You are on page 1of 8

Accurate stratigraphic prediction

from seismic
Matt Hall & Eric Trouillot, Landmark Graphics, Calgary, Canada

Introduction hour. The horizons were interpolated and smoothed and are
used for building a velocity model for depth conversions,
Interpretation of subsurface geology is greatly enhanced by and for guiding the attribute extraction algorithms along
3-D seismic data, and this accounts for its ubiquity in today’s structure. The reservoir interval is interpreted as a tidal sand-
search for hydrocarbons. Seismic interpretation has two ridge complex deposited during a lowstand and transgres-
fundamental disciplines at its core: seismic geomorphology sion in the early part of the Mississippian. Offshore shales are
and seismic stratigraphy. Plan views allow the interpreter to unconformably overlain by heterolithic lower- and then
apply principles of geomorphology, based on analogies with sandy upper-shoreface sandbodies, which are subsequently
modern sedimentary systems, to interpret depositional en- partially reworked and redeposited as tidal sand-ridges
vironments and even predict facies distributions. On the during a period of transgression. These sands are then con-
other hand, section views require stratigraphic interpretation formably overlain by the ensuing highstand shales. The sand
and give insight into stratal architecture and the temporal ridges are around 20–40 km long, 3–7 km wide, and 20–30 m
development of the depositional system. Both approaches thick, and these are the subject of our investigation.
give insight into the geological processes that formed the
hydrocarbon play (eg Posamentier 2003).
Seismic Stratigraphy
This paper explores some of the techniques available today
to take this insight to deeper levels than ever. The wide- Modelling
spread availability of powerful computers and high-
Seismic volumes are complex volumes of reflections, inter-
performance graphics have opened new doors to the seismic
ference and tuning effects. It can be difficult to know what to
interpreter, allowing him or her to test and tweak parameters
expect a given sandbody or stratigraphic geometry to look
to fine-tune an attribute, and to make unique and
like in the seismic. Seismic modelling can give the interpreter
compelling displays to help them tell their story. Many of
this insight, and allow him or her to try multiple iterations in
these techniques are especially potent if time is short: a quick,
order to best match a known response. Stratigraphic inter-
simple interpretation of a couple of easily-interpreted hori-
pretations can then be made in areas where there is little or
zons is enough input to guide a rich assemblage of sophisti-
no well control.
cated algorithms through the dataset, revealing hidden
patterns and subtle features in a matter of minutes. If more A workflow for seismic modelling to illuminate stratigraphic
time is available, techniques such as 2-D seismic modelling relationships is illustrated in Figures 2–6.
can shed light on difficult interpretation problems and give
more confidence that he or she is indeed interpreting geolog- Volume seismic attributes
ical features and not geophysical artefacts. We present exam-
ples of both of these approaches in this paper. Attributes drawn from a group of seismic traces in a 3-D
volume are capable of representing complex three-dimen-
The example seismic dataset, shown in Figure 1, is a sional reflection characteristics and relationships. For
migrated full-stack, filtered, 16-bit floating-point amplitude example, it is possible to generate a volume representing the
volume from a producing oilfield in western Canada. We amount of dip or degree of parallelism displayed by the
interpreted three horizons with sparse seed lines and an reflectors in the seismic. These are calculated directly from
autopicking tool; all three horizons were finished in under an the seismic traces, with no interpretive input (though the

Figure 1. Dip and strike seismic lines across the sand ridge feature that forms the reservoir interval. This subtle feature is discernible on the map on the right
(outlined) and has a distinctive seismic expression that reflects its complex internal character and reveals a challenging reservoir characterization problem. A
carbonate platform that existed before deposition of the reservoir sediments is a dominant feature of the map. The top of the key reservoir zone is shown as a green
seismic surface. The zone of interest is enveloped by the blue and orange seismic surfaces which were used for calculation of interval attributes and isochores. On
the map, red is shallow, blue is deep; on the seismic sections, black is a peak indicating a positive acoustic impedance contrast. [The images are from PowerView.]

Continued on Page 31

30 CSEG RECORDER March 2004


Focus Article Cont’d
Accurate stratigraphic prediction from seismic
Continued from Page 30

quality of the results are improved by the use of a three-dimen- developed reservoir interval, a thinner reservoir interval, more
sional velocity model, which normally would incorporate some heterolithic reservoir, a shaled-out reservoir section, and an
surface interpretations to guide the velocity fields). Other such eroded, unconformable section. These waveforms represent 5
attributes quantify aspects of the reflectors such as continuity, discrete classes. Every trace in the seismic volume is then class-
hummockiness, divergence and azimuth. It is even possible to ified according to which of the reference waveforms it most
combine and reprocess these attributes to produce new ones, such closely resembles; there is also a ‘zeroth’ or reject class for traces
as shaded relief images (Barnes 2003; see example in Figure 7a). which resemble none of the references. The result is a map of
These attributes all represent aspects of apparent palaeotopog- wavelet classes in which each class is represented by an integer.
raphy, which can aid in geological interpretation. This is espe-
cially true when the results are combined with other images. Two The second approach, supervised classification, automatically
effective ways of achieving this are semi-transparent overlays calculates a reference wavelet from the data before assigning
(Figure 8) and colour blending (Figure 9). As with all 3-D seismic classes. This is achieved by an iterative statistical analysis of
datasets, using these visualization techniques in a 3-D space is a subset of the data, the results of which are then applied to
more powerful still. the entire dataset. As before, the result is a map of waveform
classes, such as those shown in Figure 10, with each class
Waveform Classification being represented by an integer. The classes are organized
according to relative similarity, so class 1 more closely resembles
An important component of seismic stratigraphy is recognizing class 2 than it does class 5. This approach is especially useful if
systematic variations in the character of a seismic reflector or set there are no wells in the area of interest, or if the interpretation is
of reflectors. Such variations might reflect changes in stratal in the early stages and no firm correlations with wells have yet
geometry or lithologic stacking patterns. This can be a powerful been established.
way of elucidating subsurface stratigraphy qualitatively and,
when coupled with seismic modelling, quantitatively (eg Seismic Geomorphology
Williamson 2003). An important feature of waveform classifica-
tion is that wavelets are compared by shape, not by amplitude. Spectral decomposition
This can mean that waveform classification is less affected by
tuning effects and porosity variation, for example, than other Ordinary 3-D seismic data typically has a 60–80 Hz bandwidth, so
types of seismic attribute. it contains energy reflected from the subsurface at a wide range of
frequencies, all of which are compounded in a typical seismic
There are two fundamental approaches to waveform classifica- volume. In certain circumstances, especially subtle stratigraphic
tion: supervised and unsupervised. Supervised approaches plays, it may be helpful to see the amplitude (or phase) of reflec-
require the interpreter to select a time-window (usually centred tions at particular frequencies. These tuned amplitudes may tell
on a seismic horizon) and provide any number of reference wave- the interpreter something about the physical spacing of the
forms (as an inline/crossline location). These are often chosen to acoustic impedance contrasts in the image—in other words the
be representative of distinct lithologies at nearby wells. For bedding thicknesses (eg Partyka et al., 1999). There are also exam-
example, one might provide the location of an especially thickly- ples of mapping the attenuation of high-frequency signal by
hydrocarbons, allowing their direct detection (eg
Castagna et al., 2003). Spectral decomposition
analysis allows the interpreter to quantify amplitude
variation with frequency, and thereby gain insight
into the distribution of different stratigraphic pack-
ages and/or hydrocarbons.

There are three key spectral decomposition work-


flows. The first is to process a discrete window
around a very smooth seismic horizon interpreta-
tion, transforming the amplitude or phase data into
the frequency domain (in other words, a new
volume results, with frequency represented by the z-
direction; a typical volume might contain 100 slices,
representing amplitude or phase at 1–100 Hz). Such
a volume is called a ‘tuning cube’. There are two
helpful ways of visualizing this dataset. The first is
simply to animate through the volume, exactly as
Figure 2. This seismic section shows wells that penetrate a stratigraphic reservoir. On either side one would animate through timeslices. This yields
of the well bore are displayed a sonic curve and the corresponding synthetic. The geology has been images similar to those in Figure 11a–11c. The second
tied to the seismic aligning synthetic character with seismic signatures. The 2 wells on the right visualization technique is to blend three timeslice
clearly show matching seismic and synthetic reflections, however the well on the left does not, even images at a time, using the techniques illustrated in
though it seems to penetrate the reservoir (in circles). [Image from SeisVision, synthetics from
LogM.] Figures 11d and 11e. Such blended images can

Continued on Page 32

March 2004 CSEG RECORDER 31


Focus Article Cont’d
Accurate stratigraphic prediction from seismic
Continued from Page 31

as those established as useful or revealing in the tuning cube


workflow. The reason for using a flattened volume is to give
insight into the geological evolution of the interval. Animating
through a tuned volume shows the same reservoir subtleties
revealed in the first workflow, but now with the added dimension
of geological time.

The third workflow gives a valuable quantitative estimation of


bedding thickness in the analysis window, without the need for
accurate seismic interpretation or phase concerns. Figure 12
shows an example of such a thickness estimation map, derived
from the first spectral peak frequency attribute. This attribute is
simply the frequency of the first peak in the amplitude spectrum.
This is an important number, because it is exactly half the value of
the spacing of the interference ‘notches’ in the spectrum. The
notches are, in turn, spaced in inverse relation to bedding thick-
Figure 3. This is a stratigraphic geological cross section of the same 3 wells we see ness in the analysis window. See Partyka et al. (1999) for a full
in the seismic section from Figure 2. The sonic curves are displayed in the tracks explanation of the physics of this analysis. A simple horizon
and between the wells is an interpolated colour fill based on the sonic curve values calculation can therefore transform the first spectral peak
highlighting the slowness of the data in microseconds per metre. Reds are low frequency attribute into time thickness, or, given the interval
velocity and blues higher velocity. The low velocity sand body shown in this figure
tapers towards the left well where we had no synthetic reflection in Figure 2. velocity of the zone, vertical thickness. Since this is an inde-
[Image from GESXplorer Cross Section.] pendent measure of bedding thickness, this workflow is poten-
tially an important risk reduction tool for the explorationist. The
approach is especially powerful when coupled with synthetic
modelling and/or crossplotting the predicted thicknesses with
provide an attractive and data-rich snapshot of very subtle reser-
actual vertical thicknesses as measured in offset wells.
voir characteristics. This workflow should help establish which
features of the reservoir interval are brought out by which These workflows, well-proven by several years of application in
frequencies. the industry, can give valuable insight into the internal organiza-
tion of the reservoir interval. We have found the images and
The second workflow is to leave the data in the time domain, but
animations particularly effective communication tools, particu-
to process a series of time windows (a ‘running window’).
larly in plays with strong stratigraphic components such as
Typically, a flattened seismic volume is used, with the same
channel sands and deep marine gravity deposits.
guiding horizon as a datum. The result is a flattened volume
which is tuned to a specific frequency or set of frequencies, such

Figure 4. To get a clear understanding of the correlation between the geological stratigraphy and the seismic, modeling can be a key element. The middle image (Step 2) in
Figure 4 is a seismic trace model generated from the cross section in Figure 3. Again this model shows a leftward tapering of the same sand body identified in Figure 3. By
cropping a representation of this model and overlaying it on the seismic in the zone of interest (Step 3) you can visually validate the correlation between the geology and
the seismic; thus giving you an invaluable confidence in understanding the stratigraphic seismic signatures that are present. [Images from SeisVision, model created in
LogM.]

Continued on Page 33

32 CSEG RECORDER March 2004


Focus Article Cont’d
Accurate stratigraphic prediction from seismic
Continued from Page 32

by particular examples. The ability to view many


maps simultaneously, with a range of different colour-
bars and zoom-levels, can be a significant aid to inter-
pretation. An example is shown in Figure 7, in which
six maps are compared side-by-side. This allows the
interpreter to quickly spot correlations or discrepan-
cies between attributes which may have stratigraphic
explanations. Opacity overlays and blended images
may also help with interpretation, as previously illus-
trated in Figures 7–11.

Another technique for multi-attribute analysis, and


for correlating seismic attributes with reservoir prop-
erties or with each other, is crossplotting. An in-depth
review of this technique is beyond the scope of this
paper, but many examples can be found in the litera-
Figure 5. Well number 1 (the left-most well in Figure 2) still leaves a lot of uncertainty regarding ture. Establishing a relationship with a reservoir prop-
where the last synthetic reflector sits on the upper part of the reservoir, leaving little indication of erty is especially useful, since it allows the interpreter
where the sandbody may lie in the seismic. Is this because our sonic curve is incorrect, because we to model and predict the property directly from the
are on the edge of the reservoir? What if we increased the sand body thickness in the well—what
synthetic response would that give us? Here we have taken well number 1 and copied it 4 more seismic, with a known and quantifiable level of
times in a cross section, displaying the sonic curve and the synthetic for each well. In the middle certainty. This makes the analysis of exploration risk
well we have added about 20 metres below the top of the sandbody, having started at 5 metres first, considerably easier.
and then increased to 10 m, 15 m and finally 20 m until we got the synthetic response expected
which indicates a significant positive reflection. We can now take this cross section and model it to
generate an interpolated seismic trace model and compare it to the real seismic. [Image from LogM The Power of Volume Interpretation
Stratigraphic Model Builder.]
While this paper is focused on two dimensional views
of the seismic data, volume visualization and inter-
Instantaneous seismic attributes pretation is now well-established as an interpretation discipline.
There is an almost bewildering number of instantaneous seismic A 3-D seismic volume has not been truly interpreted until it has
attributes, each of which describes some aspect of a single trace at been volume interpreted (Figure 13). Despite this, it remains curi-
a single time or time-window. Examples include simple attributes ously ‘niche’ or somehow ‘advanced’ in the Canadian geoscience
such as phase or reflection strength, and more sophisticated community (but that’s another story!). Two basic aspects of
interval statistics such as number of zero-crossings or total time volume interpretation make it an important tool for the strati-
thickness above a given amplitude. Maps of these attributes can graphic investigator: rapid volume interpretation and surface
often be interpreted geologically, with some features drawn out visualization.

Figure 6. Step 2 is the result of the seismic trace model generated from the cross section in Figure 5. We can see the sand wedge the model has created right where we have
added 20 metres of sand in the zone of interest from the middle well. By cropping a representation of this model, capturing the sand wedge and overlaying it on the seismic
in the zone of interest (step 3) you can align the correlation polygon by matching the strong upper reflectors from the model with the strong reflectors on the seismic in the
top part of the reservoir. As a result we can see that the sand wedge matches the intermittent seismic anomalies reflecting the sand body of the reservoir. This workflow
scenario, using ‘what if’ approaches, validates the stratigraphic relationships of the geology with the seismic anomalies of a potential target. These results are invaluable in
helping an interpreter feel confident of his or her interpretation. [Images from SeisVision, model created in LogM.]

Continued on Page 34

March 2004 CSEG RECORDER 33


Focus Article Cont’d
Accurate stratigraphic prediction from seismic
Continued from Page 33

Figure 7. Multi-panel seismic attribute analysis. Commonly, the interpreter generates a large number of seismic attributes, most of which are displayed as maps. These can
be cumbersome and time-consuming to compare, but views like this one can make the job easier. The maps can be zoomed and panned together or individually, and each has
its own colour-bar. a) Shaded relief attribute: a 3-D view generated directly from the seismic, without any interpretation; b) an isochore map, the result of subtracting one
horizon from another and converting the result to thickness with interval velocity calculations—produced in a single step from the original time horizons; c) a timeslice
through the amplitude volume; d) a waveform classification map showing seismic facies tracts—each colour represents a discrete class of waveform shapes based on the result
of a standard K-means test; e) a red-green-blue blended image showing the amplitude responses at three different frequencies in a single view; f) a bedding thickness predic-
tion map generated from the first spectral peak frequency attribute; g) a vertical section through the seismic amplitude volume showing the interval of interest along the
green Top Reservoir horizon. The blue and orange horizons were used for calculating interval attributes. [The tools used are PostStack Shaded Relief, PAL Waveform
Classifier, SpecDecomp, PowerCalculator, and the image is from PowerView.]

Figure 8. Red-green-blue blended image showing three seismic volumes simulta- Figure 9. Comparison of two seismic volumes. A black-white seismic amplitude
neously. Red represents the Hummockiness attribute, green is Parallelism, and image is shown, with a seismic attribute volume overlain. The attribute is Dip
blue is Dip. The resulting colours represent the varying reflection patterns in the and is displayed with the scale shown to the right of the image. No interpreted
seismic. They could assist in solving correlation problems, or qualitatively horizons are provided to the algorithm, but a three-dimensional velocity model is
compare seismic facies. [The volumes are PostStack reflection pattern attributes, used. In this display, the Dip volume is semi-transparent, allowing the interpreter
and the image is from PowerView.] to see both volumes at the same time. This can assist with resolving correlation
issues and with qualitative stratigraphic analysis. [Dip is a PostStack reflection
pattern attribute, and the image is from PowerView.]

Continued on Page 35

34 CSEG RECORDER February 2004


Focus Article Cont’d
Accurate stratigraphic prediction from seismic
Continued from Page 34

Figure 10. Waveform classification analysis for a 40 ms window around the Top Reservoir surface using a standard K-means testing algorithm for six classes. (a) shows a
horizon-guided waveform classification map. Each colour represents a discrete class of trace shapes, based on a statistical classification of the analysis window in every trace.
Like colours represent like wavelets. The sand ridge complex is clearly differentiated from other areas of the map, with some internal differentiation also visible. (b) again
shows the classification map, but semi-transparent and overlain on an attribute representing the confidence with which the waveforms were classified, and corresponds to
the probability of the given trace being in the assigned class. Darker areas were classified with lower confidence. (c) shows the same classification map, again semi-trans-
parent, but this time overlain on a timeslice from the amplitude volume. Views like this can help the interpreter distinguish seismic facies variations from other effects and
data quality variation. [The tool used is PAL Waveform Classifier, and the images are from PowerView.]

Figure 11. Spectral decomposition analysis using blended map images. The analysis was performed using the discrete Fourier transform method on a 60 ms window centred
on the Top Reservoir horizon, with a Gaussian taper (required for eliminating spurious high-frequency responses from the edges of the window). (a)–(c) show amplitude at
30 Hz, 40 Hz and 50 Hz respectively. A bright response is represented by white. (d) is a red-green-blue (RGB) blended image comprising of the individual images above
displayed in red, green and blue respectively. The colours are additively combined to produce the full-spectrum image shown. For example, yellow hues indicate that the red
(30 Hz) map and the green (40 Hz) map have coincident high amplitude responses in that area. Likewise, white indicates that all three frequencies are responding.(e) is a
hue-saturation-brightness (HSB) blended image. The same three amplitude slices are represented by hue (colour), saturation (colour intensity), and brightness (or colour
value) respectively. The images are then combined to give the final map; a bright, saturated red colour indicates that all three frequencies are ‘bright’ in that area. The inter-
esting thing about these images is that while they represent the same datasets, they illustrate quite different characters of the reservoir interval. For example, the RGB blended
image seems to give a more definite sense of where the sweet-spot might be for the reservoir, whereas the HSB-blended image resolves the ‘Swiss-cheese’ seismic texture very
well. [The tool used is SpecDecomp, and the images are from PowerView.]

Continued on Page 36

March 2004 CSEG RECORDER 35


Focus Article Cont’d
Accurate stratigraphic prediction from seismic
Continued from Page 35

The ability to instantaneously track an amplitude anomaly or


apparent pinchout, and see the results in three dimensions along-
side key reference data such as well logs and fault interpretations
transforms the slow, methodical, line-by-line interpretation task
into a dynamic and much more creative process. Hunches can be
instantly verified, or disproven, and leads can be rapidly gener-
ated and investigated. More powerful still is the ability to bring
multiple seismic attributes to bear simultaneously on a problem.
Some great examples of applying these techniques to strati-
graphic problems are shown in Meyer et al. (2001) and Alexander
et al. (2001).

Surface visualization is an important tool for seismic geomor-


phological studies (for example, see Posamentier 2003). Viewing
angle, lighting direction, colour, texture, and vertical exaggera-
tion can be instantly adjusted to bring out the particular features
of a seismic horizon. Misties and other anomalies can be seen
Figure 12. Thickness prediction from spectral decomposition analysis. The
discrete Fourier transform is guided through the dataset by the Top Reservoir and corrected. Attributes can be overlain on structural surfaces
horizon which has had a very aggressive smoothing filter applied to it. As such, and visual correlations made. The interpreter can learn more in
the results are almost completely independent of interpretation. Because of this minutes of playing in 3-D than in days of looking at maps.
factors, spectral decomposition can be an important risk reduction tool in explo-
ration. The first spectral peak frequency of the data is a by-product of the decom-
position process, and, since it is inversely proportional to time thickness, can Conclusions
easily be transformed into a bedding thickness prediction map. This map clearly
shows the sand ridge running in a NE-SW orientation. The apparent decrease in A very rich set of tools is available to the seismic interpreter for
bedding thickness across the reservoir is explained by the discrete nature of the gaining insight into stratigraphy. Which tools one chooses to
depositional sandbody, in which the reservoir zone is typically 20–25 m thick, apply depend on the play type being investigated, the amount of
compared to the thickly-bedded shales of the ‘swale’ regions, which contain few
time available for analysis, the stage in the interpretation process,
significant acoustic impedance contrasts. [The tools used are SpecDecomp and
PowerCalculator, and the image is from PowerView.] the quality and nature of the available data, and the preferences,
experience and intuition of the individual. There is no formula
for the ‘right’ attribute for a given situation. In fact, our own
approach is to build as many models, look at as many attributes,
and try as many parameters as time allows. Time saved by auto-
matic surface interpretation and volume interpretation can be
put to good use by applying the available tools, letting the data
itself offer up the answers. This seems like a better use of time,
and technology, than the frustrating process of line-by-line inter-
pretation of difficult surfaces, the results of which are always
risky and uncertain.

Readers may be interested to know about the SEPM/Geological Society


conference on seismic geomorphology, in Houston, 10 & 11 February
2005. For more information, go to:
http://www.sepm.org/events/rconferencehome.htm R

References
Alexander, C, et al. (2001). The Plutonio discovery, Block 18, Angola—A3-D visuali-
zation and multiattribute approach to exploration success. The Leading Edge,
December 2001.
Barnes, A (2003). Shaded relief seismic attribute. Geophysics 68 (4), July-August
2003, p 1281–1285.
Figure 13. Volume visualization and interpretation give new insight into the
Castagna, J, S Sun & R Siegfried (2003). Instantaneous spectral analysis: Detection of
spatial arrangement of stratigraphic features and broaden the interpreter’s low-frequency shadows associated with hydrocarbons. The Leading Edge, February
perspective. This makes it possible to interpret seismic or plan well trajectories 2003.
interactively and dynamically, with input from the geophysicist, geologist and
Meyer, D, et al. (2000). Use of seismic attributes in 3-D geovolume interpretation. The
engineer, each of whom can see their own data in the same space. This allows them Leading Edge, December 2001.
to significantly reduce risk and make complex decisions with much more confi-
dence. Images like the one shown are replacing the venerable contour map as the Partyka, G, J Gridley & J Lopez (1999). Interpretational applications of spectral
decomposition in reservoir characterization. The Leading Edge, March 1999.
fundamental decision tool. See Meyer et al. (2001) for more examples. [The tools
used are GeoProbe and Wellbore Planner. Data courtesy of Seitel Inc.] Posamentier, H (2003). Integration of seismic geomorphology and seismic stratigraphy:
principles and applications. CSEG/CSPG Annual Convention Abstracts Volume,
June 2003.

Continued on Page 37

36 CSEG RECORDER March 2004


Focus Article Cont’d
Accurate stratigraphic prediction from seismic
Continued from Page 36
Williamson, A, et al. (2003). Quantitative interpretation of neural network seismic
facies—Oriente Basin, Ecuador. CSEG/CSPG Annual Convention Abstracts
Biographies
Volume, June 2003.
We would like to acknowledge Landmark Graphics for permission to
Wood, L, et al. (2000). Seismic attribute and sequence stratigraphic integration methods
for resolving reservoir geometry in San Jorge Basin, Argentina. The Leading Edge, publish this paper. PowerView, PowerCalculator, SpecDecomp,
September 2000. PostStack, GeoProbe, Wellbore Planner, SeisVision, LogM and
GESXplorer are trademarks of Landmark Graphics.

Eric Trouillot (etrouillot@ lgc.com) is


manager of business development
for GeoGraphix at Landmark
Graphics Canada. He received a
D.U.T. in Civil Engineering in France
where he practiced field engineering
before moving to Calgary in 1988.
Since then he has entered the high
tech world of software development
in the oil and gas industry, accumu-
lating over 16 years of experience
between Digi-Rule, GMA,
GeoGraphix and Landmark.

Matt Hall (mhall@lgc.com) gained


his B.Sc. degree in geology from the
University of Durham, UK, in 1993
and his Ph.D. in sedimentary geology
from the University of Manchester,
UK, in 1997. He joined Statoil as a
geologist in Stavanger, Norway,
working on the Tertiary of the North
Sea. In 2000 he moved to Canada and
joined Landmark Graphics, where he
is now responsible for business
development in prospect generation
systems.

March 2004 CSEG RECORDER 37

You might also like