Professional Documents
Culture Documents
Marchenko Redatuming in A Dissipative Medium: Numerical and Experimental Implementation
Marchenko Redatuming in A Dissipative Medium: Numerical and Experimental Implementation
net/publication/328157697
CITATIONS READS
6 208
5 authors, including:
Some of the authors of this publication are also working on these related projects:
All content following this page was uploaded by Tianci Cui on 14 October 2018.
James E. Rickett
Schlumberger Cambridge Research, United Kingdom
Ivan Vasconcelos
Department of Earth Sciences, Utrecht University, the Netherlands
(Dated: August 8, 2018)
Conventional Marchenko redatuming can retrieve the Green’s function at a virtual receiver inside
a heterogeneous medium from single-sided reflection data. It requires minimal a priori knowledge of
medium properties but assumes the medium to be lossless. The dissipative Marchenko scheme, on
the other hand, can retrieve the Green’s function inside a lossy medium, but requires double-sided
access to the lossy medium in oder to obtain the reflection response of its corresponding effectual
medium with negative dissipation. Here, we implement the dissipative Marchenko scheme both
numerically and experimentally, and successfully redatum full sound wavefields in a 1D waveguide.
To achieve this, we first quantify the medium attenuation using the nearly constant Q model to
facilitate numerical modeling. Then, we numerically retrieve two sets of focusing functions which
are required by the scheme, by iteratively re-emitting the time-windowed and time-reversed signal
from a single side of the dissipative and effectual media. Without any knowledge of the medium
properties, we experimentally retrieve the desired focusing and Green’s functions inside the physical
wave tube from double-sided scattering data measurements. We show that both offline processing
and real-time focusing can achieve similar accuracy. True-amplitude Green’s functions can also be
obtained but require a scaling correction. This is the first time that the dissipative Marchenko
scheme is applied to laboratory data to physically focus a wavefield inside a dissipative medium.
Reconciling numerical modeling and physical experiments is an important step in establishing the
applicability of the Marchenko methods for attenuative media.
R(x0 , x0 , ω)f1−∗ (x0 , xi , ω) = f1+∗ (x0 , xi , ω)−G+ (xi , x0 , ω), A low-noise sound wave tube is manufactured at ETH
(5) Zürich for various 1D acoustic wave propagation experi-
where ∗ denotes complex conjugation. Based on their ments. Fig. 1 shows the wave tube used for experimen-
causality properties, focusing and Green’s functions can tal implementation of the dissipative Marchenko scheme,
be obtained using an iterative procedure [1, 26, 27], al- with the x-axis dimension scaled down by a factor of four
though other approaches such as least-squares inversion for display. The tube is assembled from short cylindrical
are also possible [28]. parts along the axial direction x. The parts are made
To adapt the Marchenko method to include dissipa- by 3D printing using photopolymer VeroWhitePlus. The
tion, Slob [24] introduced a so-called effectual medium, tube is open to the atmosphere and filled with air at
which is similar to the physical medium but with negative room temperature and pressure. Acoustic impedance in
dissipation. In the following, all variables with subscript the 1D waveguide is inversely proportional to the cross-
q representing the dissipative medium and e its effectual sectional area. Hence, impedance contrasts are created
counterpart. Slob [24] derived two sets of Marchenko by varying the tube diameter along the y-axis. The mod-
equations, the first of which reads ular design of this setup enables easy modification of the
tube geometry and flexible access to the pressure field
along the tube. A National Instruments (NI) USB-6216
−
+
Rq (x0 , x0 , ω)f1q (x0 , xi , ω) = f1q (x0 , xi , ω)+G−
q (xi , x0 , ω), DAQ board generates the source signal, which is am-
(6) plified by a stereo audio amplifier (Monacor SA-100) to
3
drive one of the loudspeakers (B&C DE7) generating a where the phase velocity c(ω) and the absorption coef-
transient pressure perturbation within the tube. Three ficient α(ω) are both frequency dependent. The phase
microphones and pre-amplifiers (model PCB Piezotron- velocity determines phase dispersion according to
ics 378B13) are used to probe the pressure wavefield at
locations x0 , xi and xm , respectively. Passing through a
four-channel signal conditioner (model PCB Piezotronics 1 1 1 ω
= 1− ln , (15)
482C15), the signal returns to the NI DAQ board for ac- c(ω) c0 πQ ω0
quisition. We repeat each acquisition 50 times and stack
them to suppress random noise. where c0 is a reference velocity at a reference frequency ω0
The pressure field inside the tube is the sum of the (rad/s) or ν0 (Hz). The absorption coefficient is positive
fundamental mode and higher modes. If the source is and describes energy loss in terms of
driven at frequencies below the planar cutoff frequency,
higher modes become evanescent and decay exponentially ω
α(ω) = , (16)
[29]. For the cylindrical tube, the planar cutoff frequency 2c0 Q
νc (Hz) is computed according to
where Q is assumed to be independent of frequency over
the range of measurement. Inserting Eqs. (14), (15) and
1.841c (16) to Eq. (13), we obtain the impulse response of the
νc = , (12)
πd nearly constant Q wave propagation in a 1D homoge-
where c is the speed of sound in the wave tube and d is neous medium
the diameter of the tube cross section. With c ≈ 343 m/s
and d = 2 cm, νc ≈ 10 kHz; if d = 4 cm, νc ≈ 5 kHz. xω i ω 1
During our experiment, a Ricker wavelet with a dominant Iq (x, ω) = exp − i− ln + . (17)
c0 πQ ω0 2Q
frequency of 1 kHz is chosen to drive the loudspeaker so
that the pressure wave propagating inside the tube is Choosing ν0 to be the dominant frequency 1 kHz, we
approximately planar. estimate Q and c0 by inverting the recorded waveforms
The sound wave experiences significant attenuation in Fig. 2. The choice of this reference frequency will be
when propagating in the tube. The exact cause of at- discussed later. By means of a grid search, we scan the
tenuation has not been identified, but it is postulated to values of Q at intervals of 2 from 20 to 140 and c0 at
be from energy conversion to heat and scattering from intervals of 0.1 m/s from 339 m/s to 349 m/s. For every
imperfect coupling of every two adjacent tube parts. To pair of c0 and Q values being searched, we filter the first
study the attenuation effects, we record the direct ar- waveform using Eq. (17) and calculate the L2 -norm misfit
rivals at various source-receiver offsets after emitting a between the filtered waveform and a recorded waveform
1 kHz Ricker wavelet into a tube of 2 cm diameter. To at a later time. The optimal solution of Q and c0 should
mitigate the transfer-function effects of the loudspeaker make the forward-Q filtered first waveform fit the later
and the microphone, an inverse transfer function is pre- waveform in the least-squares sense. We first conduct
calculated and convolved with the source wavelet so that individual inversion, meaning that the misfit function is
the recorded waveforms are approximate to the shape of calculated for each waveform pair individually. Fig. 3
the Ricker wavelet. Blue traces in Fig. 2 are the recorded displays the misfit functions of fitting the first waveform
arrivals, whose amplitudes decay with the increasing off- to the third, fifth, seventh and ninth individually. They
set x. Attenuation is commonly characterized by the all look similar and converge to similar Q and c0 . Their
quality factor Q. For moderate loss, Q can be defined as standard deviations from inverting all the waveform pairs
the maximum energy stored during a cycle, divided by are correspondingly 2.9 and 0.66 m/s, implying reliable
the energy lost during the cycle [30]. To facilitate numer- inversion results. In order to get a consistent estimate,
ical modeling of the dissipation effects, we parameterize we then conduct simultaneous inversion, that is, a total
the attenuation by fitting the waveform data with the misfit function is calculated by fitting the first waveform
nearly constant Q model [30, 31]. We define the impulse to all the other eight waveforms simultaneously. Fig. 4(a)
response of a 1D plane wave propagating to the distance shows the total misfit function which converges to Q = 56
x as and c0 = 343.8 m/s. The first waveform filtered using the
inverted values of Q and c0 to all larger offsets are plotted
as dashed red traces in Fig. 2. They are seen to fit the
I(x, ω) = exp − iK(ω)x , (13) recorded waveforms very well.
Whereas in the above analysis we choose 1 kHz as the
where K(ω) is the propagation constant and it becomes reference frequency, in theory, any frequency within the
complex valued in the presence of attenuation range of measurement can be used to invert for the ref-
erence velocity at this frequency. Figs. 4(b) and 4(c)
ω are the misfit functions of simultaneous inversion using
Kq (ω) = − iα(ω), (14)
c(ω) a low (ν0 = 0.1 kHz) and a high (the Nyquist frequency
4
ν0 = 62.5 kHz) reference frequency, respectively. In con- is distant enough from the physical tube boundaries x00
trast to Fig. 4(a) where Q and c0 are constrained inde- and x0m so as to push boundary reflections to arrive at
pendently, there are trade-offs between them caused by later times. Besides the boundaries, there are three re-
the relatively low signal-to-noise ratio at low and high flectors above and one reflector below the desired focus-
frequencies. Substituting each pair of the Q and c0 esti- ing datum xi . Fig. 6b is the truncated tube model, from
mates from simultaneous inversion at the low, dominant which benchmark focusing functions can be obtained.
or high frequency into Eq. (15), we can predict a corre- We first implement the dissipative Marchenko scheme
sponding velocity dispersion relation. As Fig. 4(d) shows, by numerical modeling to illustrate the physics intu-
inversion using these three different reference frequencies itively. Impedance contrasts are modeled by varying the
yields almost the same Q value and velocity dispersion medium density according to Fig. 6(c). We use a con-
relation from this low-noise experiment. stant velocity of c0 = 343.8 m/s at ν0 = 1 kHz and
We conclude that the attenuation effects observed from an invariant Q of 56, which does not fully represent the
the 1D waveguide can be well captured by the nearly con- medium properties of the variable-diameter wave tube,
stant Q model within the experimental frequency range. but serves as a numerical reference for the experimental
More experiments show that the attenuation parameters implementation in the next section. We model acoustic
vary with the dominant signal frequency, diameter of the wave propagation in a 1D heterogeneous medium incor-
tube cross section, coupling condition of tube parts, 3D- porating the nearly constant Q attenuation using layer
printing material, etc. Note that we parameterize the matrices in the frequency domain [32, 33]. This model-
dissipation for the purpose of modeling wave propagation ing method is also adapted to model effectual wave prop-
in the dissipative and effectual media only. The dissipa- agation using Eq. (19). We adopt the complex frequency
tive Marchenko scheme is purely data driven, requiring technique [34, 35] to avoid aliasing in the time domain
no knowledge of the medium properties including the at- especially for wave propagation in the effectual medium.
tenuation parameters. Assuming that the truncated medium is known,
we model its reflection RA (x0 , x0 ) and transmission
TA (xi , x0 ) responses to calculate the benchmark focus-
IV. NUMERICAL IMPLEMENTATION ing functions by definition (Eqs. (2) and (3)) for dis-
sipative and effectual cases, respectively. Fig. 7 shows
Based on the nearly constant Q model, the effectual the benchmark focusing functions with td (x0 , xi ) denot-
medium is modeled such that its propagation constant is ing the traveltime of the direct wave transmitting from xi
the complex conjugate of the corresponding one in the to x0 . In both dissipative and effectual cases, the events
dissipative medium: have different amplitudes but the same traveltimes. They
obey the same causality properties as the lossless focus-
ω ing functions: f1+ is composed of a direct arrival f1d +
at
Ke (ω) = + iα(ω). (18) +
−td and a following coda f1m arriving between −td and
c(ω)
td ; f1− arrives between −td and td . Separated in time
Inserting Eqs. (15), (16) and (18) to Eq. (13), we model from the Green’s functions G± (xi , x0 ) arriving at and af-
the impulse response of 1D effectual wave propagation ter td , the focusing functions can be retrieved without
according to knowledge of the truncated medium.
With single-sided access to both the dissipative and
xω i ω 1 effectual full media at x0 , we retrieve the focusing func-
Ie (x, ω) = exp − i− ln − , (19) tions by iteratively re-emitting a time-reversed and time-
c0 πQ ω0 2Q windowed signal into these media. Fig. 8 shows the it-
which is the same as Eq. (17) except for the last minus erations of solving the first set of Marchenko equations
sign determining energy gain. Fig. 5 shows the modeled (6) and (7). We update the dissipative focusing functions
+ −
waveforms of propagating a 1 kHz Ricker wavelet to var- f1q (blue traces) and f1q (red traces) by iteratively inject-
ious traveltimes in a 1D homogeneous medium which is ing them into the dissipative medium (left column) and
lossless, dissipative and effectual, respectively. Instead of then the effectual medium (right column) individually.
remaining a constant waveform in the lossless medium, Numerical absorbing boundary conditions are applied at
the amplitude of the wavelet decays gradually in the the two tube ends x00 and x0m , and wavefields between
dissipative medium whereas it increases in the effectual x0 and xm are displayed only. The overall amplitude of
medium. Similar phase dispersion can be observed from the wavefields decays when propagating in the dissipative
both dissipative and effectual media that the later side medium in contrast with that in the effectual medium
lobe of the waveform is progressively larger than the ear- where wave propagation gains energy. Suppose we know
+
lier side lobe. the direct arrival of f1q at the time −td (blue trace in
The dissipative Marchenko scheme is implemented Fig. 8(a)), which is the correct initial focusing function
+ +
both numerically and experimentally on a 6.8-m-long f1qd . We first inject f1qd into the dissipative medium and
wave tube shown schematically in Fig. 6(a). The tube record its reflection response (red trace in Fig. 8(a)). We
section between x0 and xm is our target of interest, which then mute all the events arriving at and after td , reverse
5
FIG. 1. A top view of the sound wave tube setup with the x-axis dimension scaled down by a factor of four for display: the
tube section whose diameter is (a-1) 3 cm and (a-2) 4 cm (others are 2 cm), (b) the NI DAQ board, (c) the audio amplifier, (d)
the two loudspeakers, the three microphones and pre-amplifiers at (e-1) x0 , (e-2) xi and (e-3) xm , (f) the signal conditional.
FIG. 2. Direct arrivals recorded at various offsets (solid blue), FIG. 3. Misfit functions of individual inversion choosing ν0 =
and the first waveform being forward-Q filtered to all larger 1 kHz. The titles indicate which waveform pairs from Fig. 2
offsets using the optimal values of Q and c0 at ν0 = 1 kHz are used and the inverted Q and c0 values. The red star
from simultaneous inversion (dashed red). indicates the minimum of each misfit function.
FIG. 5. Modeling of propagating a 1 kHz Ricker wavelet to FIG. 7. The benchmark focusing functions calculated by def-
various traveltimes (indicated by vertical dotted black lines) inition for the numerical implementation: (a) the downgoing
in a 1D homogeneous medium which is (a) lossless, (b) dissi- and (b) upgoing focusing functions in the dissipative (blue)
pative and (c) effectual, assuming Q = 56 and c0 = 343.8 m/s and effectual (red) media. The vertical dotted black lines
at ν0 = 1 kHz. indicate traveltimes td and −td .
FIG. 6. The schematics of the wave tube: (a) the full medium,
(b) the truncated medium, (c) the medium density imple-
mented by numerical modeling. The horizontal dotted black
lines indicate the physical tube ends (x00 and x0m ), the bound-
aries of the target section (x0 and xm ) and the desired fo-
cusing datum (xi ). Solid red stars and triangles denote the
sources and receivers in the experimental implementation.
±
FIG. 8. Illustration of iteratively retrieving f1q . Two columns
with reverse polarity. See Refs. [36, 37] for details of in-
of the panels represent wavefield propagation in the dissipa-
terpreting Marchenko focusing wavefields in a 1D lossless tive medium (left) and the effectual medium (right). Three
medium. rows of the panels correspond to the first, second and fourth
iterations from top to bottom. Within each panel, the traces
Fig. 9 shows the iterations of solving the second set of being injected (blue in the left column, red in the right col-
Marchenko equations (8) and (9). The plots and inter- umn) and recorded (red in the left column, blue in the right
pretations are shown in the same way as Fig. 8 except column) at x0 are plotted atop the corresponding time-space
for interchanging the dissipative medium and the effec- wave propagation diagram. The vertical dotted black lines de-
tual medium. Starting from the correct initial focusing note the traveltimes −td and td . The horizontal dotted lines
+ +
function f1ed , which is the direct arrival of f1e at −td , denote specific depth levels: white are the four reflectors, and
+ blue are x0 , xi and xm from top to bottom. The hollow red
we iteratively update the effectual focusing functions f1e
−
(blue traces) and f1e (red traces) to retrieve the Green’s triangles indicate the virtual receivers.
functions G− +
e (xi , x0 ) and Gq (xi , x0 ).
7
+ −
FIG. 12. Iterations of retrieving f1q by offline processing and FIG. 15. Iterations of retrieving f1e by offline processing and
real-time focusing compared to the normalized benchmark real-time focusing compared to the normalized benchmark
+ −
f1q . f1q .
+
FIG. 14. Iterations of retrieving f1e by offline processing and FIG. 17. Green’s function retrieval: (a) upgoing and (b)
real-time focusing compared to the normalized benchmark downgoing dissipative Green’s functions retrieved up to over-
+
f1q . all scaling errors, (c) total dissipative Green’s functions after
scaling correction compared to the benchmark.
10
cal wavefield focusing is observed at the desired location sponse for the lossless Marchenko scheme. In both cases,
by injecting the retrieved downgoing dissipative focusing whether the Marchenko method is more practical than
function into the wave tube. With no need to charac- the existing Q estimation and compensation methods is
terize medium properties within the wave tube, up- and an open question. Even if we have double-sided access,
downgoing dissipative Green’s functions are retrieved in- we foresee problems existing in the lossless Marchenko
dividually. A scaling correction is applied so that their scheme would also challenge its extensions to dissipa-
weighted sum is verified with the physically recorded to- tive media, e.g., sampling issues, limitations in complex
tal Green’s function. media and etc. We believe the dissipative Marchenko
Reconciling numerical modeling and physical experi- scheme can also benefit from ongoing research of account-
ments is an important step in establishing the appli- ing for source wavelet estimation, free surface multiples
cability of the Marchenko methods for attenuative me- and elastic media in the lossless Marchenko scheme for
dia. Derivations of the Marchenko equations are based more practical applications.
on time reversal invariance, a property shared by the
acoustic wave equation, the electromagnetic wave equa-
tion and the Schrödinger equation [38]. We speculate
the dissipative Marchenko scheme is also applicable to VII. ACKNOWLEDGMENTS
dissipative electromagnetic waves and dissipative quan-
tum systems. Although we implement the dissipative
Marchenko scheme in 1D, the theory was established in We thank the European Union’s Horizon 2020 research
3D. With the requirement of double-sided access, the dis- and innovation programme for funding the WAVES
sipative Marchenko scheme can be applied to medical project under the Marie Sklodowska-Curie grant agree-
imaging, nondestructive testing and laboratory tests of ment No. 641943. We are grateful to Kees Wapenaar
rocks, whereas applications to seismic reflection explo- (Delft University of Technology), Patrick Elison, Miguel
ration are not straightforward. A reasonable Q model Molerón (ETH Zürich) for discussions. Reviews from
may be needed, either to compute the effectual reflec- Colin Thompson and Ali Özbek (Schlumberger Cam-
tion response for the dissipative Marchenko scheme, or bridge Research), and two anonymous reviewers are much
to compensate for the dissipation in the reflection re- appreciated.
[1] K. Wapenaar, F. Broggini, E. Slob, and R. Snieder, using Marchenko autofocusing and seismic interferome-
“Three-dimensional single-sided Marchenko inverse scat- try,” Geophysics 80, A7–A11 (2015).
tering, data-driven focusing, Green’s function retrieval, [10] C. A. da Costa Filho, G. A. Meles, and A. Curtis,
and their mutual relations,” Physical Review Letters “Elastic internal multiple analysis and attenuation using
110, 084301 (2013). Marchenko and interferometric methods,” Geophysics
[2] V. A. Marchenko, “Reconstruction of the potential en- 82, Q1–Q12 (2017).
ergy from the phases of the scattered waves,” Doklady [11] G. A. Meles, K. Wapenaar, and A. Curtis, “Recon-
Akademii Nauk SSSR 104, 695698 (1955). structing the primary reflections in seismic data by
[3] K. Wapenaar and J. Thorbecke, “Review paper: Virtual Marchenko redatuming and convolutional interferome-
sources and their responses, Part I: time-reversal acous- try,” Geophysics 81, Q15–Q26 (2016).
tics and seismic interferometry,” Geophysical Prospect- [12] G. A. Meles, C. A. da Costa Filho, and A. Cur-
ing 65, 1411–1429 (2017). tis, “Synthesising singly-scattered waves (primaries) from
[4] F. Broggini and R. Snieder, “Connection of scattering multiply-scattered data,” 79th Annual International
principles: A visual and mathematical tour,” European Conference and Exhibition, EAGE, Expanded Abstracts
Journal of Physics 33, 593–613 (2012). (2017).
[5] J. Behura, K. Wapenaar, and R. Snieder, “Autofocus [13] F. Broggini and J. O. A. Robertsson, “FD injection utiliz-
Imaging: Image reconstruction based on inverse scatter- ing the wavefields generated by Marchenko redatuming:
ing theory,” Geophysics 79, A19–A26 (2014). A target-oriented approach,” 84th Annual International
[6] K. Wapenaar, J. Thorbecke, J. van der Neut, F. Brog- Meeting, SEG, Expanded Abstracts (2014).
gini, E. Slob, and R. Snieder, “Marchenko imaging,” [14] P. Elison, D.-J. van Manen, F. Broggini, and J. O. A.
Geophysics 79, WA39–WA57 (2014). Robertsson, “Using a Marchenko-redatumed reflection
[7] S. Singh, R. Snieder, J. Behura, J. van der Neut, response as an exact boundary condition,” 78th An-
K. Wapenaar, and E. Slob, “Marchenko imaging: Imag- nual International Conference and Exhibition, EAGE,
ing with primaries, internal multiples, and free-surface Expanded Abstracts (2016).
multiples,” Geophysics 80, S165–S174 (2015). [15] M. Ravasi, I. Vasconcelos, A. Kritski, A. Curtis, C. A.
[8] K. Wapenaar, J. van der Neut, and E. Slob, “On the da Costa Filho, and G. A. Meles, “Target-oriented
role of multiples in Marchenko imaging,” Geophysics 82, Marchenko imaging of a North Sea field,” Geophysical
A1–A5 (2017). Journal International 205, 99–104 (2016).
[9] G. A. Meles, K. Löer, M. Ravasi, A. Curtis, and C. A. [16] I. Vasconcelos, M. Ravasi, A. Kritski, J. van der Neut,
da Costa Filho, “Internal multiple prediction and removal and T. Cui, “Local, reservoir-only reflection and trans-
11
mission responses by target-enclosing extended imaging,” [35] S. Mallick and L. N. Frazer, “Practical aspects of reflec-
87th Annual International Meeting, SEG, Expanded Ab- tivity modeling,” Geophysics 52, 1355–1364 (1987).
stracts (2017). [36] T. Cui, I. Vasconcelos, D.-J. van Manen, and K. Wape-
[17] J. van der Neut, M. Ravasi, Y. Liu, and I. Vasconcelos, naar, “A tour of Marchenko redatuming: Focusing the
“Target-enclosed seismic imaging,” Geophysics 82, Q53– subsurface wavefield,” The Leading Edge 37, 67a1–67a6
Q66 (2017). (2018).
[18] C. A. da Costa Filho, G. A. Meles, A. Curtis, M. Ravasi, [37] K. Wapenaar, J. Thorbecke, J. van der Neut, E. Slob,
and A. Kritski, “Imaging strategies using focusing func- and R. Snieder, “Review paper: Virtual sources and
tions with applications to a North Sea field,” Geophysical their responses, Part II: data-driven single-sided focus-
Journal International 213, 561–573 (2018). ing,” Geophysical Prospecting 65, 1430–1451 (2017).
[19] I. Vasconcelos, M. Ravasi, and J. van der Neut, [38] R. Snieder, K. Wapenaar, and U. Wegler, “Unified
“Subsurface-domain, interferometric objective functions Green’s function retrieval by cross-correlation; connec-
for target-oriented waveform inversion,” Geophysics 82, tion with energy principles,” Physical Review E 75,
A37–A41 (2017). 036103 (2007).
[20] J. van der Neut and K. Wapenaar, “Adaptive overburden [39] J. Brackenhoff, “Rescaling of incorrect source strength
elimination with the multidimensional Marchenko equa- using Marchenko redatuming,” M.Sc. thesis, Delft Uni-
tion,” Geophysics 81, T265–T284 (2016). versity of Technology (2016).
[21] K. Wapenaar, J. van der Neut, and E. Slob, “Uni- [40] C. A. da Costa Filho, M. Ravasi, and A. Curtis, “Elas-
fied double- and single-sided homogeneous Green’s func- tic p- and s-wave autofocus imaging with primaries and
tion representations,” Proceedings of the Royal Society internal multiples,” Geophysics 80, S187–S202 (2015).
A 472, 20160162 (2016). [41] C. Fleury and I. Vasconcelos, “Imaging condition for non-
[22] S. Singh and R. Snieder, “Source-receiver Marchenko linear scattering-based imaging: Estimate of power loss
redatuming: Obtaining virtual receivers and virtual in scattering,” Geophysics 77, S1–S18 (2012).
sources in the subsurface,” Geophysics 82, Q13–Q21 [42] E. Dokter, G. A. Meles, A. Curtis, and K. Wapenaar,
(2017). “Velocity analysis using surface-seismic primaries-only
[23] T. S. Becker, P. Elison, D.-J. van Manen, C. Donahue, data obtained without removing multiples,” 79th An-
S. Greenhalgh, F. Broggini, and J. O. A. Robertsson, nual International Conference and Exhibition, EAGE,
“Experimental Marchenko focusing in a variable diame- Expanded Abstracts (2017).
ter sound wave tube,” 86th Annual International Meet- [43] G. F. Margrave, “Using well logs to estimate the effect of
ing, SEG, Expanded Abstracts (2016). fine layering on Q estimation,” 85th Annual International
[24] E. Slob, “Green’s function retrieval and Marchenko imag- Meeting, SEG, Expanded Abstracts (2015).
ing in a dissipative acoustic medium,” Physical Review [44] C. Mildner, F. Broggini, J. O. A. Robertsson, D.-J. van
Letters 116, 164301 (2016). Manen, and S. Greenhalgh, “Target-oriented velocity
[25] E. Slob, K. Wapenaar, F. Broggini, and R. Snieder, analysis using Marchenko-redatumed data,” Geophysics
“Seismic reflector imaging using internal multiples with 82, R75–R86 (2017).
Marchenko-type equations,” Geophysics 79, S63–S76 [45] M. Ravasi, “Rayleigh-Marchenko redatuming for target-
(2014). oriented, true-amplitude imaging,” Geophysics 82, S439–
[26] J. H. Rose, “‘Single-sided’ autofocusing of sound in lay- S452 (2017).
ered materials,” Inverse Problems 18, 1923–1934 (2002). [46] J. W. S. B. Rayleigh, “The theory of sound (vol. 2),”
[27] J. van der Neut, I. Vasconcelos, and K. Wapenaar, “On Macmillan (1896).
Green’s function retrieval by iterative substitution of the [47] T. Rossing, “Springer handbook of acoustics,” Springer
coupled Marchenko equations,” Geophysical Journal In- (2014).
ternational 203, 792–813 (2015). [48] S. Singh, R. Snieder, J. van der Neut, J. Thorbecke,
[28] J. van der Neut, J. Thorbecke, K. Wapenaar, and E. Slob, and K. Wapenaar, “Accounting for free-surface
E. Slob, “Inversion of the multidimensional Marchenko multiples in Marchenko imaging,” Geophysics 82, R19–
equation,” 77th Annual International Conference and Ex- R30 (2017).
hibition, EAGE, Expanded Abstracts (2015). [49] E. Slob and K. Wapenaar, “Theory for Marchenko imag-
[29] M. Redwood, “Mechanical waveguides: The propagation ing of marine seismic data with free surface multiple elim-
of acoustic and ultrasonic waves in fluids and solids with ination,” 79th Annual International Conference and Ex-
boundaries,” Pergamon Press (1960). hibition, EAGE, Expanded Abstracts (2017).
[30] E. Kjartansson, “Constant Q wave propagation and at- [50] D.-J. van Manen, J. O. A. Robertsson, and A. Curtis,
tenuation,” Journal of Geophysical Research: Solid Earth “Exact wave field simulation for finite-volume scatter-
84, 4737–4748 (1979). ing problems,” The Journal of the Acoustical Society of
[31] W. I. Futterman, “Dispersive body waves,” Journal of America 122, EL115–EL121 (2007).
Geophysical Research 67, 5279–5291 (1962). [51] I. Vasconcelos, M. Ravasi, and J. van der Neut, “An
[32] D. C. Ganley, “A method for calculating synthetic seis- interferometry-based, subsurface-domain objective func-
mograms which include the effects of absorption and dis- tion for targeted waveform inversion,” 76th Annual Inter-
persion,” Geophysics 46, 1100–1107 (1981). national Conference and Exhibition, EAGE, Expanded
[33] G. F. Margrave and P. F. Daley, “VSP modeling in 1D Abstracts (2014).
with Q and buried source,” CREWES Research Reports [52] I. Vasconcelos, K. Wapenaar, J. van der Neut, C. Thom-
26 (2014). son, and M. Ravasi, “Using inverse transmission matri-
[34] M. Bouchon and K. Aki, “Discrete wave-number repre- ces for Marchenko redatuming in highly complex media,”
sentation of seismic-source wave fields,” Bulletin of the 85th Annual International Meeting, SEG, Expanded Ab-
Seismological Society of America 67, 259–277 (1977). stracts (2015).
12
[53] I. Vasconcelos and J. van der Neut, “Full-wavefield reda- WCC1–WCC26 (2009).
tuming of perturbed fields with the Marchenko Method,” [55] K. Wapenaar, J. van der Neut, E. Ruigrok, D. Draganov,
78th Annual International Conference and Exhibition, J. Hunziker, E. Slob, J. Thorbecke, and R. Snieder,
EAGE, Expanded Abstracts (2016). “Seismic interferometry by crosscorrelation and by mul-
[54] J. Virieux and S. Operto, “An overview of full-waveform tidimensional deconvolution: A systematic compari-
inversion in exploration geophysics,” Geophysics 74, son,” Geophysical Journal International 185, 1335–1364
(2011).