Beruflich Dokumente
Kultur Dokumente
INTERPRETER’S
REBECCA BUXTON LATIMER and RICK DAVISON, Jason Geosystems, Houston, Texas, U.S.
PAUL VAN RIEL, Jason Geosystems, Rotterdam, The Netherlands
Y ou have just joined a new asset tation with two imaging techniques. without the complicating factors
team or new company. You’re handed The model is simply a low acoustic caused by wavelets and adds
an area to evaluate. The data include impedance wedge embedded in a essential information from the log
Coordinated by Linda R. Sternbach
seismic, logs, and a cube of acoustic high acoustic impedance background data. The AI volume is a result of
impedance. What do you do with the (Figure 1a). Figures 1b and 1c show the integration of data from sev-
impedance data? How was it created? zero-phase seismic representations of eral different sources, typically seis-
What unique information does it pro- the model in standard wiggle trace mic, well log, and/or velocity.
vide? What pitfalls and artifacts may and color density with wiggles over- Indeed, building an impedance
be present? How do you interpret lain. Notice the tuning effects as the model is the most natural way to
this data set on a workstation that is wedge thins and the side lobe inter- integrate data and provides a
designed for seismic data? How do ference within the wedge itself. medium understood by geologists,
you report your results to manage- Figure 1d shows the results from geophysicists, petrophysicists, and
ment? Valid questions? Read on. inverting the seismic data to AI. engineers.
Inversion of seismic data into Tuning is diminished, and the false • Acoustic impedance is a rock prop-
acoustic impedance (AI) is a rapidly internal geometry is eliminated. The erty. It is the product of density and
growing field, due primarily to the resulting inverted wedge is a more velocity, both of which can be
ease and accuracy of interpretation of accurate spatial representation of the directly measured by well logging.
the impedance data. The term “inver- original model and provides absolute Seismic data is an interface prop-
sion” has the potential for a great AI values (shown in color) that match erty, a close approximation to the
amount of confusion, as it is used to the original model. convolution of a wavelet with a
mean many different things within Another compelling reason for reflection coefficient series, which
various branches of geoscience. The inverting seismic data is illustrated in reflects relative changes in acoustic
discussion in this paper will concen- Figure 2. A synthetic seismic data set impedance. AI is therefore the nat-
trate on the inversion of poststack (colored seismic with wiggles over- ural link between seismic data and
seismic traces into acoustic imped- lain) is shown in panel 2b. The syn- well data.
ance data. Even with this narrower thetic seismic is created from the • AI is closely related to lithology,
scope, the plethora of programs on acoustic impedance model in panel porosity, pore fill, and other fac-
the market today makes the com- 2d and the wavelet in panel 2a. The tors. It is common to find strong
parison of various inversion method- model contains three interfaces: 50 empirical relationships between
ologies and the determination of the ms, 135 ms, and 230 ms. Note that acoustic impedance and one or
quality of your AI cube difficult at each interface represents the same more of these rock properties. AI
best. change in absolute AI units but in models can provide the basis for
This paper will provide a descrip- varying gradational degrees. The the generation of 3-D facies mod-
tion of terminology and a basis for seismic data identify the sharp inter- els and 3-D petrophysical property
comparison of poststack acoustic face at 50 ms. They identify the top models. These volume results can
impedance inversion products, as of the second interface at 135 ms, but be ported directly into reservoir
well as give the interpreter a method- it is not apparent that the interface is simulators for flow analysis.
ology for quality control and inter- a gradational coarsening upward • AI is a layer property. Seismic
pretation of inverted data. sequence because the seismic do not amplitudes are attributes of layer
recognize the base of the event. The boundaries. As a layer property,
Benefits of impedance data. seismic fail to identify the most grad- acoustic impedance can make
Acoustic impedance (AI) is the prod- ual interface at 230 ms. Compare the sequence stratigraphic analysis
uct of rock density and P-wave veloc- seismic response with that of the more straightforward. Wavelet side
ity. This means that AI is a rock inverted traces in panel 2c. The lobes are attenuated, eliminating
property and not an interface prop- inverted trace data can effectively some false stratigraphic-like effects
erty (e.g., seismic reflection data). As model all these variations in rock as seen in Figures 1b and c.
we will illustrate, this distinction is properties because the inverted data • AI data support fast and accurate
the power of AI. Acoustic impedance utilize a complete frequency range of volume-based interpretation tech-
inversion is simply the transforma- 0-80 Hz. To summarize some advan- niques, allowing for rapid delin-
tion of seismic data into pseudoa- tages of impedance data: eation of target bodies.
coustic impedance logs at every trace. • The AI concept is readily general-
All information in the seismic data is • A good quality impedance model ized to handle the inversion of
retained. contains more information than angle or offset stack data to elastic
Figure 1 shows an acoustic seismic data. It contains all the impedance or elastic parameters.
impedance model and its represen- information in the seismic data Elastic impedance captures AVO
242 THE LEADING EDGE MARCH 2000 MARCH 2000 THE LEADING EDGE 00
information and, in conjunction
with AI, improves interpretation
power and the ability to discrimi- a)
nate lithology and fluids.
244 THE LEADING EDGE MARCH 2000 MARCH 2000 THE LEADING EDGE 00
a) b) c) d)
Downloaded 06/11/13 to 128.59.62.83. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/
Figure 2. Impedance inversion models contain more information than seismic data because they have a broader
frequency content. (b) Synthetic seismic data set based on the acoustic model in (d) and the wavelet in (a). There
are three transitions—50 ms, 135 ms, and 230 ms. Each interface represents the same change in absolute AI units
but in varying gradational degrees, representing varying dominant frequencies. The synthetic seismic identifies
the sharp interface at the top 50 ms (80 Hz), sees the top of the event at 135 ms (15 Hz), but it is not clear that it is a
gradational coarsening upward sequence and fails to recognize the most gradual interface at 230 ms (6 Hz).
Compare the seismic responses to that of the inverted traces in (c). The inverted traces are shown in color with a
black overlay. There is a significant difference in the properties of the rocks at 150 ms and 230 ms. These differ-
ences are not clear from the synthetic seismic data because the low-frequency information is missing. On the other
hand the impedance inversion model contains this necessary information.
mic data to broadband impedance data alone is clearly problematic. The data, and a set of preliminary time or
data. In this example, AI inversion change in layer “hardness” allowed depth horizons.
was applied to assist in the inter- the inverted data to image the uncon- Prior to inversion, examine the
pretation of a “hidden” channel. In formity clearly. well logs for suitable relationships
Figure 4, a seismic section is shown between measured impedance logs
in wiggle trace format. The yellow Quality control of input data. The (calculated by dividing the density by
event on the left and right of the sec- quality of the final inversion is a the sonic log) and other desirable
tion is interpreted based on well con- direct result of the quality of the input properties, such as porosity and fluid
trol and is a known unconformity. data. To objectively estimate the accu- fill. Well logs should be converted to
There are a number of ways that the racy of an AI inversion cube, the time and filtered to the approximate
yellow marker could be interpreted interpreter must be familiar with the bandwidth of the seismic to deter-
to tie between segments. The problem input data and what processes were mine if zones of interest are recog-
is resolved by looking at the acoustic applied to invert the data. A com- nizable at the frequencies expected
impedance inversion result (Figure prehensive inversion report is a pow- after inversion. All well logs should
5). Following the top of the high- erful source of information but, if not be edited for borehole effects, bal-
impedance layer (red/yellow) leaves available, some key items should be anced and classified based on qual-
little ambiguity in the answer. A examined: seismic processing infor- ity. Logs that do not tie the seismic
channel has been incised into the pre- mation, inversion algorithm, date should be investigated for problems
viously deposited high acoustic and workflow, well spud details, and in log, wavelet, or seismic data.
impedance layer. The completed log processing. Depending on the When inverting, it is generally
interpretation has been transferred inversion method, the data types may preferable to run a loosely con-
back to the seismic data in Figure 6. include poststack seismic data (full strained, trace-based inversion first.
The interpretation from the seismic fold as well as angle stacks) well-log The inversion can then be used for a
246 THE LEADING EDGE MARCH 2000 MARCH 2000 THE LEADING EDGE 00
more thorough interpretation as
shown in Figures 4-6. This initial
inversion can be followed by a more
tightly constrained or model-based
inversion as the need arises to meet a)
your particular project’s interpreta-
tion objectives. With trace-based
inversion, the process begins with the
Downloaded 06/11/13 to 128.59.62.83. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/
Quality control of the acoustic im- Figure 3. The band-limited nature of seismic data is often considered in
pedance results. Numerous AI in- terms of the high frequencies and consequent lack of resolution. However,
version algorithms are available the low frequencies missing from the seismic data are extremely important
throughout the industry. Regardless if quantitative interpretation is required. This is illustrated by filtering a
of the method used, certain quality simple impedance layer model to three different frequency ranges (a) 10-80
controls should always be carried Hz, (b) 10-500 Hz, and (c) 0-80 Hz. The inclusion of the high frequencies
out. The main tests of inversion accu- (b) allows us to interpret the location of the layer boundaries more accu-
racy are the ties between the input rately, but it is the inclusion of the low frequencies (c) that allows us to
well logs and the inversion result, obtain absolute values for use in the quantitative interpretation of the rock
and between the input seismic and properties.
the synthetic derived after the inver-
sion. Which one of these two tests is
most important as a quality check
depends on whether the inversion
is predominantly model-based or
trace-based.
Volumes created with methods
that are heavily driven by log-derived
models should match at the well loca-
tions. If the logs do not tie, then they
have been perturbed in the inversion
process, indicating a problem in the
log or seismic data or, more likely,
the initial tie of the seismic to the well
data. For these model-based inver-
sions, the match between the seismic
and the synthetic created from the
inverted results serves as an impor-
tant QC.
Trace-based inversion methods
and those model-based methods that
make limited use of well log or other
nonseismic data rely heavily on the
seismic data and should tie the seis- Figure 4. Interpretation is made easier by reference to the impedance inver-
mic. Because the well data have lim- sion model. This is clearly demonstrated with this example of a “hidden”
ited use, the logs can be used as an channel. A seismic section is shown in wiggle-trace format. The yellow
independent QC. To make a valid event on the left and right is interpreted based on well control and is a
comparison between log data and the known unconformity. A number of places exist where the yellow horizon
inverted impedance, the log data could drop down onto a lower event in the attempt to tie the horizons. The
should first be filtered to the range of correct answer appears to be a matter of interpretative judgment and
the seismic frequencies. knowledge of the regional play concepts.
248 THE LEADING EDGE MARCH 2000 MARCH 2000 THE LEADING EDGE 00
In all cases, a universal quality
check of an acoustic impedance vol-
ume is to compare the inverted AI
estimate to log data not used in the
inversion. These wells may have been
drilled later or intentionally not input
to provide “blind” accuracy tests of
the process. For such blind tests to be
Downloaded 06/11/13 to 128.59.62.83. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/
250 THE LEADING EDGE MARCH 2000 MARCH 2000 THE LEADING EDGE 00
are being treated as though they were
seismic. Once it is understood that
inverted AI data represent a true rock
property, it becomes much easier
to extend our methods of interpreta-
tion beyond traditional 2-D or 2.5-D
interpretation methods. In fact,
impedance data make true 3-D inter-
Downloaded 06/11/13 to 128.59.62.83. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/
252 THE LEADING EDGE MARCH 2000 MARCH 2000 THE LEADING EDGE 00
converted to depth and ported
directly into a reservoir flow sim-
ulator, along with AI-derived reser-
voir properties such as porosity.
attempt to broaden the spectrum for meters between wells. Geostatistical ucts, a prospect derived from your
the higher frequencies. The focus in inversion provides a powerful way to new exploration area or a develop-
these methods is on including an ini- bring in information from outside the ment or production well proposal can
tial model and stabilizing to this ini- seismic bandwidth, utilizing both well be more accurately and efficiently
tial model to cover the low-frequency control and geologic control on the evaluated and risked by the geolo-
component of the spectrum. spatial distribution of acoustic imped- gist, geophysicist, and engineer. LE
Layer-based or blocky, sparse ance.
spike, and least squares inversion all Acknowledgments: The authors thank BP
produce broadband results by virtue The future of inversion. AI inversion Amoco for permission to use and publish the
of the nature of the method itself, or provides the most straightforward seismic data in this paper.
by providing control relative to an ini- conversion from seismic reflection data
tial geologic model. They are all lim- to layer rock-property data, providing Corresponding author:
ited in their ability to reconstruct the a wide range of interpretive benefits. rebecca@houston.jasongeo.com
high frequencies. Except for RTI, all However, AI inversion is only a step-
these methods succeed, to some level, ping stone into the realm of seismic- Rebecca Buxton Latimer is chief geoscientist for
in backing out the wavelet and reduc- derived rock-property data. There are Jason Geosystems in Houston, Texas. She has
ing tuning effects. Also, the more mod- many new exciting areas of develop- been an oil-industry geophysicist for 19 years,
ern of these methods allow varying ment to investigate. The first is the including five with Jason. She was formerly
degrees of control on the use of exter- inversion of angle or offset partial stack with Amoco Production in Houston and New
nal information. In this way some of data to leverage AVO information. Orleans. Latimer received a master’s degree in
these methods can smoothly cover the Several approaches are feasible, geology/geophysics from Boston College in
spectrum from trace-based to model- including methods that invert angle- 1980.
based, making them applicable to a or offset-stacked data to elastic imped-
range of interpretation projects. ance. Interpretation of elastic imped- Rick Davison is senior project geoscientist for
New methods are also becoming ance has added benefits to acoustic Jason Geosystems in Houston, Texas. He has
available. These methods take inver- impedance interpretation. The possi- 11 years in the oil industry as a prospect devel-
sion to the next level by greatly extend- bility to combine AI and EI data, oper primarily in the Gulf of Mexico. He was
ing the use of nonseismic information enabled by a new generation of mul- formerly with ARCO and Vastar Resources in
to get broadband results. We discuss ticube volume interpretation methods, Lafayette and Houston. Davison received a doc-
two examples of these advanced meth- creates powerful possibilities for torate in earthquake seismology at Virginia
ods. enhanced interpretation and discrim- Tech.
Methods based on 3-D geologic log ination of lithology and fluids. As a
models: Modern computers allow for next step, simultaneous inversion of Paul van Riel, a founder of Jason Geosystems,
the construction of complex 3-D geo- multiple angle or offset stacks is now has been involved with inversion research, devel-
logic models using a parametric feasible. This leads to estimates of var- opment, and application for 16 years. He
approach. One example utilizes a ious combinations of elastic parame- received a master’s in physics from Delft
model based on input logs, lateral dis- ters such as P-wave and S-wave sonic, University in the Netherlands.
tribution of log weights, time struc- VP /VS, density, and Lame parameters.
ture maps, and velocity corrections to Joint interpretation of these parameters
control geologic layer thickness. This maximizes use of the available seis-
model utilizes seismic, where log infor- mic data for interpretation and for
mation is sparse, to update the lateral enhanced lithology and fluid discrim-
distribution of log weights. Such a ination.
method produces a high-resolution,
broadband output. Because the initial Conclusions. AI, being a lithologic
geologic model is heavily utilized property rather than an interface prop-
(strongly model-based), successful erty, can be used for direct geologic
application requires multiple wells interpretation. Impedance data sets
with excellent fit to the seismic and have many advantages over
good control on the geologic model. seismic. Tuning is diminished and res-
Geostatistical inversion: This inver- olution is increased; sequence strati-
sion algorithm combines geostatisti- graphic analysis is simpler because the
cal data analysis and modeling with data are now in layers, rather than
seismic inversion. In geostatistical interfaces; and wavelet side lobes are
analysis, the spatial statistics of the removed, eliminating the risk of false
data are generated. Geostatistical mod- geologic structures. Direct hydrocar-
eling simulates data at grid points bon indicators are commonly more
starting from known control points, apparent in impedance than in seismic,
typically well logs. Geostatistical mod- and rock types are easily discernible.
256 THE LEADING EDGE MARCH 2000 MARCH 2000 THE LEADING EDGE 00