1. 4. Spectral Decomposition
AASPI
Overview of AASPI accomplishments and software development
Kurt J. Marfurt
Attribute-Assisted Seismic Processing and Interpretation

2. AASPI 2014 Sponsors AASPI Welcome 2012
Our sponsors as of November 2014.

3. AASPI Software Package
AASPI Welcome 2012
AASPI Software Package
Volumetric Attributes
geometric attributes
spectral decomposition
GLCM texture and disorder attributes
Prestack and Poststack Data Conditioning
pre- and post-stack structure-oriented filtering
residual velocity analysis
non-stretch NMO
footprint suppression
acquisition patch-based coherent noise suppression
preconditioned least-squares time migration
Laplacian of Gaussian fault plane enhancement
The AASPI software package runs on both Linux and Windows. Most applications run in parallel, such that if your PC has 16 processors, it will use all of them if you ask it to. The software is broken into four general areas – volumetric attributes, data conditioning, correlation tools, and cluster analysis (i.e. looking for hidden patterns and relationships in the data). Areas in pink are major additions in 2014, though all codes are routinely modified and extended.

4. AASPI Software Package
AASPI Welcome 2012
AASPI Software Package
Correlation tools
Azimuthal intensity (“fault proximity”)
Vector correlation
Prestack Analysis tools
AVAz
Cluster Analysis
Self-organizing maps
Generative topological maps
Principal component analysis
Support vector machines
Polygonal cluster selection in histogram (SOM)
Page 2 of the previous list.

5. AASPI Software Package
Sponsors have full source code
Data formats are the same as Stanford’s SEPLib and
UT Austin/CSM’s Madagascar
Documentation online and accessible from the program Continue to add details and examples
Same applications, GUIs, and scripts run on Linux and Windows
Computationally intensive algorithm parallelized, running on multi threaded Windows PCs and across Linux clusters. (One sponsor runs across Windows PCs)

6. AASPI Welcome 2012
mcee.ou.edu/aaspi
documentation
We keep all our documentation online under mcee.ou.edu/aaspi . This makes it easy for our sponsors to access, and will also turn up in google-type searches which forms a type of subliminal advertising. In contrast, the software is password protected. Sponsors receive both executables and source code for their use and modification.

7. Most applications have a flow chart
AASPI Welcome 2012
Most applications have a flow chart
Here is a representative page from our documentation. Most applications include a flow chart.

8. Many applications now have gray “Theory” boxes
AASPI Welcome 2012
Many applications now have gray “Theory” boxes
Many documentation of many applications (and more as we proceed) include a theory section, often in a gray box, to allow the interested user to see exactly how a given code has been implemented. This theory is augmented by publications also found on the same web site, where greater detail and case studies are presented.

9. AASPI Welcome 2012
Followed by instructions defining parameters and how to complete the GUI
The documentation has a step by step instruction, much as I would teach lab exercises in a classroom environment. All software is accessed via graphics user interfaces (GUIs). When the mouse is moved over various parameters, simple help and tip comments will appear.

10. These are followed by references and examples
AASPI Welcome 2012
These are followed by references and examples
We also attempt to include case studies and references for further insight as to how the software can be used.

11. Snapshot of some recent software releases and workflows
AASPI Welcome 2012
Snapshot of some recent software releases and workflows
Here are some recent releases as of the end of 2014.

12. Multiattribute DisplayN 1
Time (s)
1 mi
Azimuth of Reflector Convergence modulated by its Magnitude
co-rendered with Variance and Amplitude
Pinchout
Karst
Unconf
Karst
Pinchout
Pinchout
Fault
Grooves
Pinchout
-180
+180
Opacity (%)
100
Converge Azim
(degrees) N S E W
Conv Mag
Vertical slices and a time slice at t=1.150 s through the azimuth of reflector convergence index plotted against hue as the background and the magnitude of reflector convergence plotted against saturation. Seismic amplitude plotted on the vertical slice, and variance plotted on the time slice against lightness. Note the W-NW trending pinch-out seen in cyan on the vertical slice that can be traced across the time slice.
Fault
0.004
100
Opacity (%)
Conv Mag
0.0
0.2
0.5
100
Opacity (%)
Variance
0.1
0.3
0.4
-2500
+2500
100
Opacity (%)
Amp

13. Correlation of production to seismic attributes: Miss. Lime Kay Co
Correlation of production to seismic attributes: Miss. Lime Kay Co., OK ZP
(g-ft/cm3-s)
60000
40000
Grayscale
Amplitude 1 -1
Pre-stack ZP impedance along Mississippi horizon shown with well bores and cumulative oil production from the Mississippi Lime posted as green circles at the top of the well bore. Amplitude (grayscale) shown in cross-line and inline. Magnitude of oil production is directly related to the size of the circle.
Figure 64. Pre-stack ZP impedance along Mississippi horizon shown with well bores and cumulative oil production from the Mississippi Lime posted as green circles at the top of the well bore. Amplitude (grayscale) shown in cross-line and inline. Magnitude of oil production is directly related to the size of the circle.

14. Correlation of AVAz and curvature?
AASPI Welcome 2012
Correlation of AVAz and curvature?
AVAz k1
Vector correlation
This image shows the results of a new algorithm that correlates two vectors, amplitude vs. azimuth and the strike and magnitude of curvature each of which may be correlated to natural fractures and horizontal stress. This example comes from Osage Co., OK.
Top Miss Lime
Osage Co., OK
Data courtesy of Spyglass LLC

15. Why is coherence so discontinuous?
Seismic Attributes
Pop-up structure (1:1 scale) D D’
-500- k1 k1
-750- k2
Fault
Fault k2 k1
-1000-
0.4
Amp
Time(ms)
pos
-1500-
-0.4 k2
neg
0.4
-1750-
This image comes from Onur Mutlu’s M.S. thesis analyzing the unconventional Chicontepec reservoir of Mexico, where he shows that the lack of fault lineaments seen in coherence is due to overprinting of the image by strong, coherent interbed multiples. All figures are displayed with 1:1 ratio. Positive curvature anomalies are observed on the upthrown blocks, whereas negative curvature anomalies are observed on the downthrown blocks.
-2000-
-0.4
-2250-
1 km
Why is coherence so discontinuous? 28

16. Poststack structure-oriented filtereing
AASPI Welcome 2012
Poststack structure-oriented filtereing …
(1, 9)
trace ,
9 x 9 covariance matrix k
k+1
k+2
k-1
k-2
time
Inline
crossline
slice dk
We have also made considerable progress in data conditioning. This image shows the computational stencils for post-stack single window structure-oriented filtering. Bo Zhang has worked on extending this algorithm to prestack migrated data.
Figure 1a.

17. Poststack structure-oriented filtereing
AASPI Welcome 2012
Poststack structure-oriented filtereing
9 x 9 covariance matrix
“eigenmap” v1
Obtain α by cross correlating with slice dk
PCA=α v1
The workflow of poststack structure-oriented filtering. In program sof3d, we compute a covariance matrix about each window oriented along structural dip, compute eigenvectors, and project them onto the data slice resulting in a principal-component filtered data volume. Such filtering preserves amplitude and suppresses cross-cutting noise. In actual implementation, we would chose the most coherent of 9 overlapping windows (for this stencil) to perform our filter.
Figure 1b.

18. Prestack structure-oriented filtereing
offset m-1 k
k+1
k+2
k-1
k-2
time
Inline
crossline
slice dk
offset m
offset m+1 …
(1,
trace ,
27)
27 x 27 covariance matrix
In prestack structure-oriented filtering we filter not only along structural dip for a given offset and azimuth, but also across adjacent offsets and azimuths. Several filters are provided, with non-linear LUM and alpha-trimmed mean filters working better for data with spikes in them.
Figure 2a.

19. “eigenmap” v1 to obtain the signal component for slice dk
Poststack structure-oriented filtereing
27 x 27 covariance matrix
“eigenmap” v1 to obtain the signal component for slice dk
Obtain α by cross correlating with slice dk
PCA=α v1
Figure 2b.

20. Build sample vectors for offset m at analysis
Structure oriented
filtered gathers
Migrated seismic
gathers
Build sample vectors for offset m at analysis
index k along local reflector’s orientation
Is there a significant
discontinuity at analysis index k?
Yes No
Obtain filtered amplitude using “eigenmap”
Stacking
Estimate the reflectors orientation
Stacked Volume
Coherence
Inline dip
Crossline dip
Estimate the discontinuity
More traces or samples
It looks complicated, but this is what is internal to prestack SOF. A workflow GUI drives the process.

21. Time-migrated gather (Barnett Shale)
AASPI Welcome 2012
Time-migrated gather (Barnett Shale)
t0 (s)
Offset - x(m)
2100
4200
0.5
1.0
2.0
1.5
0.0 -4 -2 4 2
Amp
An example of a common reflection point gather from the Fort Worth Basin (data courtesy of Devon Energy).
Figure 4a.

22. After prestack SOF Figure 4b. AASPI Welcome 2012 t0 (s) Offset - x(m)
2100
4200
0.5
1.0
2.0
1.5
0.0 -4 -2 4 2
Amp
The same data after prestack SOF using 3 inlines, 3 crosslines, and 3 offsets (i.e. a 27x27 covariance matrix)
Figure 4b.

23. Rejected noise Figure 4c. AASPI Welcome 2012 t0 (s) Offset - x(m) 2100
2100
4200
0.5
1.0
2.0
1.5
0.0 -1 1
Amp
Here is the rejected noise plotted at the same scale. So ends a short review of some of the effort presented at the end of Much greater detail can be found in the documentation, theses, dissertations, and published papers and abstracts.
Figure 4c.