jan01 seismic inversion

Upload: abbas-abdu

Post on 07-Apr-2018

231 views

Category:

Documents


0 download

TRANSCRIPT

  • 8/6/2019 Jan01 Seismic Inversion

    1/7

    FEATURE ARTICLE

    SEISMIC INVERSION THE BEST TOOL

    FOR RESERVOIR CHARACTERIZATION

    By John Pendrel, Jason Geosystems Canada

    Introduction

    The principle objective of seismic inversion is to transform

    seismic reflection data into a quantitative rock property, descrip-

    tive of the reservoir. In its most simple form, acoustic impedance

    logs are computed at each CMP. In other words, if we had drilled

    and logged wells at the CMPs, what would the impedance logs

    have looked like? Compared to working with seismic ampli-

    tudes, inversion results show higher resolution and support

    more accurate interpretations. This in turn facilitates better esti-

    mations of reservoir properties such as porosity and net pay. An

    additional benefit is that interpretation efficiency is greatly

    improved, more than offsetting the time spent in the inversion

    process. It will also be demonstrated below that inversions make

    possible the formal estimation of uncertainty and risk.

    In various forms, seismic inversion has been around as a

    viable exploration tool for about 25 years. It was in common use

    in 1977 when the writer joined Gulf Oil Companys research lab

    in Pittsburgh. During this time, it has suffered through a severe

    identity crisis, having been alternately praised and vilified. Is it

    just coloured seismic with a 90 deg phase rotation or a unique

    window into the reservoir? Should we use well logs as a priori

    information in the inversion process or would that be telling us

    the answer? When should we use inversion and when should we

    not? And what type: blocky, model-based, sparse spike? At

    Jason Geosystems we offer all these algorithms - and a few more.

    In the following, I will briefly discuss them in a somewhat quali-tative manner, keeping the equations to a minimum. After all, it

    is drilling results which count and against which any algorithm

    will ultimately be measured.

    The Inversion Method

    The modern era of seismic inversion started in the early 80s

    when algorithms which accounted for both wavelet amplitude

    and phase spectra started to appear. Previously, it had been

    assumed that each and every sample in a seismic trace represent-

    ed a unique reflection coefficient, unrelated to any other. This

    was the so-called recursive method. The trace integration method

    was a popular approximation. At the heart of any of the newer

    generation algorithms is some sort of mathematics - usually in

    the form of an objective function to be minimized. Here I will

    write that objective function, in words, rather than symbols and

    claim that it is valid for all modern inversion algorithms - blocky,

    model-based or whatever.

    Obj = Keep it Simple + Match the Seismic + Match the Logs (1)

    Lets look at each of these terms, starting with the seismic. It

    says that the inversion impedances should imply synthetics

    which match the seismic. This is usually (but not always) done

    in a least squares sense. Invoking this term also implies a knowledge of the seismic wavelet. Otherwise, synthetics could not be

    made. At this point, life would be good except for one thing. The

    wavelet is band-limited and any broadband impedances tha

    would be obtained using the seismic term only would be non-

    unique. Said another way, there is more than one inversion

    impedance solution which, when converted to reflection coeffi

    cients and convolved with the wavelet would match the seismic

    In fact there are an unlimited number of such inversions. Worse

    such one-term objective functions could even become unstable as

    the algorithm relentlessly crunches on, its sole mission in life

    being to match the seismic, noise and all, to the last decima

    place.

    Enter the simple term. Every algorithm has one. It could no

    care less about matching the seismic data, preferring instead to

    create an inversion impedance log with as few reflection coeffi-

    cients as possible. Different algorithms invoke simplicity in dif

    ferent ways. Some do it entirely outside of the objective function

    by an a priori blocky assumption. In our most-used algorithm

    it is placed inside the objective function in the form of an L-1

    norm on the reflection coefficients themselves. This is advanta

    geous since it locates all the important terms together where thei

    interactions can easily be controlled. How much simplicity i

    best? The answer is project-dependent and will be different fo

    example, for hard-contrast carbonates and soft-contrast sandand shales. We exercise control by multiplying the seismic term

    by a constant. When the constant is high, complexity rule

    When it becomes smaller, the inversion becomes more simple

    with a sparser set of reflection coefficients. I will now attach the

    name sparse spike to these types of inversions.

    What about the Match the Logs term? When turned on, i

    makes the inversion somewhat model-based. Sounds reasonable

    - the inversion impedances should agree or a least be consistent

    with an impedance model constructed from the well logs. And i

    is reasonable, as long as it is not overdone. The primary use o

    the model term should be to help control those frequencies below

    the seismic band. When it is used to add high frequency information above the seismic band, great care should be exercised

    (more on this below). We control the model contribution in a

    variety of ways both inside and outside of the objective function

    In the initial iteration of a sparse spike inversion, the model term

    is set firmly off. We need to very clearly isolate the unique and

    separate contributions of the seismic and the model. In the fina

    inversion, we may turn it on to achieve the best power matching

    between the model and seismic at the transition frequency.

    By now you might be thinking that no simple label can be

  • 8/6/2019 Jan01 Seismic Inversion

    2/7

    FEATURE ARTICLE Contd

    applied to any modern algorithm. You would be right, and as we

    will see below, the distinctions can become even more blurred.

    The Details - Low Frequencies, Constraints, QC, Annealing,

    Global Inversion

    It is important how the low frequency part of the inversion

    spectrum is computed. By low, we mean the frequencies belowthe seismic band. They are important since they are present in

    the impedance logs which we seek to emulate. Commonly, the

    low frequencies are obtained from an impedance model derived

    from the geologic interpolation of well control. They can be

    added to the inversion afterwards as an end step or within the

    objective function itself. In the latter case only very low frequen-

    cies need to come from the model. At Jason, we can do both and

    each has its advantages. Whichever method is used, you might

    now be saying that all inversions must, to some degree, be

    model-based. The writer would not dispute this assertion. The

    important point though, is that the model contribution is essen-

    tially in a different band than that of the seismic.

    There are other strategies in seismic inversion which control

    the way in which the output impedances are obtained. It is com-

    mon to define high and low limits on the output impedances.

    These are supposed to keep the inversions physical and consis-

    tent with known analogues and theories. Our implementation is

    unique in that these constraints are not a fixed percentage of the

    log impedances but can be freely defined at each horizon, vari-

    able with time. The constraints are interpolated along the hori-

    zons throughout the project, riding them like a roller coaster.

    Could the inversions be then critically dependent upon inaccu-

    rate horizons interpreted from seismic data? We address this

    potential problem in two ways. In the first pass of inversion, theconstraints are relaxed to allow for inaccuracies. The horizons

    are then re-evaluated against the initial inversion before a final

    pass with tighter constraints. Second, the re-evaluation can be

    done on an inversion without the model-based low frequencies

    added in at all. We call this the relative inversion and it is by def-

    inition, free from any inaccuracies in the input model.

    Quality control of the relative inversion comes from perhaps

    an improbable source. In our sparse spike approach, when we

    invert at the well locations, the impedance log is not made avail-

    able to the algorithm. Only the user constraints and the objective

    function settings control the output, leaving it free to disagree

    with the logs. It then follows that comparing the relative inver-sion and the band-limited impedance logs is a very powerful

    quality control tool. The corollary is that the addition of more

    logs to the inversion project increases confidence in the result

    rather than copying the answer into it.

    Some algorithms incorporate a process called simulated

    annealing. This method deliberately takes a round-about path to

    the solution of the objective function in recognition of the fact the

    solution space may not contain a single simple minimum - the

    inversion solution. If the solution space is a bumpy place, then

    simulated annealing is indeed needed to avoid being hung up on

    local minima. So, is solution space bumpy or not? It depend

    upon how the inversion problem is parameterized. In the Jason

    sparse spike algorithm family, the parameterization is by

    impedance and grid point, resulting in a solution space withou

    local minima. Some algorithms parameterize by impedance andtime. That is a more problematic parameterization due to the fac

    that the time parameter can be strongly non-linear,. The result i

    a much more complicated solution space which may necessitate

    the use of a simulated annealing algorithm to come to a mean

    ingful solution.

    Global is another term which has recently been used in con-

    junction with inversion. In Global mode, more than one trace i

    inverted at the same time within a common objective function

    The idea is that seismic noise induced variations that are not con

    sistent over a user-specified number of traces will tend to be sup

    pressed in the output impedances. The result is a smoothe

    looking inversion, which, if one has been careful, does not compromise resolution. The number of traces inverted at once varie

    with algorithms. We typically simultaneously invert for large

    slightly overlapping blocks of traces, the only limit being

    imposed by CPU memory.

    In the preceding, our discussion has been centred on inversion

    methods which emphasize sparsity as a way to invoke the keep

    it simple term. In the following, we present some field example

    of this and other inversion methods, some with potentially even

    more powerful strategies.

    EXAMPLES

    Constrained Sparse Spike Inversion

    This is the most common inversion algorithm in use by our

    clients. Most importantly, they need to understand precisely

    what statement the seismic data, itself is making abou

    impedance. Sparse Spike is ideally suited for this since, among

    its other virtues (noted above), it produces both a relative (no low

    frequencies information from the logs added) and an absolut

    (low frequencies added) version. The separate contributions o

    the logs and the seismic are then clear.

    Figure 1 shows such an inversion over a Nisku reef. On struc

    ture, the blue colours represent the low impedances of the porou

    reef which has both a west and an east build-up. Farther to theeast are low impedance shales. The plot is in an un-smoothed

    format so that the contribution of each voxel (small rectangle) can

    be more easily seen. Meaningful lateral variations in impedanc

    of 1 or 2 samples have been achieved and variations in

    impedance within the reef can be identified. The inversion vol

    ume was converted to depth using a client-provided depth

    datum, its associated time datum from the inversion and veloci

    ties from the geologic model. The depth inversion was then con

    verted to porosity using the observed relation in the logs. Map

  • 8/6/2019 Jan01 Seismic Inversion

    3/7

    FEATURE ARTICLE Contd

    of average porosity were made for several depth intervals and

    one of these is shown in Figure 2. The best porosity is concen-

    trated in the south-east reef but in a non-homogeneous way.

    Wells were drilled from these maps and confirmed the inversion

    predictions.

    In Figure 3 we show a southeast Asia clastic example from

    Latimer et al., 1999. The facies are an alternating sequence of

    sands and shales. Interpretation is problematic due to the closevertical positioning of contrasting layers within half of a wavelet

    length. The result is severe interference (tuning) and a general

    complication of the seismic section. The interpretation of the yel-

    low reservoir event is particularly difficult. Figure 4 shows the

    inversion result. It is generally more simple and the interpreta-

    tion of the yellow event is obvious. It is now interpreted as a

    sequence boundary which is overlain by an incised valley sand.

    Model-Based Inversion

    We reserve the term model-based inversion for methods that

    attempt to achieve resolution within and beyond the seismic

    band by a stronger use of the a priori impedance logs in the model

    term of equation 1. In one of the model-driven methods we sup-port, the simplicity term is implemented by limiting the solution

    space to that spanned by (perturbed) well logs. The model term

    is free to act both within the seismic band and outside of it.

    Obviously we need to have confidence in our logs and in the rela-

    tion between them and the seismic data if we are going to use

    them in such a way. This confidence is generally provided byrunning a sparse spike inversion first. As in any model-based

    approach we need to ensure that all possible facies are represent-

    ed in the logs. The Jason inversion works by adapting the initial

    model made from a simple interpolation of the input logs along

    structure to a more detailed one. At any given CMP to be invert-

    ed, an optimal weighting of the input wells is determined to min-

    imize the residual between the input seismic and the inversion

    synthetic. The whole operation is performed in a transformed

    principle components space to remove non-uniqueness resulting

    from the similarity of input wells. The final detailed model i

    also the output inversion. Since high resolution logs are input

    high resolution in the inversion is possible.

    This approach has been used successfully by Helland-Hansen

    et al., 1997 in the North Sea to highlight regions where sands

    might occur with increased probability. Figure 5 shows the inpu

    density model constructed from selected wells in principle com

    ponents space. Figure 6 shows an enlarged section of the fina

    model where the green colours represent densities indicative o

    increased sand probability. Prior to doing the inversion, this are

    had not been considered to be sand prone. A well was drilled on

    this anomaly and the inferences from the inversion were confirmed. This resulted in the development of a new productive

    Figure 1: A section through a 3D constrained sparse spike inversion of a Nisku reef. There aretwo main build-ups. Porosity is indicated by the lowest impedances and can be seen to be non-homogeneously distributed. In this un-smoothed display, each small colour square representsone voxel. There are apparently meaningful, consistent variations of 1 or 2 voxels.

    Figure 2: The impedance volume in Figure 1 has been converted to depth using a client-supplied datum and velocities from the logs. It was then transformed to porosity using the observerelation between impedance and porosity in the logs. The figure shows the average porosity inone of a series of depth intervals. Note that the porosity is not distributed uniformly withineither of the build-ups.

    Figure 3: This seismic section is from southeast Asia and represents a sequence of sands anshales. The Yellow horizon interpretation has not been completed. It is made difficult by thclose proximity of events and the structure.

  • 8/6/2019 Jan01 Seismic Inversion

    4/7

    FEATURE ARTICLE Contd

    horizon. The model-based algorithm has two flavours. The sec-ond, not illustrated here, associates character changes in (multi-

    ple) seismic data set(s) with well log attributes and obviates the

    need for a wavelet.

    Geostatistical (Stochastic) Inversion

    Geostatistical simulation differs from all of the other methods

    in one respect. There is no objective function and hence no need

    for a simplicity term to stabilize it. Rather, property solutions

    (impedance, porosity, etc.) are drawn from a probability density

    function (pdf) of possible outcomes. The pdf is defined at each

    grid point in space and time. A priori information comes from

    well logs and spatial statistical property and lithology distribu-

    tions. As in the other model-based methods, the logs are assumed

    to represent the correct solution at the well locations. It is again

    useful to run a sparse spike inversion first, to establish this.

    Historically, away from wells, geostatistics has had problems. It

    is the inversion aspect of geostatistics which has finally guaran-

    teed its use as a modern inversion tool. The geostatistical (or

    stochastic) inversion algorithm simply accepts or discards simu-

    lations at individual grid points depending upon whether they

    imply synthetics which agree with the input seismic. The deci-

    sion to accept or reject simulations can optionally be controlled

    by a simulated annealing strategy. The inversion option results

    in a tighter set of simulations, the variation of which can be used

    to estimate risk or make probability maps. The simulations canbe done at arbitrary sample intervals. Close to wells, resolution

    beyond the seismic band can reasonably be inferred. Away from

    wells, the absence of a simplicity term in the simulation and the

    statistical conditioning still hold the possibility of resolution

    beyond that of traditional inversion methods.

    In geostatistical modelling, property and indicator (facies)

    simulations can be combined to produce both property (eg.

    impedance) and facies volumes. This is illustrated in Figure 7

    from Torres-Verdin et al., 1999, which shows such an estimate

    from Argentinean data. The green patches are sand bodies from

    a single simulation. Favourable locations for new wells were

    determined by integrating the sand volume at each CMP for a se

    of simulations. The results of this development programme

    showed a definite improvement in sand detection. Accumulated

    production has been up to three times the field average in some

    instances, more than justifying the effort and expense of the

    inversion.

    Important end results of 3D Geostatistical modelling areproperty probability volumes. A set of volume simulations o

    porosity, for example, can be modelled as a Normal probability

    density function at each grid point in time and space. From

    these, volumes can be constructed giving the probability that the

    porosity lies within a specified range. Figure 8 shows an exampl

    of this for simulations of porosity over a Western Canadian

    Devonian reef. Twenty simulations were used to generate a

    probability volume for the occurrence of porosity above 10%

    Figure 4: This shows the constrained sparse spike inversion of the data in Figure 3. It is a muchsimpler section due to the attenuation of wavelet sidelobes. The completion of the Yellow hori-zon is now easy. The low impedance region above the Yellow, in the middle of the figure hasbeen interpreted to be a valley fill.

    Figure 5: This section is from a volume representing the initial density model of a model-basedinversion. It was constructed from selected well logs which had previously been transformed tprinciple components space.

    Figure 6: The density model in Figure 5 (enlarged section shown) has been modified by modelbased inversion. This final model is in fact the inversion output. It is a weighted average of thwell logs, optimized to produce a synthetic which agrees with the seismic. The low density(green) area indicates increased probability of sand. A well was subsequently drilled on thianomaly and confirmed the predictions of the inversion.

  • 8/6/2019 Jan01 Seismic Inversion

    5/7

    FEATURE ARTICLE Contd

    This volume was then viewed in 3D perspective and probabilities

    less than 80% were set to be transparent. The tops and bottoms

    of the viewable remainders were picked automatically. It is the

    thickness of one of these high-probability bodies which is

    mapped in Figure 8. The colours represent the thickness, within

    which, the probability of 10% or greater porosity exceeds 80%. In

    this way, uncertainty can be formally measured and input direct-ly into risk management analyses.

    Elastic Impedance Inversion

    AVO analysis, which has been with us for many years, has

    taken a new and dramatic turn recently with the introduction of

    the concept of Elastic Inversion. Briefly, the thought process is as

    follows. We are familiar with the transformation of reflection

    coefficients to acoustic (zero incidence) impedance. Zoeppritz

    tells us that reflection coefficients are angle-dependent. What

    then would a transformation of angle-dependent reflection coef-

    ficients to impedance represent? We call this angle-dependent

    quantity Elastic Impedance and claim that an inversion of an

    angle or offset-limited stack would measure it. Such an approach

    addresses a whole host of problems including offset-dependent

    scaling, phase and tuning (Connolly, 1999). Both traditional AVO

    methods and Elastic methods which determine P and S reflectiv-

    ities from seismic directly, do not deal with these issues correctly.

    Simultaneous Zp, Zs Inversion

    The elastic impedance inversion process can in fact be carried

    one step further with the simultaneous inversion of angle or off-

    set-limited sub-stacks. This procedure brings all the benefits of

    inversion to bear upon problems encountered in traditional seis-

    mic weighted stacking approaches to AVO analysis. The out-

    comes are inversion volumes of P-impedance, S-impedance and

    density. Density is usually ill-defined and set to be constrained

    to the relationships observed in the logs.

    In Figure 9, we present an example from Pendrel et al., 2000

    Detection of Cretaceous valley sands in the Blackfoot, Alberta

    area is problematic due to the similarity of their P impedanc

    with that of shales. In Figure 9, the near and far offset stacks have

    been inverted separately. Note that the valley sands indicated by

    the arrows are softer (lower impedance) at the far offset. Figur

    10 shows the results of the simultaneous inversion for the Upper

    Valley. The P-impedance is as much responsive to the off-valley

    shales as it is to the valley sands. The S-impedance section, on

    the other hand, highlights the highly-rigid valley sandsTransforming to the familiar Lame parameters LambdaRho and

    MuRho (Goodway et al., 1997), and then forming their ratio

    results in a good sand discriminator (Figure 11).

    Conclusions / Discussion

    I hope in the above that we have conveyed the wide range o

    possibilities in modern seismic inversion and in particular those

    available through Jason Geosystems. Interpreters need to con

    sider seismic inversion whenever interpretation is complicated

    by interference from nearby reflectors or the end result is to be a

    quantitative reservoir property such as porosity. Output in the

    format of geologic cross-sections of rock properties (as opposedto seismic reflection amplitudes) is endlessly useful as a means to

    put geologists, geophysicists, petrophysicists and engineers on

    the same page.

    It is almost always useful to do sparse spike inversion first so

    that we may be crystal clear about what the seismic is telling us

    free from bias from logs. These ideas are best expressed in th

    relative inversion. The absolute inversion brings in the logs as a

    geologic context for the relative inversion and at the same time

    broadens the inversion band. Only after this, when the relation

    Figure 7: This is a perspective view showing the results of the simulaneous geostatisticalsimulation of density and lithotype (shale, tight sand, porous sand). The green bodies arelow density sands. A set of simulations were computed and integrated at each CMP toestablish the probability of occurrence of sand. Wells drilled on these analyses have shownsignificantly higher production rates.

    Figure 8: Geostatistical simulation of porosity was done over a Western CanadianDevonian reef. At each point in time and space, the set of simulations was modelled by probability density function (pdf). Then a volume was created, the elements of which were probabilities that the porosity was in excess of 10%. Next, the probabilities themselves werviewed in 3D perspective, and values less than 80% made transparent. Finally, the remaining visible probability bodies were automatically interpreted. A small area of the final mapof probability body thickness is shown here. The colours are the thickness (ms) of bodiewhich should have greater than 10% porosity, 80% of the time.

  • 8/6/2019 Jan01 Seismic Inversion

    6/7

    FEATURE ARTICLE Contd

    between the logs and the seismic has been ascertained should we

    consider methods, which make a stronger use of the logs. Model

    based Inversion, Elastic Inversion, Simultaneous Zp, Z

    Inversion and Geostatistical Inversion are all possibilities

    depending upon the data and the goals of the project.

    I do not expect that the confusion over inversion nomencla-

    ture will disappear anytime soon. Perhaps, though, we hav

    managed to shed some light on this topic. May I suggest that we

    reserve the term, model-based for those algorithms which, in a

    significant way, use logs to modify the inversion band within or

    above seismic frequencies?

    The days of viewing seismic inversion as an extra processing

    step or subject of an isolated special study are long gone. Modern

    inversions are intimately connected to detailed and quantitative

    reservoir characterization and enhanced interpretation produc

    tivity. The process requires and integrates input from all mem

    bers of the asset team. Horizons should be re-assessed, model

    re-built, log processing reviewed and inversion steps iterated

    toward the best result. After drilling, new information should bused to create a living cube, always up-to-date with all avail-

    able information. It is this partnership directed to the solution o

    real reservoir characterization problems which leads to success.

    References

    Connolly, P., 1999, Elastic Inversion, The Leading Edge, 18 #4

    p.440

    Goodway, B., Chen, J.,Downton, J., 1997, AVO and Prestac

    Inversion, CSEG Ann. Mtg. Abs. p.148

    Helland-Hansen, D., Magnus, I., Edvardsen, A., Hansen, E., 1997Seismic Inversion for Reservoir Characterization and Well Planning in

    the Snorre Field, The Leading Edge, 16 #3, p.269

    Latimer, R.B., Davison, R., Van Riel, P., 2000,An Interpreters Guide

    to Understanding and Working with Seismic-Derived Acoustic

    Impedance Data, The Leading Edge, 19 #3, p.242

    Pendrel, J., Debye, H. Pedersen-Tatalovic, R., Goodway, B.

    Dufour, J., Bogaards, M., Stewart, R., 2000, Estimation and

    Interpretation of P and S Impedance Volumes from the Simultaneous

    Inversion of P-Wave Offset Data, CSEG Ann. Mtg. Abs. paper

    AVO 2.5

    Torres-Verdin, C., Victoria, M., Merletti, G., Pendrel, J., 1999

    Trace-Based and Geostatistical Inversion of 3-D Seismic Data for Thin

    Sand Delineation: An Application to San Jorge Basin, Argentina, The

    Leading Edge, 18, #9, p.107

    Figure 10: The near and far offset stacks from the Blackfoot seismic data were simultane-ously inverted for P-impedance and S-impedance. These figures are maps of the averages ofthese impedances in the Upper valley. Note that the S-impedance map shows the rigid sandsas relatively high impedances. The P-impedance map is sensitive to both sands and shales.

    Figure 11: The P- and S- impedance volumes from simultaneous inversion were trans- formed to the Lame constants, ??and ??(multiplied by density). Then a ratio volume wasformed by dividing them. The result is a very sensitive valley sand detector.

    Figure 9: The figures show the independent inversions of near and far offset stacks from theBlackfoot 3D survey. At Blackfoot, Glauconitic sands and shales have similar P-impedances. Note that the valley sands (arrows) appear softer at the far offsets. This is theAVO effect. The estimation of separate wavelets for each inversion means that offset-depen-

    dent scaling, phase and tuning have been optimally addressed.

  • 8/6/2019 Jan01 Seismic Inversion

    7/7

    FEATURE ARTICLE Contd

    JOHN PENDREL

    John Pendrel is Chief Geophysicist with Jason Geosystems Canada. In this positionhe is responsible for applying proprietary, leading edge technology to problems in reser-voir analysis and characterization. From 1981 to 1995, he was Sr. Geophysicist and thenManager, Geophysical Technology with Gulf Canada Resources in Calgary. He beganhis career in the oil industry in 1978 with Gulf Science and Technology Company inPittsburgh, PA, the research arm of the former Gulf Oil.

    Johns academic career included a B.Sc. at The University of Saskatchewan (1968), and

    an M.Sc. from The University of Saskatchewan, Saskatoon, (1972) where he did research

    in auroral magnetic fields. John also holds a Saskatchewan Class A teachers certificate. He obtained a Ph.D.

    (geophysics) from York University, Toronto in 1978 where his interests were in two-dimensional time series and

    spectral analysis.

    Johns early research was in the areas of pattern recognition and principal components analysis. More

    recently he has done applied research and published papers in seismic inversion, geostatistical analysis and

    AVO. Away from work he plays on an ice hockey team and volunteers with a local youth group, The Calgary

    Stampede Showband.