Académique Documents
Professionnel Documents
Culture Documents
OBSs were deployed over about half the prole, covering the
Spratly Islands area (Dangerous Grounds province), the oceanic
domain and the distal portion of the northern conjugate margin
(i.e. OBS-1 to OBS-24). The remaining OBSs were provided by the
Institute of Geology and Geophysics, Chinese Academy of Sciences,
and cover the rest of the northern margin (i.e. OBS-25 to OBS-50).
For technical reasons, the originally planned seismic reection
line across the set of OBSs could not be shot, so that previous
seismic lines around were further used to interpret the refraction
results, in particular the sub-surface structures.
The rationale of the experiment was to focus as much as possible
on the conjugate continental domains and transition zones, using a
reasonable number of OBSs. The mean distance between OBSs was
thus kept close to 18 km in the continental parts (Fig. 2). A larger
and variable spacing was used in the oceanic domain, which was
not the primary goal of the study. Thus, oceanic crustal velocities
were obtained with fewer details.
All OBSs were recovered successfully and only two did not re-
cord any data (OBS-28 and OBS-50). The source was designed to
investigate deep crustal structures and consisted of an array of Bolt
air guns with a total volume of w105 l. Shots were distance trig-
gered every 150 m (leading to a 1 min trace length). A total of ca.
7000 shots were recorded on each OBS. The sample interval used
was 1, 4 or 8 ms, depending on the OBS type. All records were
resampled to 4 ms for the nal processing. The OBS records are of a
good overall quality (Fig. 3). For each instrument, the hydrophones
and the vertical component of the geophones were used for seismic
phase identications. Clock drifts were checked and corrections
were applied. A relocation procedure was applied using water wave
travel times and multibeam bathymetry. On average, relocated
positions of the OBSs showed a drift of the order of w300 m along
inline and/or cross line directions.
2.2. Seismic phase identication
We manually picked a total of 41,100 rst refraction arrivals (Pg
phases, rays refracted in the sediments or the crust and Pn phases,
rays refractedinthe upper mantle). Pnphases weredetectedas far as
200 km offset in many cases (Fig. 3A) and occasionally beyond. A
picking error ranging from30 ms to125 ms was assigned depending
on the signal to noise ratio, using the empirical parameterization of
Zelt andForsyth(1994). A200ms pickingerror was attributedtoless
certain Pg and Pn picks. Most OBSs did not show any sedimentary
second arrival, while at near offset, the rst refraction arrivals most
often showed too high of an apparent velocity to imply a thick
sedimentary pile (Fig. 3A). However, where sedimentary basins do
exist (mainly on the northern margin) refraction phases from the
sediments are rst arrivals (Fig. 3B and C).
PmP phases (i.e. reection phases from the crust-mantle
boundary) were identied on most of the OBSs deployed above
the continental crust (Fig. 3) and correspond to a set of 6622 picks.
A 150 ms picking error was assigned for this reection phase. No
PmP arrival was recognized in the oceanic basin (Fig. 3D). For OBSs
located close to the continental slope, Pg arrivals display a strong
asymmetrical pattern suggesting a sharp transition between the
oceanic crust and the continental crust in termof velocity structure
(Fig. 3E). In most cases where PmP arrivals were picked, the Pg-Pn
crossover was recognized (known as triplication point) providing
better constrains for the determination of the Moho depth (Fig. 3C).
Some OBS data display a Pg phase at very far offset where the Pn
phase is already clearly visible, suggesting a strong lateral velocity
variation within the continental crust (Fig. 3F).
Figure 2. A) Multibeam bathymetric map along the OBS prole (see Fig. 1 for location). B) Free air gravity anomaly map (Sandwell and Smith, 2009). The positions of the OBS are
reported in yellow dots. The red open circles correspond to OBS that did not record any data. The magnetic anomalies interpreted by Briais et al. (1993) are indicated in thick dashed
lines. Bathymetric (C) and gravimetric (D) cross-sections along the OBS prole. Note that the regional trend of the free air gravity anomaly along most of the OBS prole is zero. (For
interpretation of the references to color in this gure legend, the reader is referred to the web version of this article).
T. Pichot et al. / Marine and Petroleum Geology xxx (2013) 1e17 3
Please cite this article in press as: Pichot, T., et al., Deep crustal structure of the conjugate margins of the SW South China Sea from wide-angle
refraction seismic data, Marine and Petroleum Geology (2013), http://dx.doi.org/10.1016/j.marpetgeo.2013.10.008
2.3. Modeling procedure
Different 2-D velocity modeling strategies are used for wide-
angle seismic data: forward modeling (Zelt and Smith, 1992), in-
verse modeling (Zelt and Barton, 1998; Van Avendonk et al., 1998;
Korenaga et al., 2000) or a combination of both (Zelt and Smith,
1992). We use here the joint inversion method of Korenaga et al.
(2000) that combines refraction and reection travel times. The
main reason for choosing Korenagas inversion scheme is that it
does not require strong a priori information, apart from a loose
initial velocity model. This is particularly important in our case
since we did not collect the seismic reection line. Inverting for the
depth of a oating reector (here, the Moho interface) is also
possible. The rationale is to provide a simple model that ts the
data with little risk of pre-inversion over-interpretation. All
methods (forward or inverse modeling) produce non-unique re-
sults. As such, we will later discuss the preferred results in the light
of what is known of the geology of the area.
The velocity model is parameterized as a two-dimensional
sheared mesh grid. To allow for a higher level of sub-surface de-
tails, vertical grid cells size increases from 200 m at the top to
500 m at the bottom of the model, with a constant horizontal
length of 500 m. The multibeambathymetric data denes the top of
the sheared grid. Because no seismic reection data has been ac-
quired collectively with the refraction seismic experiment, the
interface between sediments and the basement has not been
introduced in the initial velocity model. As stated in Section 2.2,
sedimentary refracted arrivals are always rst arrivals at the
Figure 3. Receiver-gather of OBS data using the vertical component of the geophone (top). First refracted (red) and reected (green) arrival picks with their corresponding error
bars (middle). Observed (colors) and modeled travel times (black dots) using the nal velocity model presented in Figure 4 (bottom). The vertical axis is in reduced travel time and
the horizontal axis represents the offset from the OBS position (in kilometers). The reduced travel time is expressed as T-offset/(reduced velocity). A refraction phase with an
apparent velocity higher than the chosen reduced velocity would have travel times decreasing with offset, while a phase with a lower apparent velocity would have travel times
increasing with offsets. (For interpretation of the references to color in this gure legend, the reader is referred to the web version of this article).
T. Pichot et al. / Marine and Petroleum Geology xxx (2013) 1e17 4
Please cite this article in press as: Pichot, T., et al., Deep crustal structure of the conjugate margins of the SW South China Sea from wide-angle
refraction seismic data, Marine and Petroleum Geology (2013), http://dx.doi.org/10.1016/j.marpetgeo.2013.10.008
continental domain water depths. As a result, the inversion should
retrieve the large sedimentary basins without difculty.
A Bouguer gravity inversion has been performed to dene a
crude Moho interface to be used as input in the initial velocity
model. The inversion procedure is using Parkers Fourier formula-
tion for gravity (Parker, 1973) widely applied in sedimentary basins
(details of our procedure can be found in Chamot-Rooke et al., 1997;
see Braitenberg et al., 2006; Nguyen and Nguyen, 2013 for the latest
applications to the SCS). Since the aim was to obtain a smooth
starting velocity model, we did not try to introduce sophisticated
corrections such as variable crustal densities due to the nature of
the crust (continental versus oceanic) or variable mantle density
due to the thermal anomaly related to the SCS opening. A standard
2700 kg/m
3
crustal density was used to rst derive a Bouguer
anomaly based on the multibeam bathymetry. The resulting
anomaly was then converted to a Moho geometry using a crust/
mantle contrast of 450 kg/m
3
. The reference depth for the inversion
was set to 20 km, so that the gravity inverted oceanic crust thick-
ness was kept close to 6 km, and the depth of the continental Moho
interface was in good agreement with previous seismic refraction
studies (Nissen et al., 1995; Pin et al., 2001; Qiu et al., 2011; Ruan
et al., 2011; L et al., 2011). The initial velocity model is
composed of a linear velocity gradient from 4 km/s below the
seaoor interface to 7.5 km/s at the crust-mantle boundary (a ve-
locity of 7.5 km/s is quite high but this avoids a strong velocity
contrast at the crust-mantle boundary in the initial velocity model).
Avelocity of 8 km/s is applied in the shallowupper mantle (Fig. 4A).
The inversion method used here requires a large set of param-
eters (detailed in Table 1; see also Korenaga et al., 2000). The key-
parameters are summarized hereafter. Both velocity and depth
perturbations are controlled by predened vertical and horizontal
correlations lengths (Toomey et al., 1994). After exploring a large set
of parameters, the preferred values for the correlation lengths are
10e30 km horizontally and 0e24 km vertically, giving a value of
Figure 3. (continued).
T. Pichot et al. / Marine and Petroleum Geology xxx (2013) 1e17 5
Please cite this article in press as: Pichot, T., et al., Deep crustal structure of the conjugate margins of the SW South China Sea from wide-angle
refraction seismic data, Marine and Petroleum Geology (2013), http://dx.doi.org/10.1016/j.marpetgeo.2013.10.008
several hundreds meters at the seaoor and nearly 12 km at 20 km
depth (the rst value is applied to the top of the grid, the second is
applied to the bottom, while linear interpolation is performed in
between).
To minimize the computation time and the memory cost of the
runs, the w1000-km-long prole was divided into two parts, a 740-
km-long northern prole (PR1), and a 458-km-long southern pro-
le (PR2). The proles share a common oceanic domain around
OBS-18 (Fig. 5).
3. Results of the velocity modeling
3.1. Robustness of the results
As mentioned above, the result of the inversion is non-unique.
The nal velocity model presented in Figure 4B has the lowest
nal root mean square (RMS) residual and reduced c
2
among the
tested models and satises the rst order geology (discussed
later). After 12 iterations, the nal RMS residual is nearly equal for
both the northern (PR1) and southern (PR2) proles with a value
of w160 ms. Note that the average picking uncertainties is slightly
higher for PR1 than for PR2 because of a lower signal-to-noise
ratio for OBSs along PR1. The nal c
2
values are 1.55 and 2.02 for
PR1 and PR2, respectively (see Table 2 for details). The expected
reduced c
2
values should be close to 1 if the post-inversion travel
time residuals were close to the initial picking uncertainties.
However, this does not take into account the aliasing caused by
the w18 km OBS spacing. Considering that the basement interface
was not constrained by seismic reection data we explain values
of c
2
above 1 by unresolved basement topography. The higher c
2
for PR2 illustrates this, as it is likely due to the combination of
lower average uncertainties and higher amplitudes of seaoor and
basement topography at short wavelengths (i.e. Dangerous
Grounds).
Figure 3. (continued).
T. Pichot et al. / Marine and Petroleum Geology xxx (2013) 1e17 6
Please cite this article in press as: Pichot, T., et al., Deep crustal structure of the conjugate margins of the SW South China Sea from wide-angle
refraction seismic data, Marine and Petroleum Geology (2013), http://dx.doi.org/10.1016/j.marpetgeo.2013.10.008
For both proles PR1 and PR2, the refracted ray coverage (Pg
and Pn) is particularly dense in the rst 10 km (Fig. 5AeB). The
refracted ray density is slightly higher in PR2 than in PR1 below
10 km. Many refracted arrivals turn into the shallow upper mantle
(Pg) giving good constraints on the velocity at the base of the
model (Fig. 5AeB).
Reected rays (PmP) are traced along sizeable portions of the
proles: w500 km for PR1 (between km 30 and 540) and 200 km
for PR2 (between km 200 and 400; Fig. 5CeD). In particular, con-
tinental slopes adjacent to the oceanic domain show good wide-
angle reection ray coverage. Rays from several OBSs cover the
crust of the ContinenteOcean Transitions (COTs) on both conjugate
margins, showing a reasonably good coverage (Fig. 5). However, the
resolution of the velocity model is limited in these areas since
almost no crustal arrival is recorded as rst arrival (crustal arrivals
are rst arrivals only at a very short offset). As a result, the obtained
nal velocity gradient is similar to the initial velocity gradient.
Velocities are also poorly constrained in the central part of the
oceanic domain, refracted rays being recorded by OBS-18 only. As a
result, the obtained nal velocity there does not really improve
compared to the initial velocity gradient. Furthermore, no reection
phase was identied at the oceanic crust/mantle boundary.
Small-amplitude and small-wavelength undulations of the
Moho were found in many of the test runs. They are actually
superimposed onto more stable long-wavelength Moho variations.
These small-wavelength undulations may be artifacts related to the
Figure 3. (continued).
T. Pichot et al. / Marine and Petroleum Geology xxx (2013) 1e17 7
Please cite this article in press as: Pichot, T., et al., Deep crustal structure of the conjugate margins of the SW South China Sea from wide-angle
refraction seismic data, Marine and Petroleum Geology (2013), http://dx.doi.org/10.1016/j.marpetgeo.2013.10.008
difculty in the inversion to weight reected phases with respect to
refracted ones. If real, a plausible interpretation is that the Moho
discontinuity is not a sharp boundary, but rather a complex tran-
sition between the lower crust and the shallow upper mantle
(Korenaga and Kelemen, 1997; Ziegler, 2001). Seismic rays would
then possibly not be reected by a simple reector, but interact
with a heterogenous velocity gradient in the vicinity of the Moho, at
the base of the lower continental crust.
3.2. Velocity structures
Velocities directly below the seaoor are well below 4 km/s, the
assigned value there in the initial velocity model. They increase
rapidly with depth in the rst few kilometers from 1.6 km/s to 5.0e
6.0 km/s (Fig. 6AeB). Despite the absence of basement interface in
the initial velocity model, known major sedimentary basins and
basement highs are fairly well imaged. These basins are, fromNorth
to South, the Quiongdongnan Basin, the Guangle Uplift (also known
as the Triton Ridge), the Zhongjannan Basin (also known as the
Triton Lower Terrace), the southwestern area of the Penxi Sea-
mounts province and isolated basins and basement highs of the
Spratly islands (Fig. 4B). Narrow sedimentary basins are also
modeled at the feet of the continental slopes of both margins. The
largest of these basins is at the foot of the northern margin.
Within the upper 5 km, 1-D velocity-depth proles (Figs. 6DeE)
show larger variations over PR1 when compared to PR2. This is
essentially due to the presence of two large sedimentary basins on
the northern margin, the Quiondongnan Basin (tr1; Figs. 4B and
6D) and the Zhongjannan Basin (tr3; Figs. 4B and 6D). In the upper
section (rst 5 km), the 1.6 km/s to 5 km/s velocity isocontours
showsmall perturbations with an associated wavelength bracketed
between 15 and 30 km(Fig. 4B). This wavelength is probably not an
artifact since it is signicantly greater than the scaling parameters
used in the inversion, the vertical and horizontal correlation
lengths at shallow depth being much smaller. The wavelength of
these shallow perturbations varies along the prole. This is not
Figure 3. (continued).
T. Pichot et al. / Marine and Petroleum Geology xxx (2013) 1e17 8
Please cite this article in press as: Pichot, T., et al., Deep crustal structure of the conjugate margins of the SW South China Sea from wide-angle
refraction seismic data, Marine and Petroleum Geology (2013), http://dx.doi.org/10.1016/j.marpetgeo.2013.10.008
simply due to the velocity perturbation inherited from the seaoor
topography propagating further down into the velocity model.
Indeed, in the Zhongjannan Basin, between OBS-41 and OBS-36, the
upper part of the sedimentary cover (1.6e4.0 km/s) does not pre-
sent such perturbation while, velocity isocontours just below
(ranging between 4.0 and 5.5 km/s) show clearly the small wave-
length perturbation (Fig. 4B). The same wavelength can be seen in
the topography of the seaoor and in the free-air gravity anomaly
(especially along PR2; Fig. 2). In the discussion, we suggest that
these structures relate to the uppermost crust horst-and-graben
small-scale fabric.
A rather sharp change in the velocity gradient occurs in all
proles at w5.0e6.0 km/s (Fig. 5AeB). At intermediate depth (be-
tween 5 and 15 km below the sea level), the nal velocity model
shows strong lateral variation for velocities ranging between
w5.7 km/s to less than 6.7 km/s. Within this velocity range, the
entire prole, including PR1 and PR2, shows alternating bodies of
high and low velocity (Fig. 4B). The Low Velocity Bodies (LVBs)
show smaller velocity values when compared to the velocity
average of both PR1 and PR2 (compare Fig. 6FeG to Fig. 6H) and to
the crustal velocity model of the average continental margin type
(Fig. 6FeG) published by Mooney et al. (1998). Six LVBs were
identied along the northern margin and two more along the
southern margin (Fig. 4B). These bodies are not regularly spaced:
along PR1, the spacing reaches w90 km to the North and decreases
to a value of 45 km toward the oceanic crust. Along PR2, spacing is
about 70 km. Velocities higher than 6.5e6.7 km/s show the same
type of undulations although with far less amplitude.
Thin (2e3 km) high velocities lenses (7.0e7.7 km/s) are imaged
in some places within the lowermost crust above the Moho inter-
face (Figs. 4B and 6FeG). Because they are thin, we call them HVL
for High Velocity Lenses rather than HVB (High Velocity Bodies).
Along the northern margin (PR1), the largest of these lenses reaches
180 km between 130 and 310 km. A HVL is also present between
390 and 500 kmbut to a less extent (w110 km). Along the Southern
margins (PR2), a HVL is recognized between LVBs A and B over a
Figure 3. (continued).
T. Pichot et al. / Marine and Petroleum Geology xxx (2013) 1e17 9
Please cite this article in press as: Pichot, T., et al., Deep crustal structure of the conjugate margins of the SW South China Sea from wide-angle
refraction seismic data, Marine and Petroleum Geology (2013), http://dx.doi.org/10.1016/j.marpetgeo.2013.10.008
length of about 80 km (Figs. 4B). At both margins, HVL are also
present at the feet of the continental slope. Although relatively
limited on the northern side (w50 km in length, between 540 and
590 km), the southern HVL seems to be continuously present from
the necking zone to the COT (between 840 and 740 km; Fig. 4B).
The Moho interface is relatively at over the entire continental
margins (according to the vertical exaggeration; Fig. 4B). Along PR1,
the Moho gently rises southeastward with an average depth below
the sea level of w21 km (between 30 and 300 km) to w19 km (at
500 km). This change in depth of the Moho interface is marked by a
more pronounced rise around w400 km. This relative shallowing
coincides with the northeastern extent of the Phu Khan Basin (see
Figs. 1 and 4B). Along PR2, the Moho depth interface increases
slightly to the northwest from 15 km to 19.5 km bellow sea level.
The mean crustal thickness varies between 17.5 km and 9 km along
the northern margin (excluding the Quiongdongnan Basin) and
from 15 km to 8.5 km, along the southern margin (Fig. 7). The
thickest sediment thickness (16e17 km) is found within the
Quiongdongnan Basin, the crust below being reduced to w5 km.
Sharper changes occur at the COTs, the Moho rising from w19 km
to 10 km for both PR1 and PR2. The crust abruptly thins over a
distance of w100 km along PR1 (between 490 and 590 km) giving
an angle of 5
.
In summary, the nal velocity model shows strong lateral ve-
locity variations within the intermediate depth continental crust
where low velocity bodies are imaged with a spacing ranging from
90 to 45 km. In a few places, high velocity lenses are imaged within
the lowermost crust, including at the ContinenteOcean Transition.
The Moho interface remains rather at over the entire continental
margins (except toward the ContinenteOcean Transition). Despite
a number of small differences between the two margins (mean
Figure 4. Preferred model for the entire OBS prole (i.e. PR1 and PR2 are merged). The positions of each OBS are reported in red. A) Initial velocity model with a linear velocity
gradient ranging from 4 km/s at the seaoor (thin black line) to 7.5 km/s at the Moho interface (heavy white line). The seaoor is extracted from multibeam bathymetry data and the
Moho interface is derived from Bouguer anomaly inversion. B) Final velocity model after 12 iterations. The heavy white line corresponds to the Moho interface constrained by PmP
arrivals. The shadow zone represents areas without ray coverage. The velocity isocontours are drawn every 0.25 km/s. The black arrows indicate the position of the 1-D velocity-
depth models presented in gure 6. The main geological provinces are annotated. Low Velocity Bodies (LVBs) are indicated by small number along the northern margin and by
letters along the southern margin. (For interpretation of the references to color in this gure legend, the reader is referred to the web version of this article).
Table 1
Tomo2D set of parameters used in this study.
Parameters Symbols Values used in
this study
Vertical correlation length for
velocity nodes
L
Vv
a
CV
b
10e30
Horizontal correlation length for
velocity nodes
L
Hv
a
CV
b
0e24
Correlation length for reector L
d
a
CD
b
7
Velocity smoothing l
v
a
SV
b
10
Depth smoothing l
d
a
SD
b
10
Velocity damping D
v
a
TV
b
10
Depth damping D
d
a
TD
b
e
Depth kernel weighting w
a
W
b
0.2
a
Refers to Korenaga et al. (2000).
b
Refers to Korenaga (2003).
T. Pichot et al. / Marine and Petroleum Geology xxx (2013) 1e17 10
Please cite this article in press as: Pichot, T., et al., Deep crustal structure of the conjugate margins of the SW South China Sea from wide-angle
refraction seismic data, Marine and Petroleum Geology (2013), http://dx.doi.org/10.1016/j.marpetgeo.2013.10.008
crustal thickness, Moho slope at the COT, spacing between the low
velocity bodies, size of the sedimentary basins), the average 1-D
velocity-depth models for each continental margin (PR1 and PR2)
are similar (Fig. 6C) suggesting a common origin.
4. Discussion
In the following section, the term northern margin and southern
margin will be preferred rather than PR1 and PR2, respectively.
4.1. Sub-surface crustal structure (sediments and upper to middle
crust)
The 1.6e5.5 km/s velocity range is consistent with seismic ve-
locities of the CenozoicePaleozoic sedimentary sequence
(including compacted sediments) reported in previous refraction
studies constrained by seismic reection proles in the SCS area
(Table 3).
The crustal structure that we nd in the uppermost part of our
section can be compared with what has been documented in the
seismic reection proles nearby (Hayes et al., 1995; Schlter et al.,
1996; L et al., 2011; Ding et al., 2013). Seismic reection data
collected in the vicinity of both conjugate margins actually show
subsurface extensional features commonly found at other rifted
continental margins such as small-scale grabens and horsts. In the
SCS, the dominant spacing between these small-scale subsurface
features is between 15 and 30 km (Hayes et al., 1995; Ding et al.,
2013). These structures, distributed over a wide area, are well
imaged at the seaoor and in the free-air gravity, so that the area
has been commonly referred to as a Basin-and-Range type of
rifted basin (Gilder et al., 1991; Zhang et al., 2009; Franke, 2012).
These small-scale normal faults are often associated with a tilt of
the velocity isocontours, not only through the entire sedimentary
cover, but also within the upper crust (Figs. 4B and 8).
Lateral velocity variations are highest at intermediate depth (5e
15 km), in the 5.7 to 6.5e6.7 km/s velocity range (Fig. 4B, variations
highlighted by the 6 km/s velocity contour). Nearby seismic
reection proles show the presence of localized large-scale deep-
rooted faults cutting through the entire crust (called through-going
crustal faults in Hayes et al., 1995) and controlling major tilted
faults blocks (Hayes et al., 1995; Hutchison and Vijayan, 2010; Ding
et al., 2013). These structures are consistent with the identied Low
Velocity Bodies (LVBs) as well as the spacing between them on the
nal velocity model (45e90 km). Along the northern margin, these
large-scale normal faults dip with an angle between 25
and 40