GEOPHYSICS, VOL. 74, NO. 6 NOVEMBER-DECEMBER 2009; P. V123V132, 12 FIGS., 1 TABLE.

10.1190/1.3245216

Five-dimensional interpolation: Recovering from acquisition constraints

Daniel Trad1

ference of data and thus is sensitive to irregular and coarse sampling

ABSTRACT Abma et al., 2007. Analysis of amplitude variations with offset and
azimuth AVO, AVAz, which we want to observe in the migrated
Although 3D seismic data are being acquired in larger vol- domain, are also affected by the presence of gaps and undersam-
umes than ever before, the spatial sampling of these volumes pling.
is not always adequate for certain seismic processes. This is There are many different approaches to tackling this problem. The
especially true of marine and land wide-azimuth acquisi- only perfect solution is to acquire well-sampled data; all other ap-
tions, leading to the development of multidimensional data proaches deal with the symptoms of the problem rather than the
interpolation techniques. Simultaneous interpolation in all problem itself, and there is no guarantee that they can adequately
five seismic data dimensions inline, crossline, offset, azi- solve it. However, given that, in the real world, we usually cannot go
muth, and frequency has great utility in predicting missing back to the field and fix the actual problem, we need to address this
data with correct amplitude and phase variations. Although issue using the processing tools at our disposal.
there are many techniques that can be implemented in five di- Most seismic algorithms implicitly apply some sort of interpola-
mensions, this study focused on sparse Fourier reconstruc- tion because they assume correctly sampled data. Typically, missing
tion. The success of Fourier interpolation methods depends samples are assumed to be zero or similar to neighboring values. The
largely on two factors: 1 having efficient Fourier transform advantage of using a separate interpolation algorithm is that more in-
operators that permit the use of large multidimensional data telligent assumptions can be made by using a priori information. For
windows and 2 constraining the spatial spectrum along di- example, sinc interpolation uses the constraint that there is no energy
mensions where seismic amplitudes change slowly so that at frequencies above Nyquist. This is more reasonable than assum-
the sparseness and band limitation assumptions remain valid. ing that the unrecorded data are zeros. Interpolation algorithms can
Fourier reconstruction can be performed when enforcing a then be viewed as methods to precondition the data with intelligent
sparseness constraint on the 4D spatial spectrum obtained constraints.
from frequency slices of five-dimensional windows. Binning Interpolation of wide-azimuth land data presents many challeng-
spatial positions into a fine 4D grid facilitates the use of the es, some quite different from those of interpolating narrow-azimuth
FFT, which helps on the convergence of the inversion algo- marine data sets. The most familiar interpolation algorithms have
rithm. This improves the results and computational efficien- been developed for marine streamer surveys. Marine data are usual-
cy. The 5D interpolation can successfully interpolate sparse ly well sampled in the inline direction and coarsely sampled in the
data, improve AVO analysis, and reduce migration artifacts. crossline direction. Many algorithms based on Fourier interpolation
Target geometries for optimal interpolation and regulariza- are quite successful at infilling the crossline direction, even in the
tion of land data can be classified in terms of whether they presence of aliasing and complex structure Schonewille et al., 2003;
preserve the original data and whether they are designed to Xu et al., 2005; Abma and Kabir, 2006; Poole and Herrmann, 2007;
achieve surface or subsurface consistency. Zwartjes and Sacchi, 2007. Land data interpolation brings addition-
al complications because of noise, topography, and the wide-azi-
muth nature of the data. In particular, the azimuth distribution re-
INTRODUCTION quires interpolation to use information from all spatial dimensions at
the same time because sampling along any particular subset of the
All current 3D seismic acquisition geometries have poor sam- four spatial dimensions is usually very poor.
pling along at least one dimension. This affects migration quality, Multidimensional interpolation algorithms have become feasible
which is based on the principle of constructive and destructive inter- even for five dimensions Trad et al., 2005. This capability raises

Manuscript received by the Editor 12 August 2008; revised manuscript received 29 May 2009; published online 25 November 2009.
1
CGGVeritas, Calgary, Alberta, Canada. E-mail: [email protected].
2009 Society of Exploration Geophysicists. All rights reserved.

V123

Downloaded 29 Jan 2010 to 80.194.194.190. Redistribution subject to SEG license or copyright; see Terms of Use at http://segdl.org/
V124 Trad

new possibilities but also brings new challenges and questions. The A related distinction is the number of dimensions that the algo-
general principle is the same: Missing data are assumed to have a rithm can handle simultaneously. Usually, the time dimension is well
similar nature to data recorded in their neighborhood, but the term sampled, so only spatial dimensions need be interpolated. Although
neighborhood can have different meanings in multiple dimen- 3D seismic data have four spatial dimensions, many traditional
sions. An additional complication for wide-azimuth data interpola- methods use data along one spatial dimension only. If the method is
tion in five dimensions is that these data are always very irregular cascaded through the different dimensions, the order of these opera-
and sparse in at least two of the four spatial dimensions because of tions becomes extremely important. However, interpolation of
acquisition and processing costs. sparse wide-azimuth data is more likely to succeed in a full 5D space
Interpolation implementations have two different aspects: the because often at every point there is at least one spatial direction
general interpolation strategy choice of spatial dimensions, win- along which seismic amplitudes change slowly. Information along
dow size, and target geometry and the mathematical engine used to this direction helps to constrain the problem along the other dimen-
predict the new traces from some kind of model. A discussion of sions where events are harder to predict.
these two aspects follows. Also, seismic amplitude variations are smoother in five dimen-
sions than they are in any projection into a lower dimensional space.
To see why, consider an analogy: Imagine the shadow of an airplane
INTERPOLATION STRATEGIES flying over a mountain range. The shadow of the airplane is a com-
plex path even if the airplane goes in a simple trajectory. Interpola-
Interpolation methods differ in complexity, assumptions, and op-
tion of the airplane flight path is much more difficult on the 2D sur-
erator size. Local methods e.g., short-length prediction filters use
face shadow than in the original 3D space. A similar argument can
simple models usually linear events to represent the data in small
be made about seismic wavefield variations in the full 5D space.
windows. Therefore, they tend to be robust, fast, adaptable, and easy
to implement. Their shortcoming is an inability to interpolate large My approach to interpolation is to work with large operators in 5D
gaps because the local information they need does not exist there windows. In practice, the data window size is often constrained by
are no data around the trace to interpolate. the processing system capabilities, particularly when using clusters
Global methods use all of the data simultaneously up to some ap- in busy computing networks. I normally apply windows of 30 30
erture limit defined by the physics of the problem and models with lines, 1000-m offsets, and all azimuths. Larger windows are occa-
many degrees of freedom because they cannot assume simple data sionally required to deal with very sparse data. The spatial dimen-
events at a large scale. They are slower, less adaptable, and harder to sions in these windows are chosen so that the data look as simple as
implement. However, they can, at least in theory, interpolate large possible along each dimension. After extensive testing in different
gaps by using information supplied from distant data. Most practical domains shot, receiver, cross spreads, and common-offset vector
methods fall between these two extremes; but the sparser the sam- domains, I have chosen the inline-crossline-azimuth-offset-fre-
pling, the larger the operator size needs to be. If the geology is com- quency i.e., midpoint, offset, and azimuth with NMO-corrected
plex, some methods with a large operator can smear geologic fea- data for the following reasons:
tures and decrease resolution. A safe choice is to work with global in-
terpolation methods that behave like local interpolators when local 1 These are the dimensions where amplitude variations are most
information is available. important structure, AVO, and AVAz. Interpolation is always
an approximation of the truth, and that approximation is better
a) c) along the dimensions where the algorithm is applied.
Receiver trace number
400 500 600 700
2 AVO and AVAz are usually slow after NMO; therefore, data
Receiver trace number
400 500 600 700 Receiver line have limited bandwidth in the Fourier spectra along these di-
0 mensions. The azimuth dimension also has the advantage of be-
ing cyclic in nature, making it particularly fit for discrete Fouri-
Time (s)

er transform representation.
1 3 The interval between samples in the inline crossline dimen-
sions i.e., midpoints is on the order of the common-midpoint
CMP bin size. In the shot or receiver domain, the sampling
2
can be as coarse as shot/receiver line sampling several CMP
bins.
b) d)
Figure 1 shows a simple synthetic experiment to demonstrate the
advantage of 5D interpolation over 3D interpolation. The original
Time (s)

traces from an orthogonal survey were replaced by synthetic seismic

1 events while preserving the original recording geometry of the trac-
es. The distance between receiver lines was 500 m, with 12 receiver
lines per shot. Every second receiver line of this synthetic data set
2
was removed, simulating a 1000-m line interval, and then predicted
with Fourier reconstruction.
Figure 1. Synthetic data: comparison of 5D versus 3D interpolation. In the first case, I interpolate on a shot-by-shot basis three dimen-
a Synthetic data, one shot window. b After removing every sec-
ond line. c Interpolation in 5D inline/crossline/offset/azimuth/fre- sions, and in the second case in the inline-crossline-offset-azimuth-
quency. d Interpolation in 3D receiver. x, receiver. y, frequency. frequency domain five dimensions. It is evident in Figure 1 that the
Traces are sorted according to receiver numbers. algorithm can reproduce all data complexity when using five inter-

Downloaded 29 Jan 2010 to 80.194.194.190. Redistribution subject to SEG license or copyright; see Terms of Use at http://segdl.org/
 Five-dimensional interpolation V125

polation dimensions, but it is unable to repeat this using only three receivers following the original design types 1a and 1b has the ad-
dimensions. Because the algorithm is exactly the same, this example vantage that original data can be preserved and interpolation is well
shows the importance of the additional information supplied by the constrained. Preserving the original data is generally safer than re-
extra dimensions for Fourier interpolation. placing all of the acquisition with interpolated data, particularly for
The actual location of the newly created traces is an important is- complex noisy data from structured areas in the presence of topogra-
sue for interpolation. I can distinguish six cases, of which only four phy. This approach works well for Kirchhoff time and depth migra-
are used for land data wide-azimuth surveys: tion. By adding new shots and receivers, the subsurface sampling
can be improved according to well-understood acquisition concepts
1 Preserving original data interpolation e.g., Cordsen et al., 2000.
Type 2a, surface-consistent interpolation with perfectly regular
a Decrease shot and receiver interval decrease bin size. shot and receiver lines, is useful for wave equation migration, inter-
b Decrease shot and receiver line interval increase offset polation of very irregular surveys, and time-lapse applications. Type
and azimuth sampling. 2b, subsurface-consistent uniform coverage of offsets and azimuths
c Make shot and receiver line interval equal to shot and re- for each CMP, is desirable for migration in general. However, this
ceiver interval fully sampled. This is a particular case of design implies a large number of shots and receivers with nonuni-
1b. form shot and receiver fold. This is a problem for ray-tracing meth-
ods and any kind of shot or receiver processing. Therefore, its appli-
2 Replacing data totally with predicted traces regularization
cation seems to be limited to time migration and, because of the large
a Target geometry regular on shot and receiver locations size of the resulting data sets, for small surveys. Probably, it can also
surface consistency. be applied well to common-offset Gaussian beam migration.
b Target geometry regular on CMP, offset, and azimuth sub- Finally, types 1c and 2c, complete coverage of shots and receiv-
surface consistency. ers, are desirable for all seismic processing, but the resulting large
c Target geometry regular on surface and in subsurface. size of the data makes it impractical.
Any of these interpolation types can be used for infilling acquisi-
Possibilities 1a, 1b, 2a, and 2b each have important applications tion gaps. A modification of type 2b from polar to Cartesian coordi-
see Table 1. Adapting to the acquired data by adding new shots and nates can be used to produce common-offset vector gathers. Types

Table 1. Types of land data interpolation and main benefits. The size of the circle is proportional to the real use in production
(based on use from 2005 to 2008). The font style in the bottom row reflects a positive (bold) or negative (italic) remark.

Main application Type

Heavy use Increasing Increasing

Occasional use inline- offset- Regularizing Regularizing
crossline azimuth shot/receiver CMP/offset/
Possible, but never sampling sampling positions azimuth positions Full sampling
used 1a 1b 2a 2b 1c and 2c
Interpolation of large gaps
Time Kirchhoff migration
Depth Kirchhoff migration
Wave-equation migration
Merging surveys with
different bin size and/or
design 2D and 3D, parallel
and orthogonal, etc.
Increase resolution for
steep dips relax antialias
filters during migration
Improve CIGs AVO,
AVAz, velocity analysis
4D applications matching
time-lapse surveys
Main use Merging Time/depth Wave Time migration Not used
Kirchhoff equation small surveys because of
migration migration high cost
Main advantage- Reliable Reliable Good Great sampling Best
disadvantage Sometimes Difficult for sampling for Expensive possible
produces topography any for shot-receiver sampling
time slice migration processing Expensive
artifacts Less reliable for all
processing

Downloaded 29 Jan 2010 to 80.194.194.190. Redistribution subject to SEG license or copyright; see Terms of Use at http://segdl.org/
V126 Trad

1c, 2b, and 2c are fully implemented and have been used in internal spatial dimension. They link the frequency slices, making the fre-
tests but have not yet been used in production projects. Notice that quency axis behave as the fifth interpolation dimension, although
one case missing in the table is to replace all data with predictions frequencies are not really interpolated.
onto a given geometry. This is the situation in 4D time lapse, where it The model m is in the -x domain x is a vector representing all
is usual to interpolate the monitor locations to match the baseline. spatial directions. If kmax is the maximum wavenumber on each di-
For typical wide-azimuth land data surveys in a complex environ- mension for the maximum dip of the data, then the case of pk 1 for
ment, the safest choice seems to be surface-consistent interpolation k kmax and pk 0 for k kmax corresponds to sinc interpolation.
1a and 1b. This allows one to preserve the original data untouched The variable in equation 1 is a hyperparameter that controls the
and to apply careful quality control QC to the new traces. Some QC balance between fitting the data and enforcing sparseness on the
parameters can be added to the headers, making it possible to discard spectrum. This parameter is eliminated by changing the cost func-
new traces with high risk or low confidence after the interpolation. tion in equation 1 to the standard form and using the residuals to de-
There are several possible quality parameters. Two QC parameters fine the number of iterations Trad et al., 2003. The actual geophysi-
that complement each other and are often useful are 1 the distance cal meaning of the spatial dimensions is irrelevant to the algorithm.
along the four spatial dimensions between the new and the original However, for the method to work well, at least one of these dimen-
traces and 2 the ratio of original to interpolated traces. sions must have a sparse spectrum or a band limited spectrum.
For the 5D configuration discussed in this paper, a meaningful cal- The multidimensional spectrum can be calculated using discrete
culation of the first parameter requires a weighted average of the dis- Fourier transforms DFTs that exactly honor space locations or the
tance along inline, crossline, offset, and azimuth. The weights de- fast Fourier transforms FFTs that require binning the data into a
pend on the structural complexity, residual moveout, and anisotropy. grid with exactly one trace per position. In practice, I define m to be a
The second parameter refers to the ratio of the number of original regular supersampled 4D grid that contains many more traces than
traces to the number of sampling points on the 4D spatial grid used in the target geometry. This allows us to use FFTs but forces us to bin
the numerical algorithm. This ratio is usually much smaller than the the data during the interpolation.
ratio of input to output traces for a given area. The bin intervals along the spatial dimensions are kept small to
avoid smearing and data distortion. The binning errors along the in-
line/crossline directions can be made negligible by subdividing
INTERPOLATION ENGINE CMP bins into subbins if necessary, but CMP grid bin size usually is
The second major component of the interpolation problem is the adequate. The binning errors along offset and azimuth dimensions
choice of a mathematical algorithm to predict new information giv- are kept small by applying NMO and static corrections before inter-
en a set of recorded traces. One method with the flexibility to adapt to polation. However, data with significant residual NMO and strong
the requirements for multidimensional global interpolation is mini- anisotropy require small bin intervals along offset and azimuth.
mum weighted norm interpolation MWNI Liu and Sacchi, 2004, Large bins reduce computation time and improve numerical stability
which extends the work from Sacchi and Ulrych 1996 to multiple but reduce precision. There is a trade-off between precision and nu-
dimensions. MWNI is a constrained inversion algorithm. The actual merical stability that requires careful parameterization and imple-
data d are the result of a sampling matrix T acting on an unknown mentation. A good rule of thumb for land surveys is to use, as offset
1 1
fully sampled data set m m and d are containers for multidimen- bin interval, a fraction of the receiver group interval e.g., 2 or 4 , de-
sional data, and T is a mapping between these two containers. creasing from near to far offsets and geologic complexity. Azimuth
The unknown interpolated data are constrained to have the same intervals are usually chosen in the range between 20 and 45, de-
multidimensional spectrum as the original data. Enforcing this con- creasing with offset and anisotropy.
straint requires a multidimensional Fourier transform, which is the DFTs can also be used for the spectrum with the advantage that
most expensive part of the algorithm. To solve for the unknown data, they do not require binning. The problem with DFTs is computation-
a cost function is defined for every frequency slice and is minimized al cost. For N variables, a 1D FFT requires computation time propor-
using standard optimization techniques. The cost function J is de- tional to N log N, but DFT requires a computation time proportional
fined, frequency by frequency, as to N2. This constraint makes the cost in two spatial dimensions pro-
portional to N4 and four spatial dimensions proportional to N8. Al-
J d Tm2 mW, 1 though numerical tricks such as nonuniform FFTs Duijndam and
Schonewille, 1999 can improve these numbers dramatically, a 4D
where indicates an 2-norm and w indicates an 2-weighted
2
DFT algorithm is quite expensive in terms of computer time and has
norm calculated as been unfeasible for production demands until now. Very recently,
mW mHF1 2 this has become possible Gordon Poole, personal communication,
n pk Fnm. 2
2009, although it demands large computer resources.
In equation 2, Fn is the multidimensional Fourier transform, with n There are many differences between working with FFTs or DFTs.
indicating the number of spatial dimensions of the data, mH the On the negative side, working with FFTs has the potential to distort
transpose conjugate of the model m, and pk the multidimensional data because of the binning. However, binning spatial coordinates is
spectrum of the unknown data. often applied in seismic processing, even by methods that can use
The multidimensional vector pk contains weight factors that give exact spatial coordinates. For example, when working on common-
freedom to the model to be large where it needs to be large. They can offset volumes, a binning along offset is applied. On the positive
be obtained by bootstrapping from the previous temporal frequency side, when working with FFTs, the results improve because the in-
in a manner similar to that done for Radon transforms Herrmann et creased speed of the iterations permits us to obtain a solution close to
al., 2000. These weights are defined in the -k domain, where is the one that would have been obtained after full convergence.
the temporal frequency and k is the wavenumber vector along each Furthermore, the nature of the system of equations solved at every

Downloaded 29 Jan 2010 to 80.194.194.190. Redistribution subject to SEG license or copyright; see Terms of Use at http://segdl.org/
 Five-dimensional interpolation V127

frequency changes, depending on whether we use regular sampling, cally without loss of precision, but T becomes nonorthogonal. In-
irregular sampling, or regular sampling as a result of binning. To un- creasing bin interval does not affect F1 n and decreases nonorthogo-
derstand why, let us incorporate the sparseness constraint into the nality on T but introduces data distortion.
operator by transforming equation 1 from the general form to the Figure 2 illustrates the effects of sampling in the matrix distribu-
standard form Hansen, 1998. By defining a new model u k, tion for the left side of the system of equation 6. Let us consider two
different cases: coarse regular sampling left column and irregular
u k pk1Fnm x, 3 sampling right column. The matrix distribution for these two cases
which is m after transforming to the -k domain and inverse weight- is shown when applying three different methods: coarse binning,
ing with pk, equation 1 becomes true locations, and fine binning.
The first row, Figure 2a and b, shows the structure of the system of
J d TF1
n pku k u k .
2 2
4 equations when coarse binning is used which allows the use of
The weighted norm w now becomes an 2-norm, and the oper- FFTs. The system of equations is quite sparse, with most elements
ator absorbs the spectral weights. This allows us to include the along the main diagonal; therefore, the optimization converges very
sparseness constraint into the operator, i.e., to modify the basis func- quickly. In the decimation case on the left Figure 2a, the secondary
tions of the transformation to include the sparseness constraint Trad peaks produced by operator aliasing are as strong as the nonalias
et al., 2003. The mapping between d and the new model u is now component. In practice, they can be taken care of by filters and boot-
performed by the operator: strapping weights from low to high frequency.
The second row, Figure 2c and d, shows the same for irregular
L TF1
n pk . 5 sampling true spatial locations. The system of equations is fully
populated because the irregularly sampled Fourier transform intro-
Solving this equation for u k requires solving the following system
duces cross-terms between the model elements the basis functions
of equations:
are nonorthogonal and convergence is slower Figure 2c and d. Op-
pkHFnTHTF1
n pk Iu k pk FnT d ,
H H H
6 erator aliasing, on the other hand, becomes less strong Figure 2c.
The third row, Figure 2e and f, shows the same for fine binning. The
where I is the identity matrix and the super index H means conjugate system of equations becomes almost fully populated again, but in
transpose. this case not because of F as before but because of T. The multidi-
Because of the large size of the system of equations in our prob- mensional case is more difficult to visualize, but the same ideas ap-
lem, on the order of 105 equations, the final full solution u k is never ply. In that case, the nonaliased directions help to constrain the solu-
achieved. Instead, an approximate solution is obtained by using an tion and attenuate the effect of aliasing.
iterative algorithm and running only a few iterations. Components of In my experience, if the bin size is not made too small, the large
u k that have a weak mapping through operator L such as low-am- computational advantage of FFT algorithms over DFTs is more ben-
plitude spectral components can be resolved with this limited num- eficial than the consequent increase on nonorthogonality on T. This
ber of iterations only if the system of equation 6 has good conver- is possible when working along spatial dimensions where the data
gence. This convergence improves as the operator L TF1 n p k be- look simple. In this case, the method can preserve localized ampli-
comes closer to orthogonal, i.e., as
tude variations better than inversion using irregularly sampled spa-
LHL I. 7 tial locations because it is possible to iterate more and to obtain a so-

The operator pk is usually a diagonal operator; therefore, conver- A=(FTH TF1)

gence depends mainly on the two operators F1 n and T, which in turn
depend on the spatial axes and the missing samples, respectively. a) 1 b) 1
10 10
The operator F1n maps the 4D spatial wavenumber k to the inter- 0.5 0.5
20 20
polated 4D spatial axis x. If x and k are perfectly regular, then
30 30
FnF1
n I. If, in addition, there are no missing traces, the left side of 10 20 30
0
10 20 30 0
equation 6 is diagonal and the system converges in one iteration. The c) 1 d) 1
wavenumber axes k one axis per spatial dimension can always be 10 10
made regular, but the axes x depend on the input data. Binning the in- 20 0.5 20 0.5
put data makes x regular. 30 0 30 0
The sampling operator T that connects interpolated data to ac- 10 20 30 10 20 30
quired data depends on the missing traces. It is orthogonal when e) 1 f) 1
there are no missing traces. Binning the data without decreasing 20 20
0.5 0.5
sampling interval does not affect T but can introduce data distortion. 40 40
Making bin intervals small to avoid data distortion introduces nonor- 60 0 60 0
20 40 60 20 40 60
thogonality into the system of equations 6, making convergence
more difficult. Figure 2. Matrix distributions for the left side of the system of equa-
As we see from this analysis, there is a trade-off between nonor- tions 6 for irregularly sampled data in two different cases, coarse
thogonality on F1n and T. Moreover, there is a trade-off between
sampling and gaps columns, using coarse binning, true locations,
binning and data distortion on one side and convergence of the sys- and fine binning rows. a Coarse binning on decimated data. b
Coarse binning on data with gaps. c True locations on decimated
tem of equations on the other. Precisely honoring spatial coordinates data. d True locations on data with gaps. e Fine binning on deci-
slows down convergence because of the nonorthogonality of F1 n . mated data. f Fine binning on data with gaps. The color represents
Alternatively, F1
n can be made regular by fine binning of x, practi- amplitudes in absolute numbers, with dark blue representing zeros.

Downloaded 29 Jan 2010 to 80.194.194.190. Redistribution subject to SEG license or copyright; see Terms of Use at http://segdl.org/
V128 Trad

lution closer to the one obtained with full convergence. This high fi- vantage of the fact that the Fresnel zone increases in size with offset.
delity for localized events makes the algorithm very useful for land A practical combination would be a hybrid method where binning is
data, where amplitude changes very quickly. At the same time, it applied for near and middle offsets and exact locations are applied
makes the method less useful in removing noise. on long offsets.
Pseudorandom noise noise that looks random but has a coherent A complete discussion of the topic is beyond the scope of this pa-
source can be propagated into the new traces, becoming locally co- per. The comments above are intended to point out the effect of sam-
herent and therefore very difficult to remove. Although this is a dis- pling in the system of equation 6 and the impact this has in predicting
advantage, it is important to realize that interpolation and noise at- localized amplitude variations in the data.
tenuation have different and sometimes conflicting goals. Noise at-
tenuation should predict only signal and should filter out noncoher- APPLICATIONS AND DATA EXAMPLES
ent events. Interpolation should predict all of the data, even if the
events are very weak or badly sampled. Undersampled events can Applications for land data interpolation usually involve increas-
look noncoherent; therefore, their preservation depends on the algo- ing inline and crossline sampling decreasing bin size and/or in-
rithm not being too selective in terms of coherence. Although simul- creasing offset and azimuth sampling increasing fold. This classi-
taneous interpolation and noise attenuation is a very desirable goal, fication is too broad, however, because there are many possible ways
better chances of success are achieved by applying noise attenuation to increase the sampling, just as there are many possible geometry
and interpolation iteratively in sequence rather than in a single pass. designs. Table 1 shows several applications classified according to
On the other hand, there are many advantages in using exact posi- the six types defined earlier. All of these cases have been used in
tions to eliminate aliasing and difficulties of binning for complex practice, but only a few of them are often required in production
structure. The first aspect is balanced by working in the full 5D space projects. In this section, we review examples for the most common
of the data. The second can be addressed in most cases by using small cases:
binning intervals.
increasing offset and azimuth sampling by decreasing shot and
Some problems appear often, however, when binning long offsets
receiver line interval to improve migration type 1b
in structured data because of rapid amplitude variations caused by
increasing offset and azimuth sampling for better velocity, AVO,
anisotropy and residual moveout. This is a problem for land and
and AVAz estimates after migration type 1b
ocean-bottom OBC data, where far offsets usually have poor sam-
increasing inline and crossline sampling to improve imaging of
pling because of the rectangular shape of shot patches. Also, this is-
steep reflectors by relaxing antialias filters in migration algo-
sue makes binning more difficult for marine data where residual mo-
rithms type 1a
veout can be very significant at long offsets. A possible solution is to
increasing inline and crossline sampling for changing natural bin
use larger bins along inline and crossline for far offsets, taking ad-
size, as in merging surveys acquired with different geometries
type 1a
a)
infilling missing shots and receivers in acquisition gaps type 1b
in this example

Increasing offset and azimuth sampling for imaging

The first example shows the benefits of interpolation for aniso-
tropic 3D depth migration in a Canadian Foothills area with signifi-
cant structure, topography, and noise. These surveys often can bene-
fit from interpolation because usually they have shot and receiver
lines acquired quite far apart because of high acquisition costs on to-
pographic areas. Foothills acquisitions on structurally complex ar-
eas, however, are difficult to interpolate because small imperfections
b) in static corrections affect coherence in the space domain. Also,
these data often are severely affected by ground roll noise, which
makes interpolation difficult particularly for shallow structures.
Figure 3a shows an orthogonal geometry vertical lines are shots,
horizontal lines are receivers. CMPs are overlain on the shot/receiv-
er locations, with their color indicating the fold. Figure 3b shows the
target geometry that contains all of the same shots and receivers as in
Figure 3a, along with the new interpolated shot and receiver lines.
Notice that these new lines follow the geometry of the original lines,
permitting us to preserve all original data because original shots and
receivers do not need to be moved.
By halving shot and receiver line intervals, the CMP fold increas-
es by a factor of four, giving a better sampling of offsets and azi-
Figure 3. Foothills survey: orthogonal geometry. Shots are located
along vertical lines; receivers are located along horizontal lines. Col- muths. This can be seen in Figure 4a and b, which shows the offset/
or represents fold. a Before interpolation. b After interpolation azimuth distribution for a group of CMPs before and after interpola-
shot and receiver line spacing decreased by a factor of two. tion. The increased sampling benefits migration because imaging al-

Downloaded 29 Jan 2010 to 80.194.194.190. Redistribution subject to SEG license or copyright; see Terms of Use at http://segdl.org/
 Five-dimensional interpolation V129

a) gorithms rely on interference to form the correct image and therefore

require a proper sampling at least two samples per cycle to work
correctly.
These benefits can be observed in the final stacked image, but they
are more obvious in common-image gathers CIGs. Figure 5 com-
pares the CIGs with and without interpolation. The better continuity
of the events will certainly bring improvements to the results of gath-
er processing, especially for AVAz and AVO analysis which are
very sensitive to acquisition footprint and automated processes
such as tomography based on residual curvature analysis. Figure 6
b) shows the image stack from 0 to 1000-m offsets. The continuity of
events has been improved over nearly all of the section.

Increasing offset-azimuth sampling for AVO

Prestack migration of the seismic data prior to performing AVO
inversion has been advocated for more than 10 years Mosher et al.,
1996. However, poor sampling typical of land seismic acquisition
makes practical implementation of this concept quite difficult.
Downton et al. 2008 demonstrate that these problems can be ad-
Figure 4. Foothills survey: offset/azimuth distribution in an area of dressed by performing interpolation prior to prestack migration, re-
the a original and b interpolated surveys. sulting in better AVO estimates. These workers performed a series of
comparisons of processing flows for AVO, including the 5D interpo-
a) o lation method presented in this paper. By calculating the correlation
between AVO estimates to well-log information for the Viking For-
mation in Alberta, they concluded that the combination of interpola-

a)

b) o

b)

Figure 5. Foothills survey: migrated gathers 3D anisotropic depth

migration from a original data and b with interpolation before Figure 6. Foothills survey: migrated stack section, 0 1000 m a
migration. without interpolation, and b with interpolation before migration.

Downloaded 29 Jan 2010 to 80.194.194.190. Redistribution subject to SEG license or copyright; see Terms of Use at http://segdl.org/
V130 Trad

tion and prestack time migration provided the best AVO estimates, spacing along lines was halved to reduce the bin size from 12.5
achieving a correlation increase from 0.39 for migration without in- 50 to 12.5 25 m. Figure 8 shows the shot locations after inter-
terpolation to 0.57 for migration after interpolation. This improve- polation. The red dots indicate the locations of the original shots, and
ment can be taken as evidence of amplitude preservation during in- the blue dots indicate the locations of the new shots.
terpolation. As a comparison, a prestack time migration stack was produced
Figure 7 shows CIGs from prestack time migration with and with- using the original acquired data; then the stack was interpolated, as
out interpolation before migration. The migration applied in this shown in Figure 9a. In Figure 9b, the prestack data were interpolated
case was type 1b, decreasing shot and receiver lines by half and in- before migration using 5D interpolation. The prestack interpolation
creasing fold by four times. Hunt et al. 2008 give a complete de- produced a data set input for migration that was better sampled than
scription of the experiment. the noninterpolated data set. This allowed the migration to operate
with greater fidelity on the steep-dip events in this case, applying
fewer antialiasing constraints. The prestack interpolation did not add
Increasing inline-crossline sampling for steep dips
information to the data, but it did allow the migration to make better
This example, also described in Gray et al. 2006, shows the ben- use of the information that was already in the data, allowing it to pro-
efits of reducing the bin size increasing inline-crossline sampling duce an image with greater structural detail.
before migration rather than afterward. The land data set in this ex-
ample was acquired over a structured area in Thailand using an or-
Increasing inline-crossline sampling for survey merging
thogonal layout. The objective of the interpolation was to obtain
more information on steep dips by including moderate- to high-fre- Often, surveys acquired with different natural bin sizes need to be
quency energy that the migration antialias filter removed from the merged into a common grid. If a survey is gridded into a bin size
original, more coarsely sampled data. For this purpose, the shot
a)
a) o b) o

Figure 7. Foothills II. CIGs from prestack time migration a without

interpolation and b with interpolation before migration.
b)

Figure 9. Thailand: prestack time migration stacks. a Interpolation

Figure 8. Thailand: shot locations after interpolation for an orthogo- performed after stacking the migrated images. b Interpolation per-
nal geometry. Red dots are original shots; blue dots are new interpo- formed before migration. The improved imaging of the steep-dip
lated shots. The two large gaps are 1000 1500 m in diameter be- event in the center of the section is evident in b. Data courtesy of
fore interpolation. PTT Exploration and Production

Downloaded 29 Jan 2010 to 80.194.194.190. Redistribution subject to SEG license or copyright; see Terms of Use at http://segdl.org/
 Five-dimensional interpolation V131

smaller than it was designed for, the CMP coverage becomes very Figure 10a and b shows one CMP before and after interpolation.
poor, affecting further interpretation even after migration. A solution Figure 11a shows a time slice from the stack of original data in the
is to use prestack interpolation to reduce the natural bin size to match 10 10-m grid. Figure 11b shows the same time slice from the stack
the merge grid. This can be achieved by increasing sampling in the of the interpolated data.
inline-crossline domain or, alternatively, by using the surface-con-
sistent approach to decrease the distance between shots and receiv-
ers along lines. Infilling large gaps
Trad et al. 2008 show a case history from the Surmont bitumen It is common for 3D acquisitions to have large gaps with missing
project in northern Alberta. In this area, nine surveys had to be shots or receivers because of inaccessibility in some areas lakes,
merged into a common 10 10-m CMP grid. Of the nine surveys in hills, population, etc.. Although it usually is impossible to infill
the project, one was acquired with a natural bin size of 15 30 m, large gaps completely, decreasing their size has a large impact on mi-
giving poor coverage when binned in the 10 10-m CMP grid used gration results. The following example shows the infilling of a large
for the merge. Furthermore, this survey was the only one in the gap produced by an open-pit coal mine in an area with structured ge-
project with a parallel design the other surveys were acquired with ology. This obstacle prevented shots and receivers from being de-
an orthogonal geometry. By adding new shots and receivers using ployed at this location during the 3D acquisition. New shots and re-
the method presented in this paper, the coarser survey was trans- ceivers were added on the border of the gap. Time migration of the
formed from a parallel geometry with 10 30-m bin size to an or-
thogonal survey of 10 10-m bin size and twice the original fold. a)
The original data were fully preserved and the numbers of shots and
receivers were each increased by three, so the final size was nine
times the original size. The interpolation allowed this survey to
merge with the other surveys in the Surmont area, avoiding the need
for reshooting.

a)
0

b)

b)

Figure 10. Surmont: comparison of a CMP a before and b after in-

terpolation. Empty traces have been added to the CMP before inter-
polation to match the traces obtained after interpolation. The CMP in Figure 11. Surmont: time slice comparison from a the stack of the
b was created by using information from many other CMPs not original data and b a time slice from the stack of the interpolated
shown in the figure. data.

Downloaded 29 Jan 2010 to 80.194.194.190. Redistribution subject to SEG license or copyright; see Terms of Use at http://segdl.org/
V132 Trad

a) b) benefits of tighter acquisition sampling patterns, higher fold, and/or

smaller bin size.
By working in five dimensions, this interpolation method can in-
crease sampling in surveys that are problematic for lower dimen-
sional interpolators. The technique might be applied to overcome ac-
quisition constraints at a fraction of field acquisition costs, merge
data sets with different bin sizes, and eliminate differences caused by
acquisition, avoiding the need to reshoot surveys. Benefits include
more reliable prestack processes: velocity analysis, prestack migra-
tion, AVO and AVAz analyses, reduction of migration artifacts, and
improved imaging of steep dips.

ACKNOWLEDGMENTS
Figure 12. Coal mine: comparison of time migration images a I would like to thank CGGVeritas for permission to publish this
without and b with interpolation for a 3D survey acquired on top of paper and CGGVeritas Library Canada, PTT Exploration and Pro-
a large gap.
duction, PetroCanada, ConocoPhillips Canada Ltd., and Total E&P
Canada Ltd. for data examples. Special thanks are owed to several
original seismic data produced the image in Figure 12a. After inter-
colleagues who helped produce the interpolation examples and who
polation, time migration produced the image in Figure 12b. The in-
provided useful ideas and discussions on interpolation over the
terpolated image shows an anticline underneath the open-pit mine
years. In particular, my thanks to Bin Liu and Mauricio Sacchi,
that was confirmed by 2D seismic and well logs acquired before the
whose work constitutes the cornerstone of this method.
existence of the mine.

DISCUSSION REFERENCES
Interpolation fills missing traces by incorporating information in Abma, R., and N. Kabir, 2006, 3D interpolation of irregular data with POCS
the data using a priori assumptions. This provides, for standard pro- algorithm: Geophysics, 71, no. 6, E91E97.
cesses, information that is already in the data but is not accessible Abma, R., C. Kelley, and J. Kaldy, 2007, Sources and treatments of migra-
tion-introduced artifacts and noise: 77th Annual International Meeting,
without these constraints. On the negative side, results from interpo- SEG, Expanded Abstracts, 23492353.
lated data sometimes can be worse than results without interpola- Cordsen, A., M. Galbraith, and J. Peirce, 2000, Planning land 3-D seismic
tion. This can happen because some processes such as stacking surveys: SEG.
Downton, J., B. Durrani, L. Hunt, S. Hadley, and M. Hadley, 2008, 5D inter-
might work better for zero traces than for wrong traces. In addition, polation, PSTM and AVO inversion for land seismic data: 70th Annual
interpolation can add spurious information in a coherent manner, a Conference and Technical Exhibition, EAGE, Extended Abstracts, G029.
Duijndam, A. J. W., and M. A. Schonewille, 1999, Nonuniform fast Fourier
problem that stacking is unable to fix. Interpolation must be applied transform: Geophysics, 64, 539551.
very carefully to ensure this does not happen. Gray, S., D. Trad, B. Biondi, and L. Lines, 2006, Towards wave-equation im-
Several factors play against interpolation because it is by nature aging and velocity estimation: CSEG Recorder, 31, 4753.
Hansen, P., 1998, Rank-deficient and discrete ill-posed problems: Numerical
an ill-conditioned problem. Not only do unrecorded samples have to aspects of linear inversion: Society for Industrial and Applied Mathemat-
be estimated, but they also must be located in a manner consistent ics.
with the rest of the data, for example, at the proper elevations with Herrmann, P., T. Mojesky, M. Magesan, and P. Hugonnet, 2000, De-aliased,
high-resolution Radon transforms: 70th Annual International Meeting,
proper static information. Usually, geometries that can benefit from SEG, Expanded Abstracts, 19531957.
interpolation do not lend themselves to good noise attenuation. Ac- Hunt, L., J. Downton, S. Reynolds, S. Hadley, M. Hadley, D. Trad, and B.
Durrani, 2008, Interpolation, PSTM, & AVO for Viking and Nisku targets
curate interpolation becomes more difficult as the structure becomes in West Central Alberta: CSEG Recorder, 33, 719.
more complex, as gaps become larger or sampling poorer, and as the Liu, B., and M. D. Sacchi, 2004, Minimum weighted norm interpolation of
signal-to-noise ratio gets lower. Therefore, careful QC is necessary seismic records: Geophysics, 69, 15601568.
Mosher, C. C., T. H. Keho, A. B. Weglein, and D. J. Foster, 1996, The impact
to select interpolated data according to some quality criteria. A use- of migration on AVO: Geophysics, 61, 16031615.
ful criterion is the minimum distance between a trace and its original Poole, G., and P. Herrmann, 2007, Multidimensional data regularization for
neighbors, but many other QC parameters can be estimated and modern acquisition geometries: 77th Annual International Meeting, SEG,
Expanded Abstracts, 25852589.
saved into headers. After interpolation, these parameters can be used Sacchi, M. D., and T. J. Ulrych, 1996, Estimation of the discrete Fourier
together to decide if a new trace is acceptable. transform A linear inversion approach: Geophysics, 61, 11281136.
Schonewille, M. A., R. Romijn, A. J. W. Duijndam, and L. Ongkiehong,
2003, A general reconstruction scheme for dominant azimuth 3D seismic
CONCLUSIONS data: Geophysics, 68, 20922105.
Trad, D., J. Deere, and S. Cheadle, 2005, Understanding land data interpola-
Wide-azimuth geometries often are undersampled along one or tion: 75th Annual International Meeting, SEG, Expanded Abstracts,
21582161.
more dimensions, and interpolation is a very useful tool to precondi- Trad, D., M. Hall, and M. Cotra, 2008, Merging surveys with multidimen-
tion the data for prestack processes such as migration, AVO, and sional interpolation: CSPG CSEG CWLS Conference, Expanded Ab-
AVAz. I have discussed a 5D interpolation technique to create new stracts, 172176.
Trad, D., T. Ulrych, and M. Sacchi, 2003, Latest views of the sparse Radon
shots and receivers for 3D land seismic data that honors amplitude transform: Geophysics, 68, 386399.
variations along inline, crossline, offset, and azimuth. Although not Xu, S., Y. Zhang, D. Pham, and G. Lambare, 2005, Antileakage Fourier trans-
form for seismic data regularization: Geophysics, 70, no. 4, V87V95.
intended to replace acquiring adequate data for processing, this tool Zwartjes, P. M., and M. D. Sacchi, 2007, Fourier reconstruction of nonuni-
is useful for overcoming acquisition constraints and for obtaining formly sampled, aliased seismic data: Geophysics, 72, no. 1, V21V32.

Downloaded 29 Jan 2010 to 80.194.194.190. Redistribution subject to SEG license or copyright; see Terms of Use at http://segdl.org/