reprocessing 3d seismic data for quantitative interpretation forth
TRANSCRIPT
UNIVERSITY OF OKLAHOMA
GRADUATE COLLEGE
REPROCESSING 3D SEISMIC DATA FOR QUANTITATIVE INTERPRETATION,
FORT WORTH BASIN, JEAN, TEXAS
A THESIS
SUBMITTED TO THE GRADUATE FACULTY
in partial fulfillment of the requirements for the
Degree of
MASTER OF SCIENCE
By
MARCUS CAHOJ
Norman, Oklahoma
2015
REPROCESSING 3D SEISMIC DATA FOR QUANTITATIVE INTERPRETATION,
FORT WORTH BASIN, JEAN, TEXAS
A THESIS APPROVED FOR THE
CONOCOPHILLIPS SCHOOL OF GEOLOGY AND GEOPHYSICS
BY
______________________________
Dr. Kurt Marfurt, Chair
______________________________
Dr. John Pigott
______________________________
Dr. Oswaldo Davugustto
© Copyright by MARCUS CAHOJ 2015
All Rights Reserved.
iv
Acknowledgements
Completing this thesis would not have been possible without the help of many
people. First, I would like to thank the ConocoPhillips School of Geology and
Geophysics at the University of Oklahoma for providing me the chance to attain a
Master’s degree in Geophysics. Dr. Kurt J. Marfurt has been fundamental in my
development as a scientist and person. Dr. John Pigott and Dr. Oswaldo Davogustto
have both been crucial in providing me with geophysical and geological insight. Next, I
would like to express gratitude to Mike Burnett of TameCat llc for providing a license
to the data that I used in this thesis. Many students, current and former, friends, and
staff at the university have also aided in my research and completion of my thesis.
Those individuals include Thang Ha, Sumit Verma, Tengfei Lin, Jie Qi, Fangyu Li, Tao
Zhao, Gabriel Machado, Bryce Hutchinson, Joseph Snyder, Oluwatobi Olorunsola,
Lanre Aboaba, Abdulmohsen AlAli, Dr. Shiguang Guo, Dr. Bradley Wallet and many
others. I cannot thank each and every one of you enough for the knowledge and
companionship you have shared with me. I would also like to thank Schlumberger for
the use of VISTA and Petrel as well as CGG for the use of HampsonRussell. Without
the availability of this software at the University of Oklahoma I could not have
completed this thesis. Lastly, I would like to thank my family for their continuous
support during the pursuit of my Master’s degree.
v
Table of Contents
Acknowledgements ......................................................................................................... iv
List of Tables .................................................................................................................. vii
List of Figures ................................................................................................................ viii
Abstract ......................................................................................................................... xvii
Chapter 1: Introduction ..................................................................................................... 1
Objectives ................................................................................................................... 3
Location ...................................................................................................................... 4
Chapter 2: Geological Background .................................................................................. 7
Regional Geology ....................................................................................................... 7
Local Geology ............................................................................................................ 8
Chapter 3: 3D Prestack Seismic Processing ................................................................... 17
Data Acquisition ....................................................................................................... 19
Data Loading and Geometry .................................................................................... 21
Trace Editing ............................................................................................................ 24
Elevation and Refraction Statics .............................................................................. 25
Deconvolution .......................................................................................................... 28
Velocity Analysis and Residual Statics .................................................................... 32
Noise Suppression .................................................................................................... 37
Prestack Time Migration .......................................................................................... 48
Matching Pursuit Normal Moveout .......................................................................... 53
Prestack Structure Oriented Filtering ....................................................................... 55
vi
Chapter 4: Comparison to Legacy 1999 Vendor Volume .............................................. 57
Data Comparison ...................................................................................................... 57
Chapter 5: Interpretation, Inversion and AVAz ............................................................. 67
Structural and Stratigraphic Interpretation ............................................................... 67
Geometric Attributes ................................................................................................ 68
Post stack Acoustic Impedance Inversion ................................................................ 69
Amplitude Variations with Azimuth ........................................................................ 70
Geobody Extraction .................................................................................................. 70
Chapter 6: Conclusion .................................................................................................... 89
References ...................................................................................................................... 90
Appendix A: Understanding processing pitfalls with seismic modeling ....................... 93
vii
List of Tables
Table 1: Survey parameters ............................................................................................ 19
Table 2: Parameters for F-K filtering ............................................................................. 38
Table 3: Prestack Kirchhoff migration parameters ........................................................ 51
viii
List of Figures
Figure 1: County map of Texas. Young County is located in north central Texas (blue
arrow) ............................................................................................................................... 5
Figure 2: Satellite image of part of North America. Texas is outlined by the blue
polygon and the location of the seismic survey is denoted by the red star. ..................... 6
Figure 3: Cross section of Late Paleozoic trough through the Texas Craton and Ouachita
foldbelt. (Walper, 1982). .................................................................................................. 9
Figure 4: Map of Late Paleozoic’s structural elements in Texas and Oklahoma. (Walper,
1982). .............................................................................................................................. 10
Figure 5: Paleogeographic reconstruction of the Fort Worth Basin and Bend Arch (top)
during the Late Cambrian and (bottom) during the Devonian (Walper, 1982). ............. 11
Figure 6: Paleogeographic reconstruction of the Fort Worth Basin and Bend Arch (top)
during the Mississippian and (bottom) during the Pennsylvanian (Walper, 1982). ....... 12
Figure 7: Stratigraphic column of the Bend Arch. Stratigraphic formations of interest
for this study are the Early Pennsylvanian Caddo Limestone (Pollastro, 2007). ........... 13
Figure 8: Paleographic map of North America during the Early Pennsylvanian. The red
star represents the approximate location of the study area (Blakey, 2011). ................... 14
Figure 9 : (top) Structural provinces of Oklahoma and Texas. (bottom) Paleographical
map of the Fort Worth Basin (Thomas 2002). ............................................................... 15
Figure 10: East to West stratigraphic cross-section of the Fort Worth Basin and Bend
Arch (Bowker, 2007). The Caddo limestone lies within the Strawn group. ................. 16
Figure 11: Flow used for seismic processing and interpretation. ................................... 18
ix
Figure 12: Fold map of the Jean survey with source receiver array overlain. The
nominal fold is 17 with a maximum fold of 36. The survey has a rectilinear brick
pattern with a few irregularities due to surface obstacles. .............................................. 20
Figure 13: Raw seismic data sorted in the shot domain. There are 10 receiver lines with
the shot occurring between line 5 and line 6. Green arrows showcase reflectors and red
arrows highlight groundroll. ........................................................................................... 22
Figure 14: Raw seismic data sorted shot verses offset domain. The green line shows the
first break picks, used for refraction statics. The orange arrows highlight coherent
seismic noise, such as groundroll and headwaves. The blue arrow shows a seismic
reflection. ........................................................................................................................ 23
Figure 15: Representative gather with no sort order. The red arrow points to a traces
that needs to be killed. The green arrow points to a once bad traces that has been killed
by the processor. ............................................................................................................. 24
Figure 16: (From left to right) Map of the source receiver elevation, Layer 1 refraction
velocity, Layer 2 refraction velocities, picked from first break picks and long
wavelength refraction corrections. ................................................................................. 26
Figure 17: A representative NMO corrected gather (left) before and (right) after
refraction and elevation statics. Arrows indicate events that are better aligned resulting
in better frequency content. ............................................................................................ 27
Figure 18: (Left) Raw seismic data before correction for spherical divergence. (Right)
Seismic data after t-power gain correcting for spherical divergence. Deeper reflectors
are now more visible; however, groundroll is consequently enhanced. ......................... 30
x
Figure 19: Shot gather of the seismic data (top) before and (bottom) after spiking
deconvolution. The frequency spectrum is flattened increasing the contribution of both
low and high frequencies. ............................................................................................... 31
Figure 20: Semblance panel used for the second iteration of velocity analysis. (Left)
Semblance panel with picks used for NMO of the seismic data. (Right) NMO corrected
seismic gathers. ............................................................................................................... 34
Figure 21: Inline of the velocity field picked after the second iteration of velocity
analysis. The velocities range from 5500 ft/s to 16,000 ft/s. The velocity field is
laterally smooth such that prestack Kirchhoff time migration will provide accurate
results. ............................................................................................................................. 35
Figure 22: Brute stacked seismic data after the first round of velocity analysis. Orange
arrows indicate groundroll while green arrows indicate reflectors. (Middle) Brute stack
after second round of velocity analysis, deconvolution and F-K filtering. (Bottom) Brute
stack after velocity analysis performed on migrated gathers. ........................................ 36
Figure 23: Seismic data sorted shot verses channel. An (middle) Ormsby bandpass
filter shows most of the noise falls below 20 Hz. (right) By removing the bottom 20 Hz
most of the groundroll is suppressed. However, significant signal is also rejected
making an inversion and quantitative measures more difficult. ..................................... 39
Figure 24: (Left) A common receiver gather sorted by field station number versus
absolute offset and (Right) the corresponding F-K spectrum. Green arrows indicate
reflectors, yellow headwaves and orange groundroll. .................................................... 40
xi
Figure 25: Seismic data sorted by receiver versus absolute offset. It is necessary to have
the data sorted in this domain to perform F-K filtering. A strong presence of ground
roll is easily identifiable as are headwaves and reverberations. ..................................... 41
Figure 26: Seismic data sorted by field station number verses absolute offset with
polygons drawn around each mode of noise that was removed independently. Because
F-K filtering operates under the assumption of linear moveout and to avoid aliasing, the
noise to be removed needed to be flattened to be about wavenumber zero. .................. 42
Figure 27: (Top) Seismic data sorted by field station number versus absolute offset.
(Middle) Muted noise to be modeled in the F-K domain. The blue pick is defined as t =
𝒓𝑽𝑳𝑴𝑶 where r is the absolute offset and VLMO is the LMO velocity (blue). (Bottom)
LMO flattened noise. ...................................................................................................... 43
Figure 28: F-K spectrums of the segmented and bandpassed groundroll. ..................... 44
Figure 29: First mode of groudnroll modeled using F-K filtering. The frequency
content is between 0-25 Hz. The bottom figure is the same as above but with inversed
linear moveout applied. This is to be subtracted from the raw data to remove this mode
of groundroll. .................................................................................................................. 45
Figure 30: Seismic data sorted field station number verses absolute offset. The top
image is after the first mode has been subtracted from the original seismic. The bottom
figure is after all four F-K modeled modes of noise defined in Table 2 have been
subtracted. ....................................................................................................................... 46
Figure 31: (left) Seismic data sorted shot versus channel. Before any noise suppression
has been undergone. The (right) same gather after all F-K modeled modes of noise
have been subtracted. The orange arrows indicate zones where groundroll and coherent
xii
noise was overbearing the reflectors. The green arrows indicate zones where the
reflectors are now more visible. ..................................................................................... 47
Figure 32: Seismic data (left) after migration and (right) after migration with far offset
stretching muted. ............................................................................................................ 49
Figure 33: Stacked seismic data (A to A’) after migration and the muting of the
stretched far offsets ......................................................................................................... 50
Figure 34: Result after reverse normal moveout (RNMO). The data needed to be
RNMO’d for velocity analysis on the migrated data and matching pursuit normal
moveout. ......................................................................................................................... 51
Figure 35: Semblance panel used for the third iteration of velocity analysis after
migration. (Left) Semblance panel with picks used for NMO of the seismic data. (Right)
NMO corrected seismic gathers. The NMO corrected gather has greater resolution than
the NMO corrected gather during the second iteration of velocity analysis. ................. 52
Figure 36: (Left) Results after the second pass of migration. (Right) Results after
matching pursuit normal moveout. Notice the far offsets are better preserved leading to
better frequency content when stacked. .......................................................................... 54
Figure 37: (top) Rejected noise after prestack structural-oriented filtering using
Principle Components. (Bottom) Reflectors after Principal Component PSOF. After
PSOF incoherent noise has been removed from the migrated gather. ........................... 56
Figure 38: Vendor processed Jean seismic survey. The amplitudes in the shallow
section have not been properly balanced. Furthermore, the seismic data has been blued
up to 225 Hz causing a ringing around reflectors. The basement at 1100 ms is not easily
visible. ............................................................................................................................ 60
xiii
Figure 39: Reprocessed Jean survey. The amplitudes have been properly balanced in
the shallow section resulting in more identifiable reflectors. Furthermore, the frequency
spectrum has been whitened to 140 Hz resulting in a more geological result. The
basement at 1100 ms is easily identifiable. .................................................................... 61
Figure 40: Shallow horizon tracked to show the improved performance of autotracking
after reprocessing. (Left) Picks performed manually by the seismic interpreter. (Middle)
Autotracked horizon using the vendor processed Jean survey with a quality factor of
0.7. (Right) Autotracked horizon using the reprocessed Jean survey with a quality
factor of 0.7. ................................................................................................................... 62
Figure 41: Deeper horizon tracked to show the improved performance of autotracking
after reprocessing. (Left) Picks performed manually by the seismic interpreter. (Middle)
Autotracked horizon using the vendor processed Jean survey with a quality factor of
0.55. (Right) Autotracked horizon using the reprocessed Jean survey with a quality
factor of 0.55. ................................................................................................................. 63
Figure 42: Timeslice through seismic amplitude at t = 540 ms. (left) Vendor processed
seismic data. (right) Reprocessed seismic data. The reprocessed seismic data has
greater continuity due to the lack of high frequencies cross cutting the timeslices. ...... 64
Figure 43: Timeslice at t = 846 ms through a coherency volume. (Upper left) Vendor
processed coherence volume. (Upper right) Reprocessed data coherence volume. ...... 65
Figure 44: Timeslice through most negative curvature at t = 840 ms. (left) Vendor
processed most negative curvature. (right) Reprocessed most negative curvature. The
reprocessed seismic data has greater continuity due to the lack of high frequencies cross
cutting the timeslices. ..................................................................................................... 66
xiv
Figure 45: Line C-C’ showing key horizons. ................................................................. 71
Figure 46: Line from A - A’ showing key horizons. ...................................................... 72
Figure 47: (left) Time structure map of the Palo Pinto stratigraphic horizon tracked
through the reprocessed Jean survey. White arrows indicate areas of the horizon
contaminated by acquisition footprint. The causes of this footprint will be investigated
further in Appendix A. (right) Time structure map of the KMA stratigraphic horizon
tracked through the reprocessed Jean survey. ................................................................ 73
Figure 48: (left) Time structure map of the Upper Caddo stratigraphic horizon tracked
through the reprocessed Jean survey. The white arrow indicates areas of the horizon
containing draping over a carbonate reef. Reefs and draping due to deposition on top of
reefs are common oil and gas exploration targets (right). Time structure map of the
Basement horizon tracked through the reprocessed Jean survey. .................................. 74
Figure 49: (left) Time structure map of the Mississippian stratigraphic horizon tracked
through the reprocessed Jean survey. (Right) Time structure map of the Middle Caddo
stratigraphic horizon tracked through the reprocessed Jean survey. The white arrow
indicates a structurally low area in the horizon that is a potential channel cross cutting
the carbonate Strawn formation. ..................................................................................... 75
Figure 50: (left) Thickness map constructed by subtracting the Strawn Unconformity
stratigraphic horizon from the Upper Caddo stratigraphic horizon. The unconformity
thickens as we traverse to the northeast, coinciding with the trend of the Bend Arch.
(Right) Thickness map constructed from subtracting the Upper Caddo from the
Basement stratigraphic horizon. ..................................................................................... 76
xv
Figure 51: (left) Timeslice through most negative curvature at t = 420 ms. (Right)
Timelesice through most positive curvature at t = 420 ms. The curvature anomalies
indicated by the white arrows are interpreted to be acquisition footprint overprinting
geology. In Appendix A, I investigate the cause of footprint in the Jean survey using
seismic modeling. ........................................................................................................... 77
Figure 52: (left) Co-rendered K1 and K2 curvature extracted along the Middle Caddo
stratigraphic horizon. (Right) Co-rendered K1 and K2 curvature and energy ratio
similarity coherence attribute extracted along the Middle Caddo stratigraphic horizon.
The white arrow indicates a potential incised channel. .................................................. 78
Figure 53: Well tie and synthetic seismogram for Well #1 in the Jean survey. For the
best correlation between the seismic and the well logs, the statistical wavelet was
rotated to -170 degrees. With no density log available, one was estimated using the
Gardner’s equation (1974). ............................................................................................. 79
Figure 54: Well tie and synthetic seismogram for Well #2 in the Jean survey. For the
best correlation between the seismic and the well logs, the statistical wavelet was
rotated to -170 degrees. With no density log available, one was estimated using the
Gardner’s equation (1974). ............................................................................................. 80
Figure 55: Vertical section through the acoustic impedance inversion volume. The
section bisects Well #1 which has the acoustic impedance from the well logs projected
upon it. The impedance from the well to the impedance generated from the seismic
have a good correlation. ................................................................................................. 81
Figure 56: (left) Acoustic impedance extracted along the Upper Caddo horizon. The
Upper Caddo is a limestone resulting in high acoustic impedances. (Right) Acoustic
xvi
impedance co-rendered with the Upper Caddo time structure map. The majority of the
low impedance areas correlate to low structural relief zones. ........................................ 82
Figure 57: (left) Acoustic impedance extracted along the Middle Caddo horizon. The
Upper Caddo is a limestone resulting in high acoustic impedances. (Right) Acoustic
impedance co-rendered with the Middle Caddo time structure map. Low impedance
appears to be found along the channel feature seen in this horizon and the carbonate reef
in the southeast corner of the survey. ............................................................................. 83
Figure 58: (left) Acoustic impedance co-rendered with K1 and K2 curvature extracted
along the Upper Caddo horizon. (Right) Acoustic impedance co-rendered with K1 and
K2 curvature extracted along the Middle Caddo horizon............................................... 84
Figure 59: (left) Timeslice at t = 750 ms through the anisotropy azimuth volume.
(Right) Co-rendered anisotropy azimuth and density. .................................................. 85
Figure 60: Timeslice at t = 960 ms through the anisotropy azimuth volume. (Right) Co-
rendered anisotropy azimuth and density. ...................................................................... 86
Figure 61: Chair display from A to A’. Vertical section is seismic amplitude and
timeslice is K2 curvature. The arrow indicates a potential channel in the Middle Caddo
that is extracted using geobodies. ................................................................................... 87
Figure 62: Chair display showing geobody extraction of channel after Wallet (2014).
The vertical sections are seismic amplitude and the horizontal slice is the Middle Caddo
time structure map. The channel geobody extracted by thresholding K2 curvature is
shown in red. .................................................................................................................. 88
xvii
Abstract
The purpose of this study is to use state-of-the-art 3D seismic processing and
interpretation tools to bring a modern perspective to legacy data. Having access to a
relatively small, 10 mi.2, 3D seismic survey acquired in the early 1990’s and a number
of wells within the adjacent area, I reprocess the raw shot gathers with a more modern
seismic processing workflow. With careful attention to velocity analysis, techniques
for preserving the frequency content while mitigating noise, a newly developed prestack
Kirchhoff migration algorithm coupled with prestack structural-oriented filtering and an
abundance of time, I am able to improve the seismic image quality thus boosting the
interpretability of the data. The primary reflectors of interest are representative of
Pennsylvanian fluvial-deltaic sandstone and conglomerate and the Strawn formation
carbonate reefs. Given the relatively shallow depth of these reflectors and the expected
compaction, poststack inversion, AVAz analysis and direct hydrocarbon indicators
could prove to be valuable tools for the interpreter. Geometric attributes are used to
assist in the development of a stratigraphic and structural framework. Geobody
extraction is utilized for 3D mapping and visualization of fluvial-deltaic channels,
potential bypassed pay. I use poststack inversion to interpolate porosity and fluid
saturation measured at over 50 wells within the survey. With the renewed interest in
shallower targets in the Fort Worth Basin improved imaging techniques and the use of
modern interpretation tools will have great importance.
1
Chapter 1: Introduction
The Bend Arch-Fort Worth Basin Province is located in southwestern Oklahoma
and north-central Texas (Figure 1). Although a gray oil field, much light has been shed
on it by hydrocarbon exploration companies in recent years, with the resurgence of
drilling activity attributed to horizontal drilling and hydraulic fracturing. The basin has
a number of potential stacked plays ranging from the Ordovician Ellenburger to the late
Pennsylvanian Cisco group. With a seismic survey atop of the Bend Arch, in my thesis,
I reprocess this legacy data with the intent of increasing signal to noise ratio and also
preserving higher frequencies to allow for better imaging capabilities. Because of more
robust processing workflows and advancements in imaging algorithms, reprocessing of
legacy seismic data has proven an invaluable technique for re-investigating hydrocarbon
plays (Aisenberg, 2013). Upon completion of the reprocessing I compute seismic
attribute and poststack inversion volumes in order to identify and better map features
within the hydrocarbon bearing Strawn formation. Considerable work has been
performed over the attribute expression of the Barnett Shale and Ellenburger dolomite
in the Bend-Arch Fort Worth basin province identifying prospective locations and
geohazard recognition (Fernandez, 2013). However, little has been reported on the
Strawn and Cisco groups. Legacy data and reprocessed data attribute analysis and
interpretations are compared to ensure that significant improvements have been made.
This thesis begins with Chapter 2 where I provide an overview of the geology in
the area surrounding the Jean seismic survey. I begin by providing various maps and
satellite images allowing the reader to determine the survey’s approximate location.
Next, I move on to the regional geology and local geology, specific to the producing
2
intervals of the survey. Chapter 3 provides a general overview of the processing steps
and parameters used through 3D prestack Kirchhoff time migration of the seismic
survey. Chapter 4 includes a comparison of my results to the 3D seismic data processed
by a commercial vendor. Chapter 5 shows the results of a poststack inversion and
AVAz. I also provide an interpretation of the structural and stratigraphic framework
using geometric attributes. I conclude in, Chapter 6 summarizing the critical points
found within each chapter.
3
Objectives
The primary objective of this study is to reprocess a legacy 3D seismic data set
and bring it up to modern 3D seismic data quality. Upon completion of reprocessing,
seismic attributes, inversion and AVAz are used to map and identify sweet spots in the
Pennsylvanian units in the Jean survey of Young County, Texas.
- Processing Objectives:
Human Intensive Objectives
o Careful attention to trace editing
o Proper static corrections
o Detailed removal of linear noise
o Diligent velocity analysis
Computationally Intensive Objectives
o Prestack Kirchhoff time migration
o Prestack structure oriented filtering
o Matching pursuit normal moveout
- Interpretation Objectives:
Map key reflectors corresponding to producing lithological units
Use geometric attributes to understand morphology of structure and stratigraphy
o Dip and azimuth
o Coherency
o Curvature
Compute a poststack inversion volume to locate porosity and fluid saturation
Use amplitude variations with azimuth (AVAz) to identify anisotropic zones
Extract geobodies for 3D visualization of channels
4
Location
The data license used for this project was provided by Mike Burnett of
TameCAT llc., an independent oil and gas company in Norman, OK. The data
relinquished for this project included 2- 8 mm 5GB tapes containing poststack seismic
data each with different processing procedures applied by a commercial vendor, 2- 8
mm 5GB tapes including raw prestack seismic data to be reprocessed, 4 floppy disks
containing observer and field notes, 5 LAS logs and 60+ paper well logs from multiple
wells drilled within and adjacent to the seismic survey. The 5 LAS files contained p-
wave sonic logs while the paper logs, being older wells, only contained SP, gamma ray
and resistivity. There are no shear sonic logs in the area of the seismic survey.
The seismic survey, approximately 10 sq. miles in size, was acquired in Young
County, Texas. Figure 1 is a county map of Texas, Young County can be found in
north central Texas. Figure 2 is the location of the seismic survey on a satellite image.
5
Figure 1: County map of Texas. Young County is located in north central Texas
(blue arrow)
6
Figure 2: Satellite image of part of North America. Texas is outlined by the blue
polygon and the location of the seismic survey is denoted by the red star.
7
Chapter 2: Geological Background
Regional Geology
The seismic survey in this study lies on the eastern edge of Young County,
Texas. It is in a transition zone between the Fort Worth Basin and the Bend Arch,
located in North Central Texas and Southern Oklahoma. The surrounding regional
geological features around the Fort Worth Basin and Bend Arch consist of the Muenster
Arch, Ouachita Structural Belt and Llano Uplift . Figure 4 shows the structural
elements in the area around the Fort Worth Basin during the Late Paleozoic. These
structures were created during the collision of North Africa and North America during
the early to middle Paleozoic (Flippin, 1982).
The Fort Worth Basin is an asymmetrical peripheral wedge-shaped foreland
basin (Dickenson, 1976) and is approximately 15,000 mi2 in area (Montgomery, 2005).
The basin was created due to the advancing Ouachita structural belt during the
formation of Pangaea in the Late Paleozoic, when North Africa collided with North
America. In Figure 3 we see a cross-section through the Fort Worth Basin area during
the Late Paleozoic. In Figures 5 and 6 we see the evolution of the Fort Worth Basin and
surrounding structural elements as North Africa approaches and collides with North
America.
The Ouachita Structural Belt formed during the Late Paleozoic; however, it is
currently only found in the subsurface with the exception of southeastern Oklahoma,
northwest Arkansas and west Texas. This orogenic system is believed to be formed
similarly to the Appalachian Mountains in western North America, during the creation
of Pangaea.
8
The Bend Arch is a northward plunging broad anticlinal structure extending to
the north from the Llano Uplift and terminating in the regional syncline south of the
Red River Uplift (Evanoff, 1976). The Bend Arch was a relatively stable area during
the formation of the Fort Worth Basin to the east and the Permian Basin to the south
west.
Local Geology
The stratigraphy of the area around the survey entails rock units from Cambrian
to Permian age. Figure 7 shows the generalized stratigraphy for the Fort Worth Basin
and Bend Arch. The primary stratigraphic units of petroleum exploration are the
Desmoinesian Caddo limestone. In Figure 8, the approximate location of the survey is
marked by a red star on a Blakey reconstruction of the time of deposition of the Caddo
limestone. Reefing is common in the Caddo limestone along a North-South trend line
from Archer to Brown County. The Caddo limestone ranges from 100 ft to 800 ft and
can be correlated as far as the Permian Basin. Oil and gas are not only found in reefs
but also formations above the reefs due to the draping effect capturing upward
migrating hydrocarbons (Evanoff, 1976).
Sessions of upwarp and downwarp during the Ouachita Orogeny lead to several
erosional unconformities in the Bend Arch bedding. The most notable unconformity,
found at the top of the Ordovician Ellenburger, has Mississippian beds deposited on top
of it (Walper, 1982). Figure 9 shows the paleogeographic features of the Late
Paleozoic. Figure 10 shows a representative east to west geological cross section across
the Fort Worth Basin and Bend Arch; the red star represents the approximate location of
the seismic survey.
9
Figure 3: Cross section of Late Paleozoic trough through the Texas Craton and
Ouachita foldbelt. (Walper, 1982).
10
Figure 4: Map of Late Paleozoic’s structural elements in Texas and Oklahoma.
(Walper, 1982).
11
Figure 5: Paleogeographic reconstruction of the Fort Worth Basin and Bend Arch
(top) during the Late Cambrian and (bottom) during the Devonian (Walper,
1982).
12
Figure 6: Paleogeographic reconstruction of the Fort Worth Basin and Bend Arch
(top) during the Mississippian and (bottom) during the Pennsylvanian (Walper,
1982).
13
Figure 7: Stratigraphic column of the Bend Arch. Stratigraphic formations of
interest for this study are the Early Pennsylvanian Caddo Limestone
(Pollastro, 2007).
14
Figure 8: Paleographic map of North America during the Early Pennsylvanian. The
red star represents the approximate location of the study area (Blakey, 2011).
15
Figure 9 : (top) Structural provinces of Oklahoma and Texas. (bottom)
Paleographical map of the Fort Worth Basin (Thomas 2002).
16
Fig
ure
10
: E
ast
to W
est
stra
tigra
phic
cro
ss-s
ecti
on o
f th
e F
ort
Wort
h B
asin
and B
end A
rch (
Bo
wker
, 2007).
The
Cad
do l
imes
tone
lies
wit
hin
the
Str
awn g
roup.
17
Chapter 3: 3D Prestack Seismic Processing
The goal in reprocessing the Jean survey, provided by Mike Burnett of
TameCAT LLC, was to use modern processing workflows to improve the data fidelity
beyond that provided by the commercial vendor in 1999. The original processing
focused on improving seismic resolution using well based data bluing factors. My goal
is to apply an amplitude friendly workflow that preconditions the data for more
quantitative interpretations, i.e. seismic inversion, than the commercial vendor.
Such improvement was accomplished with more time dedicated to statics,
velocity analysis, frequency preservation and noise removal and also having a more
geologically centered processing workflow in mind. Because of the amount of time
spent reprocessing this data set and the amount learned in doing so, this chapter of my
thesis is very dear (both affectionately and costly) to me and forms the bulk of my
effort.
The reprocessing workflow can be outlined as follows:
-Loading seismic data,
-Defining geometry and trace editing,
-Performing refraction and elevation statics,
-Suppressing linear noise,
-Picking velocities and calculating residual statics,
-Applying prestack time migration,
-Applying prestack structure oriented filtering and,
-Stacking the prestack time migrated gathers.
Figure 11 summarizes the above steps in the format of a workflow.
18
Fig
ure
11
: F
low
use
d f
or
seis
mic
pro
cess
ing a
nd i
nte
rpre
tati
on
.
19
Data Acquisition
The Jean survey was acquired and originally processed in 1999 by Western
Geophysical. Table 1 shows a detailed outline of the survey parameters. The source
was 5 kg of dynamite buried at 80 ft. The sources and receiver line spacing was 2200 ft
and the source and receiver interval was 220 ft. The acquisition follows a standard
rectilinear seismic survey array with the exception of small deviations due to surface
obstacles. Figure 12 shows the source and receiver deployment with shot points in red
and receiver stations in blue and green. The shot indicated by the red star is recorded by
the green receivers, defining the live receiver patch. For any shot the maximum
number of receiver lines turned on was 10. The signal was recorder for 3001 ms with a
1 ms sampling time leading to a Nyquist frequency of 500 Hz. The targets of interest
are relatively shallow, with the producing formations around 700 to 800 ms and
basement reflection at 1100 ms.
Survey Name: Jean 3D
Receiver Spacing: 220 ft
Receiver Line Spacing 1320 ft
Source Spacing: 220 ft
Source Line Spacing: 1320 ft
Source Type: Dynamite buried 80 ft.
Trace Length: 3000 ms
Sample Rate: 1 ms
Bin Size: 110 X 110 ft
Azimuth: 1 degree East of North
Number of Total Traces: 390,403
Nominal Fold: 17
Table 1: Survey parameters
20
Figure 12: Fold map of the Jean survey with source receiver array overlain. The
nominal fold is 17 with a maximum fold of 36. The survey has a rectilinear
brick pattern with a few irregularities due to surface obstacles.
21
Data Loading and Geometry
The first step to processing is loading the seismic data properly and checking the
geometry. Geometry is quality controlled by picking first breaks (later be used for
refraction statics) and defining the binning, azimuth and inline/crossline spacing. I
began with 2- 8mm 5 GB tapes containing raw prestack seismic data. The azimuth of
the survey was 1 degree east of north and the binning used was the natural binning, of
110 ft. After binning, the fold, azimuth of traces and offset can be computed (Figure
12). The nominal fold of the Jean survey is 17 with the maximum of 36. This relatively
low fold made improving the signal to noise ratio a priority if more quantitative
measurements were to be attempted. The seismic survey has relatively wide azimuth,
making AVAz a likely quantitative measurement. The offsets of the survey range from
0 to 14,000 ft. Figure 13 is representative of a typical patch in the shot domain. In the
patch, ten receiver lines can be seen with split spread, meaning the shot occurred near
the middle of the receiver line. The inset schematic in Figure 13 represents the
approximate location of the shot with respect to the receiver lines. Seismic reflectors
are shown by the green arrows where coherent noises, such as groundroll, are shown by
red arrows. With no noise removal or amplitude balancing, the reflectors are very
difficult to identify at this point in the processing flow.
Figure 14 is a representative shot sorted by offset with the first breaks in light
green. These first breaks are used for refraction statics, an essential procedure for
aligning reflectors and improving the frequency content after stacking the seismic data.
22
Fig
ure
13
: R
aw s
eism
ic d
ata
sort
ed i
n t
he
shot
do
mai
n.
Ther
e ar
e 10 r
ecei
ver
lin
es w
ith
th
e sh
ot
occ
urr
ing
bet
wee
n l
ine
5 a
nd l
ine
6. G
reen
arr
ow
s sh
ow
case
ref
lect
ors
and r
ed a
rro
ws
hig
hli
ght
gro
undro
ll.
23
Fig
ure
14
: R
aw s
eism
ic d
ata
sort
ed s
hot
ver
ses
off
set
dom
ain. T
he
gre
en l
ine
show
s th
e fi
rst
bre
ak p
icks,
use
d f
or
refr
acti
on s
tati
cs. T
he
ora
nge
arro
ws
hig
hli
ght
coher
ent
seis
mic
nois
e, s
uch
as
gro
undro
ll a
nd
hea
dw
aves
. T
he
blu
e ar
row
show
s a
seis
mic
ref
lect
ion.
24
Trace Editing
Trace editing was a very time consuming but crucial part of the processing for
this data set. With a nominal fold of 17, suppression of coherent and incoherent noise is
crucial. I could not rely on the power of the stack to attenuate noise from bad coupling
or broken geophones. Therefore, killing such bad traces before proceeding to other
processing steps is critical. To do so I had to manually inspect all trace in three
different sort domains, shot, receiver and not sort order. Figure 15 shows a
representative gather with no sort order, highlighting the typical removal of ‘bad’ traces.
Figure 15: Representative gather with no sort order. The red arrow points to a traces
that needs to be killed. The green arrow points to a once bad traces that has been killed
by the processor.
25
Elevation and Refraction Statics
Both elevation and refraction statics were computed and applied as part of the
processing procedure. The goal of elevation statics is to project the seismic data onto a
fixed datum, typically below the lowest part of the weathering zone. The aim of
refraction statics is to determine and then correct for the velocity and thickness of the
shallowest two layers in the substrate and to correct for the weathering zone (Russell
and Russell, 1989). Figure 16 shows a map of the source and receiver elevation the
elevation static corrections and the velocity for the first and second layer used in the
weathering zone correction. Figure 17 shows a gather sorted by common midpoint.
From this figure we can see the importance of proper statics corrections. Green arrows
indicate events that are better aligned after statics and consequently have greater
frequency content after stacking (Stein et al., 2009). Elevation statics and refraction
statics are important to correct for large scale undulations on a seismic survey array,
whereas residual statics are important for trace by trace corrections.
26
Fig
ure
16
: (F
rom
lef
t to
rig
ht)
Map
of
the
sourc
e re
ceiv
er e
levat
ion,
Layer
1 r
efra
ctio
n v
eloci
ty,
Layer
2
refr
acti
on v
eloci
ties
, pic
ked
fro
m f
irst
bre
ak p
icks
and l
ong w
avel
ength
ref
ract
ion
co
rrec
tio
ns.
27
Figure 17: A representative NMO corrected gather (left) before and (right) after
refraction and elevation statics. Arrows indicate events that are better aligned resulting
in better frequency content.
28
Deconvolution
The mathematical basis for deconvolution is well documented and routinely
used by most geophysicists (Cary, 2001). In my application, I use deconvolution to
compress the source wavelet to more accurately approximate the reflectivity spike
series. Shallow multiples form the weathering zone form part of the effective wavelet
and are also suppressed. The spectrum is flattened such that low frequency groundroll
is somewhat attenuated (Sheriff, 2004). Deconvolution can be time consuming, with
multiple parameters to test, including window length and area, operator length and
alternative pre-whitening factors.
Before deconvolution could be applied the data had to be scaled for geometric or
spherical divergence energy losses with depth. For the Jean survey a t-powered
exponential gain of 1.8 worked well. From my experimentation I found it important to
apply a t-gain before deconvolution before an additional surface consistent gain, and not
the other way around. Figure 18 shows the data before and after t-gain. Note that
reflectors in the deeper section that were originally not seen are now visible. However,
the amplitudes of groundroll are also more pronounced after correcting for spherical
divergence.
Upon completion of applying geometry, trace editing, refraction and elevation
statics and amplitude scaling for geometric attenuation I performed an initial
deconvolution. Because the data were acquired with dynamite as opposed to Vibroseis,
no phase conversion or matching needed to be performed because the data were already
minimum phase. This initial deconvolution involved a channel deconvolution as
opposed to a surface consistent deconvolution, the reason for this being that the
29
overbearing presence of ground roll had too great an affect while shaping the
deconvolution operator. Therefore, surface consistent deconvolution was not applied to
the data. The initial channel deconvolution was a spiking deconvolution, had an
operator length of 120 ms, a pre-whitening factor of 5% and was applied to the entire
trace. Figure 19 show the results before and after the initial channel deconvolution.
Take notice to the increase in the frequency content (Figure 19 inset image). The
resulting frequency spectrum is also flatter; therefore, spectral whitening or bluing was
not performed.
30
0
1500
Tim
e (m
s)
350
1100
Before Amplitude Gain After Amplitude Gain
Figure 18: (Left) Raw seismic data before correction for spherical divergence. (Right)
Seismic data after t-power gain correcting for spherical divergence. Deeper reflectors
are now more visible; however, groundroll is consequently enhanced.
31
Figure 19: Shot gather of the seismic data (top) before and (bottom) after spiking
deconvolution. The frequency spectrum is flattened increasing the contribution of
both low and high frequencies.
32
Velocity Analysis and Residual Statics
Velocity analysis is the most human intensive part of the processing workflow.
Figure 20 shows a representative semblance panel and NMO corrected CMP gather
from running the Interactive Velocity Analysis flow. Before I began velocity analysis I
first ran some initial conditioning. Since semblance represents reflectors by hyperbola,
velocity analysis is applied after gaining the data, spiking deconvolution, elevation
statics, refraction statics, and bandpass filtering. The bandpass applied for velocity
analysis was an Ormsby filter with corner frequencies of 17-22-80-95 Hz. I determined
that with this Ormsby I could eliminate large portions of noise and only keep reflectors
for velocity analysis. Subsequent processing will use a broader spectrum of the data.
Velocity analysis is an iterative process, having to be applied on finer and finer
grids until the residual statics converge, ideally to 0 ms. Residual static shifts that are
incorrect decrease the power of the CMP stack (Ronen and Claerbout, 1985). Having a
decreased CMP stack power is synonymous with lower resolution and destruction of
high frequency signals. The order for velocity analysis is as follows:
Compute semblance on a user defined grid,
Pick semblance panel that flatten seismic primary reflectors,
NMO correct seismic data with velocity field from picked semblance panel,
Compute residual statics, and
Stack NMO corrected gathers with the residual statics applied.
For my first pass of velocity analysis I used a 20x20 inline-crossline grid. I also
used a 10x10 super-gather, which essentially boosts the number of traces in an
33
ensemble while picking semblances by combining 100 CMP’s together. Supergathers
improve the signal to noise ratio of the semblance calculation but laterally smear
geology. Using this velocity field I flattened my seismic gathers using an NMO
correction and computed residual statics. After the first pass of velocity analysis the
residuals had converged to an RMS value of 3 ms, indicating that I needed to perform
additional iterations of velocity analysis. On the second pass of velocity analysis I used
a 10x10 inline-crossline grid. Upon completion of residual statics after this pass the
residuals had converged to less than an RMS value of 1 ms, which is acceptable for my
sampling increment of 1 ms. Figure 21 shows an inline of the velocity field after the
second iteration of velocity analysis; note that the velocity field is very smooth laterally.
Figure 22 shows a NMO corrected and stacked inline after the first pass of velocity
analysis. This stack was performed before deconvolution and noise suppression. The
effects of groundroll can be seen by the orange arrows. Figure 22 (middle) shows a
stack after deconvolution, F-K filtering and the second iteration of velocity analysis.
We can see greater frequency content and continuity of reflectors.
Upon completion of the prestack Kirchhoff migration, velocity analysis was run
again on the reverse normal moveout migrated gathers. Figure 22 (Bottom) shows a
stack after velocity analysis on migrated gathers. This is discussed later in more detail
under the Prestack Time Migration section of this thesis.
34
Figure 20: Semblance panel used for the second iteration of velocity
analysis. (Left) Semblance panel with picks used for NMO of the seismic
data. (Right) NMO corrected seismic gathers.
200
400
1000
Tim
e (m
s)
800
Velocity Offset
35
2500
2000
3000
18000
12000
6000 Velocity (ft/s)
1000
500
0 Crossline 0 90 30 60
Tim
e (m
s)
Figure 21: Inline of the velocity field picked after the second iteration of velocity
analysis. The velocities range from 5500 ft/s to 16,000 ft/s. The velocity field is
laterally smooth such that prestack Kirchhoff time migration will provide accurate
results.
36
Figure 22: Brute stacked seismic data after the first round of velocity analysis.
Orange arrows indicate groundroll while green arrows indicate reflectors. (Middle)
Brute stack after second round of velocity analysis, deconvolution and F-K filtering.
(Bottom) Brute stack after velocity analysis performed on migrated gathers.
Amplitude
37
Noise Suppression
Figure 23 shows the data and noise under a suite of frequency ranges. It is
evident from Figure 23 (right) that most of the Ground roll is below 20 Hz; however,
using a bandpass filter to reject this noise removal would significantly degrade
subsequent post stack inversion. Therefore, other options for linear noise removal were
examined. The first methodology was the AASPI coherent noise suppression purposed
by VERMA (2014). However, from my findings and those of Ha (2014) on a Texas
panhandle data volume, Verma’s AASPI coherent noise suppression workflow works
best for broad band noise. In my case, with the ground roll below 20 Hz, I decided to
implement the F-K filter for noise suppression. F-K filtering separates various types of
noise from signal based on their different moveout (Alvarez de la Hoz, 1995). Figure
24 shows the F-K spectrum of my data. Before an F-K transform is performed it is
quintessential that the lateral and vertical amplitudes of the seismic data be roughly
balanced. In my case I used exponential t-powered gain and surface consistent scaling.
F-K filtering can often be a very tedious process having to be applied for each type of
noise individually and for both the source and receiver domains (Vermeer, 2012).
The general steps for F-K noise removal include:
-Sorting the data in source versus absolute offset (Figure 25),
-Determining the spatial range of the noise and isolating it with a mute,
-Finding the frequency range of the noise within the muted area (Figure 26),
-Flattening the gathers with respect to moveout to avoid aliased noise and signal,
-Modeling the isolated area in the F-K domain (Figure 28),
38
-Unflattening F-K modeled noise (Figure 29),
-Subtracting the F-K modeled noise from the original data (Figure 30),
-Looping for all modes of noise (groundroll, headwaves and reverberations) and
-Repeating the flow for the data sorted by receiver verse absolute offset.
Table 2 shows the parameters used for F-K modeling the different types of noise:
Polygon Color Frequency Linear Moveout
Blue 0 - 40 Hz 10,480 ft/s
Yellow 0 - 22 Hz 8,755 ft/s
Green 0 - 25 Hz 6075 ft/s
Red 0 - 25 Hz No LMO Used Table 2: Parameters for F-K filtering
Figure 31 shows the seismic data sorted by shot (left) before F-K filtering and
(right) after F-K filtering. We can see that most of the groundroll is suppressed and
reflectors are more identifiable.
39
Fig
ure
23
: S
eism
ic d
ata
sort
ed s
hot
ver
ses
chan
nel
. A
n (
mid
dle
) O
rmsb
y b
andpas
s fi
lter
show
s m
ost
of
the
nois
e fa
lls
bel
ow
20 H
z. (r
ight)
By r
emovin
g t
he
bott
om
20 H
z m
ost
of
the
gro
undro
ll i
s su
ppre
ssed
. H
ow
ever
, si
gnif
ican
t si
gn
al
is a
lso r
ejec
ted m
akin
g a
n i
nver
sion a
nd q
uan
tita
tive
mea
sure
s m
ore
dif
ficu
lt.
Time (ms)
50
0
15
00
Sig
nal
Nois
e
Nois
e
40
Fig
ure
24
: (L
eft)
A c
om
mon r
ecei
ver
gat
her
sort
ed b
y f
ield
sta
tion n
um
ber
ver
sus
abso
lute
off
set
and (
Rig
ht)
the
corr
espondin
g F
-K s
pec
trum
. G
reen
arr
ow
s in
dic
ate
refl
ecto
rs,
yel
low
hea
dw
aves
and o
ran
ge
gro
undro
ll.
41
Fig
ure
25
: S
eism
ic d
ata
sort
ed b
y r
ecei
ver
ver
sus
abso
lute
off
set.
It
is
nec
essa
ry t
o h
ave
the
dat
a so
rted
in t
his
dom
ain
to p
erfo
rm F
-K f
ilte
rin
g.
A s
trong p
rese
nce
of
gro
und r
oll
is
easi
ly i
den
tifi
able
as
are
hea
dw
aves
an
d r
ever
ber
atio
ns.
Nois
e
Nois
e
42
Fig
ure
26
: S
eism
ic d
ata
sort
ed b
y f
ield
sta
tion n
um
ber
ver
ses
abso
lute
off
set
wit
h p
oly
go
ns
dra
wn
aro
un
d e
ach
mod
e of
nois
e th
at w
as r
emo
ved
ind
epen
den
tly. B
ecau
se F
-K f
ilte
ring o
per
ates
und
er t
he
assu
mpti
on o
f li
nea
r
mov
eout
and t
o a
vo
id a
lias
ing, th
e nois
e to
be
rem
oved
nee
ded
to b
e fl
atte
ned
to b
e ab
out
wav
enum
ber
zer
o.
43
Figure 27: (Top) Seismic data sorted by field station number versus absolute offset.
(Middle) Muted noise to be modeled in the F-K domain. The blue pick is defined as
t = 𝒓
𝑽𝑳𝑴𝑶 where r is the absolute offset and VLMO is the LMO velocity (blue).
(Bottom) LMO flattened noise.
44
Fig
ure
28
: F
-K s
pec
trum
s of
the
segm
ente
d a
nd b
and
pas
sed
gro
un
dro
ll.
45
Fig
ure
29
: F
irst
mode
of
gro
udn
roll
model
ed u
sing F
-K f
ilte
ring. T
he
freq
uen
cy c
onte
nt
is b
etw
een
0-2
5 H
z. T
he
bott
om
fig
ure
is
the
sam
e as
above
but
wit
h i
nver
sed l
inea
r m
oveo
ut
appli
ed. T
his
is
to b
e su
btr
acte
d f
rom
the
raw
dat
a to
rem
ove
this
mode
of
gro
undro
ll.
46
Fig
ure
30
: S
eism
ic d
ata
sort
ed f
ield
sta
tion n
um
ber
ver
ses
abso
lute
off
set.
T
he
top
im
age
is a
fter
the
firs
t m
ode
has
bee
n s
ubtr
acte
d f
rom
the
ori
gin
al s
eism
ic. T
he
bott
om
fig
ure
is
afte
r al
l fo
ur
F-K
model
ed m
odes
of
nois
e
def
ined
in T
able
2 h
ave
bee
n s
ubtr
acte
d.
Am
pli
tud
e
47
Figure 31: (left) Seismic data sorted shot versus channel. Before any noise
suppression has been undergone. The (right) same gather after all F-K modeled
modes of noise have been subtracted. The orange arrows indicate zones where
groundroll and coherent noise was overbearing the reflectors. The green arrows
indicate zones where the reflectors are now more visible.
48
Prestack Time Migration
Land seismic surveys often have irregularities in their acquisition. Because of this fact
and the relatively smooth velocity field, I performed a prestack Kirchhoff migration
described by Perez and Marfurt (2008). Seismic migration moves dipping events to
their correct locations and focuses the energy from diffractions. A good migration will
provide a higher resolution image and move out of plane reflections to their proper
location.
After suppression of coherent noise described in the previous section I ran my
first pass of migration. This migration used the velocity field that was created after the
second round of velocity analysis, on a 10x10 inline-crossline grid. Figure 32 shows
the results after migration (left) without a mute and (right) with a mute to remove any
reflectors that were stretched at far offsets. Figure 33 shows a stack of the migrated
data after muting the stretched offsets.
The next step was to reverse normal moveout (RNMO) the migrated data using
the velocity field used to migrate the seismic data. Upon completion of RNMO I
repicked the velocity field on the migrated data following the Deregowski loop (1990).
Figure 34 shows the reverse normal moveout gathers. Repicking velocities on the
migrated data led to greater clarity of events and better resolution of the now flattened
events. Figure 35 shows a semblance panel and normal moveout corrected seismic data
on the migrated data. Table 3 contains the parameters for the prestack Kirchhoff time
migration.
49
Am
pli
tud
e
Fig
ure
32
: S
eism
ic d
ata
(lef
t) a
fter
mig
rati
on a
nd (
right)
aft
er m
igra
tion w
ith f
ar o
ffse
t st
retc
hin
g m
ute
d.
50
Figure 33: Stacked seismic data (A to A’) after migration and the muting of
the stretched far offsets
Amplitude
51
Min.
Frequency
Max.
Frequency
No.
Offset
Bins
Max.
Offset
No.
Azimuths
Migration
Aperture
0 Hz 140 Hz 60 14000 ft. 8 7000 ft.
Table 3: Prestack Kirchhoff migration parameters
Figure 34: Result after reverse normal moveout (RNMO). The data needed to be
RNMO’d for velocity analysis on the migrated data and matching pursuit normal
moveout.
Offset 0 60 45 15
Tim
e (m
s)
0
1500
500
1000
Amplitude
52
Figure 35: Semblance panel used for the third iteration of velocity analysis
after migration. (Left) Semblance panel with picks used for NMO of the
seismic data. (Right) NMO corrected seismic gathers. The NMO corrected
gather has greater resolution than the NMO corrected gather during the
second iteration of velocity analysis.
53
Matching Pursuit Normal Moveout
Using a new velocity field picked on the migrated data we recomputed a
prestack Kirchhoff migration, using the same input seismic data as the first time. In
order to preserve the fidelity of far offset I performed matching pursuit normal moveout
(MPNMO). The purpose of MPNMO is to minimize the stretching and subsequent
decrease of high frequencies that occurs at far offsets during migration (Zhang, 2014).
Figure 36 (left) shows the output after MPNMOing the data. We can see when
compared to the results after migration (Figure 36 (right)) that the far offsets, especially
in the shallow section have been preserved. The result contains better frequency
resolution and gives the interpreter the ability to perform more quantitative measures,
such as AVO and inversion. Even after MPNMO some of the farthest offsets are still
not preserved. Because of this we still have to apply a mute to the far offset before we
can stack the seismic data.
54
Fig
ure
36
: (L
eft)
Res
ult
s af
ter
the
seco
nd p
ass
of
mig
rati
on. (
Rig
ht)
Res
ult
s af
ter
mat
chin
g p
urs
uit
norm
al m
oveo
ut.
N
oti
ce t
he
far
off
sets
are
bet
ter
pre
serv
ed l
eadin
g t
o b
ette
r fr
equ
ency c
onte
nt
wh
en s
tack
ed.
Am
pli
tud
e
55
Prestack Structure Oriented Filtering
The purpose of prestack structural-oriented filtering (PSOF) is to remove
incoherent noise that may be cross cutting reflectors (Davogustto, 2011). To do so we
must first stack the seismic data and compute dip, azimuth and similarity. I chose to
compute energy ratio similarity because the results are dependent on both the amplitude
and waveform of the traces in the analysis window. The next step is to use the prestack
structural oriented filter. The inputs for this program are the prestack migrated volume,
the inline and crossline dip and the similarity attribute. Within the AASPI software
there are four methods for computing prestack structural oriented filtering. After
running all of them I decided the Principle-component-filter gave the best results.
Principle-component filtering works by generating a covariance matrix based on
coherency values within a structural dip oriented analysis window. From this
covariance matrix the principle-components can be determined. With the first
principle-component being representative of most coherent part of the data, the seismic
reflectors, and the later component being representative of the least coherent parts of the
data, the noise. The first principle component is then kept while the other components
are rejected, resulting in the suppression of cross cutting incoherent noise.
Figure 37 (top) shows the rejected noise and (bottom) the results of prestack
structure oriented filtering.
56
Tim
e (m
s)
0
1500
500
1000
Offset 0 60 45 15
Offset 0 60 45 15
Tim
e (m
s)
0
1500
500
1000
Figure 37: (top) Rejected noise after prestack structural-oriented filtering using
Principle Components. (Bottom) Reflectors after Principal Component PSOF.
After PSOF incoherent noise has been removed from the migrated gather.
Amplitude
57
Chapter 4: Comparison to Legacy 1999 Vendor Volume
In order to ensure that significant improvements have been made by
reprocessing the Jean survey a seismic attribute comparison was made. Seismic
processing has not only made improvements in seismic image quality and attribute
image quality but also improved utility in common interpretation functions, such as
horizon autotracking.
The major advantage I possessed while reprocessing the Jean survey was an
understanding of the geology from the previous vendor processed seismic data.
Knowing the time of reflectors of interest allowed me to concentrate velocity analysis
and migration to that particular area, thus, giving an advantage over the commercial
vendor. Other advantages over the commercial vendor are recent technological
advancements, such as prestack structure-oriented filtering to remove incoherent noise
and matching pursuit non-stretch normal moveout to preserve far offsets. Furthermore,
understanding how previous processing steps hindered the subsequent attribute
interpretation, I attempted to bypass such steps to increase the utility of seismic
attributes and quantitative interpretation measures.
Data Comparison
The first step in comparing the vendor processed to the reprocessed is to check
the frequency spectrum. Figures 38 and 39 show a representative vertical slice through
the vendor processed and reprocessed data volumes, respectively. From Figure 38 it is
evident that the commercial vendor blued the seismic data up to 225 Hz during
processing, resulting in a ringing around the reflectors that does not appear geological.
The amplitudes have not been vertically balanced resulting in artifacts in the shallow
58
section. Furthermore, many key reflectors, such as the Strawn Unconformity at t = 800
ms and the basement at t = 1100 ms, are not easily identifiable. In Figure 39, through
the reprocessed data, the key reflectors are more easily identifiable. The frequency
spectrum has been whitened to 140 Hz resulting in a more geological image. In the
reprocessed data, the amplitudes were balanced vertically before migration resulting in
reflectors in the shallow section. Furthermore, the Strawn Unconformity and basement
are continuous and can be tracked with more ease.
Associated with a better seismic image, the performance of key seismic
interpretation tools such as the autotracker has been improved. Figures 40 and 41 show
the improvements on tracking reflectors in the shallow and deep portion of the survey,
respectively. In Figure 40 (left) we see the manual picks by the interpreter used for
autotracking in the shallow section. In the middle figure, we see the result of using the
autotracker in the vendor processed data with a quality factor of 0.7. Due to the high
frequencies added during bluing of the spectrum, areas of autotracking failure are
evident as holes in the horizon. In the right figure, we see the result of the autotracker
on the reprocessed seismic data with the same quality factor of 0.7. Using the
reprocessed data we see fewer holes in the horizon, thereby, improving the utility of the
autotracker. Figure 41 shows the improved utility of the autotracker on a deeper
seismic reflector, requiring more picks by the interpreter. The results are similar to the
shallower horizon, having fewer holes in the tracked horizon while using the
reprocessed seismic survey.
Figure 42 shows a timeslice at t = 540 ms through the seismic amplitude data
volumes processed by the vendor and reprocessed data. Due to the high frequency
59
noise created by bluing the spectrum, the (left) timeslice through the data processed by
the venodor possesses very incoherent or choppy amplitudes. In contrast, the (right)
timeslice through the reprocessed data is much more continuous because artificial high
frequencies were not generated during moderate spectral whitening.
Reprocessing of the Jean survey also resulted in greater utility from the
subsequent attribute interpretation. In Figure 43 we see a timeslice at t = 846 ms
through a coherence volume. Figure 44 shows a timeslice at t = 840 ms through the
curvature attribute of the vendor processed and reprocessed data. Coherency and
curvature, for both the vendor processed and reprocessed, were computed with the same
parameters. In the vendor processed (left) little geological detail can be gathered from
either attribute. However, on the reprocessed (right) data set we see both a coherency
anomaly and a most negative curvature anomaly that appears to be geological i.e. an
incised channel.
60
Figure 38: Vendor processed Jean seismic survey. The amplitudes in the shallow
section have not been properly balanced. Furthermore, the seismic data has been blued
up to 225 Hz causing a ringing around reflectors. The basement at 1100 ms is not easily
visible.
61
Figure 39: Reprocessed Jean survey. The amplitudes have been properly balanced in
the shallow section resulting in more identifiable reflectors. Furthermore, the frequency
spectrum has been whitened to 140 Hz resulting in a more geological result. The
basement at 1100 ms is easily identifiable.
62
Fig
ure
40
: S
hal
low
hori
zon t
rack
ed t
o s
how
the
impro
ved
per
form
ance
of
auto
trac
kin
g a
fter
rep
roce
ssin
g. (L
eft)
Pic
ks
per
form
ed m
anual
ly b
y t
he
seis
mic
inte
rpre
ter.
(M
idd
le)
Auto
trac
ked
ho
rizo
n u
sin
g t
he
ven
do
r pro
cess
ed J
ean s
urv
ey
wit
h a
qual
ity f
acto
r of
0.7
. (
Rig
ht)
Auto
trac
ked
hori
zon u
sing t
he
repro
cess
ed J
ean s
urv
ey w
ith a
qual
ity f
acto
r of
0.7
.
63
Fig
ure
41
: D
eeper
ho
rizo
n t
rack
ed t
o s
how
the
impro
ved
per
form
ance
of
auto
trac
kin
g a
fter
rep
roce
ssin
g. (L
eft)
Pic
ks
per
form
ed m
anual
ly b
y t
he
seis
mic
inte
rpre
ter.
(M
idd
le)
Auto
trac
ked
ho
rizo
n u
sing t
he
ven
dor
pro
cess
ed J
ean s
urv
ey
wit
h a
qual
ity f
acto
r of
0.5
5. (
Rig
ht)
Au
totr
acked
hori
zon u
sing t
he
repro
cess
ed J
ean s
urv
ey w
ith a
qual
ity f
acto
r of
0.5
5.
64
Figure 42: Timeslice through seismic amplitude at t = 540 ms. (left) Vendor processed
seismic data. (right) Reprocessed seismic data. The reprocessed seismic data has
greater continuity due to the lack of high frequencies cross cutting the timeslices.
A A’
Amp.
Amplitude
65
Figure 43: Timeslice at t = 846 ms through a coherency volume. (Upper left) Vendor
processed coherence volume. (Upper right) Reprocessed data coherence volume.
A A’
Amplitude
Coh.
66
Figure 44: Timeslice through most negative curvature at t = 840 ms. (left) Vendor
processed most negative curvature. (right) Reprocessed most negative curvature. The
reprocessed seismic data has greater continuity due to the lack of high frequencies cross
cutting the timeslices.
A A’
Curv.
Amplitude
67
Chapter 5: Interpretation, Inversion and AVAz
Upon completion of an iteration of the Deregowski loop, prestack Kirchhoff
migration into eight azimuthally migrated volumes, reverse normal moveout, matching
pursuit non-stretch normal moveout and prestack structure oriented filtering the data
were stacked for attribute, inversion and AVAz interpretation.
Structural and Stratigraphic Interpretation
With the reprocessed Jean survey stacked I loaded the seismic data set into a
commercial software package to generate a synthetic seismogram and tie the seismic
data to the well logs. Wells ties are essential in order to know which seismic reflectors
correlate to a lithological unit. Using the well ties I was able to track key stratigraphic
horizons across the Jean survey. Figure 45 is an E-W vertical section through the
reprocessed Jean survey with interpreted horizons. The Upper Caddo occurs at
approximately t = 800 ms and the basement reflection occurs at approximately t = 1100
ms. Figure is a N-S vertical section through the reprocessed Jean survey.
Figures 47, 48 and 49 show the time structure maps of the horizons shown in
Figure 45 and 46. In Figure 47 (left) we see the Palo Pinto horizon. It is heavily
contaminated with acquisition footprint, which could be misinterpreted as geology. In
Appendix A the source of the acquisition footprint in the Jean survey is examined via
seismic modeling. In Figure 48 (left) we see the Upper Caddo. A structural high can be
found in the southeast corner (white arrow). This structural high is interpreted to be a
reef or draping caused by reefing in a lower formation. Figure 49 (right) is a time
structure map of the Middle Caddo horizon. In this figure a narrow, sinuous structural
68
low can be seen (white arrow). This is interpreted to be a potential channel cross
cutting the limestone Caddo formation (Figure 61).
Figure 50 (left) shows the thickness of the Strawn Unconformity generated
subtracting the top of the Strawn Unconformity from the top of the Upper Caddo. The
sediment of the unconformity thickens to the northeast, following the overall trend of
the Bend Arch. Figure 50 (right) shows the thickness of the formations from the Upper
Caddo to the Basement. From this figure it is evident that these units thicken in the
southeast corner of the seismic survey, around the interpreted location of the Caddo
reef.
Geometric Attributes
With the stacked reprocessed Jean survey a suite of geometric attributes was
generated. These attributes include dip and azimuth, most positive, most negative
curvature, k1 and k2 curvature, Sobel filter similarity and energy ratio similarity.
Curvature was the most useful attribute during interpretation, allowing for the extraction
of subtle morphologies along horizons and timeslices (Roberts, 2001). Using these
attributes I was able to better understand the expression and further strengthen my
interpretation of key stratigraphic features. Figure 51 shows a timeslice at t = 420 ms
through the (left) most negative curvature and (right) most positive curvature.
Undulations can be seen in the curvature responses that are attributed to acquisition
footprint and not geology. Appendix A shows the source of the acquisition footprint in
the Jean survey to be due to residual groundroll not properly removed by f-k filtering.
Figure 52 (left) shows co-rendered, k1 and k2 curvature and (right) co-rendered k1 and
69
k2 curvature and Sobel filter similarity extracted on the Middle Caddo formation. A
white arrow indicates a potential channel incised through this horizon.
Post stack Acoustic Impedance Inversion
Due to the absence of a shear sonic log within the survey or in a near offset well
only a post stack inversion could be computed, instead of a prestack inversion. To
perform the post stack inversion I generated synthetic seismograms and well ties for 5
wells within the survey that had P-sonic well logs. Because no density logs were
available, density was estimated from Gardner’s equation (1974). Figures 53 and 54
show synthetic seismograms and well ties for two wells within the Jean survey. A
correlation coefficient of 0.669 and 0.637 was achieved for the two wells displayed.
After the 5 wells were tied a model based post stack acoustic impedance
inversion was ran. Figure 55 shows a representative vertical section through Well #1.
With the acoustic impedance from the well logs projected on the wellbore, the inversion
through the seismic data appears to match. Figure 56 (left) shows the acoustic
impedance inversion extracted on the Upper Caddo and (right) low acoustic impedance
co-rendered on the time structure map. From this figure it appears that low impedance
values correlate to structural lows on the Upper Caddo horizon. Figure 57 (left) shows
the acoustic impedance extracted along the Middle Caddo and (right) low acoustic
impedance co-rendered on the time structure map. Figure 58 shows k1 and k2 curvature
co-rendered with low acoustic impedance on the (left) Upper Caddo and (right) Middle
Caddo.
70
Amplitude Variations with Azimuth
In order to detect any anisotropy I computed amplitude variations with azimuth
or AVAz. In order to perform AVAz the seismic data has to be azimuthally migrated. I
chose to migrate the Jean survey into 8 azimuths. After migration the seismic data
could then be stacked, each azimuthal volume individually, and flattened upon a
reference horizon. Next AVAz can be computed by cross-correlating the same trace
from each stacked azimuthally migrated volume.
Figure 59 shows a timeslice at t = 750 ms through (left) azimuth of the AVAz
program and (right) azimuth co-rendered with anisotropy density. There is strong
anisotropy at 90 degrees in the southern portion of the survey. Figure 60 shows a
timeslice at t = 960 ms through (left) azimuth of the AVAz program and (right) azimuth
co-rendered with anisotropy density.
Geobody Extraction
For better visualization of the potential channel in the Middle Caddo indicated
by curvature and coherence I used the method proposed by Wallet (2014). Using this
method allows for three dimensional interpretation of the channel by thresholding k2
curvature and then extracting adjacent voxels with similar values.
Figure 61shows a 3D display with the vertical section being seismic amplitude
and the timeslice at t = 860 ms being K2 curvature through the Middle Caddo. The
channel in the Middle Caddo is indicated by the arrow. Figure 62 shows a 3D chair
display of the channel geobody extracted by thresholding k2 curvature co-rendered with
the Middle Caddo time structure map.
71
Fig
ure
45
: L
ine
C-C
’ sh
ow
ing k
ey h
ori
zons.
A-A’
72
Fig
ure
46:
Lin
e fr
om
A -
A’
show
ing k
ey h
ori
zons.
C-C’
73
Fig
ure
47
: (l
eft)
Tim
e st
ruct
ure
map
of
the
Pal
o P
into
str
atig
raphic
ho
rizo
n t
rack
ed t
hro
ugh
the
repro
cess
ed J
ean s
urv
ey. W
hit
e ar
row
s in
dic
ate
area
s of
the
ho
rizo
n c
onta
min
ated
by
acquis
itio
n f
ootp
rint.
T
he
cause
s o
f th
is f
ootp
rint
wil
l be
inves
tigat
ed f
urt
her
in A
ppen
dix
A.
(rig
ht)
Tim
e st
ruct
ure
map
of
the
KM
A s
trat
igra
ph
ic h
ori
zon t
rack
ed t
hro
ugh t
he
repro
cess
ed
Jean
surv
ey.
74
Fig
ure
48
: (l
eft)
Tim
e st
ruct
ure
map
of
the
Upper
Cad
do s
trat
igra
phic
hori
zon t
rack
ed t
hro
ugh
the
repro
cess
ed J
ean
surv
ey. T
he
whit
e ar
row
indic
ates
are
as o
f th
e hori
zon c
onta
inin
g
dra
pin
g o
ver
a c
arbonat
e re
ef. R
eefs
and d
rapin
g d
ue
to d
eposi
tion o
n t
op o
f re
efs
are
com
mon
com
mon o
il a
nd g
as e
xplo
rati
on t
arget
s (r
ight)
. T
ime
stru
cture
map
of
the
Bas
emen
t h
ori
zon
trac
ked
thro
ugh t
he
rep
roce
ssed
Jea
n s
urv
ey.
75
Fig
ure
49
: (l
eft)
Tim
e st
ruct
ure
map
of
the
Mis
siss
ippia
n s
trat
igra
phic
hori
zon t
rack
ed
thro
ugh t
he
repro
cess
ed J
ean s
urv
ey.
(R
ight)
Tim
e st
ruct
ure
map
of
the
Mid
dle
Cad
do
stra
tigra
phic
hori
zon t
rack
ed t
hro
ugh t
he
rep
roce
ssed
Jea
n s
urv
ey. T
he
wh
ite
arro
w
indic
ates
a s
truct
ura
lly l
ow
are
a in
the
hori
zon t
hat
is
a pote
nti
al c
han
nel
cro
ss c
utt
ing t
he
carb
onat
e S
traw
n f
orm
atio
n.
76
Fig
ure
50
: (l
eft)
Thic
knes
s m
ap c
onst
ruct
ed b
y s
ubtr
acti
ng t
he
Str
awn
Un
con
form
ity s
trat
igra
ph
ic
hori
zon f
rom
the
Upper
Cad
do s
trat
igra
phic
hori
zon.
The
unco
nfo
rmit
y t
hic
ken
s as
we
trav
erse
to
the
nort
hea
st,
coin
cidin
g w
ith t
he
tren
d o
f th
e B
end A
rch.
(R
ight)
Thic
knes
s m
ap c
onst
ruct
ed f
rom
subtr
acti
ng t
he
Upper
Cad
do f
rom
the
Bas
emen
t st
rati
gra
phic
hori
zon.
77
Fig
ure
51
: (l
eft)
Tim
esli
ce t
hro
ugh m
ost
neg
ativ
e cu
rvat
ure
at
t =
420 m
s. (
Rig
ht)
Tim
eles
ice
thro
ugh m
ost
posi
tive
curv
ature
at
t =
420 m
s. T
he
curv
ature
anom
alie
s in
dic
ated
by t
he
whit
e ar
row
s ar
e in
terp
rete
d t
o b
e
acquis
itio
n f
ootp
rint
over
pri
nti
ng g
eolo
gy. I
n A
pp
endix
A,
I in
ves
tigat
e th
e ca
use
of
footp
rint
in t
he
Jean
surv
ey
usi
ng s
eism
ic m
odel
ing.
78
Fig
ure
52
: (l
eft)
Co
-ren
der
ed K
1 a
nd K
2 c
urv
ature
ex
trac
ted a
lon
g t
he
Mid
dle
Cad
do s
trat
igra
phic
hori
zon. (
Rig
ht)
Co
-
rend
ered
K1 a
nd
K2 c
urv
ature
and e
ner
gy r
atio
sim
ilar
ity c
oher
ence
att
rib
ute
ex
trac
ted a
lon
g t
he
Mid
dle
Cad
do
stra
tigra
phic
hori
zon. T
he
whit
e ar
row
indic
ates
a p
ote
nti
al i
nci
sed c
han
nel
.
79
Fig
ure
53
: W
ell
tie
and s
yn
thet
ic s
eism
ogra
m f
or
Wel
l #1 i
n t
he
Jean
surv
ey.
For
the
bes
t co
rrel
atio
n b
etw
een
the
seis
mic
and t
he
wel
l lo
gs,
the
stat
isti
cal
wav
elet
was
rota
ted t
o -
17
0 d
egre
es. W
ith n
o d
ensi
ty l
og a
vai
lable
,
one
was
est
imat
ed u
sin
g t
he
Gar
dner
’s e
qu
atio
n (
1974).
80
Fig
ure
54
: W
ell
tie
and s
yn
thet
ic s
eism
ogra
m f
or
Wel
l #2 i
n t
he
Jean
su
rvey
. F
or
the
bes
t co
rrel
atio
n
bet
wee
n t
he
seis
mic
and t
he
wel
l lo
gs,
the
stat
isti
cal
wav
elet
was
ro
tate
d t
o -
17
0 d
egre
es. W
ith
no
den
sity
lo
g a
vai
lable
, one
was
est
imat
ed u
sing t
he
Gar
dner
’s e
qu
atio
n (
197
4).
81
Fig
ure
55
: V
erti
cal
sect
ion t
hro
ugh t
he
acoust
ic i
mped
ance
inv
ersi
on v
olu
me.
T
he
sect
ion b
isec
ts
Wel
l #1 w
hic
h h
as t
he
acoust
ic i
mped
ance
fro
m t
he
wel
l lo
gs
pro
ject
ed u
po
n i
t. T
he
imped
ance
from
the
wel
l to
the
imped
ance
gen
erat
ed f
rom
the
seis
mic
hav
e a
go
od
co
rrel
atio
n.
82
Fig
ure
56
: (l
eft)
Aco
ust
ic i
mped
ance
ex
trac
ted a
long t
he
Upp
er C
addo h
ori
zon. T
he
Upper
Cad
do i
s a
lim
esto
ne
resu
ltin
g i
n h
igh a
coust
ic i
mped
ance
s. (R
ight)
A
coust
ic i
mp
edan
ce c
o-r
end
ered
wit
h t
he
Upper
Cad
do t
ime
stru
cture
map
. T
he
maj
ori
ty o
f th
e lo
w i
mped
ance
are
as c
orr
elat
e to
low
str
uct
ura
l re
lief
zones
.
83
Fig
ure
57
: (l
eft)
Aco
ust
ic i
mped
ance
ex
trac
ted a
long t
he
Mid
dle
Cad
do h
ori
zon. T
he
Upper
Cad
do i
s
a li
mes
tone
resu
ltin
g i
n h
igh a
coust
ic i
mped
ance
s. (R
ight)
A
coust
ic i
mped
ance
co
-ren
der
ed w
ith t
he
Mid
dle
Cad
do t
ime
stru
cture
map
. L
ow
im
ped
ance
app
ears
to b
e fo
und a
long t
he
chan
nel
fea
ture
see
n
in t
his
hori
zon a
nd t
he
carb
onat
e re
ef i
n t
he
south
east
corn
er o
f th
e su
rvey.
84
Fig
ure
58
: (l
eft)
Aco
ust
ic i
mped
ance
co
-ren
der
ed w
ith K
1 a
nd K
2 c
urv
atu
re e
xtr
acte
d a
lon
g
the
Upper
Cad
do h
ori
zon. (
Rig
ht)
A
coust
ic i
mped
ance
co
-ren
der
ed w
ith K
1 a
nd K
2 c
urv
ature
extr
acte
d a
lon
g t
he
Mid
dle
Cad
do h
ori
zon.
85
Fig
ure
59
: (l
eft)
Tim
esli
ce a
t t
= 7
50 m
s th
rou
gh t
he
anis
otr
op
y a
zim
uth
volu
me.
(R
ight)
Co-r
ender
ed a
nis
otr
op
y a
zim
uth
and d
ensi
ty.
86
Fig
ure
60
: T
imes
lice
at
t =
960 m
s th
rough t
he
anis
otr
op
y a
zim
uth
volu
me.
(R
ight)
C
o-r
ender
ed
anis
otr
op
y a
zim
uth
and d
ensi
ty.
87
Fig
ure
61
: C
hai
r dis
pla
y f
rom
A t
o A
’.
Ver
tica
l se
ctio
n i
s se
ism
ic a
mp
litu
de
and
tim
esli
ce i
s K
2
curv
ature
. T
he
arro
w i
nd
icat
es a
pote
nti
al c
han
nel
in t
he
Mid
dle
Cad
do t
hat
is
extr
acte
d u
sing g
eobo
die
s.
88
Fig
ure
62
: C
hai
r dis
pla
y s
how
ing g
eobod
y e
xtr
acti
on o
f ch
annel
aft
er W
alle
t (2
014).
T
he
ver
tica
l se
ctio
ns
are
seis
mic
am
pli
tude
and t
he
hori
zonta
l sl
ice
is t
he
Mid
dle
Cad
do t
ime
stru
ctu
re m
ap. T
he
chan
nel
geo
bod
y e
xtr
acte
d b
y t
hre
shold
ing K
2 c
urv
atu
re i
s sh
ow
n i
n r
ed.
89
Chapter 6: Conclusion
Reprocessing the legacy Jean 3D seismic surveys in Young County, Texas can
provide new insight into the geology within the survey. Utilizing modern processing
workflows, along with careful attention to trace editing, statics and velocity analysis and
advancements in processing and imaging technology such as prestack Kirchhoff
migration, non-stretch NMO, and prestack structure oriented filtering can improve the
subsequent interpretation. Furthermore, reprocessing with a focus on preserving the
data for quantitative interpretation will increase the amount of information that can be
gained from the 3D seismic data volume.
I have shown that by not bluing the frequency spectrum, as the commercial
vendor had done, we preserve the frequency content of the data allowing for subsequent
attribute and quantitative interpretation. Curvature and coherence attributes show
geology after reprocessing where as previously the only showed high frequency noise.
Key reflectors, such as the Caddo limestone and the basement, are easier to interpret
and quantitative measures, such as AVAz and acoustic impedance inversion can be
utilized.
With the quantitative interpretation of the reprocessed data, we are able to
identify a previously unseen channel in the Caddo formation. We are also able to detect
low impedance zones trending with the channel, interpreted to be channel sands.
90
References
Aisenberg, M., 2013, the value of reprocessing legacy data: A case study of Bois D’Arc
a Mississippi play in northeastern Oklahoma: Master’s Thesis, University of Oklahoma,
Norman, Oklahoma.
Blakey, R., 2011, Colorado Plateau stratigraphy and geology and global and
regional paleogeography: Northern Arizona University of Geology,
http://www2.nau.edu/rcb7/RCB.html. accessed 04/10/2015
Bowker, K.A., 2007, Recent development of the Barnett Shale play, Fort Worth Basin:
West Texas Geological Society Bulletin, 42, 4-11.
Cahoj, M. P., S. Verma, B. Hutchinson, J. Qi and K.J. Marfurt , 2015, Pitfalls in seismic
processing: part 2 velocity analysis sourced acquisition footprint, Submitted for 85th
Annual International Meeting, SEG, Expanded Abstracts.
Chopra, S., G. Larsen, 2000, Acquisition footprint – its detection and removal: CSEG
Recorder 25, 16-20.
Cvetkovic M., N. Pralica, S. Falconer, K. J. Marfurt, and S. C. Pérez, 2008, Comparison
of some algorithms for acquisition footprint suppression and their effect on attribute
analysis, 78th Annual International Meeting, SEG, Expanded Abstracts, 2637-2641.
Davogustto, O., 2011, Removing footprint from legacy seismic data volumes: Master’s
Thesis, University of Oklahoma, Norman, Oklahoma.
Dickinson, W.R., 1976, Plate tectonic and hydrocarbon accumulations: AAPG
Coninuing Education Course Note Series. 1, 62.
Deregowski, S., 1990, Common-offset migrations and velocity analysis: First
Break, 6, 225-234.
Dowdell, B.L, 2013, Prestack seismic analysis of a Mississippi lime resource play in the
midcontinent, U.S.A.: Master’s Thesis, University of Oklahoma, Norman, Oklahoma.
Evanoff, B.G., 1976, The Bend Arch: Abilene Geological Society Geological
Contributions, 11-13.
Fernandez, A.A., 2013, 3D seismic attribute expression of the Ellenburger Group karst-
collapse features and their effect on the production of the Barnett Shale, Fort Worth
basin, Texas: Master’s Thesis, University of Oklahoma, Norman, Oklahoma.
Flippin, J., 1981, Stratigraphy, Structure, and Economic Aspects of Paleozoic strata,
Erath County: Dallas Geological Society.
91
Gardner, G.H.F., Gardner, L.W., and Gregory, A.R., 1974, Formation velocity and
density – the diagnostic basics for stratigraphic traps: Geophysics, 39, 770-780
Guo, S., B. Zhang and K.J. Marfurt, 2012, Noise suppression using preconditioned
least-squares prestack time migration: Application to the Mississippi limestone: 82nd
Annual International Meeting of SEG Expanded Abstract, 1-6.
Ha, T., 2014, Seismic Reprocessing and Interpretation of a shallow “BURIED HILL”
play: Texas Panhandle : M.S. Thesis, The University of Oklahoma.
Hampson, D., AVO, 1991, Inversion, theory and practice: The Leading Edge, 6,
39-42.
Hill S., M. Shultz, and J. Brewer, 1999, Acquisition footprint and fold‐of‐stack plots:
The Leading Edge, 18, 686-695.
Hoz, G.A., K.L. Larner, 1995, A comparison of Moveout-based approaches to
suppression of ground-roll and multiples: Colorado School of Mines.
Marfurt K. J., R. M. Scheet, J. A. Sharp, and M. G. Harper, 1998, Suppression of the
acquisition footprint for seismic sequence attribute mapping: Geophysics, 63, 1024-
1035.
Marfurt, K. J., and T. M. Alves, 2015, Pitfalls and limitations in seismic attribute
interpretation of tectonic features: Interpretation, 3, SB5-SB15.
Montgomery, S.L., D.M. Jarvie, K.A. Bowker and R.M. Pollastro, 2005, Mississippian
Barnett shale, Fort Worth basin, north-central Texas: Gas-shale play with multi-trillion
cubic foot potential: AAPG Bulletin, 89, 155-175.
Partyka, G., J. Gridley, and J. A. Lopez, 1999, Interpretational applications of spectral
decomposition in reservoir characterization: The Leading Edge, 18, 353-360.
Perez, G., K. J. Marfurt, 2008, New azimuthal binning for improved delineation of
faults and fractures: Geophysics, 73, 7-15.
Pollastro, R.M., 2007, Total petroleum system assessment of undiscovered resources in
the giant Barnett Shale continuous (unconventional) gas accumulation, Fort Worth
Basin, Texas: AAPG Bulletin, 91, 551-578.
Roberts, A., 2001 Curvature attributes and their application to 3D interpreted horizons.
First Break, 19, 85-99.
92
Ronen, J., J. Claurbout, 1985, Surface-consistent residual statics estimation by stack-
power maximization, http://sepwww.stanford.edu/theses/sep42/42_17.pdf. Accessed 30
March 2015.
Russell, B.H., 1989, Statics corrections – a tutorial: Canadian Geophysical Society,
http://74.3.176.63/publications/recorder/1989/03mar/mar1989-statics-corrections.pdf.,
accessed 08 October, 2014.
Sheriff, R.E., 2004, What is deconvolution?,
http://www.searchanddiscovery.com/pdfz/documents/2004/sheriff/images/sherrif.pdf.ht
ml. Accessed 13 February 2015.
Stein, J.A., T. Langston, and S.E. Larson, 2009, A successful statics methodology for
land data: The Leading Edge, 222-227
Thomas, J., 2003, Integrating synsedimentary tectonics with sequence stratigraphy to
understand the development of the Fort Worth Basin,
http://rockfractureandstress.com/Burnett%20Ranch/INTEGRATING%20SYNSEDIME
NTARY%20TECTONICS%20WITH%20SEQUENCE%20STRAT%20TO%20UNDE
RSTAND%20THE%20%20DEVELOPMENT%20OF%20THE%20FORT%20WORT
H%20BASIN.pdf. Accessed 24 January 2015
Wallet, B. C., 2014, Seismic attribute expression of fluvial-deltaic and turbidite
systems, PhD dissertation, University of Oklahoma.
Walper, J.L., 1982, Plate tectonic evolution of the Fort Worth basin: Dallas Geological
Society.
Verma, S., S. Guo, and K. J. Marfurt, 2014, Prestack suppression of high frequency
ground roll using a 3D multiwindow KL filter: Application to a legacy Mississippi
Lime survey, 84th Annual International Meeting, SEG, Expanded Abstracts, 4274-4278.
Vermeer, G.J.O, 2012, 3D seismic survey design: SEG, Tulsa, Oklahoma.
Yilmaz, O., 2000, Seismic data analysis: SEG, Tulsa, Oklahoma.
Zhang, B., 2014, Long offset seismic data analysis of resource plays: PhD dissertation,
The University of Oklahoma.
93
Appendix A: Understanding processing pitfalls with seismic modeling
The following appendix is part of two expanded abstracts submitted for the 85th Annual
SEG meeting. My co-authors on these abstracts are Sumit Verma and Bryce
Hutchinson.
Pitfalls in seismic processing: part 1 groundroll sourced acquisition footprint
Summary
Whether it is in reference to the limitations of interpretation or associated with
seismic processing, usage of the phrase acquisition footprint is never in a positive
context. Footprint contaminates both time structure map and impedance inversion.
Although common, footprint is often poorly understood. Footprint is more common in
older, lower fold surveys. Part of this mystery is due to the division of labor in most
exploratory companies. Processing is usually conducted by specialists in a service
company, while attribute analysis is conducted by interpreters (often geologists) in an
oil company. Often, younger interpreters have never processed 3D seismic data, while
younger processors have never analyzed attributes. As a part of a reprocessing effort
for quantitative interpretation analysis, Cahoj (2015) encountered severe footprint
masking his shallow exploration target. We attempt to modify his processing workflow
to ameliorate the footprint lead to an effort to understand its cause, at least for this
survey. Upon completion of seismic processing we are left with a stacked version of our
synthetic data in which we can compute seismic attributes. We show that the
subsequent attribute interpretation is greatly affected by footprint caused by residual
groundroll. Lastly, we show an attribute interpretation corresponding to real 3D seismic
dataset and conclude that many artifacts seen in the dataset, often labeled under the
94
broad category of acquisition footprint, are actually residual groundroll not properly
removed during the processing flow. Because out of plane groundroll can have
hyperbolic moveout common noise removal techniques, such as F-K filtering, that
operate under the assumption of modeling noise with different linear moveouts, fail.
Introduction
Acquisition footprint refers to the imprint of acquisition geometry seen on
seismic amplitude timeslices and horizons. Acquisition footprint can obstruct not only
classical seismic interpretation but also affect interpretation based on seismic attributes
(Marfurt and Alves 2015, Marfurt et al., 1998). Seismic attributes, especially coherence
and curvature, often exacerbate the effect of footprint making their utility diminish
(Marfurt and Alves 2015; Verma et al., 2014).
With footprint being such a common problem its occurrence and formation are often
poorly understood (Chopra and Larsen, 2000). Although many methodologies have
been developed to remove linear coherent noise and acquisition footprint (Cvetkovic et
al., 2008 and Marfurt et al., 1998), little has been done in the way of illustrating its
occurrence via modeling. Hill et al. (1999) investigated acquisition footprint is caused
by inaccurately picked NMO velocity. Although groundroll is one of the prime causes
of acquisition footprint, the footprint pattern caused by the presence of groundroll has
not been modeled and documented.
One of the main causes of seismic acquisition footprint is sparse spatial sampling. It is
particularly challenging to remove aliased groundroll. Because of this the residual
groundroll’s occurrence on the stacked seismic data can be strong enough to influence
the interpretation. We study a low fold legacy seismic survey of North Central Texas
95
and observed acquisition footprint with the North-South lineaments (Figure 1a) aligned
with the receiver lines. We investigate what can cause such footprint to be present in
our dataset; in this paper we present the findings.
Motivation
We observed north–south acquisition footprint present on the curvature attribute
shown in Figure 1a. The presence of this acquisition footprint hindered our attribute
assisted interpretation. Because of this we had an incentive to understand its origin. We
hypothesis that this acquisition footprint could have three potential sources:
1) Inadequate removal of groundroll,
2) NMO far offset stretch, and
3) Improper velocity analysis
In this paper we decide to investigate the effect of inadequately removed groundroll.
In Part 2 (Cahoj et al., 2015) of this abstract we will try to understand the effect of
NMO stretching and incorrect velocity analysis on our seismic interpretation. Equipped
with an actual seismic dataset with acquisition footprint, we are able to construct a
synthetic analogue.
Methodology
Seismic modeling
The objective of this model is to see the effect of residual groundroll on stacked
seismic data after processing and its relation with reflectors.
To do so we created a simple 3D flat layer seismic model with four layers. The
acquisition geometry is shown in Figure 2, with 6 receiver lines and 9 shot lines. Each
receiver line contains 60 receiver groups totaling 360 geophones, and each shot line
96
contains 18 sources totaling 162 shots. The model has a strong presence of broad
bandwidth (0-50Hz) dispersive groundroll. We generated two separate models, one for
groundroll using an elastic modeling approach with only the weathering layers and a
second model with four layers using an acoustic modeling approach. We added these
two models to simulate the final 3D acquisition geometry for our study.
Seismic processing
The seismic processing can be broken into 7 steps.
1) Importing the synthetic seismic data
2) Defining the geometry
3) Sorting the data by absolute offset
4) Identifying the noise corridor with a mute and finding its respective linear
moveout velocity
5) Model the noise in the F-K domain
6) Inverse linear moveout and subtraction
7) NMO correction and stacking the synthetic data
Figure 3a shows a common shot the synthetic sorted by absolute offset. It is easy to
identify the lower velocity groundroll crosscutting and overbearing the reflectors.
Figure 3b shows the groundroll modeled by a standard F-K noise filtering procedure
and Figure 3c shows the results after the modeled groundroll is subtracted from the
input model. In this figure we see that most of the high amplitude groundroll has been
removed and the reflectors, once overprinted, are now visible. Upon completion of
groundroll removal the synthetic data were NMO corrected and stacked (Figure 4a).
Attribute interpretation
97
We computed a suite of seismic attributes using a commercial software package
on both the modeled synthetic seismic data and the actual seismic data. Such attributes
included dip and azimuth, energy ratio similarity and curvature. With these attributes
we were able to determine footprint’s response from improperly removed groundroll.
Using the modeled seismic data we were able to make an analogue to actual seismic
data to compare groundroll’s response and effect on interpretation.
Results
Figure 4a shows the inline of the stacked synthetic seismic data. The
undulations in the shallow section are the responses of constructively and destructively
interfering groundroll not properly removed by F-K filtering. Figure 4b shows the
corresponding inline through the actual seismic data. It is evident that similar
undulations exist in the shallow section of the real seismic data.
Figure 5a is a timeslice at t=1.320s through the most negative curvature
response of the stacked synthetic seismic data. We find that the response of curvature,
an attribute commonly used to map folds, flexures and deformation about faults, is
greatly contaminated by the inadequately removed groundroll. Figure 1a shows the
corresponding timeslice at t=0.410s through the most negative curvature of the real
seismic data; containing a similar footprint expression.
Figure 6a shows a horizon tracked through the 2nd layer in the synthetic dataset.
Because the layers were modeled to be horizontal we expect a uniform surface at a
constant depth. However, we can see rectilinear features, particularly strong in the
East-West direction. These features can also be seen in Figure 6b, the real seismic data.
Conclusions
98
Our analysis indicates that the undulations caused by residual groundroll will be
present on the seismic, having strongest amplitude near the surface and attenuating with
depth.
We conclude that inadequately removing groundroll can result in erroneous and
more difficult interpretations. Furthermore, seismic attributes, often used by less
experienced interpreters to accelerate there interpretations, are not immune to
acquisition footprint caused by groundroll. In many cases, seismic attributes exacerbate
the effects of this noise.
99
4000ft
4000ft
Figure A-2: Timeslice at t=0.41s through coherence
volume from real seismic dataset. The North-South
lineaments are aligned with the receiver lines. These
artifacts are weaker at depth but overprint the
objective at t=1.0s.
Figure A-1: Timeslice at t=0.41s through most
negative curvature volume from real seismic dataset.
The North-South lineaments are aligned with the
receiver lines. These artifacts contaminate attribute
volumes.
100
0 2000 12000
6000
10000 East-West (ft.)
No
rth
-So
uth
(ft.
)
1000
Figure A-3: The synthetic model’s geometry. Sources are in red and
receivers are in green. The geometry is perfectly rectilinear which is
not the case with actual seismic data due to surface obstructions.
101
Figure 3. Shot vs absolute offset sorted (a) modeled seismic data
with four reflectors and groundroll with a large bandwidth (0-50Hz).
b) F-K modeled groundroll to be removed from the modeled seismic
data (a). (c) Result of subtracting F-K modeled groundroll (b) from
modeled seismic (a). Notice large amounts and high amplitude
groundroll is removed, but residual remains.
Positive
Negative
Figure A-4:
102
0.5
0.0
1.0
1.5
2.0
Time (s)
0.45
0.1
0.25
a)
b)
Figure A-5: (a) Inline through the synthetic seismic data.
(green horizon is displayed in Figure 6a) ( b) Inline of
real seismic data (yellow horizon is displayed in Figure
6b) . Notice the undulation anomalies caused by
inadequately removed groundroll in Figure 6a and similar
undulation features can be seen in Figure 6b most likely
caused by groundroll.
103
a)
b)
Figure A-6: (a) Timeslice at t=1.320s through most negative curvature of
the synthetic seismic data. Notice the undulation anomalies caused by
inadequately removed groundroll. (b) Coherence at t=1.320s of the
synthetic seismic data. Similar undulation features can be seen causing
lateral discontinuity in the reflectors.
104
4000ft
a)
b)
Figure A-7: Horizons tracked through (a) synthetic data, displayed on
Figure 4a as green horizon. (b) real seismic data, displayed on Figure 4a
as yellow horizon. The linear striations (red arrows) are due to
residual groundroll overprinting P-wave reflections.
105
Pitfalls in seismic processing: part 2 velocity analysis sourced acquisition footprint
Summary
Seismic interpretations are often highly subjective and depend on the
imagination of the geologist or geophysicist constructing them. Because of this nature
an interpretation can be no better than the data under investigation or the interpreter’s
ability to understand and delineate features created from the processing procedure. If
acquired improperly or processed poorly the geological interpretation will have
shortcomings as well. Such processing errors could be the product of improper statics,
poor velocity analysis and not adequately removing coherent noise. The result of these
errors, if not mistaken as geology, is often classified under the broad category of
acquisition footprint.
In this paper we use synthetic seismic data composed of four reflectors to
investigate the effects of poor velocity analysis and normal moveout stretching. We
compare the results from the synthetic seismic dataset to a real 3D seismic dataset. We
show an attribute interpretation of both datasets and how inaccurate processing can lead
to fallacious claims about the geological background.
Introduction
Processing procedures can greatly affect the reliability of conventional
interpretation and the utility of seismic attribute interpretation. Because of this seismic
modeling can be an invaluable tool in quality checking your seismic processing job.
Very few efforts have been made to explain acquisition footprint and processing
generated noise using synthetic models. Hill et al. (1999) discussed acquisition footprint
caused by inaccurately picked NMO velocity. Ha (2014) used seismic modeling in an
106
attempt to better understand the response of a fractured granitic basement. He also used
elastic modeling to identify coherent seismic noise, such as groundroll. With the insight
gained from seismic modeling he was able to better identify and eliminate coherent
noise during seismic processing.
Seismic attributes, especially coherence and curvature, often exacerbate the
effects of inaccurate processing procedures (Verma et al. 2014; Marfurt and Alves,
2015). Because attributes are popular, particularly among younger interpreters, as a
method to hasten interpretations this could lead to pitfalls in our geologic model
(Marfurt and Alves, 2015).
One of the key factors affecting the resolution of seismic data is velocity
analysis. With improper velocity analysis and without muting NMO stretch the
frequency content of the reflectors can be greatly deteriorated. Furthermore, it can
create pseudo-geological artifacts that could lead to an incorrect interpretation. Using a
small 3D seismic land dataset from North Central Texas, we investigate the origin of
features in our final stack and attribute volumes after reprocessing this legacy dataset.
Motivation
We processed a small 3D land seismic dataset with a conventional workflow. Upon
completion we arrived at the conclusion that the geophysical interpretation contained
features that did not match the a priori geological background. With the geological
background being confirmed by vast numbers of wells in the adjacent area drilled over
many decades of oil and gas exploration. This led us to conclude that the cause of the
incorrect interpretation was at the fault of erroneous processing parameters. We
formulate that the processing job could be affected by any of the three factors:
107
1) NMO far offset stretch
2) Improper velocity analysis
3) Inadequate removal of groundroll
In this paper we investigate the effects of NMO stretching and improper velocity
analysis on synthetic seismic data. In Part 1 (Verma et al., 2015) of this abstract we will
try to understand the effect of groundroll on our seismic interpretation.
We then make an analogue to anomalies seen in our actual seismic data. In doing so we
are able to better understand how processing can affect the geological interpretation.
We are also able to connect common pitfalls in seismic processing using seismic
models.
Methodology
Seismic modeling
We created a simple 3D isotropic seismic model with four layers. The
acquisition geometry is shown in Figure 1, with 6 receiver lines and 9 shot lines. Each
receiver line contains 60 receivers totaling 360 geophones, and each shot line contains
18 sources totaling 162 shots.
Seismic processing
The seismic processing can be broken into 5 steps:
1) Importing the synthetic seismic data
2) Defining the geometry
3) Velocity analysis
4) NMO correction
5) Stacking the synthetic data
108
Figure 2a shows the raw synthetic seismic data sorted in shot versus offset. The
four hyperbolic reflectors in the model are clearly identifiable. Figure 2b shows the
semblance panel and Figure 2c shows the respective NMO corrected gather of the
picked semblances (in white). In this figure we see that the processor has picked the
velocities on the semblance panel to be too fast for reflector 2 and 4. The result is an
undercorrected NMO corrected gather. Figure 2d shows a semblance panel; Figure 2e
the respective NMO corrected gather; and Figure 2f the NMO corrected gather with a
30% stretch mute. However, in this case the processor has selected the semblances to
be too fast in reflector 1 and 3 resulting in an overcorrection. With these inaccurately
picked velocity models we NMO correct and stack the synthetic seismic model.
Attribute interpretation
We computed a suite of seismic attributes using a commercial software package
on both the modeled synthetic data and the real seismic data. Such attributes included
dip and azimuth, coherence and curvature. We then analyzed how improper velocity
analysis and NMO stretching affect attribute response. Using the information gained
from the synthetic model we were able to use our real 3D seismic data as an analogue to
better understand the influence of processing on interpretation.
Results
Figure 3a shows an inline through the processed and stacked four layered synthetic
seismic dataset. We can see quasi-hyperbolic artifacts caused by the constructive and
destructive interference of interfaces not properly flattened during velocity analysis.
The blue arrows point to the artifacts on reflector 1 and 3. This inline is constructed
from the velocity panel and NMO corrected gather shown in Figure 2d and e.
109
Figure 3b shows an inline through the processed and stacked four layered modeled
seismic dataset. In this figure we have applied a 30% stretch mute, to mute far offset
stretching phenomena. The quasi-hyperbolic artifacts caused by the constructive and
destructive interference of interfaces not properly flattened during velocity analysis are
smaller, however still identifiable. The blue arrows point to the artifacts on reflector 1
and reflector 3. This inline is constructed from the velocity panel and NMO corrected
gather shown in Figure 2d and f.
Figure 3c shows an inline through the processed and stacked four layered modeled
seismic dataset. This data underwent velocity analysis that was intentionally picked too
fast. The resulting constructive and destructive interference patterns from the reflectors
results in hyperbolic anomalies. The blue arrows point to the artifacts on reflector 2 and
reflector 4. This inline is constructed from stacking the velocity panel and NMO
corrected gather shown in Figure 2b and c.
Figure 4a shows undulatory patterns in the real data similar to those seen in our
synthetic dataset. Figure 4b shows the attribute expression (most negative curvature)
through a timeslice at 850ms through the modeled seismic data. Compared with Figure
4c, the attribute expression of the actual seismic data at 365ms, we see that although
both have perturbations in their expression, they appear to have different patterns.
Although it appears that poor velocity analysis and NMO stretching can cause artifacts
in the seismic data, attribute interpretation seems to show that the cause of our footprint
comes from another source.
110
Conclusions
From interpreting the models, we initially conclude that inaccurate velocity analysis and
not removing NMO stretching can result in incorrect geological interpretations. While
looking at an inline it appears that our real seismic dataset suffers from NMO velocities
picked to be too slow. This can be seen as the undulations in the shallow sections that
could be interpreted as karst features or pock marks. However, analyzing the attribute
expression of both the modeled data and the real seismic data we see different patterns.
This leads us toward the conclusion that velocity analysis may not be the source of our
footprint. In Pitfalls in seismic processing: part 1, we analyze groundroll as a potential
source of acquisition footprint. We believe that in our real 3D seismic dataset, the
expression of the footprint is more aligned with that of groundroll than velocity
analysis. We also deduce that footprint and noise from groundroll or velocity analysis
can be deciphered by other means. From our observations groundroll’s expression will
start strongest at the surface and attenuate with depth. However, inaccurate velocity
analysis will only display features within the interval that the velocity is mis-picked.
Picking a correct velocity in the vertically adjacent section will result in reflectors with
no artifacts. Lastly, improper velocity analysis can create patterns that even more
experienced interpreters could perceive as geology as seen in Figure 4a.
111
East-West(ft.)
No
rth
-So
uth
(ft.
)
12000 10000 2000 0.0
6000
1000
Figure A-1: The acquisition geometry for the synthetic model. Sources
are in red and receivers are in green. The geometry is perfectly rectilinear
which is not the case with actual seismic data due to surface obstructions.
112
Figure A-2: (From left to right, top to bottom) Input modeled seismic data with four
hyperbolic reflectors. Semblance panels for raw gather. NMO corrected gathers for
semblance picked too fast. Semblance panels for raw gather. NMO corrected gather
for semblance picked too slow. Previous NMO corrected gather with 30% stretch
mute.
113
Figure A-3: (a) Modeled seismic data with NMO velocity picked too
slow. (b) is (a) with 30% stretch mute. (c) Modeled seismic data with
NMO velocity picked too fast.
114
Figure A-4: (a) Actual seismic data with hyperbolic artifacts that could be
the effect of inaccurate velocity analysis (blue arrows). (b) Timeslice
through most negative curvature of the modeled seismic data. (c) Timeslice
through most negative curvature of actual 3D seismic data. Notice the
discrepancy of the patterns of curvature between (b) and (c).
b)
a)
c)