Journal of Geophysical Research: Solid Earth

RESEARCH ARTICLE Three-dimensional Vp and Vp/Vs models in the Coso

10.1002/2014JB010992 geothermal area, California: Seismic characterization
Key Point: of the magmatic system
• Three-dimensional Vp and Vp/Vs
models and earthquake relocations in Qiong Zhang1 and Guoqing Lin1
the Coso geothermal field
1 Division of Marine Geology and Geophysics, Rosenstiel School of Marine and Atmospheric Science,

University of Miami, Miami, Florida, USA

Supporting Information:
• Readme
• catalog
• Figure S1 Abstract We combine classic and state-of-the-art techniques to characterize the seismic and volcanic
• velfile_fine_vp features in the Coso area in southern California. Seismic tomography inversions are carried out to map the
• velfile_fine_vpvs variations of Vp, Vs, and Vp/Vs beneath Coso. The velocities in the top layers of our model are correlated
with the surface geological features. The Indian Wells Valley, with high silica content sediment strata, shows
Correspondence to: low-velocity anomalies up to 3 km depth, whereas the major mountain ranges, such as the south Sierra
Q. Zhang, Nevada and the Argus Range, show higher velocities. The resulting three-dimensional velocity model is
[email protected]
used to improve absolute locations for all local events between January 1981 and August 2011 in our study
area. We then apply similar-event cluster analysis, waveform cross correlation, and differential time
Citation: relocation methods to improve relative event location accuracy. A dramatic sharpening of seismicity
Zhang, Q., and G. Lin (2014),
Three-dimensional Vp and Vp/Vs patterns is obtained after using these methods. We also estimate high-resolution near-source Vp/Vs
models in the Coso geothermal ratio within each event cluster using the differential times from waveform cross correlation. The in situ
area, California: Seismic character- Vp/Vs method confirms the trend of the velocity variations from the tomographic results. An anomalous
ization of the magmatic system,
J. Geophys. Res. Solid Earth, 119, low-velocity body with low Vp, Vs, and Vp/Vs ratios, corresponding to the ductile behavior underlying the
4907–4922, doi:10.1002/2014JB010992. Coso geothermal field from 6 to 12 km depth, can be explained by the existence of frozen felsic magmatic
materials with the inclusion of water. The material is not likely to include pervasive partial melt due to a lack
Received 27 JAN 2014 of high Vp/Vs ratios.
Accepted 16 MAY 2014
Accepted article online 21 MAY 2014
Published online 9 JUN 2014 1. Introduction
The Coso geothermal field (CGF) is located between the Sierra Nevada batholith and the Basin and Range
Province in southeastern California (Figure 1a). As one of the largest geothermal fields in the US, it has been
exploited to generate power through over 100 production wells since 1987 [Adams et al., 2000]. The maxi-
mum heat flow of the geothermal field was estimated to be 10 times the value from the background Basin
and Range [Combs, 1980] (Figure 1b). A crustal magma body has been assumed to provide the primary heat
source for the present surface geothermal system [Combs, 1980; Bacon et al., 1980; Duffield et al., 1980].
Several lines of evidences favor the existence of a magma body beneath Coso. First, Coso is in a transten-
sional tectonic regime and the extension facilitates the ascent of magma. Coso sits in an extensional
stepover between the dextral striking faults to the north and south. The geothermal field is surrounded by
three main valleys with a series of strike-slip and normal faults [Reasenberg et al., 1980]. Lithospheric exten-
sion causes crustal thinning beneath the geothermal field, which favors the intrusion of basalt from the
mantle into the crust, sustaining a long-lived magma reservoir in the crust [Duffield et al., 1980]. The crustal
thinning at Coso is supported by the geochemical result of the isotopic composition of rocks, which are
strongly influenced by the asthenosphere [Monastero et al., 2005].
The geothermal field is characterized by young volcanic rocks, mainly Pleistocene rhyolite dome, flanking
basalt flows and cinder cones. The active volcanoes erupted around 4 Ma, and the Pleistocene bimodal
eruptions continued from 1.04 to 0.03 Ma, while the latest active volcanism is estimated to be 0.03 Ma
[Duffield et al., 1980; Manley and Bacon, 2000]. It is believed that the magmatic activity beneath the dome
field triggered the Coso geothermal system around 0.2 Ma [Adams et al., 2000; Duffield et al., 1980]. Combs
[1980] showed that the geothermal field is associated with 38 rhyolite domes, surface hydrothermal man-
ifestations, and higher heat flows than the surrounding area. The highest heat values correspond to the
geological features, such as the Sugarloaf Mountain (SM, the largest rhyolite dome in the area), the Devil’s

ZHANG AND LIN ©2014. American Geophysical Union. All Rights Reserved. 4907
 Journal of Geophysical Research: Solid Earth 10.1002/2014JB010992

(a) Kitchen (DK), and the Coso Hot Springs

(CHS) with the presence of hot springs
and fumaroles. The present geothermal
production area lies in a 6 × 10 km2
36.5
north-south trending zone between
the SM and CHS [Bishop and Bird, 1987;
Feng and Lees, 1998; Fialko and Simons,
2000]. The well log data up to 500 m
show that the high-temperature system
has been developed beneath the flank
35.5
of the present production area, with
temperatures up to 328◦ C [Adams et al.,
2000]. The production wells are con-
centrated above 3 km depth. To 10 km
south of the SM and CHS, the area has
not been exploited widely, although
34.5
−119 −118 −117 the heat flow value is as high as
120 mW/m2 (Figure 1b).
(b) Exploring the existence of a resid-
36.2 ual magma chamber beneath the
geothermal field has been the focus
of numerous geophysical, geochem-
36.0 ical, and geological studies [e.g.,
Duffield et al., 1980; Wu and Lees, 1996;
Manley and Bacon, 2000; Lees, 2002;
Wilson et al., 2003; Monastero et al.,
35.8
2005]. The extensive studies reveal
that a long-lived rhyolitic magma reser-
35.6
voir presumably emplaced around
−118.2 −118 −117.8 −117.6 −117.4 −117.2 0.03 Ma and provided the heat flux for
Figure 1. Geologic map of the Coso area. Coso sits under the transition
the overlying hydrothermal system.
zone between the strike-slip San Andreas Fault and the extensional However, the accurate location of the
Basin and Range. It is bounded on the north by the northwest striking magma reservoir varies between differ-
Owens Valley Fault while in the south is dominated by the dextral Little ent studies. Most previous geochemical
Lake fault zone (LLFZ) and Airport Lake fault zone (ALFZ). (a) Map show- and petrological studies suggest that the
ing tectonic features and our study areas I and II. Blue dots represent
top of the magma begins below 6 km
the earthquakes between 1981 and 2011 recorded by the Southern
California Seismic Network (SCSN) stations shown by the red triangles. depth, based on analyses of mineral-
Black lines denote Quaternary faults. Our study area II for the fine-scale ogy and ages of erupted volcanic rocks
model marked by the red box encloses the geothermal field and the [e.g., Duffield et al., 1980; Bacon et al.,
surrounding areas. In the inset map, the red star shows the location of 1981; Manley and Bacon, 2000]. The grav-
the Coso area in California and the rectangle shows our study area I.
ity data also suggest a low-resistivity
(b) A close-up of the study area II. Black dots denote the grid nodes for
inversion. Blue dots are the master events for the tomographic inver- and low-density zone 5 km beneath
sions. Diamonds are colored by the heat values from Combs [1980] the CGF [Wamalwa et al., 2013]. With
and Saltus and Lachenbruch [1991]. High heat flow values within the the advantage of mapping the sub-
geothermal field are present in the Sugarloaf Mountain (SM), the Devil’s surface structure, seismic studies have
Kitchen (DK), and the Coso Hot Springs (CHS). The geothermal field is
also pointed out that the anoma-
surrounded by three main valleys, the Rose Valley (RV), the Coso Wash
(CW), and the Indian Wells Valley (IWV). Other abbreviations are SSNFZ, lous body exists around 6 km depth
Southern Sierra Nevada Frontal Fault Zone; WCF, Wilson Canyon Fault; from the teleseismic receiver func-
and AHF, Ash Hill Fault. tion analysis [Wilson et al., 2003] and
the ambient noise tomography [Yang
et al., 2011]. However, these large-scale studies failed to map the lateral variation of the anomalies on a
fine scale comparable to the geothermal field. The local earthquake body wave tomography study by
Hauksson and Unruh [2007] also detected a low P wave velocity zone ranging from 5 to 10 km depth
beneath Coso, but they argued that the slightly varied ratios of the P and S wave velocity (Vp/Vs) cannot
support the existence of a magma body shallower than 10 km.

ZHANG AND LIN ©2014. American Geophysical Union. All Rights Reserved. 4908
 Journal of Geophysical Research: Solid Earth 10.1002/2014JB010992

0 In this study, we apply local earthquake

body wave tomography to invert for
three-dimensional (3-D) P wave velocity
5 (Vp) and Vp/Vs model in the Coso area. The
objective is to image the high-resolution
subsurface velocity structure underneath

Depth (km)
10 Hauksson2000
the Coso geothermal field and outline
the location, geometry, and depth of the
Hauksson2007
magma body. Combining these models
Lin2007
with other geophysical studies, we try to
15
minimum model understand how the geothermal and mag-
matic systems operate to generate the
surface manifestations. By analyzing our
20
4.5 5 5.5 6 6.5 7 tomography models and distribution of
P wave velocity (km/s) earthquake relocations, we aim to deter-
Figure 2. Comparison of different 1-D velocity models. Blue line mine the accurate location of the magma
shows the model by Hauksson [2000] for southern California, which body and answer the unsolved questions
was used as the starting model to generate the 1-D Vp model for related to the magmatic system, such as
the Coso region by Hauksson and Unruh [2007] shown by the black the upper bound of the reservoir depth, the
line. Our minimum Vp model for the Coso region denoted by the
interface between silicic and mafic magma,
red line is produced by using the model of Lin et al. [2007a] for
southern California as the starting model (pink line). and whether a large magma reservoir serves
as the main heat source or several small heat
sources exist.

2. Data and Methods

2.1. Data Set
We develop the regional and local velocity models for Coso (Figure 1) using data from the Southern
California Earthquake Data Center (SCEDC). The regional study area I includes Coso and the adjacent areas,
such as the Garlock Fault, the Sierra Nevada, and the Death Valley, and is referred to as the Coso region.
Study area II focuses on the vicinity of the Coso geothermal field. We obtain P and S wave first arrival times
and waveform data in our study area I from 159,295 events between January 1981 and August 2011 from
the SCEDC. Seismic stations are more densely distributed in the vicinity of the geothermal area than other
parts of study area I, which ensures the resolution and reliability of the tomographic models for our study
area II (Figure 1a). The completeness magnitude for the entire study area II is estimated to be M 1.3, and
the maximum magnitude is Mw 5.75. The waveform data are resampled at 100 Hz sample rate and filtered
by applying a bandpass 1–10 Hz for waveform cross correlation. These data-processing steps are similar to
those in previous studies for southern California [e.g., Shearer et al., 2005; Lin et al., 2007a].
In local earthquake tomography, well-recorded and evenly distributed events are usually used in the inver-
sion. We choose 1893 master events for study area I from the entire data set by applying the criteria of 5 km
spacing in horizontal plane and 2 km in depth between master events and each event having more than
14 P and 9 S picks. As a result, a total number of 40,859 P and 20,321 S wave phase picks are selected for the
inversion for the regional model. To invert for a finer model for study area II, we select 1263 master events
with 25,248 P picks and 13,884 S picks with the criteria of 2 km spacing in horizontal and 1 km in depth
between master events and each master event having more than 14 P picks and 9 S picks (Figure 1b).
2.2. One-Dimensional Starting Model
Three-dimensional velocity model inversions typically start with one-dimensional (1-D) models. We use the
layer-average velocity model for southern California by Lin et al. [2007a] as the starting model and invert
for the minimum 1-D model for Coso by using the program VELEST [Kissling et al., 1994, 1995]. The model
damping parameters are chosen based on the trade-off tests of optimizing data misfit and model variance.
We set the station corrections to be 0 during the inversion to avoid trade-off between the velocity struc-
ture and station corrections. The model variance from our minimum model is decreased by 35% from the
starting model, and the data variance is reduced by 26%. Hauksson and Unruh [2007] also inverted for a 3-D
velocity model for Coso using with the average model of Hauksson [2000] for southern California as the 1-D
starting model. Compared with the 1-D model by Hauksson and Unruh [2007], our minimum model is 3–14%

ZHANG AND LIN ©2014. American Geophysical Union. All Rights Reserved. 4909
 Journal of Geophysical Research: Solid Earth 10.1002/2014JB010992

Depth= 3 km faster above 9 km depth, which mainly

results from distinct starting models, data
Vp % coverages, and different inversion param-
14
12 eters. These 1-D models are shown in
10
8 Figure 2.
6
4 2.3. Three-Dimensional Tomographic
2
0
Inversion
−2 The goal of our study is to obtain an accu-
−5

s Range
Coso rate velocity model for interpreting the

evada
−8
crustal structure beneath the geothermal

Valley
−11
−14

In Sierra N
field. The inversion for Vp/Vs ratios, which

Argu
−17

lls
−20 is indicative of both lithology and rheology

We
of subsurface materials, depends on both P

an
ns
tai
di
n and S wave ray paths. Since there are fewer

Ea hea
ou
eM

S
S wave picks than P wave picks and the

ste r Z
i t
an
Gr

rn one
Mojave quality of S wave data is not as good as

Ca
those of P wave, Vs models are usually

lifo
rn
Desert poorly resolved compared to Vp and the

ia
method of deriving Vp/Vs ratio from Vp
divided by Vs is not reliable [Thurber and
Eberhart-Phillips, 1999]. In this study, we
solve the Vp/Vs model directly by using the
Figure 3. Our regional-scale velocity model for study area I at 3 km
depth. Vp perturbations are shown relative to the layer-average S-P travel time differences. In the simul2000
value. The thick black line encloses the area with the diagonal ele- algorithm [Thurber, 1983; Thurber and
ment of the resolution matrix greater than 0.1, which is considered Eberhart-Phillips, 1999], ray paths are
well resolved. Most of the observed velocity anomalies correlate selected from the fastest travel time
well with the known geological features. For example, the Sierra
between the source and receiver and cal-
Nevada, the Argus Range, the Granite Mountains, and the core of
Mojave Desert are represented by high-velocity anomalies, whereas culated by approximate ray tracing. These
the Eastern California Shear Zone and the Indian Wells Valley show ray paths are curved nonplanar [Eberhart-
low-velocity anomalies. Coso is located at the north end of the Phillips and Michael, 1998], and the S ray
Indian Wells Valley. paths are approximated by P paths. The
algorithm is a damped least squares inver-
sion, and the optimal damping parameters
are chosen based on the trade-off curve of data misfit and model variance [Eberhart-Phillips, 1986]. We run
a series of inversions using different damping parameters and choose the damping values of 150 for Vp and
80 for Vp/Vs in the inversion for the regional model.
We first solve the velocity model for study area I and then invert for the final 3-D velocity model for study
area II, the vicinity of the CGF. The inversion follows two steps: (1) start with the minimum 1-D model of
the Coso region and invert for the 3-D Vp and Vp/Vs models in study area I with a coarser uniform horizon-
tal grid spacing of 6 km; and (2) use the resulting 3-D model for the Coso region to invert for the final 3-D
model of the CGF (study area II) with a finer horizontal spacing of 3 km. By doing this, the final model for the
CGF shows more detailed velocity anomalies than the regional model. A constant Vp/Vs ratio of 1.73 is used
for the starting Vp/Vs model inversion for the region based on the Wadati diagram [Kisslinger and Engdahl,
1973]. The depth distribution of seismicity suggests that the seismicity is focused in the upper 15 km depth,
and we set up the depth layers at −5, 0, 3, 6, 9, 12, 15, 20, and 25 km. Note that in this study all depths are
relative to mean sea level.

3. Results
3.1. Regional Coarse Model
The regional model is obtained after six iterations when the reduction of data variance becomes insignifi-
cant. Compared to the initial models, the data variance is reduced by 69% and 45% for final Vp and Vp/Vs
models, respectively. The root-mean-square (RMS) of the travel time residuals is reduced from 0.24 to 0.11 s.
We show a representative velocity image at 3 km depth in Figure 3 with the Vp perturbations relative to the
layer-average value of the inverted model. The inverted Vp anomalies reflect the near-surface geological

ZHANG AND LIN ©2014. American Geophysical Union. All Rights Reserved. 4910
 Journal of Geophysical Research: Solid Earth 10.1002/2014JB010992

(a) SCSN (b) 3D

(d) This
(c) HYS study

Figure 4. Map view of different catalogs in the Coso region. The red dots represent the common events that are relo-
cated by different methods. (a) Southern California Seismic Network (SCSN) catalog, covering events in Coso from 1981
to 2011. Grey dots show the events that are not relocated but remain in the original SCSN catalog. (b) Three-dimensional
relocations produced by the simul2000 algorithm [Thurber, 1983; Thurber and Eberhart-Phillips, 1999]. (c) HYS catalog, the
relocation catalog for southern California between 1981 and 2011 by Hauksson et al. [2012]. (d) Highly correlated events
in this study. See Figure 1 for the abbreviations.

features. The Indian Wells Valley and the Eastern California Shear Zone mainly exhibit low Vp. The moun-
tain areas, such as the Sierra Nevada, the Argus Range, and the Granite Mountain with the Mesozoic granitic
rocks, and the stable Mojave Desert show high Vp anomalies.
3.2. Earthquake Relocation
After obtaining the regional 3-D velocity model, we invert for the 3-D earthquake relocations in the
simul2000 by fixing the velocity model. The absolute earthquake relocations have been improved by tak-
ing into account the biasing effect of velocity structure. To improve the accuracy of relative earthquake
locations, we implement the waveform cross correlation, similar-event cluster analysis, and differential time
relocation methods by Lin et al. [2007b] to relocate all the events in the Coso region.
Both the absolute and relative locations are improved compared to the SCSN catalog locations. The
improvement is demonstrated by the sharpening of the relocated seismicity in Figure 4. Our relocations are
consistent with the previous relocation catalogs for southern California, such as the SHLK catalog [Shearer
et al., 2005], the LSH catalog [Lin et al., 2007b], and the latest HYS location catalog by Hauksson et al. [2012]
for events between 1981 and 2011. We use similar criteria for cross correlation to those for the LSH catalog
such as the correlation coefficient cutoff, station-event distance range, and minimum average of the max-
imum correlation coefficient. The main difference between our catalog and previous ones stems from the
different absolute locations, and we use the produced 3-D locations from simul2000 (Figure 4b). Our catalog
chooses 0.65 as correlation coefficient cutoff and encompasses the highly correlated local events in Coso,
whereas the HYS catalog uses the correlation coefficient cutoff of 0.6.
Quantitative estimates of location uncertainties indicate that both absolute and relative location accuracies
are significantly improved. The absolute location error estimates are provided by the simul2000 algorithm
[Thurber and Eberhart-Phillips, 1999]. The mean horizontal uncertainty is 120 m, and the vertical uncertainty
is 300 m. The relative location uncertainties are estimated by a bootstrap method [Efron and Gong, 1983],

ZHANG AND LIN ©2014. American Geophysical Union. All Rights Reserved. 4911
 Journal of Geophysical Research: Solid Earth 10.1002/2014JB010992

(a) Depth= 0 km (b) Depth= 3 km

40 40

30 30

20 20
Y (km)

10 10
3.0
0 0

−10 −10

−20 −20 2.0

−40 −20 0 20 40 −40 −20 0 20 40

(c) Depth= 6 km (d) Depth= 9 km

40 40 1.0

30 30

20 20
Y (km)

0.0
10 10
Spread Values
0 0 for Vp

−10 −10

−20 −20
−40 −20 0 20 40 −40 −20 0 20 40
X (km) X (km)

Figure 5. Distribution of the spread values for the final Vp model in study area II. Colored nodes denote spread values below 3. Values greater than 3 are not
shown. Black lines denote the diagonal elements of the resolution matrix greater than 0.1, which are also shown in Figures 6–11.

similar to Lin et al. [2007b]. After resampling for 15 times, we obtain the median of the relative location
uncertainties of 11 m in horizontal and 22 m in depth. In study area I, about 66% of events fall into 1225
clusters with at least five events. Three distinct clusters with more than 5000 events are observed from the
distribution of the relocated hypocenters within the Rose Valley, to the east of the CGF, and around the
Airport Lake Fault Zone, respectively.

3.3. Model Resolution Tests

Several parameters are used to examine the model quality, which mostly depends on the geometry and
density of rays. Ray-dependent measurements and synthetic tests are conducted together to assess the res-
olution of the tomographic models. In this paper, we focus different resolution estimates on our study area II.
We plot resolution estimates to assess the ray coverage including the derivative weighted sum (DWS) values,
diagonal elements of resolution matrices (RDE) and spread function. The DWS values indicate the weighted
ray density directly. The weight scheme is based on the distance of rays from each node. Based on the dis-
tribution of DWS values (supporting information Figure S1), high numbers of the DWS values (>2000) with
the maximums of 16,000 for P rays and 10,000 for S-P rays are present in the vicinity of the Coso geothermal
field and the adjacent areas, which make up of our study area II. The deficiency of the DWS measurement is
that the ray directions are not taken into account; thus, smearing cannot be estimated.
The model resolution matrix gives the information for all the nodes, and each row represents the averaging
vector of a model parameter. The RDE reflects the resolution for each node and provides relative measures
of the ability of the data for detecting anomalies in different locations. Resolutions greater than 0.1 are con-
sidered to indicate good quality of the models in this study based on the synthetic data tests. However, the
RDE values depend mainly on the grid spacing and damping parameters.

ZHANG AND LIN ©2014. American Geophysical Union. All Rights Reserved. 4912
 Journal of Geophysical Research: Solid Earth 10.1002/2014JB010992

(a) Depth= 0 km (b) Depth= 3 km

40 40

30 30

20 20
Y (km)

10 10 3.0

0 0

−10 −10
2.0
−20 −20
−40 −20 0 20 40 −40 −20 0 20 40

(c) Depth= 6 km (d) Depth= 9 km

1.0
40 40

30 30

20 20
0.0
Y (km)

10 10 Spread Values
for Vp/Vs
0 0

−10 −10

−20 −20
−40 −20 0 20 40 −40 −20 0 20 40
X (km) X (km)

Figure 6. Distribution of the spread values for the final Vp/Vs model in study area II. Colored nodes denote spread values below 3. Black lines denote the RDE
values greater than 0.1.

The spread function investigates the dependence of a model parameter on the other grid nodes. Ideally,
the velocity at each node is independent of other nodes, but common ray geometries can link neighboring
model parameters and can lead to artificial smearing of anomalies across multiple nodes. Thus, the spread
value will be zero in an ideal case. For the real data sets, we regard spread function values smaller than 3
as good values indicating that the peaked resolution is achieved at the grid and the distant nodes have
no significant contribution. We choose the cutoff of 3 to represent low-spread values with good resolution
because the distributions of the spread function values smaller than 3 for the Vp and Vp/Vs models are con-
sistent with the RDE contours with resolutions greater than 0.1 (Figures 5 and 6). The cutoff of 2.5 or 3 has
been chosen by previous tomography studies [e.g., Sherburn et al., 2006; Reyners et al., 2006] to show the
nodes without too much smearing.
Resolution tests, i.e., checkerboard tests in this study, are also performed to compare with the
ray-dependent measurements. Checkerboard models are constructed to assess the amount of image blur-
ring. Five percent of synthetic velocity perturbations are assigned to blocks with dimensions of 6 × 6 km
at all layers. The synthetic travel times are then inverted to recover the velocity anomalies using the same
parameterization method as in the real data set. The synthetic anomalies are reconstructed well underneath
the geothermal field, especially at depths of 3 km and 6 km for both Vp and Vp/Vs models (Figures 7 and 8).
The RDE contour lines with resolutions greater than 0.1 are correlated with the well-resolved checkerboard
patterns. For Vp/Vs, the ability of reconstructing checkerboard patterns is similar to that of Vp.
We assess the reliability of the velocity features by considering the RDE, spread values, and well-resolved
checkerboard patterns. The area around the geothermal field shows reliable resolutions, especially at
depths of 3 and 6 km. The RDE contour lines correlate mostly with the spread values and checkerboard test

ZHANG AND LIN ©2014. American Geophysical Union. All Rights Reserved. 4913
 Journal of Geophysical Research: Solid Earth 10.1002/2014JB010992

(a) Depth= 0 km (b) Depth= 3 km results. In the following map views and
cross sections, we use the RDE contour
40 40
lines of 0.1 to enclose the well-resolved
30 30
20
area of the velocity structure.

Y (km)
20
10 10
3.4. Map Views of Final
0 0 %Vp
−10
5 Velocity Models
−10 4
−20 −20 With the regional 3-D model as input,
3
−40 −20 0 20 40 −40 −20 0 20 40 2 we invert for the velocity structure in
1 study area II. Figures 9 and 10 show
(c) Depth= 6 km (d) Depth= 9 km 0
−1 the velocity images above 9 km depth,
40 40 −2 below which the seismicity is sparse
30 30 −3
−4 and rays are insufficient to resolve the
20 20
Y (km)

−5 velocity structure well. The lateral het-

10 10
0 0 erogeneities are large at near-surface
−10 −10 depths of 0 and 3 km for both Vp and
−20 −20 Vp/Vs. At the surface, a notable feature
−40 −20 0 20 40 −40 −20 0 20 40 of low Vp (ranging between 3.6 and
X (km) X (km)
4.6 km/s) is observed within and around
(e) Depth= 0 km (f) Depth= 3 km the CGF. The nearby faults and valleys
show low Vp, such as the Little Lake Fault
40 40
Zone (LLFZ), the Airport Lake Fault Zone
30 30
(ALFZ), and the Indian Wells Valley (IWV).
20 20
Y (km)

10 10 The intensely high Vp anomalies lie

0 0 %Vp along the Southern Sierra Nevada Frontal
5
−10 −10 4 Fault (SSNFZ) and the Argus Range. At
−20 −20 3 3 km depth, the southeastern part of
−40 −20 0 20 40 −40 −20 0 20 40 2
1 the CGF exhibits high Vp. Low Vp val-
(g) Depth= 6 km (h) Depth= 9 km 0 ues are prominent in the upper 3 km
−1
−2 near the IWV and the adjacent LLFZ and
40 40
30 30
−3 ALFZ. The ALFZ forms pull-apart basins
−4
20 20 and cuts the IWV [Monastero et al., 2005].
Y (km)

−5
10 10 The low Vp anomaly is consistent with
0 0 the geological study that the 3 km high
−10 −10 silica-content sediment exists beneath
−20 −20
the Valley [Monastero et al., 2002]. Below
−40 −20 0 20 40 −40 −20 0 20 40
X (km) X (km) 6 km depth, the Vp lateral variations
become small and a relatively low Vp is
Figure 7. Checkerboard test for the Vp model with blocks of 6 × 6 km. apparent at the CGF.
The velocity perturbations relative to the initial 1-D model are shown
Spatial correlation between the Vp and
in gray scale. (a–d) The true velocity model. (e–h) The inverted velocity
model. White lines denote the RDE values greater than 0.1. Vp/Vs anomalies is observed for most
parts of our study area II. In the Argus
range with strong high Vp, high Vp/Vs
anomaly is seen at the surface. However, we do not see high Vp/Vs corresponding to the high Vp beneath
the SSNFZ. There is a variation of Vp/Vs anomalies across the CGF in the upper 3 km depth. The northern
part of the geothermal field shows low Vp/Vs, whereas the other areas within and in proximity to the CGF
show high Vp/Vs. The model in the vicinity of the CGF mainly shows low Vp/Vs features below 6 km depth,
but a notable high Vp/Vs body (1.78–1.80) is observed around the ALFZ at 6 km depth.

3.5. Cross Sections of Final Velocity Models

The geometrical shapes of the velocity anomalies are easier to track in cross sections. Figure 11 shows the
cross sections of the inverted velocity models beneath the CGF and surrounding geologic regions. The Vs
model is obtained from Vp divided by Vp/Vs model. We combine the Vp, Vp/Vs, and Vs models with the relo-
cated seismicity to find out the prominent features around the geothermal field. Beneath profiles B, D, and
E, which pass through the CGF, an extensive velocity anomaly of low Vp, low Vs, and low Vp/Vs from 6 to

ZHANG AND LIN ©2014. American Geophysical Union. All Rights Reserved. 4914
 Journal of Geophysical Research: Solid Earth 10.1002/2014JB010992

(a) Depth= 0 km (b) Depth= 3 km 12 km depth is observed. The size

of this anomaly is about 10 km in
40 40
30 30 lateral and 6 km in depth. The dis-
20 tribution of the relocated seismicity

Y (km)
20
10 10 %Vp/Vs underneath the CGF indicates that the
0 0 5
4
brittle-ductile depth is around 5 km
−10 −10
3 compared to 10 km in adjacent areas.
−20 −20 2
−40 −20 0 20 40 −40 −20 0 20 40 1
This shallower brittle-ductile transi-
0 tion depth has also been mentioned by
(c) Depth= 6 km (d) Depth= 9 km −1
previous studies [e.g., Monastero et al.,
−2
40 40 −3 2005; Hauksson and Unruh, 2007].
30 30 −4
−5
The anomalous zone of low-velocity
20
Y (km)

20
10 10 anomalies between 6 and 12 km is
0 0 aseismic. The combination of the duc-
−10 −10 tile behavior and low-velocity features
−20 −20 suggests that the circumstance in this
−40 −20 0 20 40 −40 −20 0 20 40
X (km) X (km)
depth range is different and we will
discuss the possible conditions in the
(e) Depth= 0 km (f) Depth= 3 km
following section.
40 40
30 30 Comparing profiles D and E, we can see
20 that the Vp/Vs ratio varies at the surface
Y (km)

20
10 10 %Vp/Vs
5
inside the CGF, while profile D passes
0 0
−10
4 through the geothermal exploitation
−10
3
−20 −20 2
area and E is roughly 7 km away. The
−40 −20 0 20 40 −40 −20 0 20 40 1 Vp/Vs ratio is lower in the northern
0
part, the main exploitation areas with
(g) Depth= 6 km (h) Depth= 9 km −1
−2 steam and hot water at the surface,
40 40 −3
30 −4
than the southern part of the geother-
30
20 −5 mal field. The other striking features
Y (km)

20
10 10 include low Vp, low Vs, and low Vp/Vs
0 0 at the upper 3 km of the LLFZ, the IWV,
−10 −10 and the ALFZ. The Vp/Vs ratio changes
−20 −20
to around 1.81 at depth of 6 km
−40 −20 0 20 40 −40 −20 0 20 40
X (km) X (km) beneath the ALFZ and the adjacent
eastern part. Some small velocity bod-
Figure 8. Checkerboard test for the Vp/Vs model with blocks of 6 × 6 km.
ies of low Vp/Vs are also visible in cross
(a–d) The true velocity model. (e–h) The inverted velocity model. White
lines denote the RDE values greater than 0.1. sections b and c (Figure 11), which may
come from smearing of the low Vp/Vs
for the CGF.
3.6. In Situ Vp/Vs Ratios in Similar-Event Clusters
In order to complement our tomographic results, we estimate in situ Vp/Vs ratios within similar earthquake
clusters using the demeaned P and S wave differential times from waveform cross correlation by applying
the technique presented in Lin and Shearer [2007]. This technique assumes that the scale length of changes
in Vp/Vs ratios is greater than the size of the similar-event clusters and all the correlated events within each
individual compact cluster have the same local Vp/Vs ratio. It provides higher resolution for near-source
Vp/Vs ratios than typical tomographic inversion methods by using high-precision differential times and a
robust misfit function method, and it has been applied to study the near-source structure for the entire
southern California [Lin and Shearer, 2009], the rupture zone of the 1989 Loma Prieta earthquake [Lin and
Thurber, 2012], and Mammoth Mountain at the southwest rim of Long Valley caldera [Lin, 2013]. We applied
this approach to all the 1225 similar-event clusters in study area I and estimated standard uncertainties in
the in situ Vp/Vs ratios. These uncertainties are computed using the bootstrap approach [e.g., Efron and
Gong, 1983], in which the pairs of differential P and S times in the same cluster are randomly resampled 1000
times. In order to obtain the most robust results, we select 227 event clusters with uniformly distributed
events and uncertainties of Vp/Vs ratios less than 0.03.

ZHANG AND LIN ©2014. American Geophysical Union. All Rights Reserved. 4915
 Journal of Geophysical Research: Solid Earth 10.1002/2014JB010992

Figure 9. Map view of the final Vp model for study area II. Vp perturbations are shown relative to the layer-average
values, which are also given for each slice. Dashed circle represents the location of the Coso geothermal field. Black dots
show the grid nodes used in the tomographic inversions. Two red stars represent the geologic sites, SM and CHS. White
lines denote the RDE values greater than 0.1. See Figure 1 for the abbreviations of the major geological faults and valleys
around the geothermal field.

(a) Depth= 0 km (b) Depth= 3 km

CHS
SM CGF

Vp/Vs
WCF 1.95
AL
LL

FZ
FZ

1.90
IWV
1.80

1.76
(c) Depth= 6 km (d) Depth= 9 km
km 1.72

0 10 20
1.60

1.45

Figure 10. Map view of the final Vp/Vs model for study area II. The symbols are the same as in Figure 9.

ZHANG AND LIN ©2014. American Geophysical Union. All Rights Reserved. 4916
 Journal of Geophysical Research: Solid Earth 10.1002/2014JB010992

Figure 11. Cross sections of the Vp, Vp/Vs, and Vs structures across and in proximity to the CGF. The geothermal field is
denoted by the dashed circle with the red stars marking the locations of the SM and CHS. The five profiles are shown
by the black lines with profiles B, D, and E passing through the geothermal field and A and C off the CGF. Relocated
earthquakes within 5 km of both sides of the profiles are projected to the Vs cross sections, denoted by grey dots. White
contour lines represent the RDE values greater than 0.1. See Figure 1 for the abbreviations.

To compare with our tomographic results beneath the CGF, we project the in situ Vp/Vs ratios along profile
B in Figure 11. The in situ Vp/Vs ratios vary from 1.5 to 1.8 for the 36 clusters around the CGF (Figure 12).
The mean value of six clusters with slightly higher Vp/Vs (> 1.7) is 1.786, and these clusters mainly focus
around the area with Vp/Vs ratio of 1.77 inverted from the tomography. The other 30 clusters with low in
situ Vp/Vs ratios (≤ 1.7) are mainly distributed along the low-velocity zone within 20 km away from the CGF
resolved by the tomographic inversion. To further verify whether the observed low Vp/Vs anomalies from
6 to 12 km from tomography is reliable, we plot the in situ Vp/Vs ratios for each earthquake near the CGF
within similar-event clusters within this depth range. The median value of the in situ Vp/Vs ratios is 1.656,
consistent with the estimate of 1.667 from our tomography model. Therefore, the near-source in situ Vp/Vs
supports our tomographic results.

ZHANG AND LIN ©2014. American Geophysical Union. All Rights Reserved. 4917
 Journal of Geophysical Research: Solid Earth 10.1002/2014JB010992

(a) 4. Discussion
0
With the finer 3-D seismic velocity
model, we resolve the subsurface

Depth(km)
5
structure beneath the geothermal
field. The velocity of seismic waves
10
can be affected when they prop-
<=1.7
>1.7 agate through different materials
15
0 10 20 30 40 50 60 70
due to changes in lithology, mineral
composition, presence of fracture,
Distance(km)
temperature, pore pressure, and
Vp/Vs fluid saturation. Here we combine
1.45 1.63 1.67 1.71 1.75 1.79
the velocity structure, the seismic-
(b)
6 ity distribution, and the geological
and geochemical knowledge to
7 infer the dominating factors for the
observed anomalies.
8 4.1. Geothermal Reservoir
Depth (km)

Our tomographic results map the

9
feature of the geothermal reservoir
in the shallow depth. In map views
10
shown by Figures 9a and 10a, low
In−situ Vp/Vs Vp and low Vp/Vs anomalies domi-
11 Tomography
Median of In−situ nate the geothermal production area
Median of Tomo at the surface. Our velocity struc-
12
1.5 1.6 1.7 1.8 ture reflects the vapor-dominated
Vp/Vs geothermal field in high-temperature
system. Because seismic velocities
Figure 12. (a) In situ Vp/Vs ratios for the clusters around the CGF. Clusters
are projected along profile B in Figure 11. The tomography results are the reduce with increased temperatures
same as in Figure 11b2. Clusters with in situ Vp/Vs below 1.7 are shown [e.g., Mueller and Raab, 1997; Fielitz,
by red dots, and clusters with higher Vp/Vs are shown by green dots. (b) 1976; Sato et al., 1989; Ito et al., 1979],
Comparison of the in situ Vp/Vs and tomography results for each event we observe low Vp at the surface.
near the CGF between 6 and 12 km depth. Black dots show the in situ
Similar low Vp/Vs features within the
Vp/Vs ratios, and red dots denote tomography results. The median values
of the Vp/Vs at 1 km depth intervals are given by the stars. production area were observed by
previous studies [Walck, 1988; Lees
and Wu, 2000] and were interpreted
to be affected by vapor.
4.2. Magmatic System
Although upwelling magma beneath the geothermal field has been proposed in previous studies [e.g., Wu
and Lees, 1996, 1999; Lees and Wu, 2000; Lees, 2002], the velocity anomalies we observe beneath the CGF are
low Vp, low Vs, and low Vp/Vs between 6 and 12 km depth, which may be a candidate for magmatic materi-
als but refute the possibility of a magma chamber with a large amount of partial melt. The low-velocity zone
with ductile behavior is the typical feature associated with magma, but the low Vp/Vs ratios suggest a lack
of pervasive partial melt. To validate the low Vp/Vs structure, we conduct the recovery test for the observed
low Vp/Vs anomaly from the real data inversion underneath the CGF shown in Figure 11b2. A cross section
of the true and recovered Vp/Vs structure is shown in Figure 13. The anomalous body between 6 and 12 km
depth is recovered well, although slight horizontal smearing is observed.
We take into account the chemical composition of magma from other studies in Coso to further detect if the
anomalous velocity zone can refer to magmatic material. It has been argued that different silica contents
of magma, crack aspect ratios, and volume percentage of magma can make magmatic system behave with
both low or high Vp/Vs ratios [Nakajima et al., 2001; Hauksson and Unruh, 2007; Patane et al., 2006]. In our
tomographic results, the Vp and Vs anomalies show similar lateral and vertical geometry. This means that
the lithology influences both Vp and Vs but affects Vp more to result in low Vp/Vs. The anomalies do not
vary with depth, and we may assume a conductive temperature gradient and hydrostatic pore pressure

ZHANG AND LIN ©2014. American Geophysical Union. All Rights Reserved. 4918
 Journal of Geophysical Research: Solid Earth 10.1002/2014JB010992

0 in this anomalous zone. The mineral com-

position and fluid content may play an

Depth(km)
5 important role in affecting the seismic
velocity ratio among all the other factors.
10
The study by Christensen [1996] suggests
that the ratio is decreased with increased
15
30 20 10 0 10 20 30 silica contents for the rocks with 55–75%
Distance(km) of silica contents. Previous studies [e.g.,
Vp/Vs Sanders et al., 1995; Nakajima et al., 2001;
1.50 1.62 1.66 1.70 1.72 1.74 1.76 Lin and Shearer, 2009] reported that cracks
0 with a small volume percent of H2 O could
decrease Vp, Vs, and Vp/Vs. From the esti-
Depth(km)

5 mate of Nakajima et al. [2001], 2% of water

with aspect ratio of 0.1 can reduce Vp from
10
the original 6.31 km/s to 6.1 km/s and Vp/Vs
15 as low as 1.67 compared to the reference
30 20 10 0 10 20 30 value of 1.71, which is consistent with our
Distance(km)
observations in the upper and middle crust.
Figure 13. Recovery test for the Vp/Vs anomaly in the CGF. (top) The The presence of gas can also decrease Vp
true model and (bottom) the recovered model. The zero distance and Vp/Vs ratios because of the very low
refers to the location of the CGF. The Vp/Vs anomaly beneath the bulk modulus [Husen et al., 2004; Lin, 2013].
CGF corresponds to the anomaly at depths between 6 and 12 km
The geochemical study by Manley and
shown in Figure 11b2. The anomaly is recovered well in depth with
some horizontal smearing. Bacon [2000] pointed out that the erupted
magma in Coso was felsic with 66.5 wt %
of silica contents and saturated with an
H2 O-rich fluid, with the presence of the
vapor phase in melt inclusions. The per-
centage of H2 O in the melt inclusion can
be around 4.5–6.2 wt % in different dome
groups [Manley and Bacon, 2000], which is
Coso steam, water larger than the typical percentage of gas
N Range
CHS dissolved in magma, 0.2–3 wt %. Thus, we
DK Airport
SM Lake F
ault suggest that the low Vp, low Vs, and low
Vp/Vs body lacking of seismicity may indi-
0 km
cate that the molten material is felsic, rich in
gas, or with the inclusion of water. Although
5 km we cannot link the large quantities of hot
water near the surface to the presence of
water at 6 km depth, it is possible that water
10 km is rich in the anomalous velocity zone and
high content of water can facilitate magma
15 km
crystallization [Ritchey, 1980], which will
H2O
further lower Vp/Vs ratios.
Basaltic dike We show a conceptual model for subsurface
20 km structure in Figure 14. The content of the
magma reservoir erupted magma suggests that the material
is felsic, which is much more viscous than
Figure 14. Conceptual model of the subsurface structure for the mafic magma. The felsic magma originates
Coso geothermal field. The schematic diagram shows the main from the basaltic magma in deeper depth,
features derived from our results: (1) low-velocity zone beneath which ascended to crystallize and expelled
the geothermal field shown by Vp cross section; (2) shallow gas to induce fluid accumulation. Between
brittle-ductile depth (around 5 km depth) beneath the geother-
6 and 12 km depth beneath the CGF, the
mal field; (3) crystallizing magma body with the inclusion of
water between 6 and 12 km depth; and (4) water that may result magmatic material is likely to exist in the
from the underlying crystallizing basaltic magma. Lateral scale form of magma mush, with a small percent
is approximate. of melt trapped among small crystals. The

ZHANG AND LIN ©2014. American Geophysical Union. All Rights Reserved. 4919
 Journal of Geophysical Research: Solid Earth 10.1002/2014JB010992

silica-rich magma mush might also contain a high portion of water. Our model shows the velocity anomaly
up to 12 km depth, and it could be the interface of the silicic magma and mafic magma or the mafic magma
could be deeper. However, the ray coverage is insufficient to resolve the structure well at those depths.
4.2.1. Comparison With Previous Seismic Studies
The low-velocity body in the middle crust below the geothermal field has been observed in previous tomog-
raphy studies [e.g., Reasenberg et al., 1980; Walck and Clayton, 1987]. In the study by Hauksson and Unruh
[2007], variation in Vp/Vs model is too small to confirm the existence of the magma body and they interpret
the low Vp and normal Vp/Vs features to be the possibility of brine. We include more S wave arrival times
for the inversions and our Vp/Vs model shows low value of 1.667 beneath the geothermal field, which may
be associated with the magmatic system. Other factors causing differences between these two models are
the two-step inversion scheme and grid spacing of 3 km in this study. The 10 km grid used in Hauksson and
Unruh [2007] may hide some local anomalies with short wavelengths. Our Vp model is generally consis-
tent with their model but shows more variations in the vicinity of the geothermal field. Some features are
revealed by both models, such as the low Vp and low Vp/Vs anomalies beneath the IWV.
A low Vp anomaly beneath the CHS with the top at 5 km depth was detected by calculating teleseismic
receiver functions [Wilson et al., 2003]. They interpreted this as the top of the magma chamber and esti-
mated the amount of melts to be 1.5–5% by assuming a Vp/Vs ratio of 2.5. Their Vp/Vs model was obtained
by dividing the preferred 1-D P and S velocity models, which were used for the receiver function analysis.
Our high-resolution Vp/Vs model shows that the ratio is estimated to be around 1.667 and does not agree
with such an amount of melt.
The 3-D shear wave velocity model resolved by ambient noise tomography revealed a low shear velocity
zone between 6 and 12 km depth beneath the Coso geothermal field [Yang et al., 2011]. Their tomographic
results showed Vs structures at grid spacing of 0.25◦ × 0.25◦ . The low Vs body was estimated to be at the
size of 0.5◦ × 0.5◦ , which covers an area larger than the entire geothermal field. Our Vp and Vp/Vs models
show that the velocity varies inside the geothermal field and the prominent low Vp, low Vs, and low Vp/Vs
anomalies exist in the similar depth range as they indicated.

5. Conclusion
We developed 3-D high-resolution Vp and Vp/Vs models for the Coso region in southern California. A
by-product of this study is a high-precision earthquake relocation catalog. We focus our interest on explor-
ing the 3-D velocity structure in the vicinity of the Coso geothermal field. The tomographic results reveal
the features of the geothermal reservoir and the magmatic materials. Our observations suggest a lack of
a large amount of remnant melt from the Pleistocene volcanic activity underneath the geothermal field,
but the detected low-velocity anomalies could be associated with the magmatic system. The tomography
results reveal a low-velocity body of 10 km diameter in the depths between 6 and 12 km. Another technique
of estimating near-source in situ Vp/Vs was used to compare with our tomography models. The low in situ
Vp/Vs around the geothermal field is consistent with the low-velocity anomalies derived from the tomogra-
phy models. We interpret the 6 × 10 km2 low-velocity anomaly to be a region of hot and weak felsic rocks,
trapped with a series of small silicic magma chambers, sills, or dikes under the cooling phase with inclu-
sion of water. The data cannot show good resolutions below 15 km depth, and whether the low-velocity
anomalies extend to the deeper depth needs further investigation with more seismic data.
Acknowledgments
We thank the SCEDC for maintaining
the seismic data and making them References
available. We are grateful to Steven
Sherburn and an anonymous reviewer Adams, M., J. Moore, S. Bjornstad, and D. Norman (2000), Geologic history of the Coso geothermal system, Trans. Geotherm. Resour.
for their constructive comments. The Counc., 24, 205–210.
work has benefited from discussions Bacon, C., W. Duffield, and K. Nakamura (1980), Distribution of Quaternary rhyolite domes of the Coso range, California: Implications for
with J. Ole Kaven, Mel Erskine, Michael extent of the geothermal anomaly, J. Geophys. Res., 85(B5), 2425–2433.
Hasting, and Zhongwen Zhan. We Bacon, C., R. Macdonald, R. Smith, and P. Baedecker (1981), Pleistocene high-silica rhyolites of the Coso volcanic field, Inyo County,
give our special thanks to Ralf Weger California, J. Geophys. Res., 86(B11), 10,223–10,241.
for the petrology discussion. Funding Bishop, B. P., and D. K. Bird (1987), Variation in sericite compositions from fracture zones within the Coso hot springs geothermal system,
for this research was provided by the Geochim. Cosmochim. Acta, 51(5), 1245–1256.
National Science Foundation grant Christensen, N. I. (1996), Poisson’s ratio and crustal seismology, J. Geophys. Res., 101(B2), 3139–3156.
EAR-1045856. Plots are made using Combs, J. (1980), Heat flow in the Coso geothermal area, Inyo County, California, J. Geophys. Res., 85(B5), 2411–2424.
the public GMT (Generic Mapping Duffield, W., C. Bacon, and G. Dalrymple (1980), Late Cenozoic volcanism, geochronology, and structure of the Coso Range, Inyo County,
Tools) software. California, J. Geophys. Res., 85(B5), 2381–2404.

ZHANG AND LIN ©2014. American Geophysical Union. All Rights Reserved. 4920
 Journal of Geophysical Research: Solid Earth 10.1002/2014JB010992

Eberhart-Phillips, D. (1986), Three-dimensional velocity structure in northern California Coast Ranges from inversion of local earthquake
arrival times, Bull. Seismol. Soc. Am., 76(4), 1025–1052.
Eberhart-Phillips, D., and A. J. Michael (1998), Seismotectonics of the Loma Prieta, California, region determined from three-dimensional
Vp, Vp/Vs, and seismicity, J. Geophys. Res., 103(B9), 21,099–21,120.
Efron, B., and G. Gong (1983), A leisurely look at the bootstrap, the jackknife, and cross-validation, Am. Stat., 37(1), 36–48.
Feng, Q., and J. Lees (1998), Microseismicity, stress, and fracture in the Coso geothermal field, California, Tectonophysics, 289(1), 221–238.
Fialko, Y., and M. Simons (2000), Deformation and seismicity in the Coso geothermal area, Inyo County, California: Observations and
modeling using satellite radar interferometry, J. Geophys. Res., 105(B9), 21–21.
Fielitz, K. (1976), Compressional and shear wave velocities as a function of temperature in rocks at high pressure, in Explosion Seismology
in Central Europe, pp. 40–44, Springer, New York.
Hauksson, E. (2000), Crustal structure and seismicity distribution adjacent to the Pacific and North America plate boundary in southern
California, J. Geophys. Res., 105(B6), 13,875–13,903.
Hauksson, E., and J. Unruh (2007), Regional tectonics of the Coso geothermal area along the intracontinental plate boundary in central
eastern California: Three-dimensional Vp and Vp/Vs models, spatial-temporal seismicity patterns, and seismogenic deformation, J.
Geophys. Res., 112, B06309, doi:10.1029/2006JB004721.
Hauksson, E., W. Yang, and P. Shearer (2012), Waveform relocated earthquake catalog for southern California (1981 to June 2011), Bull.
Seismol. Soc. Am., 102(5), 2239–2244.
Husen, S., R. Smith, and G. Waite (2004), Evidence for gas and magmatic sources beneath the Yellowstone volcanic field from seismic
tomographic imaging, J. Volcanol. Geotherm. Res., 131(3), 397–410.
Ito, H., J. DeVilbiss, and A. Nur (1979), Compressional and shear waves in saturated rock during water-steam transition, J. Geophys. Res.,
84(B9), 4731–4735.
Kissling, E., W. Ellsworth, D. Eberhart-Phillips, and U. Kradolfer (1994), Initial reference models in local earthquake tomography, J. Geophys.
Res., 99(B10), 19,635–19,646.
Kissling, E., U. Kradolfer, and H. Maurer (1995), Program VELEST user’s guide—Short introduction. Institute of Geophysics, ETH Zurich.
Kisslinger, C., and E. Engdahl (1973), The interpretation of the Wadati diagram with relaxed assumptions, Bull. Seismol. Soc. Am., 63(5),
1723–1736.
Lees, J. (2002), Three-dimensional anatomy of a geothermal field, Coso, southeast-central California, Geol. Soc. Am. Mem., 195, 259–276.
Lees, J., and H. Wu (2000), Poisson’s ratio and porosity at Coso geothermal area, California, J. Volcanol. Geotherm. Res., 95(1–4), 157–173.
Lin, G. (2013), Seismic investigation of magmatic unrest beneath Mammoth Mountain, California, USA, Geology, 41, 847–850.
Lin, G., and P. Shearer (2007), Estimating local Vp/Vs ratios within similar earthquake clusters, Bull. Seismol. Soc. Am., 97(2), 379–388.
Lin, G., and P. M. Shearer (2009), Evidence for water-filled cracks in earthquake source regions, Geophys. Res. Lett., 36, L17315,
doi:10.1029/2009GL039098.
Lin, G., and C. H. Thurber (2012), Seismic velocity variations along the rupture zone of the 1989 Loma Prieta earthquake, California, J.
Geophys. Res., 117, B09301, doi:10.1029/2011JB009122.
Lin, G., P. M. Shearer, E. Hauksson, and C. H. Thurber (2007a), A three-dimensional crustal seismic velocity model for southern California
from a composite event method, J. Geophys. Res., 112, B11306, doi:10.1029/2007JB004977.
Lin, G., P. Shearer, and E. Hauksson (2007b), Applying a three-dimensional velocity model, waveform cross correlation, and cluster
analysis to locate southern California seismicity from 1981 to 2005, J. Geophys. Res., 112, B12309, doi:10.1029/2007JB004986.
Manley, C., and C. Bacon (2000), Rhyolite thermobarometry and the shallowing of the magma reservoir, Coso volcanic field, California, J.
Petrol., 41(1), 149–174.
Monastero, F., A. Katzenstein, J. Miller, J. Unruh, M. Adams, and K. Richards-Dinger (2005), The Coso geothermal field: A nascent
metamorphic core complex, Geol. Soc. Am. Bull., 117(11–12), 1534–1553.
Monastero, F. C., J. D. Walker, A. M. Katzenstein, and A. E. Sabin (2002), Neogene evolution of the Indian Wells Valley, east-central
California, in Geologic Evolution of the Mojave Desert and Southwestern Basin and Range, edited by A. F. Glazner, J. D. Walker, and J. M.
Bartley, Geol. Soc. Am. Mem. 195, 199–228, Boulder, Colo.
Mueller, H., and S. Raab (1997), Elastic wave velocities of granite at experimental simulated partial melting conditions, Phys. Chem. Earth,
22(1), 93–96.
Nakajima, J., T. Matsuzawa, A. Hasegawa, and D. Zhao (2001), Three-dimensional structure of Vp, Vs, and Vp/Vs beneath northeastern
Japan: Implications for arc magmatism and fluids, J. Geophys. Res., 106(B10), 21,843–21,857, doi:10.1029/2000JB000008.
Patane, D., G. Barberi, O. Cocina, P. De Gori, and C. Chiarabba (2006), Time-resolved seismic tomography detects magma intrusions at
Mount Etna, Science, 313(5788), 821–823.
Reasenberg, P., W. Ellsworth, and A. Walter (1980), Teleseismic evidence for a low-velocity body under the Coso geothermal area, J.
Geophys. Res., 85, 2471–2483.
Reyners, M., D. Eberhart-Phillips, G. Stuart, and Y. Nishimura (2006), Imaging subduction from the trench to 300 km depth beneath the
central North Island, New Zealand, with Vp and Vp/Vs, Geophys. J. Int., 165(2), 565–583.
Ritchey, J. L. (1980), Divergent magmas at crater lake, Oregon: Products of fractional crystallization and vertical zoning in a shallow,
water-undersaturated chamber, J. Volcanol. Geotherm. Res., 7(3), 373–386.
Saltus, R. W., and A. H. Lachenbruch (1991), Thermal evolution of the Sierra Nevada: Tectonic implications of new heat flow data,
Tectonics, 10(2), 325–344.
Sanders, C., S. Ponko, L. Nixon, and E. Schwartz (1995), Seismological evidence for magmatic and hydrothermal structure in Long Valley
caldera from local earthquake attenuation and velocity tomography, J. Geophys. Res., 100(B5), 8311–8326.
Sato, H., I. S. Sacks, and T. Murase (1989), The use of laboratory velocity data for estimating temperature and partial melt fraction in the
low-velocity zone: Comparison with heat flow and electrical conductivity studies, J. Geophys. Res., 94(B5), 5689–5704.
Shearer, P., E. Hauksson, and G. Lin (2005), Southern California hypocenter relocation with waveform cross-correlation, part 2: Results
using source-specific station terms and cluster analysis, Bull. Seismol. Soc. Am., 95(3), 904–915.
Sherburn, S., R. White, and M. Chadwick (2006), Three-dimensional tomographic imaging of the Taranaki volcanoes, New Zealand,
Geophys. J. Int., 166(2), 957–969.
Thurber, C., and D. Eberhart-Phillips (1999), Local earthquake tomography with flexible gridding, Comput. Geosci., 25(7), 809–818.
Thurber, C. H. (1983), Earthquake locations and three-dimensional crustal structure in the Coyote lake area, central California, J. Geophys.
Res., 88(B10), 8226–8236.
Walck, M. (1988), Three-dimensional Vp/Vs variations for the Coso region, California, J. Geophys. Res., 93(B3), 2047–2052.
Walck, M., and R. Clayton (1987), P wave velocity variations in the Coso region, California, derived from local earthquake travel times, J.
Geophys. Res., 92, 393–405.

ZHANG AND LIN ©2014. American Geophysical Union. All Rights Reserved. 4921
 Journal of Geophysical Research: Solid Earth 10.1002/2014JB010992

Wamalwa, A. M., K. L. Mickus, L. F. Serpa, and D. I. Doser (2013), A joint geophysical analysis of the Coso geothermal field, south-eastern
California, Phys. Earth Planet. Inter., 214, 25–34, doi:10.1016/j.pepi.2012.10.008.
Wilson, C., C. Jones, and H. Gilbert (2003), Single-chamber silicic magma system inferred from shear wave discontinuities of the crust and
uppermost mantle, Coso geothermal area, California, J. Geophys. Res., 108(B5), 2226, doi:10.1029/2002JB001798.
Wu, H., and J. Lees (1996), Attenuation structure of Coso geothermal area, California, from wave pulse widths, Bull. Seismol. Soc. Am.,
86(5), 1574–1590.
Wu, H., and J. M. Lees (1999), Three-dimensional P and S wave velocity structures of the Coso geothermal area, California, from
microseismic travel time data, J. Geophys. Res., 104(B6), 13,217–13,233.
Yang, Y., M. Ritzwoller, and C. Jones (2011), Crustal structure determined from ambient noise tomography near the magmatic centers of
the Coso region, southeastern California, Geochem. Geophys. Geosyst., 12, Q02009, doi:10.1029/2010GC003362.

ZHANG AND LIN ©2014. American Geophysical Union. All Rights Reserved. 4922