On the Depth of Earthquakes in the Delaware Basin:

A Case Study along the Reeves–Pecos County Line

Yixiao Sheng*1 , Karissa S. Pepin2 , and William L. Ellsworth2

Abstract
Increased rates of seismicity in the Delaware basin, Texas, accompanying unconventional
petroleum development have created intensive interest in determining their cause.
Detailed and accurate spatial distribution of seismicity and focal mechanisms are critical
components for understanding the underlying industrial processes responsible for induc-
ing seismicity. We focus on a highly seismically active area straddling the Reeves–Pecos
County line where two TexNet stations sit atop the seismicity, which includes 21
ML 3+ events from 2017 to 2020 (Advanced National Seismic System Comprehensive
Earthquake Catalog). Short epicentral distance enables us to reliably estimate the hypo-
central depth using seismic phase picks and standard location methods. We use a deep-
learning-based method to detect earthquakes and time the phase arrivals. Hypocentral
locations computed in a velocity model constrained by local well data reveal that the seis-
micity concentrates between 1.5 and 2.5 km below ground in the Delaware Mountain Cite this article as Sheng, Y., Pepin, K.
S., and Ellsworth, W. L. (2022). On the
Group, the primary wastewater disposal zone at this location. Waveform inversions
Depth of Earthquakes in the Delaware Basin:
for the moment tensor and focal depth independently confirm the shallow depths. A Case Study along the Reeves–Pecos
The moment tensor solutions define critically stressed high-angle normal faults, sug- County Line, The Seismic Record. 2(1),
gesting a causal connection between injection and seismicity. 29–37, doi: 10.1785/0320210048.

Supplemental Material

Introduction mechanisms responsible for Delaware basin seismicity are

The Delaware basin of west Texas and southwestern New under debate. Both hydraulic fracturing during well stimula-
Mexico comprises the western portion of the Permian basin, tion and wastewater injection have been proposed (Lomax
and contains some of the deepest sedimentary deposits in and Savvaidis, 2019; Deng et al., 2020; Skoumal et al., 2020;
the world. It is also one of the largest oil and gas fields in Dvory and Zoback, 2021; Zhai et al., 2021).
the world and, as of 2020, was producing over 4 million barrels Seismic monitoring within the basin significantly improved
per day of oil (U.S. Energy Information Administration starting in 2017 with the installation of the TexNet array
Drilling Productivity Report 2/2021). The transformation of (Savvaidis et al., 2019). However, due to unique station siting
the Delaware basin from a field in decline to a giant is due challenges, the large area to be monitored, and complex wave
entirely to production from shale source rocks, developed propagation within the basin, the reported hypocenter depths
using horizontal drilling and “multiphase” hydraulic fracturing vary over a wide range, from very near the surface to deeper
technology. As development proceeded, the southern portion
of the Delaware basin in Texas experienced increasing rates of
1. Université Grenoble Alpes, Université Savoie Mont Blanc, CNRS, IRD, IFSTTAR, ISTerre, Grenoble,
seismicity beginning in about 2009 that continue to the present France, https://orcid.org/0000-0002-6109-0084 (YS); 2. Department of Geophysics, Stanford
(Frohlich et al., 2020; Skoumal and Trugman, 2021). Seismicity University, Stanford, California, U.S.A., https://orcid.org/0000-0002-5465-0776 (KSP);
https://orcid.org/0000-0001-8378-4979 (WLE)
is not homogeneously distributed within the basin, with most
*Corresponding author: [email protected]
events in the southern Texas portion of the basin (Savvaidis
© 2022. The Authors. This is an open access article distributed under the terms of the CC-BY
et al., 2019; Fig. 1). Although it is commonly accepted that license, which permits unrestricted use, distribution, and reproduction in any medium, provided the
the earthquakes are induced by oilfield activity, the original work is properly cited.

https://www.seismosoc.org/publications/the-seismic-record/ • DOI: 10.1785/0320210048 The Seismic Record 29

Downloaded from http://pubs.geoscienceworld.org/ssa/tsr/article-pdf/2/1/29/5519513/tsr-2021048.1.pdf

by Univ Industrial de Santander/Biblioteca user
 than 12 km, with uncertainties as large as 4–5 km (Lomax and Figure 1. Study region in the Delaware Basin, west Texas. (a) Regional map
Savvaidis, 2019). Such ambiguity poses a challenge for deter- showing county boundaries for Reeves and Pecos Counties, and seis-
micity from the TexNet catalog, (b) TexNet seismic stations, and (c) wells
mining the mechanisms responsible for inducing the seismicity in study area. See figure for explanation of the different well types.
and consequently for identifying effective mitigation measures.
This study aims to resolve the large depth uncertainty, with
a focus on a subarea in the Delaware basin where the U.S. Comprehensive Earthquake Catalog reports 21 M L ≥3 earth-
Highway 285 S crosses the Reeves–Pecos County line (Fig. 1). quakes within 5 km of the intersection and 30 within 10 km.
This small area, approximately 15 km on a side, is among the The study area was selected owing to the presence of two sta-
most seismically active in Delaware basin. In the four years tions—PB04 and PB16—that sit on top of the earthquakes.
spanning 2017–2020, the Advanced National Seismic System They enable a detailed examination of hypocentral depths.

https://www.seismosoc.org/publications/the-seismic-record/ • DOI: 10.1785/0320210048 The Seismic Record 30

Downloaded from http://pubs.geoscienceworld.org/ssa/tsr/article-pdf/2/1/29/5519513/tsr-2021048.1.pdf

by Univ Industrial de Santander/Biblioteca user
 (a) In addition, sonic logs and formation depths from three deep
0
1000
wells in the study area provide detailed control on the velocity
Ochoan salt structure from the surface to the crystalline basement (Fig. 2).
2000

Delaware
Mountain All the depths discussed in this manuscript are measured rel-
Group
ative to the ground surface.
3000

Bone Springs
Depth (m)
4000

Wolfcamp Data and Materials

Seismic data
5000

Penn Shale
Continuous seismic waveforms for the ten closest publicly
6000

available TexNet stations (Fig. 1b, Table S1) in operation from

Precambrian
1 August 2019 to 23 July 2020 are analyzed. The study period
7000

Basement

0 2 4 6 8 begins on the start date for station PB16, located near the
P-wave velocity (km/s)
(b) center of the study area. Installation of this station gives us
30

two near-source stations, the other being PB04. The waveforms

PB05 were analyzed using EQTransformer (EQT, Mousavi et al.,
25

PB19
2020), an attentive, deep-learning model designed to both
Epicentral distance (km)

detect earthquakes and time P and S arrivals. EQT analyzes

20

PB15
seismograms in a manner analogous to a human analyst, look-
15

ing for amplitude and frequency characteristics of an earth-

quake in addition to picking P and S arrivals. We set
10

PB14
PB18
relatively low probability thresholds of 0.3 for event detection
PB04
and 0.1 for both P and S phase picking, because EQT is robust
5

PB16
to false positives (Mousavi et al., 2020). EQT detected over
0

0 2 4 6 8 10 3000 earthquakes, with the requirement that each event was

Seconds after .O.T detected by at least three stations. The majority (∼85%) of
(c) these detections correspond to earthquakes outside of the study
30

area and were not considered further.

PB05
25

PB19

Subsurface data
Epicentral distance (km)
20

PB15 Within the study area the Delaware basin contains over 6 km
of sediments. Stratigraphic information was obtained from
15

DrillingInfo (now Enverus), along with technical details on

10

PB14
PB18
all wells in the area. The main stratigraphic units of interest
PB04
are the Upper Permian Ochoan Series evaporites (salt), located
5

PB16 between approximately 1.0 and 1.5 km depth below the local
land surface; the Delaware Mountain Group, the primary
0

0 2 4 6 8 10
wastewater disposal formation, between 1.5 and 2.5 km depth;
Seconds after .O.T
the Bone Springs Formation between 2.5 and 3.25 km, which
Figure 2. (a) Sonic velocities from three wells in study area. Well API 42- contains a few producing horizontal wells; and the Wolfcamp
38910429 as recorded in gray and averaged in 500 ft intervals in blue. Formation between 3.25 and 4.6 km depth where over 90% of
Averaged velocities for API 42-38910428 and API 42-38931103 in red
and green, respectively. Derived velocity model used in this study is shown
the production wells are located, all bottoming above 3.6 km
in black along with formation names. (b) Record section of seismograms below surface.
of the Mw 2.8 earthquake at 17:40 on 5 August 2020. O.T on the x-axis Between 2014 and the end of our observation period in mid-
stands for origin time. (c) Synthetic record section for the same earth-
2020, 77 horizontal wells were drilled, fracked, and put into
quake using focal parameters from Table 1. Open circles on record
section plots indicate time of first-arriving P wave. Note the extreme loss operation (Fig. 1c). These wells produced over 1.5 million bar-
of amplitude of the initial arrival at distances beyond 10 km. rels of oil and five times that amount of water, almost all of

https://www.seismosoc.org/publications/the-seismic-record/ • DOI: 10.1785/0320210048 The Seismic Record 31

Downloaded from http://pubs.geoscienceworld.org/ssa/tsr/article-pdf/2/1/29/5519513/tsr-2021048.1.pdf

by Univ Industrial de Santander/Biblioteca user
 which was disposed of by injection into the Delaware Mountain between 17 and 26 km, the first-arriving P-wave emerges so
Group. Of the 77 production wells, 8 were completed during the weakly from the noise to render its timing unreliable. Because
period for which we have seismic data. These wells locate in the we will be using first-arriving P waves to locate the earthquakes,
northeastern quadrant of the study area and are identified in we cannot rely on automatic picks made at epicentral distance
Figure 1c. Twelve disposal wells are in the study area. They beyond about 10 km and consequently restrict our hypocenter
all inject into the Delaware Mountain Group in the depth inter- analysis to the four closest stations—PB04, PB14, PB16, and
val from 1585 to 2440 m. We also obtained sonic logs for three PB18 (Fig. 1b). Stations at greater distance are used to model
vertical wells in the study area (Fig. 1c). The logged intervals low-frequency seismograms for moment tensor analysis
span the depth range from 10 to 6700 m below land surface.
Depth of Seismicity
Crustal structure and wave propagation We approach the problem of absolute depth determination
A local 1D layered velocity model was developed using sonic logs using three complementary methods: constraints from S–P
from three deep wells in the study area (Fig. 1c). Median veloc- times; iterative hypocenter solutions computed using P- and
ities computed every 500 ft (152.4 m) for each well show similar S-arrival times; and moment tensor modeling of low-frequency
depth profiles (Fig. 2a). The general trend of the velocities fea- body and surface waves.
tures a rapid increase with depth in the upper 1000 m before
jumping to approximately 6 km/s between 1000 and 1500 m. S–P times
This high velocity layer is the Ochoan salt. Beneath the salt, veloc- The observation of the S–P time interval at a single station con-
ities precipitously drop to around 4 km/s and continue to decline strains the maximum focal depth for the event in a vertically
over the next 2 km before rising in the deep part of the basin. The stratified earth model when the velocity structure is known.
lowest velocities correspond to the Wolfcamp Formation—the The earthquake can be no deeper than the depth correspond-
principal unconventional production formation in the southern ing to the S–P time from a hypothetical source located directly
portion of the Delaware basin. Basement is encountered at about below the station. We use this fact together with the 300 S–P
6 km depth in the study area. For purposes of event location, times measured by EQT at stations PB04 and PB16 and the
average compressional velocities were extracted in intervals well-controlled velocity model to explore the range of maxi-
corresponding to the formation depths, simplifying the model mum focal depths. In Figure 3a, the cumulative distribution
to 13 homogeneous layers overlying the basement (Fig. 2a of S–P times, sorted from the shortest to the longest, are com-
and Table S2). We were unable to find dipole sonic logs near pared with the corresponding depth for a source vertically
the study area. As a consequence, the shear-wave velocity is below the station. We find that 72% of the events can be
unknown. We assumed a constant V P =V S ratio, which was esti- no deeper than the base of the Delaware Mountain Group
mated as 1.89 using EQT picks. The detailed analysis is given in at 2.5 km below ground level. In addition, 98% of the events
the supplemental material available to this article (Text S1). can be no deeper than the top of the Wolfcamp at 3.25 km. If
The presence of the salt layer exerts a major impact on wave the shear-wave velocity in any of the layers were higher (lower
propagation from sources located below it to the surface. Ray V P =V S ratio) than assumed, the maximum depths would
paths for sources located below the salt would bend toward deepen slightly. The results for a constant V P =V S ratio of
the surface if the velocities increased monotonically with depth. 1.73 are also given in Figure 3a for comparison.
However, the shallow high-velocity salt layer bends rays incident It should be noted that only 1% of the measured S–P times
from below toward the horizontal upon entering the salt. require a hypocenter in the Ochoan salt, directly above the
Depending on the incidence angle, the ray refracts horizontally Delaware Mountain group, although more of such shallow
within the layer. Figure 2b,c compare record sections for the events are possible, because the S–P time only restricts the
M w 2.8 earthquake at 17:40 on 5 August 2020, with synthetic maximum depth. To better constrain the depths, it is necessary
seismograms computed using the 1D model in Table S2. The to determine hypocentral locations for individual events.
circle on each trace marks the theoretical arrival time of the
P wave. In both the data and the synthetics, an impulsive P-wave Hypocenter solutions
arrival is observed at the four closest stations to the epicenter, all The S–P time analysis shows that many of the earthquakes
within 10 km distance. At four slightly more distant stations, locate at shallow depth, within about 1 km of the base of the

https://www.seismosoc.org/publications/the-seismic-record/ • DOI: 10.1785/0320210048 The Seismic Record 32

Downloaded from http://pubs.geoscienceworld.org/ssa/tsr/article-pdf/2/1/29/5519513/tsr-2021048.1.pdf

by Univ Industrial de Santander/Biblioteca user
 (a) (b)

Ochoan salt at 1.5 km. As discussed earlier, horizontal refrac- Figure 3. (a) Cumulative distribution of S–P times from the shortest to the
tion in the salt complicates accurate timing of the initial P wave longest at the TexNet stations PB04 and PB16. Solid and dashed lines
indicate S–P travel times for vertical incidence from a source at the named
for such events, and we use only the four closest stations to
formation boundaries using different V P =V S ratios. Different colors are
locate the hypocenters. We further restrict the events to those used for different formation boundaries. (b) Depth of seismicity. The
with at least five phases. This guarantees that at least one S block histogram on the left depicts the centroid depths from the moment
phase is included in the solution. tensor analysis. The smoothed histogram shows the hypocentral depths
obtained based on seismic travel time, and the dashed line gives the
We select program VELEST (Kissling et al., 1994) to deter- TexNet depth distribution. The TexNet catalog was last accessed in 27
mine the hypocenters. VELEST traces rays in 1D models with October 2021 (Table S4).
embedded low- and high-velocity zones. Table S3 lists the
101 best-determined earthquakes, with the selection criteria
including short epicentral distance (minimum station distance lies within the Delaware Mountain Group, between 1.5 and
≤ 2 × depth), sufficient observations (≥5 phases), and small 2.5 km depth, with the remainder approximately equally dis-
root mean square error (≤0.05). Standard errors for the east, tributed between the shallower Ochoan salt and the deeper
north, and depth components of the solutions averaged 160, Bone Springs formation (Fig. 3b).
110, and 180 m, respectively (Table S3, the fifth, fourth, and
third columns from the right). The depth distribution of the Moment tensor solutions
events was estimated from the locations using a nonparametric Independent estimates of focal depth are obtained by modeling
kernel density estimator. Estimates made with the bandwidth the waveforms of both long-period body and surface waves.
set to the mean depth error or using the “rule-of-thumb” band- We select nine earthquakes larger than M L 2.5 occurring
width estimator (Silverman, 1986) produce comparable results, between August 2019 and August 2020, and determined their
the former being smoother. Approximately 75% of the density focal mechanisms and hypocentroids using the generalized cut

https://www.seismosoc.org/publications/the-seismic-record/ • DOI: 10.1785/0320210048 The Seismic Record 33

Downloaded from http://pubs.geoscienceworld.org/ssa/tsr/article-pdf/2/1/29/5519513/tsr-2021048.1.pdf

by Univ Industrial de Santander/Biblioteca user
 Table 1
Earthquake Hypocentroids, Moment Magnitudes, and Fault Planes

Origin Time Latitude Longitude Centroid Depth Strike Dip Rake

Number (yyyy/mm/dd hh:mm:ss.ss) (°) (°) (km) (°) (°) (°) Mw ML*
1 2019/08/03 03:05:08.70 31.1693 −103.2690 2.4 ± 0.1 152 82 −77 2.95 3.2

2 2019/12/28 10:54:42.52 31.1584 −103.2608 1.8 ± 0.2 146 68 −80 2.90 3.0
3 2020/01/04 18:23:33.70 31.1320 −103.2411 2.0 ± 0.2 150 70 −82 2.70 2.8

4 2020/01/30 00:04:15.38 31.1133 −103.2707 1.4 ± 0.1 326 75 −83 2.84 2.8
5 2020/01/30 15:58:50.28 31.1118 −103.2706 1.4 ± 0.1 327 74 −82 3.18 3.2

6 2020/05/21 11:50:18.28 31.1171 −103.2287 1.6 ± 0.2 326 70 −81 2.89 2.9
7 2020/07/22 17:01:33.25 31.1718 −103.2707 1.6 ± 0.1 336 63 −76 3.18 3.3

8 2020/08/05 17:14:47.84 31.1398 −103.2468 2.0 ± 0.1 166 81 −65 2.81 3.0
9 2020/08/06 14:01:38.00 31.1038 −103.2615 1.6 ± 0.1 338 68 −78 2.76 2.7

*M L is obtained from the Texnet catalog, accessed in 27 October 2021.

and paste (gCAP) method (Zhu and Ben-Zion, 2013). The for the preferred solution and moment tensor, and misfit as
gCAP method cuts the seismograms into Pnl and surface-wave a function of depth at the preferred location appears in
segments for inversion, and allows flexible time-shifts to Figure S11. Figure S12 illustrates that the centroid depths
account for imperfect Green’s functions and event locations. are stable with respect to varying epicenters. All of the events
Because we are modeling small events, we filter the waveforms have normal-faulting mechanisms striking northwest–south-
into relatively high-frequency bands to reduce the low fre- east, with one high-angle dip between 63° and 82° (Table 1
quency noise. We choose a 0.3–0.6 Hz frequency band for and Fig. 4). Centroid depths range between 1.4 and 2.6 km,
Pnl wave and 0.2–0.4 Hz for surface wave. agreeing well with the hypocenter depths discussed previously.
We solve for the earthquake hypocentroid location, in addi- All of the centroids are consistent with faulting in the Delaware
tion to the moment tensor. We set the initial epicenter indepen- Mountain Group within their uncertainties, although it should
dent of the location determined from arrival times by first be noted that the two shallowest events (events 4 and 5 in
estimating the earthquake back azimuth at nearby stations Table 1) formally locate in the Ochoan salt (Fig. 3b).
(PB04, PB14, PB16, and PB18) through analysis of the P-wave On average, the centroid depths are very similar to the
particle motion. The back azimuths are used to define a search hypocentral depths. The 5 August 2020 M w 2.8 earthquake
region divided into regular grid with a spacing of 0.2 km. The shown in the record section of Figure 2b has a hypocentral
grid spacing is smaller than the shortest wavelength considered. depth of 1.85 km, agreeing well with the moment tensor cent-
We take each grid center as a potential epicenter to perform roid depth of 2 km. The S–P time at the nearest station PB16 is
moment tensor inversion. At each grid point, we vary the focal 0.54 s, which rules out a hypocenter as deep as the Wolfcamp
depth from 1 to 3.2 km at a 0.2 km interval. The preferred hypo- formation. If the event were vertically below station PB16, it
center has the smallest misfit between the observed seismograms would locate near the base of the Bone Springs at 3.2 km depth.
and the synthetics. We limit stations considered in the inversion A depth of 3.2 km at that epicenter predicts an S–P time 0.27 s
to those within 30 km from the epicenter for surface-wave mod- longer than what is observed at station PB04.
eling and those within 12 km for the Pnl phase. The selected Pnl
and surface-wave phases are weighted equally. Discussion
Table 1 summarizes the results for the nine earthquakes. Three independent analyses of earthquake depths in our study
Details on the solutions appear in Figures S2–S10, including arrive at the same conclusion: the earthquakes occur at shallow
the comparison between observed and synthetic waveforms depth, principally within the Delaware Mountain Group,

https://www.seismosoc.org/publications/the-seismic-record/ • DOI: 10.1785/0320210048 The Seismic Record 34

Downloaded from http://pubs.geoscienceworld.org/ssa/tsr/article-pdf/2/1/29/5519513/tsr-2021048.1.pdf

by Univ Industrial de Santander/Biblioteca user
 the third is associated with the centroid locations of events
numbers 4, 5, and 9 (Table 1) to the southwest of the pro-
posed graben. Slip in their model is confined to the
Delaware Mountain Group, between 1.5 and 2.5 km below
ground level. Dip-slip magnitudes reach up to 27.5 cm,
greatly exceeding the slip we can attribute to the earthquakes.
Indeed, only about 3% of the moment of the deformation
model can be accounted for by the earthquakes. The impli-
cation is clear: most of the fault slip occurs aseismically. Pepin
et al. (2021) also noted the coincidence of saltwater disposal
wells actively injecting wastewater into the Delaware
Mountain Group near the activated faults.
Dvory and Zoback (2021) investigated the pore pressure
and geomechanical constraints on induced seismicity in the
Delaware Basin. They found the Delaware Mountain Group is
in a state of normal-faulting equilibrium for high-angle planes
aligned with the SH max direction. Pressure changes resulting
from disposal of wastewater by injection into the Delaware
Figure 4. Map showing focal mechanism solutions from moment tensor
Mountain Group are capable of inducing seismic activity,
analysis. Numbers refer to Table 1. Epicenters of the best-determined
earthquakes are shown by gray dots. Vertical deformation from and are likely responsible for the observed deformation and
Interferometric Synthetic Aperture Radar analysis of Pepin et al. (2021) is seismicity in the study area. Charzynski et al. (2019) analyzed
shown in color. Fault lines are obtained from Hennings et al. (2021), while seismic reflection data in a nearby location in the Delaware
high and moderate stand for different confidence levels. The study area is
the same as Figure 1c, with the origin at (31.0842, −103.3321).
basin. They identified high-angle, rootless faults as shallow gra-
ben features in the Delaware Mountain Group and hypoth-
esized that these features could be responsible for both
where disposal of co-produced formation water occurs by seismic and aseismic slip.
injection. We found no events in the basement or deeper than The coincidence of the seismicity, aseismic deformation
the level of production wells, and none of the best-determined (Dvory et al., 2021; Pepin et al., 2021), and wastewater injec-
hypocenters locates in the Wolfcamp Formation. Epicenters of tion to a common, restricted depth interval, where geomechan-
the earthquakes cluster along the prominent northwest-trend- ical analysis predicts northwest-striking, high-angle normal
ing subsidence lineation determined from Interferometric faults are critically stressed, strongly suggests that preexisting
Synthetic Aperture Radar (Fig. 4; Pepin et al., 2021), as do faults are being activated by reducing the effective normal
the centroids determined in the moment tensor analysis. stress via pore pressure elevation. Although the study area con-
Focal mechanism solutions show normal-faulting events on tains multiple horizontal production wells (∼3.6 km below the
northwest–southeast-striking fault planes (Fig. 4), consistent ground surface), all located below the depth of seismicity, we
with the newly mapped shallow faults (Hennings et al., 2021). found no evidence of changes in seismicity coincident with the
Their orientation and sense of slip agree well with the maxi- hydraulic fracturing of the eight wells (shown in Fig. 1c) com-
mum horizontal stress orientation in this area (SH max ; Lund pleted during the study period. As noted earlier, they are all
Snee and Zoback, 2018), which aligns with the fault-plane located in the northeastern corner of the study area, where
strike directions. we found no earthquakes. Consequently, we feel confident that
In a companion study, Pepin et al. (2021) attributed sur- hydraulic fracturing played no role in inducing these earth-
face deformation in the study area to slip on normal faults in quakes.
the Delaware Mountain Group. Their final model of vertical
and east–west deformation included three parallel, high-angle Conclusions
normal faults. Two form a graben structure aligned with We utilize three independent analyses and confirm that the
the prominent depression and seismicity in Figure 4, and earthquakes along the Reeves–Pecos county line are shallow.

https://www.seismosoc.org/publications/the-seismic-record/ • DOI: 10.1785/0320210048 The Seismic Record 35

Downloaded from http://pubs.geoscienceworld.org/ssa/tsr/article-pdf/2/1/29/5519513/tsr-2021048.1.pdf

by Univ Industrial de Santander/Biblioteca user
 The shallow seismicity is triggered by wastewater injection into through isolating densely fractured intervals correlative to
the Delaware Mountain Group. Other studies of seismicity in seismically mapped shallow Graben features in the Delaware
mountain group, SEG Global Meeting Abstracts, 4126–4141, doi:
these counties, including Savvaidis et al. (2020), Skoumal et al.
10.15530/urtec-2019-1037.
(2020) and Skoumal and Trugman (2021), also concluded from Deng, F., T. Dixon, and S. Xie (2020). Surface deformation and
a variety of lines of evidence that most of the seismicity is driven induced seismicity due to fluid injection and oil and gas extraction
by injection, without localizing it to the shallow injection zone. in Western Texas, J. Geophys. Res. 125, no. 5, doi: 10.1029/
Seismicity temporally associated with hydraulic fracturing oper- 2019JB018962.
ations also occurs (Lomax and Savvaidis, 2019; Skoumal et al., Dvory, N. Z., and M. D. Zoback (2021). Prior oil and gas production
can limit the occurrence of injection-induced seismicity: A case
2020; Trugman and Savvaidis, 2021). The question remains
study in the Delaware Basin of western Texas and southeastern
open, however, if seismicity elsewhere in Reeves and Pecos New Mexico, USA, Geology 49, no. 10, 1198–1203, doi:
Counties is as limited in depth extent as in the study area or 10.1130/G49015.1.
extends to greater depths and, particularly, into the basement. Dvory, N. Z., Y. Yang, and E. M. Dunham (2021). Rate-and-state
The challenge for ongoing research will be to determine accurate modeling of injection-induced Aseismic slip in the Delaware
focal depths. As we learned in this study, this is possible despite Basin constrains fault-zone pore pressure changes, presented at
the 2021 Fall Meeting, AGU, New Orleans, Louisiana, 13–17
challenges presented by wave propagation conditions, when sta-
December, Abstract S21A-05.
tions are located at close epicentral distances and when the Frohlich, C., C. Hayward, J. Rosenblit, C. Aiken, P. Hennings, A.
crustal model includes local well control. Savvaidis, C. Lemons, E. Horne, J. I. Walter, and H. R. DeShon
(2020). Onset and cause of increased seismic activity near
Data and Resources Pecos, West Texas, United States, from observations at the
The seismic data are from the Texas Seismological Network Lajitas TXAR Seismic Array, J. Geophys. Res. 125, no. 2,
e2019JB017737 .
(DOI: https://doi.org/10.7914/SN/TX) and is publicly available
Hennings, P., N. Dvory, E. Horne, P. Li, A. Savvaidis, and M. Zoback
on Incorporated Research Institutions for Seismology (IRIS). (2021). Stability of the fault systems that host-induced earthquakes
The well data are obtained from DrillingInfo (now Enverus, in the Delaware Basin of West Texas and Southeast New Mexico,
https://www.enverus.com, last accessed July 2020). The supple- Seism. Record 1, no. 2, 96–106, doi: 10.1785/0320210020.
mental material for this article includes descriptions on esti- Kissling, E., W. L. Ellsworth, D. Eberhart-Phillips, and U. Kradolfer
mating the V P =V S ratio, tables for the seismic stations, the (1994). Initial reference models in local earthquake tomography, J.
Geophys. Res. 99, no. B10, 19,635–19,646.
1D velocity profile, the hypocentral solutions, and TexNet
Lomax, A., and A. Savvaidis (2019). Improving absolute earthquake
earthquake catalog, and figures for the V P =V S ratio analysis location in West Texas using probabilistic, proxy ground-truth sta-
and the moment tensor solutions. tion corrections, J. Geophys. Res. 124, 11,447–11,465, doi: 10.1029/
2019JB017727.
Declaration of Competing Interests Lund Snee, J.-E., and M. D. Zoback (2018). State of stress in the
The authors acknowledge that there are no conflicts of interest Permian Basin, Texas and New Mexico: Implications for induced
seismicity, The Leading Edge 33, no. 2, 82–160.
recorded.
Mousavi, S. M., W. L. Ellsworth, W. Zhu, L. Y. Chuang, and G. C.
Beroza (2020). Earthquake transformer—An attentive deep-learn-
Acknowledgments ing model for simultaneous earthquake detection and phase pick-
The authors thank Ruijia Wang, Rob Porritt, and two anony- ing, Nat. Commun. 11, no. 2, 3952, doi: 10.1038/s41467-020-
mous reviewers for their constructive comments. This work 17591-w.
was supported by the Stanford Center for Induced and Pepin, K. S., W. L. Ellsworth, Y. Sheng, and H. A. Zebker (2021).
Shallow aseismic slip in the Delaware basin determined by
Triggered Seismicity (SCITS) and Yixiao Sheng was also sup-
Sentinel-1 InSAR, Earth Sp. Sci. Ooen Ar. doi: 10.1002/
ported by the Department of Energy (Basic Energy Sciences; essoar.10508252.1.
Award DE-SC0020445). Savvaidis, A., A. Lomax, and C. Breton (2020). Induced seismicity in the
Delaware Basin, West Texas, is caused by hydraulic fracturing and
References wastewater disposal, Bull. Seismol. Soc. Am. 110, no. 5, 2225–2241.
Charzynski, K., K. Faith, Z. Fenton, A. Shedeed, M. McKee, S. Bjorlie, Savvaidis, A., B. Young, G.-C. Huang, and A. Lomax (2019). TexNet:
and M. Richardson (2019). Delaware basin horizontal Wolfcamp A statewide seismological network in Texas, Seismol. Res. Lett. 90,
case study: Mitigating H2S and excessive water production no. 4, 1702–1715.

https://www.seismosoc.org/publications/the-seismic-record/ • DOI: 10.1785/0320210048 The Seismic Record 36

Downloaded from http://pubs.geoscienceworld.org/ssa/tsr/article-pdf/2/1/29/5519513/tsr-2021048.1.pdf

by Univ Industrial de Santander/Biblioteca user
 Silverman, B. W. (1986). Density Estimation, Chapman and Hall, Zhai, G., M. Shirzaei, and M. Manga (2021). Widespread deep
London, United Kingdom. seismicity in the Delaware Basin, Texas, is mainly driven by shal-
Skoumal, R. J., and D. T. Trugman (2021). The proliferation of low wastewater injection, Proc. Natl. Acad. Sci. Unit. States Am.
induced seismicity in the Permian Basin, Texas, J. Geophys. Res. 118, no. 20, doi: 10.1073/pnas.2102338118.
126, no. 6, doi: 10.1029/2021JB021921. Zhu, L., and Y. Ben-Zion (2013). Parametrization of general seismic
Skoumal, R. J., A. J. Barbour, M. R. Brudzinski, T. Langenkamp, and J. potency and moment tensors for source inversion of seismic wave-
O. Kaven (2020). Induced seismicity in the Delaware Basin, Texas, form data, Geophys. J. Int. 194, no. 2, 839–843.
J. Geophys. Res. 125, no. 2, doi: 10.1029/2019JB018558.
Trugman, D. T., and A. Savvaidis (2021). Source spectral properties of
earthquakes in the Delaware Basin of West Texas, Seismol. Res. Manuscript received 3 December 2021
Lett. doi: 10.1785/0220200461. Published online 27 January 2022

https://www.seismosoc.org/publications/the-seismic-record/ • DOI: 10.1785/0320210048 The Seismic Record 37

Downloaded from http://pubs.geoscienceworld.org/ssa/tsr/article-pdf/2/1/29/5519513/tsr-2021048.1.pdf

by Univ Industrial de Santander/Biblioteca user