8

Downloaded from https://agupubs.onlinelibrary.wiley.com/doi/ by Mit Libraries Serials & Journa, Wiley Online Library on [02/11/2022].
See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
8
High-Resolution Shallow Structure at Brady Hot Springs Using

Ambient Noise Tomography (ANT) on a Trenched
Distributed Acoustic Sensing (DAS) Array
Xiangfang Zeng1,2, Clifford H. Thurber2, Herbert F. Wang2, Dante Fratta3, and Kurt L. Feigl2
ABSTRACT
An ~8700 m fiber-optic cable is installed in surface trenches at the Brady Hot Springs geothermal site in Nevada,
USA. A distributed acoustic sensing (DAS) system is applied to use this fiber-optic cable to record seismic ambi-
ent noise. Noise cross-correlation functions (NCFs) of channel pairs on one linear DAS segment and on two in-
line DAS segments are obtained, from which the Rayleigh wave signal emerges. The dispersive group and phase
velocities in a high-frequency band of 2–18 Hz are measured with two different methods (multichannel analysis of
surface waves (MASW) and frequency-time analysis (FTAN)) and then used to invert shear wave velocities at
shallow depths. The obtained velocity model successfully reveals low-velocity zones (LVZs) in warm ground
areas. Our results demonstrate that using a DAS array with ambient noise tomography (ANT) to image
near-surface seismic structure is feasible.
8.1. INTRODUCTION and anisotropy) are sensitive to lithology, including

porosity, fracture network, and other reservoir character-
Seismic tomography is one of the most important istics (e.g., Vécsey et al., 1998; Zhang & Lin, 2014), seis-
approaches for imaging Earth’s interior. It has been mic tomography can be effectively applied in geothermal
widely used in seismic studies and resource exploration, reservoir imaging. One of the key factors of seismic
especially in the oil and gas industry. The potential of seis- tomography is that resolution is strongly dependent on
mic tomography in geothermal reservoir imaging and seismic ray coverage. Recently, dense arrays have been
monitoring has also been demonstrated in numerous stud- used to provide high-resolution images of volcanic and
ies. Since the seismic velocity parameters (Vp, Vs, Vp/Vs, geothermal systems (e.g., Lehujeur et al., 2018; Wang
et al., 2017). Previous studies reveal that low-velocity
anomalies correlate well with surface geology at shallow
1
State Key Laboratory of Geodesy and Earth’s Dynamics, depths, whereas porosity plays an important role at
Innovation Academy for Precision Measurement Science and greater depths (Lehujeur et al., 2018).
Technology, Chinese Academy of Sciences, Wuhan, China To further understand mechanisms and evaluate poroe-
2
Department of Geoscience, University of Wisconsin– lastic properties of a geothermal reservoir, a three-part
Madison, Madison, Wisconsin, USA seismic acquisition system was deployed at the Brady
3
Department of Civil and Environmental Engineering, Univer- Hot Springs geothermal field in Nevada for 15 days in
sity of Wisconsin–Madison, Madison, Wisconsin, USA March 2016 (Feigl & PoroTomo Team, 2018). In general,
Distributed Acoustic Sensing in Geophysics: Methods and Applications, Geophysical Monograph 268, First Edition.
Edited by Yingping Li, Martin Karrenbach, and Jonathan B. Ajo-Franklin.
© 2022 American Geophysical Union. Published 2022 by John Wiley & Sons, Inc.
DOI:10.1002/9781119521808.ch08
101
Downloaded from https://agupubs.onlinelibrary.wiley.com/doi/ by Mit Libraries Serials & Journa, Wiley Online Library on [02/11/2022]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
102 DISTRIBUTED ACOUSTIC SENSING IN GEOPHYSICS
both active source (controlled explosions, vibroseis 1600

trucks, etc.) and passive source (naturally occurring noise 18B-31
18D-31
or cultural noise) can be used for tomography. Active-
1200
source data consisting of body-wave and surface-wave 81B-1
signals excited by a vibroseis truck have been analyzed N
and used in previous tomography studies (Matzel et al., 800
Y (m)
2017; Parker et al., 2018; Song et al., 2018). One pas-
sive-source method is based on the analysis of continuous
400
ambient noise that is generated by natural and anthropo-
genic activities, such as ocean waves, wind, and traffic.
Although the origin of ambient noise tomography 0 56A-1 56-1
(ANT) came into existence as early as the 1950s, its great 46-1
47C-1
progress was achieved in the 2000s with more available
high-quality continuous records (Campillo & Roux, –800 –400 0 400 800 1200
X (m)
2015). The ANT technique makes it possible to construct
high-resolution models of shallow shear wave structures Figure 8.1 Seismic acquisition systems in the Brady Hot Springs
without costly active-source operation (Dou et al., 2017; geothermal site in Nevada. The trenched DAS cable is shown by
Zeng, Lancelle, et al., 2017). The feasibility of using the blue line. Black solid circles denote geophones. Wells are
DAS for ANT has been tested successfully with a small denoted by various colored solid circles: Injection wells (red),
data set acquired in September 2013 at a test site in Garner observation wells (green), and production wells (blue). Gray
Valley, California (Zeng, Lancelle, et al., 2017). lines represent faults in Faulds et al. (2017). The green open
In this study, we use continuous records from a surface circles denote the channel pair on two in-line segments,
DAS array to compute high-quality NCFs of channel whereas the red ones denote the channel pair along one
segment. The highway and service road are shown by a black
pairs on individual linear segments and on pairs of two
dashed line and a pink line, respectively.
in-line, but separated, segments. The Rayleigh wave sig-
nal is analyzed with MASW and FTAN methods to
obtain frequency-dependent group and phase velocity
measurements. Then, the shallow shear wave structure The DAS array was installed very near the surface of a
is inverted from the group and phase velocities. The 1500 × 500 × 400 m target volume (natural laboratory) on
inverted model is compared with models obtained from the main area of subsidence, in trenches that are about
other methods and surface geology. 30–50 cm deep and backfilled with loose soil. Since
DAS is sensitive to the strain along the axis of the fiber-
optic cable, geometry was designed in a zigzag shape to
8.2. DATA AND METHODS capture signals from vibroseis sources from different
directions as much as possible (Figure 8.1). However,
The power plant at the Brady Hot Springs geothermal the channels around the corner suffered interference from
field has been operating since 1992. Cold fluid is injected two segments, we deleted 10 channels around each corner.
at about 300 m depth in two wells in the northeast part of Gauge length was set at 10 m with a channel spacing of
the site (Figure 8.1), and the production wells are situated 1 m (Parker et al., 2014). To reduce storage and compu-
in the southwest part. An elliptical area of subsidence was tational cost, the raw data were resampled from 1000 to
observed with satellite interferometric synthetic aperture 100 Hz.
radar, and inverse modeling suggests it is the result of vol- According to a power spectral density map from the
umetric contraction at shallow depths (Ali et al., 2016). geophone array, there are three dominant ambient noise
The seismic acquisition systems deployed at the Brady sources at the Brady Hot Springs geothermal site, which
Hot Springs geothermal field included a surface geophone include traffic on the highway, traffic on a service road
array consisting of 238 short-period three-component sen- crossing the target area from southwest to northeast,
sors, an ~8600 m DAS array installed in surface trenches, and pumping noise from the injection wells (Zeng, Thur-
and an ~350 m DAS array installed in an observation well. ber, et al., 2017). Due to the varying coupling between the
Two seismic tomography techniques with different DAS cable and the soil, the noise level across the DAS
sources were planned. The active source was a 440 kN array is complicated (Figure 8.2). For example, one seg-
force vibroseis truck that occupied 191 locations within ment (x ~ 300 m, y ~ 900 m) crossing a ditch was exposed
and surrounding the DAS array during the four stages to the air, and the dangling of the cable generated an
of the experiment. Planned passive source included traffic extremely strong noise signal. The interrogator was
and more general ambient noise. installed in a mobile laboratory that was a modified
HIGH-RESOLUTION SHALLOW STRUCTURE ON A TRENCHED DAS ARRAY 103
(a) (b)
CH 0100 ⨯10–6
0.01 1
Magnitude (dB)
0.005 0.8
0.6
0
0.4
–0.005 0.2
–0.01 0
0 5 10 15 20 25 30 0 10 20 30 40 50
Time (s) Frequency (Hz)
(c) 5 Hz (d) 10 Hz (e) 15 Hz

1600 1600 –90
1400 1400 –100
1200 1200 –110
1000 1000
–120
800 800
Y (m)
dB
–130
600 600
–140
400 400
–150
200 200
–160
0 0
–170
0 200 400 0 200 400 0 200 400
X (m) X (m) X (m)
Figure 8.2 (a) Thirty seconds of raw data recorded by Channel 0100 in the southwest part of the study area. (b) The
spectrum of the waveform shown in (a). (c)–(e) Power spectral density of every 15th channel at 5, 10, and 15 Hz.
shipping container. The shaking of the mobile laboratory functions over a long time period are stacked to obtain
due to door closing, chairs moving, and other activities final NCFs.
also caused strong coherent signals across the array. Signal-to-noise ratio of the NCFs increases steadily
The strongest noise is in the range of 5–35 Hz, and several with increasing length of the stacking time period and
spectral peaks are also observed in the spectrum. then converges at some point that is the optimal stacking
Several standard processing steps are applied to the raw time period (Seats et al., 2012). One straightforward way
data (Zeng, Lancelle, et al., 2017; Figure 8.3). As men- to determine the optimal stacking time period is using a
tioned earlier in this chapter, the shaking of the interroga- cross-correlation analysis. The NCF of a long time period
tor causes a coherent noise signal across the whole array, is used as the reference trace. Since traffic is one of the
the array mean and segment mean are subtracted at first. most important noise sources, the individual NCFs show
Strong transient signals are reduced with temporal and diurnal variations (Zeng, Lancelle, et al., 2017). We use
frequency domain normalization (Bensen et al., 2007). the NCFs of 10 hours, including daytime and nighttime,
Then, the processed waveforms of two receivers in a short as reference traces. The cross-correlation coefficients
time window (30 s) are used to compute individual NCFs. between NCFs of time periods ranging from 1 to 10 hours
To remove the effect of uneven source distribution and and the reference trace are computed. The point at which
enhance signal-to-noise ratio, individual cross-correlation the cross-correlation coefficient curve reaches a particular
Array mean removement Figure 8.4a). The second example is Channel Pair
CH0110-CH0640 on two in-line segments (Figure 8.1),
Decimation
in this case, and it is parallel to the highway. The conver-
gence rate is similar to the first example (Figure 8.4b).
Therefore, we chose 4 hours (half during nighttime and
Cutting into short time windows
half during daytime) as the stacking time period for all
channel pairs used in this study.
Time domain normalization Although the body-wave signal in NCFs has been
reported (e.g., Zhan et al., 2010), the surface-wave signal
Frequency domain normalization is generally much clearer and its dispersion curve or wave-
form can be used as the observed data in tomography
Cross-correlation function (e.g., Lee et al., 2015; Shapiro et al., 2005). After extract-
computation ing the surface-wave signal, the surface-wave phase and/
or group velocity dispersion curves need to be determined.
tf-PWS stacking The MASW method, based on slant stacks in the fre-
quency domain, is utilized to determine phase velocity
between channels along the same segment of the fiber-
MASW analysis
optic cable (denoted by red open circles in Figure 8.1).
Group velocity is measured for channels on two in-line
Vs model inversion segments (denoted by green open circles in Figure 8.1)
with multiple filtering technique (MFT), which has been
Figure 8.3 The data preparation and modeling process used in
widely used since the 1960s (Dziewonski et al., 1969).
this study.
A series of narrow band-pass filters that are defined as
Gaussian functions in the frequency domain are applied
to the waveform, and then travel times of the maximum
threshold is chosen. Two examples of different channel energy for each frequency are picked for group velocity
pairs are shown in Figure 8.4. The first example is Chan- computation. Group velocities are computed by channel
nel Pair CH0462-CH0580 (Figure 8.1) that is on the same separation divided by the respective travel times. The dis-
segment, which is perpendicular to the highway. The persion curves obtained from these steps are used to deter-
cross-correlation coefficient reaches over 0.8 when the mine layered shear wave velocity structures using two
number of stacked traces exceeds 450 (~3.75 hours; linearized inversion methods for different dispersion
CH 0462 - 0580 CH 0462 - 0580 CH 0110 - 0640 CH 0110 - 0640

(a) 1 (b) (c) 1 (d)
0.9 0.9
1140 1140
0.8 1026 0.8 1026
0.7 912 912
0.7
798 798
Cross Coef.
Cross Coef.
0.6 684 0.6 684

0.5 570 0.5 570
456 456
0.4 0.4
342 342
0.3 228 0.3 228
0.2 114 114
0.2
0.1 0.1
0 0
0 500 1000 1500 –4 –2 0 2 4 0 500 1000 1500 –4 –2 0 2 4
# of stacked traces Time (s) # of stacked traces Time (s)
Figure 8.4 (a) Correlation coefficient to the reference trace vs. the number of stacked traces for Channel Pair
CH0462-CH0580 (denoted by red open circles in Figure 8.1). (b) The reference NCF (blue), time frequency-
phase weight stacking NCF (red), and different time period NCFs (black). (c) and (d) Correlation coefficient to
the reference trace vs. the number of stacked traces for Channel Pair CH0110-CH0640 (denoted by purple
open circles in Figure 8.1).
curves. The final three-dimensional (3D) model is interpo- Distance bin stacking
lated from multiple layered models.
8.3. NCF RESULTS 419.4
The Rayleigh wave signal can be observed on both ver- 419.4

tical-vertical and radial-radial component NCFs of tradi-
tional three-component seismic sensors, where radial 419.5
direction is along the line connecting the two receivers
in a laterally homogenous structure. Since each DAS 419.5
channel is equivalent to one horizontal component geo-
phone, there are two possible channel pairs that could 419.6
be used: two channels on the same linear segment or chan-
419.6
nels on two in-line segments with cable orientations that
are the same as the path between the two channels. The 419.7
first case includes 61 segments, whereas the second case
includes 51 segment pairs in total. 419.8
One example of a channel pair on one linear segment is
shown in Figure 8.5. The NCF shows strong asymmetry 419.8
that reflects more energy coming from the highway side.
The Rayleigh wave signal clearly emerges, and the strong- 419.9
est energy propagates with a velocity of ~300 m/s. For the
419.9
channel pairs on two in-line segments, the greater offset
makes it difficult to extract a clear signal due to strong 420.0
scattering. Therefore, the array interferometry technique,
which has been successfully used to extract weak signals 420.1
from NCFs (Lin et al., 2013; Nakata et al., 2015), is
adopted to improve signal quality. This method requires
computing NCFs of all possible receiver pairs across
arrays (segments in our case) and then stacking NCFs
of receiver pairs in particular distance bins. To reduce
computational cost, only 10% of all channel pairs between
two in-line segments are computed and the NCFs in 10 m
–4 –2 0 2 4
distance bins are stacked into one trace. Figure 8.6 shows
Time (s)
DASH CC SEG CH0462 - 0588 Figure 8.6 Individual NCFs (black) in a distance bin and the
120 Highway side stacked trace (red). The number on each trace indicates the
offset in meters.
100
80 NCFs of 13 channel pairs with the offset ranging from

Offset (m)
419.4 to 420.1 m, and the stacked traces. Rayleigh wave

60 signals on both negative and positive lags are enhanced,
while random noise is substantially suppressed. The
40 cross-segment NCFs in the southeast part of the study
area are shown in Figure 8.7. The Rayleigh wave signal
20
can be traced up to 600 m, making it possible to measure
dispersion information in a lower frequency band.
–2 –1 0 1 2
Time (s)
8.4. DISPERSION MEASUREMENT RESULTS
Figure 8.5 Record section of NCFs between channel pairs
along one segment (denoted by red open circles in The MASW method is used to measure Rayleigh wave
Figure 8.1). The dashed red lines denote a velocity of 300 m/s phase velocities from NCFs of channel pairs along single
and one sample trace is shown in blue. linear segments. Since the MASW method measures the
600
550
500
Offset (m)
450
400
350
300
250
–4 –2 0 2 4
Time (s)
Figure 8.7 Record section of NCFs between channel pairs along two in-line segments (denoted by green open
circles in Figure 8.1). The dashed red lines denote a velocity of 300 m/s and sample traces are shown in blue.
phase difference between receivers, phase velocity is con- compare with the result obtained with active source sig-
trolled by the seismic structure beneath the segment. nals (Song et al., 2018). Dispersive trends agree well
The denser spacing theoretically reduces aliasing in the and the difference between the two results is less than
wavenumber domain, meaning it is possible to see rela- 25 m/s. The histogram shown in Figure 8.8b suggests that
tively high-frequency components. On the other hand, the number of picks reaches a peak around 10 Hz, corre-
the length of the segment also limits both the resolution sponding to a wavelength of about 30 m.
in the wavenumber domain and the maximum wavelength The distance bin stacked NCFs mainly reflect average
that controls the maximum depth of investigation. An seismic structure of the whole path rather than that
empirical criterion is the maximum wavelength is about beneath a given segment. The phase velocity measured
twice the segment length, whereas the maximum depth with the MASW method will be affected by the structures
of investigation is very close to the segment length beneath both segments, which has been averaged by dis-
(Park & Carnevale, 2010). One MASW picking example tance bin stacking. Therefore, the multiple-filter tech-
is shown in Figure 8.8. The MASW picked dispersion nique is used to measure group velocities. Similar to the
curves for the profile shown in Figure 8.5 were used to MASW method, the greater offset makes it possible to
(a) (b)
600 250
2500
500 200
2000
Phase velocity (m/s)
400 150
Energy
1500
Count
300 1000 100
200 500 50
100 0
0 5 10 15 4 6 8 10 12 14 16 18
Frequency (Hz) Frequency (Hz)
Figure 8.8 (a) One MASW measurement example. The color represents stacking energy, and the picked velocities
are marked with white circles at the energy maximum. (b) A histogram of MASW measurements.
(a) (b)
FT analysis of Seg38 - Seg40
600 140
0.06
120
500
0.05
100
Group velocity (m/s)
400 0.04
80
Energy
Count
0.03
60
300
0.02 40
200
0.01 20
100 0
2 4 6 8 10 2 4 6 8 10 12 14
Frequency (Hz) Frequency (Hz)
Figure 8.9 (a) One MFT measurement example. The color represents stacking energy, and the picked velocities are
marked with white circles. (b) A histogram of MFT measurements.
measure dispersion in a lower frequency band (Bensen Dispersion data between 2 and 18 Hz are sensitive to
et al., 2007). As is shown in Figure 8.9, peak frequency shear wave velocity at a shallow depth. The rule of
with MFT measurements is around 5 Hz, which is much thumb in surface-wave inversion is that the dispersion
lower than the frequency for the maximum MASW of the fundamental mode Rayleigh wave is sensitive to
measurements. a depth of about 0.5–0.6 times the wavelength. As an
Finally, 1,409 phase velocities and 768 group velocities example, we used a typical model to compute example
are measured and used in the next inversion step. sensitivity kernels for our data set (Figure 8.10). Group
(a) Vs Model (b) Sensitivity (c) Sensitivity

0 0 0
0.22
0.45
10 10 0.2 10
0.4
20 20 0.18 20
0.35
30 30 0.16 30
0.3
0.14
40 40 40
0.25
Depth (m)
dC/dVs
dU/dVs
0.12
50 50 50 0.2
0.1
60 60 60 0.15
0.08
0.1
70 70 70
0.06
0.05
80 80 80
0.04
0
90 90 0.02 90
–0.05
100 100 100

200 400 600 800 5 10 15 20 5 10 15 20
Speed (m/s) Frequency (Hz) Frequency (Hz)
Figure 8.10 (a) The layered model used in sensitivity kernel computation. (b) Sensitivity kernels of Rayleigh wave
phase velocity. (c) Sensitivity kernels of Rayleigh wave group velocity.
velocity is more sensitive than phase velocity and the Temperature Z = 5.0 m
(a) (b)
maximum depth of meaningful sensitivity is about 1600 1600
30–40 m in this model.
1400 1400
8.5. SHEAR WAVE VELOCITY MODEL 1200 1200
The last step of ANT is the inversion of dispersion curves 1000 1000
to construct layered models at sample points. The sample
Y (m)
Y (m)
point is defined as the center point of the segment for the 800 800
same segment data set and the middle point of the two seg-
ments for the two in-line segment data set. Figure 8.11a 600 600
shows the sample points provided by two data sets. The
NCFs of channel pairs of the two in-line segments sample 400 400
the warm ground zone that supplement the NCFs of chan-
nel pairs along one segment. Two inversion methods were 200 200
implemented to solve shear wave velocity profiles at sample
points. The surface wave analysis, modelling and inversion 0 0
(SWAMI) code (Constable et al., 1987; Lai, 1998), based 0 200 400 0 200 400
X (m) X (m)
on Occam’s inversion, is used to invert phase velocities,
whereas group velocities are inverted with a linearized
inversion code from the Computer Programs in Seismology 8 10 12 14 16 18 200 400 600
software suite (Herrmann, 2013). The pseudo-3D shear °C Vs (m/s)
wave velocity model is interpolated from all layered models Z = 10.0 m Z = 30.0 m
(c) (d)
(Figures 8.11b–d). 1600 1600
At very shallow depth (Z = 5, 10 m), a LVZ is observed
in the warm ground zone (Figures 8.11b and 11c). The 1400 1400
area around injection wells 18–31 and 18D–31 also shows
a relatively low velocity. At greater depth, a velocity con- 1200 1200
trast emerges across the fault zone in the southern part.
Other tomography techniques were also used to image 1000 1000
the 3D seismic structure at the Brady Hot Springs geo-
Y (m)
Y (m)
thermal site. The shot interferometry (SI) technique with 800 800
geophone array data provided 3D Vp and Vs models
(Matzel et al., 2017). Parker et al. (2018) used P-wave arri- 600 600
val times from a vibroseis truck on DAS and geophone
arrays to invert a 3D Vp model. Figure 8.12 shows a com- 400 400
parison among these three models at 20 m depth. Gener-
ally, the SI Vs model shows a higher velocity that might be 200 200
due to different initial models and/or inversion strategy.
The LVZ near the injection wells is imaged in the two 0 0
Vs models, but it is not clear in the Vp model. The other 0 200 400 0 200 400
LVZ in the southwest part (x ~ 400, y ~ 100–400) is evi- X (m) X (m)
dent in our Vs and Vp models.
Distributed temperature sensing (DTS) measurements
reveal the spatial surface temperature variation that is 200 400 600 200 400 600
Vs (m/s) Vs (m/s)
strongly affected by the fumaroles and fracture density.
Therefore, surface temperature can be regarded as an Figure 8.11 (a) Daily average surface temperature of 14 March
index of porosity in the natural laboratory. Since seismic 2016. (b) Shear wave velocity at 5 m depth. (c) Shear wave
velocity decreases with increasing porosity, an anticorre- velocity at 10 m depth. (d) Shear wave velocity at 30 m
lation between surface temperature and seismic velocity depth. The fumaroles are shown by black triangles, and gray
is expected. Our result reflects this relationship well lines denote faults. Open circles on panel (a) denote sample
(Figures 8.11a and 8.11b). points (red: cross-segment; blue: same segment).
this study Matzel et al. Parker et al.

Vs Z = 20.0 m Vs Z = 20.0 m Vp Z = 20.0 m
(a) (b) (c)
1600 1600 1600
1400 1400 1400
1200 1200 1200
1000 1000 1000
Y (m)
Y (m)
Y (m)
800 800 800
600 600 600
400 400 400
200 200 200
0 0 0
0 200 400 0 200 400 0 200 400
X (m) X (m) X (m)
200 400 600 350 400 450 500 1000 1500
Figure 8.12 Velocity models at 20 m depth. (a) Vs model in this study, (b) Vs model from Matzel et al. (2017), and (c)
Vp model from Parker et al. (2018). The open circles on panel (a) denote the sample points of this study. Source:
Based on (b) Matzel et al. (2017); (c) Parker et al. (2018).
Due to the array geometry, lower frequency surface features and matches some features seen in models
waves that are sensitive to deeper structure were not obtained with the use of other methods.
extracted in this study. With longer straight segments, this
method could be used to image structure to about 100 m
depth (e.g., Zeng et al., 2019). In this study, only a small ACKNOWLEDGMENTS
portion of all possible channel pairs are used, which limits
spatial sampling and model resolution. The reason is Ray- This study was funded in part by the Office of Energy
leigh and Love waves are mixed on the NCFs of channel Efficiency and Renewable Energy (EERE), U.S. Depart-
pairs for two segments that are not in a line. A new ment of Energy, under Award Number DE-EE0006760.
method has been proposed to handle this challenge X. Zeng acknowledges support from the Natural Science
(Song et al., 2020), which will expand usable data set Foundation of China (41974067) and the LU JIAXI
and help to improve resolution. International Team Program supported by the K.C.
Wong Education Foundation and the Chinese Academy
of Sciences (GJTD-2018-12).
8.6. CONCLUSIONS
In this study, the NCFs of channel pairs in individual REFERENCES

linear DAS segments and pairs of in-line segments are
computed following standard ambient noise processing Ali, S. T., Akerley, J., Baluyut, E. C., Cardiff, M., Davatzes, N.
procedures. Distance bin stacking is used to enhance sig- C., Feigl, K. L., et al. (2016). Time-series analysis of surface
deformation at Brady Hot Springs geothermal field
nal-to-noise ratio of NCFs for channel pairs belonging to
(Nevada) using interferometric synthetic aperture radar.
two in-line segments at the greater offset. Combining the Geothermics, 61, 114–120.
two data sets improves coverage and makes it possible to Bensen, G. D., Ritzwoller, M. H., Barmin, M. P., Levshin, A.
build a 3D Vs model. Both group and phase velocities of L., Lin, F., Moschetti, M. P., et al. (2007). Processing seismic
Rayleigh waves in a high-frequency band (2–18 Hz) are ambient noise data to obtain reliable broad-band surface
used to invert shear wave velocity structure at shallow wave dispersion measurements. Geophysical Journal Interna-
depth. The velocity model correlates with some surface tional, 169(3), 1239–1260.
Campillo, M., & Roux, P. (2015). 1.12—crust and lithospheric Parker, T., Shatalin, S. V., & Farhadiroushan, M. (2014). Dis-
structure—seismic imaging and monitoring with ambient tributed acoustic sensing—a new tool for seismic applications,
noise correlations. Treatise on Geophysics, 391–417. First Break, 32(2), 61–69.
Constable, S. C., Parker, R. L., & Constable, G. G. (1987). Parker, L., Thurber, C., Zeng, X., Li, P., Lord, N., Fratta, D.,
Occam’s inversion: a practical algorithm for generating et al. (2018). Active-source seismic tomography at the Brady
smooth models from electromagnetic sounding data. Geo- geothermal field, Nevada, with dense nodal and fiber-optic
physics, 52, 289–300. seismic arrays. Seismological Research Letters, 89(5),
Dou, S., Lindsey, N., Wagner, A. M., Daley, T. M., Freifeld, B., 1629–1640.
Robertson, M., et al. (2017). Distributed acoustic sensing for Seats, K. J., Lawrence, J. F., & Prieto, G. A. (2012). Improved
seismic monitoring of the near surface: A traffic-noise interfer- ambient noise correlation functions using Welch’s method.
ometry case study. Scientific Reports, 7(1), 11620. Geophysical Journal International, 188(2), 513–523.
Dziewonski, A. M., Bloch, S., & Landisman, M. (1969). A tech- Shapiro, N., Campillo, M., Stehly, L., & Ritzwoller, M. (2005).
nique for the analysis of transient seismic signals. Bulletin of High-resolution surface-wave tomography from ambient seis-
the Seismological Society of America, 59, 427–444. mic noise. Science, 307(5715), 1615–1618.
Faulds, J. E., Ramelli, A. R., Gardised, L. J., Coolbaugh, M. Song, Z., Zeng, X., Thurber, C., Wang, H., & Fratta, D. (2018).
F., & Green, H. L. (2017). Preliminary geologic map of Imaging shallow structure with active-source surface wave sig-
the Bradys geothermal area, Churchill County, Nevada. nal recorded by distributed acoustic sensing arrays. Earth-
Nevada Bureau of Mines and Geology, Open-File quake Science, 31, 207–213.
Report, 12-5. Song, Z., Zeng, X., & Thurber. C. (2021). Surface-wave disper-
Feigl, K. L., & PoroTomo Team. (2018). Overview and sion spectrum inversion method applied to Love and Rayleigh
preliminary results from the PoroTomo project at Brady Hot waves recorded by DAS, Geophysics, 86(1), EN1–EN12.
Springs, Nevada: Poroelastic tomography by adjoint inverse Vécsey, G., Holliger, K., Pratt, R. G., Dyer, B. C., & Green, A.
modeling of data from seismology, geodesy, and hydrology. G. (1998). Anisotropic seismic tomography of a potential hot
Proceedings: 43rd Workshop on Geothermal Reservoir Engi- dry rock reservoir before and during induced pressurization.
neering, Stanford University, Stanford, California, 12–14 Geophysical Research Letters, 25(11), 1991–1994.
February 2018. Wang, Y., Lin, F.-C., Schmandt, B., & Farrell, J. (2017). Ambi-
Herrmann, R. B. (2013). Computer programs in seismology: An ent noise tomography across Mount St. Helens using a dense
evolving tool for instruction and research. Seismological seismic array. Journal of Geophysical Research Solid Earth,
Research Letters, 84, 1081–1088. 122, 4492–4508. doi:10.1002/2016JB013769
Lai, C. G. (1998). Simultaneous inversion of Rayleigh phase Wang, H. F., Zeng, X., Miller, D. E., Fratta, D., Feigl, K. L.,
velocity and attenuation for near-surface site characterization Thurber, C. H., & Mellors, R. (2018). Ground motion
(PhD dissertation), Georgia Institute of Technology. response to a Ml 4.3 earthquake using co-located distributed
Lee, E.-J., Chen, P., Jordan, T. H., Maechling, P. B., Denolle, acoustic sensing and seismometer arrays. Geophysical Journal
M. A. M., & Beroza, G. C. (2015). Full 3D tomography for International, 213(3), 2020–2036.
crustal structure in southern California based on the scatter- Zeng, X., Lancelle, C., Thurber, C., Fratta, D., Wang, H., Lord,
ing-integral and the adjoint-wavefield methods. Journal of N., et al. (2017). Properties of noise cross-correlation func-
Geophysical Research Solid Earth, 119(8), 6421–6451. tions obtained from a distributed acoustic sensing array at
Lehujeur, M., Vergne, J., Schmittbuhl, J., Zigone, D., & Chenadec, Garner Valley, California. Bulletin of the Seismological Soci-
A. L. (2018). Reservoir imaging using ambient noise ety of America, 107(2), 603–610.
correlation from a dense seismic network. Journal of Geophysi- Zeng, X., Thurber, C., Fratta, D., Wang, H., Fratta, D., Matzel,
cal Research Solid Earth, 123, 6671–6686, doi:10.1029/ E., & PoroTomo Team (2017). High-resolution shallow struc-
2018JB015440. ture revealed with ambient noise tomography on a dense array.
Lin, F.-C., Tsai, V. C., Schmandt, B., Duputel, Z., & Zhan, Z. Proceedings: 42nd Workshop on Geothermal Reservoir Engi-
(2013). Extracting seismic core phases with array interferom- neering. Stanford University, Stanford, California, 13–15
etry. Geophysical Research Letters, 40(6), 1049–1053. February 2017.
Matzel, E., Zeng, X., Thurber, C., Luo, Y., & Morency, C. Zeng, X., Wang, B., Li, X., Xu, W., Xu, S., & Song, Z. (2019).
(2017). Seismic interferometry using the dense array at the Shallow Vs structure imaged with the ambient noise recorded by
Brady geothermal field. Proceedings: 42nd Workshop on Geo- a telecommunication fiber-optic cable in urban area. Paper pre-
thermal Reservoir Engineering. Stanford University, Stan- sented at 2019 Seismological Society of America Annual
ford, California, 13–15 February 2017. Meeting, Seattle, WA, USA, 23–26 April 2019.
Nakata, N., Chang, J. P., Lawrence, J. F., & Boué, P. (2015). Zhan, Z., Ni, S., Helmberger, D. V., & Clayton, R. W. (2010).
Body wave extraction and tomography at Long Beach, Cali- Retrieval of Moho-reflected shear wave arrivals from ambient
fornia, with ambient noise interferometry. Journal of Geo- seismic noise. Geophysical Journal International, 182(1), 408–420.
physical Research Solid Earth, 120(2), 1159–1173. Zhang, Q., & Lin, G. (2014). Three-dimensional Vp and Vp/Vs
Park, C., & Carnevale, M. (2010). Optimum MASW survey- models in the Coso geothermal area, California: Seismic char-
revisit after a decade of use. GeoFlorida 2010: Advances in acterization of the magmatic system. Journal of Geophysical
Analysis, Modeling and Design. ASCE Conference Proceed- Research: Solid Earth, 119(6), 4907–4922.
ings, 20–24 February 2010, West Palm Beach, FL.

8

Uploaded by

Copyright:

Available Formats

8

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

8

Uploaded by

Copyright:

Available Formats

Downloaded from https://agupubs.onlinelibrary.wiley.com/doi/ by Mit Libraries Serials & Journa, Wiley Online Library on [02/11/2022].

High-Resolution Shallow Structure at Brady Hot Springs Using

8.1. INTRODUCTION and anisotropy) are sensitive to lithology, including

both active source (controlled explosions, vibroseis 1600

and used in previous tomography studies (Matzel et al., 800

(c) 5 Hz (d) 10 Hz (e) 15 Hz

1400 1400 –100

1200 1200 –110

CH 0462 - 0580 CH 0462 - 0580 CH 0110 - 0640 CH 0110 - 0640

0.6 684 0.6 684

8.3. NCF RESULTS 419.4

The Rayleigh wave signal can be observed on both ver- 419.4

80 NCFs of 13 channel pairs with the offset ranging from

419.4 to 420.1 m, and the stacked traces. Rayleigh wave

300 1000 100

(a) Vs Model (b) Sensitivity (c) Sensitivity

100 100 100

8.5. SHEAR WAVE VELOCITY MODEL 1200 1200

this study Matzel et al. Parker et al.

1400 1400 1400

1200 1200 1200

1000 1000 1000

600 600 600

400 400 400

200 200 200

200 400 600 350 400 450 500 1000 1500

In this study, the NCFs of channel pairs in individual REFERENCES

You might also like