Seismic Attribute Mapping of Structure and

Stratigraphy

Unit 5: Attributes and the Seismic Processor

Course Outline
Introduction
Basic concepts
Multiattribute display
Spectral decomposition
Geometric attributes
Dip and azimuth
Coherence
Curvature and reflector shape
Lateral changes in amplitude and pattern recognition
Attributes and the seismic interpreter
Structural deformation
Clastic environments
Carbonate environments
Shallow stratigraphy and drilling hazards
Reservoir heterogeneity
Attributes and the seismic processor
Influence of acquisition and processing
Structure-oriented filtering and image enhancement
Prestack geometric attributes
Use of Attributes for Quality Control of
Processing
Figure 5.1 (top)

Field Survey
data data Use of
Merge seismic coherence in
processing
and survey data

Compute field
statics or
QC
refraction statics

Test and apply

deconvolution

Generate and pick

1st pass velocities

Generate initial Calculate

migrate
stack volume coherence

Compute and apply Calculate

migrate
Residual statics coherence
Figure 5.1 (bottom)

Use of
coherence in
processing
QC
Compute and apply
Residual statics

Generate and pick

3-D DMO velocities

Apply 3-D DMO

Calculate
and output stacks or migrate
coherence
offset stacks

Calculate
Apply 3-D migration migrate
coherence
Figure 5.2

Use of coherence in processing QC

Brute stack Residual statics Final migration
Coherence Time Slice Seismic Time Slice
Figure 5.3

Use of coherence in processing QC

Coherence on migrated Coherence on migrated
volume with original volume with improved
velocities velocities
Figure 5.4
Use of coherence in choice of migration velocity

90% Velocity 100% Velocity

Slices through volumes

migrated at 3 different
velocities

110% Velocity
Figure 5.5
Use of coherence in choice of migration velocity

90% Velocity 100% Velocity

Use of coherence as a
migration velocity analysis
tool. Zone circled in red are
better focused than those
circled in yellow.

110% Velocity
Sensitivity of Attributes to Acquisition
Footprint
Figure 3.13b
Impact of acquisition on dip
5 km magnitude
B

Dip
(s/km)
0.00

0.06
Common causes of acquisition footprint

Problems due to acquisition program

• Non-uniform offsets and azimuths in bins
• Non-uniform backscattered noise suppression
• Obstacles such as lakes, villages, or platforms
• Currents and tides
Problems due to processing
• Incorrect velocities
• Migration operator aliasing
Figure 5.6
Acquisition footprint: Delaware Basin, NM, USA
A A
0.0

0.5
Time (s)

Yates

Grayburg

1.0

Gloreitta
Abo
1.5 Wolfcamp
Figure 5.7
Impact of footprint on conventional interpretation products
2 km A

Time (s)

0.6
Time/structure
(little impact of footprint)

0.7
A

Amp
0

Amplitude extraction
(strong footprint)

A
high
Figure 5.8 Impact of footprint on coherence

0.4 s 0.6 s

Typically, footprint heals with

depth
• Examine shallow slices for
footprint pattern
• Interpret slices at deeper
zones of interest with this
pattern in mind
0.8 s
Figure 5.9 Impact of footprint on other attributes

Most negative curvature Most positive curvature

Footprint can impact

amplitude, waveform,
(t=0.8 s) and apparent dip!

EW component coherent energy gradient

 Reducing Acquisition Footprint
through Reprocessing the Prestack Data
Figure 5.10

Original data. Note the E-W acquisition footprint

seismic
coherence

1272 ms 1316 ms 1332 ms

(Chopra et al., 2002)
Figure 5.11

After trace balancing, statics, and improved velocity

analysis to minimize acquisition footprint.
seismic
coherence

1316 1316
ms ms

1272 ms 1316 ms 1332 ms

(Chopra et al., 2002)
Figure 5.12a
Coherence on original data

Time = 0.940 s
(Famini et al, 2003)
Figure 5.12b
Coherence on reprocessed data

Time = 0.940 s
(Famini et al, 2003)
Post stack solutions to acquisition footprint

kx-ky filters applied to time slices

3D dip filters
fxy filters
Eigenstructure (SVD) filters
Figure 5.16
Acquisition footprint suppression via
adaptive filtering

Before footprint removal

After footprint removal

Rejected noise

(Drummond et al., 2001)

Figure 5.17
Acquisition footprint suppression using
5 km
eigenvectors of the noise

Before filtering

After ‘ truncated
SVD filtering’

Seismic Time slices at 0.320 s

(al-Bannagi et al., 2005)
New Figure
Eigenfaces

Avg 1st 2nd 3rd 4th

Average Face
14th

Training Faces Eigenvectors

Each human face has a given cross correlation coefficient with the
eigenvectors (eigenfaces).

3 or 4 coefficients describe most of the variation

(http://cs.haifa.ac.il/~dkeren/IP/short-lecture10.pdf)
Figure 5.18

Acquisition footprint suppression using

5 km eigenvectors of the noise
1.5

2.0
Time (s)

Before filtering
2.5

3.0
1.5

2.0 After ‘ truncated

Time (s)

SVD filtering’
2.5

3.0 (al-Bannagi et al., 2005)

Figure 5.19

Acquisition footprint suppression using

eigenvectors of the noise

Before filtering After ‘ truncated

SVD filtering’

(al-Bannagi et al., 2005)

Influence of Migration Aperture
Figure 5.33a Limits on lateral resolution due to
number of Fresnel zones used
Possible aperture
(survey limits)

Desired
aperture

First Fresnel
zone

(Kozlov et al., 2004)

Figures 5.33b-d

‘Ideal’ migration:
Use all Fresnel
0 5
First Zones
Fresnel
Zone
Conventional
‘efficient’
0 migration:
Use 2 Fresnel
Zones

Scattering
5 Object Imaging
Mute 1st
Fresnel zone

(Kozlov et al., 2004)

Figure 5.34

Scattering Scattering Object

‘Ideal’ migration: Objection Imaging: Imaging:
Use all Fresnel Gentle mute of 1st Harsh mute of 1st
Zones Fresnel zone Fresnel zone

(Kozlov et al., 2004)

Figure 5.35

Depth slices through Depth slices through

coherence volume seismic data after
generated from scattering object
‘ideal’ migration imaging

842 m

1762 m

(Kozlov et al., 2004)

Influence of Increasing Spectral
Content
Figures 5.36a and b

Impact of impedance inversion on resolution

Time slice through seismic Time slice through impedance

Figures 5.36c and d

Impact of impedance inversion on resolution

Time slice through Time slice through

coherence generated from coherence generated from
seismic impedance
Course Outline
Introduction
Basic concepts
Multiattribute display
Spectral decomposition
Geometric attributes
Dip and azimuth
Coherence
Curvature and reflector shape
Lateral changes in amplitude and pattern recognition
Attributes and the seismic interpreter
Structural deformation
Clastic environments
Carbonate environments
Shallow stratigraphy and drilling hazards
Reservoir heterogeneity
Attributes and the seismic processor
Influence of acquisition and processing
Structure-oriented filtering and image
enhancement
Prestack geometric attributes
5b: Structure-oriented filtering and image
enhancement
After this section you should be able to:
• Choose a filtering technique that follows structure and
suppresses cross-cutting noise.

• Identify the limits of attribute analysis on data that have

been poorly imaged.

• Recognize the need for special filters to preserve narrow

lineaments such as fractures on attribute volumes.

• Evaluate the application of image processing to enhance

attribute volumes.
Figure 5.39

Running Window Smoothing (RWS)

Output = Input * Gate function

Input

Gate

Output

(Luo et al., 2001)

Figure 5.40

Effect of Running Window Smoothing

Amp of
Input
Filtered
Input Output
Amp
of
Filtered
Output

Inline or crossline no.

(Luo et al., 2001)

Figure 5.41

Structure-oriented filtering (using diffusion equation)

Time (s)

Original migrated data

Time (s)
Time (s)

Without edge preservation With edge preservation

(Hoecker and Fehmers, 2002)

Figure 5.42
Structure-oriented filtering
(using diffusion equation)
Input seismic
volume

estimate the reflector orientation

(dip and azimuth)

estimate major discontinuities

using a coherence or
curvature estimation

yes Is there a significant no

discontinuity
In the analysis window?

Smooth along
don’t smooth yes
dip/azimuth

yes no no
Do you wish to Structure-
More traces or samples?
iterate again? Oriented
filtered
volume
Figure 5.43

1.0
before after Structure
-oriented
1.1 filtering
Time (s)

data
1.2

1.3

autotracked
volumes

Filtering preserves major discontinuities yet allows

use of fast autotrackers
(Hoecker and Fehmers, 2002)
Figures 5.44a and b
Arabian Peninsula
Impact of structure-oriented filtering on
seismic data
0.0 1 km
Time (s)

0.5

1.0

Before After

(Masaferro et al, 2003)

Figures 5.55c and d

Coherence applied to data before and after

structure-oriented filtering (t=0.440 s)
4 km

Before After

(Masaferro et al, 2003)

Figure 5.45

Running-window vs. edge-preserving (Kuwahara) smoothing

Signal Signal + noise
1.5 1.5

Filter
1.0 1.0 length

Amp
Amp

0.5 0.5

0.0 0.0

Running-window Edge-preserving
smoothing smoothing
1.5 1.5

1.0 1.0
Amp

Amp
0.5 0.5

0.0 0.0

(Luo et al., 2001)

Figure 5.46a
Amplitude Time Slice 1.026 s
(Saudi Arabia)

(Luo et al., 2002)

Figure 5.46b
Aramco amplitude gradient edge detection
(without edge-preserving smoothing)

(Luo et al., 2002)

Figure 5.46c
Aramco amplitude gradient edge detection
(with edge-preserving smoothing)

(Luo et al., 2002)

Figure 5.46c
Marfurt’s interpretation

(Luo et al., 2002)

Figure 5.55a Limitations to structure-oriented filtering
(Fort Worth Basin, USA)

Coherence on original seismic data

Figure 5.55b Limitations to structure-oriented filtering
(Fort Worth Basin, USA)

Incoherent
collapse
features are
suppressed!

Coherence on seismic data after 2 passes of filtering

New Figure Limitations to structure-oriented filtering
(South Marsh Island, Gulf of Mexico, USA)

Fractional
Derivative of
Amplitude

t = 1.296 s

(Data courtesy of Fairfield)

New Figure Limitations to structure-oriented filtering
(South Marsh Island, Gulf of Mexico, USA)
Fractional
Derivative of
Amplitude
after 2 passes of
structure-
oriented filtering

Incoherent
collapse
features are
suppressed!

t = 1.296 s

(Data courtesy of Fairfield)

Figures 5.56a and c
Impact of migration velocity on attributes
coherence time slice at 0.600 s
On seismic data after 2
passes of principal
On original seismic data
component filtering

(Marfurt et al., 2002)

Figures 5.56b and d
Impact of migration velocity on attributes
coherence time slice at 1.500 s
On seismic data after 2
passes of principal
On original seismic data
component filtering

(Marfurt et al., 2002)

Figure 5.57
Impact of migration velocity on seismic
vertical slices
2 km
0A A A A A

1
Time (s)

3
Migrated data After 3 passes Rejected noise
of principal
component
filtering
(Marfurt et al., 2002)
Figure 5.58
Impact of migration velocity on coherence
vertical slices
2 km
A A
0A A

1
Time (s)

3
Coherence on Coherence on
migrated data filtered data

(Marfurt et al., 2002)

Figure 5.59a

Impact of migration velocity on attributes

(dip, azimuth, coherence time slice at 0.750 s)
+ crossline

+ inline
- inline
no dip

azimuth

max dip
-crossline
A A’
Figure 5.60
Impact of migration velocity on seismic
vertical slices
6 km

1.0
Time (s)

1.5

Input data After 3 passes of edge-

(no smoothing) preserving smoothing
(Data courtesy of OPEX)
 Objectives of image enhancement
(applied after attribute generation)
• To improve the overall signal-to-noise ratio

• To enhance features of particular interest

(typically faults)

• To convert these images into fault planes,

fracture density measures, or other ‘objects’ useful
to the interpreter, thereby accelerating the
interpretation
Figures 5.62 and 5.63

Image enhancement for fault extraction

5 km
G G
0.0 1. Coherence computation

2. Suppression of horizontal discontinuities

3. Image dilation

4. Image erosion.
Time (s)

5. Merge with seismic in workstation

4.0

Vertical slices

(Barnes, 2005)
Figures 5.64-5.65
Image enhancement for fault extraction
5 km G

Time slices

G
1. Coherence computation
2. Suppression of horizontal discontinuities
3. Image dilation

4. Image erosion.
5. Merge with seismic in workstation (Barnes, 2005)
Course Outline
Introduction
Basic concepts
Multiattribute display
Spectral decomposition
Geometric attributes
Dip and azimuth
Coherence
Curvature and reflector shape
Lateral changes in amplitude and pattern recognition
Attributes and the seismic interpreter
Structural deformation
Clastic environments
Carbonate environments
Shallow stratigraphy and drilling hazards
Reservoir heterogeneity
Attributes and the seismic processor
Influence of acquisition and processing
Structure-oriented filtering and image enhancement
Prestack geometric attributes
Offset-limited volumes
 Depth
Figure 5.69

Time

Structural and stratigraphic effects

Amplitude vs. Offset

Lithology and fluid effects

Figure 5.70
South Marsh Island, Gulf of Mexico

Coherence on far offsets

1 km

Coherence on near offsets

Figure 5.72
South Marsh Island, Gulf of Mexico

NS coherent amplitude
gradient on far offsets

1 km

NS coherent amplitude
gradient on near offsets
Figure 5.73

Tertiary Gas
Reservoir
(North Sea)

Coherence on near
stack

Coherence on full stack

Coherence on far stack

Azimuth-limited volumes
Figure 5.75

Coherence applied to Azimuthally-Limited

Volumes

All-azimuth volume Limited-azimuth volume

Time slices at 1.514 s

Figure 5.78a

Azimuthal velocity anisotropy vs. induced

fractures (Fort Worth Basin, Texas, USA)
vfast-vslow
(m/s)
1000

500
0.5 km

Poor well (fractures parallel)

(Simon, 2005)
Figure 5.78b
Azimuthal velocity anisotropy vs. induced
fractures (Fort Worth Basin, Texas, USA)
vfast-vslow
(m/s)
1000

500
0.5 km

Good well (orthogonal fracture sets)

(Simon, 2005)
Figure 5.79 Azimuthal velocity anisotropy vs. most positive
curvature (Fort Worth Basin, Texas, USA)

vfast-vslow
(m/s)
1000

1 km
500

(Simon, 2005)
Compressional vs. Converted Wave
Volumes
Figure 5.80
Compressional (PP) wave images
(Barinas Basin, Venezuela)

Coherence EW coherent NS coherent

amplitude amplitude
gradient gradient

Horizon slices along reservoir

(D’Agosto, 2003)
Figure 5.81 Fractured Core
Converted (PS) wave images
(Barinas Basin, Venezuela)
2 km
N

Coherence EW coherent NS coherent

amplitude amplitude
gradient gradient

Horizon slices along reservoir

(D’Agosto, 2003)
5c: Prestack Geometric Attributes

In Summary:
• In general, attributes computed from near-offset seismic volumes are more
sensitive to structure and stratigraphy.

• In general, attributes computed from far-offset seismic volumes are more

sensitive to lithology and hydrocarbons.

• Azimuthally limited volumes can improve fracture illumination by

• decreasing smear due to not accounting for velocity anisotropy, and
• eliminating seismic azimuths that do are relatively insensitive to the
strike of the fractures.

• Velocity anisotropy can be due to either fractures or stress; in general,

reflector curvature has little correlation to present-day stress.

• Since PP and PS reflectivities are different, attributes computed from PP

and PS volumes illuminate different geologic features.
 Advantages and disadvantages of
seismic attributes

• data compression (onto a target sample or horizon)

• feature extraction (of a given physical or user-defined
model)
• feature communication/visualization
• information loss (some would call this interpretation!)
• dependent on data acquisition, processing and imaging
• QC on data acquisition, processing, and imaging
Seismic Attribute Mapping of Structure and
Stratigraphy
Sponsored by