ORIGINAL RESEARCH PAPER
Using overlapping sonobuoy data from the Ross Sea to construct
a 2D deep crustal velocity model
M. M. Selvans ? R. W. Clayton ? J. M. Stock ?
R. Granot
Received: 13 June 2011 / Accepted: 1 December 2011
 Springer Science+Business Media B.V. 2011
Abstract Sonobuoys provide an alternative to using long
streamers while conducting multi-channel seismic (MCS)
studies, in order to provide deeper velocity control. We
present analysis and modeling techniques for interpreting
the sonobuoy data and illustrate the method with ten over-
lapping sonobuoys collected in the Ross Sea, offshore from
Antarctica. We demonstrate the importance of using the
MCS data to correct for ocean currents and changes in ship
navigation, which is required before using standard methods
for obtaining a 1D velocity profile from each sonobuoy. We
verify our 1D velocity models using acoustic finite-differ-
ence (FD) modeling and by performing depth migration on
the data, and demonstrate the usefulness of FD modeling for
tying interval velocities to the shallow crust imaged using
MCS data. Finally, we show how overlapping sonobuoys
along an MCS line can be used to construct a 2D velocity
model of the crust. The velocity model reveals a thin crust
(5.5 ? 0.4 km) at the boundary between the Adare and
Northern Basins, and implies that the crustal structure of the
Northern Basin may be more similar to that of the oceanic
crust in the Adare Basin than to the stretched continental
crust further south in the Ross Sea.
Keywords Sonobuoy  Multi-channel seismic  Ross Sea 
Finite-difference  2D velocity model  Crustal structure
Introduction
Using sonobuoys to investigate deep crustal structure
Sonobuoys provide a means of obtaining long offsets for
deep velocity analysis while conducting multi-channel
seismic (MCS) studies. They are preferable to using a long
streamer in locations with difficult open water conditions,
such as in the Ross Sea and Southern Ocean, where
research cruises regularly encounter sea ice, icebergs, and
intense storms. Ocean-bottom seismometers (OBS) provide
another method for collecting active seismic data with
large offsets, but are more expensive than sonobuoys and
are challenging to recover in the conditions described
above. Sonobuoys are also an ideal method for collecting
large offset seismic data in other settings where ship nav-
igation is constrained, such as locations with ship traffic
and narrow bodies of water such as fjords.
Marine refraction seismology has long been recognized
as an essential technique for mapping the sedimentary rock
and basement structure of the Antarctic margins (e.g.,
Ewing and Heezen 1956; see also Anderson 1999 and
references therein). We present techniques for interpreting
deep crustal structure from overlapping sonobuoy data
collected in the northwestern Ross Sea. We first demon-
strate the necessity and method for correcting for the effect
of currents and changes in ship navigation, prior to using
standard processing techniques to obtain 1D velocity
models from the sonobuoys. Second, we explore the ben-
efits of using finite-difference (FD) method modeling of
sonobuoys to tie interval velocities to the shallow crustal
M. M. Selvans  R. W. Clayton  J. M. Stock
Seismological Laboratory, California Institute
of Technology, 1200 E. California Blvd., MC 252-21,
Pasadena, CA 91125, USA
M. M. Selvans (&)
Center for Earth and Planetary Studies, National Air
and Space Museum, Smithsonian Institution, MRC 315,
Washington, DC 20560, USA
e-mail: selvans@gps.caltech.edu
R. Granot
Department of Geological and Environmental Sciences,
Ben-Gurion University of the Negev, Beer Sheva, Israel
123
Mar Geophys Res
DOI 10.1007/s11001-011-9143-z
structure imaged with MCS data. Finally, we apply these
methods to a set of ten overlapping sonobuoys in order to
construct a 2D velocity model of the deeper crustal struc-
ture along an MCS line.
Modeling methods for seismic refraction data
The FD method is useful for modeling propagation of
seismic waves, including turning waves and their reflec-
tions and conversions at interfaces, for complex subsurface
structure. Ray tracing is similarly useful, but has difficulty
predicting the amplitudes of refracted energy, with the
consequence that it is often difficult to judge the impor-
tance of some of the predicted arrivals. One advantage of
using the FD method to model seismic refraction data is the
ability to model the entire wave front from source to
receiver, which allows us to both tie the shallow structure
detected with MCS data to particular velocity horizons and
to determine the deeper crustal velocity structure.
Shipp and Singh (2002) used a 2D elastic finite-differ-
ence forward model and an iterative full wavefield inver-
sion scheme to best fit their streamer data (with a 12 km
Fig. 1 a Sonobuoy processing methods are developed for data
collected in the northwestern Ross Sea (black box), offshore from
Antarctica. b We interpret a subset of the active seismic data collected
in the northwestern Ross Sea during research cruise NBP0701, shown
on top of multibeam bathymetry (NBP0701 Data Report 2007). Ten
sonobuoys are closely spaced along multi-channel seismic (MCS)
Line 14, which runs from deep water in the Adare Basin onto the
continental margin in the Northern Basin
Mar Geophys Res
123
offset), solving for velocity structure down to 4 km below
the seafloor. This approach is computationally demanding,
with residuals from the comparison of data and the model
being back propagated for every time step, which is why
they were constrained to analysis of relatively shallow
crust. Jones et al. (2007) performed a similar analysis on
streamer data (with offsets of 15 and 18 km); they addi-
tionally derived 1D velocity models from the intercept-
time-slowness (s-p) domain, and downward continued the
data in order to create an image of the layers at depth. They
were able to image the velocity structure to a depth of
6 km.
In order to conduct a marine seismic refraction
experiment with an offset larger than tens of kilometers,
Ritzmann et al. (2004) used OBS instruments. They used
a ray tracing model to reproduce the data, and found that
crustal thickness varied widely, from 32 km for the con-
tinental crust of Svalbard to as little as 2 km (excluding
the *2 km of sediment on top of the basement) near the
margin between continental and oceanic crust. Mantle
velocities are generally [8.0 km/s, but were found to be
as low as 7.7 km/s west of Molloy transform fault, where
the mantle is likely serpentinized (Ritzmann et al. 2004).
Also offshore from Svalbard, Geissler and Jokat (2004)
Fig. 2 Raw data for Sonobuoys
1 (a) and 4 (b) on Line 14 show
key features used to determine
the 1D velocity profile at these
locations. Time is equivalent to
two-way travel time, and
distance is relative to the airgun
source. The direct arrival comes
in first in time for smaller
offsets, while energy from head
waves traveling along layer
interfaces at depth comes in first
for larger offsets; the slopes
of these linear features are
determined by the interval
velocities in the crust. The
hyperbolic reflection from the
seafloor comes in second for
smaller offsets and provides
a time constraint used to
determine the water layer
thickness. Later hyperbolic
features are multiples of the
seafloor reflection. A bandpass
filter of 5, 15, 35, and 40 Hz
is applied to the data
Mar Geophys Res
123
used sonobuoy data as 1D velocity profile ??pseudo-
boreholes?? along an MCS line to obtain deeper crustal
structure.
Collecting sparse sonobuoy data along MCS lines and
modeling the resulting profiles with ray tracing methods
has been used to investigate deep crustal structure along
Antarctic continental margins as well, for example offshore
from Wilkes Land (Close et al. 2009) and Enderby and
Mac. Robertson Lands (Stagg et al. 2004). In this study, we
combine the analysis of sonobuoy data in the s-p domain
with FD modeling of the waveform to derive a sequence of
overlapping 1D velocity models, which are then interpreted
into a 2D deep crustal velocity model.
Seismic data in the Ross Sea
We will illustrate the analysis procedure for sonobuoy data
collected in the Adare Trough region of the Ross Sea,
offshore from Antarctica (Fig. 1a). The Adare Trough, a
dead mid-ocean spreading ridge, lies in the deep water of
the Adare Basin and trends southward toward the Northern
Basin, which is located up on the continental shelf
(Fig. 1b). How seafloor spreading was linked to extension
in the West Antarctic Rift System to the south is of major
interest, yet still poorly resolved. Shallow structure is well
imaged by MCS data (Granot et al. 2010). However,
understanding the tectonics in the area requires knowledge
Fig. 3 We correct the
sonobuoy shot spacing to
account for ocean currents and
changes in ship navigation. A
best possible linear moveout
(l mo) of 1570 m/s is applied to
the direct wave in the raw data
for Sonobuoy 1 (a), and
displayed in terms of the
resulting reduced travel time.
The distances between shots are
corrected (b), based on picks
made along the direct wave,
producing a linear direct wave
with the correct water velocity
of 1450 m/s (as determined
from multi-channel seismic
data). Sonobuoy 4 has a linear
moveout of 1520 m/s before the
correction (c), and 1450 m/s
after (d). See Table 1 for direct
wave velocities prior
to the correction
Mar Geophys Res
123
of the deep velocity structure, for example the Moho depth
along the continental shelf break (see Selvans et al. (sub-
mitted) for an analysis of all sonobuoy data, and implica-
tions for tectonics).
We collected seismic reflection and refraction data
during research cruise NBP0701 on board the R/VIB
Nathaniel B. Palmer. Sonobuoy data were collected in the
Adare and Northern Basins, with maximum offsets from
the ship of 20?30 km. Sonobuoys presented here were
deployed with a regular spacing of *15 km (Fig. 1b inset)
in order to obtain overlapping records, since data was
generally returned from as far as 20?30 km from the ship.
Most sonobuoy studies in the western Ross Sea have
focused on basins south of the Northern Basin, and have
used linear moveout and ray tracing methods to determine
sediment velocity gradients and basement depth (Houtz
and Davey 1973; Cooper et al. 1987; Cochrane et al. 1995).
Ocean bottom seismic refraction studies have investigated
the deeper crustal structure in the central and southern Ross
Sea, using ray tracing and amplitude modeling methods
(Tre?hu et al. 1993; Trey et al. 1999). Crustal thickness,
defined as the depth to velocities of [8000 m/s, is as little
as 16 km beneath sedimentary basins underlain by thinned
crust and magmatic intrusions, and 21 km beneath inter-
vening basement highs (Trey et al. 1999).
MCS data from NBP0701 contain resolvable primaries
from up to 2.3 s travel time (i.e., ??two-way travel time??)
below the seafloor in the Adare and Northern Basins,
revealing variable sediment thickness (Granot et al. 2010).
Migrated MCS data delineate the structure and deformation
of the sedimentary units in the Adare and Northern Basin,
but cannot tie velocity values to the subsurface layers, and
cannot detect crustal structure below the basement rock.
Methods for sonobuoy analysis and interpretation
Trace spacing adjustment and construction of 1D
velocity profiles
MCS data were collected using a 1.2 km, 48-channel
streamer; this short streamer was required due to regular
interactions with sea ice. Data presented here were
obtained using a 6-element Bolt-gun array with a total
capacity of 34.8 liters, with a typical source spacing of
approximately 40 m. We present data analysis methods,
and results, for ten overlapping sonobuoy profiles along
MCS Line 14, which ran from the deep water of the Adare
Basin south onto the shallow-water shelf of the Northern
Basin (see Fig. 1b).
These sonobuoys were deployed approximately every
15 km, and returned data for 20?30 km of offset. While the
source moved away from the sonobuoy launch position at a
nearly uniform speed and direction, sonobuoy drift after
launch (due to ocean currents) and slight variations in ship
navigation led to varying shot spacing in the refraction data
set. We correct for these effects before analysis, modeling,
and interpretation of the data. The sonobuoy data provide
details of the crustal structure to greater depth than the
MCS data along the same line, and directly detect interval
velocities; MCS data are useful in determining the sound
speed (seismic velocity) of the water layer and the shal-
lowest of the rock layers. Consistent with the generally flat
horizons observed in the MCS data, we assume plane
layers for all velocity horizons.
We show the processing steps needed for accurate
interpretation of the sonobuoy data, starting from the
raw data for the deep-water Sonobuoy 1 (Fig. 2a) and
the shallower-water Sonobuoy 4 (Fig. 2b). We highlight
the low-frequency reflected and refracted energy from the
seafloor and within the crust by applying a tapered band-
pass filter to the data, set to 5, 15, 35, and 40 Hz (i.e., data
with frequencies between 15 and 35 Hz are passed through
with their amplitudes unaltered, and data in the ranges
5?15 and 35?45 Hz are passed through with amplitudes
that decrease to zero at the edges of the filter). Several key
features of the data are obvious in these sonobuoy images:
(1) the direct wave, from source to receiver, is the first
arrival at smaller offsets (e.g., 0?8500 m in Fig. 2a), (2) the
reflection from the seafloor comes in next for smaller off-
sets (e.g., 0?6500 m in Fig. 2a), and due to its strength as
well as noise in the data no reflections from deeper layers
are observed at these offsets (subsequent reflections are
seafloor multiples), (3) head waves, refracted energy from
layer interfaces at depth, come in as first arrivals for larger
Table 1 Direct wave velocities, as determined using a best possible
linear moveout fit to the feature, are listed for Sonobuoys 1?10
(S1?S10) on Line 14
Direct wave velocity
before correction (m/s)
S1 1570
S2 1570
S3 1480
S4 1520
S5 1330
S6 1050
S7 1180
S8 1200
S9 1500
S10 1580
The slope of the direct wave is corrected by adjusting the distance
between traces, such that the sonobuoy data has the correct water
velocity value of 1450 m/s
Mar Geophys Res
123
offsets (e.g., 8,500?19,000 m in Fig. 2a). The offset range
of each arrival type varies with water depth and subsurface
layer thicknesses (see Fig. 2b).
We correct sonobuoy shot spacing so that the direct
wave has a slope of 1450 m/s, as determined by applying a
linear moveout to the direct wave in the MCS data (accu-
rate to ?50 m/s). This value for the seismic velocity of the
water layer is consistent along the MCS line, and is
therefore taken as the default value for all sonobuoys; it is
also consistent with measurements of sound speed in water
taken during the cruise. Data from the one expendable
sound velocimeter (XSV) show water velocities within
1441?1451 m/s (to 600 m depth), while expendable
bathythermographs (XBTs, which measure temperature
with depth and use surface measurements of salinity to
compute sound velocity) show water velocities of
1444?1480 m/s at the northern end of MCS Line 14, and
1446?1448 m/s at the southern end of Line 14. These
sensor measurements of sound speed in water are accurate
to ?0.25 m/s.
In order to correct the shot spacing, we adjust the dis-
tance coordinate of each trace, such that the direct wave is
linear and has the correct slope in the distance versus time
plot. The direct wave of Sonobuoy 1 is shown with a linear
moveout applied before (Fig. 3a, with a direct wave
velocity of 1570 m/s) and after (Fig. 3b, with a direct wave
Fig. 4 Linear moveout (l mo)
is applied to the refracted
energy in the distance-adjusted
version of Sonobuoy 1, to
directly measure interval
velocities and associated times
(s). We detect layers with
velocities of 3900 m/s (a),
4400 m/s (b), 5600 m/s (c), and
8000 m/s (d), with head waves
lying within the arrows; the
8000 m/s layer is also shown as
an inset. The latter value is the
only instance of a velocity that
can be interpreted as the Moho.
Refracted energy from the
seafloor is not observed (an
interval velocity of 2200 m/s for
the uppermost sediment layer is
determined from multi-channel
seismic data)
Mar Geophys Res
123
velocity of 1450 m/s) this correction is made; the adjust-
ment for Sonobuoy 4 makes a similar direct wave velocity
adjustment (Fig. 3c, d). Other sonobuoys have larger dis-
crepancies between the direct wave velocities before and
after the correction (see Table 1). Sonobuoy 3 requires the
smallest correction based on water velocity (average ori-
ginal shot spacing of 38.4 ? 1.6 m, and average corrected
shot spacing of 37.5 ? 2.1 m), while Sonobuoy 6 requires
the largest (similarly, 32.1 ? 2.6 m, and when corrected
43.7 ? 3.5 m). This correction will be more straightfor-
ward for sonobuoys with reliable GPS; in that case, the
sonobuoy location will be at least as well known as that of
the MCS streamer.
We directly detect layer velocities, and their associated
times, by applying linear moveout to head waves in the
corrected sonobuoy data. We detect between three and six
distinct layers at depth for each sonobuoy with this method,
as shown in detail for Sonobuoys 1 and 4 (Figs. 4 and 5,
respectively). This same approach yielded a consistent
velocity of 2000 m/s for the shallowest rock layer in seven
out of the ten sonobuoys; in other cases this seafloor head
wave is not visible. Even when it is visible, it is usually
brief (*10 traces) and so less distinct than most other
refractors we measure. For this reason, we verify this layer
velocity by applying a normal moveout to the hyperbolic
arrival of the seafloor reflection in the raw MCS data, and
Fig. 5 Linear moveout (l mo)
is applied to the refracted
energy in the stretch-corrected
version of Sonobuoy 4, directly
measuring interval velocities
and their associated times. We
detect layers with velocities of
2300 m/s (a), 3300 m/s (b),
4400 m/s (c), and 4800 m/s (d),
with head waves lying within
the arrows. Refracted energy
from the seafloor is not
observed (an interval velocity of
2000 m/s for the uppermost
sediment layer is determined
from multi-channel seismic
data)
Mar Geophys Res
123
use it as the first rock layer velocity for all ten sonobuoys;
one exception is Sonobuoy 1, where analysis of MCS data
provides a first rock layer velocity of 2200 m/s. Layer
velocities for all sonobuoys are listed in Table 2.
The velocities and their associated times are used to
calculate the 1D velocity profile for each sonobuoy (e.g.,
Fowler 1990, p. 119?123). Firstly, the water depth and first
rock layer thickness are calculated using velocities (v) and
reflection times (t) from the MCS data (i.e. h1 = 0.5(t1)(v1)).
Secondly, the thicknesses of deeper layers (i = 2, 3, etc. are
rock layers) are determined using standard seismic refrac-
tion analysis:
hi ?
vi
2 cos ?hi;i?1?
si1 
Xi1
j?1
2hj
vj
cos ?hj;i?1?
$ %
where ha,b = arcsin(va/vb) and s is the reduced travel time
associated with the layer velocity, obtained when linear
moveout is applied to the sonobuoy data. Additionally, a
direct image of the s-p curve can be obtained by a radon
transform of the data (McMechan and Ottolini 1980); an
example is shown in Fig. 6.
Layer thicknesses are summed to determine layer
depths, resulting in a 1D velocity profile for the sonobuoy.
Velocity profiles for Sonobuoys 1 and 4 are plotted with
respect to depth below the seafloor for ease of comparison
(Fig. 7), revealing the unusually high-velocity detail in
Sonobuoy 1, with a maximum velocity of 8000 ? 100 m/s
at 5.5 ? 0.4 km crustal depth. Table 2 lists the details of
the velocity models for all ten sonobuoys.
Finite-difference method modeling and depth migration
We verify our 1D velocity models via finite-difference
(FD) method modeling of the sonobuoys, and by imaging
the subsurface through depth migration of the data. Our FD
model solves the acoustic wave equation in an elastic
medium (Vireaux 1986), using the velocity models calcu-
lated above to reproduce the sonobuoy data (Fig. 8). The
model solutions are 2nd order in time and 8th order in
space; the resulting lack of dispersion of the wave front
through the medium allows for accurate timing. The high
spatial accuracy is required because of the wide range of
velocities in the model (1500?8000 m/s). We assume layer
densities of 2600 kg/m3 for velocities from 2000 to
4000 m/s; 2700 kg/m3 for velocities from 4000 to 6000 m/s;
and 2800 kg/m3 for velocities of 6000 m/s and greater;
Table 2 Velocity model values for Sonobuoys 1?10 (S1?S10) on Line 14 (also see Fig. 10b), and sonobuoy distance along the line
1 (H2O) 2 3 4 5 6 7 8 Distance along the line (km)
S1: d
v
0
1450
1960
2200
2950
3900
4090
4400
5850
5600
7500
8000
0
S2: d
v
0
1450
1780
2000
2320
2200
2520
2800
2910
3700
3660
4400
5000
5000
13
S3: d
v
0
1450
1510
2000
2130
2900
2620
3500
3060
4200
3680
4800
5130
5600
27
S4: d
v
0
1450
1160
2000
1460
2300
1800
3300
2600
4400
2990
4800
38
S5: d
v
0
1450
520
2000
1200
2800
1680
3300
2440
4100
2930
4400
3690
4700
52
S6: d
v
0
1450
500
2000
1010
2500
1270
2800
1800
3500
2500
4100
3040
4500
66
S7: d
v
0
1450
470
2000
1320
3000
2320
4300
2820
4500
78
S8: d
v
0
1450
440
2000
1030
2500
1390
3000
2420
4300
3040
4700
3560
4900
88
S9: d
v
0
1450
470
2000
1020
2400
1250
2900
2020
4000
2700
5000
107
S10: d
v
0
1450
460
2000
1030
2500
1230
2900
1790
3500
2260
4300
2660
4500
3350
4900
114
Sonobuoy 1 lies in the deep water of the Adare Basin, while Sonobuoys 5?10 lie in the shallow water of the Northern Basin. Columns contain the
depth (d, in m) and velocity (v, in m/s) for each layer, as derived from the sonobuoy data. Note that the first rock layer velocity for Sonobuoy 1 is
2200 m/s, while the rest have a velocity of 2000 m/s for this layer; this value is obtained by performing a normal moveout on multi-channel
seismic data at the sonobuoy location. Velocities are resolved to ?100 m/s, while depths below the seafloor have resolutions of ?0.1?0.4 km,
increasing with depth
Mar Geophys Res
123
these densities are consistent with sedimentary, upper crust
(e.g., basement rock), and lower crustal rock respectively
(Jones 1999, p. 64).
By reproducing the sonobuoy data without noise, we
obtain reflection times associated with each head wave.
These allow us to tie the shallow structure observed in the
stacked and migrated MCS data to particular velocity
horizons (Fig. 9). We are therefore able to interpret the
shallow structure revealed by MCS data in terms of directly
observed interval velocities. We additionally extend the
depth to which the crustal velocity structure is delineated,
below the acoustic basement that limits MCS penetration.
We also verify our 1D velocity models by depth
migrating the sonobuoy data (Fig. 10). First a radon
transform is applied to the distance-adjusted sonobuoy
data (e.g., Jones 1999, p. 74?75), which maps the data
into slowness (1/v)?intercept time (s) space (Fig. 6). The
transformed data is then migrated in depth, using the
finite-difference version of the wave-equation method
(Clayton and McMechan 1981), based on the 1D velocity
model.
The subsurface image is in terms of velocity versus
depth, and for Sonobuoy 1 (Fig. 10a) shows flat shallow
layers below the water layer, as observed in the stacked and
Fig. 6 Applying a radon
transform to the model (a) and
data (b) for Sonobuoy 1 reveals
the ellipses that image each
velocity layer in s-p space. The
cusp at the beginning of each
ellipse is defined in terms of the
slowness (p) of the layer (or 1/v,
where v is the velocity), where
layers 1?6 are labeled (1450,
2200, 3900, 4400, 5600, and
8000 m/s respectively). Note
that cusps of layers 1, 4, and 5
are relatively low amplitude,
and that layer 6 is hidden by the
multiple of layer 1
Mar Geophys Res
123
migrated MCS data (Fig. 9a); in general, flat interfaces in
depth-migrated sonobuoy images indicate use of an accu-
rate velocity profile. Flat shallow interfaces in the migra-
tion data for Sonobuoy 4 (Fig. 10b) validate its respective
velocity model. Slight tilting in the deeper layers indicates
small lateral variations in velocity that cannot be properly
modeled with 1D analysis.
Constructing a 2D velocity model
We interpolate velocity values between the 1D velocity
models from overlapping sonobuoys, using them as
??pseudo-borehole?? data, to construct a 2D velocity model
along Line 14 (Fig. 11). Since the sonobuoys record a
greater offset than their spacing and are reveal similar 1D
velocity profiles, we interpolate between sonobuoys to
construct a 2D velocity profile across the continental shelf
break. We use a Delaunay triangulation algorithm to
perform the interpolation, the vertices being location
(depth below seafloor and distance along the line) and
velocity. The 1D models include maximum velocities of
4500?8000 m/s, to maximum depths of 2.7?7.5 km below
sea level (i.e., 1.8?5.5 km below seafloor; see Table 2
and Fig. 11b, respectively). Velocities are resolved to
?100 m/s, while depths below the seafloor have resolu-
tions of ?0.1?0.4 km, increasing with depth.
Results
By interpolating between 1D velocity profiles for ten
overlapping sonobuoys, we obtain a 2D velocity model
along 115 km of a seismic reflection line that crosses the
continental shelf break, with a maximum depth of 5.5 km
below the seafloor under Sonobuoy 1 (Fig. 11). We
determine the crustal structure to a greater depth than is
possible with the MCS data alone, and provide interval
velocities for interpretation of the subsurface. We obtain
accurate velocities by correcting the sonobuoy data before
linear moveout analysis, since changing currents and ship
navigation can alter the apparent velocities of head waves
coming in from interfaces at depth. Other sonobuoy studies
in the Ross Sea have detected sound speeds in water of
1430?1540 m/s, close to the true value (compare to our
spread of 1050?1580 m/s, see Table 1), and so have not
needed to make this distance adjustment before performing
data analysis (e.g., Houtz and Davey 1973; Cooper et al.
1987; Cochrane et al. 1995).
We are able to tie our 1D velocity models to layers
imaged in the shallow crust using MCS because we use
an FD model to reproduce the obvious features in the
sonobuoy data (Fig. 8), which allows us to determine the
reflection time associated with each head wave. These
reflection times are directly compared to MCS images of
the subsurface, to place accurate constraints on layer
Fig. 7 Velocity models are
constructed for Sonobuoys 1
(a) and 4 (b) using the velocities
and associated times obtained
through standard linear moveout
and normal moveout methods
(see text for detail). The
velocity of 8000 m/s suggests
that the Moho is only 5.5 km
below the seafloor beneath
Sonobuoy 1. See Table 2 for
velocity model details
Mar Geophys Res
123
velocities. This provides a method for determining sedi-
ment thickness, whose structure is otherwise interpreted
(from MCS data) entirely in time space (Granot et al.
2010).
Due to the experimental design and processing methods
we employ, we are able to accurately determine crustal
structure in the Adare and Northern Basins to a greater
depth than is possible with the MCS data. Overlapping
sonobuoys allow us to construct a 2D velocity profile of the
crust more cheaply than is possible with ocean-bottom
seismometers. Since the NBP0701 cruise deployed simi-
larly spaced sonobuoys along other MCS lines in the
southern Adare Basin and northern part of the Northern
Basin, these methods can be used to construct a pseudo-3D
interpretation of the crustal structure in locations where
MCS lines cross (see Selvans et al. (submitted)).
Discussion and conclusions
Obtaining sonobuoy FD models that accurately reproduce
the main features being analyzed allows for further inter-
pretation of MCS data than is otherwise possible. MCS
data image shallow structure in great detail, but cannot
directly measure layer velocities at depth. Matching the
reflection times from each layer of known velocity in the
FD model to the MCS image provides the velocity model
details needed for depth migration of MCS data.
Fig. 8 Using the velocity
model derived from the
distance-adjusted data for
Sonobuoy 1 (a), we reproduce
the key features in the data
using a finite difference model
(b). The finite difference model
allows head wave features to be
tied to their associated
reflections, which come in after
the seafloor reflection and are
masked in the data. The
corrected data for Sonobuoy 4
(c) and resulting finite
difference model (d) highlight
the usefulness of modeling in
shallow-water locations for the
identification of multiples
Mar Geophys Res
123
Modeling sonobuoys using FD has further potential for
sonobuoy data analysis, since refinements in the model can
reveal more details about the crustal structure. For instance,
models could be altered to include head waves that occa-
sionally come in late as a result of interbed multiples,
further refining the 1D velocity model at that sonobuoy
location. An elastic version of the model could be used to
identify the converted phases in the data; an elastic FD
model was run for Sonobuoys 1 and 4 to confirm that none
of the head waves used in this analysis are converted
phases.
Using an elastic FD model would also constrain the
allowable range of shear wave velocities, based on the
amplitude of those phases. Elastic FD models with both
vs = (1/H3)vp and a vs = (1/2H3)vp have converted phases
with amplitudes similar to those of the main head wave
features of the sonobuoy; these phases are not observed in
the data, indicating that shear wave velocities are small.
Modeling the full range of possible shear wave velocities
would help in identifying the crustal lithology, below the
layers that can be mapped in MCS data over long distances
and then tied to borehole results (e.g., Granot et al. 2010).
Modeling of overlapping, and particularly reversed
sonobuoys could additionally include lateral variations in
layer velocity and layer thickness and a smoothly varying
profile with depth, in order to determine whether these
secondary effects are observed in the sonobuoy data.
We show the necessity of adjusting distance between
traces for sonobuoy studies, since ocean currents and
changing speed and direction of the ship can alter the direct
wave slope such that it does not accurately reflect the sound
speed in water of 1450 m/s (see Table 1). Prior to the
correction, direct wave velocities were as much as 130 m/s
higher (Sonobuoy 10) and 400 m/s lower (Sonobuoy 6)
than the true sound speed in water. Directly measuring
layer velocities at depth by applying a linear moveout to
head waves in the uncorrected sonobuoy data will be off
from the true velocities of those layers by similar amounts,
resulting in errors larger than the measurement uncertainty
of ?100 m/s in most cases (seven out of ten sonobuoys).
In order to use the reflected energy features in the
sonobuoy analysis, this error cannot be corrected through a
simple stretch in the time dimension, based on the best fit
direct wave velocity, since a hyperbolic feature will not
retain its proper shape if a linear stretch is applied. Gen-
erally, in cases where direct wave velocities differed from
the true sound speed in water by more than the uncertainty
of the velocity measurement, sonobuoy studies have
reported inaccurate layer velocities.
Finally, our analysis is greatly aided by the experiment
design. Collecting overlapping sonobuoy data along sev-
eral MCS lines during research cruise NBP0701 allows us
to construct 2D velocity models deep within the crust, to a
greatest depth of 5.5 km below the seafloor. This depth is
Fig. 9 Multi-channel seismic (MCS) data from the same location as
Sonobuoy 1 (a) can be compared to the finite difference model (b) in
order to determine the layer velocities within the reflection data. In
the model, reflections from which head waves originate are labeled
with large arrows, while multiples are labeled with small arrows. Note
that the first layer at depth (3900 m/s) corresponds to sediment layers
imaged in the MCS data, the 4400 m/s seems to be the top of the
basement rock, 5600 m/s is within the basement rock, and the
8000 m/s layer is deep in the basement rock (below the seafloor
multiple). Sonobuoy data also determine layer velocities below the
acoustic basement, which is the limit to which reflection data can
delineate subsurface structure
Mar Geophys Res
123
equivalent to 5.0 s travel time below the seafloor, deeper
than the maximum depth of 2.3 s below the seafloor at
which basement rock is imaged with MCS data (Granot
et al. 2010). We regularly determine layer velocities of
4500?5000 m/s, at 2.0?3.0 km below the seafloor, and are
able to see that the velocity contours fit to these data are at
consistent depths along the 115 km of MCS Line 14. Being
able to use the depth penetration and ability to directly
measure layer velocities that sonobuoys provide in order to
construct a 2D velocity model is a particularly useful
analytical technique for locations such as the Ross Sea,
where sea ice conditions preclude the use of a longer MCS
streamer.
Having a deep 2D velocity profile along Line 14 of the
NBP0701 MCS data allows us to begin examining the
crustal structure along the margin between the Adare and
Northern Basins (Fig. 11a). Two distinct features stand out.
One is the unusually thin crust under Sonobuoy 1: the data
Fig. 10 Depth migration is applied to the distance-adjusted data for
Sonobuoy 1 (a) and Sonobuoy 4 (b), using their respective 1D
velocity models. Parallel bedding, as observed in the multi-channel
seismic data (e.g., Fig. 8a), indicates a good model fit to the data;
slight tilting in deeper layers indicates small lateral variations in
velocity that cannot be properly modeled with 1D analysis
Mar Geophys Res
123
reveal a maximum velocity of 8000 m/s at 5.5 km below
the seafloor, which is interpreted as the Moho. Since there
is 2.1 km of overlying sediment at this location (as indi-
cated by comparing reflection time of the 4400 m/s layer in
the FD model with that of acoustic basement as imaged
with MCS data (see Fig. 9), and referring to the 1D
velocity model of Sonobuoy 1 for the depth of that layer),
the igneous crustal thickness is 3.4 km.
Although unusual, similarly thin crustal thicknesses are
observed in the Ross Sea (4?27 km; see Anderson (1999)
and references therein), and in continental margins of East
Antarctica (\4 km in continental crust offshore from
Fig. 11 The 2D velocity
structure along Line 14 is
determined to a depth of 5.5 km
below the seafloor. Seafloor
bathymetry (a) provides context
for the 2D velocity model
(b) interpolated from velocity
models for ten overlapping
sonobuoys (c). Layer interfaces
observed in the data are
indicated with an ??x?? (see
Table 2 for details); vertical
exaggeration is 1:8. Sediment
layers (*2000?4000 m/s)
thicken slightly from the Adare
Basin (Sonobuoy 1) to the
Northern Basin (Sonobuoys
5?10), as expected for moving
up a shelf break onto a
sediment-filled basin.
Interestingly, deeper velocity
contours remain approximately
flat along the line, whereas if the
shelf break were the transition
from oceanic to continental
crust, we would expect these
contours to deflect down under
the Northern Basin.
Uncertainties on 2D velocity
contours are ?200 m/s and
?100?500 m (increasing
with depth)
Mar Geophys Res
123
Wilkes Land (Close et al. 2009), and 4 km in oceanic crust
offshore from Enderby Land (Stagg et al. 2004)). Results
from processing all NBP0701 sonobuoys, including the
interpolation of additional 2D velocity models, are dis-
cussed in the context of other tectonic settings similar to
that of the Adare and Northern Basin region by Selvans
et al. (submitted).
The other obvious feature is that all well-constrained
velocity contours (2000?5400 m/s) are approximately flat,
considering the 8:1 vertical exaggeration (Fig. 11), indi-
cating that it is unlikely for there to be significant local
relief on layer interfaces within the crust. Velocity contours
at a transition between oceanic and continental crust would
be expected to deflect downward under the thicker conti-
nental crust, suggesting that the continental shelf break
between the Adare and Northern Basins (obvious in
bathymetry) may not be the location of the transition
between crustal types. As also suggested by Cande and
Stock (2006), who discussed the possibility of the Northern
Basin crust being ?transitional,? based on analysis of
magnetic anomaly data, we find that the crustal structure
may be continuous across the shelf break.
Acknowledgments We would like to thank Captain Mike Watson,
the crew, and the Raytheon Polar Services Corporation technical staff
on board the Nathaniel B. Palmer. This study was supported by
National Science Foundation grants OPP04-40959 (S. Cande) and
OPP-0440923 and OPP-0944711 (J. Stock and R. Clayton).
References
Anderson JB (1999) Antarctic marine geology. Camb University
Press, Cambridge
Cande SC, Stock JM (2006) Constraints on the timing of extension in
the Northern Basin, Ross Sea. 9th International symposium on
Antarctic earth science proceeding
Clayton R, McMechan G (1981) Inversion of refraction data by wave
field continuation. Geophysics 46:860?868
Close DI, Watts A, Stagg H (2009) A marine geophysical study of the
Wilkes Land rifted continental margin, Antarctica. Geophysical
J Int 177(2):430?450
Cochrane GR, De Santis L, Cooper AK (1995) Seismic velocity
expression of glacial sedimentary rocks beneath the Ross Sea
from sonobuoy seismic-refraction data, geology and seismic
stratigraphy of the Antarctic margin. Antarctic Res Ser 68:
261?270
Cooper A, Davey F, Cochrane G (1987) Structure of extensionally
rifted crust beneath the western Ross Sea and Iselin Bank,
Antarctica, from sonobuoy seismic data. The Antarctic conti-
nental margin: geology and geophysics of the Western Ross Sea.
Published by the Circum-Pacific Council of Energy and Mineral
Resources, Earth Science Ser 5B:93?117
Ewing M, Heezen BC (1956) Some problems of Antarctic submarine
geology. In: Carey AP, Gould LM, Hulbert EO, Odishaw H,
Smith WE (eds) Antarctica in the IGY. Geophys Monogr 1,
pp 75?81
Fowler CMR (1990) The solid earth: an introduction to global
geophysics. Camb University Press, CambridgeFig. 11 continued
Mar Geophys Res
123
Geissler WH, Jokat W (2004) A geophysical study of the northern
Svalbard continental margin. Geophysical J Int 158(1):50?66
Granot R, Cande S, Stock J, Davey F, Clayton R (2010) Postspreading
rifting in the Adare Basin, Antarctica: regional tectonic conse-
quences. Geochem Geophys Geosyst 11(8):Q08005. doi:10.1029/
2010GC003105
Houtz R, Davey F (1973) Seismic profiler and sonobuoy measure-
ments in Ross Sea, Antarctica. J Geophys Res 78(17):3448?3468
Jones EJW (1999) Marine geophysics. Wiley, London
Jones GD, Barton PJ, Singh SC (2007) Velocity images from stacking
depth-slowness seismic wavefields. Geophys J Int 168(2):583?
592
McMechan GA, Ottolini R (1980) Direct observation of a p-s curve in
a slant stacked wave field. Bull Seismol Soc Am 70(3):775
NBP0701 Data Report (2007) prepared by Ayers J. Available at http://
www.marine-geo.org/tools/search/data/field/NBPalmer/NBP0701/
docs/NBP0701Report.htm
Ritzmann O, Jokat W, Czuba W, Guterch A, Mjelde R, Nishimura Y
(2004) A deep seismic transect from Hovga?rd Ridge to north-
western Svalbard across the continental ocean transition: a
sheared margin study. Geophys J Int 157(2):683?702
Selvans MM, Stock JM, Clayton RW, Cande S, Davey F, Granot R
(submitted Sept. 30, 2011) Deep crustal structure of the Adare
and Northern Basins, Ross Sea, Antarctica, from sonobuoy data.
J Geophys Res
Shipp RM, Singh SC (2002) Two dimensional full wavefield inversion
of wide aperture marine seismic streamer data. Geophys J Int
151(2):325?344
Stagg HMJ, Colwel J, Direen N, O??Brien P, Bernardel G, Borissova I,
Brown B, Ishirara T (2004) Geology of the continental margin of
Enderby and Mac. Robertson Lands, East Antarctica: insights
from a regional data set. Marine Geophys Res 25(3):183?219
Tre?hu A, Behrendt JC, Fritsch J (1993) Generalized crustal structure
of the Central basin, Ross Sea, Antarctica. In: Damaske D,
Fritsch J (eds) German Antarctic North Victoria Land Expedition
1988/1989. Bundesanstalt fur Geowissenschaften und Rohstoffe,
Hannover, pp 291?311
Trey H, Cooper AK, Pellis G, della Vedova B, Cochrane G,
Brancolini G, Makris J (1999) Transect across the West
Antarctic rift system in the Ross Sea, Antarctica. Tectonophysics
301(1?2):61?74
Vireaux J (1986) P-SV wave propagation in heterogeneous media:
velocity-stress finite difference method. Geophysics 51:889?901
Mar Geophys Res
123