Interpolation, PSTM, & AVO for Viking

LUNCHEON
and Nisku targets in West Central Alberta
Lee Hunt, Scott Reynolds, Scott Hadley, and Mark Hadley
Fairborne Energy Ltd.

Jon Downton, Daniel Trad, and Bashir Durrani

CGGVeritas, Calgary, Alberta, Canada

Summary use a 5D interpolation prior to migration to minimize these

artifacts. We chose two current exploration and development
West Central Alberta has a thick stratigraphic section (up to targets to illustrate the effectiveness of this new method of
5000m) with multiple exploration and development targets in AVO processing: the Viking, and the Nisku.
conventional clastic, carbonate, tight sand, and shale reser-
voirs. Compressional and extensional structural features are The Viking is a prolific zone that is deep, structured, and very
common throughout this large region, particularly further thin relative to tuning thickness. We attempted to exploit the
west. The area has extensive 3D seismic coverage, but that AVO response of this gas-charged zone on migrated gathers
coverage typically has coarse source and receiver line to obtain a more accurate estimate of Phi-H than we could
spacing. AVO attributes have superior exploration and devel- achieve with conventional amplitude mapping. In order to
opment utility for many of the targets as compared to evaluate the performance of the new interpolation workflow,
conventional stacked data, but the use of AVO on any of these we performed our AVO analysis with a number of other rele-
zones are challenged by the difficulty in extracting accurate vant and established competing techniques. This interpola-
AVO results. One of the variables that adversely affect the tion migration flow was the most effective flow relative to
accuracy of the AVO extractions is the coarseness of the 3D other methods tried on this data, and represents an important
seismic geometry. This coarse 3D geometry causes migration new paradigm for land AVO analysis. This assertion is vali-
artifacts on our migrated gathers that arise due to the insuffi- dated by 29 well ties and the resultant maps which are consis-
cient sampling of the seismic wavefield. This issue led us to tent with the structural erosional model of the Viking.
Continued on Page 10

Lee Hunt currently consults for Fairborne Energy and Artek Exploration. He graduated from the
University of Alberta with a B.Sc. in geophysics in 1990, after which he started his career working for
PanCanadian Petroleum Ltd. The latter two thirds of his career has largely been at small junior compa-
nies. His experience ranges from interpretation to managing a business unit, but he has always been
interested in working with others to improve his technical abilities to drill successful wells. He and his
co-authors won Best Abstract for the 2008 CSEG convention, and he was also co-recipient of the 2000
CSEG Best Paper Award. Lee has contributed to the CSEG in the past by acting as the 2001 CSEG
Convention Technical Chairman, as well as being one of the co-creators of the CSEG MLA. He is a
supporter of APEGGA, and was one of the participants in the creation of APEGGA’s Q.I. Practise
Standard.

Jon Downton is currently the Research Manger of the Canadian processing and imaging product line
for CGGVeritas. His research is in the area of estimating rock and fluid properties from seismic data,
including AVO, inversion and controlled amplitude processing. He obtained his Ph.D. from the
University of Calgary in 2005 and his B.Sc. in geophysics from the University of Alberta in 1985. He has
worked for numerous processing and software development companies including ITA, Landmark,
Integra Geoservices, Scott Pickford/Core Lab, Paradigm and now CGGVeritas. He is a member of the
SEG, EAGE, APEGGA and the CSEG for which he is currently the Vice President.

★ ★ ★ ★ ★ ★

DECEMBER LUNCHEON JANUARY LUNCHEON

DATE: December 15, 2008 Tuesday, January 20, 2008
TIME: 11:30 A.M. Lunch The Effectiveness of Seismic Fracture
LOCATION: Telus Convention Centre, Calgary
Detection Techniques in Tight Reservoirs –
examples from WCSB
TICKETS: Contact CSEG office Christian Abaco
TELEPHONE: 403-262-0015 or Fax: 403-262-7383 EnCana Corp., Calgary

December 2008 CSEG RECORDER 7

Luncheon Cont’d
Interpolation, PSTM, & AVO…
Continued from Page 7

Continued on Page 11

10 CSEG RECORDER December 2008

 Luncheon Cont’d
Interpolation, PSTM, & AVO…
Continued from Page 10
The Nisku carbonate play was also subjected to our new interpolation migration method. Evaluating the performance of the new para-
digm is aided in this case by the fact that one of the 3D seismic surveys in the area has been the subject of a previously published AVO
case study on the Nisku. As such, we have an excellent control study to compare against our new method. We generate AVO Inversion
results that behave in a more physically reasonable manner in cross plot space, tie the wells better, and produce cleaner, more geolog-
ically consistent maps than the previous study was able to produce.

By illustrating our new interpolation migration method of AVO processing on these two very different targets, we demonstrate the
general importance of stabilizing prestack migration with this kind of support on structured, coarse 3D seismic surveys. These exam-
ples offer some insight into the future direction of other, related methods of analysis such as Azimuthal AVO (AVAz), whose sampling
is much worse and could be expected to benefit from this kind of imaging support.

Introduction: interpretive
challenges
Seismic processing and interpretation
is at once a scientific, and a hopeful
endeavour. The presence of noise,
inadequacies of the acquisition
parameters, interpretive or physical
non-uniqueness within the resolution
at hand, and sometimes even the lack
of sufficient geologic sampling to
adequately understand all aspects of
the problem, all serve to add uncer-
tainty to our ability to confidently
interpret seismic data. Our approach
must always be as scientific as the
shortcomings of our data allow, and
constantly engage in the pursuit of
improving our method against these
same limitations. Whenever we
attempt a new technology to attack or
mitigate against the ill-posed nature
of our interpretive pursuit, there is a
certain element of hope that these
efforts can enable us to do better.

The interpretive examples shown in

this paper are somewhat disparate:
the thin gas-charged clastic Viking
zone and the Nisku reef of West
Central Alberta. These zones make
interesting representatives of a large
suite of similar or related interpretive
problems within the Western
Canadian Sedimentary Basin. The
Viking is very thin, but is over pres-
sured and gas filled. The Nisku is
very deep, and while thicker, has a
lithologic uniqueness problem with
some of its basinal equivalents. Both
are set in a structural regime and are
covered by 3D seismic data. In
forming the geophysical element of a
strategy around how to explore and
develop these zones, we first consid-
ered the various seismic attributes
that we might extract from the data.

There have been many recent efforts

to categorize the various attributes in
order to make this exercise easier.

Continued on Page 12

December 2008 CSEG RECORDER 11

Luncheon Cont’d
Interpolation, PSTM, & AVO…
Continued from Page 11
Table 1 (page 10) represents a classification scheme inspired by methods of AVO analysis. Much current research is being
and heavily borrowed from the work of Barnes (2001), Chambers directed into this worthwhile pursuit, with this paper directly
(2002), Wen (2004), and Chopra (2006). The classification scheme concerned with improving AVO analysis through a method of
attempts to break the attributes into types of information, interpolation.
certainty of their information, kind of geophysical problem they
address, and physical meaningfulness. The color overlay indi- Interpolation is useful and relevant because most AVO-sensitive
cates qualitatively the level of discrimination the attribute might plays need to be accurately imaged in space as well as time,
be able to provide. A darker green color indicates a higher degree requiring prestack migration. The more structurally complex the
of discrimination. It is clear that depending on the particular area, the more important this consideration is, although even for
exploration problems at hand, the useful or important attributes stratigraphic plays migration is important to collapse the Fresnel
will change. It is also clear, however, that certain of these attrib- zone. Despite these considerations it has been questionable
utes contain more fundamental and broadly useful information whether land seismic data processed specifically for AVO
than the others. The most fundamental attributes are, in fact, the analysis should be prestack migrated due to the migration arte-
rock property type attributes that are estimated through AVO facts introduced as a result of the sparse acquisition typical of
analysis. This makes sense when we consider that the stacked land acquisition. Land 3D seismic surveys are commonly shot
trace is a complex mixture of contributions from the AVO attrib- with coarse, irregular geometries. This lack of regularity causes
utes. One might argue that shape and texture attributes are also migration artefacts on the imaged gathers which can be a serious
fundamental in some ways, but the AVO attributes attempt to contributor to the overall noise level. By interpolating the data in
represent the rock properties directly, and are therefore the most an AVO-preserving manner prior to migration, we mitigate or
universally useful elements of information that may be estimated reduce these artefacts with minimal harm to the sharpness of the
from seismic data. The well known problem with these attributes image. Other methods of noise reduction commonly require
is that their certainty is poor. One might argue that some amount mixing, averaging, or stacking after imaging, which has the
of effort towards AVO analysis should always be employed on consequence of smearing structural information.
seismic data, except for the difficult, ubiquitous, problem of the We will illustrate and support these statements in our paper
difficulty in accurately extracting this data. This uncertainty through our discussion of the Viking and the Nisku exploration
almost always forces us to retreat to using the more robust and development problems. In the Viking we will employ a
stacked trace as our fundamental unit to extract the other attrib- rigorous method of comparing the usefulness of the interpola-
utes from, even though we know that the AVO attributes would tion method as compared to the older, established ways of
make more physical sense as the starting point. If having better handling the data. Our comparison is less rigorous in the Nisku,
AVO attributes would help us on our interpretive endeavours but we will utilize the available well control for validation
generally, then it makes sense to continue to improve our purposes and investigate other indications and effects of noise
on the results. We will also discuss issues of quality controlling
the interpolation method, and address future work.

It should be pointed out that much of the following work is taken

directly from Hunt et al. (2008) and Reynolds et al. (2008), of
which this paper is a compilation.

The Viking challenge

Viking Formation reservoirs in West Central Alberta represent
attractive exploration targets. Reservoirs are comprised of
shoreface sandstone assemblages that often retain 12-14 percent
porosity over 0 to 7 meters thickness, and occur at depths greater
than 2800m. The sandstones have low permeabilities (< 1 mD) but
are commonly overpressured and gas-bearing with typical recov-
Figure 1. Stratigraphic cross section of two key wells within the 3D volume. Log erable resource of 8 Bcf and 80,000 bbls condensate per section.
displays include gamma ray, density and neutron porosity curves. Well A (good
reservoir) associated with preserved upper to middle shoreface deposits shows a 5 m The structural setting for the area includes both extensional and
thick section of sandstone reservoir with greater than 6% density porosity. compressional tectonic elements.

Figure 2. 0 to 35 degree AVO models created for Well A (good reservoir) and Well B (absent reservoir), respectively.

Continued on Page 13

12 CSEG RECORDER December 2008

 Luncheon Cont’d
Interpolation, PSTM, & AVO…
Continued from Page 12
Figure 1 is a schematic cross section of two wells within the 3D stacked section. When porous and gas-charged, a well resolved
survey. These wells were selected for the modeling work as they Viking zone may have a Type IIa AVO response, where the weak
had full wireline log suites including shear and compressional peak goes towards smaller amplitudes with increasing offset.
sonic and core data. The section is hung on the Base Fish Scales Log data from Well A and B illustrate this response with the good
(BFS) and shows two basin-wide transgressive events labeled quality reservoir in Well A versus the absent reservoir in Well B.
VE3 and VE4 (after Boreen and Walker, 1991).
This response suggests that an advantage may lie with the
The preserved porosity in Well A is associated with the upper utilization of this contrasting AVO effect. The structural setting
shoreface deposits below the VE3 surface. demanded the use of PSTM gathers for the analysis.
Unfortunately, the 3D coverage in the area has coarse source and
In Well B the VE3 surface is underlain by tight lower shoreface
receiver line geometries, which give rise to gross irregularities in
and basinal deposits with no reservoir quality. Tectonism during
the offset and azimuthal distribution of the data. Upon PSTM,
Viking time may have caused reactivation of basement-rooted
these irregularities cause distortions which adversely affect the
faults which created an irregular sea floor topography that was
AVO analysis.
differentially eroded during basin-wide transgressive events
resulting in the complex aerial distribution of preserved porosity.
In general, porous and permeable progradational shoreface The Nisku challenge
deposits are preserved in structural lows.
In 2005 Pelletier and Gunderson published a thoughtful, petro-
Despite the coverage of 3D seismic, accurate reservoir prediction physically driven, work on AVO Inversion for a Nisku target in
is very challenging. The productive reservoir, which may be the Brazeau area. They attempted to discriminate low impedance
entirely or partially eroded, is never greater than 1/15 of a wave- basinal Cynthia shales from porous Nisku Reefs using AVO
length in thickness. These variables contribute to a poor correla- inversion on a 3D seismic survey. Their AVO Inversion tech-
tion between seismic amplitude and measures of porosity niques followed a rationale derived from petrophysical analysis
thickness. Recalling Figure 1, the Viking can be described as a of Lamé’s parameters from wireline data in the area. The work
more competent material sitting under a less competent half- appeared to be successful in that it seemed to identify 4 known
space (the BFS zone). Thus the Viking is a peak on a zero phase reefal wells, and also identified 2 new prospects. There were no

Continued on Page 14

December 2008 CSEG RECORDER 13

Luncheon Cont’d
Interpolation, PSTM, & AVO…
Continued from Page 13
basinal wells available on the 3D survey at that time. We use the these new improvements add enough value to warrant repro-
modifier “appeared” simply because the amount of drilling cessing of even recently analyzed data.
control remained statistically insufficient for greater confidence.
In the time since that paper was published, 10 new wells have Exploration for gas-bearing Nisku reefs in the Brazeau area repre-
become available, and several new technologies have been sents a classic challenge for seismic lithologic prediction. The
adapted to improve the veracity of AVO Inversion products. Of basinal Cynthia member is stratigraphically equivalent to the
particular interest is the fact that Pelletier and Gunderson’s Nisku reefal carbonates. When the Cynthia member is sufficiently
analysis was performed on CDP gathers rather than image
gathers. Since then, new 5D amplitude-preserving interpolation
algorithms have become available. These interpolation algo-
rithms can minimize migration artefacts allowing us to improve
our AVO estimates. We compare the AVO analysis based on the
original and reprocessed flow including interpolation and
PSTM. We will utilize the new drilling data to help us objectively
quantify the improvements to be gained by using our interpola-
tion migration method of AVO analysis, and discuss whether

Figure 3. A stratigraphic cross section representing several key Nisku facies: a

porous reef versus argillaceous or gamma ray “hot” basin. GR, DT, DS, and DN
Table 2. Key Nisku facies, quantified facies codes, and qualitative descriptions of represent gamma ray, compressional slowness, shear slowness, and bulk density,
their rock properties. respectively.

Figure 4a. Old Lambda*Rho versus Mu*Rho cross plot from Pelletier and Gunderson. Figure 4b. New Lambda*Rho versus Mu*Rho cross plot. Each data point in Figure
4b is coloured according to its gamma ray value as illustrated by the colour bar on the right side of the figure.

Continued on Page 15

14 CSEG RECORDER December 2008

 Luncheon Cont’d
Interpolation, PSTM, & AVO…
Continued from Page 14
argillaceous, it has a strong gamma ray
response, a low rigidity, and may
appear very similar to the porous
Nisku reef on conventional stacked
seismic. In this case, our challenge
would be the separation of higher
rigidity, lower Lambda*Rho reef from
lower rigidity shale – a classic
carbonate AVO Inversion problem. The
Cynthia member can also be more
calcareous as well, which would repre-
sent a different problem seismically. In
such a case, wireline data would see
less of gamma ray response, and the
rigidity would be higher – potentially
higher than the reefal facies. Figure 3
represents these facies differences as
measured by wireline logs.

Pelletier and Gunderson (2005)

focussed more on the calcareous
element of the basinal response. With
further drilling since their publication,
we have been able to sample the poten-
tial lithologic outcomes more
completely, and can address a wider
range of reefal and basinal possibilities.
Table 2 represents the rock properties
we would expect from the key end
members we wish to represent.

Figure 4a is a Lambda*Rho versus

Mu*Rho cross plot from Pelletier and
Gunderson (2005). As stated earlier,
critical sampling of the argillaceous
basin was not available at the time of
that publication, and the facies was
therefore not fully represented. In
Figure 4b we illustrate a Lambda*Rho
versus Mu*Rho cross plot calculated
from currently available wireline logs
on or near the 3D. In this case, we used
data points from 3 reefal wells and an
argillaceous basinal well that discrimi-
nates the lower rigidity we expect
from that facies. The augmented petro-
physical sampling that is currently
available should improve our ability to
discriminate facies in cross plot spaces
estimated from the 3D seismic.

It is clear that accurate estimates of

Mu*Rho are key in order to carry out
the necessary discrimination. Lambda*
Rho is also important, but is subject to
smaller experimental error than the
estimate of Mu*Rho. This is a conse-
quence of the fact that Mu*Rho is
calculated solely from the S-impedance
reflectivity (Goodway, 2001) which has
larger uncertainty in the presense of
noise than the P-impedance reflectivity

Continued on Page 16

December 2008 CSEG RECORDER 15

Luncheon Cont’d
Interpolation, PSTM, & AVO…
Continued from Page 15
(Downton and Lines, 2001). By interpolating and then prestack to be useful on many stratigraphic prospects. Yi, Downton, and
migrating the seismic data, the signal-to-noise level is improved. Xu (2007) describe the crucial challenges involved in successfully
This is a result of the better sampling of the wavefield prior to the estimating the elastic parameters. Those challenges include the
migration, resulting in less migration artifacts and noise. need to obtain migrated gathers that have sufficient resolution
Pelletier and Gunderson’s work was carried out on CDP gathers and high S/N ratios. In these examples, the structural deforma-
at a time when this kind of interpolation migration flow was not tion appears to be significant enough to require that AVO be
available. We will evaluate how much of an improvement comes performed on PSTM gathers. The nominal source and receiver
from using the new method over the old by comparing the orig- line spacing of this 3D survey is 600m, and the data is noisy and
inal Pelletier and Gunderson AVO attributes to the same attrib- bandlimited to about 55Hz. This acquisition geometry was not
utes extracted using our new method. Validation is carried out sampled sufficiently fine enough so that the wavefield would
using the current, improved, well control in the area. constructively and destructively interfere to image all the reflec-
tors with a good S/N ratio.
Theory
To address this issue a 5D interpolation method (Trad, 2007) based
Fractional elastic parameters such as the compressional (Rp) and on Minimum Weighted Norm Interpolation (MWNI) (Liu and
shear reflectivity (Rs) may be estimated from the prestack seismic Sacchi, 2004) was employed to regularize the data prior to migra-
data by AVO Inversion such as the two-term Gidlow et al. (1992) tion. The goal of this interpolator is to detect poorly sampled
equation R(q)=Rp sec2 q – 8Rs g2 sin2 q, (where q is the average coherent information and make it available for other processing
angle of incidence and g is the average S-wave / P-wave velocity techniques like migration.The 5D interpolation is performed by
ratio. There exists a wide variety of AVO attributes that could solving a large inverse problem in the inline-crossline-offset-
have been used for mapping and validation in this paper. For the azimuth and frequency domain. In this case the desired model is
Viking zone, we chose a very simple parameter, the damped Rp a super-sampled seismic dataset, the data is the original seismic
to Rs ratio, while we chose Goodway’s (2001) Lamé parameters dataset, and the linear model is the sampling operator. The
for the Nisku zone. Unfortunately, all of these parameters are resultant super-sampled seismic data contains data for every
notoriously difficult to estimate with the fidelity and resolution possible inline, crossline, offset, and azimuth combination. In

Figure 5. Ostrander gather at well A (a) & B (b), interpolated Ostrander gather at well A (c) & B (d), PSTM Ostrander gather at well A (e) & B (f), interpolated PSTM
Ostrander gather at well A (g) & B (h). The Viking is at 1790 ms at well A and 1821 ms at well B.

Figure 6. Ostrander gathers at the argillaceous basinal well 10-17. 6a is the 1x1 CDP gather, 6b is the 5x5 CDP gather such as was used by Pelletier and Gunderson. 6c is
a 1x1 PSTM gather. 6d is the interpolated PSTM gather without superbinning, and represents the new method. The Nisku zone is comprised of the trough (in blue) just
above 2.1s, to the base Nisku peak (in red) which is just below 2.1s.

Continued on Page 17

16 CSEG RECORDER December 2008

 Luncheon Cont’d
Interpolation, PSTM, & AVO…
Continued from Page 16
practice this creates so much data that it becomes impossible to any new prospective well, we loaded the Phi-H value as a top
deal with, so the algorithm only outputs a representative subset of into each well on the 3D. We then extracted our Rp/Rs ratio
this. In this case we output twice as many shot and receiver lines attribute at each well point and plotted the attribute versus Phi-
as the original geometry, thus preserving the original geometry H. To quantify any improvements that our new method might
and data as a subset of the new volume. As the original data are yield, we created several versions of the data. Each version was
preserved, it is easy to verify that the original AVO trend is processed identically to the others- except for the noted change
preserved. Further, the minimum norm constraint applied in the (Table 3). By comparing results of our extracted AVO attribute
spatial frequency domain tends to ensure that the amplitude (Rp/Rs) at the well control for each of the versions, we hope to be
changes slowly in all four spatial dimensions including offset and able to comment on the usefulness of our new paradigm. The
azimuth while honouring the original data. correlation coefficient (CC) of this regression was used as the
primary method of validation, and is denoted as CC in Table 3
The high dimensional nature of this method permits the use of below. Note that the interpolated PSTM result was the best. The
information from different spatial directions simultaneously PSTM versions were generally better than those not imaged.
and hence to infill geometries with very sparse sampling, Interestingly, super-binning did not uniformly yield a better CC.
beyond what is possible when working in a lower dimensional Perhaps the trade-off of footprint noise suppression versus struc-
space. However, this capability should not be abused, therefore tural and stratigraphic smearing compromised this technique.
traces created at a large distance from the acquired data are
automatically discarded. This distance is measured in the five
dimensional space where the interpolation takes place. Also 5D
windows which do not have a required minimum number of
live traces are not interpolated. Other typical quality control
techniques are limited offset/azimuth stacks, which may be
viewed as timeslices in addition to inline and crossline direc-
tions, and prestack gather inspection to check for amplitude and
traveltime consistency.

Gather comparisons
Table 3. The versions of the data that the AVO attribute was extracted from and
Let us compare the effectiveness of the new interpolation regressed with Phi-H. CC is the correlation coefficient of that regression.
method on individual gathers first for the Viking, and secondly,
for the Nisku zones. Figure 7 compares a portion of the interpolated PSTM attribute
map to the 5x5 superbinned CDP method, the 1x1 PSTM method,
Figure 5 (a, b, c, d) shows a comparison of the original data at the
and the 5x5 PSTM method. Low ratio values are in green to red.
Viking zone’s well locations A & B versus the interpolated data
Although the CC values are an objective measure of the veracity
at the same locations. The AVO trend of the data has been
of the AVO attribute, they can be hypersensitive to outliers. This
preserved and is similar to the models shown in Figure 2.
is why a comparison of the quality of the map results and even
Similarly, Figure 5 (e, f, g, h) compares the prestack migrated
the gathers are of significant additional value. For example, even
(uninterpolated) data to the PSTM gathers after interpolation.
though the 5x5 CDP method has a similar CC to the 5x5 PSTM
For all the cases the interpolated results have a superior S/N
method, it is clear from the map comparison that the 5x5 PSTM
ratio than the non-interpolated gathers. This is partly due to the
result is superior. In fact, the map comparsion reveals that the
higher fold and, in the case of the interpolated PSTM gathers,
CDP method is grossly inferior to all the PSTM (migrated)
better sampling of the wavefield prior to the migration resulting
results. Wells A and B are also noted, and are discriminated quite
in less migration noise.

Figure 6 is an Ostrander gather taken at an argillaceous basinal

well location. As Pelletier and Gunderson used super-binned
CDP gathers for their AVO analysis, we illustrate a CDP gather
in Figure 6a versus a super-binned CDP gather in Figure 6b. Our
new method is illustrated by comparison with a PSTM gather in
Figure 6c and the interpolated PSTM we advocate in Figure 6d.
It is clear that the interpolated results have a superior S/N ratio.
This is due to the higher fold in the non-migrated examples. In
the case of the interpolated PSTM gathers, better sampling of the
wavefield results in less migration noise. Note that all the
gathers shown have been partially offset-stacked with the same
offset binning so that the figures may be easily compared.

Viking test case: 3D seismic with significant

well control
The 3D that we used for this test had some 29 Viking penetra- Figure 7. RpRs Ratio attributes. Top left -interpolated PSTM, top right – 5x5 CDP,
tions. Since the primary geologic goal was to predict Phi-H in bottom left – 1x1 PSTM, bottom right – 5x5 PSTM.

Continued on Page 18

December 2008 CSEG RECORDER 17

Luncheon Cont’d
Interpolation, PSTM, & AVO…
Continued from Page 17
well in all maps. The low Rp/Rs ratio trend has a spatially sharp, and surgical nature of proper use of such Lamé parameters for
structurally influenced shape. This spatial sensitivity may be the lithologic prediction, some caution should be taken in evalu-
strongest influence on the dissappointing results of superbinning ating our technique with simple correlation techniques.
in this example. Stability enhancement is better achieved prior to Nevertheless, this comparison is instructive and supportive of
migration. In the regression and on the maps, it is clear that the our assertions.
new interpolated PSTM method yielded the best results. The
dataset was also used to succesfully predict the results of two 2. Comparison of cross plots of LMR Ratio versus Mu*Rho
new wells denoted with rig symbols on the maps. and Lambda*Rho versus Mu*Rho applied to a map view.
These maps are compared and it is shown that the maps
resulting from the new method are much cleaner and more
The Nisku test case geologically reasonable than the old maps (illustrations not
The 3D that we used for this test had some 14 Nisku penetra- shown here).
tions. Of key importance is the fact that 3 of the Nisku penetra- 3. The seismic cross sections of Lamé parameters (not illus-
tions are basinal, and 1 penetration is a tight reef. Each of these trated here) demonstrate that the new method has yielded
wells is given a lithology code as defined in Table 2. We have much cleaner, higher signal-to-noise ratio inversions.
Pelletier and Gunderson’s Lamé parameter attributes from their
original work, and we have the new Lamé attributes extracted 4. The shape and scatter appear in the cross plot domains to
from the same 3D with our new interpolation migration tech- also be supportive of our new method. Noise or error in the
nique. We compare the quality of these results in several ways: estimation process contributes to greater scatter of the
Mu*Rho axis in particular. These cross plots are compared
1. The correlation coefficient of (chiefly) the LMR Ratio and in Figure 8, below. We know from Figure 4 that we should
(secondarily) the other Lamé parameters to the lithology code expect an upper right to lower left set of trends in this
that we give to each well on the 3D. This objective, quantifi- particular cross plot space, although mineralogical
able measure is illustrated below, in Table 4, and supports complexities can cause a widening of the trends in the
improved predictive ability with the new interpolation orthogonal direction. It is clear that the old method has a
method. Given the complexity of the mineralogy in the area, non-physical trend orthogonal to our expectations, and in
excess of what we might expect from the influence of
mineralogical variation. The new method yields a more
physically meaningful cross plot.

Table 4. Nisku zone correlations of lithology code versus Lamé parameters.

Figure 8. Lambda rho versus Mu rho cross plot comparison for the Nisku zone, between the old method and new method respectively.

Continued on Page 19

18 CSEG RECORDER December 2008

 Luncheon Cont’d
Interpolation, PSTM, & AVO…
Continued from Page 18

Conclusions successful noise attenuation and static Goodway, B., 2001, AVO and Lamé constants for rock
parameterization and fluid detection: CSEG RECORDER,
corrections. Therefore, the existence of Vol. 26, No. 6, pp. 39-60.
The new interpolation migration method efficient interpolation tools does not Hunt, L., H., Hadley, S., Hadley, M., Downton, J.,
of AVO analysis yielded superior results diminish the importance of acquiring Durrani, B., 2008, Interpolation, PSTM, AVO, and a
in our comparative work on the Viking, data with the best possible sampling. In Thin Gas Charged Viking Shoreface in West Central
and the Nisku. In both cases, the signal-to- Alberta, CSEG Annual Convention.
fact, interpolation could be taken into
noise ratio of the new method was illus- Li, Y., Downton, J., and Xu, Y., 2007, Practical aspects of
account during acquisition not to save AVO modeling: TLE, March, 26, No. 3, pages 295-311.
trated by a variety of methods. The Viking acquisition costs, but to acquire addi- Liu, B. and Sacchi, M., 2004, Minimum Weighted Norm
zone had the most well control, and was tional data that can be used by interpola- Interpolation of Seismic Records, Geophysics 69, 7.
compared more rigorously than the tion to infill areas where shots or Pelletier, H., and Gunderson, J., 2005, Application of
Nisku. The well control on our 3D receivers cannot be deployed. It is better rock physics to an exploration play: A carbonate case study
allowed us to validate our method for the from the Brazeau River 3D: TLE 24, No 5, 516-519.
to acquire actual data whenever possible
Viking objectively, including a quantita- (and economic) due to potential non- Reynolds, S., Hunt, L., Hadley, S., Hadley, M.,
Downton, J., Durrani, B., 2008, Are key technological
tive measure of the relationship between uniqueness, than to rely too heavily on changes of the last three years enough to warrant repro-
Phi-H and an AVO attribute. Of the many interpolation. cessing? Interpolation and AVO Inversion for the Nisku at
variations of the seismic data that we Brazeau, CSEG Annual Convention.
compared, the interpolated PSTM gathers Future work is focused on using interpo- Trad, D., 2007, A Strategy for wide-azimuth Land Data
lation to precondition the data for Interpolation: 77th Annual International Meeting,
without superbinning were unequivo- SEG, Expanded Abstracts, 946-950.
cally the best. Interestingly, the variations azimuthal migrations and analysis (e.g.
Wen, R., 2004, 3D Modeling of Stratigraphic
that had PSTM were consistently superior azimuthal AVO and velocity analysis). Heterogeneity in Channelized Rservoirs: Methods and
to those which were not migrated. Despite This will place greater demands on the Applications in Seismic Attribute Facies Classification,
interpolation and seismic data acquistion. CSEG RECORDER, Vol 29, No. 3, pp. 38-45.
our original opinion that imaging would
R Geomodeling Technology Corp, 2006, VisualVoxAt
be important, we were nevertheless training manuals, Introduction to VisualVoxAt.
surprised by how important it was on this
data. Interpolation was also shown to be a Acknowledgements
better method of stabilization than simply
superbinning the PSTM gathers. In fact, We thank Scott Cheadle, Kendall Rogers,
despite giving the map views a cleaner Xiaowei Luo and John Zhang
appearance, the net effect of super- (CGGVeritas), Cameron Demmans, Brad
binning did not consistently help our Molnar (Fairborne) for their work on this
cross correlations- a result that surprised project, CGGVeritas Library Canada and
us. Comparisons of gather quality are Fairborne Energy Ltd for allowing us to
consistent with the quantitative measures, show this information.
and support the interpolation PSTM flow
that we advocate. Cross plot comparsions References
of Lamé parameters for the Nisku zone Barnes, A. E., 2001, Seismic Attributes in your facies,
also appeared to physically indicate the CSEG RECORDER, pp. 41-47, September Issue.
superiority of the new method. Boreen, T. and Walker, R.G., 1991, Definition of
allomembers and their facies assemblages in the Viking
AVO attributes are generally of great Formation, Willesden Green area, Alberta; Bull. Can.
Petrol. Geol., v39, p123-144.
fundamental importance in seismic inter-
Chambers, Richard, and Yarus, Jeffrey, 2002,
pretation, but their accuracy limits their Quantitative Use of Seismic Attributes for Reservoir
wider use. This new interpolation Characterization, CSEG RECORDER, Vol. 27, No. 6,
pp. 14-25.
method yields measurable improvements
to the fidelity of AVO extractions on two Chopra, S, and KJ. Marfurt, 2006, Seismic Attributes for
Prospect Identification and Reservoir Characterization,
zones in related 3D data, and appears to SEG publication.
be an important step in enabling wider Downton, J., and Lines, L., 2001, AVO feasibility and
use of AVO attributes in exploration and reliability analysis: CSEG RECORDER, Vol. 26, No 6, p
66-73.
development.
Gidlow, P.M., Smith, G.C., and Vail, P.J., 1992,
It is important to point out that successful Hydrocarbon detection using fluid factor traces, a case
study: How useful is AVO analysis?: Joint SEG/EAEG
interpolation requires the capability of summer research workshop, Technical Program and
detecting data coherency across spatial Abstracts, 78-79.
dimensions, which heavily depends on

December 2008 CSEG RECORDER 19