FEATURE ARTICLE

EFFECT OF WELL CONTROL ON

CONSTRAINED SPARSE SPIKE
SEISMIC INVERSION
JOHN PENDREL and PAUL van RIEL

ABSTRACT • Inclusion of geophysical constraints from known information or

Modern seismic inversion methods transform seismic reflection analogues
data to pseudo acoustic impedance logs. Implicit in this approach is • Calibration to well log data
an integration of low frequency model information with higher fre- • Improved interpretability of seismic horizons
quencies from the seismic. This is necessary since the low frequen-
cies present in logs and therefore required by the inversion, are • Increased bandwidth in the inversion output
absent in the seismic data. We obtain the low frequency impedances • Attenuation of random noise
from a 3D solid model defined by interpreted horizons and
impedance logs. It is important to understand the effects of this a The problem of transforming seismic reflection data to acoustic
priori log information on the final inversion. We must guard against impedances has been addressed by us elsewhere (Pendrel & Van
using the logs to provide the answer. Our goal is to place the trans- Riel, 1997). Here we focus on the problem of appropriately repre-
formed high frequency seismic information in a reasonable geolog- senting the low frequencies in the final inversion. These are missing
ic setting defined by the low frequency model. in the seismic reflection data and are ill-constrained in the trans-
formed seismic. We can obtain reliable and reasonable low frequen-
We investigate these ideas using a western Canadian example, a cies from a 3D solid model created from impedance logs and
small 8600 bin 3D survey over a Devonian reef. Within it are located interpreted horizons. The result is a broad band inversion repre-
19 wells, each with a set of sonic and density logs. First, we compute senting a set of pseudo acoustic impedance logs at each CMP. We
an inversion assuming knowledge of only the most distant off-reef present a methodology for doing this and demonstrate the proce-
well. In this case, the low frequency model is clearly erroneous. Next dure using data acquired over a Western Canadian Devonian reef.
we re-do the inversion, adding a single on-reef well to the model.
Finally, we assume knowledge of all on-reef wells. We conclude that It is a common concern that the results of seismic inversion sim-
significant information about the reef anomaly resides in the seismic ply mirror the input well logs at and near well locations. We use a
domain. However, lower frequencies from the model are useful in constrained sparse spike inversion algorithm (Debeye and Van Riel,
providing a correct geological context and also necessary to calibrate 1990) to investigate the effects of adding new logs to the model,
the seismic information. The modelling procedure is evolutionary, seeking to separate the individual contributions of logs and seismic
new well logs being used to update both the model and the inver- to the final inversion. In addition to computing the full inversion, we
sion. In this way, the inversion can become a viable tool in both produce an auxiliary data cube representing the seismic only com-
exploration and production environments. ponent. This result is referred to as the relative inversion as it con-
tains no low frequencies. It will be shown that the fine detail in the
INTRODUCTION full inversion is contained in the relative inversion. The low fre-
The goal of seismic inversion is to transform input seismic data quencies from the model are used to place the relative inversion
to pseudo acoustic impedance logs at each CMP. The development information in the correct geologic setting. When logs from many
of inversion methods which account for both wavelet amplitude wells are available, the geologic setting can be determined quite
and phase has resulted in an increased role for inversions as a diag- accurately. Alternately, in a rank exploration environment, it may
nostic tool in the exploration for hydrocarbons. The benefits of seis- represent the best guess of the geologist and geophysicist. Then, sev-
mic inversion include eral geologic scenarios may be investigated and models and inver-
sions computed for each. As more wells are drilled, the model is
• Compensation for and reduction of the effects of wavelet tuning updated and the geologic setting becomes more accurate. It is true
• Presentation of output as geologic layers rather than reflection however, that the low frequency component can never be deter-
edges mined uniquely at a given CMP until a well has been drilled there.
While the model may be updated over time, the relative inversion
• Merging of low frequency geologic and geophysical information
remains largely invariant. Changes will only result from updating
with seismic data
• Modelling and inclusion of layer stratigraphy
Continued on Page 20

18 CSEG Recorder December, 2000

FEATURE ARTICLE Continued
EFFECT OF WELL CONTROL ON CONSTRAINED SPARSE SPIKE SEISMIC INVERSION
Continued from Page 18

the global sparse spike constraints or upgrading the wavelet, in METHODOLOGY

response to new well information. Creating an Earth Model
The first step in model building is to design the structure. This is
GEOLOGY done by providing two pieces of information - the interpreted hori-
The geology is that of a Western Canadian Devonian reef. Figure zons and the model “framework”. The framework, in the form of a
1 shows the reef structure and the locations of the wells. The width spreadsheet, describes the ordering of the horizons in space and
of the main build-up is about 1 km. Surrounding the main reef is a time and their behaviour at faults. The horizons, which can include
porous apron which, away from the reef core interfingers with interpreted faults, provide structure information. Together, these
tighter facies. A lateral salt is present which is dissolved close to the form a blueprint for the model. Stratigraphy within the layers is
main reef build-up. There were several episodes of dissolution. specified as onlap, offlap, reef, channel, etc. The model is completed
Some were compensated by the basin material and some were not. by populating it with geophysical information, usually input in the
This has had some influence in the small scale structure of the off- form of well log data. We are most interested in impedance since this
reef events. The off-reef basin facies are dolomitic and contain thin is needed to complete the low frequency portion of the seismic
anhydrites which onlap against the reef. The considerable velocity inversion.
contrast between the porous reef core and the high impedance off-
reef facies has resulted in push-down of the reef platform and lower The logs are usually provided in depth and the horizons in time.
horizons. Other major reflectors are a shale-carbonate transition Thus it is necessary to create a time-depth transformation if these
above the reef and a second deeper laterally continuous salt under two pieces of information are to be rationalized. Input sonic logs are
the reef platform. integrated, hung on an input time datum (one of the input horizons)
and drift-corrected to tie the time horizons. Upon completion, the
model exists in both time and depth and time-to-depth and depth-
to-time transformations are possible. Interpolation of the input log
information between wells is done along layers, respecting both
stratigraphy and faults. One of several simple schemes such as lin-
ear interpolation or kriging is generally used. This is more than ade-
quate for the task of computing a low frequency background model
for the inversion.

Constrained Sparse Spike Inversion

Constrained sparse spike inversion models the input seismic
data as the convolution of the seismic wavelet with a reflection coef-
ficient series. Wavelet analysis is accomplished by computing a filter
which best shapes the well log reflection coefficients to the input
seismic at the well locations. Because the seismic wavelet is band-
limited, the problem is non-unique. There are many reflection coef-
ficient series which, when convolved with the seismic wavelet,
reproduce the input seismic to an arbitrary degree of accuracy. To
find the best geophysical and geological solution from the large
number of available mathematical solutions, we must impose other
Figure 1: Structure of the “Reef” Horizon and the locations of the vertical and conditions. These are provided by geophysical constraints which
horizontal wells. describe how the impedance can vary laterally away from the wells.
A “fairway” of allowed impedances is defined, which limits the
variability of inversion impedances and automatically rejects many
good mathematical solutions which are geophysically or geological-
ly unjustifiable. The impedance fairway is interpolated throughout
the model along the model horizons, in turn making the seismic con-
straints somewhat dependent on the model.

The final step in the seismic inversion is to define a transition fre-

quency, below which information in the final inversion is provided
by the model. The model can be computed using logs from a single
well or many, depending upon the circumstances. Above the transi-

20 CSEG Recorder December, 2000

 FEATURE ARTICLE Continued
EFFECT OF WELL CONTROL ON CONSTRAINED SPARSE SPIKE SEISMIC INVERSION

tion frequency, information in the inversion comes from the con- tive inversion containing only frequencies at and above the transi-
strained sparse spike output as described above. tion frequency. Sections along the key line 40 are shown in Figures 4
and 5 for the relative and full inversions, respectively. Note that the
relative inversion does indicate a build-up and evidence of local low
RESULTS
impedances. The high impedance anhydrites lapping up against the
Data reef are strong indicators of structure. The build-up itself would cer-
The project data consisted of 100 lines, each with 90 CMP’s. The tainly be evident from the seismic reflection data too, although the
bin size was 40 * 40 m. Time migration before stack had been done interpretation of the low impedance layers would be less so. Even
and care taken to preserve true
amplitudes. Nineteen vertical wells,
each with sonic and density logs
were used and porosity logs were
available for many. Several horizon-
tal wells were also available for use
as quality control. The key interpret-
ed horizon was the “Reef” event
which was picked along the top of
the build-up until the salt was
encountered whereupon the salt
base was followed. Other key hori-
zons were the shale carbonate inter-
GEDCO
face above the reef, the reef platform
and the tops of both salts. The inter-
preted horizons along the key line 40 NEW 2/3 Ad
are shown in Figure 2.
Film Supplied
Modelling Using Only an Off-Reef
Well
Initially, we do not assume a
knowledge of either the reef or even
its event horizon. We use only the
distant off-reef well, “A”. This well
encountered salt and platform facies
with no reef build-up at all. The salt
in this initial model was allowed to
collapse, not against a reef but erro-
neously against the regional plat-
form. With this incomplete
information, a 3D impedance model
was constructed, a line through
which is shown in Figure 3.

The wavelet was estimated from

a comparison of the well “A”
impedance log and seismic traces
near the “A” location. After an
inspection of the amplitude spec-
trum of the wavelet, the transition
frequency was set at 10 Hz where the
wavelet power rolled off. In the final
inversion, frequencies below the
transition frequency are taken from
the model. We also produced a rela-

December, 2000 CSEG Recorder 21

FEATURE ARTICLE Continued
EFFECT OF WELL CONTROL ON CONSTRAINED SPARSE SPIKE SEISMIC INVERSION

though the reef has been placed in an incorrect off-reef geologic set- of a reef horizon is also now assumed. We used the same reef hori-
ting, the existence of anomalous activity in the full inversion (Figure zon as shown in Figure 2. In the early stages of development, it
5) is still evident. might be more uncertain. This will not affect our conclusions. We
expect to fine tune the horizons as new wells become available, thus
uniquely positioning key layers in time and space. The updated
model is shown in Figure 6. Since only a single well lies within the
reef core, the on-reef impedance is laterally homogeneous except for
structure effects. The corresponding full and relative inversions are
shown in Figures 7 and 8. Now the geologic setting has been ren-
dered with considerable accuracy. This with just two wells. The
meaning of the fine seismic detail has been made clear and more
reliable judgements about the locations of future wells made. In this
way sweet spots within the reef can be detected and exploited.

Figure 2: The original interpreted horizons along the key line 40. Lateral salts
dissolve against the reef. Initially, we do not assume existence of the reef hori-
zon. Inversion analysis is done using only the off-reef well,“A”.

Figure 4: Relative (seismic frequencies only) inversion using constraints inter-

polated along the solid model defined only by the off-reef well “A”and assum-
ing no knowledge of the reef horizon. Both a build-up and anomalous low
impedance activity are indicated near the middle of the section.

Figure 3: Line 40 through the 3D impedance model constructed from only a

knowledge of the off-reef well,“A”.The reef horizon has been left out, resulting
in the salt erroneously collapsing upon the platform. No reef facies have been
modelled.

Modelling using the Off-Reef Well and One On-Reef Well

Given the above results, an explorationist might choose to
hypothesize the existence of a reef, interpret a potential reef horizon
and then place a pseudo well within it. The pseudo well would con-
tain reef-like impedances. The model could then be re-built, and the Figure 5: The full inversion using the no-reef model of Figure 3. Although the
inversion re-run to test this reef hypothesis. Geologic settings high frequency component from the inverted seismic data has been placed in
notwithstanding, most geophysicists would be drawn to the activi- the wrong geologic setting, a build-up and regions of relatively low impedance
ty between CMP’s 42 and 52 and indeed well “G” was drilled near- are still indicated.
by. We now assume existence of each of the wells “A” and “G”, and
then re-build the model and re-compute the inversion. The existence We note that the relative inversions in Figures 8 and 4 are virtu-
ally identical. They would only have been different if new well infor-

22 CSEG Recorder December, 2000

 FEATURE ARTICLE Continued
EFFECT OF WELL CONTROL ON CONSTRAINED SPARSE SPIKE SEISMIC INVERSION

mation had revised the expectation of the range of impedance val- To answer these types of questions, apron impedance maps can be
ues in the region of interest. Then, the constraint fairway defining made. This was done here, by copying the reef apron event below
the constrained sparse spike solution space would have been updat- itself, re-building the model and then measuring the average
ed. Other differences could have resulted from re-estimating the impedance in the layer between the reef and copied-reef horizons.
wavelet using the “G” impedance log. We did not do this since well Since this layer is conformable with the apron horizon, the average
“A” was one of the few to encounter the lateral salt. This resulted in of the inversion within it is a measure of the lateral variation in aver-
a highly characteristic impedance log and less non-uniqueness in age apron impedance. We first mapped the inversion constrained by
the estimation of the wavelet. a model containing logs from well “A” plus all the on-reef wells.
Figure 9 is this map for a sub-apron layer 4 ms thick, starting at the
Reef event. Apparently, regions of low impedance surround the reef.
However, on-reef wells have little to say about the apron facies.
Except for some knowledge of structure, we have essentially
returned to our single-well scenario. Furthermore, that single well
(“A”) did not encounter near-reef apron facies.

Figure 6: A section through the model impedance cube designed from the off-
reef well “A”and the on-reef well,“G”. A knowledge of the reef horizon has also
been assumed and added to the model.The reef horizon and the model would
normally be updated with each well drilled.

Figure 8: Line 40 through the relative inversion with low frequencies

defined by the model constructed from wells “A” and “G”. Note the close simi-
larity between this relative inversion and that in Figure 4, obtained from only
a knowledge of well “A”.

Rather than beginning to add wells one at a time, we look ahead

to the present day mature environment and compute a comparable
map for an inversion computed from all the wells. It is shown in
Figure 10. Porosity development is evident to the south below well
P, to the west at well L and to the north and northwest. The good
porosity near the reef core is in agreement with horizontal well
observations. By comparing Figures 9 and 10, it can be seen that the
Figure 7: Line 40 through the full inversion with low frequencies defined addition of off-reef wells has added information and highlighted the
by the model constructed from wells “A” and “G”. The high frequency seismic windward north east as the most prospective area. It is true that we
information has now been placed into a very reasonable geologic setting. The cannot know the low frequency component of the impedance pre-
reef core is seen to be inhomogeneous with areas of enhanced low impedance cisely until a well is drilled. But as more wells are drilled in the pro-
(porosity). ject area, the model can be upgraded and confidence increased.

Modelling using Well “A” and All the Reef Wells

For completeness, we have shown in Figure 11 the model frame-
We now assume some passage of time. In our exploration model work corresponding to a section through the final 3D model. The
several on-reef wells have been drilled. After each, the reef horizon broad colours are meant to indicate the layer boundaries. The struc-
was re-interpreted and new impedance logs were used to upgrade ture of the finer banding internal to each layer is indicative of the
the model. Subsequent inversions were used as an aid in determin- stratigraphy. In most layers, the stratigraphy is conformable with the
ing future drilling locations. Eventually, interest was directed to the bounding horizons. Exceptions are the Salt layer which is con-
apron facies. Were there regions of enhanced porosity in the apron? formable only with its upper boundary and the reef interior which

December, 2000 CSEG Recorder 23

FEATURE ARTICLE Continued
EFFECT OF WELL CONTROL ON CONSTRAINED SPARSE SPIKE SEISMIC INVERSION

has been defined to respect both the reef structure and the observed
push-down. Figure 12 is the same section through the final
impedance model. Lateral variations in impedance arise from both
structural and stratigraphic effects. At CMP’s where no log is pre-
sent, nearby logs were stretched or squeezed on a layer-by-layer
basis to account for structure and then subjected to a weighted aver-
age to arrive at the model impedance. A simple distance-weighted
average sufficed although the weights could have been set arbitrar-
ily as required. The final full inversion using all the wells is shown
in Figure 13.

Figure 11: A stratigraphic section through the 3-D geologic model.The colours
indicate the different layers defined by the interpreted horizons and the
framework table.The banding within each layer describes the layer stratigra-
phy which governs how lateral well log interpolation is done.

Figure 9: Average inversion impedance in the first 4 ms below the reef hori-
zon using the model defined only from the distant well “A” and all of the on-
reef wells.There are several areas around the core which might be prospective
for low porosity.

Figure 12: Line 40 through the 3D geologic model. The impedance logs have
been interpolated according to the framework defined by the interpreted hori-
zons. None of the fine detail within the reef will be transmitted to the final
inversion. The model provides only low frequencies so that the transformed
seismic can be placed in a geologic context.

Figure 13: Line 40 through the final inversion cube.This result has been com-
Figure 10: Average impedance in a layer 4 ms thick immediately below the
puted using a model defined by all the vertical wells. All of the detail within
Reef horizon using a model constructed from all 19 vertical wells. Porosity
the reef has come from the input seismic data. The model has simply provid-
development is evident to the west, south and northwest of the reef core.
ed a reef-like geologic setting.

24 CSEG Recorder December, 2000

 FEATURE ARTICLE Continued
EFFECT OF WELL CONTROL ON CONSTRAINED SPARSE SPIKE SEISMIC INVERSION

CONCLUSIONS and DISCUSSION were necessary to define the basic geology. As off-reef wells were
We have demonstrated that seismic inversion can be useful in made available, their contributions to the net porosity could be
both exploration and production environments. The high frequency accounted for and regions of enhanced porosity located. It is not
seismic band component of the inversion remains remarkably con- always possible that extensive drilling will have resulted in a thor-
stant from rank exploration to mature production. The reason for ough a priori understanding of the play. But when such knowledge
this is simple. The inversion result is constrained only to lie within is available, it should be used to build an accurate and detailed
the impedance constraint boundaries. It is not constrained to be sim- model. We have used the abundance of wells not to force the answer
ilar to any particular impedance log, even when inverting at that in the seismic band but to improve our model and the knowledge of
CMP corresponding to the well location. Therefore, a comparison the low frequencies which we extract from it. This idea is illustrated
with log impedances is a very powerful QC tool. Significant changes in Figure 14.
could only result from an improvement and general tightening of
the sparse spike constraints and re-estimation of the wavelet. These It is perhaps worth emphasizing the importance of formally com-
can be implemented as more wells are drilled and there is a better puting both the relative and full inversions explicitly. Whether the
understanding of the expected range of impedances. The model and play in question is new or mature, there is always the danger of a
the full inversion are dramatically improved when wells have been mis-interpretation. The relative inversion is mostly immune to this
drilled through all representative facies. In this case, only two wells condition. When horizons are at all suspect, the constraints are sim-
ply relaxed a little. It is a different story in the case of the full inver-

TDROV + Stratimagic =
Reservoir Description
www.arciscorp.com

Structure Map
ARCIS Processing is pleased to Structure map of top reef
announce Mr. Alan Richards has using Stratimagic horizon
propagator tool
joined the division as of November Reservoir Quality
1, 2000. Mr. Richards has over Reservoir quality from
twenty-six years of experience pro- TDROV impedance
cessing 2D data, large 3D projects
and structural data sets in areas Surface Attributes
both domestic and international. Azimuth surface attribute
Alan will hold the position of pro- from Stratimagic
ject management and Technical
support and will be involved with
2D/3D land projects both structural CGG Canada Services Ltd.
and stratigraphic. Alan will further Seismic Facies Suite 700, 404-6th Avenue S.W.
enhance Arcis Processing ongoing Seismic facies of reef Calgary, Alberta T2P 0R9
using Stratimagic neural (403) 266-1011
commitment to increasing its tech- network technology [email protected]
nical capability and experience with
key individuals. Alan can be
reached at 298-4786 or e-mail at Processing and Reservoir Services
[email protected].

December, 2000 CSEG Recorder 25

FEATURE ARTICLE Continued
EFFECT OF WELL CONTROL ON CONSTRAINED SPARSE SPIKE SEISMIC INVERSION

sion. Mis-interpretations will change the model and the way in

which the low frequencies contribute. Given the large amount of
power in the low frequencies of logs and inversions, the effects can
be dramatic (compare Figure 5 and 7). Fortunately, these problems
can be easily ameliorated or a least risked by a simple comparison of
the relative and full inversions.

JOHN PENDREL BIOGRAPHY

John Pendrel is Chief Geophysicist

with Jason Geosystems Canada. In this
position he is responsible for applying
proprietary, leading edge technology to
problems in reservoir analysis and char-
acterization. From 1981 to 1995, he was Figure 14: Detail in the full version comes from the transformed seismic band.
Sr. Geophysicist and then Manager, It is usually possible to gain a few Hz on the high side as illustrated. Low fre-
Geophysical Technology with Gulf quencies from the geologic model fill the band below the seismic and provide
Canada Resources in Calgary. He began his career in the oil a geologic setting. In the relative inversion, the Model Band is not included.
industry in 1978 with Gulf Science and Technology Company
in Pittsburgh, PA, the research arm of the former Gulf Oil.

John’s academic career included a B.Sc. at The University ACKNOWLEDGEMENTS

of Saskatchewan (1968), and an M.Sc. from The University of
Saskatchewan, Saskatoon, (1972) where he did research in The authors gratefully acknowledge Husky Oil Ltd. and Mobil
auroral magnetic fields. John also holds a Saskatchewan Oil Canada for their generous donation of these data.
Class “A” teacher’s certificate. He obtained a Ph.D. (geo-
physics) from York University, Toronto in 1978 where his
interests were in two-dimensional time series and spectral
analysis. REFERENCES
Debeye, H.W.J., van Riel, P. (1990): Lp-norm deconvolution,
John’s early research was in the areas of pattern recogni-
Geophys. Prosp. 38,381
tion and principal components analysis. More recently he has
done applied research and published papers in seismic inver- Pendrel, J.V., van Riel, P. (1997): Methodology for Seismic
sion, geostatistical analysis and AVO. Away from work he Inversion - A Western Canadian Reef Example, CSEG Recorder,
plays on an ice hockey team and volunteers with a local youth 22, #5, p.5
group, The Calgary Stampede Showband.

Jason Geosystems, 410 505 8th Ave. SW,

Calgary, AB, Canada T2P 1G2
Tel: (403) 263-3340 Fax: (403) 263-3285
Paul van Riel Biography

Paul van Riel is a founder & managing director of Jason

Geosystems B.V. in The Netherlands. Prior to that, he was
founder & project leader at the Delft University Princeps oil
company consortium. Paul’s interests lie in the field of seismic
inversion and reservoir characterization. In his present role he
has responsibilities for implementing these processes world-
wide.

Van Riel, Paul.

Jason Geosystems B.V.
Plaza Bldg. Weena 598
3012 CN Rotterdam
The Netherlands
31-10-280-1515 (V); 31-10-280-1511 (F); pvanriel@ jason.nl

26 CSEG Recorder December, 2000

 GRAPEVINE EVENTS CALENDAR
CSEG Technical Luncheon
January 17, 2001 Westin Hotel, Calgary
February 5, 2001 Westin Hotel, Calgary

PEICE
January 26, 2001 Luncheon ‘Confronting People’
February 5, 2001 Sulphur Recovery Fundamentals
February 6-7, 2001 Introduction to the Petroleum Industry
February 16, 2001 Luncheon ‘Family’
For more information or to register call the Petroleum Institute for Continuing Education (PEIC) at visit the organizers
(403) 284-1250 or register on line at www.montnet.com

GeoQuest Forum – Technology Conference

April 4-6, 2001

PETRO CHEM EXPO 2001 - Petrochemical, Petroleum, Pipeline, Refining, Equipment and Services Show
September 19-20, 2001 Westerner Park, Red Deer, Alberta
For more information on workshops and conference please contact: Unni Soelberg-Claridge,
Ste. #100, 4500 - 16 Avenue N.W., Calgary, Alberta T3B 0M6
Phone: (403) 247-0321, Fax: (403) 247-4811, email: [email protected], website: www.eventplant.net.

ASCOPE 2001 - Conference and Exhibition

November 5-7, 2001 Putra World Trade Centre, Kuala Lumpur

To ensure timely publication of your materials, all materials must be sent two months in advance of any critical dates.

CORRECTION
The following two figures were inadvertently omitted in the December, 2000 feature article “Effect of Well Control
on Constrained Sparse Spike Seismic Inversion”. Figures 2 and 12 were incorrectly shown twice.

A complete and correct version of the paper can be obtained by e-mailing John Pendrel ([email protected]) .
Please specify whether you would like a digital or hard copy version. The paper can also be downloaded from Jason’s
ftp site (calgary.jasongeo.com; userid: recorder; password: booboo). Our apologies.
Editors

Figure 3 Figure 13

50 CSEG Recorder January, 2001