEL ET AL.: INDUCED SEISMICITY CENTRAL CA
Abstract. Changes in seismicity rates, whether of tectonic or induced
origin, can readily be identified in regions where background rates are low,
but are difficult to detect in seismically active regions. We present a novel
method to identify likely induced seismicity in tectonically-active regions
based on short-range spatio-temporal correlations between changes in fluid
injection and seismicity rates. The method searches through the entire pa-
rameter space of injection rate thresholds and determines the statistical sig-
nificance of correlated changes in pumping and seismicity rates. Applying
our method to Kern County, central California, we find that most earth-
quakes within the region are tectonic, however, fluid injection contributes
to seismicity in four different cases. Three of these are connected to produc-
tive earthquake sequences, with events above M4. Each of these sequences
followed an abrupt increase in monthly injection rates of at least 15,000 m3.
The probability that the seismicity sequences and the abrupt changes in in-
jection rates in Kern County coincide by chance is only 4%. The identified
earthquake sequences display low Gutenberg-Richter b-values of ∼0.6–0.7
and at times systematic migration patterns characteristic for a diffusive
process. Our results show that injection-induced pressure perturbations can
influence seismic activity at distances of 10 km or more. The large extent of
diffusive processes in tectonically active regions may be facilitated by com-
plex local geology and faults. Our study provides the first comprehensive,
statistically robust assessment of likely injection induced seismicity within
a large, tectonically active region.
D R A F T April 12, 2015, 11:17am D R A F T
GOEBEL ET AL.: INDUCED SEISMICITY CENTRAL CA X - 3
1. Introduction
The dramatic increase in the amount of waste-water that is produced as by-product
of reservoir stimulation and hydro-carbon extraction is a growing concern for earthquake
hazard in the U.S. Much of this waste-water is reinjected into high permeable formations
via waste-water disposal (WD) wells, at times leading to fault slip and noticeable seismic
activity [e.g. Ellsworth, 2013; Frohlich and Brunt , 2013]. Most major oilfields in central
California exhibited a rapid increase in fluid-injection rates by more than a factor of 2
between 2001 and 2013 (Figure 1), however, potential seismogenic consequences have not
been studied up to now. At a regional scale, the number of earthquakes above M2 in
central California shows no apparent correlation to the change in injection rates (Figure
1). The purpose of this study is to investigate whether small-scale induced earthquake
sequences may exist, hidden within high background seismicity rates (λ0). (Note that
the term ’induced’ in our study implies an anthropogenic component involved in a seis-
micity sequence. A clear distinction between triggering and inducing earthquakes is not
attempted here. A more extensive discussion on earthquake triggering and contributions
from different types of forcing is provided in Section 6.3.)
At present, the understanding of when and how fluid injection operations induce earth-
quakes is incomplete, which makes it challenging to devise a systematic method to identify
induced seismicity. Induced seismicity is commonly assumed to be a result of pore-pressure
increase [Kim, 2013], poro-elastic loading [Segall et al., 1994], or the elastic stress changes
resulting from large volumes of injected and extracted fluids [Segall , 1989] or a combina-
tion of these effects. In addition, several studies suggest that well operational parameters
D R A F T April 12, 2015, 11:17am D R A F T
X - 4 GOEBEL ET AL.: INDUCED SEISMICITY CENTRAL CA
significantly influence the potential of inducing earthquakes. These parameters include
injection rates [Frohlich, 2012], well-head pressures [Keranen et al., 2013], total injec-
tion volumes per well and formation [McGarr , 2014], and net-production rates [Brod-
sky and Lajoie, 2013]. Local crustal conditions [Frohlich and Brunt , 2013; Van der Elst
et al., 2013], the presence of critically stressed faults [Deichmann and Giardini , 2009], and
static stress triggering [Keranen et al., 2013; Sumy et al., 2014] may further contribute to
injection-related seismic hazard. More recently, reports of likely induced earthquakes due
to high-volume fluid injection into vertically-confined aquifers above basement lithological
units are becoming more numerous [e.g. Horton, 2012; Kim, 2013; Keranen et al., 2013].
Fluid injection and migration may occur over long periods and affect large areas thereby
altering regional forcing rates and principal stress orientations [Hainzl and Ogata, 2005;
Martınez-Garzon et al., 2013; Schoenball et al., 2014]. The likelihood of inducing earth-
quakes may be further enhanced if fluids can migrate beyond the intended geological
formation, for example, observed in Ohio [Kim, 2013] and Oklahoma [Keranen et al.,
2013]. In these regions, the identification of possibly induced events is facilitated by low
background seismicity rates and few naturally-emerging seismicity clusters.
Many previous studies focused on qualitative criteria for the identification of injection-
induced seismic events in areas with low background seismicity rates such as the central
U.S. [e.g. Davis and Frohlich, 1993; Frohlich, 2012; Frohlich and Brunt , 2013; Keranen
et al., 2013; Rubinstein et al., 2014]. These criteria include the spatial-temporal prox-
imity between injection and seismic activity (including depth) and a noticeable change
in seismicity rates. In California, only few studies attempted to link earthquake activ-
ity to fluid injection that was not associated with geothermal reservoirs, highlighting the
D R A F T April 12, 2015, 11:17am D R A F T
GOEBEL ET AL.: INDUCED SEISMICITY CENTRAL CA X - 5
challenges connected to identifying induced earthquakes in tectonically active regions [e.g.
Teng et al., 1973; Kanamori and Hauksson, 1992]. The extensive fluid injection activity
and high seismic activity, recorded over more than 40 years, make central California an
excellent location to expand on and quantify previously suggested criteria.
This study is structured as follows: We first provide on overview of tectonic setting and
utilized data-sets. We then perform an exploratory examination of seismicity and injection
data to identify injection patterns that can be parameterized for an automated analysis.
We describe a new method for an uniform, quantitative assessment of potentially induced
seismicity and apply it to Kern County, central CA. Lastly, we show the migration char-
acteristics and frequency-magnitude distributions of likely induced seismicity sequences
identified with our method.
2. Tectonic setting and data
Our study is focused on central California and includes the Southern San Joaquin basin,
as well as regions of active tectonic faulting to the west, east, and south of the basin.
The tectonic deformation west of the basin is dominated by strike-slip faulting along
the San Andreas fault, while the east shows mainly normal faulting associated with the
Breckenridge fault. The tectonics south of the basin are characterized by a series of strike-
slip and thrust faults, including the Wheeler Ridge, Pleito and White Wolf faults [e.g.
Hardebeck , 2006; Yang and Hauksson, 2013; Unruh et al., 2014]. The latter produced the
largest magnitude earthquake within the study area, the 1952 Mw7.5 Kern County event
(Figure 2). The recent, larger magnitude earthquakes to the north and south of the basin,
none of which substantially changed the long-term background seismicity rates within the
basin, are highlighted in Figure 1.
D R A F T April 12, 2015, 11:17am D R A F T
X - 6 GOEBEL ET AL.: INDUCED SEISMICITY CENTRAL CA
We analyzed the Advanced National Seismic System earthquake (ANSS) catalog be-
tween 1975 and 2014 which includes ∼17,000 events from both the Southern and Northern
California Seismic Networks. The completeness magnitude, Mc, of the ANSS catalog is
∼2.0 within the region, determined by inverting for the lower magnitude cut-off that min-
imizes the misfit between observed and modeled power-law distributions [Clauset et al.,
2009; Goebel et al., 2014a]. Mc varies in space and time, likely reaching higher values
to the north and within the central-part of the basin where the station density is low.
However, the influence of Mc variations on the present study is small because we compare
relative changes in seismicity characteristics on small spatial-temporal scales as discussed
below. In addition to the ANSS catalog, we also use the waveform relocated catalog by
Hauksson et al. [2012] to obtain more accurate relative location-uncertainties and focal-
depth-estimates in areas of denser station configurations south of the San Joaquin basin.
Within the basin, large azimuthal gaps and event-station distances complicate depth-
estimates which is discussed in more detail in the Suppl. Material Section S4.
In the following, we correlate changes in seismicity and fluid injection rates (Vinj) in
waste-water disposal (WD) wells. The latter have been archived by the Division of Oil,
Gas, and Geothermal Resources (DOGGR) of the California Department of Conservation
since 1977 1. In addition to Vinj, the DOGGR also archives well-head pressures. The
pressure data were not included in the present study for two reasons: (1) The well-head
pressure record is fragmentary and the statistical basis of the monthly documented values
are not disclosed by the well-operator and may vary for different wells. (2) The well-head
pressures may not be representative of reservoir pressures which are strongly influenced
by structural heterogeneity [e.g. Hsieh and Bredehoeft , 1981].
D R A F T April 12, 2015, 11:17am D R A F T
GOEBEL ET AL.: INDUCED SEISMICITY CENTRAL CA X - 7
The present analysis encompasses ∼1400 WD wells in the study area south of the
Kern county border (see Figure 2). Kern County is the largest oil-producing county in
California, contributing more than 75% of California’s total oil production and hosting
more than 80% of the production and hydraulically fractured wells. Here, each barrel
of produced oil is accompanied by ∼5–15 barrels of produced water [CA Department of
Conservation, 2012] highlighting that large-volumes of waste-water have to be re-injected
regularly. In addition, water disposal can also include fluids from secondary reservoir
stimulation like hydraulic fracturing. In the following, Vinj is given in thousands of barrel
per month (kbbl/mo), and 1 kbbl/mo is equivalent to ≈160 m3/mo.
3. Initial observations of M > 3 earthquakes close to fluid injection wells
We performed a preliminary, exploratory examination of the injection and seismicity
data resulting in an initial set of possibly induced earthquakes in Kern County. This
preliminary data set was then used to guide the formulation of a more rigorous identifi-
cation algorithm in Section 4. For the initial and following analysis, we chose earthquake
sequences containing events above a target magnitude of MT = 3, which are events that
can be felt by the nearby population.
During the exploratory data examination, we identified eight regions containing earth-
quakes above MT with epicenters located within oil-field boundaries or within a 10 km
radius of active injection wells (Figure 2). For each earthquake above MT, we deter-
mined whether a well within 10 km of the epicenters had fluid injection rates (Vinj) above
100 kbbl/mo at the time of the event. The choice for the initial injection rate thresh-
old of 100 kbbl/mo was guided by previous observations of likely induced seismicity [e.g.
Frohlich, 2012]. We then visually examined the time series for possible correlations be-
D R A F T April 12, 2015, 11:17am D R A F T
X - 8 GOEBEL ET AL.: INDUCED SEISMICITY CENTRAL CA
tween seismicity rates (λ) within a 10 km radius from the well and Vinj. This examination
revealed a variety of episodes of changes in injection rates within the spatio-temporal
proximity of seismic activity, which generally showed two types of characteristics.
1) We observed earthquakes above MT that were associated with local injection maxima
(e.g. in 1991 and 2000 in Figure 3a). These maxima in injection rates were preceded by
several months to years of sustained increase in injection rates. In addition, one region
showed a long-term positive correlations between λ and Vinj superimposed on the short-
term correlation (Figure 3b).
2) We observed seismic activity close to injection sites with abrupt jumps in Vinj without
long-term sustained increase. Such a site is for example located at the southern end of
the Lost Hills oil field which in turn is located to the north-west of Kern County (Lh in
Figure 2). The injection well Lh was active between 1986 and 1995, and displays a strong
short-term correlation between a rapid increase in injection rates in February 1988 and
a ML4.2 earthquake sequence that occurred at a distance of less than ∼0.5 to 1.5 km on
February 22, 1988.
In the following, we use these preliminary observations to guide the formulation of more
rigorous identification algorithm which includes a search through the entire parameter
Sumy, D. F., E. S. Cochran, K. M. Keranen, M. Wei, and G. A. Abers (2014), Observations
of static Coulomb stress triggering of the November 2011 M5.7 Oklahoma earthquake
sequence, J. Geophys. Res., doi:10.1002/2013JB010612, (in press).
Teng, T. L., C. R. Real, and T. L. Henyey (1973), Microearthquakes and water flooding
in Los Angeles, Bull. Seism. Soc. Am., 63 (3), 859–875.
Unruh, J., E. Hauksson, and C. H. Jones (2014), Internal deformation of the southern
Sierra Nevada microplate associated with foundering lower lithosphere, California, Geo-
phere, 10, doi:10.1130/GES00936.1.
Utsu, T. (1992), On seismicity, in Report of the Joint Research Institute for Statistical
Mathematics, vol. 34, pp. 139–157, Institute for Statistical Mathematics, Tokyo.
Van der Elst, N. J., and E. E. Brodsky (2010), Connecting near-field and far-field earth-
quake triggering to dynamic strain, J. Geophys. Res., 115 (B7).
Van der Elst, N. J., H. M. Savage, K. M. Keranen, and G. A. Abers (2013), Enhanced
remote earthquake triggering at fluid-injection sites in the midwestern United States,
D R A F T April 12, 2015, 11:17am D R A F T
X - 38 GOEBEL ET AL.: INDUCED SEISMICITY CENTRAL CA
Science, 341 (6142), 164–167.
van Stiphout, T., and Zhuang, J., and Marsan, D. (2012), Seismicity declustering, commu-
nity online resource for statistical seismicity analysis, 10 doi:10.5078/corssa-52382934.
Wiemer, S., and M. Wyss (1997), Mapping the frequency-magnitude distribution in as-
perities: An improved technique to calculate recurrence times?, J. Geophys. Res., 102,
15,115–15,128.
Yang, W., and E. Hauksson (2013), The tectonic crustal stress field and style of faulting
along the Pacific North America Plate boundary in Southern California, Geophys. J.
Int., 194 (1), 100–117.
Zhang, Y., et al. (2013), Hydrogeologic controls on induced seismicity in crystalline base-
ment rocks due to fluid injection into basal reservoirs, Groundwater, 51 (4), 525–538.
D R A F T April 12, 2015, 11:17am D R A F T
GOEBEL ET AL.: INDUCED SEISMICITY CENTRAL CA X - 39
2001
1983Coalinga
Parkfield2004
Wheeler Ridge2005
Cent
ral C
alifo
rnia,
M≥2
Central U
S, M≥3
Injection Rate central CA
Figure 1: Comparison of seismic activity above M2 between central California (red
curve) and the central U.S. (green curve). The dashed lines represent the average seismic-
ity rates before 2001, and the blue curve shows the cumulative waste-water injection rate
in millions of barrels (Mbbl instead of industry-standard MMbbl) in central California.
Major historic earthquakes in greater Kern County after 1980 are highlighted by arrows
and include the 1983 Mw6.4 Coalinga, the 2004 Mw6.0 Parkfield and the 2005 Mw4.6
Wheeler Ridge earthquakes. The cumulative earthquake numbers for central California
exclude the areas of the Coalinga and Parkfield main- and aftershocks. Many oilfields in
central California shows a strong increase in injection activity after ∼2001, however, an
increase in seismic activity is conspicuously absent.D R A F T April 12, 2015, 11:17am D R A F T
X - 40 GOEBEL ET AL.: INDUCED SEISMICITY CENTRAL CA
Kern CountyTulare County
OilfieldsPFE
CLE
KCE
SAF
SAFSAF
SAF-P
GFWWF
WRF
BRF
SAF-C
PF
LhKrBv
Tj
Ed
Ms
Bu
Eh
Figure 2: Seismicity above M3 (squares and stars), faults (red lines) and oilfields (green
polygons) within San Joaquin Basin. Major faults include the Wheeler Ridge (WRF),
Pleito (PF), Garlock (GF), White Wolf (WWF), and Brecken Ridge faults (BRF) as well
as the Parkfield (SAF-P) and Carrizo (SAF-C) segment of the San Andreas fault. The
three largest earthquakes (red beach-balls) are the 1952 Mw7.5 Kern County (KCE), the
2004 Mw6.0 Parkfield (PFE), and the 1983 Mw6.4 Coalinga event (CLE). The regions
with M > 3 earthquakes within 10 km of WD wells close to 8 active oil fields, i.e., the
Edison (Ed), Kern River (Kr), Bellevue (Bv), Lost Hills (Lh), Elk Hills (Eh), Buena
Vista (Bu), Midway Sunset (Ms) and Tejon (Tj) fields, are highlighted by blue circles.
Magnitudes here and in the following figures are local magnitudes. The following analysis
is limited to the area of Kern County within the red polygon. The larger region in this
figure is displayed to capture some of the major historic seismic activity.
D R A F T April 12, 2015, 11:17am D R A F T
GOEBEL ET AL.: INDUCED SEISMICITY CENTRAL CA X - 41
0.8
0.4
0.0
a
b
Bv S
eism
icit
y R
ate
[10-
3 km
-2m
o-1
]
Figure 3: a) Injection rates, Vinj (blue) and seismicity rates, λ (black) for events above
M2 and within 10 km of the well as well as events above M2.5 (colored squares). The
vertical positions of the squares correspond to relative differences in magnitudes. λ and
Vinj display a noticeable long-term correlation. b) Temporal variations in cross-correlation
coefficients for Vinj and λ within a sliding time window of 2 yr and a 10 km radius from
injection site Bv in Figure 2. Higher frequency variations in seismicity and injection rates
were removed using a 5-point median-filter, prior to correlating the two time series.
D R A F T April 12, 2015, 11:17am D R A F T
X - 42 GOEBEL ET AL.: INDUCED SEISMICITY CENTRAL CA
Injection Rate
CumulativeEarthquakes
Lh
Figure 4: Injection rates (blue), cumulative earthquake number (black) and earthquake
magnitudes (colored squares) within a 10 km radius of injection well Lh. (See also Figure
10c for short term behavior close to injection).
D R A F T April 12, 2015, 11:17am D R A F T
GOEBEL ET AL.: INDUCED SEISMICITY CENTRAL CA X - 43
Table 1: Identification criteria for likely induced seismicity sequences.
Description of criteria Parameter Section
1 Episodes of changes in injection rates and seismic ac-tivity above MT are associated in space and time.
r, ∆t 4.3
2 The probability that the number of events above MT
are part of the background seismicity is low. Thisprobability is computed for events above MT withinr, and ∆t based on the background rates within theregion.
Ppoi 4.4
3 The probability that episodes of changes in injectionrates coincide with earthquakes above MT by chanceis low, given the local injection and seismic activity.
Pran 4.5
4 The seismicity rate is significantly higher after, com-pared to before an episode of injection rate change.Seismicity rate changes are determined using the R-statistic of Felzer and Brodsky [2005] and Van der Elstand Brodsky [2010].
R 4.6
A detailed description of each criteria and connected parameters can be found in the
sections listed in the last column. The here described identification method of likely
induced seismicity is referred to as ’Objective Induced Seismicity Correlation (OISC)
Method’. A step-by-step summary of the algorithm is provided in Section 4.7 and in a
flow chart within the Appendix.
D R A F T April 12, 2015, 11:17am D R A F T
X - 44 GOEBEL ET AL.: INDUCED SEISMICITY CENTRAL CA
Trigger b100 kbbl/mo
Trigger a100 kbbl a a b
Figure 5: Example of injection rates at a WD well that can be categorized into two
type-a triggers and one type-b trigger. Type-a triggers are defined as a gradual increase
in Vinj whereas type-b triggers are defined as an abrupt in monthly injection rates (see
text for details).
D R A F T April 12, 2015, 11:17am D R A F T
GOEBEL ET AL.: INDUCED SEISMICITY CENTRAL CA X - 45
BackgroundEvents
PossiblyInducedEvents
Figure 6: Seismicity-density of stacked, candidate induced earthquake sequences, 300
days before and after injection rate peaks of wells with the highest spatial-temporal cor-
relations of the initial analysis scheme. The background seismic activity is highlighted
by black markers, possibly induced events by green dots as well as red circles (M > 3)
and rectangles (M >4). The criteria for initial candidate event selection are described in
Section 3 . The ML4.2 event close to Lh (see Figure 4) is easily recognizable at the origin
of injection.
Event beforeTrigger
Event afterTrigger
Trigger
t1 t2
Figure 7: Sketch of the time-interval ratio computed from the periods between the
trigger-onset and the last earthquake before and after.
D R A F T April 12, 2015, 11:17am D R A F T
X - 46 GOEBEL ET AL.: INDUCED SEISMICITY CENTRAL CA
Total Number of WD Wells
3 detections 4 detections
Wel
ls w
ith S
eism
ic A
ctivi
ty
Wel
ls w
ith In
jecti
on A
ctivi
ty
Injection Rate [kbbl/mo] Injection Rate [kbbl/mo]
Type-a Trigger Type-b Trigger
Figure 8: Number of wells that show type-a (left) and type-b (right) triggers. The
total number of WD wells is ∼1400 (red dashed line) . The number of injection sites that
also exhibit nearby seismic activity is highlighted by black, dashed curves. The number
of injection sites associated with earthquake sequences according to the OISC-method
are highlighted by green curves. The dashed horizontal lines highlight the number of
wells that were consistently correlated with earthquake sequences over the widest range
of injection thresholds. Note that the red dashed lines and red curves are plotted at a
different scale represented by the right y-axis.
D R A F T April 12, 2015, 11:17am D R A F T
GOEBEL ET AL.: INDUCED SEISMICITY CENTRAL CA X - 47
Lh88Kr87
Kr85
Tj05
Lha,b2
Lhb
Lha1
KrbKra
Tjb
Krb
Kr87
Kr85
Bv
1
SHmax
Time from Trigger [yr]
Figure 9: Locations of injection wells with type-a (black-bordered triangles) and
b (white-bordered triangle) triggers and nearby seismicity. Mainshocks above M3 of
earthquake sequences that were identified by the objective induced seismicity correlation
method are labeled according to injection site and year of trigger occurrences. Seismicity
(dots) is colored according to time after trigger-onset and events above M3 are highlighted
by red squares (see legend). The orange markers highlight the region of long-term corre-
lation close to injection site Bv. Inset: Temporal migration of Kr85 and Kr88 seismicity
sequences. Seismicity is colored according to time from trigger-onsets at the closest injec-
tion sites. SmaxH orientation is taken from Yang and Hauksson [2013]. The injection rates
for wells Bv and Lh1b are shown in Figure 3 and Figure 4.
D R A F T April 12, 2015, 11:17am D R A F T
X - 48 GOEBEL ET AL.: INDUCED SEISMICITY CENTRAL CA
Table 2: Results of probabilistic assessment for well locations and nearby seismicity