SEGGEO_714_SI1-SI4.

qxd 7/13/06 3:43 PM Page SI1

Volume 71 July-August 2006 Number 4

GEOPHYSICS

Introduction to the supplement on seismic interferometry

In 1968, J. F. Claerbout derived a remarkable relation be- interferometric-imaging method by an elegant theory based
tween the transmission and reflection responses of a horizon- on stationary phase analysis (Schuster, 2001; Schuster et al.,
tally layered medium, bounded by a free surface (Claerbout, 2004).
1968). He showed that autocorrelation of the transmission re- Schuster’s coworkers at the University of Utah, notably
sponse is equal to the reflection response plus its time-reversed Jianhua Yu and Jiaming Sheng, successfully applied his
version (plus an impulse at time zero). method to various types of data, including shot records, VSP
This implies that when one measures the response of a data, and drill-bit data.
plane-wave source in the subsurface by a geophone at the free In the meantime, the Delft Applied Geophysics group de-
surface, the reflection response is obtained simply by taking veloped a theory based on seismic reciprocity, which formally
the causal part of the autocorrelation of the observed response. generalizes Claerbout’s relation between transmission and re-
Primary as well as multiple reflections are recovered correctly flection responses to 3D inhomogeneous acoustic and elastic
by that procedure. The source wavelet in the recovered re- media (Wapenaar et al., 2002). Draganov et al. (2003) con-
flection response is equal to autocorrelation of the source sig- firmed this theory with numerically modeled data in laterally
nal in the subsurface. varying media.
Hence, if one would measure the response of a band-lim- Mathias Fink at University of Paris VII pioneered an initial
ited white-noise source in the subsurface, the autocorrelation independent field of research in the early 1990s, making use
would give the impulsive reflection response, convolved with of the invariance of the wave equation to time reversal.
a band-limited delta function. This shows that noise observed Through various physical-modeling tests using ultrasonic
at the surface can be turned into a signal with information about transducers, Fink’s group showed that strongly scattered wave-
the subsurface. The principle of using passive-noise mea- fields could be time-reversed and back-propagated through
surements to derive the reflection response and subsequently the complex medium to result in a focused wavefield (Fink,
form an image of the earth’s interior was called acoustic day- 1997). Researchers around the world in different disciplines
light imaging. were amazed at the robustness of the time-reversal process in
Later, Claerbout conjectured that his relation could be gen- cases governed by severe multiple scattering, and this led to
eralized for offset measurements in 3D inhomogeneous me- a renewed interest in the use of the multiply scattered coda
dia, i.e., that by crosscorrelating noise traces recorded at two (Snieder and Scales, 1998; Snieder et al., 2002).
locations on the surface, one can construct the wavefield that Since the beginning of the new millennium, various research
would be recorded at one of the locations as if there were a groups in other fields of science have discovered indepen-
source at the other. Since its conception, several attempts have dently that crosscorrelation of signals at two different receivers
been made to make this idea work on real data, some more in an acoustic diffuse field yields the response at one of the re-
successful than others (Scherbaum, 1987a, 1987b; Cole, 1995; ceiver positions as if there were a source at the other. This work
Daneshvar et al., 1995; Rickett and Claerbout, 1999). The first was pioneered by Weaver and Lobkis (2001, 2002) in ultra-
convincing results have been obtained by solar seismologists sonics and by Campillo and Paul (2003) in seismology. These
(Duvall et al., 1993). and other researchers speak of Green’s function reconstruc-
In the exploration-geophysics community, the research on tion. The assumed diffusivity in their theory can be caused by
retrieving information from crosscorrelations received new multiple scattering among heterogeneities in a disordered
momentum after a sabbatical stay of Gerard T. Schuster at the medium, reverberations in an enclosure with an irregular
Stanford Exploration Project in 2000. He applied the correla- bounding surface, a random distribution of uncorrelated noise
tion method not only to passive data but also to exploration sources, or any combination of those causes. Roux and Fink
seismic data with man-made sources. Schuster introduced the (2003) obtained similar results for underwater acoustics, as-
concept of interferometric imaging, which involves an inte- suming a wavefield consisting of orthogonal modes in a wave-
gration of crosscorrelation and migration. He supported his guide.

1
Department of Geotechnology, Delft University of Technology, Delft, the Netherlands, E-mail c.p.a.wapenaar@tudelft.nl.
2
Department of Geotechnology, Delft University of Technology, Delft, the Netherlands, E-mail d.s.draganov@tudelft.nl.
3
WesternGeco Oslo Technology Centre, Solbraveien 23, 1383 Asker, Norway, E-mail jrobertsson@oslo.westerngeco.slb.com.

SI1
SEGGEO_714_SI1-SI4.qxd 7/13/06 3:43 PM Page SI2

SI2 Seismic interferometry

Since 2003, researchers in different fields have become verge to the precise Green’s function. The authors review re-
aware of one another’s results and started to make links among cent measurements and recent developments in theory and as-
the different approaches. In time-reversed acoustics, Derode cribe the differences to incomplete convergence in some cases
et al. (2003) showed that recording a time-reversed wavefield and to nonfully diffuse fields in other cases.
at a second point other than the original source point yields Larose et al. give an overview of the theoretical founda-
the Green’s function between the two points. tions for passive imaging techniques based on correlation of
Equivalently, the central idea in all cases of interferometry random wavefields. The authors present applications of the
is that the crosscorrelation of signals at two observation points method to ultrasonic nondestructive testing and seismic to-
tends to yield the impulse response between those points, mography.
which explorationists call the reflection response and physi- Gerstoft et al. use crosscorrelations of seismic-noise data
cists and seismologists call the Green’s function. from 151 stations in southern California to extract group ve-
The main differences between the various approaches are locities of surface waves between station pairs for determin-
found in the underlying assumptions: Whereas explorationists ing surface-wave velocity structure.
consider deterministic media, the main underlying assump-
tion in the theory of Weaver and Lobkis (2001) and others is Green’s function reconstruction:
that the wavefield is diffuse because of any of the causes men- Deterministic wavefields
tioned above. The approaches converge for the situation of un-
correlated sources in a deterministic medium, where uncor- Wapenaar and Fokkema derive exact representations of
related sources should be interpreted in a broad sense, in- Green’s functions between any two points in an arbitrary in-
cluding transient sources sufficiently separated in time. homogeneous medium in terms of crosscorrelations of wave-
Since 2003, the research on retrieving new responses from field observations at those points. These representations form
crosscorrelations has taken an enormous flight, in exploration a theoretical basis for seismic interferometry.
geophysics and in ultrasonics, seismology, and underwater Van Manen et al. propose an efficient and flexible inter-
acoustics. During the SEG annual meeting in 2004, Gerard ferometric modeling scheme for wave propagation in arbitrary
Schuster, who was then the editor of SEG, asked us to com- inhomogeneous elastic media. After systematic illumination
pile a supplement for GEOPHYSICS, dedicated to this emerging of the medium from a surrounding surface, Green’s functions
branch of science. The result is the 21 papers published in this between arbitrary points in the volume can be computed us-
supplement. ing only crosscorrelation and summation.
We decided to name the supplement Seismic Interferome- Draganov et al. use numerical experiments to show how
try. The term interferometry is borrowed from radio astron- to reconstruct the reflection response from crosscorrelation of
omy, in which it refers to crosscorrelation methods applied to the transmission response from deterministic media. The au-
radio signals from distant objects. thors show this for acoustic and elastic media for the case of
To stimulate cross-fertilization among seismic exploration, separate measurements from transient subsurface sources as
ultrasonics, and seismology, we invited researchers from out- well as for the situation of simultaneously acting white-noise
side seismic exploration as well and asked them to explain sources in the subsurface.
their methods and indicate possible applications for applied Fan et al. develop an algorithm to remove free-surface mul-
geophysics. tiples for teleseismic transmission and to construct reflection
Many of the papers in this supplement have been presented responses. This approach integrates the one-way reciprocity
at the workshop titled “Seismic Interferometry, Daylight Imag- and the inverse-scattering series in the teleseismic framework.
ing and Time-Reversal,” organized in connection with the SEG Van Wijk demonstrates a controlled ultrasonic laboratory
annual meeting in 2005.4 experiment that provides detailed analysis of retrieving a band-
We have grouped the papers into three main categories (the limited estimate of the Green’s function between receivers in
same as in the workshop). These are (1) Green’s function re- an elastic medium. Instead of producing a formal derivation,
construction, (2) redatuming, and (3) imaging. Any ordering this article refers to a series of intuitive operations, common
of papers is to some extent subjective and arbitrary. Within to geophysical data processing, to understand the practicality
those categories, where appropriate, we have subdivided into of seismic interferometry.
methods employing diffuse wavefields versus those for which Poletto and Petronio discuss the use of transmitted waves
the medium and wavefield are assumed to be deterministic. using autocorrelogram interferometry techniques with a tun-
We hope you enjoy reading this supplement just as much nel-boring machine (TBM) as a seismic source. The approach,
as we enjoyed preparing it. which offers the advantage of obtaining reflections from the
transmitted (front) wavefield, is used to improve prediction of
Green’s function reconstruction: fractures ahead of the TBM.
Diffuse wavefields Mercier et al. present a novel deconvolution technique to
improve retrieval of the Green’s functions from passive-source
Weaver and Lobkis show that practical passive imaging data, particularly teleseismic data. They successfully demon-
using correlations of diffuse fields is good, but laboratory and strate application of this method to data from three stations of
field measurements sometimes fail to fully or faithfully con- the Canadian National Seismic Network.

4
The presentations of the above-mentioned workshop can be found at http://www.geos.ed.ac.uk/homes/acurtis/SEG2005Interferometry.html.
SEGGEO_714_SI1-SI4.qxd 7/13/06 3:43 PM Page SI3

Seismic interferometry SI3

Redatuming: Deterministic wavefields Xiao et al. describe how VSP interferometric imaging of
transmitted PP and PS waves can delineate the flanks of salt
Schuster and Zhou review the theory underlying reda- bodies. This method does not require the migration-velocity
tuming methods which effectively time-shift traces using nat- model of the salt or upper sediments to image the salt flank.
ural or model-based traveltimes. The model-based traveltimes Berkhout and Verschuur show that multiples tradition-
are computed from an a priori velocity model. The correla- ally have been considered as noise and are discarded after re-
tion-based redatuming methods use natural phase information moval. The authors argue that multiple reflections contain a
in the data to time-shift weighted traces so that they appear to wealth of information that can be used in seismic processing
have been generated by sources (or recorded by geophones) to improve resolution of reservoir images beyond current ca-
shifted in location. The correlation-based redatuming meth- pability.
ods differ in their choice of trace weights and are superior to Shragge et al. detail how shot-profile migration can be tai-
model-based methods because they do not require a velocity lored to image teleseismic wavefield-coda data based on in-
model and they eliminate statics at the source and/or receiver terferometric principles. The authors (1) develop a 2.5D imag-
locations. ing procedure that enables kinematic and structural imaging
Snieder et al. show that correlation of single-reflected using recorded transmission and free-surface reflected passive
waves can be used to recover primary reflections. In addition, wavefields and (2) demonstrate its effectiveness by migration
artifacts that behave as spurious multiples may be introduced. of the IRIS_PASSCAL CASC-1993 data set.
Weglein et al. present several distinct approaches that de-
rive from inverse-scattering series concepts that input mea- References
sured reflection data and predict the wavefield at depth and
the transmitted wavefield. These maps are realizable without Campillo, M., and A. Paul, 2003, Long-range correlations in
the traditional need for subsurface information or phase as- the diffuse seismic coda: Science, 299, 547–549.
sumptions about the reflection data. Claerbout, J. F., 1968, Synthesis of a layered medium from its
Bakulin and Calvert propose a new way of imaging in acoustic transmission response: Geophysics, 33, 264–269.
complex geologic media by placing downhole geophones be- Cole, S., 1995, Passive seismic and drill-bit experiments us-
low the most complicated part of the overburden. By measur- ing 2-D arrays: Ph.D. thesis, Stanford University.
ing transmission responses and applying time-reversal logic, Daneshvar, M. R., C. S. Clay, and M. K. Savage, 1995, Pas-
the surface shots are redatumed to downhole locations with- sive seismic imaging using microearthquakes: Geophysics,
out the overburden velocity model, thus explaining the term 60, 1178–1186.
virtual source method. Derode, A., E. Larose, M. Tanter, J. de Rosny, A. Tourin, M.
Campillo, and M. Fink, 2003, Recovering the Green’s
Imaging: Diffuse wavefields function from field-field correlations in an open scatter-
ing medium (L): Journal of the Acoustical Society of
Fink shows that a time-reversal mirror acts as an antenna America, 113, 6, 2973–2976.
that uses complex environments to appear wider than it is. The Draganov, D., K. Wapenaar, and J. Thorbecke, 2003, Synthe-
author investigates this property for various media. sis of the reflection response from the transmission re-
Borcea et al. describe a coherent interferometric approach sponse in the presence of white noise sources: 65th An-
for imaging in clutter, in which they migrate crosscorrelations nual International Conference and Exhibition, EAGE, Ex-
of the traces over appropriately chosen space-time windows tended Abstracts, P218.
rather than the traces themselves. The space-time windowing
Duvall, T. L., S. M. Jefferies, J. W. Harvey, and M. A. Pomer-
leads to a statistical smoothing of the data and therefore to re-
antz, 1993, Time-distance helioseismology: Nature, 362,
liable results in clutter, at the cost of some blurring, and the
430–432.
optimal trade-off between stability and resolution can be
achieved adaptively during image formation. Fink, M., 1997, Time reversed acoustics: Physics Today, 50,
Artman shows that direct migration of passive seismic- 34–40.
field data can produce appropriate images of the subsurface Rickett, J., and J. Claerbout, 1999, Acoustic daylight imaging
even while simultaneously processing wavefields from mul- via spectral factorization: Helioseismology and reservoir
tiple sources without demanding that they be random (uncor- monitoring: The Leading Edge, 18, 957–960.
relable) time series. The author uses synthetic data and a me- Roux, P., and M. Fink, 2003, Green’s function estimation us-
ter-scale experiment to show the efficacy of the approach. ing secondary sources in a shallow water environment:
Journal of the Acoustical Society of America, 113, 3, 1406–
Imaging: Deterministic wavefields 1416.
Scherbaum, F., 1987a, Seismic imaging of the site response
Zhou et al. present a reduced-time migration method and using microearthquake recordings: Part I. Method: Bul-
an interferometric migration method with seminatural Green’s letin of the Seismological Society of America, 77,
functions for seeing beneath salt when the migration velocity 1905–1923.
in the salt and above is not well known. Synthetic and field ———, 1987b, Seismic imaging of the site response using
CDP data tests show that by picking reflection traveltimes from microearthquake recordings: Part II: Application to the
a reference layer beneath salt, both methods can significantly Swabian Jura, southwest Germany, seismic network: Bul-
mitigate kinematic defocusing effects caused by errors in the letin of the Seismological Society of America, 77, 1924–
overburden velocity model. 1944.
SEGGEO_714_SI1-SI4.qxd 7/13/06 3:43 PM Page SI4

SI4 Seismic interferometry

Schuster, G. T., 2001, Theory of daylight/interferometric imag- Wapenaar, K., D. Draganov, J. Thorbecke, and J. Fokkema,
ing: Tutorial: 63rd Annual International Conference and 2002, Theory of acoustic daylight imaging revisited: 72nd
Exhibition, EAGE, Extended Abstracts, A32. Annual International Meeting, SEG, Expanded Abstracts,
Schuster, G. T., J. Yu, J. Sheng, and J. Rickett, 2004, Interfer- 2269–2272.
ometric/daylight seismic imaging: Geophysical Journal Weaver, R. L., and O. I. Lobkis, 2001, Ultrasonics without a
International, 157, 838–852. source: Thermal fluctuation correlations at MHz frequen-
Snieder, R. K., and J. A. Scales, 1998, Time-reversed imaging cies: Physical Review Letters, 87, 134301-1–134301-4.
as a diagnostic of wave and particle chaos: Physical Re- ———, 2002, On the emergence of the Green’s function in
view E, 58, 5668–5675. the correlations of a diffuse field: Pulse-echo using ther-
Snieder, R., A. Grêt, H. Douma, and J. Scales, 2002, Coda mal phonons: Ultrasonics, 40, 435–439.
wave interferometry for estimating nonlinear behavior in
seismic velocity: Science, 295, 2253–2255.