Physics of the Earth and Planetary Interiors 230 (2014) 114

Contents lists available at ScienceDirect

Physics of the Earth and Planetary Interiors

journal homepage: www.elsevier.com/locate/pepi

Crust and upper mantle structure of the New Madrid Seismic Zone:
Insight into intraplate earthquakes
Chuanxu Chen a,b,c,, Dapeng Zhao a,, Shiguo Wu b
a
Department of Geophysics, Tohoku University, Sendai 980-8578, Japan
b
Institute of Oceanography, Chinese Academy of Science, Qingdao, China
c
University of Chinese Academy of Sciences, Beijing, China

a r t i c l e i n f o a b s t r a c t

Article history: We determine a 3-D P-wave velocity model of the crust and upper mantle down to 400 km depth to
Received 18 July 2013 investigate structural heterogeneity and its inuences on the generation of intraplate earthquakes in
Accepted 23 January 2014 the New Madrid Seismic Zone. We used 4871 high-quality arrival times from 187 local earthquakes
Available online 18 February 2014
and 30,846 precise travel-time residuals from 1041 teleseismic events recorded by the EarthScope/USAr-
Edited by Vernon F. Cormier
ray Transportable Array. Our results show that, beneath the Reelfoot rift, a signicant low-velocity (low-
V) zone exists in the upper mantle down to 200 km depth, with a large volume of 200  200  150 km3.
Keywords:
The origin of the low-V zone may be related to the passage of the Bermuda hotspot and the stalled
New Madrid Seismic Zone
Seismic tomography
ancient Farallon slab materials foundering in the mantle transition zone. This low-V zone may have rel-
Low-velocity zone atively low shear strength and act as a viscously weak zone embedded in the lithosphere, being apt to
Intraplate earthquakes concentrate tectonic stress and transfer stress to the seismogenic faults in the upper crust, leading to
Farallon slab the large intraplate earthquakes in the New Madrid Seismic Zone.
Seismotectonics 2014 Elsevier B.V. All rights reserved.

1. Introduction
The New Madrid Seismic Zone (NMSZ) is seismically the most
active region in the Central and Eastern United States and an ideal
area to study intraplate earthquakes. A sequence of at least three
large earthquakes (M P 7.0) occurred here in 18111812 (e.g.,
Hough et al., 2003; Johnston and Schweig, 1996), and palaeo-seis-
mic records show evidence of large earthquakes about 500 years
apart in the past 2000 years (e.g., Kelson et al., 1996; Tuttle et al.,
2002). The distribution of local earthquakes recorded since 1974
delineates three linear faults in the NMSZ (see Fig. 1): (1) the NE-
trending Cottonwood Grove-Blytheville Arch fault along the cen-
tral Reelfoot rift, (2) the NW-trending Reelfoot Fault, and (3) the
NNE-trending New Madrid North Fault (Johnston and Schweig,
1996). The activation of these mid-continental faults and their con-
trols on duration of the seismic activity remain poorly understood.
One of the fundamental questions is: what makes the NMSZ differ-
ent from the surrounding intraplate areas in North America, espe-
cially the areas within the same geologic settings?
Corresponding authors. Address: Department of Geophysics, Tohoku University,
Sendai 980-8578, Japan. Tel.: +81 22 225 1950; fax: +81 22 264 3292 (C. Chen).
E-mail addresses: chencxiocas@gmail.com (C. Chen), zhao@aob.gp.tohoku.ac.jp
(D. Zhao).
Several models have been proposed to explain the New Ma-
drid seismic activity, such as a sinking mac body in a reacti-
vated lower crust (Pollitz et al., 2001), relaxation of a weak
zone within an elastic lithosphere (Kenner and Segall, 2000; Liu
and Zoback, 1997), unloading from deglaciation (Wu and John-
ston, 2000) or river incision (Calais et al., 2010), and a localized
mantle ow induced by deep mantle dynamics (Forte et al.,
2007). In this work, we mainly test the model of a weak zone
within the elastic lithosphere, because such a weak zone may ex-
hibit seismic-velocity anomalies that may be detected by seismic
tomography. This model postulates a weak zone within the lower
crust and upper mantle, and the intraplate earthquakes are
caused by stress transferred to the seismogenic faults from the
mechanically weak zone (Kenner and Segall, 2000). However,
there has been no clear evidence for such a weak zone and no
obvious mechanism for the weakening (e.g., Calais et al., 2010)
shown in the previous studies.
In this work we present a high-resolution three-dimensional (3-
D) P-wave velocity model of both the crust and upper mantle un-
der the NMSZ and its surrounding areas using a new and high-
quality data set of both local earthquakes and teleseismic events
recorded by the USArray. The present results shed new light on
the crustal and upper-mantle structure and mechanism of the
intraplate earthquakes in the NMSZ.

http://dx.doi.org/10.1016/j.pepi.2014.01.016
0031-9201/ 2014 Elsevier B.V. All rights reserved.
2 C. Chen et al. / Physics of the Earth and Planetary Interiors 230 (2014) 114

39N

Illinois

38N

Missouri Kentucky
37N 1843

1812 New Madrid

1812

36N 1811 ift

Arkansas o tR
lfo
1895
R ee
Greenbrier Tennessee
6.0
35N 5.0
4.0
Mississippi 3.0
Mississippi Embayment 2.0

93W 92W 91W 90W 89W 88W

Fig. 1. Seismicity in the New Madrid Seismic Zone and adjacent areas. The red circles show the earthquakes compiled by the Center for Earthquake Research and Information
catalogues from April 1974 to December 2012. The three black stars show the estimated epicenters of the large earthquakes (Mw P 7.0) which occurred in 1811 to 1812
(Johnston and Schweig, 1996; Hough et al., 2003). The two diamonds show the estimated epicenters of the 1843 Marked Tree earthquake (M 6.3) and the 1895 Charleston
earthquake (M 6.6). The black dashed line shows the estimated boundary of the Mississippi embayment, and the two blue lines delimit the Palaeozoic Reelfoot rift. The blue
triangle shows the town New Madrid and Greenbrier. The inset map shows the location of the present study area (red box). Note that the cluster of seismicity near Greenbrier
was associated with 2 earthquake swarms in 1981 and 2001 and recent seismicity was associated with waste water injection wells from hydrofrac operations in the gas elds
there. (For interpretation of the references to color in this gure legend, the reader is referred to the web version of this article.)

2. Data and method
Two sets of seismic data are used to conduct tomographic inver-
sions. One data set consists of arrival times from local earthquakes
which occurred in and around the NMSZ (Fig. 2). The other is rela-
tive travel-time residuals from teleseismic events (Fig. 3). We used
the rst P-wave arrival times of both local and teleseismic events,
which were collected precisely by the staff of the USArray Network
Facility, and the recording period was from November 2010 to
October 2012. During this period 63 seismic stations were densely
and uniformly distributed in the study area (34.239.2N, 87.4
93.0W), and the station spacing is about 50 km (Fig. 2).
We selected the local earthquakes carefully based on the fol-
lowing criteria: (1) only events occurring during the local night
time (02:0012:00 UTC) were selected in order to exclude local
quarry blasts which were conducted only in daytime; (2) each
event was recorded by over 6 seismic stations; (3) the event mag-
nitude (Mb) is P1.5; (4) before conducting tomographic inversion
we relocated these local events using the Geiger method (Geiger,
1912; Zhao et al., 1992) and selected the events with uncertainties
of <2 km in epicentral location and <8 km in focal depth. As a re-
sult, our data set contains 2500 rst P-wave and 2371 S-wave arri-
val times from 187 local earthquakes (Fig. 2).
We also collected 1041 teleseismic events with magnitudes
M P 4.5. Each event has at least 10 recordings, and the average
number of data per event is up to 30. The distances between the
teleseismic events and the center of the study area are restricted
to 30100 (Fig. 3), in order to avoid inuences of the complex
structures at the bottom of the mantle and those in the upper man-
tle outside the present study region (Zhao et al., 1994, 2013). The
azimuthal coverage of the teleseismic events is fairly complete
(Fig. 3). In total, we used 31,846 P-wave arrival times from the
1041 teleseismic events selected.
Relative travel-time residuals of the teleseismic events are com-
puted and used in the tomographic inversion in order to avoid the
effects of the complex structures in the crust and upper mantle un-
der the hypocenters of every teleseismic events (Zhao et al., 1994).
We rst compute raw travel-time residuals by subtracting the the-
oretical travel times from the observed ones, and the 1-D iasp91
Earth model (Kennett and Engdahl, 1991) is adopted to calculate
the theoretical travel times. Then we calculate the mean residual
for each event averaged over all the recording stations. Finally, rel-
 C. Chen et al. / Physics of the Earth and Planetary Interiors 230 (2014) 114
39N
38N
37N
36N
35N
93W 92W 91W 90W 89W 88W
Fig. 2. Epicentral distribution of 187 local crustal earthquakes (M > 1.5; green
circles) which are used in the tomographic inversion. The red triangles denote the
63 USArray seismic stations installed from November 2010 to October 2012. (For
interpretation of the references to color in this gure legend, the reader is referred
to the web version of this article.)
ative travel-time residuals are obtained by removing the mean
residual from every raw residual. Although the global mantle het-
erogeneities outside the study area may also affect the teleseismic
residuals, their inuences on the relative residuals of every telese-
ismic events are very small, and so would not affect the nal tomo-
graphic result (Zhao et al., 2013).
Fig. 4a shows the distribution of mean relative travel-time
residuals at every stations, which are obtained by averaging the
relative residuals over all the teleseismic events. The mean relative
travel-time residuals exhibit a spatial pattern of delayed arrivals at
stations in the Mississippi Embayment and the southern part of
Illinois, and early arrivals at stations in Ozark plateau and the geo-
logic highland region at east of the Mississippi Embayment in Ten-
nessee. This distribution pattern of relative residuals indicates that
low-velocity (low-V) anomalies exist under the Mississippi Embay-
ment and the New Madrid Seismic Zone.
The area with delayed arrivals coincides with the Mississippi
Embayment (Fig. 4a), raising a possibility that low-V sediments
may affect the tomographic images in the lower crust and even
the upper mantle as an artifact of the 3-D velocity inversion. We
made crustal correction to evaluate the effects on the relative tra-
vel-time residuals caused by the lateral crustal heterogeneity. The
details of the crustal correction, including the choice of crustal
model and the calculation of relative travel-time residuals, are
shown in the supplementary materials. Fig. 4c shows the distribu-
tion of crustal corrections at every stations. The positive values are
distributed within the Mississippi Embayment, and the value in-
creases where the sediment becomes thicker (Fig. 6a). Fig. 4b
shows the distribution of mean relative travel-time residuals after
the crustal correction. As compared with the result before the crus-
tal correction, the general trend of the mean residual distribution
of the teleseismic events remains the same, whereas there are
some differences in the amplitude, indicating that the low-V sedi-
ment alone is not sufcient to explain the observed delayed arriv-
als within the Mississippi Embayment. The delayed arrivals after
3
the crustal correction require low-V anomalies in the upper man-
tle, as shown in our nal inversion result (Fig. 13).
The local and teleseismic joint inversion method (Zhao et al.,
1994, 2012b) is adopted to study the 3-D P-wave velocity structure
of the crust and upper mantle under the NMSZ and surrounding
areas. An efcient 3-D ray tracing technique (Zhao et al., 1992) is
used to calculate the theoretical travel times and ray paths. For
the local events, the hypocenters are relocated iteratively in the
inversion process, and their raw travel-time residuals are used in
the tomographic inversion. For the teleseismic events, relative tra-
vel-time residuals are used in the inversion so as to minimize the
errors introduced by the hypocentral mislocations, origin times
and the path effects outside the target volume. The LSQR algorithm
(Paige and Saunders, 1982) with damping and smoothing regular-
izations (Zhao et al., 1994, 2012b) is used to solve the large and
sparse system of observation equations which relate the travel-
time residuals to the unknown hypocentral and velocity
parameters.
3. Model parameterization and inversion
An optimal 1-D velocity model, including depth variations of
major velocity discontinuities in the crust under the study region,
can minimize the overall mist between the model predictions and
observations. In this study, we constructed an initial velocity mod-
el (Fig. 5) for tomographic inversion by combining a 1-D crustal
velocity model of Chiu et al. (1992) with the iasp91 Earth model
(Kennett and Engdahl, 1991). The crust contains 5 layers with con-
stant velocity and depth variations of three velocity discontinuities
which are the bottom of the alluvial sediment located in the Mis-
sissippi Embayment, the Conrad discontinuity, and the Moho dis-
continuity (Table 1). The geometries of the three discontinuities
are extracted from the newly integrated Central United States
Velocity Model (CUSVM) version 1 (Ramirez-Guzman et al., 2012,
http://earthquake.usgs.gov/research/cus_seisvelmodel/). P-wave
velocity of the alluvial sediment is very low (1.82 km/s), and its
thickness varies from 0 to 1.4 km (Fig. 6a), resulting in a large tra-
vel-time variation (Zhang et al., 2009). The Conrad depth ranges
from 14 to 27 km (Fig. 6b), whereas the Moho depth ranges from
37 to 46 km (Fig. 6c).
A 3-D grid is set up in the modeling space to express 3-D veloc-
ity variations. The grid intervals in the horizontal and vertical
directions affect the distribution of hit counts (i.e., number of rays
passing around each grid node) in the model, and so ultimately af-
fect the resolution of the tomographic image. After many tests, we
set the bottom of the modeling space at 1000 km depth, for which
the root-mean-square (RMS) travel-time residual becomes the
minimum. We nd that the optimal grid for our data set is as fol-
lows: in the horizontal direction the grid interval is 0.4 (approxi-
mately 40 km), while in the vertical direction 7 layers are set at 5,
20, 40, 80, 120, 160 and 200 km depths, and 13 layers with a ver-
tical interval of 50100 km are set in the depth range of 200
1000 km. Fig. 8 shows plan views of the hit counts at nine depths
in the crust and upper mantle, demonstrating the ray path cover-
age beneath the study area. The ray path coverage in the upper
mantle is very good except for the edge portions. The ray sampling
becomes lower in the crust, in particular, to the southeast of the
NMSZ, due to the lack of local earthquakes there (Fig. 2).
Before the 3-D velocity inversion, we rst relocated the 187 lo-
cal earthquakes using all the P and S wave arrival times to obtain
more accurate hypocentral parameters for the tomographic inver-
sion. The RMS travel-time residual for the local events is reduced
from 0.634 to 0.486 s after the relocation (Fig. 9). Fig. S1 shows
changes in the hypocenter locations of the local events before
and after the relocation. The changes in the epicenters are rela-
4 C. Chen et al. / Physics of the Earth and Planetary Interiors 230 (2014) 114

W E
30

60

90

10
0

S
Fig. 3. Epicenter locations of 1041 teleseismic events (blue crosses) used in this study. The concentric circles correspond to epicentral distances of 30, 60, 90 and 100 degrees.
The thin green lines denote the plate boundaries. The red square shows the present study area. (For interpretation of the references to color in this gure legend, the reader is
referred to the web version of this article.)

tively small (0.52 km), while the changes in the focal depths are
slightly larger, being 15 km.
We conducted many tomographic inversions with different val-
ues of the damping and smoothing parameters (Fig. S2 and Fig. S3).
Fig. 7a and b show the trade-off curves for the variance of velocity
perturbations and RMS travel-time residuals. Considering the bal-
ance between the reduction of travel-time residuals and the
smoothness of the 3-D velocity model (Eberhart-Phillips, 1986),
we found that the optimal values of the damping and smoothing
parameters are 5.0 and 500.0, respectively. The nal P-wave RMS
travel-time residual is reduced from 0.468 to 0.331 s after the joint
inversion of the local and teleseismic data.
4. Resolution tests
We conducted two kinds of resolution tests to evaluate the ade-
quacy of ray coverage and reliability of the tomographic result. One
is checkerboard resolution tests (CRTs) for evaluating the spatial
resolution of the entire tomographic image (Zhao et al., 1992).
The other is synthetic tests to conrm the main features of the
tomographic result (Zhao et al., 1992, 2012b).
In the CRTs, alternative positive and negative velocity anoma-
lies (4%) were assigned to the 3-D grid nodes in the modeling
space. Then we calculated synthetic travel times for the checker-
board model with the same locations of earthquakes and stations
as those in the real data set. The synthetic data were then inverted
using the same starting velocity model and inversion algorithm as
those for the real data. Fig. 10 shows the CRT results at nine repre-
sentative depths with a grid interval of 0.4 (approximately
40 km). The CRT results are straightforward and easy to distin-
guish. The spatial resolution is high for most of the study region
(Fig. 10ci), though it is slightly lower at the edge and southeastern
corner of the study area in the upper crust (Fig. 10a and b), due to
the lack of local earthquakes there (Fig. 2) during the short record-
ing period. We also conducted a CRT by adding random errors hav-
ing a normal distribution with a standard deviation of 0.1 s to the
synthetic data to simulate the picking errors in the observed arrival
times. The test result (Fig. 11) shows that the checkerboard pattern
is well recovered for most of the study region, though the anomaly
amplitude is not fully recovered in some areas.
 C. Chen et al. / Physics of the Earth and Planetary Interiors 230 (2014) 114 5

93 92 91 90 89 88 93 92 91 90 89 88
39 39

38 38

37 37

36 36

35 35

(a) (b)
Before crustal correction After crustal correction
39

38

37

0.45s 0.45s
36

35

(c)
93 92 91 90 89 88
Crustal corrections
Fig. 4. (a) Distribution of raw relative travel-time residuals for the teleseismic events, calculated with respect to the network average. (b) Distribution of relative travel-time
residuals after crustal corrections. (c) Distribution of crustal corrections (see text for details). Red crosses and blue dots represent delayed and early P-wave arrivals,
respectively. The scale for the travel-time residuals is shown below (b). (For interpretation of the references to color in this gure legend, the reader is referred to the web
version of this article.)

Table 1
1-D crustal velocity model adopted (modied from Vlahovic et al., 2000; Chiu et al.,
1992).

Layer Vp (km/s) Vs (km/s) Bottom depth (km)

L1 1.8 0.72 Sediment [01.6]
L2 6.02 3.48 2.65
L3* 4.83 2.79 4.65
L4 6.17 3.56 Conrad [1430]
L5 6.65 3.84 Moho [3648]
*
L3 at 2.654.65 depth is a low-velocity layer separating Paleozoic carbonate and
clastic sedimentary rocks (Vp = 6.02 km/s) from the crystalline upper crust
(Vp = 6.17 km/s).

15). The procedure of the synthetic tests is the same as that of

Fig. 5. The 1-D P-wave velocity model which is taken as the starting velocity model
for the 3-D tomographic inversion. The blue box shows the crust and uppermost
mantle; its details are shown in Table 1. (For interpretation of the references to
color in this gure legend, the reader is referred to the web version of this article.)
We carried out synthetic tests (Fig. 12) to further conrm the
main features of the obtained tomographic images (Figs. 13 and
the checkerboard resolution tests; the only difference between
them is the input model. As shown in Fig. 12a and b, the input
model of the synthetic test contains four anomalies to simulate
the main features of the nal tomographic images: a high-velocity
(high-V) anomaly at 050 km depth, a low-V anomaly at 70
170 km depth beneath the NMSZ, and two high-V anomalies at
70170 km depths to the northwest of the NMSZ (Fig. 12a and
b). The test results show that the pattern and amplitudes of the
four velocity anomalies in the input model are generally well
6 C. Chen et al. / Physics of the Earth and Planetary Interiors 230 (2014) 114

93 92 91 90 89 88 93 92 91 90 89 88
39 (a) 39 (b)

19
38 38

37 37

24
36 36
23
22

0.1
21
35 35

0.2
20
0.3

0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 14 16 18 20 22 24 26 28 30

Sediment depth (km) Conrad depth (km)
93 92 91 90 89 88
39 (c)

38
43

37 40

36
43

35

36 38 40 42 44 46 48
Moho depth (km)
Fig. 6. Depth distribution of (a) the bottom of the Mississippi Embayment, (b) the Conrad and (c) the Moho discontinuities in the study area, taking from the newly integrated
central US velocity model (Ramirez-Guzman et al., 2012). The depth scale is shown below each map.

recovered, though slight smearing occurs in the vertical direction
(Fig. 12c).
These checkerboard and synthetic tests demonstrate that the
spatial resolution of our tomographic images is at least 40 km
and the major velocity anomalies in the tomographic images are
reliable features.
5. Results and discussion
Our nal tomographic images are shown in Figs. 1315 and
Fig. S4. A NE-SW trending low-V anomaly is revealed in the upper-
most crust beneath the Reelfoot rift (Fig. 13a), especially around
the New Madrid city, which is similar to the image at the same depth
in a recent double-difference tomography (Fig. 9a in Dunn et al.
(2013)). High-V anomalies are revealed in the middle crust
(Fig. 13b) beneath the NMSZ, which are similar to those imaged by
a previous local and teleseismic tomography (Zhang et al., 2009). A
high-V zone is visible in the lower crust (Fig. 13c), which is also
revealed by a recent surface-wave tomography (Fig. S1 in Pollitz
and Mooney, 2013) and previous seismic refraction studies (Mooney
et al., 1983; Catchings, 1999). These high-V anomalies might be the
expression of an inferred rift pillow (Mooney et al., 1983; Stuart
et al., 1997) or a sinking mac body (Pollitz et al., 2001) in the lower
crust. The most notable feature in the upper mantle at depths of
50200 km (Figs. 13df, 14 and 15) is the contrast between the
low-V anomalies below the NMSZ and high-V anomalies to the
northwest of the NMSZ. The Western boundary of the Reelfoot rift
coincides with the boundary between the southeastern low-V zone
and the northwestern high-V zone, and the northern end of the Reel-
foot rift overlies the northern end of the upper-mantle low-V zone
below the NMSZ (Fig. 13d and e). Our inferred pattern of velocity
anomalies in the upper mantle is also similar to those obtained by
surface-wave imaging (Pollitz and Mooney, 2013) and other P-wave
tomography results (Zhang et al., 2009), though there are slight
differences due to the difference in the methodology and ray-path
coverage. Vertical cross-sections (Fig. 14) and a 3-D view (Fig. 15)
of our tomographic images show that the upper-mantle low-V zone
 C. Chen et al. / Physics of the Earth and Planetary Interiors 230 (2014) 114 7

0.36
70
(a) 00
(b)
0.35 50
50
RMS travel time residual (sec)

RMS travel time residual (sec)

0.35
00
20
30 0.34
0.34
00
10
20
0.33 0.33 0
50
10
0.32
5 0.32 0
20
1
0.31 0
10
0.31
50
0.30
0 1 2 3 4 5 6 7 8 9 10 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5
RMS velocity perturbation (%) RMS velocity perturbation (%)

Fig. 7. (a, b) Trade-off curve for the variance of velocity perturbations and root-mean-square (RMS) travel-time residuals. The numbers beside the circles in (a) denote values
of the damping parameter, while those in (b) denote values of the smoothing parameter. The black circles denote the optimal damping parameter in (a) and smoothing
parameter in (b).

is located at 50200 km depths with a large volume of about revealed by a seismic anisotropy study (Fig. 4b in Deschamps et al.
200  200  150 km3. The lower boundary of this low-V zone is (2008), and Fig. 4a in Yuan and Romanowicz (2010)), implying that
roughly consistent with the lithosphereasthenosphere boundary this low-V anomaly is located in the lithosphere.

92 90 88 92 90 88 92 90 88

38 38

36 36

5 km 20 km 40 km
92 90 88 92 90 88 92 90 88

38 38

36 36

80 km 120 km 160 km
92 90 88 92 90 88 92 90 88

38 38

36 36

200 km 250 km 300 km

0 500 1000 1500 2000 2500

Hit count

Fig. 8. Distribution of hit counts (number of rays passing around each grid node) at nine depths. The hit-count scale is shown at the bottom.
8 C. Chen et al. / Physics of the Earth and Planetary Interiors 230 (2014) 114
800
1D model
before relocation
600
Number of data
400
200
0 0
3 2 1 0 1 2
traveltime residual (s)
800
3D model
before relocation
600
Number of data
400
200
0 0
3 2 1 0 1 2
traveltime residual (s)
Fig. 9. (a, b) Distribution of travel-time residuals before and after the relocation of local earthquakes, using the 1-D velocity model shown in Fig. 5. (c, d) Distribution of travel-
time residuals before and after the relocation of local earthquakes, using the nal 3-D velocity model obtained by this study.
5.1. Structural heterogeneity in the crust and upper mantle
The existence of high-V zones in the crust and a low-V zone in
the upper mantle down to 200 km depth beneath the NMSZ makes
the Reelfoot rift different from the other ancient rifts in the central
US, such as the Midcontinent Rift, the Rome trough and the Bir-
mingham trough. The latter ones are associated with relatively
low velocity in the crust and uppermost mantle (<65 km), without
distinct velocity contrast in the lithospheric mantle (Pollitz and
Mooney, 2013). The Reelfoot rift formed within the Eastern Ryo-
lite-Granite Province in the Early-to-Middle Cambrian (520
500 Ma), since then this region and its surrounding areas have
experienced four major tectonic events which have possibly mod-
ied the lithospheric mantle and been responsible for the unique
low-V anomaly in the upper mantle under the NMSZ: the Allegha-
nian Orogeny in the Permian Period (300250 Ma), the lithosphere
thickening beneath the Grenville-Appalachian orogen to the south-
east of the NMSZ during the post-orogenic period (160125 Ma)
(Deschamps et al., 2008), the passage of the Bermuda hotspot be-
neath the Mississippi Embayment ca. 100 Ma (Van Arsdale and
Cox, 2007), and the rst-stage subduction of the Farallon slab
reaching the central-eastern America ca. 55 Ma (Sigloch et al.,
2008).
McKenna et al. (2007) argue that there is no compelling case for
assuming that the NMSZ is signicantly hotter than its surrounding
areas according to the surface heat-ow data, thus the low-V
anomaly cannot, at least not solely, be explained with current ther-
mal anomalies. A possible scenario for the mid-lithospheric low-V
zone is the presence of water and/or melts, which may result in
800
(a) 1D model (b)
after relocation
600
400
200
3 3 2 1 0 1 2 3
traveltime residual (s)
800
(c) 3D model (d)
after relocation
600
400
200
3 3 2 1 0 1 2 3
traveltime residual (s)
partial melting (e.g., Thybo and Perchuc, 1997; Thybo, 2006; Sato
et al., 1989) or metasomatism of a mantle peridotite (Pollitz and
Mooney, 2013; Frost, 2006) at the pressure and temperature con-
ditions of the observed low-V anomaly in the lithosphere, whereas
this scenario raises a question on the source of such water and/or
melts. The passage of the Bermuda hotspot might have played such
a role and resulted in mantle compositional changes in two ways:
(1) geneartion of minerals (for example, antigorite, phlogopite,
chlorite and talc) by metasomatic alternation, which can reduce
seimic velocity; (2) inltration of primitive mantle which is rela-
tively dense relative to the surroundring mantle, accompanied
with the ascending fertile melts from the hotspot (Pollitz and Moo-
ney, 2013).
In this work we envisage another source of uids and melts,
which is the stalled Farallon slab material foundering on the
670 km discontinuity (Sigloch et al., 2008). Regional and global
tomographic studies show that the ancient Farallon slab has sub-
ducted down to the lower mantle under North America and part
of it has stalled in the mantle transition zone (e.g., Ren et al.,
2007; Sigloch et al., 2008; Zhao et al., 2013). This deeply subducted
Farallon slab may have also contributed to the surface dynamics
(Forte et al., 2007) and hydrated the lithosphere and asthenosphere
under the central and eastern U.S., and the slab has been proposed
as a source of volatile-enrichment (Van der Lee et al., 2008). It is
possible that deep dehydration of the Farallon slab has resulted
in upwelling of hot asthenospheric materials, similar to that in
the big mantle wedge above the stagnant Pacic slab under East
Asia (e.g., Zhao et al., 2007; Tian and Zhao, 2011, 2013; Wei
et al., 2012; Zhao and Tian, 2013). The dipping low-V zone
 C. Chen et al. / Physics of the Earth and Planetary Interiors 230 (2014) 114 9

Fig. 10. Results of a checkerboard resolution test for nine representative layers. The depth of each layer is shown below each map. In this test the lateral grid interval is 0.4
(40 km). The open and solid circles denote low and high velocities, respectively. The velocity perturbation (in %) scale is shown at the bottom.

(Fig. 14b,d and Fig. S4b, c), which is visible in the upper mantle
from 400 km depth and connects with the low-V anomaly be-
neath the NMSZ, may be associated with such a wet and hot
upwelling ow.
5.2. Insight into intraplate earthquakes
The existence of a signicant low-V zone in the upper mantle
under the NMSZ is a very prominent feature, and the large intra-
plate earthquakes are spatially correlated with this low-V anomaly
(Fig. 14). The Reelfoot rift overlies the gap between the low-V zone
and the high-V zone at its northwestern side (Figs. 14b and 15).
This spatial coincidence implies that the intraplate seismicity at
the NMSZ is mechanically related to the underlying low-V zone
in the upper mantle, which is consistent with the notion of stress
concentration near rheological boundaries (e.g., Artyushkov,
1973; Zhang et al., 2009).
The lithosphere to the northwest of the Reelfoot rift is rela-
tively strong in strength, which might represent the dry and de-
pleted cratonic lithosphere. In contrast, in the upper-mantle
10 C. Chen et al. / Physics of the Earth and Planetary Interiors 230 (2014) 114

Fig. 11. The same as Fig. 11, but random errors with a standard deviation of 0.1 s are added to the synthetic arrival times.

39

A B A B
38 0 0 4%
4%
A 3%
4 1
50 50
37 2%
Depth (km)

3
Depth (km)

4% 4%
100 100 1%
dVp

2 3
4% 0%
36
1 150 150 1 %
4 2
2 %
35 B 200 200
3 %
(a) (b) (c)
250 250 4 %
93 92 91 90 89 88 0 50 100 150 200 250 300 350 400 450 500 550 0 50 100 150 200 250 300 350 400 450 500 550
Distance (km) Distance (km)

Fig. 12. (a, b) Map view and a vertical cross-section of the input model of a synthetic test (see text for details). The location of the vertical cross-section is shown in (a). (c)
Output result of the synthetic test. The velocity perturbation scale (in %) is shown on the right.
 C. Chen et al. / Physics of the Earth and Planetary Interiors 230 (2014) 114 11

92 90 88 92 90 88 92 90 88

38 38

36 36

(a) 5 km (b) 20 km (c) 40 km

92 90 88 92 90 88 92 90 88

38 38

36 36

(d) 80 km (e) 120 km (f) 160 km

92 90 88 92 90 88 92 90 88

38 38

36 36

(g) 200 km (h) 250 km (i) 300 km

6 % 3 % 0 % 3% 6%
dVp
Fig. 13. Map views of the obtained P-wave tomography at nine layers in the crust and upper mantle. The depth of each layer is shown below each map. Red and blue colors
denote low and high velocities, respectively. The velocity perturbation scale is shown at the bottom. The black dashed lines show the boundaries of the Reelfoot rift. The black
stars and diamonds denote the large intraplate earthquakes as shown in Fig. 1. (For interpretation of the references to color in this gure legend, the reader is referred to the
web version of this article.)

low-V zone beneath the Reelfoot rift, elevated water and/or melts
content, also possible production (layered phyllosilicates) of the
metasomatism (Pollitz and Mooney, 2013), may reduce the vis-
cosity and strength of the lithosphere and form a weak zone
embedded within the elastic lithosphere. This weak zone may
promote the occurrence of intraplate earthquakes in two ways.
One is that such a weak zone may relax stress when localized
perturbation happens, for example, a thermal or uid pressure
perturbation or a transient change in regional stress, and the
stress may ultimately be transferred to the seismogenic zone
and result in stress concentration in the upper crust, as being
demonstrated by a numerical modeling study (Kenner and Segall,
2000). The other aspect is that the accumulation of uid and/or
melts associated with this low-V zone may hydrate the litho-
sphere and result in lithospheric thinning, making it possible to
develop a shear zone whereas a homogeneously dry lithosphere
12 C. Chen et al. / Physics of the Earth and Planetary Interiors 230 (2014) 114

A (a) A C (c) C
0 0

100 100

Depth (km) 200 200

300 300

400 400
0 100 200 300 400 500 0 100 200 300 400

B (b) B D (d) D
0 0

100 100
Depth (km)

200 200

300 300

400 400
0 100 200 300 400 500 0 100 200 300 400
(e)
39
C
38 B A

37
6 % 3 % 0% 3% 6%
D D dVp
36

35

A C B
93 92 91 90 89 88

Fig. 14. (ad) Vertical cross-sections of P-wave tomography. Red and blue colors denote low and high velocities, respectively. The velocity perturbation scale (in %) is shown
below (d). The black curved lines denote the Conrad and Moho discontinuities. The locations of the proles are shown in (e). White stars denote the ve large historic
earthquakes as shown in Fig. 1, whereas the open circles show the smaller local earthquakes which occurred within a 30-km width of each prole. (For interpretation of the
references to colour in this gure legend, the reader is referred to the web version of this article.)

cannot (Regenauer-Lieb et al., 2001). The relatively thin (1) Our tomographic images conrm the existence of the postu-
lithosphere in the shear zone is an effective mechanism for stress lated weak zone under the NMSZ in the form of a pro-
concentration around a strike-slip fault zone when a localized nounced low-V zone at 50200 km depth, which is deeper
stress perturbation happens. Fluids and/or melts related low-V and larger than the presumed weak zone.
zones are also revealed in source areas of many large crustal (2) The lithospheric modication under the Reelfoot rift may be
earthquakes in Japan, China and India (e.g., Zhao et al., 2002; related to the passage of the Bermuda hotspots ca. 100 Ma,
Mishra and Zhao, 2003; Huang and Zhao, 2009; Cheng et al., as well as subsequent deep dehydration of the ancient Faral-
2011). Interestingly, recent studies show that even the distribu- lon slab, which may have caused the low-V zone under the
tion and generation of deep moonquakes are inuenced by struc- NMSZ.
tural heterogeneities in the lunar mantle (Zhao et al., 2008, (3) The large crustal earthquakes in the NMSZ are generally
2012a), suggesting similar seismogenic processes in both Earth located close to the boundary between the low and high
and Moon. velocity anomalies in the lower crust and upper mantle,
where the lithospheric strength is weak. Tectonic stress
may concentrate there and be transferred to the upper crust,
6. Conclusions leading to the large intraplate earthquakes.

We determined a detailed 3-D P-wave velocity model of the

crust and upper mantle down to 400 km depth under the NMSZ
and its surrounding areas using a larger number of high-quality ar-
rival-time data from local earthquakes and teleseismic events
simultaneously. Main ndings of the present work are summarized
as follows:
Acknowledgements
We thank the data center of the US EarthScope/Transportable
Array for providing the high-quality seismic data used in this
 C. Chen et al. / Physics of the Earth and Planetary Interiors 230 (2014) 114
if
tR
oo
lf
ee
R
39
38
37
36
-90
-91
-92
35-93
Fig. 15. A three-dimensional view of main features of the velocity structure in the study region. Red and blue colors denote low and high velocity anomalies, respectively. The
isosurface of velocity anomalies is rendered where P-wave velocity perturbation is 3% and 3% from 5 to 400 km depth. The yellow lines denote the USA state boundaries.
(For interpretation of the references to colour in this gure legend, the reader is referred to the web version of this article.)
study. We appreciate the helpful discussions with Drs. Y. Tian, W.
Wei and X. Liu. Most of the gures were made using GMT (Wessel
and Smith, 1998). This work was supported by a grant from Japan
Society for the Promotion of Science (Kiban S-11050123) to D.
Zhao, the Global-COE Program of Earth and Planetary Sciences of
Tohoku University, and the Strategic Priority Research Program of
Chinese Academy of Science (Grant No. XDA11030102). An anony-
mous reviewer and Prof. V. Cormier (the Editor) provided thought-
ful review comments and suggestions which have improved this
paper.
Appendix A. Supplementary data
Supplementary data associated with this article can be found, in
the online version, at http://dx.doi.org/10.1016/j.pepi.2014.01.016.
References
Artyushkov, E.V., 1973. Stresses in the lithosphere caused by crustal thickness
inhomogeneities. J. Geophys. Res. 78, 76757708.
Calais, E., Freed, A., Van Arsdale, R., Stein, S., 2010. Triggering of New Madrid
seismicity by late-Pleistocene erosion. Nature 466, 608611.
Catchings, R.D., 1999. Regional Vp, Vs, Vp/Vs, and Poissons ratios across earthquake
source zones from Memphis, Tennessee, to St. Louis, Missouri. Bull. Seismol.
Soc. Am. 89, 15911605.
Cheng, B., Zhao, D., Zhang, G., 2011. Seismic tomography and anisotropy in the
source area of the 2008 Iwate-Miyagi earthquake (M 7.2). Phys. Earth Planet.
Inter. 184, 172185.
Chiu, J., Johnston, A., Yang, Y., 1992. Imaging the active faults of the central New
Madrid seismic zone using PANDA array data. Seism. Res. Lett. 63, 375393.
Deschamps, F., Lebedev, S., Meier, T., Trampert, J., 2008. Stratied seismic
anisotropy reveals past and present deformation beneath the East-central
United States. Earth Planet. Sci. Lett. 274, 489498.
Dunn, M., DeShon, H.R., Powell, C.A., 2013. Imaging the New Madrid Seismic Zone
using double-difference tomography. J. Geophys. Res. 118, 54045416.
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, 10251052.
Forte, A., Mitrovica, J., Moucha, R., Simmons, N., Grand, S., 2007. Descent of the
ancient Farallon slab drives localized mantle ow below the New Madrid
Seismic Zone. Geophys. Res. Lett. 34, L04308.
13
t
5
100
m)
200
th ( k
Dep
300
400
-88
-89
Frost, D.J., 2006. The stability of hydrous mantle phases. Rev. Mineral. Geochem. 62,
243271.
Geiger, L., 1912. Probability method for the determination of earthquake epicenters
from the arrival time only. Bull. St. Louis. Univ. 8, 6071.
Hough, S.E., Seeber, L., Armbruster, J.G., 2003. Intraplate triggered earthquakes:
observations and interpretation. Bull. Seismol. Soc. Am. 93, 22122221.
Huang, J., Zhao, D., 2009. Seismic imaging of the crust and upper mantle under
Beijing and surrounding regions. Phys. Earth Planet. Inter. 173, 330348.
Johnston, A.C., Schweig, E.S., 1996. The enigma of the New Madrid earthquakes of
18111812. Ann. Rev. Earth Planet. Sci. 24, 339384.
Kelson, K.I., Simpson, G.D., VanArsdale, R.B., Haraden, C.C., Lettis, W.R., 1996.
Multiple late Holocene earthquakes along the Reelfoot fault, central New
Madrid Seismic Zone. J. Geophys. Res. 101, 61516170.
Kenner, S.J., Segall, P., 2000. A mechanical model for intraplate earthquakes:
application to the New Madrid Seismic Zone. Science 289, 23292332.
Kennett, B., Engdahl, E., 1991. Traveltimes for global earthquake location and phase
identication. Geophys. J. Int. 105, 429465.
Liu, L., Zoback, M.D., 1997. Lithospheric strength and intraplate seismicity in the
New Madrid seismic zone. Tectonics 16, 585595.
McKenna, J., Stein, S., Stein, C.A., 2007. Is the New Madrid Seismic Zone hotter
and weaker than its surroundings? Special papers. Geol. Soc. Am. 425, 167175.
Mishra, O.P., Zhao, D., 2003. Crack density, saturation rate and porosity at the 2001
Bhuj, India, earthquake hypocenter: a uid driven earthquake? Earth Planet. Sci.
Lett. 212, 393405.
Mooney, W.D., Andrews, M.C., Ginzburg, A., Peters, D.A., Hamilton, R.M., 1983.
Crustal structure of the northern Mississippi Embayment and comparison with
other continental rift zones. Tectonophysics 94, 327348.
Paige, C.C., Saunders, M.A., 1982. LSQR: an algorithm for sparse linear equations and
sparse least squares. Assoc. Comput. Math. Software 8, 4371.
Pollitz, F.F., Kellogg, L., Brgmann, R., 2001. Sinking mac body in a reactivated
lower crust: a mechanism for stress concentration at the New Madrid Seismic
Zone. Bull. Seismol. Soc. Am. 91, 18821897.
Pollitz, F. F., Mooney, W.D., 2013. Seismic structure of the Central US crust and
shallow upper mantle: Uniqueness of the Reelfoot Rift. Earth Planet. Sci. Lett.
http://dx.doi.org/10.1016/j.epsl.2013.05.042.
Ramirez-Guzman, L., Boyd, O.S., Hartzell, S., Williams, R.A., 2012. Seismic velocity
model of the central United States (Version 1): description and simulation of
the 18 April 2008 Mt. Carmel, Illinois, earthquake. Bull. Seismol. Soc. Am. 102,
26222645.
Regenauer-Lieb, K., Yuen, D.A., Branlund, J., 2001. The initiation of subduction:
criticality by addition of water? Science 294, 578580.
Ren, Y., Stutzmann, E., Van Der Hilst, R.D., Besse, J., 2007. Understanding seismic
heterogeneities in the lower mantle beneath the Americas from seismic
tomography and plate tectonic history. J. Geophys. Res. 112, B01302.
Sato, H., Sacks, I.S., Murase, T., 1989. The use of laboratory velocity data for
estimating temperature and partial melt fraction in the low-velocity zone:
comparison with heat ow and electrical conductivity studies. J. Geophys. Res.
94, 56895704.
14 C. Chen et al. / Physics of the Earth and Planetary Interiors 230 (2014) 114
Sigloch, K., McQuarrie, N., Nolet, G., 2008. Two-stage subduction history under
North America inferred from multiple-frequency tomography. Nat. Geosci. 1,
458462.
Stuart, W.D., Hildenbrand, T.G., Simpson, R.W., 1997. Stressing of the New Madrid
Seismic Zone by a lower crust detachment fault. J. Geophys. Res. 102, 27623
27633.
Thybo, H., Perchuc, E., 1997. The seismic 8 discontinuity and partial melting in
continental mantle. Science 275, 16261629.
Thybo, H., 2006. The heterogeneous upper mantle low velocity zone.
Tectonophysics 416, 5379.
Tian, Y., Zhao, D., 2011. Destruction mechanism of the North China Craton: insight
from P and S wave mantle tomography. J. Asian Earth Sci. 42, 11321145.
Tian, Y., Zhao, D., 2013. Reactivation and mantle dynamics of North China Craton:
insight from P-wave anisotropy tomography. Geophys. J. Int. 195, 17961810.
Tuttle, M.P., Schweig, E.S., Sims, J.D., Lafferty, R.H., Wolf, L.W., Haynes, M.L., 2002.
The earthquake potential of the New Madrid Seismic Zone. Bull. Seismol. Soc.
Am. 92, 20802089.
Vlahovic, G., Powell, C.A., Chiu, J., 2000. Three-dimensional P wave velocity
structure in the New Madrid Seismic Zone. J. Geophys. Res. 105, 79998011.
Van Arsdale, R.B., Cox, R.T., 2007. The Mississippis curious origins. Sci. Am. 296 (1),
7682.
van der Lee, S., Regenauer-Lieb, K., Yuen, D.A., 2008. The role of water in connecting
past and future episodes of subduction. Earth Planet. Sci. Lett. 273, 1527.
Wei, W., Xu, J., Zhao, D., Shi, Y., 2012. East Asia mantle tomography: new insight into
plate subduction and intraplate volcanism. J. Asian Earth Sci. 60, 88103.
Wessel, P., Smith, W., 1998. New, improved version of generic mapping tools
released. Eos Trans. AGU 79, 579.
Wu, P., Johnston, P., 2000. Can deglaciation trigger earthquakes in North America?
Geophys. Res. Lett. 27, 13231326.
Yuan, H., Romanowicz, B., 2010. Lithospheric layering in the North American craton.
Nature 466, 10631068.
Zhang, Q., Sandvol, E., Liu, M., 2009. Lithospheric velocity structure of the New
Madrid Seismic Zone: a joint teleseismic and local P tomographic study.
Geophys. Res. Lett. 36, L11305.
Zhao, D., Tian, Y., 2013. Changbai intraplate volcanism and deep earthquakes in East
Asia: a possible link? Geophys. J. Int. 195, 706724.
Zhao, D., Hasegawa, A., Horiuchi, S., 1992. Tomographic imaging of P and S wave
velocity structure beneath northeastern Japan. J. Geophys. Res. 97, 19909
19928.
Zhao, D., Hasegawa, A., Kanamori, H., 1994. Deep structure of Japan subduction zone
as derived from local, regional, and teleseismic events. J. Geophys. Res. 99,
2231322329.
Zhao, D., Mishra, O., Sanda, R., 2002. Inuence of uids and magma on earthquakes:
seismological evidence. Phys. Earth Planet. Inter. 132, 249267.
Zhao, D., Maruyama, S., Omori, S., 2007. Mantle dynamics of western Pacic to East
Asia: new insight from seismic tomography and mineral physics. Gondwana
Res. 11, 120131.
Zhao, D., Lei, J., Liu, L., 2008. Seismic tomography of the Moon. Chin. Sci. Bull. 53,
38973907.
Zhao, D., Arai, T., Liu, L., Ohtani, E., 2012a. Seismic tomography and geochemical
evidence for lunar mantle heterogeneity: comparing with Earth. Global Planet.
Change 90, 2936.
Zhao, D., Yanada, T., Hasegawa, A., Umino, N., Wei, W., 2012b. Imaging the
subducting slabs and mantle upwelling under the Japan Islands. Geophys. J. Int.
190, 816828.
Zhao, D., Yamamoto, Y., Yanada, T., 2013. Global mantle heterogeneity and
its inuence on teleseismic regional tomography. Gondwana Res. 23,
595616.