Download as pdf or txt
Download as pdf or txt
You are on page 1of 5

Imaging faults in 3D seismic volumes

Geoffrey A. Dorn, Benjamin Kadlec and Patty Murtha, TerraSpark Geosciences LLC
Downloaded 04/03/14 to 190.198.34.207. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/

Summary estimates of the discontinuity or edges present in the


seismic data. Upper case Coherence™ will refer to the
Faults interpretation is required step in most seismic patented Coherence technology from Amoco).
interpretations. Unlike horizons, which are imaged by
reflections in the seismic volume, faults are typically Figure 2 shows a coherence-class attribute (Horizon Edge
imaged as discontinuities in the data. For many years faults Stacking) on a section that corresponds to the seismic data
were interpreted by visual recognition of discontinuities in shown in Figure 1. Figure 2 illustrates the limitations of
the seismic data, which were then manually interpreted as this type of attribute. Typically faults are reasonably well
fault cuts on seismic sections. The purpose of our work was imaged on time or depth slices. However, the imaging on
to develop an attribute algorithm which images faults with vertical sections is discontinuous. Attributes that image
sufficient accuracy, precision, and continuity to enable discontinuities are also extremely sensitive to noise in the
automatic extraction or autotracking of faults in a 3D seismic volume. The quality of fault imaging is sufficient to
attribute volume, much like horizons have been autotracked be an assistive technology in fault interpretation, but is
for years in 3D seismic volumes. insufficient to support autotracking of faults in 3D. As a
result faults still have to be interpreted manually – a very
Introduction time-consuming, tedious and imprecise process.

Figure 1 shows a vertical section from a 3D survey in the


Gulf of Mexico, offshore Louisiana. Most faults, as
illustrated in this figure, are represented as discontinuities
in the seismic volume. In early 3D interpretation systems
faults were identified by the interpreter as discontinuities in
the seismic data, and individual fault cuts were manually
interpreted on a sparse set of inlines and crosslines.

Figure 2: Line 184 from a coherence-class attribute


(Horizon Edge Stacking). Note the discontinuous imaging
of numerous steeply dipping normal faults (arrows).

Several attempts have been made to develop techniques to


take a coherence class attribute as input, and generate an
output volume with fault imaging that supports 3D
Figure 1: Line 184 from a survey in the Grand Isle area, autotracking. Most notable among these are AFE™
offshore Louisiana, Gulf of Mexico. Note the numerous (Automated Fault Extraction, Crawford and Medwedeff,
steeply dipping normal faults (arrows) indicated by 1999; Dorn and James, 2005), Ant Tracking (Randen, et al,
discontinuities in the seismic reflections. 2001), and Hough Transforms (Al bin Hassan and Marfurt,
2003). In ideal situations (high signal, low noise, steeply
dipping faults) all three of these techniques showed some
Volumetric imaging of faults based on attributes that image success in improving the imaging of faults, but all three
discontinuities in the seismic volume began at several approaches have problems in real-world data with noise,
energy companies in the early to mid-1990s. The most shallower fault dips (e.g., dips shallower than 75o), and
well-known of these research efforts was at Amoco complex fault patters.
(Bahorich et al., 1995), and resulted in the development of
Coherence. (Note – following Marfurt and Chopra, 2007, The resolution of these fault-imaging issues has motivated
the lower case “coherence” is uses in this paper to refer to the development of a workflow and an advanced version of
the entire class of attributes that represent volumetric the AFE algorithm.

© 2012 SEG DOI http://dx.doi.org/10.1190/segam2012-1538.1


SEG Las Vegas 2012 Annual Meeting Page 1
Imaging faults in 3D seismic volumes

Advanced AFE Workflow and Process The vertical striping on the section associated with
Downloaded 04/03/14 to 190.198.34.207. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/

acquisition footprint has been removed and random noise


The workflow combines improved imaging of faults in the has been reduced, resulting in improved imaging of faults
input discontinuity volume, with a new Advanced AFE in the seismic amplitude data.
algorithm that is designed to accurately and precisely image
faults regardless of dip, image intersecting faults, and Properly managing coherent and random noise in the input
improve resolution and signal strength in the output fault volume substantially improves the imaging of the faults
probability volume. Together they combine to produce with a discontinuity attribute. Figure 4 shows the result of
results that support 3D fault autotracking. applying Horizon Edge Stacking to the seismic volume
represented by the seismic data in Figure 3.
The workflow consists of the following steps:
• Coherent noise (acquisition footprint) removal Compare Figures 4 and 2. The same vector dip volume was
• Create a vector dip attribute volume used in both cases, and the same input parameters were
• Dip compensated random noise removal used for the Horizon Edge Stack process. The difference in
• Create a dip compensated discontinuity volume fault imaging is entirely associated with the presence of
• Apply Advanced AFE process random and coherent noise in the input seismic volume for
Figure 2, and the relative absence of coherent and random
A process based on a de-striping algorithm from satellite noise in the input seismic volume for Figure 4.
image processing (Crawford and Medwedeff, 1999) was
used to remove coherent noise from the migrated seismic
volume. The input parameters include the direction of
striping, and the wavelength of the striping. If the footprint
is comprised of multiple wavelengths, or has components
in both the inline and crossline directions, the process is
applied several times in sequence to remove the entire
footprint pattern.

A volumetric estimate of strike and dip was then created


using a structure tensor analysis based algorithm (e.g.,
Chopra and Marfurt, 2007). A dip compensated edge
preserving random noise filter was then applied to the Figure 4: Line 184 from a coherence-class attribute
volume produced by the coherent noise removal process. In (Horizon Edge Stacking) applied after coherent and random
this instance, a simple small 5x5x1 horizon dip oriented noise have been removed from the seismic volume. Note
median filter was used. Figure 3 shows the result of this the imaging of the faults is improved, but still is too
processing applied to the data in Figure 1. discontinuous for autotracking.

The discontinuity volume represented by the section in


Figure 4 was then used as input to the Advanced AFE
process. Advanced AFE processes the input discontinuity
volume in three steps:

• Azimuth scan on time or depth slices


• Dip scan on vertical slices
• Planarity scan

Since discontinuity attributes tend to image faults better on


horizontal slices through the volume than on vertical slices,
Figure 3: Line 184 after footprint removal and horizon dip AFE processing begins with an azimuth scan at each
compensated random noise removal. Note the elimination sample on each horizontal slice in the volume. At each
of the vertical striping associated with acquisition footprint voxel, the process searches for the azimuth at which the
and the reduction in random noise level compared with strongest signal is achieved for a lineament passing through
Figure 1. that sample. The signal strength is output to a Line

© 2012 SEG DOI http://dx.doi.org/10.1190/segam2012-1538.1


SEG Las Vegas 2012 Annual Meeting Page 2
Imaging faults in 3D seismic volumes

Enhanced attribute volume, and the Azimuth is output to a


Fault Azimuth volume.
Downloaded 04/03/14 to 190.198.34.207. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/

In the second step, at each voxel in the Line Enhanced


volume, the azimuth is obtained from the Fault Azimuth
volume, and a dip scan is performed in the direction
perpendicular to the local azimuth. Again the process
searches for the dip at which the strongest signal is
achieved for a lineament passing through that sample. The
signal strength is output to a Vertical Enhanced attribute
volume, and the Dip is output to a Fault Dip volume.

In the final step, at each voxel in the Vertical Enhanced


attribute volume, the azimuth and dip calculated in the Figure 6: Line 184 showing corendered seismic data after
previous two steps is used to define the orientation of a noise management and Fault Enhance attribute.
dipping plane through the voxel. A planarity test is
performed where the output Fault Enhanced (fault Figures 7a-c are volume rendered versions of Fault
probability) value is proportional to the signal strength on Enhance, Fault Azimuth and Fault Dip volumes generated
the dipping plane through the voxel. for this 3D survey. Each rendering extends from 1.2s to 3s
vertically.
In all three steps the user specifies an operator length to be
used in the lineament or planarity calculations. Generally
azimuths between 0 and 180o, and dips between 0 and 90o
are scanned. Any discontinuity attribute can be used as
input (e.g., Coherence, Horizon Edge Stack, or Ant Track
volumes). The process allows for azimuth and dip ranges to
be constrained if desired. The output of the process is a
volumetric estimate of the probability of a fault at each
sample (Fault Enhance volume), and volumetric estimates
of Fault Azimuth and Fault Dip. (a)

Figure 5 shows line 184 from the Fault Enhanced attribute


volume. Fault imaging resolution, continuity and signal-to-
noise ratio have been greatly improved over what was
present in the input Horizon Edge Stack volume. Figure 6
shows line 184 from the Fault Enhanced attribute volume
co-rendered with seismic amplitude. This figure
demonstrated that the faults imaged in the Fault Enhance
attribute volume are properly positioned relative to the
input seismic data.

(b) (c)
Figure 7: 3D volume rendering of (a) Fault Enhance
attribute, (b) Fault Azimuth attribute, and (c) Fault Dip
attribute volumes.

Extracting Fault Surfaces

Once faults have been imaged sufficiently well, the faults


surfaces may be extracted from the volume using either ot
two approaches. The interpreter may autotrack individual
faults in the Fault Enhance volume in much the same way
that horizons are autotracked in 3D. Alternatively, fault
Figure 5: Line 184 from the Fault Enhanced volume. cuts may be automatically extracted on all time or depth

© 2012 SEG DOI http://dx.doi.org/10.1190/segam2012-1538.1


SEG Las Vegas 2012 Annual Meeting Page 3
Imaging faults in 3D seismic volumes

slices in the Fault Enhance volume, and then processed to


associate the independent fault cuts into fault surfaces.
Downloaded 04/03/14 to 190.198.34.207. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/

In Figure 8 a seed point has been placed by the interpreter


on the fault of interest in the Fault Enhance volume, and an
autotracking algorithm has picked a portion of one of the
imaged faults. The interpreter selects the faults of interest,
controls the extent of autotracking, and the linkage between
faults through autotracking parameters. This interactive
approach has the advantage of providing the interpreter
with more direct input to the fault extraction process,
however, in a highly faulted volume, it can still be fairly
time consuming. Figure 10: 3D view of the 71 largest fault surfaces in the
seismic volume. These surfaces were created by extracting
fault cuts from the Fault Enhance volume and clustering the
cuts into surfaces.

Summary and Conclusions

Fault imaging based on discontinuities in the seismic


volume has proven to be a valuable assistive technology to
support manual interpretation of faults. We have
demonstrated a workflow to control noise present in the
seismic volume to improve the imaging of faults in a
coherence-class attribute. The Advanced AFE process uses
the resulting discontinuity attibute volume as input and
Figure 8: 3D view of Fault Enhanced volume with an improves the 3D imaging of faults sufficiently to enable 3D
autotracked fault surface. autotracking and automated extraction of 3D fault surfaces.
The resulting surfaces honor the imaging of faults with all
A second approach is illustrated in Figures 9 and 10. Figure of the detail represented in the 3D seismic volume.
9 shows a set of fault cuts extracted from time-slices in the
Fault Enhance volume. For clarity, only extracted cuts for a This approach to fault imaging and interpretation enables
limited range of time slices is displayed. When the fault interpretation of faults that honor all of the detail in the 3D
cuts are extracted, they are independent polylines that are seismic volume. It can substantially reduce the time and
not yet associated into fault surfaces. These fault cuts are human effort involved in interpreting complex fault
then processed through a clustering algorithm. The result of patterns in exploration and development. The fault azimuth
this clustering process on the fault cuts for the entire and dip attribute volumes provide attributes that can be
volume of data is shown in Figure 10 for the 71 largest used both to assist in reliable extraction of fault surfaces
faults in the volume. from the volume, and to enable sorting of resulting surfaces
by azimuth, and dip in addition to size.

Acknowledgements
This work has been conducted as part of the GIVC
Consortium operated by TerraSpark Geosciences LLC. BP,
BHP Billiton, Chevron, ConocoPhillips, Repsol and Stone
Energy support this industry-funded consortium. Their
support, insight and encouragement are greatly appreciated.
The work on Advanced AFE has particularly benefited
from discussions and interaction with Jenny Thompson and
Anastasia Miranova (ConocoPhillips), Barton Payne
(Chevron), William Butler (Shell). AFE™ is a patented
process and is a trademark of TerraSpark Geosciences
Figure 9: 3D view of extracted fault cuts over a portion of LLC.
the Fault Enhanced volume.

© 2012 SEG DOI http://dx.doi.org/10.1190/segam2012-1538.1


SEG Las Vegas 2012 Annual Meeting Page 4
http://dx.doi.org/10.1190/segam2012-1538.1

EDITED REFERENCES
Note: This reference list is a copy-edited version of the reference list submitted by the author. Reference lists for the 2012
SEG Technical Program Expanded Abstracts have been copy edited so that references provided with the online metadata for
Downloaded 04/03/14 to 190.198.34.207. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/

each paper will achieve a high degree of linking to cited sources that appear on the Web.

REFERENCES
AlBinHassan, N. M., and K. J. Marfurt, 2003, Fault detection using Hough transforms: 73rd Annual
International Meeting, SEG, Expanded Abstracts, 1719–1721.
Bahorich, M. S., J. A. Lopez, N. L. Haskell, S. E. Nissen, and A. Poole, 1995, Stratigraphic and structural
interpretation with 3-D coherence: 65th Annual International Meeting, SEG, Expanded Abstracts,
97–100.
Chopra, S., and K. J. Marfurt, 2007, Seismic attributes for prospect identification and reservoir
characterization: SEG.
Crawford, M., and D. A. Medwedeff, 1999, Automated extraction of fault surfaces from 3-D seismic
prospecting data: U. S. Patent 5,987,388.
Dorn, G. A., and H. E. James, 2005, Automatic fault extraction in 3-D seismic interpretation: 67th
Conference and Exhibition, EAGE, Extended Abstracts, F035.
Randen, T., S. I. Pedersen, and L. Sonneland, 2001, Automatic extraction of fault surfaces from three-
dimensional seismic data: 71st International Meeting, SEG, Expanded Abstracts, 551–554.

© 2012 SEG DOI http://dx.doi.org/10.1190/segam2012-1538.1


SEG Las Vegas 2012 Annual Meeting Page 5

You might also like