Content uploaded by Spahr Webb
Author content
All content in this area was uploaded by Spahr Webb on May 27, 2014
Content may be subject to copyright.
Fine-scale seismic structure of the shallow volcanic crust
on the East Pacific Rise at 9°50
0
N
Robert A. Sohn
Woods Hole Oceanographic Institution, Woods Hole, Massachusetts, USA
Spahr C. Webb
Lamont-Doherty Earth Observatory, Palisades, New York, USA
John A. Hildebrand
Scripps Institution of Oceanography, University of California, San Diego, La Jolla, California, USA
Received 23 April 2004; revised 7 September 2004; accepted 20 September 2004; published 9 December 2004.
[1]We use a combination of body wave and interface wave observations from an
on-bottom seismic refraction survey to constrain the fine-scale seismic structure of the
upper crust in a 33 km field area centered on the East Pacific Rise at 950
0
N. We
detonated 18 explosive shots (18 sources) in a circular pattern (1.5 km radius) on the
rise axis and recorded seismic arrivals with eight ocean bottom seismometers (eight
receivers). We observed 30– 40 Hz compressional body waves from all shots (144 P
waves) and 1–3 Hz Stoneley (interface) waves on a subset of source-receiver pairs
(58 interface waves). Using a station correction inversion, we find that roughly half of the
variance in the Pwave first-arrival times results from lateral variations in the thickness of
the surface low-velocity layer (SLVL), a layer of extremely porous lava and basalt breccia
with an average Pwave velocity of 2.2 km s
1
. The SLVL thickness increases from
<20 m along the axial summit trough (AST) to 120 m at near-axis lava depocenters,
which are not symmetric about the rise axis. Depocenters are located 0.5 km to the west
and 1.5 km to the east of the rise axis. Tomographic inversion of the Stoneley wave first
arrivals reveals that shear velocities in the SLVL covary with the layer thickness,
exhibiting a similar asymmetric pattern, with shear velocities increasing from 320 m s
1
near the AST to 520 m s
1
at the near-axis depocenters. Our analysis demonstrates that
the seismic characteristics of the extrusive layer near the rise axis are related primarily
to volcanic features and processes. The thickness and velocity of the SLVL are low on the
axis and within channel networks that deliver lava flows away from the axis and then
increase rapidly at the distal ends of the channels where the lavas are deposited. We find
that azimuthal anisotropy exerts only a weak influence on our Pwave first-arrival
times, which we model as weak (4%) seismic azimuthal anisotropy in the upper dikes with
a fast axis oriented N23–32W. We find no evidence for seismic azimuthal anisotropy
in the extrusive layer. INDEX TERMS:3025 Marine Geology and Geophysics: Marine seismics
(0935); 3035 Marine Geology and Geophysics: Midocean ridge processes; 7220 Seismology: Oceanic crust;
KEYWORDS:volcanic, seismic, structure
Citation: Sohn, R. A., S. C. Webb, and J. A. Hildebrand (2004), Fine-scale seismic structure of the shallow volcanic crust on the East
Pacific Rise at 950
0
N, J. Geophys. Res.,109, B12104, doi:10.1029/2004JB003152.
1. Introduction
[2] Magmatic processes at oceanic spreading centers
emplace a thin layer of lava (extrusives) at the very top of
the crustal section. These high-porosity lavas provide a
record of volcanic activity that can be used to study the
space-time patterns of eruptions at mid-ocean ridges
(MORs) [e.g., Christeson et al., 1994a, 1996; Harding et
al., 1993; Hooft et al., 1996; Kent et al., 1994], and they are
also an important physical and chemical interface for
hydrothermal convection [e.g., Alt,1995;Sleep, 1991].
Seismic methods play a key role in extrusive layer studies,
as they can provide estimates of the layer thickness and a
means to investigate structural variations related to volca-
nism, hydrothermal activity, and fracturing. While ship-
based (i.e., sea surface) multichannel seismics are perhaps
the most common method for measuring the extrusive layer
thickness, fine-scale structural variations within the extru-
sives are best imaged with on-bottom seismic surveys that
JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 109, B12104, doi:10.1029/2004JB003152, 2004
Copyright 2004 by the American Geophysical Union.
0148-0227/04/2004JB003152$09.00
B12104 1of17
provide high-fidelity measurements of internal refractions
and reflections. A fully on-bottom experiment configuration
is necessitated by the geometry of seismic propagation in
the shallow crust and overlying water column. The extru-
sive layer is much thinner than the overlying water column
(typically 1/20th), and in young oceanic crust the Pwave
propagation velocities are only slightly greater (less than a
factor of 2) than those in water. As a result, refractions from
the extrusive layer are obscured by the high-amplitude
direct water wave in most seismic records acquired using
near-surface seismic sources [e.g., Raitt, 1963].
[3] High-fidelity measurements of extrusive layer refrac-
tions become possible when seismic sources (and receivers)
are placed directly on the seafloor. While the technical
aspects and logistics of placing repeatable seismic sources
on the deep seafloor are formidable, this technique arguably
provides the best means to image the seismic structure of
the upper few hundred meters of oceanic crust. Body wave
refractions from on-bottom source-receiver pairs with slant
range offsets of less than 500 m in young oceanic crust
refract internally within the upper few hundred meters of
oceanic crust and provide a robust means to estimate
seismic velocities just beneath the seafloor-ocean interface.
[4] In this paper we present results from an on-bottom
seismic refraction experiment on the East Pacific Rise
(EPR) at 950
0
N designed to study the fine-scale seismic
structure underlying the high-temperature vent fields found
along the rise axis in this area. Our study is similar to the
Near Ocean Bottom Explosives Launcher (NOBEL) East
Pacific Rise Refraction Observations (NERO) experiment
conducted by G. M. Purdy and G. J. Fryer on the EPR axis
at 930
0
N(36 km south of our study area) [Christeson et
al., 1994a] in that we use on-bottom refraction methods to
resolve fine-scale variations in the seismic structure of
young crust emplaced on the axis of the fast-spreading
EPR. The primary differences, apart from locale, are:
(1) We arranged our sources and receivers in a 3-dimensional
(3-D) pattern as opposed to discrete lines for 2-D imaging;
(2) we employed four-component (three geophone channels
plus one hydrophone) ocean bottom seismometers (OBSs)
as opposed to single-channel ocean bottom hydrophones
(OBHs); and (3) we detonated our sources directly on
the seafloor as opposed to within the water column. The
latter two considerations proved to be especially important.
By placing our sources directly on the seafloor and by
measuring horizontal particle motion, we were able to
generate and observe very slowly propagating Stoneley
waves (waves that propagate at the interface of two homo-
geneous half spaces, often observed at solid-liquid inter-
faces) at the seafloor interface that provide excellent
constraints on the Svelocity structure of the shallow crust.
These are the first observations of Stoneley waves in
seismic layer 2, and they allow us to resolve lateral
variations within the Svelocity structure of mid-ocean ridge
lavas for the first time, a key development since Svelocity
is highly sensitive to key rock mechanical properties such as
porosity and shear modulus.
2. Experiment and Observations
[5] We deployed nine four-channel (three geophones
plus one hydrophone) ocean bottom seismometers (OBSs)
in a small-aperture network on the rise axis of the EPR at
950
0
N during the Temporal Observations of Eruption
Seismicity (TOES) experiment from March to June 1995
(Figure 1). This portion of the rise crest was volcanically
active in 1991/1992 [Haymon et al., 1993; Rubin et al.,
1994] and is the site of several well-studied high-temperature
hydrothermal vent fields [e.g., Fornari et al., 1998b;
Von Damm et al., 1995]. The principal objective behind
the OBS deployment was to study microearthquake activ-
ity beneath the high-temperature vent fields, and those
aspects of the experiment are reported by Sohn et al.
[1998, 1999].
[6] We supplemented the microearthquake survey by
conducting an on-bottom seismic refraction survey. Eigh-
teen explosive shots were detonated in a radial pattern
within and around the OBS network (Figure 1). Each shot
comprised 4 kg of high-density pentolite, three detraprime
boosters, and a Dupont E119 blasting cap to initiate
detonation. The shots were lowered to the seafloor on a
wire from the R/V New Horizon and were detonated via an
electronic trigger from a GPS clock. Returns from a 12 kHz
pinger, located 50 m above the explosive package, were
used to ensure that each shot was detonated directly on the
seafloor rather than in the water column. The seafloor
positions of the explosive shots and OBSs and the detona-
tion times of the shots were estimated with a joint inversion
[Creager and Dorman, 1982] using arrival times of direct
and surface-reflected water waves (generated by the shots,
recorded by the OBSs). The OBS positions were addition-
ally constrained by acoustic returns from a shipboard
transponder survey. The delay between the programmed
versus actual detonation time for each shot was typically
120–130 ms.
[7] Seismic arrivals from the explosive shots were
recorded by broadband seismometers with a flat accelera-
tion response from 0.016 to 40 Hz and were digitized at
rates of 128 Hz (vertical geophone channel and hydro-
phone) and 64 Hz (horizontal geophone channels). Record
sections from the survey contain solid Earth phases and a
series of water column multiples (Figure 2). The first solid
Earth phase is a direct Pwave (30–40 Hz) that refracts
within the surface low-velocity layer (SLVL) and upper
dikes. Direct crustal Pwaves were observed for all (144)
source-receiver pairs. The crustal Pwave is followed by the
direct water wave, and in some cases SLVL multiples are
observed between the direct crustal and water column
Pwaves (Figure 2a). The water column Pwave is followed
by slowly propagating shear crustal body waves and an
interface (Stoneley or Scholte) wave (Figure 2b). The
crustal Swave can be difficult to discern in the records
because it often arrives in the direct water phase wave train
and has a relatively low signal level. The Stoneley waves,
which were observed on 40% (58) of the source-receiver
pairs, have frequencies of 3 Hz and are associated
primarily with Swaves turning at shallow depths in
the extrusive layer. Stoneley waves have previously been
observed in marine sediments [e.g., Davies, 1965; Jensen
and Schmidt, 1986; Schreiner et al., 1991; Whitmarsh and
Lilwall, 1982], but this is the first time, to our knowledge,
that they have been observed in seafloor lavas. In this paper
we present analysis and inversions of first arrivals for
the crustal Pwave and the Stoneley wave observations.
B12104 SOHN ET AL.: SEISMIC STRUCTURE OF EAST PACIFIC RISE CRUST
2of17
B12104
Analysis of Swave refractions and synthetic fitting of the
Stoneley waveforms will be reported in a subsequent paper.
3. PWave First-Arrival Data, Models,
and Inversions
3.1. Average Velocity Model
[8] First-arrival travel time picks for the direct crustal
Pwaves and Stoneley waves are shown in Figure 3a. Pick
uncertainties are 10 ms for the Pwaves and 15–50 ms
for the Stoneley waves. We fit the Pwave first-arrival data
with a simple, 1-D model consisting of a 100 m thick
surface layer with a nearly constant V
p
of 2.2 km s
1
,a55m
thick high-gradient layer (47 s
1
), and a bottom layer with
an average V
p
of 4.9 km s
1
that extends from a depth of
155 m to the limit of ray coverage at 400 m (Figure 3b).
We adopt the nomenclature of Christeson et al. [1994a] and
refer to the top layer in our model as the SLVL to
distinguish it from layer 2A, which also includes the
underlying high-gradient layer. Thus the SLVL in our 1-D
‘‘average’’ model is 100 m thick, while layer 2A is 155 m
thick.
[9] Previous studies have shown that the thickness of
seismic layer 2A doubles within 1 –2 km of the rise axis,
so we developed a more complicated starting velocity
model by parameterizing the results of Christeson et al.
[1994a]. The off-axis thickness of layer 2A in this model
increases by both a thickening of the SLVL and a thickening
(weakening) of the underlying gradient. We generated a
2.5-D velocity model for the study area incorporating these
effects, assuming that the structural variations are symmet-
ric about the rise axis (i.e., lateral variations in the across-
axis direction only), and by ‘‘hanging’’ the velocity model
from the local bathymetry. We used forward modeling
(geometric ray tracing) to ‘‘tune’’ (i.e., perturb layer thick-
nesses and velocities) the Christeson et al. [1994a] results to
our data and ultimately arrived at a model with residuals
that have essentially the same variance as residuals from our
simple, two-layer model, even though the seismic structure
and ray paths in the two models are significantly different
(Figure 4).
[10] The similarity of the residual variance for these two
models illustrates the intrinsic nonuniqueness of velocity
models based solely on arrival time data and provides some
insight into modeling issues for this kind of analysis. The
experiment geometry limits resolution in that ray coverage
is not uniform beneath the study area, which allows for
some trade-offs between SLVL thickness and the velocity of
Figure 1. Experiment configuration and study area bathymetry (contours in meters). Seismic sources
are shown with gray circles. Seismic receivers (OBSs) are shown with black circles. Locations of three
vent fields known to be active during the experiment are shown with stars. Centerline of axial summit
trough (AST) from DSL-120 side-scan imagery (D. Fornari et al., The axial summit trough of the East
Pacific Rise 909
0
–59
0
N: New insights from DSL-120 sidescan and ABE sonar surveys, manuscript in
preparation, 2004, hereinafter referred to as Fornari et al., manuscript in preparation, 2004) is shown with
the heavy black line.
B12104 SOHN ET AL.: SEISMIC STRUCTURE OF EAST PACIFIC RISE CRUST
3of17
B12104
the underlying layer across the rise axis, but perhaps the
most important issues pertain to heterogeneity length scale
and structural asymmetry. As we will show, the shallow
seismic structure is not symmetric about the rise axis in our
study area, which limits the variance reduction that can be
achieved by our symmetric models. We also find that there
is significant structural heterogeneity in our study area with
length scales that are too small to be resolved with our
observations, which limits the variance reduction that can be
achieved by any constrainable model. That both of
our starting models achieve roughly the same variance
reduction is essentially a random result that arises from
the fact that neither model includes short-wavelength
heterogeneities that produce much of the arrival time
variance nor the structural asymmetry about the rise axis
that exerts a large influence on arrival times.
[11] We lack the sort of 2-D lines that could potentially be
used to formally discriminate between the two starting
models, but these types of lines were in fact the basis for
the results of Christeson et al. [1994a] that we used to
parameterize the variable 2A model. This suggests that the
variable 2A model may provide better insight into the
volcanic lithology near the rise axis, but rather than choose
between the starting models, we conducted all of our Pwave
analyses for both models in parallel. This allows us to assess
the sensitivity of our results to our choice of starting model
and provides perspective regarding the uniqueness of our
modeling and inversions. As we will show, our primary
Figure 2. Pseudorecord sections (i.e., record section compiled from complete data set, not a section
along a line) plotted so as to emphasize first arrivals used in analysis. (a) Raw vertical geophone
components, with Pbody wave and direct water wave. (b) Horizontal component band pass filtered from
2 to 5 Hz. Dashed line represents Stoneley wave average first-arrival velocity of 420 m s
1
. Stoneley
waves and first water column multiple are shown.
B12104 SOHN ET AL.: SEISMIC STRUCTURE OF EAST PACIFIC RISE CRUST
4of17
B12104
results are insensitive to our choice of starting model,
which ultimately affects the amplitude, but not so much
the nature, of the extrusive layer thickening predicted by our
analysis.
3.2. Station Corrections and SLVL Thickness
[12] We used the travel time residuals between our
Pwave first-arrival observations and predictions from the
1-D average and variable 2A thickness starting models
(Figures 4c and 4g) to constrain lateral variations in
the seismic structure of the shallow crust. We initially
attempted to parameterize inversions using lateral velocity
perturbations along interfaces of constant depth, which is a
standard approach for local seismic tomography, but we
were unable to significantly reduce the data variance
following this approach. After a great deal of trial and error
we began to realize that the velocity contrast between the
SLVL and the underlying layer is so strong that the critical
modeling issue is to properly position this interface beneath
the stations (sources and receivers), meaning that the
variance in our arrival time data primarily results from
thickness variations in the SLVL rather than lateral velocity
variations per se.
[13] Most of the Pwave ray paths are nearly vertical
within the SLVL, which allows us to model residuals by
defining a ‘‘station correction’’ [e.g., Aki and Richards,
1980, p. 610] that accounts for the travel time anomaly
generated by SLVL thickness variations beneath each
source and receiver. We modeled the travel time residual
for each source-receiver pair as the sum of the station
corrections for the source and the receiver. We then formu-
lated an inversion to estimate the station correction for each
source and receiver.
[14] We have one station correction value for each of the
18 sources and 8 receivers, yielding 26 model parameters
(m,261), and 144 data values representing the first-
arrival residuals for each source-receiver pair with respect to
the starting models (d, 144 1). The Green’s matrix (also
known as forward or sensitivity matrix) that expresses the
relationship between the data and model has one row for
each observation and one column for each model parameter
(G, 144 26). Thus
d¼Gm;ð1Þ
where each row of Ghas a value of 1 in the columns
corresponding to the source and receiver station correction
parameter for each row of dand zeros everywhere else. The
inversion must be damped because of the low condition
number of the sparse matrix G, which conceptually arises
from the fact that the prediction error is insensitive to the
addition of an arbitrarily large constant value to all station
corrections. We therefore estimate the model parameters
using damped least squares such that
^
m¼GTGþe2I
1GTd;ð2Þ
where eis the damping factor and Iis the identity matrix.
[15] Applying this method, we obtain a station correction
estimate for each source and receiver (Figure 5), yielding a
variance reduction of 51% for the 1-D starting model and
47% for the variable 2A starting model, with a nominal 1s
uncertainty of 6 ms. We obtain similar station corrections
(i.e., SLVL thickness perturbations) for both starting
models (Table 1). The station corrections must be associated
Figure 3. (a) First-arrival travel time picks for Pwaves (open circles) and Stoneley waves (open
squares), with average Stoneley wave first-arrival propagation time of 420 m s
1
shown as a dashed line.
(b) Average Pvelocity versus depth profile, with SLVL and layer 2A identified.
B12104 SOHN ET AL.: SEISMIC STRUCTURE OF EAST PACIFIC RISE CRUST
5of17
B12104
primarily with variations in SLVL thickness because the
sign and magnitude of the values are not consistent with
variations in SLVL velocity. If the station corrections
resulted from lateral velocity variations, then our results
would require that V
p
within the SLVL decrease from a
value of 4kms
1
on the rise axis to values of 1.5 km
s
1
at distances of 2 km from the rise axis. This result is
incompatible with a large body of research demonstrating
Figure 4. Starting model comparison of (a–d) one-dimensional, two-layer model and (e – h) variable
2A model. In Figures 4a and 4e, across-axis section shows model velocities as contours in m s
1
, with
SLVL and layer 2A identified. In Figures 4b and 4f, across-axis depth section shows ray paths in the
starting models. Figures 4c and 4g show fit of the travel time observations (open circles) versus model
predictions (closed circles) with corresponding variance. Figures 4d and 4h show fit of the travel time
observations (open circles) versus model predictions (closed circles) after application of station
corrections with corresponding variance.
B12104 SOHN ET AL.: SEISMIC STRUCTURE OF EAST PACIFIC RISE CRUST
6of17
B12104
Figure 5. Station correction results and SLVL thickness for (a) the 1-D starting model, (b) the variable 2A starting model,
and (c) the variable 2A starting model incorporating effects of SLVL velocity increasing away from the rise axis. Station
corrections (ms) from inversion are shown as text labels inside gray circles (sources) and black circles (receivers). Final
SLVL thickness estimates after application of the station corrections are plotted according to a common color map to allow
for direct comparison. High-temperature vent fields known to be active in 1995 are shown as white stars. Centerline of axial
summit trough (AST) from DSL-120 side-scan imagery (Fornari et al., manuscript in preparation, 2004) is shown with a
heavy black line. Color contours are estimated via nearest-neighbor interpolation as described in section 5.4. Note that the
color contours for Figures 5b and 5c are nearly identical, demonstrating that SLVL velocity changes away from the rise axis
have a negligible effect on our layer thickness estimates.
B12104 SOHN ET AL.: SEISMIC STRUCTURE OF EAST PACIFIC RISE CRUST
7of17
B12104
that the velocity of seismic layer 2A increases away from
the rise axis, and it is also incompatible with our Stoneley
wave analysis (section 4), which also shows that velocities
in our study area increase away from the rise axis.
[16] By process of elimination we therefore associate our
station corrections with SLVL thickness variations. The
simplest approach for mapping station corrections into
thickness variations is to assume that the average Pwave
velocity remains constant (Figures 5a and 5b). The thick-
ness estimates, however, are relatively insensitive to the
sorts of subtle velocity variations that would be expected
within 2 km of the rise axis. We included the effect of
SLVL velocities increasing away from the rise axis into our
mapping of station corrections into thickness estimates by
incorporating our Svelocity inverse results from section 4
and assuming that V
s
/V
p
is constant (Figure 5c). Thickness
estimates that include lateral velocity variations are difficult
to distinguish from estimates obtained assuming constant
velocities (cf. Figures 5b and 5c).
[17] Our estimate of the SLVL thickness for a given site
at 950
0
N depends on our choice of starting model, but
the overall pattern does not. Both inversions require that
the station corrections, and hence the SLVL thickness,
increase moving away from the rise axis. And in both
inversions we find that the thickening pattern is not
symmetric about the rise axis, but rather that the SLVL
thickens more rapidly on the west compared with the east
side of the rise axis (Figure 6). This marked asymmetry is
correlated with the local bathymetry, which falls off more
rapidly to the west side of the rise axis, suggesting that the
extrusive layer thickens primarily via magma effusing
from the AST and then flowing downhill in channels, as
we will describe in section 5.
3.3. Azimuthal Anisotropy
[18] Many studies have found evidence for seismic
azimuthal anisotropy within oceanic crust on the basis of
Swave particle motions [e.g., Stephen, 1981, 1985], Swave
splitting [e.g., Almendros et al., 2000; Stephen, 1981], and
azimuthal variations in Pvelocity [e.g., Dunn and Toomey,
2001; Shearer and Orcutt, 1985; Sohn et al., 1997; White
and Whitmarsh, 1984]. Here we examine whether or not
any of the remaining variance in our first-arrival data can
be explained with azimuthal variations in propagation
velocity. A simple plot of source-receiver azimuth versus
residual amplitude (Figure 7) reveals that there is not a
strong correlation between these two parameters. We used
least squares methods to fit a cos (2q) curve to these data
and found that there may be a small anisotropic signal in
the observations, but it is largely masked by larger ampli-
tude variations caused by short-wavelength heterogeneities.
The least squares inversion solves for the amplitude and
phase of the cos (2q) curve and the associated standard
errors, yielding an amplitude of 3.6 ± 2.45 10
3
s with
a fast azimuth oriented N23W±19for the 1-D
model and an amplitude of 4.5 ± 2.5 10
3
s with a
fast azimuth oriented N32W±15for the variable
2A model.
[19] These results are somewhat equivocal in that the
amplitude parameter estimates are only slightly larger than
the standard errors. However, the cos (2q) curve fit is not a
formal inversion for seismic anisotropy because it does not
account for body wave polarization variations along the
seismic ray paths, nor does it account for vertical variations
of anisotropic parameters within the crust. Given that
extrusive basalts are known to have dramatically different
fracture patterns than intrusives [e.g., van Everdingen, 1995;
Wilkens et al., 1991], it seems logical to expect that the
seismic anisotropy within our study area is vertically
stratified. To investigate this possibility, we formulated an
inversion that allows for estimation of anisotropic amplitude
at four depths (0, 100, 200, and 500 m) below the seafloor
[Sohn et al., 1997]. The fast azimuth of anisotropy is a prior
constraint that we fixed at N23W (the result of the cos (2q)
curve fit for the 1-D starting model), and the prior variance
at each node was set at (200 m s
1
)
2
.
[20] The inversion was performed using residuals after
station corrections for both the 1-D and variable 2A starting
models. The results (Table 2) indicate that statistically
significant azimuthal anisotropy is only resolved at the
200 m depth node, corresponding to the top of the dikes.
A negative result is returned for the shallowest layer (0 m
depth), meaning that the inverse has resolution at this depth
but fails to find any evidence for anisotropy. Equivocal
results are returned for the 100 and 500 m depth nodes,
meaning that the inverse does not have sufficient resolution
at these depths.
[21] Incorporation of weak (4 –5%) azimuthal anisotropy
(1-D model, 97 m s
1
; variable 2A model, 123 m s
1
, half
amplitude) into our model at a depth of 200 m below the
seafloor results in an additional variance reduction of 5%
and 8% for the 1-D and variable 2A models, respectively.
Anisotropy is therefore a subtle component of this data set,
Table 1. Station Correction Results
Station
a
1-D Starting Model
Correction, ms
Variable 2A Starting
Model Correction, ms
R2 14 14
R3 4 3
R4 16 12
R5 81
R6 12 9
R7 21 21
R8 22
R9 39
S1 12 9
S2 22
S3 63
S4 22
S5 17 16
S6 87
S7 10 8
S8 67
S9 6 5
S10 35
S11 2 0
S12 910
S13 11
S14 45
S15 8 7
S16 20 26
S17 5 3
S18 27
a
Stations are identified with source/receiver type and station ID number.
For example, R2 is receiver 2, and S10 is source 10. See Figure 1 for station
locations.
B12104 SOHN ET AL.: SEISMIC STRUCTURE OF EAST PACIFIC RISE CRUST
8of17
B12104
and we find no evidence for anisotropy in the SLVL. The
total variance reduction for the Pwave first-arrival model-
ing (station corrections plus anisotropy) is 56% for both
models, leaving almost half of the initial variance from our
starting models unaccounted for. It appears as though
almost all of the variance in the Pwave arrival times is
acquired as the waves propagate through the SLVL. SLVL
thickness variations resolvable with our station correction
analysis account for about half of this variation, while the
other half appears to result from unresolvable short-wave-
length (<100 m) heterogeneities.
4. Stoneley Wave Data and SVelocity Structure
[22] We picked first arrivals for 58 Stoneley waves and
used these data to formulate a tomographic inversion for
lateral variations in Swave propagation speed within the
SLVL within our study area. Stoneley wave first arrivals are
shallow turning Sbody waves or Swave multiples (the
Stoneley wave in a half-space model travels slightly slower
than the Swave), and our inversion therefore constrains
velocities in the uppermost section of the extrusive pile.
Stoneley wave propagation velocities exhibit considerable
variability, ranging from 275 to 590 m s
1
with a mean
value of 420 m s
1
(Figure 3). These values are consistent
with, though on the low side of, previous estimates of
Swave velocities for young crust on the EPR based on
Pto Sconversions at the base of layer 2A [Christeson et al.,
1997] and the amplitude of SLVL multiples [Christeson et
al., 1994a].
[23] We modeled the Stoneley wave ray paths as straight
lines in a uniform half space (V
s
= 420 m s
1
) between the
source and receiver and parameterized lateral velocity
variations as a truncated set of Fourier coefficients [e.g.,
Sohn et al., 1997]. Prior spectral variances for the model
parameters were weighted to impose smoothness constraints
and to minimize sidelobes associated with spectral trunca-
tion. The physical size of the model domain in the inversion
was increased by 50% in both the x(across-axis) and
y(along-axis) directions so that the implicit periodicity of
the spectral parameterization did not affect our results.
[24] The inversion seeks to minimize an objective func-
tion representing the weighted sum of the data misfit and the
size of the model parameters (perturbations),
dG^
mðÞC1
dd dG^
mðÞþ
^
mC1
mm
^
m¼minimum;ð3Þ
where dis the residual of each observation with respect to
an average velocity of 420 m s
1
, and C
dd
and C
mm
are the
prior covariance matrices for the data and model, respec-
tively. This equation is solved by differentiating and setting
the derivative to zero, resulting in
^
m¼GTC1
dd GþC1
mm
GTC1
dd d:ð4Þ
The parameter estimates are then obtained using a damped,
least squares inverse.
[25] We arrived at a final model by systematically in-
creasing the complexity until the variance reduction tapered
Figure 6. Seismic cross sections from the along-axis high at 950.0
0
N after station corrections for
(a) the 1-D starting model and (b) the variable 2A thickness starting model. Heavy black line is seafloor
topography along the cross section, with depth in m below sea surface. Contours represent Pwave
velocity in m s
1
.
B12104 SOHN ET AL.: SEISMIC STRUCTURE OF EAST PACIFIC RISE CRUST
9of17
B12104
off. The model complexity is determined by the number of
Fourier coefficients used to parameterize the velocity per-
turbations in the along- and across-axis directions. We
found that a systematic increase in Swave velocities away
from the rise axis is a common feature of all the models we
generated and is therefore considered to be a robust feature
of the data. Our final model (Figure 8) uses three across-axis
and eight along-axis Fourier coefficients and achieved a
variance reduction of 74%, with formal errors as shown in
Figure 8b. The nominal spatial resolution of this tomogram
is 1.3 km across axis and 450 m along axis.
[26]Swave velocities in our model increase from mini-
mum values of 320 m s
1
near the rise axis to maximum
values of 520 m s
1
at distances of 1–2 km from the AST.
Interestingly, the Svelocity pattern is also asymmetric about
the rise axis, similar to the SLVL thickness estimates from
the station corrections, with Svelocities increasing more
rapidly to the west compared with the east of the rise axis.
This covariation (Figure 9) has implications for volcanic
processes near the rise axis, as we will describe below. Here
we note that the sense of the arrival time residuals that
generate the perturbation patterns in the SLVL thickness and
Svelocity inversions is exactly opposite. In other words,
Pwave arrivals to off-axis instruments are generally slow
arrivals compared with on-axis instruments, whereas Swave
arrivals to off-axis instruments are generally fast. This
provides confidence that the covariation between SLVL
thickness and Svelocity is analytically robust and not an
artifact of timing or data handling errors.
[27] Our results suggest that Svelocities in the SLVL also
vary along the ridge axis, with a decrease of 10% moving
away from the along-axis high at 950.1
0
N in either direc-
tion (Figure 8). We utilized a variety of along-axis param-
eterizations to investigate this apparent trend and found that
Table 2. Results From Depth-Stratified Anisotropy Inversions
a
Depth Below
Seafloor, m
1-D Model Variable 2A Model
Anisotropy
Amplitude, m s
1
Anisotropy
Uncertainty, m s
1
Anisotropy
Amplitude, m s
1
Anisotropy
Uncertainty, m s
1
015 98 9 125
100 8 155 16 162
200 97 70 123 98
500 18 198 13 157
a
Anisotropy parameters were estimated at four depth intervals for two different isotropic seismic velocity models. Posterior
uncertainties are based on prior uncertainties of 200 m s
1
. The inversion can best resolve anisotropy in the 200 m depth interval,
where most rays turn and have nearly horizontal ray paths. The inversion also has limited resolution at 0 m where most ray paths
are oriented 30from the vertical, but 8 – 10 short-range ray paths are essentially horizontal.
Figure 7. Travel time residuals after application of station corrections versus source-receiver azimuth.
Data are shown as black circles. Cardinal direction north is shown with gray vertical band at 0. Best fit
(least squares) cos (2q) curve is plotted as a black sinusoidal line.
B12104 SOHN ET AL.: SEISMIC STRUCTURE OF EAST PACIFIC RISE CRUST
10 of 17
B12104
while it is not as robust as the across-axis variations, it is a
feature that appears in most of the models we produced. We
discuss the possible implications of these results for volca-
nic and hydrothermal processes in section 5.
5. Discussion
[28] Our results are consistent with previous seismic
studies of young crust on the EPR in that we find similar
average seismic velocities in the shallow crust and a
systematic evolution in seismic structure moving away
from the rise axis [e.g., Christeson et al., 1996; Harding
et al., 1993]. We provide new insight, however, by
providing fine-scale imagery of the SLVL thickness and
Svelocity variations within a 33 km region centered
on the along-axis high at 950
0
N. The volcanic and
hydrothermal features of our study area have been
extremely well characterized by a relatively long history
Figure 8. Svelocity inverse results. (a) Svelocity tomogram plotted for pixels with posterior error
<100 m s
1
. Velocity perturbations from the inversion are added to the mean value of 420 m s
1
to generate
absolute values of shear velocity as shown. Locations of seismic sources (gray circles), receivers (black
circles), known high-temperature vent fields active in 1995 (black stars), and AST centerline (heavy black
line) are shown for reference. (b) Map of posterior model error with same legend as Figure 8a.
B12104 SOHN ET AL.: SEISMIC STRUCTURE OF EAST PACIFIC RISE CRUST
11 of 17
B12104
of submersible observations, side-scan imagery, and near-
bottom magnetic surveys [e.g., Fornari et al., 2004;
Schouten et al., 2003; Von Damm, 2000], which provides
excellent context for our results and allows us to carefully
examine the relationship between these processes and our
seismic results.
5.1. Evolution of SLVL Seismic Properties
and Relationship With Volcanism
[29] The interaction between cold seawater and basaltic
liquids emplaced on the seafloor during volcanic eruptions
at oceanic spreading centers generates a porous and perva-
sively fractured lava pile with very low seismic propagation
velocities. There is a strong seismic velocity contrast
between the lavas and the underlying intrusives (dikes),
which is particularly strong in young crust at fast-spreading
centers where volcanism dominates the crustal accretion
process. This contrast is immediately evident in our average
Pvelocity model (Figure 3b) where a thin (55 m) high-
gradient layer separates extrusive material with Pvelocities
of 2.2 km s
1
from intrusive material with Pvelocities of
4.8 km s
1
. Following Christeson et al. [1994a], we refer
to the low-velocity layer at the top of our model as the
SLVL, and we use the SLVL thickness as a conservative
proxy for extrusive layer thickness within our study area
because Pvelocities within the SLVL are too low to be
associated with intrusive material. Previous studies have
commonly used seismic layer 2A as a proxy for extrusive
layer thickness [e.g., Christeson et al., 1994a, 1996;
Harding et al., 1993; Hooftetal., 1996; Kent et al.,
1994], but here we use the SLVL because, while the
lithologic interpretation of the velocity gradient at the base
of seismic layer 2A can be debated, there is no question that
the SLVL, with Pvelocities <2.5 km s
1
, is composed of
extrusives. Our extrusive layer thickness estimates are
therefore conservative in that some fraction of the material
in the underlying gradient at the base of layer 2A may be
extrusive in nature, as well.
[30] The principal result of our study is the observation
that the thickness and Svelocity of the SLVL, and therefore
the extrusive pile, are closely coupled and that they covary
in a systematic (though asymmetric) pattern about the rise
axis (Figure 9). Inspection of the local bathymetry reveals
that the rise axis itself is also asymmetric (Figure 1), and all
three of these data sets (SLVL thickness, SLVL Svelocity,
and bathymetry) have the same sense of asymmetry: The
thickness of the extrusive pile, Svelocities of the lava pile,
and the seafloor depth all increase more rapidly to the west
compared with the east of the AST. Close inspection of the
microbathymetry and side-scan sonar imagery [Fornari et
al., 2004] reveals that the correlation between SLVL thick-
ness and volcanic features is surprisingly specific. In the
immediate vicinity of the AST the ridge has a scallopy
texture representing discrete flow fronts and lava channel
networks, and this texture is evident in the SLVL thickness
map (Figure 10), providing compelling evidence that the
Figure 9. Covariation of SLVL thickness and Svelocity. Thickness is shown as contours (m), and
Svelocity is shown in colors. AST centerline is shown as a heavy black line for positional reference along
with three main regions of high-temperature venting (north to south equals M, Q, minus Bio9, P minus tube
worm pillar, Y) shown as stars. Note how SLVL thickness contours mimic scallopy lava flow fronts in
bathymetric and side-scan imagery, particularly on the west side of the AST.
B12104 SOHN ET AL.: SEISMIC STRUCTURE OF EAST PACIFIC RISE CRUST
12 of 17
B12104
station corrections accurately render fine-scale thickness
variations with the lava pile.
[31] The combination of submersible observations,
microbathymetric surveys, side-scan sonar surveys, and
near-bottom magnetometer surveys [Fornari et al., 2004;
Schouten et al., 2003] has revealed the presence of lava
channel networks that provide pathways for lavas to flow up
to 1.5 km away from the AST and onto the rise flanks.
Some of these channels are clearly evident in microbathy-
metric data, but others are more difficult to image because
they are shells covered with a thin roof and thus can only be
unequivocally identified by visual observations. The chan-
nels terminate rather abruptly (within hundreds of meters)
on the steeper west side of the AST, but on the east side the
channels are longer and in at least one case appear to reach
all the way to the first inward-facing fault scarp located
1.5 km from the rise axis (Figure 10). The scarp is 3m
high and runs roughly parallel to the rise axis but becomes
buried at the edge of the lava channels, at 950
0
8
00
N. Our
results demonstrate that the extrusive layer thickness and
Svelocity remain low (40 m and 370 m s
1
, respectively)
within this channel and then increase abruptly (>120 m and
500 m s
1
, respectively) at the distal end where the growth
fault is buried. Minimum Svelocities near the rise axis are
as low as 320 m s
1
, which is somewhat lower than
previous values derived from Pto Sconversions at the base
of layer 2A (400 –800 m s
1
)[Christeson et al., 1997] and
the amplitude of SLVL multiples (<750 m s
1
)[Christeson
et al., 1994a]. This small discrepancy is probably real in that
the previous constraints from both the Pto Sconversions
and SLVL multiples constrain whole layer average veloci-
ties, whereas our surface wave observations are most
sensitive to Svelocities at the very top of the SLVL, where
velocities are likely to be somewhat lower than layer
averages.
[32] Comparison of our results with near-bottom bathym-
etry geophysical imagery of the rise axis therefore reveals
that the seismic characteristics of the SLVL near the rise
axis are tightly coupled with volcanic features and processes.
To first order we see that the thickness and Svelocity of the
extrusive layer remain low in areas where channels transport
lava away from the rise axis and then increase simultaneously
as the lavas pile up in near-axis depocenters. This observa-
tion lends support to the idea that eruptions primarily initiate
within the AST and that lava subsequently flows downhill
once the AST walls are breached [e.g., Fornarietal.,
1998a]. The magnitude of extrusive layer thickening
deduced from our results depends on our choice of
starting model for the station correction inversion. The
minimum variation estimate from our 1-D starting model
inversion has the SLVL thickness increasing from 60 m
on the rise axis to 140 m at the near-axis depocenters.
The maximum variation estimate from our variable 2A
starting model has SLVL thickness increasing from <20 m
Figure 10. SLVL thickness results from the variable 2A thickness starting model (same as Figure 5b but
with amplified color scale) with overlay of geological features interpreted from side-scan sonar imagery
[Fornari et al., 2004], submersible observations (D. Fornari and M. Tivey, personal communication,
2004), and near-bottom microbathymetry and magnetics surveys [Schouten et al., 2003]. Multibeam
bathymetric map with 10 m contour interval [Cochran et al., 1999] is shown for reference. Eastern and
western edges of AST are based on DSL-120 side-scan sonar images (Fornari et al., manuscript in
preparation, 2004). Lava channel networks on the east side of the AST are confirmed with visual
observations from submersible dives. Channel networks on the west side are inferred from side-scan
sonar and magnetics data. Arrows indicate lateral lava flow paths inferred from the channel network
geometry.
B12104 SOHN ET AL.: SEISMIC STRUCTURE OF EAST PACIFIC RISE CRUST
13 of 17
B12104
on the rise axis to 120 m at the near-axis depocenters. We
are formally unable to discern between these two possibil-
ities, but other seismic studies of layer 2A on the EPR that
were configured to make measurements along 2-D flow
lines and isochrons [e.g., Christeson et al., 1994a, 1994b,
1996, 1997; Hussenoeder et al., 2002] suggest that the
results from the variable 2A starting model may be more
representative of the true lithology.
[33] We cannot discern any correlation between the
seismic characteristics of the SLVL and hydrothermal
processes. There are three major high-temperature vent
fields in our study area, and they do not appear to exert
any influence on our seismic data. Thus any influence that
hydrothermal processes may be exerting on the mechanical
and hence seismic properties of the extrusive layer has a
length scale that is too small to be observed by our
experiment (<100 m). This is consistent with the fact that
high-temperature vent fields on the EPR have very small
seafloor expressions (tens of meters or less) and suggests
that processes such as pore space infilling from interaction
with cold fluids and mineral deposition are occurring at
correspondingly small length scales within the lavas. This
does not mean that hydrothermal alteration and mineral
deposition are not an important processes for long-term
‘‘aging’’ and seismic evolution of layer 2A but rather
suggests that these processes are not important factors in
the seismic velocities of very young lavas at fast-spreading
centers and/or that focused mineral deposition from high-
temperature vent fields has length scales on the order of a
few tens of meters.
5.2. Seismic Azimuthal Anisotropy
[34] Our analysis suggests that azimuthal anisotropy in
the uppermost dike section exerts some impact on our Pwave
arrival times but that this amounts to only 5% of the total
data variance. We find evidence for weak (4–5%) azimuthal
anisotropy within the uppermost dikes, which is consistent
with our understanding of dike fracture patterns. Individual
dikes are tabular features to first order and are emplaced as
thin ‘‘blades’’ of magma perpendicular to the axis of least
compressive stress [e.g., Rubin, 1992]. At mid-ocean ridges
the axis of least compressive stress is generally perpendic-
ular to the strike of the rise axis, such that dikes are
emplaced in along-axis orientations. The fractures that
define the margins of individual dikes are therefore prefer-
entially aligned parallel to the rise axis [e.g., van Everdingen,
1995], which can generate seismic azimuthal anisotropy
with a fast axis oriented along the strike of the rift
zone.
[35] Our anisotropy amplitude estimate of 4 –5% for the
upper dikes is similar to values reported for the middle and
lower dikes on the basis of Pwave arrival time residuals
from air gun refraction surveys on the EPR at 930
0
N[Dunn
and Toomey, 2001] and the Mid-Atlantic Ridge at 35N
[Barclay et al., 1998] and is slightly lower than the value of
6% reported for the middle and lower dikes at the Juan de
Fuca Ridge [Sohn et al., 1997]. The air gun surveys cannot
resolve anisotropy in the upper 400 m of the crustal
section for the ray geometry and propagation reasons
described in section 1 [e.g., Raitt, 1963], and therefore
our results illuminate the air gun survey results by con-
straining azimuthal anisotropy within their ‘‘blind spot.’’
[36] The fast azimuth of anisotropy in our inversions is
aligned at an absolute orientation of between N23–32W.
Interestingly, this is aligned with the strike of the local rise
axis as opposed to the larger-scale plate boundary, which
strikes at N9W. This suggests that the strike of the dike
section, and hence the local stress field, deviates by 14
with respect to the larger-scale plate boundary. We can only
speculate regarding the possible causes of this rotation, but
we note that the Lamont seamount chain intersects the EPR
at 950
0
N, raising the possibility that interaction between the
melting anomaly that generates the seamount chain and the
spreading center has locally perturbed the stress field [e.g.,
Pockalny et al., 1997].
[37] Our depth-stratified anisotropy inversion returns a
negative result for the extrusives, arguing against seismic
azimuthal anisotropy in this upper layer. If we take this
result at face value, then our results suggest that the
aggregate fracture set that controls seismic velocity in the
extrusives is not preferentially aligned. This is consistent
with extrusive velocities being controlled by thin, high
aspect ratio fractures resulting from rapid thermal contrac-
tion and flow deformation, as opposed to larger scale
tectonism. Larger-scale fissures, such as the AST itself,
which are obviously preferentially aligned with the rise axis
and penetrate into the extrusive layer, do not appear to exert
a significant impact on average velocities. More work may
be required, however, to verify this null result since we have
but a handful of rays that refract entirely within the SLVL.
Our negative anisotropy result for the extrusive layer con-
trasts with the study of McDonald et al. [1994] on the Cleft
segment of the Juan de Fuca Ridge, which reported an
extrusive layer anisotropy of 19% on the basis of analysis
of Pwave residuals in air gun refraction data. Our results
are probably more robust in that our on-bottom experiment
configuration provides first-arrival observations at very
short source-receiver offsets for phases that propagate
nearly horizontally within the extrusive layer, which is ideal
for constraining anisotropy, but we cannot rule out the
possibility that extrusive layer anisotropy is a site-specific
characteristic. More studies are clearly needed to determine
the nature of vertically stratified anisotropy within layer 2
and to assess variability between geologic/tectonic environ-
ments (i.e., fast versus slow or intermediate spreading).
5.3. Elastic Characteristics of the Extrusive Layer
[38] The elastic/mechanical characteristics of the extru-
sive layer are important for understanding the relationship
between crustal deformation and seafloor geology and for
understanding how tidal loading and stress perturbations
from nearby earthquakes may influence hydrothermal cir-
culation within the shallow crust. The elastic characteristics
of a solid material are uniquely determined by three
parameters: V
p
,V
s
, and r(density). In this study we have
constrained V
p
and V
s
, and the average density of the
extrusive layer at 950
0
N has been previously constrained
by seafloor [Stevenson and Hildebrand, 1996] and near-
bottom [Cochran et al., 1999] gravimetric measurements.
As a result, we are in a unique position to estimate the
elastic characteristics of the extrusive layer at this site,
which are shown in Table 3.
[39] The most striking aspect of these estimates is the
extremely low shear modulus of 425 MPa that arises from
B12104 SOHN ET AL.: SEISMIC STRUCTURE OF EAST PACIFIC RISE CRUST
14 of 17
B12104
the low Svelocities we measured from the interface waves.
This indicates that the top of the extrusive layer near the rise
axis is a combination of massively porous channels and
essentially unconsolidated basaltic rubble. This layer has
almost no shear strength, indicating that the rock behaves
similar to a fluid (Poisson’s ratio equals 0.48). The lateral
variations in Svelocity we have imaged translate into
corresponding variations in shear modulus. Our results
indicate that the shear modulus increases from a value of
247 MPa on the rise axis to a value of 602 MPa 1.5 km
from the rise axis. The shear strength of the extrusive layer
therefore more than doubles over this length interval. The
very low shear modulus values we obtain for the extrusive
layer preclude the development of significant levels of
differential stress or strain in the extrusives, and we may
expect this layer to deform more like a pile of bricks rather
than a coherent solid matrix. The shear modulus results also
have implications for poroelastic modeling of the effects of
ocean tides on hydrothermal circulation, which is sensitive
to the elastic parameters of the extrusive layer [e.g., Jupp
and Schultz, 2004; Wang and Davis, 1996]. Our constraints
on lateral variability of elastic parameters may eventually
prove useful for future modeling efforts when they begin to
address 2- and 3-dimensional systems.
5.4. Inverse Resolution and Regularization
[40] We made seismic velocity models from three different
data inversions: (1) station correction inversion, (2) depth-
stratified azimuthal anisotropy inversion, and (3) Stoneley
wave Svelocity inversion. These inversions have disparate
resolutions and regularization schemes, as we describe here.
As described in section 3.2, station corrections are effectively
point measurements, and there is no formal regularization
scheme applied to the inversion. Ray paths beneath each
source and receiver form an inverted cone within the SLVL,
and the length scale of this cone represents an effective
regularization since the estimates are formulated by averag-
ing over all of the ray paths in a cone. The radius of a cone
beneath a given station depends on the SLVL thickness and
the propagation angle of the seismic rays therein. In this
experiment the nominal propagation angle within the SLVL
is 30for most rays (i.e., turning in the upper dikes), and if
we take the SLVL as ranging from 20 to 120 m, then we find
that the station corrections represent thicknesses averaged
over lateral length scales of 12 –70 m. For graphical pur-
poses we produced contour maps of SLVL thickness by
generating a Cartesian grid with nodes spaced every 80 m in
the north-south and east-west directions and then interpolat-
ing between the two nearest stations within 600 m of the grid
node. The 600 m distance threshold was determined via trial
and error.
[41] The depth-stratified anisotropy inversion does not
incorporate lateral variations, and the resolution is therefore
limited to the vertical axis. We parameterized the vertical
variations as triangle functions, which assume that the
magnitude of anisotropy varies linearly between vertical
nodes. The spatial resolution is controlled by the number
and location of the depth interfaces chosen for parameter-
izing the inversion. Our choice of depth nodes was based
primarily on the relationship of the ray path geometry for
the two starting models with the inferred lithologic stratifi-
cation (Figures 4b and 4f) and a desire to minimize the
number of degrees of freedom in the inverse. Here again we
used trial and error to arrive at a final parameterization. The
ability of the inverse to resolve anisotropy at each discrete
depth node is quantified by the diagonal elements of the
posterior covariance matrix or model error (Table 2).
[42] The lateral resolution of the Stoneley wave first-
arrival inversion for Svelocity structure is a function of the
number of Fourier coefficients used to parameterize varia-
tions in the along- and across-axis directions. In contrast to
the station corrections, the Svelocity models therefore do
have prior smoothness constraints with explicit length
scales determined by the wavelengths of the Fourier coef-
ficients and by spectral weighting via specification of prior
model variances.
[43] The net result is that our models of Svelocity
and SLVL thickness have dramatically different spatial
parameterizations and smoothness constraints. Thickness
estimates can vary dramatically from one nearby station
to the next (e.g., OBS 7 and shot 17), whereas Svelocities
cannot, because they are regularized as described above.
For example, the minimum axis of the Svelocity model is
offset a few hundred meters east of the AST (Figure 8) such
that Svelocities are higher in the AST than they are a few
hundred meters to the east. This is almost certainly not
the case but rather reflects the effect of regularization since
the inverse is trying to model the effects of structural
asymmetry about the rise axis with a finite set of Fourier
coefficients.
6. Summary
[44] We use observations from an on-bottom seismic
refraction survey to constrain the fine-scale seismic struc-
ture of the upper 250 m of volcanic crust on the EPR at
950
0
N. Our principal results are that
[45] 1. The seismic structure of our study area is capped
by a low-velocity layer (SLVL) with average Pvelocities of
2.2 km s
1
and with Svelocities of 420 m s
1
near the
top of the layer. We are able to model 50% of the
variability in Pwave first-arrival times with respect to our
starting models as SLVL thickness variations and 74%
of the Stoneley wave first-arrival times as later velocity
variations.
[46] 2. SLVL thickness increases systematically off axis.
In a conservative (minimum variation) model the thickness
increases from 60 m on the rise axis to 140 m at a distance
of 1.5 km from the axis. In a more aggressive model that
incorporates constraints from other seismic experiments, the
thickness increases from <20 m on the rise axis to 120 m at
1.5 km off axis.
Table 3. Elastic Parameters Derived for the Extrusive Layer at
950
0
N on the Basis of Average Values of V
p
= 2.2 Versus
0.42 km s
1
and Density = 2410 kg m
3
Parameter Value
Shear modulus, m=rb
2
425 MPa
Lambda parameter, l=a
2
r2m10.8 GPa
Poisson’s ratio, s=l/[2(l+m)] 0.48
Bulk modulus, k=l+(2m/3) 11.1 GPa
Young’s modulus, E = m[(3l+2m)]/(l+m) 1.3 GPa
B12104 SOHN ET AL.: SEISMIC STRUCTURE OF EAST PACIFIC RISE CRUST
15 of 17
B12104
[47]3.Svelocity covaries with SLVL thickness, with
minimum velocities of 320 m s
1
on the rise axis
increasing to 500 m s
1
at 1–1.5 km away from the
AST. The patterns of SLVL thickness and Svelocity are
both asymmetric about the AST and increase more rapidly
on the west side of the rise axis where the bathymetric
gradient is steeper.
[48] 4. The SLVL thickness estimates are correlated with
seafloor volcanic features imaged by complementary near-
bottom geophysical surveys and closely mimic the scallopy
lava flow pattern observed in microbathymetric and side-
scan sonar surveys. This supports the concept that these
features are channels that provide primary pathways for
lavas erupting out of the AST and then subsequently
flowing downhill. The SLVL thickness is observed to
increase abruptly at lava depocenters located at the distal
end of the channels.
[49] 5. There is weak (4 –5%) azimuthal anisotropy
within the upper dikes with a fast axis oriented with the
local ridge strike at N23–32W. No evidence for azimuthal
anisotropy is found in the SLVL.
[50] 6. The shear modulus of the SLVL varies from
250 MPa in channel networks on the rise axis to 600 MPa
at the near (1.5 km) axis lava depocenters.
[51] 7. We are unable to account for 45% of the
variance in the Pwave first-arrival data, indicating that
heterogeneities with length scales less than the resolving
power of our study (100 m) exert a significant influence
on seismic propagation within the upper 250 m of crust.
[52]Acknowledgments. The authors thank Jacques Lemire and Tom
Deaton for engineering support for the ocean bottom seismometers and
John Boaz for technical support at sea. The manuscript has benefited
significantly from discussions with Gail Christeson, Suzanne Carbotte,
Alistair Harding, Maurice Tivey, Dan Fornari, and Hans Schouten and from
thoughtful reviews by Andrew Barclay, Fred Simons, and an anonymous
reviewer.
References
Aki, K., and P. G. Richards (1980), Quantitative Seismology, Theory and
Methods, W. H. Freeman, New York.
Almendros, J., A. H. Barclay, W. S. D. Wilcock, and G. M. Purdy (2000),
Seismic anisotropy on the Juan de Fuca Ridge from observations of
shear-wave splitting, Geophys. Res. Lett.,27, 3109 – 3112.
Alt, J. C. (1995), Subseafloor processes in mid-ocean ridge hydrothermal
systems, in Seafloor Hydrothermal Systems: Physical, Chemical, Biolog-
ical, and Geological Interactions,Geophys. Monogr. Ser., vol. 91, edited
by S. E. Humphris et al., pp. 85 – 114, AGU, Washington, D. C.
Barclay, A. H., D. R. Toomey, and S. C. Solomon (1998), Seismic structure
and crustal magmatism at the Mid-Atlantic Ridge, 35N, J. Geophys.
Res.,103, 17,827 – 17,844.
Christeson, G. L., G. M. Purdy, and G. J. Fryer (1994a), Seismic constraints
on shallow crustal emplacement processes at the fast spreading East
Pacific Rise, J. Geophys. Res.,99, 17,957 – 17,973.
Christeson, G. L., W. S. D. Wilcock, and G. M. Purdy (1994b), The shallow
attenuation structure of the fast-spreading East Pacific Rise near 930
0
N,
Geophys. Res. Lett.,21, 321 – 324.
Christeson, G. L., G. M. Kent, G. M. Purdy, and R. S. Detrick (1996),
Extrusive thickness variability at the East Pacific Rise, 910N: Con-
straints from seismic techniques, J. Geophys. Res.,101, 2859 – 2873.
Christeson, G. L., P. R. Shaw, and J. D. Garmany (1997), Shear and com-
pressional wave structure of the East Pacific Rise, 9–10N, J. Geophys.
Res.,102, 7821 – 7835.
Cochran, J. R., D. J. Fornari, B. J. Coakley, R. Herr, and M. A. Tivey
(1999), Continuous near-bottom gravity measurements made with a
BGM-3 gravimeter in DSV Alvin on the East Pacific Rise crest near
931
0
N and 950
0
N, J. Geophys. Res.,104, 10,841 – 10,861.
Creager, K. C., and L. M. Dorman (1982), Location of instruments on the
seafloor by joint adjustment of instrument and ship positions, J. Geophys.
Res.,87, 8379 – 8387.
Davies, D. (1965), Dispersed Stoneley waves on the ocean bottom, Bull.
Seismol. Soc. Am.,55, 903 – 918.
Dunn, R. A., and D. R. Toomey (2001), Crack-induced seismic anisotropy
in the oceanic crust across the East Pacific Rise (930
0
N), Earth Planet.
Sci. Lett.,189, 9 – 17.
Fornari, D. J., R. M. Haymon, M. R. Perfit, T. K. P. Gregg, and M. H.
Edwards (1998a), Axial summit trough of the East Pacific Rise 9–10N:
Geological characteristics and evolution of the axial zone on fast spread-
ing mid-ocean ridges, J. Geophys. Res.,103, 9827 – 9855.
Fornari, D. J., T. M. Shank, K. L. Von Damm, T. K. P. Gregg, M. Lilley,
G. Levai, A. Bray, R. M. Haymon, M. R. Perfit, and R. A. Lutz (1998b),
Time-series temperature measurements at high-temperature hydrothermal
vents: East Pacific Rise 949
0
Nto951
0
N: Monitoring dike intrusion and
crustal cracking events, Earth Planet. Sci. Lett.,160, 419 – 431.
Fornari, D., et al. (2004), Submarine lava flow emplacement at the East
Pacific Rise 950
0
N: Implications for uppermost ocean crust stratigraphy
and hydrothermal fluid circulation, in Mid-Ocean Ridges: Hydrothermal
Interactions Between the Lithosphere and Oceans,Geophys. Monogr.
Ser., vol. 148, edited by C. R. German, J. Lin, and L. M. Parson,
pp. 187 –218, AGU, Washington, D. C.
Harding, A. J., G. M. Kent, and J. A. Orcutt (1993), A multichannel
seismic investigation of upper crustal structure at 9N on the East Pacific
Rise: Implications for crustal accretion, J. Geophys. Res.,98, 13,925 –
13,944.
Haymon, R. M., et al. (1993), Volcanic eruption of the mid-ocean ridge
along the East Pacific Rise crest at 945 – 52
0
N: Direct submersible ob-
servations of seafloor phenomena associated with an eruption event in
April, 1991, Earth Planet. Sci. Lett.,119 , 85 – 101.
Hooft, E. E. E., H. Schouten, and R. S. Detrick (1996), Constraining crustal
emplacement processes from the variation in seismic layer: 2. A thickness
at the East Pacific Rise, Earth Planet. Sci. Lett.,142, 289 – 309.
Hussenoeder, S. A., R. S. Detrick, G. M. Kent, H. Schouten, and A. J.
Harding (2002), Fine-scale seismic structure of young upper crust at
1720
0
S on the fast spreading East Pacific Rise, J. Geophys. Res.,
107(B8), 2158, doi:10.1029/2001JB001688.
Jensen, F. B., and H. Schmidt (1986), Shear properties of ocean sediments
determined from numerical modeling of Scholte wave data, in Ocean
Seismo-Acoustics: Low-Frequency Underwater Acoustics, edited by
T. Akal and J. M. Berkson, pp. 683 – 692, Plenum, New York.
Jupp, T. E., and A. Schultz (2004), A poroelastic model for the tidal mod-
ulation of seafloor hydrothermal systems, J. Geophys. Res.,109, B03105,
doi:10.1029/2003JB002583.
Kent, G. M., A. J. Harding, J. A. Orcutt, R. S. Detrick, J. C. Mutter, and
P. Buhl (1994), Uniform accretion of oceanic crust south of the Garret
transform at 1415
0
S on the East Pacific Rise: Implications for crustal
accretion, J. Geophys. Res.,99, 9097 – 9116.
McDonald, M. A., S. C. Webb, J. A. Hildebrand, B. D. Cornuelle, and C. G.
Fox (1994), Seismic structure and anisotropy of the Juan de Fuca Ridge
at 45N, J. Geophys. Res.,99, 4857 – 4873.
Pockalny, R. A., P. J. Fox, D. J. Fornari, K. C. Macdonald, and M. R. Perfit
(1997), Tectonic reconstruction of the Clipperton and Siqueiros Fracture
Zones: Evidence and consequences of plate motion change for the last
3Myr,J. Geophys. Res.,102, 3167 – 3181.
Raitt, R. W. (1963), The crustal rocks, in The Sea: The Earth Beneath the
Sea, edited by M. N. Hill et al., pp. 85– 102, John Wiley, Hoboken, N. J.
Rubin, A. M. (1992), Dike-induced faulting and graben subsidence in
volcanic rift zones, J. Geophys. Res.,97, 1839 – 1858.
Rubin, K. H., J. D. Macdougall, and M. R. Perfit (1994),
210
Po-
210
Pb dating
of recent volcanic eruptions on the sea floor, Nature,368, 841 – 844.
Schouten, H., M. A. Tivey, D. Fornari, D. Yoerger, A. Bradley, M. Edwards,
and P. Johnson (2003), Central anomaly magnetization high: Constraints
on the volcanic construction and architecture of young upper oceanic
crust, EPR 9–10N, in Ridge 2000 Events, edited by D. Hassler,
C. Fisher, and S. Givens, pp. 30–34, Ridge 2000 Off., University Park,
Pa.
Schreiner, A. E., L. M. Dorman, and L. D. Bibee (1991), Shear wave
velocity structure from interface waves at two deep water sites in the
Pacific Ocean, in Shear Waves in Marine Sediments, edited by J. M.
Hovem, R. Stoll, and M. Richardson, pp. 231 –238, Kluwer Acad.,
Norwell, Mass.
Shearer, P. M., and J. A. Orcutt (1985), Anisotropy in the oceanic litho-
sphere: Theory and observations from the Ngendie seismic refraction
experiment in the southwest Pacific, Geophys. J. R. Astron. Soc.,80,
493 – 526.
Sleep, N. H. (1991), Hydrothermal circulation, anhydrite precipitation, and
thermal structure at ridge axes, J. Geophys. Res.,96, 2375 – 2387.
Sohn, R. A., S. C. Webb, J. A. Hildebrand, and B. D. Cornuelle (1997),
Three-dimensional tomographic velocity structure of the upper crust,
coaxial segment, Juan de Fuca Ridge: Implications for on-axis evolution
and hydrothermal circulation, J. Geophys. Res.,102, 17,679 – 17,695.
B12104 SOHN ET AL.: SEISMIC STRUCTURE OF EAST PACIFIC RISE CRUST
16 of 17
B12104
Sohn, R. A., D. Fornari, K. L. Von Damm, J. A. Hildebrand, and S. C.
Webb (1998), Seismic and hydrothermal evidence for a cracking event on
the East Pacific Rise crest at 950
0
N, Nature,396, 159 – 161.
Sohn, R. A., J. A. Hildebrand, and S. C. Webb (1999), A microearthquake
survey of the high-temperature vent fields on the volcanically active East
Pacific Rise, J. Geophys. Res.,104, 25,367 – 25,378.
Stephen, R. A. (1981), Seismic anisotropy observed in upper oceanic crust,
Geophys. Res. Lett.,8, 865 – 868.
Stephen, R. A. (1985), Seismic anisotropy in the upper oceanic crust,
J. Geophys. Res.,90, 11,383 – 11,396.
Stevenson, J. M., and J. A. Hildebrand (1996), Gravity modeling of a
volcanically active site on the East Pacific Rise axis, Tectonophysics,
254, 57 – 68.
van Everdingen, D. A. (1995), Fracture characteristics of the sheeted dike
complex, Troodos ophiolite, Cyprus: Implications for permeability of
oceanic crust, J. Geophys. Res.,100, 19,957 – 19,972.
Von Damm, K. L. (2000), Chemistry of hydrothermal vent fluids from 9–
10N, East Pacific Rise: ‘‘Time zero,’’ the immediate posteruptive period,
J. Geophys. Res.,105, 11,203 – 11,222.
Von Damm, K. L., S. E. Oosting, R. Kozlowski, L. G. Buttermore, D. C.
Colodner, H. N. Edmonds, J. M. Edmond, and J. M. Grebmeier (1995),
Evolution of East Pacific Rise hydrothermal vent fluids following a
volcanic eruption, Nature,375, 47 – 50.
Wang, K., and E. E. Davis (1996), Theory for the propagation of tidally
induced pore pressure variations in layered subseafloor formations,
J. Geophys. Res.,101, 11,483 – 11,495.
White, R. S., and R. B. Whitmarsh (1984), An investigation of seismic
anisotropy due to cracks in the upper oceanic crust at 45N, Mid-Atlantic
Ridge, Geophys. J. R. Astron. Soc.,79, 439 – 468.
Whitmarsh, R. B., and R. C. Lilwall (1982), A new method for the deter-
mination of in-situ shear-wave velocity in deep-sea sediments, paper
presented at Oceanology International 1982, Spearhead Exhibitions,
Kingston-upon-Thames, U. K.
Wilkens, R. H., G. J. Fryer, and J. Karsten (1991), Evolution of porosity
and seismic structure of upper oceanic crust: Importance of aspect ratios,
J. Geophys. Res.,96, 17,981 – 17,995.
J. A. Hildebrand, Scripps Institution of Oceanography, University of
California, San Diego, 9500 Gilman Dr., La Jolla, CA 92093-0205, USA.
(jah@mpl.ucsd.edu)
R. A. Sohn, Woods Hole Oceanographic Institution, Woods Hole,
MA 02543, USA. (rsohn@whoi.edu)
S. C. Webb, Lamont-Doherty Earth Observatory, 61 Route 9W, PO Box
1000, Palisades, NY 10964-8150, USA. (scw@ldeo.columbia.edu)
B12104 SOHN ET AL.: SEISMIC STRUCTURE OF EAST PACIFIC RISE CRUST
17 of 17
B12104