Hybrid Seismic Attributes for identifying geological Features

and fractures intensity using North Sea Seimic data

SHIHATAA, M. I.

STO Company, Cairo, Egypt

Abstract
Seismic attributes used to identify and isolate important geological features
from seismic data, while no unique attribute is expected to perfectly identify the
targeted object, various attributes contributing to the same purpose should be utilized
simultaneously when performing detection. In this work we present new hybrid
attributes generated by combining various seismic attributes to enhance identifying of
interested geological features from seismic data, by combining different spectral
bands frequencies to increase signal-to-noise ratios, one of new hypride attributes
average SD(spectral decompositions ) attributes, this attributes generated by
combination divergent types of seismic attributes to eliminate noises effect and reduce
effect of un wanted geological feature, average SD attribute used to generate
similarity attribute to improve shallow channel detection and guidance to determine
gas migration pass, it is important to combine faults attributes with amplitude
attributes to identify faults trends, To validate the proposed method we use the
volume of the Netherlands offshore F3 block downloaded from the Open Seismic
Repository, average SD deliver promising results for both shallow and deep thin
geological features interpretation because it combine different bands frequencies in
one volume. Furthermore, the results show that average SD attributes can use for
predict gas migration path and faults attributes help for identify shallow minor faults.

Introduction

Seismic attributes are defined as any measure of seismic data that helps to
visually enhance or quantify features of interest. A good seismic attribute is either
directly sensitive to the desired geologic feature or reservoir property or allows us to
define the structural or depositional environment and thereby enables us to infer some
features or properties of interest (Chopra and Marfurt, 2007). In the last decades
numerous published works have documented the successful use of seismic attributes
to explore for hydrocarbon-bearing sediments and to extract key information about
their lithology and their different saturating fluids (Hardage et al., 1996a; Chopra and
Marfurt, 2007; Chen et al., 2008).

Spectral-domain seismic data attributes have been useful for some applications in
hydrocarbon-reservoir characterizations. For example, Dilay and Eastwood (1995)
analyze seismic data in the spectral domain for monitoring bitumen production by
cyclic steam stimulation (steam injection)1 at Cold Lake,Alberta, Canada. Partyka et
al. (1999) discuss spectral-decomposition analysis and interpretation of 3D seismic
data. Extracting the spectral components at different dominant frequencies may
provide more precise perspectives of given geologic structures. For example, the
thickness of a channel and its spectral amplitude are strongly correlated (Laughlin et
al., 2002). spectral decomposition could be used to image hydrocarbon sands a
certain frequency bands (Burnett etmal., 2003; Sinha et al., 2003). The seismic
response of a given geologic feature is expressed differently at different spectral
bands. Often, a particular frequency component carries the information regarding
structure and stratigraphy. Spectral decomposition methods map 1D signal into the 2D
time and frequency plane, generating amplitude and phase spectral components
(Castagna et al., 2003). Sun et al. (2010) use discrete frequency coherence cubes in
fracture detection and find that high-frequency components can provide greater detail
Combination spectral decomposition. Farfour and Youn (farfour and youn, 2012) used
frequency decomposition for delineating stratigraphic traps and identifying subtle
frequency variations caused by hydrocarbons. The application of complex spectral
coherence shows that it is useful for detecting different-scale structural and stratigraphic
discontinuity features (Li and Lu, 2014).

In this work, we used different hybrid attributes to identify important

geological features that hard to determine by unique attribute, average SD attributes
has been developed based on seismic spectral decomposition analysis, this method
was started by removing high and low frequencies noises depend on our targets
frequencies band and used mean smooth filter to reduce effect of foot print noises, our
first target to generate new hybrid attribute (average SD) to identify thin shallow
channels trend, first step depend on determine channel dominant frequency using
tuning thickness analyses for extracted wavelet. Then generate spectral band
frequencies around dominant frequency .Finally, average SD attribute was generated
to enhance thin channel interpretation. Calculation similarity attribute by average SD
shows that it is useful for enhancing thin geological features interpretation and obtains
promise results for shallow and deep geological features interpretation. In order to
evaluate the proposed method, we use the volume of the Netherlands offshore F3
block downloaded from the Opendtect website and compare the obtained results with
normal amplitude and spectral decomposition attributes, we conclude that this new
simple average attributes help to identify thin channels with different frequencies
bands.

2
Geologic Background and Seismic Data
F3 is a block in the Dutch sector of the North Sea (Figure 1). The block is
covered by 3D seismic that was acquired to explore for oil and gas in the Upper-
Jurassic – Lower Cretaceous strata, which are found below the interval selected for
this demo set (Figure 2) . The upper 1200ms of the demo set consists of reflectors
belonging to the Miocene, Pliocene, and Pleistocene. The large-scale sigmoidal
bedding is readily apparent, and consists of the deposits of a large fluviodeltaic
system that drained large parts of the Baltic Sea region (Sorensen, 1997; Overeemetal,
2001).

The structural and depositional development of the southern North Sea basin
has been well documented. At the large scale the Southern North Sea sedimentary
basin can be seen as a basin dominated by rifting during most of the Mesozoic with a
Cenozoic post rift sag phase. Rifting already started in the Triassic, and culminated in
the Jurassic and Early Cretaceous with the various Kimmerian extensional tectonic
phases related to the opening of the Atlantic Ocean. Active rifting was followed by a
post-rift sag phase from Late Cretaceous to Present, which was mostly characterized
by tectonic quiescence and subsidence of the basin, with the exception of a few
compressional tectonic pulses during the Late Cretaceous and Tertiary. During most
of the post-rift phase the basin accumulated thick sedimentary mega-sequence (
Schroot, B.M., 2002)

Figure 1. Satellites map of F3 a block in the Dutch sector of the North

Sea.

3
 Figure 2. Netherlands offshore sector. Showing license blocks. Locations of 2D and
3D Survey.

Only in the very south the Pliocene-Pleistocene is overlying much older

Tertiary deposits. In the same area crag-like deposits were very locally deposited
Pliocene-Pleistocene times, similar to those presently outcropping in East Anglia
(Cameron et al, 1989a). Coastlines shifted back and forth over the Netherlands North
Sea and surrounding areas from the end of the Pliocene onwards (Sha, 1991) leading
to a variety of sedimentary environments and grain sizes deposition

Gological iterpretattions

Geological crosection between wells used to detect lateral lithology variation

between wells , figure 3 display crosection between wells and fill by color to help to
detect change in lithology and identify uplifting result from salt diaper , salt diaper
upliftening effect in migration of gas in shallow depth. Figure 4 show Depth
crosection fill by GR log change in facies between sand and shale in shallow .

4
 Figure 3. Depth crosection using formation tops grids to identify uplifting effect of salt
dome.

Figure 4. Depth crosection fill by GR log change in facies between sand and shale in
shallow .

5
Average SD (spectral decomposition) Attributes Workflow

Spectral decomposition was expected to reveal stratigraphic features of the

channel that could not be seen in seismic images. To accomplish this, different
frequencies were calculated for a single time slice at this interval (Farfour and Youn ,
2012). Over the last decades, several studies have demonstrated that spectral
decomposition can provide more interpretable results if it is integrated with edge
attributes. To handle this problem, it is prefer to divide seismic data to several spectral
bandwidth and average the best three frequencies bands to generate new hybrid
average SD attribute (Figure 5).
Similarity is an ideal attribute in mapping lateral variation in waveform within
defined time window; but it is relatively insensitive to amplitude change. In a very
thin bed reservoir, the below tuning implies that the waveform stabilizes and only
seismic amplitude changes; thus, similarity is not the appropriate attribute. On the
other hand, spectral decomposition is known to be a good indicator of amplitude
change

Figure 5. An example broadband trace (left), its spectrogram (middle) with

the limiting frequencies indicated in white and the band-limited reconstructions
(right) for the three frequency bands.( Lowell, J., Eckersley, A., Kristensen, T.,
Szafian, P.,2014)

Average SD attributes depend on detecting dominant frequency for interested

geological features time window, first similarity attribute was used to identify shallow
channels time window from 0.8 s to 1.04s, then Dominant frequency found by tuning

6
 thickness analysis using extracted wavelet surrounded interested channel interval.
Finally, average three frequencies combined around dominant frequency to generate
Average SD attributes that reduce effect conflict of other uninterested spectral band
and eliminated noises effect of other bands.

Figure 6(a) shows Survey spectrum at survey time interval where there are
different band widths interfere with interested channel band width frequencies. Figure
3(b) presents Survey spectrum at channel interval shows dominant frequency around
60 Hz. A noticeable decrease from 90 to 60Hz is associated to high Frequency
attenuation and absorption while traveling to deeper formations

a)
b)

Figure 6. a)Survey spectrum at survey time interval, b) Survey spectrum at channel

interval shows dominant frequency around 60 Hz.

A layer is called a thin layer when 1 < λ/d ≤ 4, and an ultra-thin layer when,
λ/d > 4, where λ is the dominant wavelength within the layer and d is the layer
thickness (Liu and Smith 2003). Tuning Analysis allows analyzing tuning thickness
from frequencies content of the wavelet. Geologic layers did not identified at one
frequency/wavenumber or in a broadband display may be prominent at the specific
tuning frequency that relates to the actual layer thickness. It is important to understand
that spectral decomposition can reveal the acoustic response related to certain
thicknesses. The interpreter must determine whether this spectral decomposition
acoustic response relates to actual bed thickness. Tuning and survey spectrum analysis
was run at this channel interval inferred that the dominant frequency was around
60Hz (Figure 7).
Where tuning thickness = 1/4* λ
Actual time thickness need for tuning thickness = 1/4* P
So p = 4*(actual time/2) =4*.008/2= 0.16 s (1)

7
FD = 1/p = 1/.0 16= 62.5 Hz (2)
Where λ = wave length, p = periodic time, FD= dominant frequency.
Spectral decomposition calculated for different bands width frequencies
around dominant frequency, tuning curve analysis used wavelet extracted around time
window of shallow channel from 0.8 to 1.04 s, Tuning and survey spectrum analysis
was run at this channel interval inferred that the dominant frequency was around
60Hz from equation (1) and (2) where P is periodic time and FD represent dominate
frequency.

Figure 7. Tuning analysis for extraction wavelet.

I3D (Illuminator-3D) attributes application

A variety of different seismic attributes, such as Symmetry and Similarity for
example, can reveal and display fault patterns in a formation. However, actual fault
patterns in a formation may not be continuous, and a single fault may appear as a
combination of seemingly isolated parts. In addition, horizontal footprints may coexist
in the fault attributes in great numbers further obscuring the faults. Fault analysis can
be done more easily if isolated parts of a single fault can be connected together into a
single piece, while footprints of low dips can be removed. The I3D algorithm (patent
pending) performs these operations which enhance the fault image in all spatial
directions. I3D Energy, Dip, and Azimuth are generated to represent the fault
distribution patterns in the fault attribute volume. Enhancing the fault attributes
improves automatic and manual fault extraction workflows, regardless of the fault
attributes that are being enhanced. Figures 8 shows the resulting attributes present
smoother and cleaner curve lines or plane patterns of sharper contrast with additional
dip and azimuth information. .

8
One unique feature of this attributes is that it does not require a spatial context
window. It is inspired by the neuronal mechanisms of the primary visual cortex for
orientation perception (Yingwei Yu, Cliff Kelley, and Irina Mardanova,2013).

The orientation energy E reflects the strength of orientation features. The low
values of orientation energy mean that there are fewer oriented patterns in the
neighborhood, while the stronger ones mean the orientation feature is more salient in
the context. Figure 9 shows an example of the orientation vector field (OVF)

Figure 8: Rotational Symmetry in a 3D Seismic Volume

Figure 9. Orientation Vector Field near a Salt Dome. The orientation vectors (red) are
plotted on top of the seismic image in a region near the salt dome. The magnitudes of
the vectors are normalized (modified after Yu, Kelley and Mardanova 2013)

9
Results

Compared results of spectral decomposition frequencies confirm our proposal

dominant frequency where edge of channel have been enhanced after extracted SD
frequency 62 Hz, Figure 10 compares between normal amplitude and amplitude for
SD frequency 62 Hz at same time slice there are improve in edge of channel and
increase resolution of reflectors.

Figure 10. Normal amplitude slice at 1.036 s (left), amplitude for SD frequency 62 Hz
(right).

Average SD attributes calculated by combining best three frequencies around

thin channel dominant frequency to enhance channels edge and depend on determine
dominant frequencies by tuning analysis (Figure 11), this attribute combine different
band frequencies to enhance thin channels, Figure 11 shows average SD merge the
best three frequencies around dominant frequency (55-64-70 Hz) at 1.036 s to enhance
channel edge compare with normal amplitude at same time slice, this attribute add
valuable geological information
Similarity is an ideal attribute in mapping lateral variation in waveform within
defined time window; but it is relatively insensitive to amplitude change. In a very
thin bed reservoir, the below tuning implies that the waveform stabilizes and only
seismic amplitude changes; thus, similarity calculated by normal amplitude is not the
appropriate attribute. On the other hand, spectral decomposition is known to be a
good indicator of amplitude change, to handle this problem, average SD
attributes used to calculate similarity attributes to enhance subtle channel detection
better than using normal amplitude, combination best detected three frequencies
bands that reduce random noises and maximize amplitude for interested stratigraphic
features and reduce effect of conflict of unwanted signal
10
. Figure 12 compares between similarity generated by normal amplitude and average SD
attribute, subtle thin channels system easily identified in right image especially in middle
and in east part. In other side, noises and unwanted bands signals reduce channels system in
left image.
a
. g
s

Figure 11. Amplitude slice at 1.036 s for SD frequency 62 Hz (left), Average

attributes time slice at 1.036 enhance channel image (right).

Figure 12. Comparing between similarity attribute calculated along normal amplitude
(left) and similarity attributes calculated along average SD (55-64-70 Hz) attributes
(right) with white black color.

11
 There are another important advantage for average SD attributes it can use for
DHI and reduce noise for similarity attributes results, figure 13 compare between
amplitude and average SD attributes for arbitrary line between wells, left map
present amplitude attributes where it is hard to identify shallow gas indication,
right line represent average SD attributes average four bandwidth frequencies (8,
25, 40, 60 Hz) that determined from spectral analysis for seismic cube.

Figure 13. Normal amplitude attributes for arbitrary line between wells (left), average
SD attributes for arbitrary line between wells (right).

Figure 14 show usage of average SD attributes to identify shallow bright spot that
hard to detect by normal amplitude, right map represent amplitude time slice at 0.624,
left map show average SD attributes at 0.624 where two black circler isolate two
important bright spots that hard to detect in right map.

Figure14: Normal amplitude time slice at 0.624 (left), average SD attributes time
slice at 0.624 (right).
12
it is important to eliminate effect of noise in the similarity attributes results to enhance
geological features detection, calculated similarity attributes using average SD give
good result for fault detection and reduce effect of noise , figure 15 show comparison
between similarity attributes calculation

Figure15. Similarity attribute calculated using normal amplitude(left), similarity

attribute using average SD (right).

Figure 16 represent comparison between symmetry attributes and new fault

attributes, right figure represents symmetry attributes time slice at 0.624 where it is
hard to identify faults because noises effect on results, left figures represent new fault
I3D illuminators energy attributes where it enhance fault image and reduce effect of
noise because it depend on orientation pattern is analyzed in frequency domain, and
inspired by the neuronal circuits in the biological brain.

Figure 16. Symmetry attribute at time slice 0.640 (left), I3D energy attribute
calculated from symmetry attribute (right).

13
Blend fault attributes I3D energy with edge attributes enhance fault image, multi-
attributes help to identify faults trends and reduce risk of seismic interpretation
(Figures 17).

Figure 17. Blend fault attributes I3D energy with amplitude attributes to enhance
faults interpretation and aid to identify edge of gas chimney and reduce risk.

Dip maximam similarity atteibute is very important to identify geobodies

with highly dip and high contrast between surrounding lithology, there are a lot of
geobodies effected by gas migration from gas chimney unfortunately geometric
attributes alone hard to identify lithology change or predict gas accumulation but can
identify edge of geobodies and edge of gas accumulated, it is important to combine
physical with geometric attributes for identify geobodies and lithology change
(Figure18,Figure19).

14
Figure 18. Shallow geological features may be indicate for gas migrated and
accumulated.

Figure 19. Blending average energy with similarity attributes indicate shallow gas
accumulated.

15
 Fractures detection is very important, it is difficult to detect fracture trend from
seismic data because of small resolution of seismic data and seismic noise effect
resulting from fracture, preconditioning seismic date used to overcome noise effect
and enhance seismic resolution.

The I3D Scan algorithm convert I3D Energy volumes to new fault segments
volume using a set of image recognition algorithms depend on thin segments methods
, improve connectivity between small segments, and remove noise from the fault
images identified by the I3D algorithms .Figure 20 represent time slice at 1.366 to
explain fracture effect but it is difficult to detect fracture trend, figure 21explain how
blending techniques can help to detect fault trends by bled fault scan attributes with
normal amplitude.

:
Figure 20. Amplitude slice at 1.366 S represent fracture distribution gas chimney

16
 Figure 21. Blend new scan attribute with amplitude

Figure 22 explain effect of change amplitude color bare enhance blending between fault
scan attributes and amplitude and improve fracture and faults detections.

Figure 22. Blend new scan attribute with amplitude with different color bare

17
Displaying fault scan attributes in 3D is important to detect faults and fractures , fault
scan attributes connect fracture with each other that help to detect fracture network
.Figure 23 represent fault scan attributes in 3D to detect faults and fractures trends.

Figure 23. Displayin fault scan attributes in 3D to detect fractures and faults
trends.

Transparent technique is one of important methods to display seismic attribute in 3D ,

using opacity methods are important to remove unwanted values from seismic attributes
and render important values for structure or stratigraphic anomalies. Figure 24 explain
how opacity methods enhance fault scan attributes displaying in 3D, fault scan
attributes are important to detect faults and fractures trends in 3D view .

18
 Figure 24. explain how opacity change to enhance fault scan attributes in 3D.

Fault scan attributes can render in 3D and control in opacity view to detect different
geological features, figure 25 show fault attributes render in 3D view to detect salt dome
and fault trends.

Figure 25. Fault attributes render to identify salt and faults trends in 3D
19
 Figure 26 show extraction bright spot geobodies and faults attributes rendering to
detect salt dome and faults, figure 27 illustrate salt dome extraction and
interpreted faults validate with faults attributes

Figure 26. Bright spot geobodies (blue arrows) extraction and detect faults trends (purple arrows)
and salt dome (white arrow) .

Figure 27. Bright geobodies extraction and salt dome geobodies extraction combine with
interpreted faults
20
 Conclusion
average SD attribute used to enhance similarity attributes results and improve seismic
interpretations for shallow, it is important to merge different bands frequencies
cubes in one volume, to handle this problem, average SD attribute was created to sum
absolute values for three bands frequencies and generate one volume for important
bands frequencies, this new hybrid attribute eliminated foot noises effect and reduce
effect of un wanted geological feature, average SD attribute used to generate similarity
attribute to improve shallow channel detection and guidance to determine boundary of
deep reservoir, average SD deliver promising results for both shallow and deep
geological interpretation because it combine different bands frequencies in one volume.

Fault and fracture attributes are important to detect fractures trends and enhance
identifying faults network , combination fault attributes like fault scan with other
physical attributes give promising result to identify how fracture network and study
fracture effect in reservoir.

References

Burnett, M. D., J. P. Castagna, E. Méndez-Hernández, G. Z. Rodríguez, L. F.

García, J. T. M. Vázquez, M. T. Avilés, and R. V. Villaseñor, 2003: Application of
spectral decomposition to gas basins in Mexico: The Leading Edge, 22, 1130–1134.

Cameron, T.D.J., Laban, C. & Schüttenhelm, R.T.E., 1989: a. Upper Pliocene and
Lower Pleistocene stratigraphy in the Southern Bight of the North Sea. In: Henriet,
J.P. & De Moor, G. (eds) The Quaternary and Tertiary Geology of the Southern
Bight, North Sea. Belgian Geological Survey, Brussels: 97-110.
Castagna, J., S. Sun, and R. Siegfried, 2003: Instantaneous spectral analysis:
Detection of low-frequency shadows associated with hydrocarbons: The Leading
Edge, 22, 120–127.
Chopra, S., Marfurt, K.J., 2007:Seismic Attributes for Prospect Identification and
ReservoirCharacterization. Society of Exploration Geophysicists, Tulsa, OK (456
pp.).
Dilay, A., and J. Eastwood, 1995:Spectral analysis applied to seismic monitoring of
thermal recovering: The Leading Edge, 14, 1117–1122.
21
 Farfour, M.; Yoon, W.J.; Jo, Y,2012: Spectral decomposition in illuminating
thin sand channel reservoir, Alberta, Canada. Can. J. Pure Appl. Sci. 6(2), 1981–1990

Hardage, B.A., Carr, D.L., Lancaster, D.E., Simmons, J.L., Hamilton, D.S.,
Elphick, R.Y., Oliver,K.L., Johns, R.A., 1996a: 3D seismic imaging and
seismic attribute analysis of genetic sequences deposited in low
accommodation conditions. Geophysics 61,1351–1362.
Laughlin, K., P. Garossino, and G. Partyka, 2002:Spectral decomposition applied
to 3D: AAPG Explorer, 23,28–31.

Li, F. Y and Lu, W. K., 2014: Coherence attribute at different spectral

scales, Interpretation, 2(1), SA99-SA106.
Liu, Y.; Schmittz, D,2003: Amplitude and AVO responses of a singlethin bed.
Geophysics 68(4), 1161–1168.

Lowell, J., Eckersley, A., Kristensen, T., Szafian, P.,2014: Improvements to

Frequency Decomposition Methodologies for Use with Broad Bandwidth Seismic
Datasets, EAGE Conference & Exhibition 2014 Amsterdam RAI, The
Netherlands,
76th
.

Sha, L.P.(ed.), 1991: Quaternary Sedimentary Sequences in the southern North Sea
basin, Final discipline rept. of the project: The Modelling And Dynamics Of The
Quaternary Geology Of The Southern North Sea And Their Applications To
Environmental Protection And Industrial Developments, CEC DGXII, Scientific
Programme Contract No. SCI*-128-C 9EDB: 135 pp., app
Sinha, S.; Routh, P.; Anno Castagna, J.P,2003.: Spectral decomposition of seismic
data with continuous-wavelet transform. Geophysics 70(06), PP. 19–25

Schroot, B.M., 2002: North Sea shallow gas as a natural analogue in

feasibility studies on CO2 sequestration. In: Extended Abstracts of the 64th EAGE
Meeting and Technical Exhibition, Florence, paper H010: 4 pp.
Sørensen, J. C., Gregersen, U., Breiner, M., and Michelsen, O., 1997:High-
frequency sequence stratigraphy of upper cenozoic deposits in the central and
southeastern north sea areas. Marine and PetroleumGeology, 14(2):99–123.
Szafian,p
.
Partyka, G. A., J. Gridley, and J. Lopez, 1999: Interpretational applications of
spectral decomposition in reservoir characterization: The Leading Edge,
18,
353–360. 22
Yu, Y., C. Kelley, and I. Mardanova,2013 : Volumetric seismic dip and azimuth
estimation with 2D log-gabor filter array, SEG Technical Program Expanded
Abstracts 2013: pp. 1357-1362.

23
24