GEOPHYSICS, VOL. 80, NO. 6 (NOVEMBER-DECEMBER 2015); P. B167–B176, 8 FIGS., 2 TABLES.

10.1190/GEO2014-0558.1
Downloaded 10/02/15 to 132.203.227.62. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/

Case History

Ambient seismic noise tomography at Ekofisk

Sjoerd A. L. de Ridder1 and Biondo L. Biondi2

ABSTRACT sources. These sources were dominated by Scholte waves trav-

We have studied an approximately 40-hr recording from the
Life of Field Seismic array installed over Ekofisk field to assess
whether the recorded ambient seismic energy was suitable for
passive seismic surface-wave interferometry. Passive seismic
interferometry aims to retrieve virtual seismic sources by cross-
correlation. We have found that the noise recorded by the pressure
sensors between 0.4 and 1.4 Hz consisted predominantly of
Scholte-wave microseism energy. The noise incidence directions
have an almost uniform distribution over the azimuth, enabling
the synthesis of nearly symmetric estimated Green’s functions
between all stations in the array that formed virtual seismic
eling along the seafloor, and we sought to determine which features
in the subsurface can be imaged by straight-ray group-velocity
tomography. We located a high-velocity anomaly in the center
of Ekofisk’s production-induced subsidence bowl, surrounded
by lower velocities. This pattern seemed to result from produc-
tion-induced seafloor subsidence that altered the near-surface
shear strengths. A dispersion analysis showed that the Scholte-
wave virtual seismic source exhibited an approximate penetration
depth to 600 m below the seafloor. These results are significant
because they demonstrated that recordings made at the ocean-
bottom cable array at Ekofisk field in the absence of seismic shoot-
ing can be used to image and monitor the near surface.

INTRODUCTION A theory named passive seismic interferometry uses so-called

The Ekofisk field in the southern part of the Norwegian North
Sea is one of the first developed oil-producing fields in the North
Sea (Hermansen et al., 1997). Rapid pressure depletion in the early
phase of production and weakening due to subsequent water injec-
tion caused more than 9 m of seafloor subsidence over the Ekofisk
field (Hermansen et al., 1997; Lyngnes et al., 2013). Since 2010,
Ekofisk field has had a Life of Field Seismic (LoFS) 4C optical
sensor array of ocean-bottom cables (OBCs) buried approximately
2 m into the sand of the seafloor (Eriksrud, 2010). The main ob-
jective of this installation is to record production-related time-lapse
surveying during active seismic surveying (Folstad et al., 2010).
Because the array is permanently installed and capable of recording
continuously, we study the potential of imaging using the ambient
seismic noise recorded in the absence of seismic shooting.
noise recordings of the ambient seismic field to turn seismic stations
into virtual seismic sources by crosscorrelation. Aki (1957) first
derives (from a modal formulation) how to retrieve the dispersion
of surface waves from the crosscorrelations of a circle of stations.
Claerbout (1968) concludes that the 1D autocorrelation of transmis-
sion responses would yield the reflection response (using 1D reci-
procity theorems) and conjectures an extension to three dimensions
by crosscorrelations. This was formally proven for the elastody-
namic case by Lobkis and Weaver (2001) and Weaver and Lobkis
(2002) based on normal mode expansions and by Wapenaar (2003,
2004) and Wapenaar and Fokkema (2006) using 3D reciprocity
theorems.
Baskir and Weller (1975), Scherbaum (1987), Cole (1995), and
Daneshvar et al. (1995) are among the first to test Claerbout’s re-
lationship between the reflection response and the autocorrelation

Manuscript received by the Editor 26 November 2014; revised manuscript received 7 May 2015; published online 23 September 2015.
1
Formerly Stanford University, Department of Geophysics, Stanford, California, USA; presently University of Edinburgh, School of GeoSciences, Edinburgh,
UK. E-mail: s.deridder@ed.ac.uk.
2
Stanford University, Department of Geophysics, Stanford, California, USA. E-mail: biondo@stanford.edu.
© 2015 Society of Exploration Geophysicists. All rights reserved.

B167
 B168 De Ridder and Biondi

of the transmission response at an exploration scale. Breakthrough
success came with the application in crustal-scale seismology start-
ing with Campillo and Paul (2003) and Shapiro and Campillo
(2004). These successes led to a resurgence of research in explora-
tion-scale attempts (Artman, 2007; Draganov, 2007; Draganov et al.,
Downloaded 10/02/15 to 132.203.227.62. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/
has yielded images of the near surface using the Scholte-wave group
and phase velocities between 0.5 and 1.75 Hz (De Ridder and Del-
linger, 2011; De Ridder and Biondi, 2013; Mordret et al., 2013a,
2013b, 2013c). The question arises whether the successes at Valhall
can be replicated at other permanent OBC installations.
The studies at Valhall and Astero use energy from the double-

2009; Edme and Halliday, 2011; Ruigrok et al., 2011; Ruigrok and
Wapenaar, 2012). Notably, Stewart (2006) retrieve the Scholte
waves from an ocean-bottom cable (OBC) recording of microseism
noise in the Gulf of Mexico. Other applications are found in solar
physics (Duvall et al., 1993; Rickett and Claerbout, 1999) and in
laboratory acoustics (Weaver and Lobkis, 2001, 2002).
A study of an ocean-bottom-node recording over the Astero field
yielded images of group velocities at frequencies between 0.18 and
0.4 Hz, with correlation to known structures (Bussat and Kugler,
2011). Permanently installed OBC ambient noise over Valhall field
frequency microseism band excited by ocean swell noise (Longuet-
Higgins, 1950). In marine environments, this energy is dominated by
Scholte waves (Olofsson, 2010). Over land, this energy peak is com-
posed of surface and body waves incident from coasts and shallow
seas (Ruigrok et al., 2011). Permanently deployed OBCs could be
used to monitor the near surface for time-lapse changes using ambi-
ent seismic noise (De Ridder et al., 2014; Mordret et al., 2014).
The near surface is an imaging challenge (Butler, 2005). Knowl-
edge of the near surface helps to resolve wavefield imaging challenges
at depth that are caused by near-surface transmission complexities
(Cox et al., 1999). Recently, the applied seismology community has
taken an interest in the use of surface waves for imaging of the near
surface because Scholte waves have proven to be very sensitive to
stresses in the near surface and respond to near-surface geomechan-
ical changes induced by hydrocarbon production at depth (Wills
et al., 2008; Hatchell et al., 2009). Accurate knowledge of the near
surface may also help scientists to better understand geohazards in-
volved with hydrocarbon exploration and production (Landrø and
Amundsen, 2011; Barkved, 2012).
We present a study of almost 40 hr of data recorded by the pres-
sure sensors of the LoFS at Ekofisk field. First, we analyze the noise
characteristics with special attention to the microseism noise. Then,
we crosscorrelate the microseism noise and judge the quality of the
retrieved virtual seismic Scholte-wave sources. Finally, we image
Scholte-wave group-velocities for five central frequency ranges,
from 0.4 to 1.4 Hz, by straight-ray tomography.

Figure 1. Map of station locations in Ekofisk’s LoFS array (coor- AMBIENT SEISMIC FIELD RECORDED BY LoFS
dinates in ED50/universal transverse mercator[UTM] Zone 31N). AT EKOFISK FIELD
Each black dot denotes a station. The stations used to create the
spectrogram in Figure 2 and for the beam steering results in Figure 3 In this section, we analyze the characteristics of the microseism
are shown in blue. noise recorded by the pressure sensors of Eko-
fisk’s LoFS array. Ekofisk’s LoFS array uses
fiber-optic technology (Eriksrud, 2010). The
pressure sensors consist of a fiber wrapped
around a mandrel. The mandrel responds to
pressure changes lengthening the fiber, which
is measured by an interferometer (Amundsen
and Landrø, 2009). The exact response function
was not known to the authors, but the phase re-
sponse cancels by a crosscorrelation. Figure 1
contains a map with the LoFS station locations.
The inline and crossline station spacings are ap-
proximately 50 and 300 m, respectively. The ar-
ray comprises 3966 stations. For this study, we
obtained a data set spanning almost 43.5 hr, start-
ing on 24 October 2011, at Coordinated Univer-
sal Time (UTC) 00:17:00. There are three gaps in
Figure 2. Spectrogram showing spectral amplitudes versus time for the duration of the the recording lasting a combined total of approx-
entire recording used in this study. Dashed lines indicate the frequency regimes in which imately 3.5 hr.
swell noise, microseism noise, and operational noise dominate. Note that the microseism The transient nature of the ambient seismic
energy and swell-noise energy grow stronger during 24 October 2011 and remained strong
during 25 October. Blue and red curves display measurements made by a weather obser- field is observed in the spectrogram shown in
vation station at Ekofisk field (Weather Underground Inc., 2013). The blue curve shows Figure 2. The spectrogram shows how the time-
barometric pressure at sea level (in bar). The red curve shows wind speed (in kmph). frequency spectrum of the recorded seismic
 Seismic noise tomography at Ekofisk
energy changes as a function of the recording time. This spectro-
gram is computed using the Fourier transformation of 2.5-min
recording windows with 50% overlap. The spectrogram was aver-
aged over a subset of 119 stations located in a radius of 750 m of
UTM (514 km east, 6261 km north), depicted as blue stations in
Downloaded 10/02/15 to 132.203.227.62. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/
Figure 1.
Analysis of Figure 2 identifies three frequency regimes. Below
0.15 Hz, the recordings are dominated by ocean swells. These are
wind-generated gravity waves in the sea (Munk, 1950). Between 0.35
and 1.35 Hz, microseism energy dominates the recordings. Micro-
seism noise is caused by constructively interfering ocean swells that
give rise to pressure variations on the seafloor. These pressure var-
iations excite interface waves traveling along the seafloor, Scholte
and Love waves, at approximately twice the frequency of the swell
noise (Longuet-Higgins, 1950). Notice that the microseism noise in-
creases when the ocean swells grow stronger. Above 1.5 Hz, a variety
of predominantly field-operational noise sources dominate the re-
cordings. This study focuses on the microseism noise because
Scholte waves travel along the seafloor and provide subsurface sen-
sitivity to the top few hundred meters immediately below the seafloor.
The spectrogram in Figure 2 reveals that the swell noise and
swell-noise-generated microseism energy grew stronger during
24 October and remained strong on 25 October. We downloaded
observations from a weather station at Ekofisk field (Weather
Underground, Inc. 2013), and we display the recorded barometric
pressure at sea level and wind speeds in Figure 2. During 24 and
25 October, we see a gradual decrease in pressure, indicating the
emergence or arrival of a low-pressure system, which is indicative
of rougher weather conditions. This observation is corroborated by
the increase in wind strengths during the 24th, and wind generally
remains strong during the 25th. This correlation between weather
conditions and microseism energy levels at Ekofisk field is not guar-
anteed because sea-swell-excited microseism energy could be gen-
erated quite far from Ekofisk field.
Beam steering is one technique used to reveal the azimuth and
apparent slowness of energy incident on an array (Cole, 1995; Rost
and Thomas, 2002). Beam steering is carried out using the same 119
stations used for the spectrogram (the blue stations in Figure 1). The
recordings are first filtered for the microseism energy between 0.55
and 0.65 Hz using a Hann taper in the frequency domain. A narrow-
frequency band was needed because velocity dispersion causes
blurring in the beam steering results. The data are then transformed
to the τ‐p domain by a slant stack. The absolute value is smoothed
by a triangle over 10 min, forming a smooth movie of amplitudes
over px and py as a function of τ. Figure 3 contains 15 frames of
the movie formed by averaging 2.5 min at 3-hr intervals. High am-
plitudes (white) indicate the slowness and azimuth of incoming
energy, whereas low amplitudes (black) indicate the absence of en-
ergy. We observe a circle at absolute slowness jsj ≈ 1.9 ms∕m, cor-
responding to a velocity of jsj−1 ≈ 525 m∕s. These are the Scholte
waves, which travel in many directions and compose the micro-
seism noise of the ambient seismic field. Notice that the noise is
remarkably omnidirectional. Some energy at the center of the plots,
especially visible at later times, indicates that there is a small
amount of energy arriving with a very high apparent velocity. One
possible explanation of this energy may be that it represents a small
fraction of body-wave energy recorded by the array. Bussat and
Kugler (2011) observe an acoustically guided wave (P-wave)
frequencies as low as 0.6 Hz, but in much deeper waters. Mordret
B169
et al. (2013a) observe the acoustically guided mode in Valhall OBC
recordings, but at greater than 10 Hz.
PASSIVE SEISMIC INTERFEROMETRY
Passive seismic interferometry refers to the creation of virtual
seismic sources, i.e., retrieval of an estimated Green’s function
(EGF), by crosscorrelating noise recordings made at different sta-
tions. Ekofisk’s LoFS array has 3966 stations that record particle
velocity and pressure approximately 2 m below the seafloor. Here,
crosscorrelations between the pressure sensor recordings made at all
stations were computed. Let Gp̄;q̄ ðxA ; xB ; ωÞ denote the frequency-
domain elastodynamic Green’s function of a composite-receiver re-
cording pressure p̄ at xA due to a composite source representing a
volume injection rate q̄ at xB . An estimate of the Green’s function
Gp̄;q̄ ðxA ; xB ; ωÞ and its reciprocal Gp̄;q̄ ðxB ; xA ; ωÞ can be retrieved
as follows (Wapenaar and Fokkema, 2006):
hp̄ðxA ; ωÞp̄ ðxB ; ωÞi ∝ fGp̄;q̄ ðxA ; xB ; ωÞ
þ Gp̄;q̄ ðxB ; xA ; ωÞgSðωÞ; (1)
where the crosscorrelated signals p̄ðxA ; ωÞ and p̄ðxB ; ωÞ denote the
pressure recordings made at xA and xB (the master station), respec-
tively. Complex conjugation is denoted by * and corresponds to a
time reversal in the time domain, and hi denotes a spatial ensemble
average. The power spectrum of the noise source signals is denoted
by SðωÞ. For a direct noise crosscorrelation to yield the Green’s func-
tion, we require the ambient seismic field to be in a state of energy
equipartitioning (Lobkis and Weaver, 2001). However, that state is
not sufficient, we also require all noise sources to act incoherently
(Snieder et al., 2010).
Under ideal circumstances, the crosscorrelation signal yields an
equivalence of the superposition of the causal Green’s function
and its anticausal reciprocal counterpart, forming a purely symmetric
signal. An estimate for the phase of the Green’s function can be found
by applying the Heaviside step function to the crosscorrelation signal
either before or after symmetrizing (Wapenaar and Fokkema, 2004).
The asymmetric part of the crosscorrelation provides a measure of the
deviation from idealized ambient seismic noise characteristics. The
asymmetric part of the crosscorrelation signal is the difference be-
tween the causal and anticausal Green’s functions, which are re-
trieved from waves traveling in opposite directions between the
two stations (Snieder, 2004; Bensen et al., 2007). As such, the asym-
metry in the crosscorrelation also provides an indication of the noise
source distribution around the station pair.
VIRTUAL SOURCES FROM EKOFISK’S
LoFS ARRAY
To extract the microseism noise and compress the data volume,
the recorded pressure data were first filtered using a frequency-do-
main taper with a flat response for 0.4–1.3 Hz, and a Hann taper
extending from 0.35 to 1.35 Hz. Filtering was done in segments
of 3 min, plus 20 s of overlap, and the data were restitched (by aver-
aging in the overlap portion) after filtering and downsampling. Seg-
ments containing noise bursts and spikes were discarded by tapering
and substituting zeros. All available data were then crosscorrelated
in 10 blocks of 4 hr. For each station pair, the crosscorrelations were
stacked to form a virtual seismic survey with virtual sources at all
 B170 De Ridder and Biondi
stations in Ekofisk’s LoFS array. Figure 4 shows an example of a
virtual seismic source. Figure 4a to 4e contains the symmetric part
of the crosscorrelation signal, ℜfp̄ðxA ; ωÞp̄ ðxB ; ωÞg, whereas Fig-
ure 4f to 4j contains the antisymmetric part ℑfp̄ðxA ; ωÞp̄ ðxB ; ωÞg.
At later time snaps, the lower frequencies precede the arrival of the
higher frequencies due to dispersion.
Downloaded 10/02/15 to 132.203.227.62. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/
There is very little coherent energy in the antisymmetric part, and
the background correlation fluctuations in the antisymmetric part
are on the order of the background correlation fluctuations in the
symmetric part. The beam-steering results indicate that the noise
incidence directions, especially when averaged over a long time,
have uniform distribution over the azimuth. But the lack of (almost)
any antisymmetric part in the EGFs shows there are few (to none)
dominating and coherently acting noise sources. Thus, the micro-
seism noise at Ekofisk field is very suitable for creating Scholte-
wave virtual seismic sources by seismic interferometry. We expect
that crosscorrelating and stacking of more recorded data (had it been
available) would increase the signal-to-noise-ratio.
A dispersion image is formed by taking the amplitude in the Ra-
don domain. We sorted all the EGFs as a function of absolute offset
(Figure 5a) and then performed a slant stack to the τ‐p domain.
Next, the data are transformed from the τ‐p domain to the ω‐p do-
main by Fourier transformation. The amplitudes are balanced over
frequencies for clarity to produce Figure 5b.
In Figure 5a, we observe that the crosscorrelations are dominated by
a dispersive wave arrival interpreted to be an interface wave, specifi-
cally, Scholte waves (Scholte, 1942a, 1942b). The fundamental
Scholte-wave mode is the only mode visible in Figure 5b. It is a dis-
Figure 3. Results from beam steering of data filtered between 0.55 and 0.65 Hz, selecting microseism noise. Each image contains a beam-
steering result from data 3 hr apart and indicates the slowness and azimuth of incoming Scholte-wave energy averaged over 10 min.
persive wave mode; i.e., the wave-speed varies with frequency. A sum-
mary of observed phase velocities as a function of frequency and their
corresponding wavelengths is shown in Table 1. This table tells us that
the microseism energy is not aliased in the inline direction, but it be-
comes aliased in the crossline direction greater than 0.75 Hz.
The Scholte waves emitted by the virtual seismic sources propa-
gate along the seafloor with wavelengths approximately between
1460 m at 0.4 Hz and 340 m at 1.2 Hz (from Figure 5b). These
long wavelengths provide sensitivity away from the seabed,
i.e., in depth. Lower frequencies will be more sensitive to deeper
depths.
STRAIGHT-RAY TOMOGRAPHY FOR
GROUP-VELOCITY MAPS
Surface waves extracted using passive seismic interferometry are
traditionally studied to analyze their phase and group velocity (Bar-
min et al., 2001; Sabra et al., 2005; Shapiro et al., 2005; Gerstoft
et al., 2006; Yao et al., 2006). These studies approximate surface
waves as waves propagating in two dimensions through maps of
apparent velocity (Wielandt, 1993). Here, we will invert the
Scholte-wave seismic sources retrieved in the previous section
by straight-ray group-velocity tomography (Barmin et al., 2001).
To this purpose, we picked group traveltimes of the surface waves
for a series of frequencies, by picking the peak of the envelope of
the symmetrized EGFs after band-pass filtering (Levshin et al.,
1989; Bensen et al., 2007). Our band-passes had a flat response over
an 0.2-Hz interval and extended as Hann tapers over an additional
 Seismic noise tomography at Ekofisk B171

0.2 Hz on either side. These are wider band-passes than the more
conventionally used Gaussian tapers, resulting in measurements addi-
tionally averaged over frequency but more stable. We use the ratio
between the maximum of the envelopewithinanestimated linear move-
out window to the average of the envelope outside the window as an
Downloaded 10/02/15 to 132.203.227.62. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/
ments (Shapiro and Singh, 1999). We anticipate this to be small
given the relatively constant spectrum within the microseism band
(Figure 2).
We pose a regularized straight-ray inverse problem; see, for ex-
ample, Barmin et al. (2001). First, we subtract the contribution of
the average slowness from the data. The average slowness m0 is

signal-to-noise-ratio quality factor. The spectrum of the ambient

noise and the neglected station response function may bias the measure- computed using the traveltime picks t ¼ ðt1 ; t2 ; : : : ; tN ÞT :

a) b) c) d) e)

f) g) h) i) j)

Figure 4. (a-e) Snap shots for symmetric and (f-j) antisymmetric parts of the EGFs. Correlation lags are (a and f) 0 s, (b and g) 4 s, (c and h) 8 s,
(d and i) 12 s, and (e and j) 16 s.

a) b)

Figure 5. (a) Offset gather for crosscorrelations between all station pairs with midpoints within UTM (513–514 km east, 6269–6270 km
north). (b) Dispersion image generated by transforming the gather in (a) to the ω‐p domain and balancing the amplitude over frequencies.
 B172 De Ridder and Biondi

1X N
ti and by 140 (north–south) grid cells, 100 by 100 m wide. The grid
m0 ¼ : (2) cell size is below the resolution of the wavelength. Approximately
N i¼1 Δxi
half the model space is covered by the array, so we only have ap-
proximately 6300 model parameters to constrain. This means that
The contribution of the average slowness to the traveltime measure- the inverse problem is overdetermined and we can discard a signifi-
ments is subtracted from the traveltimes, yielding traveltime resid-
Downloaded 10/02/15 to 132.203.227.62. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/

cant portion of our data.

uals Δt ¼ ðΔt1 ; Δt2 ; : : : ; ΔtN ÞT :
Δt ¼ t − m0 Δx; (3)
where Δx ¼ ðΔx1 ; Δx2 ; : : : ; ΔxN ÞT is a vector containing the off-
sets Δx for each specific traveltime pick. The traveltime residuals
are predicted from a perturbation slowness model Δm by a straight-
ray tomography operator F as
Δt ¼ FΔm: (4)
This operator is simply a slowness integration kernel, in which each
row contains the distances through each model cell of a straight line
between a specific source-receiver couple. A conjugate-direction
algorithm is used to find the minimum of the following misfit
function:
JðΔm; ϵÞ ¼ kFΔm − Δtk22 þ ϵk∇2 Δmk22 ; (5)
2 2 2
where kk2 denotes an L norm. We use the ∇ operator as a regu-
larization to force a smooth model (a model styling objective), also
known as second-order Tikhonov regularization (Aster et al., 2005).
After inversion, the slowness map can be recovered from the per-
turbation map by adding the average slowness m ¼ Δm þ m0 .
There are several strategies to determine regularization strength.
Usually, the regularization strength ϵ is picked as a value at which
we find an optimal trade-off between the data-fitting objective and
the model styling objectives (Aster et al., 2005). Here, we found that
a small regularization strength resulted in a strong acquisition im-
print appearing in the inverted model. We choose a regularization
strength that minimizes this acquisition footprint, and our strategy is
further explained below in the resolution analysis.
The LoFS array has 3966 stations providing for almost 7.9 mil-
lion unique raypaths. The model space is formed by 90 (east–west)
Table 1. Measurements of phase velocity and wavelength as a function of
frequency by picking maxima in Figure 5b.
Frequency (Hz) 0.4 0.5 0.6 0.7 0.8 0.9 1.0 1.1 1.2
Phase velocity (m∕s) 584.1 565.6 525.2 490.2 466.4 448.0 432.5 416.7 407.2
Wavelength (m) 1460 1131 875.3 700.3 583.0 497.8 432.5 378.8 339.3
Table 2. Number of traveltime picks used per inversion and their corresponding
average velocity.
Frequency range (Hz) 0.4–0.6 0.6–0.8 0.8–1.0
Number of traveltime picks 1212778 5528579 5434445 3579255 552305
Average group velocity (m∕s) 378.8 349.4 328.0
To omit the poorer quality data, we select traveltime picks with an
signal-to-noise-ratio of more than two. To avoid station couples for
which the causal and anticausal Green’s functions are not well-
separated, we only selected traveltime picks for interstation offsets
exceeding 1500 m (Bensen et al., 2007). This is smaller than the
ideal three wavelengths, especially for the lowest of frequencies.
But the crosscorrelation quality and limited size of the array do
not allow for a larger exclusion zone. We took two measures to dis-
card outlier data. First, we discard the traveltime perturbations falling
under the 5th and greater than the 95th percentiles. Second, we solve
each tomography twice. The first time with a regularization strength ϵ
three orders of magnitude larger than the regularization strength that
we finally determined to be appropriate. Then, we discard the trav-
eltime picks with misfits greater than the 95th percentile, and we
solve the inverse problem again.
This whole procedure is repeated for five frequency ranges:
0.4–0.6, 0.6–0.8, 0.8–1.0, 1.0–1.2, and 1.2–1.4 Hz. In Table 2,
the number of selected traveltime picks and their corresponding
average velocities m−1 0 are shown for each frequency range.
The upper frequency range, 1.2–1.4 Hz, is pushing it a little over
the range for which we have well-formed Scholte waves, deter-
mined from the dispersion analysis in the previous section. This
is also reflected by the significant decrease in accepted picks for
this frequency range.
Figure 6 contains the five maps found by solving the inverse
problem. We observe a high-velocity anomaly in the center of the
array, surrounded by a lower velocity region. The high-velocity
anomaly coincides with the center of the seafloor subsidence bowl.
Under the southern end of the array, where the magnitude of the
gradient of the seafloor is smaller, we find higher velocities again.
Ekofisk’s LoFS has a dense concentration of stations resulting in
a very good ray coverage and measurements throughout the domain
of imaging (Figure 7a). Figure 6c is the ray coverage for all trav-
eltime picks used for the group-velocity tomog-
raphy between 0.8 and 1.0 Hz.
One measure for resolution is a checkerboard
test (Lévěque et al., 1993). For a known model,
usually a checker pattern, computing the perfect
traveltime data and running the inversion give an
indication of which anomalies could potentially
be recovered given the ray coverage. We attempt
to recover the checkerboard grid pattern shown
in Figure 7b given the raypath coverage and
the regularization strength for the group-velocity
tomography between 0.8 and 1.0 Hz. The differ-
ence between causal and anticausal traveltime
picks, a proxy for the noise in the data, was
added to the computed traveltime in the true
model. The edges of the recovered checkerboard
1.0–1.2 1.2–1.4 grid cells are smoothed by the regularization
(Figure 7c). The squares are not always well re-
311.3 301.2 solved at the edges of the array due to biased azi-
muthal coverage of the traveltime picks.
 Seismic noise tomography at Ekofisk B173

The optimization was run by a conjugate directions solver. Soft-
ware with this aim typically require a forward and adjoint operator
(Claerbout and Fomel, 2014), and our implementation was no ex-
ception. The forward operator projects the model space to the data
space, and the adjoint operator projects the data space to the model
space. This allows us to see the effect of the chosen regularization
Downloaded 10/02/15 to 132.203.227.62. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/
ing experiments on four recordings made during October, Decem-
ber, January, and February at Valhall field (Mordret et al., 2013a; De
Ridder, 2014). Thus, we suspect that this may be representative of
microseism-energy conditions in the North Sea during fall and win-
ter, and potentially year around.
Omnidirectional microseismic noise translates into omnidirec-

parameter by projecting the final data misfit with the adjoint of
the tomographic operator (Figure 7d) (Claerbout and Fomel,
2014). Most of the energy in the residual is an imprint of the acquis-
ition geometry. There is also some energy that corresponds to
geologically reasonable features. However, decreasing the regulariza-
tion strength would make the acquisition imprint appear in the
inverted model.
DISCUSSION
We find that to the first order, microseism noise is equally inci-
dent from all directions during the duration of the recording (Fig-
ure 3). This omnidirectional microseism noise is very favorable for
seismic interferometry (Weaver and Lobkis, 2002; Wapenaar and
Fokkema, 2006). Similar observations are made using beam-steer-
tional virtual seismic sources (Snieder, 2004). The resulting
virtual seismic sources at Ekofisk field, where no balancing
was applied before stacking crosscorrelations from different time
windows, are almost perfectly symmetric. There is almost no
observable antisymmetric part to the crosscorrelation stacks,
and the background correlation fluctuations in the antisymmetric
part of the crosscorrelations are of the same strength as in the sym-
metric part. These background fluctuations will diminish when
crosscorrelating and stacking more ambient seismic recordings
(Bensen et al., 2007), provided they are made under similar
conditions.
Virtual seismic sources are dominated by a single dispersive
wave mode, corresponding to the fundamental Scholte-wave mode.
Scholte waves are well-formed between 0.4 and 1.2 Hz, and they
have wavelengths as short as 340 m at the upper end of the fre-

a) b) c)

d) e)

Figure 6. Straight-ray tomography maps for group velocities with center frequencies of (a) 0.4–0.6 Hz, (b) 0.6–0.8 Hz, (c) 0.8–1.0 Hz, (d) 1.0–
1.2 Hz, and (e) 1.2–1.4 Hz. Blue lines indicate 2-m contour levels of the bathymetry.
 B174 De Ridder and Biondi

quency band and as long as 1460 m at the lower end. Based on
an approximate penetration depth of 40% of the wavelength (Low-
rie, 2007), they should exhibit depth sensitivity to approximately
600 m below the seafloor. Which means that these waves can be
Downloaded 10/02/15 to 132.203.227.62. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/
of interest to image and monitor the near surface over a producing
reservoir. Studies at Valhall have shown that Scholte waves retrieved
from noise correlations are sensitive to small time-lapse changes (De
Ridder et al., 2014; Mordret et al., 2014).

Figure 7. Resolution indicators for the Scholte- a) b)

wave group-velocity images at the central fre-
quency range 0.8–1.0 Hz. (a) Cumulative ray length
through each cell, (b) model checkerboard grid, and
(c) retrieved checkerboard grid using the raypath
coverage at (a). (d) Adjoint of the tomographic op-
erator applied to final data misfit.

c) d)

a) b) c)

Figure 8. Bathymetry at the Ekofisk field. The blue lines indicate 2-m contour levels of the bathymetry, and the LoFS array station locations
are indicated with yellow dots. (a) Bathymetry in meters with respect to sea level. (b) Magnitude of the gradient of the bathymetry. (c) Laplacian
evaluated on the bathymetry.
 Seismic noise tomography at Ekofisk
Group-velocity images generally show a high-velocity anomaly
coinciding with the center of the seafloor subsidence. At the south-
ern end of the array, there is a high-velocity region again. We con-
sidered whether the high anomaly could be caused by the flat
seafloor approximation in the tomography. But a subsidence of
9 m over an approximately 2-km horizontal distance (Figure 8a)
Downloaded 10/02/15 to 132.203.227.62. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/
disproves that a velocity anomaly of more than 20 m∕s cannot
be caused by the flat-earth approximation in the tomography. This
pattern holds over multiple central frequency ranges, which can be
an indication that it is not reflective of lithology but of stress pat-
terns that carry across lithologic boundaries.
In the group-velocity maps (Figure 6e), one can see that the low
velocities form a ring surrounding the seafloor subsidence (Figure 8).
We generally find higher Scholte-wave group velocities within the
subsidence bowl, and lower Scholte-wave group velocities at the
edge of the subsidence bowl. Within the subsidence bowl, the Lap-
lacian of the seafloor shape is positive whereas along the edges of
the array it is negative (Figure 8c). These areas correspond to where
the subsidence causes relative compression and relative decompres-
sion of the stress state. This affects the near-surface shear strengths
and thus the Scholte wave velocity. Similar anomalies have been
found at Valhall field from critically reflected P-waves (Hatchell
et al., 2009). Weakening of the shear strength in the radial directions
with respect to the center of the subsidence bowl also causes strong
azimuthal anisotropy in surface-wave phase velocities over Ekofisk
field (De Ridder et al., 2015).
Crosscorrelating longer time recordings should increase the signal-
to-noise-ratio, especially at the upper end of the frequency range. Dif-
ferent noise conditions, potentially under heavy weather, can increase
the signal in the lower and upper ends of the microseism frequency
range. Depending on the exact processing and inversion schemes,
group-velocity images from ambient seismic noise are remarkably
stable from recordings as short as a day (De Ridder and Bi-
ondi, 2013).
CONCLUSIONS
We have shown that the microseism noise in the ambient seismic
field between 0.4 and 1.4 Hz, as recorded by Ekofisk’s LoFS array,
is suitable for retrieval of Scholte waves by seismic interferometry.
The microseism energy at Ekofisk field in this recording is particu-
larly uniformly distributed over the azimuth and thus is ideal for
seismic interferometry. Dispersive virtual seismic sources emitting
Scholte waves between 0.4 and 1.4 Hz are retrieved by passive seis-
mic interferometry. The interface waves should exhibit a depth sen-
sitivity to approximately 600 m below the seafloor. The Scholte-
wave group-velocity tomography locates a high-velocity anomaly
in the center of the array, surrounded by a lower velocity region.
The high-velocity anomaly coincides with the center of the
seafloor subsidence bowl. The ring of lower velocities corresponds
with high magnitudes of the bathymetry gradient. Under the
southern end of the array, we find higher velocities again. This
behavior may reflect overburden stress states caused by decades
of production and reservoir depletion.
ACKNOWLEDGMENTS
We thank ConocoPhillips Skandinavia AS and the PL018 Part-
nership (Total E&P Norge AS, ENI Norge AS, Statoil Petroleum
B175
AS, and Petoro AS) for access to Ekofisk’s LoFS data and for per-
mission to publish these results. Many thanks go to O. Knoth, A.
Tura, A. Bertrand, R. Kazinnik, and L. Vedvik of ConocoPhillips
for helpful discussions and suggestions. We thank S. Levin for his
expertise on SEG-D formats and for help with reading the data and
meticulously decoding the header fields. This work was carried out
as part of a Ph.D. program at Stanford University, and S. de Ridder
thanks the sponsors of the Stanford Exploration Project and the
Global Climate and Energy Project at Stanford for financial support.
We acknowledge the two anonymous reviewers and the editors for
constructive comments and critique that substantially improved the
quality of this manuscript.
REFERENCES
Aki, K., 1957, Space and time spectra of stationary stochastic waves, with
special reference to microtremors: Bulletin of the Earthquake Research
Institute, 35, 415–456.
Amundsen, L., and M. Landrø, 2009, Measuring seismic with light: Geo-
Expro, 6, no. 2.
Artman, B., 2007, Passive seismic imaging: Ph.D. thesis, Stanford Univer-
sity.
Aster, R. C., B. Borchers, and C. H. Thurber, 2005, Parameter estimation
and inverse problems: Academic Press.
Barkved, O. I., 2012, Seismic surveillance for reservoir delivery from a prac-
titioner’s point of view: EAGE Publications bv.
Barmin, M. P., M. H. Ritzwoller, and A. L. Levshin, 2001, A fast and reliable
method for surface wave tomography: Pure and Applied Geophysics, 158,
1351–1375, doi: 10.1007/PL00001225.
Baskir, E., and C. Weller, 1975, Sourceless reflection seismic exploration:
Geophysics, 40, 158–159.
Bensen, G. D., M. H. Ritzwoller, M. P. Barmin, A. L. Levshin, F. Lin, M. P.
Moschetti, N. M. Shapiro, and Y. Yang, 2007, Processing seismic ambient
noise data to obtain reliable broad-band surface wave dispersion measure-
ments: Geophysical Journal International, 169, 1239–1260, doi: 10.1111/j
.1365-246X.2007.03374.x.
Bussat, S., and S. Kugler, 2011, Offshore ambient-noise surface-wave
tomography above 0.1 Hz and its applications: The Leading Edge, 30,
514–524, doi: 10.1190/1.3589107.
Butler, D. K., ed., 2005, Near-surface geophysics: SEG.
Campillo, M., and A. Paul, 2003, Long-range correlations in the diffuse seis-
mic coda: Science, 299, 547–549, doi: 10.1126/science.1078551.
Claerbout, J., and S. Fomel, 2014, Geophysical image estimation by exam-
ple: Lulu.com.
Claerbout, J. F., 1968, Synthesis of a layered medium from its acoustic trans-
mission response: Geophysics, 33, 264–269, doi: 10.1190/1.1439927.
Cole, S. P., 1995, Passive seismic and drill-bit experiments using 2-D arrays:
Ph.D. thesis, Stanford University.
Cox, M. J. G., E. F. Scherrer, and R. Chen, 1999, Static corrections for seis-
mic reflection surveys: SEG.
Daneshvar, M. R., C. S. Clay, and M. K. Savage, 1995, Passive seismic im-
aging using microearthquakes: Geophysics, 60, 1178–1186, doi: 10.1190/
1.1443846.
De Ridder, S., and J. Dellinger, 2011, Ambient seismic noise eikonal tomog-
raphy for near-surface imaging at Valhall: The Leading Edge, 30, 506–
512, doi: 10.1190/1.3589108.
De Ridder, S. A. L., 2014, Passive seismic surface-wave interferometry for
reservoir-scale imaging: Ph.D. thesis, Stanford University.
De Ridder, S. A. L., and B. L. Biondi, 2013, Daily reservoir-scale subsurface
monitoring using ambient seismic noise: Geophysical Research Letters,
40, 2969–2974, doi: 10.1002/grl.50594.
De Ridder, S. A. L., B. L. Biondi, and R. G. Clapp, 2014, Time-lapse seismic
noise correlation tomography at Valhall: Geophysical Research Letters,
41, 6116–6122, doi: 10.1002/2014GL061156.
De Ridder, S. A. L., B. L. Biondi, and D. Nichols, 2015, Elliptical-aniso-
tropic eikonal phase velocity tomography: Geophysical Research Letters,
42, 758–764, doi: 10.1002/2014GL062805.
Draganov, D., 2007, Seismic and electromagnetic interferometry — Retriev-
ing the earth’s reflection response using crosscorrelation: Ph.D. thesis,
Delft University of Technology.
Draganov, D., X. Campman, J. Thorbecke, A. Verdel, and K. Wapenaar,
2009, Reflection images from ambient seismic noise: Geophysics, 74,
no. 5, A63–A67, doi: 10.1190/1.3193529.
Duvall, T. L. Jr., S. M. Jefferies, J. W. Harvey, and M. A. Pomerantz, 1993,
Time-distance helioseismology: Nature, 362, 430–432, doi: 10.1038/
362430a0.
 B176 De Ridder and Biondi
Edme, P., and D. Halliday, 2011, Extracting reflectivity response from point-
receiver ambient noise: 81st Annual International Meeting, SEG, Ex-
panded Abstracts, 1602–1607.
Eriksrud, M., 2010, Towards the optical seismic era in reservoir monitoring:
First Break, 28, 105–111.
Folstad, P. G., L. Amundsen, and M. Landrø, 2010, Monitoring of the Eko-
fisk Field: GeoExpro, 7, 72–74.
Gerstoft, P., K. G. Sabra, P. Roux, W. A. Kuperman, and M. C. Fehler, 2006,
Downloaded 10/02/15 to 132.203.227.62. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/
Green’s functions extraction and surface-wave tomography from micro-
seisms in southern California: Geophysics, 71, no. 4, SI23–SI31, doi: 10
.1190/1.2210607.
Hatchell, P. J., P. B. Wills, and C. Didraga, 2009, Production induced effects
on near-surface wave velocities at Valhall: 71st Annual International
Conference and Exhibition, EAGE, Extended Abstracts, T016.
Hermansen, H., L. K. Thomas, J. E. Sylte, and B. T. Aasbøe, 1997, Twenty
five years of Ekofisk reservoir management: SPE Annual Technical Con-
ference and Exhibition, 38927.
Landrø, M., and L. Amundsen, 2011, Using water layer normal modes to
detect shallow gas and CO2 leakage: EAGE Workshop on Permanent
Reservoir Monitoring (PRM), Extended Abstract, 10341.
Lévčque, J.-J., L. Rivera, and G. Wittlinger, 1993, On the use of the checker-
board test to assess the resolution of tomographic inversions: Geophysical
Journal International, 115, 313–318, doi: 10.1111/j.1365-246X.1993
.tb05605.x.
Levshin, A. L., T. B. Yanovskaya, A. V. Lander, B. Bukchin, M. P. Barmin,
L. I. Ratnikova, and E. N. Its, 1989, Seismic surface waves in a laterally
inhomogeneous earth: Kluwer.
Lobkis, O. I., and R. L. Weaver, 2001, On the emergence of the Green’s
function in the correlations of a diffuse field: Journal of the Acoustical
Society of America, 110, 3011–3017, doi: 10.1121/1.1417528.
Longuet-Higgins, M. S., 1950, A theory of the origin of microseisms: Philo-
sophical Transactions of the Royal Society of London A, 243, 1–35, doi:
10.1098/rsta.1950.0012.
Lowrie, W., 2007, Fundamentals of geophysics 2nd ed.: Cambridge Univer-
sity Press.
Lyngnes, B., H. Landa, K. Ringen, and N. Haller, 2013, Life of Field Seis-
mic at Ekofisk — Utilizing 4D seismic for evaluating well target: 75th
Conference and Exhibition, EAGE, Extended Abstracts, We 12 09.
Mordret, A., M. Landès, N. M. Shapiro, S. C. Singh, P. Roux, and O. I.
Barkved, 2013a, Near-surface study at the Valhall oil field from ambient
noise surface wave tomography: Geophysical Journal International, 193,
1627–1643, doi: 10.1093/gji/ggt061.
Mordret, A., N. M. Shapiro, and S. Singh, 2014, Seismic noise-based time-
lapse monitoring of the Valhall overburden: Geophysical Research Let-
ters, 41, 4945–4952, doi: 10.1002/2014GL060602.
Mordret, A., N. Shapiro, S. Singh, P. Roux, and O. I. Barkved, 2013b, Helm-
holtz tomography of ambient noise surface wave data to estimate Scholte
wave phase velocity at Valhall Life of the Field: Geophysics, 78, no. 2,
WA99–WA109, doi: 10.1190/geo2012-0303.1.
Mordret, A., N. Shapiro, S. Singh, P. Roux, J.-P. Montagner, and O. I.
Barkved, 2013c, Azimuthal anisotropy at Valhall: The Helmholtz equa-
tion approach: Geophysical Research Letters, 40, 2636–2641, doi: 10
.1002/grl.50447.
Munk, W. H., 1950, Origin and generation of waves: Council of Wave
Research: Proceedings of the First Conference on Coastal Engineering,
1–4.
Olofsson, B., 2010, Marine ambient seismic noise in the frequency range 1–
10 Hz: The Leading Edge, 29, 418–435, doi: 10.1190/1.3378306.
Rickett, J., and J. Claerbout, 1999, Acoustic daylight imaging via spectral
factorization: Helioseismology and reservoir monitoring: The Leading
Edge, 18, 957–960, doi: 10.1190/1.1438420.
Rost, S., and C. Thomas, 2002, Array seismology: Methods and applica-
tions: Reviews of Geophysics, 40, 1008, doi: 10.1029/2000RG000100.
Ruigrok, E., X. Campman, and K. Wapenaar, 2011, Extraction of P-wave
reflections from microseisms: Comptes Rendus Geoscience, 343, 512–
525, doi: 10.1016/j.crte.2011.02.006.
Ruigrok, E., and K. Wapenaar, 2012, Global-phase seismic interferometry
unveils P-wave reflectivity below the Himalayas and Tibet: Geophysical
Research Letters, 39, L11303.
Sabra, K. G., P. Gerstoft, P. Roux, W. A. Kuperman, and M. C. Fehler, 2005,
Surface wave tomography from microseisms in Southern California: Geo-
physical Research Letters, 32, L14311, doi: 10.1029/2005GL023155.
Scherbaum, F., 1987, Levinson inversion of earthquake geometry SH-trans-
mission seismograms in the presence of noise: Geophysical Prospecting,
35, 787–802, doi: 10.1111/j.1365-2478.1987.tb02258.x.
Scholte, J. G., 1942a, On the Stoneley wave equation I: Proceedings of the
Koninklijke Nederlandse Akademie van Wetenschappen, 45, 20–25.
Scholte, J. G., 1942b, On the Stoneley wave equation II: Proceedings of the
Koninklijke Nederlandse Akademie van Wetenschappen, 45, 159–164.
Shapiro, N. M., and M. Campillo, 2004, Emergence of broadband Rayleigh
waves from correlations of the ambient seismic noise: Geophysical Re-
search Letters, 31, L07614, doi: 10.1029/2004GL019491.
Shapiro, N. M., M. Campillo, L. Stehly, and M. H. Ritzwoller, 2005, High-
resolution surface-wave tomography from ambient seismic noise: Sci-
ence, 307, 1615–1618, doi: 10.1126/science.1108339.
Shapiro, N. M., and S. K. Singh, 1999, A systematic error in estimating
surface-wave group-velocity dispersion curves and a procedure for its cor-
rection: Bulletin of the Seismological Society of America, 89, 1138–1142.
Snieder, R., 2004, Extracting the Green’s function from the correlation of
coda waves: A derivation based on stationary phase: Physics Review E,
69, 046610, doi: 10.1103/PhysRevE.69.046610.
Snieder, R., Y. Fan, E. Slob, and K. Wapenaar, 2010, Equipartitioning is not
sufficient for Green’s function extraction: Earthquake Science, 23, 403–
415, doi: 10.1007/s11589-010-0739-1.
Stewart, P., 2006, Interferometric imaging of ocean bottom noise: 76th An-
nual International Meeting, SEG, Expanded Abstracts, 1555–1559.
Wapenaar, K., 2003, Synthesis of an inhomogeneous medium from its acous-
tic transmission response: Geophysics, 68, 1756–1759, doi: 10.1190/1
.1620649.
Wapenaar, K., 2004, Retrieving the elastodynamic Green’s function of an
arbitrary inhomogeneous medium by cross correlation: Physics Review
Letters, 93, 254301, doi: 10.2528/PIER09041004.
Wapenaar, K., and J. Fokkema, 2004, Reciprocity theorems for diffusion,
flow, and waves: Journal of Applied Mechanics, 71, 145–150, doi: 10
.1115/1.1636792.
Wapenaar, K., and J. Fokkema, 2006, Green’s function representations for
seismic interferometry: Geophysics, 71, no. 4, SI33–SI46, doi: 10.1190/1
.2213955.
Weather Underground, Inc., 2013, The weather channel companies: http://
www.wunderground.com/, accessed 15 October 2013.
Weaver, R. L., and O. I. Lobkis, 2001, Ultrasonics without a source: Thermal
fluctuation correlations at MHz frequencies: Physics Review Letters, 87,
134301, doi: 10.1103/PhysRevLett.87.134301.
Weaver, R. L., and O. I. Lobkis, 2002, On the emergence of the Green’s func-
tion in the correlations of a diffuse field: Pulse-echo using thermal phonons:
Ultrasonics, 40, 435–439, doi: 10.1016/S0041-624X(02)00156-7.
Wielandt, E., 1993, Propagation and structural interpretation of non-plane
waves: Geophysical Journal International, 113, 45–53, doi: 10.1111/j
.1365-246X.1993.tb02527.x.
Wills, P. B., P. J. Hatchell, and S. J. Bourne, 2008, Time-lapse measurements
of shallow horizontal wave velocity over a compacting field: 70th Interna-
tional Conference and Exhibition, EAGE, Extended Abstracts, G039.
Yao, H., R. D. Van Der Hilst, and M. V. De Hoop, 2006, Surface-wave array
tomography in SE Tibet from ambient seismic noise and two-station
analysis — I. Phase velocity maps: Geophysical Journal International,
166, 732–744, doi: 10.1111/j.1365-246X.2006.03028.x.