Ministry of Higher Education

And Scientific Research

University of Baghdad
College of Science
Department of Geology

Noise Attenuation of 3D Seismic Cube in

Al-Samawah Area South-west of Iraq
A thesis
Submitted to the College of Science, University of Baghdad, in
partial fulfilment of the requirements for the degree of Master in
Geology (Geophysics)

By

Ahmed Hussein Ali

B.Sc., University of Baghdad, 2016

Supervised by

Prof. Dr. Ali M. Al-Rahim

2020 AD 1440 AH
 Dedicated to

My Father
Long life

My Mother
Merciful and forgiveness

My Brothers and Sisters

Love and Loyalty

Ahmed
 ACKNOWLEDGMENTS

I would like to express my appreciation and deep gratitude to my

supervisor (Dr. Ali M. Al-Rahim), for his aid, patience and
encouragement at all stages of this work.

My best thanks to the dean of the college of science, head of

geology department, and postgraduate unit of department in
Baghdad University and all teachers who guided me to the right
way in bachelor and master stages.

I am so grateful to the Oil Exploration Company- department of

processing and its managers for providing me all the facilities
and information that helped me in performing this research.

My deep gratitude to the team members who trained with them in

the Oil Exploration Company, especially to Miss. Rehab Fadel
for helping me and providing me all I need over the course of the
year. Also special thanks to Mr. Osama, Mr. Ahmed and Mr.
Yasser for helping me with my project.
 Abstract
The aims of this study is to enhance signal to noise ratio of seismic data using
different techniques and filters, in order to attenuate as much as possible of coherent
and random noise without affecting the signal.

Two types of seismic data were used and processed in this thesis, synthetic
seismic data that was generated using Seismic Unix programs, and 3D real data of
Al-Samawa area which lies 280 km south-east of Baghdad. The 3D seismic data
were recorded by the third seismic crew of Iraqi Oil Exploration Company in 2015.

The synthetic seismic data were generated using Seismic Unix based on inline
stacked section of 3D seismic data of Al-Samawa. First, an earth model was built,
then 120 shots of seismic data were generated from the model, that was processed
later through different filters after amplitude correction and random noise addition
of 10% signal to noise ratio (10% signal + 90% noise), since synthetic data don’t
contain any type of noise. Three different filters were used to remove the previously
added random noise, which is sufxdecon, sufilter and suconv filters. Applying these
filters after different parameters and tests gave an almost bad results, although a large
amount of noise was removed, but also a significant part of the signal was also
removed, mostly due to sefxdecon filter, since it was the main filter and the other
filter used to remove the residual noise. After many experiments, it was found that
applying normal move-out to the shot gather before removing noise will give a
bigger seperation between the signal and noise, resulting in better filtering and
solving the problem of the removed signal.

The second part of this thesis will deal with the processing and noise attenuation
of the 3D data of Al-samawa area. To improve signal to noise ratio, many filters
were applied using Omega software. First, reformatting was applied to convert the
data to Omega format, then gemetry update was held to update the headers.
Geometrical spread compensation was applied to compensate the lost amplitudes
due to spherical diversion.

The noise attenuation part started with Anomalous Amplitude Attenuation

(AAA) filter to remove anomalous amplitude random noise and abnormal
amplitude spikes. Due to the large amount of random noise in the data, the filter
was applied many times with different parameters, three times in the first pass and
six times in the second pass, after applying Surface Constant Amplitude
Compensation (SCAC). Tau-p Linear Noise Attenuation Filter (TPLNA) used to
remove coherent linear noise, by transforming the data into Tau-p domain after
applying NMO to the shot gather then inversing it after manually rejecting the
noise.
F-X Coherent Noise Suppression (FXCNS) filter was used to eliminate Ground-
Roll using frequency and velocity parameters. After dividing the shot into multiple
frequency panels, it was found that the Ground-Roll lies approximately between
frequency range of (3-40 to 42) Hz, the filter gave very good result in regarding to
remove the ground-roll. The residual linear noise was removed using F-K filter, by
transforming into frequency-wavenumber domain then using fan filter to manually
rejecting the noise.
Finally, Radon filter was used to remove the parabolic noise that formed the
shape of Kris-cross noise. By applying the radon transform to target these types of
noise and rejecting it manually. NMO was applied prior F-K and Radon filter and
inversed after rejecting the noise. The final results were excellent and shows a
significant enhancement to the data which can be ascertained by the signal to noise
spectrum and the time slice.
 Table of Content
Chapter 1: Introduction Page No.
1.1 Preface. 1
1.2 Geographical location and coordinates of the study 2
area.
1.3 Geology of the study area. 3
1.3.1 Surface geology. 3
1.3.2 Subsurface geology. 4
1.4 Tectonic setting of the study area. 6
1.5 The processing systems. 7
1.6 Previous seismic survey of the study area. 7
1.7 Previous Studies Outside of Iraq. 8
1.7.1 Previous studies using the 3-D seismic reflection 8
survey & processing technique.
1.7.2 Previous study using seismic Unix in seismic data 9
processing.
1.8 Previous Studies in Iraq 11
1.8.1 Previous Studies Using the 3-D Seismic Reflection 11
Survey & Processing Technique.
1.8.2 Previous Studies Concerning Signal to Noise 12
Enhancement
1.8.3 Previous Study Using Seismic Unix in Seismic Data 13
Processing.
1.9 Aims of the study. 14
1.10 Procedure of achievement of the study. 14
Chapter 2: Theoretical Background
2.1 Preface. 16
2.2 Spectral analysis. 17
2.3 Fourier transform. 17
2.4 Sampling rate. 18
2.5 Seismic data processing steps. 19
2.5.1 Reformatting. 19
2.5.2 Geometry definition. 20
2.5.3 Spherical diversion compensation. 20

I
 2.5.4 Noise attenuation. 20
2.5.5 CMP sorting. 20
2.5.6 Velocity analysis. 21
2.5.7 NMO-correction. 23
2.5.8 Stacking. 24
2.6 Noise in seismic data. 25
2.6.1 Signal to noise ratio (S/N). 26
2.6.2 Types of seismic noise. 27
2.6.2.1 Coherent noise. 28
2.6.2.2 Random noise. 34
2.7 3D data filtering 35
2.7.1 Anomalous Amplitude Attenuation (AAA) Filter. 35
2.7.2 Tau-p Linear Noise Attenuation (TPLNA) Filter. 38
2.7.3 F-X Coherent Noise Suppression (FXCNS) Filter. 41
2.7.4 Frequency-Wavenumber (F-K) Filter. 43
2.8 Synthetic seismic data processing. 45
2.8.1 Advanced Seismic Unix (SU) concepts. 45
2.8.2 Trace headers. 48
2.8.3 Synthetic seismic data filtering 49
2.8.3.1 SUFXDECON. 49
2.8.3.2 SUFILTER. 50
2.8.3.3 SUCONV. 52
Chapter 3: Modeling and Processing Al-Samawah
Strata Using Seismic Unix.
3.1 Preface. 54
3.2 Seismic Unix programs. 55
3.2.1 Editing, sorting and manipulation programs. 55
3.2.2 Filtering, transforms and attributes. 56
3.2.3 Gain, NMO, stack and standard processes. 56
3.2.4 Model building and data acquisition. 56
3.2.5 Graphics. 57
3.2.6 Import / Export. 57
3.3 Model building. 57
3.4 Synthetic seismic data acquisition. 63
3.5 Amplitude correction. 65
3.6 Random noise addition. 67
3.7 Random noise attenuation. 68
3.8 Sort from shot gather to CMP gather. 78
3.9 Velocity analysis and normal move-out. 79
II
 Chapter 4: Noise Attenuation for 3D Al-Samawah
Cube.
4.1 Preface. 84
4.2 Acquisition parameters. 84
4.3 Work flow 85
4.4 Reformat. 86
4.5 Geometry update. 86
4.6 Geometrical Spreading (Spherical Diversion) 86
Compensation.
4.7 Noise attenuation. 88
4.7.1 Anomalous Amplitude Attenuation (AAA). 88
4.7.2 Tau-p Linear Noise Attenuation (TPLNA). 94
4.7.3 F-X Coherent Noise Suppression (FXCNS). 99
4.7.4 Anomalous Amplitude Attenuation (AAA) 2nd pass. 106
4.7.5 Multi-Channel Dip Filter (F-K Filter). 109
4.7.6 Radon (Kris-Cross) Removal. 115
4.8 The Final Results. 116
Chapter 5: Conclusions and Recommendations
5.1 Conclusions. 123
5.2 Recommendations. 125
References 126
Appendix A 138

List of Figures
Figure Page No.
1.1 Survey location for the study area. 3
1.2 The geological map of the study area. 4
1.3 The geological column of Al-Samawa well-1(Sa-1). 5
1.4 Tectonic map of the study area. 6
2.1 An example of Fourier transformation. 18
2.2 General seismic data processing sequence. 19

III
2.3 (a) Seismic data in shot-receiver (s, g) coordinates. 21
(b) Seismic data processing in midpoint-offset
coordinates.
2.4 (A) synthetic CMP gathers (B) semblance scans for 22
CMP gathers. Black curves indicate velocity picks.

2.5 Seismic data that is sorted by CMP then corrected 23

by NMO.

2.6 (a) CMP gather of single event with a 2264 m/s 24

velocity (b) NMO corrected gather by using the best
velocity (c) over corrected gather by using low
velocity 2000 m/s (d) under corrected gather using
high velocity 2500 m/s.
2.7 The stacking process after sorting to CMP and 25
applying NMO.
2.8 Fold of coverage in seismic survey. 26
2.9 Midpoint in seismic survey. 27
2.10 A typical distribution of signal and noise. 28
2.11 Ground roll in 3D shot gather. 29
2.12 Head wave (refracted) and direct wave in seismic 30
data acquisition.
2.13 Direct (D) and refracted (R) waves in successive shot 31
gathers.
2.14 Seismic data acquisition. Travel paths of the primary 32
reflections (solid lines) and several multiple
reflections (dashed lines).
2.15 Primary and multiple reflection in a shot gather. The 32
multiples normally have a higher curvature than the
primaries.
2.16 Seismic data with air wave noise. 33
2.17 An example of back scattered wave in seismic data. 34
2.18 Random noise (spikes) in seismic record. 35
2.19 Work flow of AAA filter. 37
2.20 The Tau-p transform. 41
2.21 A wave travelling at an angle α to the vertical will 44
pass across an in-line spread of surface detectors at a
velocity of v/sina.

IV
2.22 An f–k plot for a typical shot gather containing 44
reflection events and different types of noise.
2.23 Comparing Events in time (t-x) domain and in 45
frequency wavenumber (f-k) domain.

2.24 Schematic illustration of pass-bands of different 52

filter designs in the frequency domain. (A) Low-pass,
(B) high-pass, (C) band-pass, and (D) notch filters. f1
and f2 represent cut-off frequencies.
3.1 Stacked section 45485 of Al-Samawa area used for 60
model building.
3.2 Nine homogeneous isotropic layers model that 61
simulate the stacked section of Al-Samawa area.
3.3 Shots number (1,60,120) acquired from the nine 64
layers earth model.
3.4 Applying gain to a shot gather (A) shot without gain, 66
(B) shot after applying gain (tpow=1), (C) shot after
applying gain (tpow=2), (D) shot after applying gain
(tpow=3).
3.5 Added random noise of signal / noise = 10% to a shot 67
gather.
3.6 Applying Sufxdecon on a shot gather: (1) shot before 69
filtering (2) after filtering (A) medium filtering (B)
severe filtering.
3.7 shot gather spectrum (A) before filtering (B) after 70
filtering.
3.8 Frequency-Offset spectrum for shot gather after 71
applying Sufxdecon.
3.9 applying Sufilter on a shot gather (A) before filtering 72
(B) after filtering (the passed data) (C) after filtering
(the rejected data).
3.10 Applying Suconv with (4,2,4,2) filter on a shot gather 73
(A) before convolution, (B) after convolution.
3.11 Noise attenuation on a shot gather (A) before noise 74
attenuation (B) after noise attenuation.
3.12 Applying noise attenuation on a shot gather (A) 77
before NMO (B) after NMO (C) after inverse NMO.
3.13 Fold values for the CMP gathers data. 79

V
3.14 Velocity analysis of cdp 60 where (1) is the 80
semblance plot for cdp 60 and (2) is the picked
velocities.
3.15 Stacked section after random noise attenuation. 81
3.16 Stacked section before and after noise attenuation, 82
(A) before removing noise and (B) after noise
removal.
3.17 A comparison between the 9 layers earth model and 83
the stacked section acquired from it.
4.1 The result of applying GSC on a raw shot gather, (A) 87
before compensation (B) after compensation.
4.2 The result of applying GSC on a stacked inline 87
section, (A) before compensation (B) after
compensation.
4.3 The final result of applying AAA on a shot: (A) 91
before applying the filter (B) after applying the filter
(C) the difference (the removed noise).
4.4 The final result of applying AAA on a shot: (A) 92
before applying the filter (B) after applying the filter
(C) the difference (the removed noise).
4.5 The final result of applying AAA on a stacked inline 93
section 45660: (A) before applying the filter (B) after
applying the filter (C) the difference (the removed
noise).
4.6 Procedure steps for TPLNA were (A) a noisy shot; 95
(B) applying NMO (C) transforming to the tau-p
domain; (D) muting or rejecting the noise manually;
(E) transforming back to the time domain; (F)
applying inverse (NMO) and (G) The removed linear
noise (A-F).
4.7 The result of applying Tau-P linear noise attenuation 97
filter (TPLNA) on a shot (A) before filtering (B) after
filtering (C) the difference
4.8 The result of applying Tau-P linear noise attenuation 98
filter (TPLNA) on an inline stacked section 41090
(A) before filtering (B) after filtering (C) the
difference.

VI
4.9 Frequency panels for the Ground-roll where the 100
range of frequencies in (A) is (0-120), (B) is (3-10),
(C) is (10-20), (D) is (20-30), (E) is (30-40), (F) is (40-
50), (G) is (50-60) and (H) is (60-70).
4.10 Linear velocity picking in which green line represent 101
the high pass velocity and blue line represent the low
pass velocity.
4.11 The result of applying FXCNS on a shot (A) before 103
filtering (B) after filtering (C) the difference.
4.12 The result of applying FXCNS on Stacked inline 104
section 45330 (A) before filtering (B) after filtering
(C) the difference.
4.13 Quality control of FXCNS filter on a shot (A) before 105
filtering (B) after filtering (C) the difference
(D) F-K spectrum for the difference.
4.14 The final result of applying AAA 2nd pass on a shot 110
(A) before filtering (B) after filtering (C) the
difference.
4.15 Procedure steps for F-K filter. Where (A) shot in 112
time domain (B) After applying NMO (C) After
transform the data to f-k domain and rejecting
linear noise (D) After transform the data to time
domain (E) After apply inverse NMO.
4.16 The result of applying F-K filter to a shot (A) Before 113
filtering (B) After filtering (C) The difference.
4.17 The result of applying F-K filter to a stacked inline 114
section 44145. Where (A) before filtering (B) after
filtering (C) the difference.
4.18 The result of applying RWD on a shot (A) before 117
filtering (B) after filtering (C) the difference.
4.19 The result of applying RWD on stacked inline 118
section 45565. Where: (A) before filtering (B) after
filtering (C) the difference.
4.20 Full noise attenuation on a shot gather (A) before 119
noise attenuation (B) after noise attenuation.
4.21 Full noise attenuation on staked inline section 45830. 120
Where: (A) before noise attenuation (B) after noise
attenuation.

VII
4.22 Signal to noise ratio spectrum (A) before noise 121
attenuation (B) after noise attenuation.
4.23 Time slice at 3500 MS in which: (A) before noise 122
attenuation and (B) after noise attenuation.

List of Tables
Table Page No.
1.1 Coordinates of 3D seismic survey of the study area. 2
2.1 Seismic Unix keys and their definitions. 48
3.1 The processing steps. 55
3.2 sfill variables. 62
3.3 Values of sfill in the model. 62
3.4 Acquisition parameters for the synthetic seismic data 63
survey.
3.5 Time-Velocity picking values. 76
4.1 The acquisition parameters. 84
4.2 work flow for noise attenuation of Al-Samawa 3D 85
seismic data.
4.3 AAA threshold factor parameters for pass 2 and 3. 89
4.4 Tau-P (Radon) transformation parameters. 96
st
4.5 FXCNS 1 pass parameters. 101
nd
4.6 FXCNS 2 pass parameters. 102
2nd
4.7 Threshold parameters for AAA filter, pass (1-5). 108
4.8 Radon (RWD) parameters. 115

VIII
Introduction
 CHAPTER ONE

1.1 Preface
Geophysical data processing is using of computers to analysis the geophysical
data. A major task in geophysics is to determine the constitution of the earth's interior
as much as possible, and because direct penetration is impractical or difficult.
Geophysical surveys are made, and the task now is to study and make inferences
from these surveys and measurments . By applying systematically, the laws of
physics and the principles of statistics (Claerbout. 1985).
The term (noise) used in seismology is generally applied to all types of disruption
that may interfere. Seismic noise is split into two primary types in exploration
seismology; coherent and incoherent noise, coherent noise on shot records are
seismic events generated by the source. They consist primarily of surface waves
(ground-roll) and air waves of relatively small bandwidth with low frequency range.
Unlike coherent noise, seismic events with unexpected amplitude and onset are
considered incoherent noise or random noise, which is basically of random nature
(Al-Sadi. 2017).
Random noise often coexists with coherent noise in seismic field data. Noise
attenuation plays an important role in data processing to improve the signal content
of the data. Coherent noise characteristics are significantly different from random
noise. A general strategy needs filtering methods based on noise characteristics, and
noise suppression must be applied individually for each type of noise. (Chiu. 2013).
Generally, part of the noise attenuation which generated from source is done in
the field. Then, a specific decision are taken in the field to enhance the signal-to-
noise ratio for seismic data, while the noise remains are done in the processing center
using certain methods such as filtering, muting, and synthesized traces (Al-
Adhami,1998).

1
 CHAPTER ONE

Coherent noise can be due to back scattering, multiples and near surface ghosting,
there are different methods to attenuate them such as different types of demultiples,
deconvolution (both prestack and poststack) (Stork et al., 2006). Random noise can
be reduced with different methods such as FX-deconvolution which depends on a
lot of parameters that can help to first recognize them and then remove or reduce
their effects (Larner et al., 1983). This thesis will focus on attenuation of the types
of noise found in the 3D seismic data of the study area.

1.2 Geographical Location and Coordinates of the Study Area:

Al-Samawa lies 280 km south of Baghdad. It’s the modern capital of the Al
Muthanna Governorate, which is located midway between Baghdad and Basra as
shown in figure (1.1), starting with inline 39800 and ending with inline 48690 with
an increment of (5). The coordinates in (U. T. M) for the 3D seismic survey of the
study area are shown in Table (1.1).

Table 1.1: Coordinates (U.T.M.) of 3D seismic survey of the study area

(Majeed 2016).
Corner X-Coordinate Y-Coordinate Inline Crossline
A 500000.00 3464000.00 39800 9250
B 534000.00 3423000.00 39800 17270
C 479000.00 3447000.00 48690 39895
D 511000.00 3406000.00 48690 46930

2
 CHAPTER ONE

Figure 1.1: Survey location for the study area.

1.3 Geology of the Study Area

1.3.1 Surface Geology
The study area and the surrounding area are entirely covered by sedimentary
rocks of the Cenozoic Era, ranging in age from Early Eocene up to Recent
(Quaternary sediments) figure (1.2). The area is a desert land covered mostly with
Eocene sediments (river sediments) which is consisting of silt and mud deposit
(Euphrates Formation deposit of Lower Miocene). As well as limestone with thin
layers of mud, sand and gypsum (Al-Damma Formation deposit of Upper Eocene)
(Sissakian et al, 1997).

3
 CHAPTER ONE

Figure 1.2: The geological map of the study area.

1.3.2 Subsurface Geology

Generally, in order to identify the subsurface layers and Formations, their ages,
lithological description and sequences that is forming the geological column of Al-
Samawah area. It requires relying on the available information of Al-Samawah well-
1 (Sa-1) figure (1.3), which represent the geological column for the study area.
Which is containing many geological Formations starting from Al-Dammam
Formation (Eocene) at depth of 120 m and ending with Al-Batmah Formation
(Lower Jurassic) at depth of 3839 m (O. E. C, 2015).

4
 CHAPTER ONE

Figure 1.3: The geological column of Al-Samawa well-1(Sa-1) after (O. E. C, 2015).

5
 CHAPTER ONE

1.4 Tectonic Setting of the Study Area

The study area lies within the eastern border of the stable shelf (Salman zone),
which represents the boundary between the Southern Desert (stable shelf) and the
scope of Mesopotamia (unstable shelf) figure (1.4). In general, the area affected by
the three major groups of faults which have the direction (N-S) belonging to the
Early Cambrian period and group of faults oriented (NE-SW) belonging to late
Cambrian period and these two groups are of deep faults which could reach the
basement rocks and continues to influence ongoing due to movement resulting from
the Arabian plate. While the third group (NW-SE), which belongs to Tertiary period
are the most important faults because of their clear impact on earth surface. The
faults in the study area were formed as a consequence of vertical block movement
(Buday and Jassim, 1987).

Figure 1.4: Tectonic map of the study area.

6
 CHAPTER ONE

1.5 The Processing Systems:

Two types of data will be processed:

- Synthetic data generated and processed using open source Seismic Unix
processing system.
-
- Real data processed by Schlumberger Omega (2015) seismic data
processing software which is a flexible, scalable system that allows for
processing and imaging of a single workstation up to massive compute
clusters, or from a single 2D line to an immense 3D seismic survey.

1.6 Previous Seismic Survey of the Study Area

 The first seismic survey was done by Basra petroleum company (BPC)
between (1957 – 1958), with a single coverage recording, in which one of its
important result is the existing of two enclosure geological structures in the
area, (O.E.C, 2015).
 The survey of Al-samawa – Shawya area that was carried out between (1975
– 1976) by the third seismic crew of the Iraqi oil exploration company,
(O.E.C, 2015). Which resulted in two geological structures named Al-
Samawah structures, the first one Al-Samawah west structure and the other
Al-Samawah south structure which containing on one or more enclosures.

7
 CHAPTER ONE

1.7 Previous Studies Outside of Iraq:

1.7.1 Previous Studies Using the 3-D Seismic Reflection Survey & Processing
Technique
 Corsmit et al, (1988), performed acquisition, processing and interpretation of
high-resolution 3- D reflection seismic on a tidal flat: Their small-scale survey
was conducted on a tidal flat in the Netherlands.
 Martinez et al, (2000), used method based on 3D Weighted Slant Stacks (3D
WSS). The technique exploits the concept of 3D local slant stacks to accomplish
the noise attenuation. The 3D WSS algorithm is robust in the presence of different
types of noise, making it attractive over other noise attenuation methods.
 Rajput and Rajput, (2006), applied noise attenuation techniques using frequency‐
shot‐receiver (f‐x‐y) prediction filters for detection of noise which is applied in
f‐x domain to the noisy frequencies of the shots.
 Henley, (2007), applied a radial trace filtering method on 3D seismic data,
attempting to attenuate coherent noise. The method is based on an interpolated
mapping of the original data traces to the radial trace domain. He found that
receiver line gathers provide the best sampling of coherent noise, and provide the
best domain for attenuating this noise.
 Al-Dossary and Wang, (2009), proposed a new filter to estimate incoherent noise
in three dimensional (3D) post‐stack seismic data. The resultant noise estimate
can be used to assess data quality for seismic interpretation and the spatial
distribution of noise.
 Yu, (2012), used multidomain 3D spatial filter that was designed in the τ-ρ-y
domain to attenuate strong seismic interference noise. This technique has been
applied effectively and extensively over more than 60,000 km2 Gulf of Mexico
data with successful results.

8
 CHAPTER ONE

 Roizman, (2013), described a method to reduce pre-stack noise in the domain of

Common Offset Vectors (COV) applied to 3D land seismic data of orthogonal
geometry and wide azimuth.
 Liu, (2013), proposed a novel method for 3D seismic random noise attenuation
by applying noncausal regularized nonstationary autoregression (NRNA) in f–x–
y domain. The key idea of his paper is to consider that the central trace can be
predicted by all around this trace from all directions in 3D seismic cube.
 Huang et al, (2016), applied Multichannel singular spectrum analysis (MSSA) to
attenuate random noise in 3D seismic data. He derived a new formula of low-
rank reduction, which is more powerful in distinguishing between signal and
noise compared with traditional TSVD.
 Anvari et al, (2017), used a new version of Tensor principal component analysis
(TPCA) for denoising 3D seismic data in which, tensor data split into a number
of blocks of the same size. And applied the method on synthetic and real 3D
seismic data.

1.7.2 Previous Study Using Seismic Unix in Seismic Data Processing

 Bording, (1995), used Seismic Unix to test the different geological environments
by creating models (forward method) of them and calculate their seismic
response with comparative of inversion method.
 Galway, (2000), used Seismic Unix to process marine seismic data in order to
combine it with sonar data of the area by using the GIS system. He finds that this
increase the performance of interpretation and facilitate the storage of the data in
interactive environment.

9
 CHAPTER ONE

 Forel et al, (2005), used Seismic Unix to develop processing techniques for
inverse and foreword seismic data. The book is very interesting and has many
shell scripts implement seismic data modeling and inversion.
 Soupios and Toth, (2005), used Seismic Unix to process shallow seismic
reflection marine data in order to get information about the stratigraphy of the
sedimentary layers.
 Mukherjee and Banerjee, (2006), used Seismic Unix to create synthetic seismic
data in order to test and compare different migration techniques. They find that
Fast Fourier Finite Difference as well as Finite Difference techniques provide
better imaging of the reflectors as compared to f-k band Gazdag and Phase shift
migration.
 Öhman et al, (2006), used Seismic Unix two-dimensional reflection seismic
processing methods to refraction seismic data obtained from Finland to locate the
ideal sealing trap for nuclear fuel disposal and to ensure no faults occur in it.
 Schleicher (2012) us Seismic Unix to reprocess 2D line data obtained from
Alaska. The 2D line data were already processed by commercial software. He
found that the results are similar to the results obtained in 1981 using GSI's
(Geophysical Service Incorporated Company) proprietary software.
 Torres and Goyes (2013) use Seismic Unix to introduce seismic forward
modeling as a powerful tool to investigate the seismic wave propagation in
structural geologic cross section. He finds events with high dips generate pitfalls
in seismic sections, these can be corrected with depth migration and velocity
model better.
 Lorenzo, etal., (2014) use Seismic Unix to make velocity analysis of seismic
refraction data and compare the result with depth model of flood protection.

10
 CHAPTER ONE

 Yagci (2016) use Seismic Unix to reprocess multi-channel land seismic data in
order to enhance signal to noise ratio and determining the correct processing
techniques and parameters. He finds that butterworth minimum phase filter was
good to solved the low amplitude, spiking deconvolution and the prediction error
filter recovered the low amplitudes on velocity semblances and increased the
accuracy of velocity picking.

1.8 Previous Studies in Iraq:

1.8.1 Previous Studies Using the 3-D Seismic Reflection Survey & Processing
Technique

 Al-Sadi, (1992), 3-D Seismic Survey. He briefly described the 3-D seismic
technique (definition, the need of the 3-D survey, data acquisition, and data
processing).
 Awaina, (1996), used the 3-D seismic reflection data for Kifl oil field to study
the Zubair Formation and how the oil is trapped in it.
 O.E.C, (2009), report on 3-D seismic reflection survey, quality control and
field processing of Gharraf field which was carried out by the Iraqi seismic
crew-02. It showed the very good quality for field shot and continuity for
reflectors in brute stack using Focus as field processing system.
 Fadhil, (2010), used the 3-D seismic reflection data for Kifl oil field to study
the seismic stratigraphy and structure of the lower Cretaceous Formations,
and to build depositional model.
 O.E.C. (2010), (2011b) & (2012a), reports on 3-D seismic survey, quality
control and field processing of Nasiriya oil field in 2010, Badra oil field in
2011 and Margan-West Kifl in 2012 which were carried out by the Iraqi

11
 CHAPTER ONE

seismic crew-1. They showed the importance of field processing system

(Geocluster & Geoland) in quality control and field processing.
 O.E.C. (2011c), (2012b) & (2013), reports on 3-D seismic survey, quality
control and field processing of Akkas field in 2011, Mansuria field in 2012
and Anah-Graben in 2013 which were carried out by the Iraqi seismic crew-
03.They showed the importance of field processing system (Geocluster &
Geoland) in quality control and field processing.
 Muhsin, (2012), used 3-D Seismic reflection data for Balad oil field to study
the seismic structural of subsurface and to build geophysical and geological
models as a resultant of 3-D seismic interpretation.

1.8.2 Previous Studies Concerning Signal to Noise Enhancement

 Al-Adhami, (1998), He showed the importance of the utilization of analysis and

figures to ensure best S/N ratio, using analytical study of some seismic noise tests
at selected area in Iraq.
 Kareem, (2014), they used thirteen kinds of filters to process the field data in
Amara oil field in south- eastern Iraq to imprrove the signal to noise ratio some
of these filters are anti-alias filter to avoid the aliasing during resampling from 2
to 4 MS, amplitude equalization to recover the seismic data to their original
shape, the f-k filter to remove the surface waves which have low apparent
velocity.
 Aswad, (2016), performed A detailed seismic processing study using (2D)
seismic data of west Luhais area, for the aim of enhancing the seismic sections.
The treatment processes that proceeded is Reformate, geometry update,

12
 CHAPTER ONE

exponential gain, spherical divergent and many techniques to remove noise such
as Anomalous Amplitude Noise Attenuation (AAA), F-X Coherent Noise
Suppression (FXCN) and F-K filters, using Omega Schlumberger seismic data
processing software. While the processes that followed static correction process
include muting, sorting information, deconvolution, normal move out and
migration.
 Ibrahime, (2019), applied different filters to enhance signal to noise ratio of 2D
seismic lines of Al-Diwanya area in southeastern Iraq, using the Geovation
system to read and process the field data. Such as Frequency Dependent Noise
Attenuation filter (FDNAT), Adaptive Ground Roll Noise Attenuation filter
(AGORA), the Radial Trace Mix filter (RADMX) and the Dip Dependent
Median filter (DDMED). Resulting a very good signal to noise enhancement.

1.8.3 Previous Study Using Seismic Unix in Seismic Data Processing

 Treky, (2018), used open software Seismic Unix (SU) application to process the
data of ground penetrating radar (GPR). And concluded that the Seismic Unix
had a high flexibility in the conversion and the preparing for the futuristic
processing of the non-seismic data like GPR data.

13
 CHAPTER ONE

1.9 Aims of the Study

Synthetic data:

Enhance signal to noise ratio by applying different filters and techniques on

the synthetic seismic data using Seismic Unix processing system. And
showing the noise attenuation influence on the synthetic data before and after
stacking.

3D real data:
Enhance signal to noise ratio by applying different filters and techniques to
attenuate random and coherent noise on the 3D seismic data of Al-Samawah
area using Omega Schlumberger processing system. And showing the noise
attenuation influence on the 3D seismic data before and after stacking.

1.10 Procedure of Achievement of the Study:

Synthetic data:

1- Build a geological model based on stacked section of the 3D data of Al-

Samawa area.
2- Generate seismic data from the previously built model.
3- Add ratio of random noise to the generated seismic data.
4- Use different filters to remove the random noise and enhancing signal to noise
ratio.

14
 CHAPTER ONE

3D real data:

1- Upload the raw 3D seismic data in (SEGY) format to reformat it in the

internal format of Omega software.
2- Geometry building for the shot and receiver of the 3D survey.
3- Apply gain application to compensate the lost energy.
4- Test and apply of random noise filters on shot gather data.
5- Finalize the random noise filtering by checking on stack data.
6- Test and apply coherent noise filters on shot gather data.
7- Final checking of coherent noise filtering results on stack data.

15
Theoretical
Background
 CHAPTER TWO

2.1 Preface
In most seismic reflection survey, the mechanically produced low-energy
acoustic waves are directed to the subsurface. The energy travels through the
weathering zone and into the rock layers. Due to acoustic impedance (density x P-
velocity) at each interface, some part of energy is reflected back to the surface from
different layers of rocks beneath. Contrasts usually occur when the physical
properties of rock layers change. Sensitive instruments called seismic receivers
capture the upcoming waves. They record the amplitude of the waves and time
(Two- Way-Travel time TWT), which is travel through the rock layers and back to
the surface. These recordings are then taken to various adjustments done by
computers, and the data are then transformed into visual images that give a picture
of what the subsurface looks like beneath the survey area. So, the seismic survey is
used to get a picture of the structure and the nature of the rock layers in the
subsurface indirectly (Bacon et al., 2003).

Before it can be interpreted for the subsurface structure, several processing steps
must be taken through the reflecting seismic data. Many factors affect the source
signal on its way down and back to the receivers. The purpose of the processing is
to correct these effects as much as possible, leaving only interest data to be
interpreted (Arora et.al 2011).

Signal and a noise component are always found in seismic. As a general definition
we can say that any recorded energy which interferes with the desired signal can be
considered as noise. The variety of noise kinds often makes it a difficult process to
separate signal and noise. For high-quality imaging, effective noise attenuation
and/or removal is essential. From an industrial point of view, it is also desirable that
de-noising algorithms should work on many types of similar noise without the need
for time consuming parameter adjustments (Elboth et.al 2008).

16
 CHAPTER TWO

Consequently, the seismic exploration consists of three main stages: data

acquisition, processing, and interpretation. This chapter focuses on the 3D noise
attenuation which is part of the processing step, as well as locking into different type
of noise in 3D seismic data and how to attenuate it using different filters and
techniques.

2.2 Spectral Analysis

Spectral analysis used with time series is one of the most widely used methods
for data analysis in geophysics, oceanography, atmospheric science, astronomy,
engineering (all types). Time series described through spectral analysis by
comparing them to sines and cosines (Percival 2000).

2.3 Fourier Transform

The Fourier transformation is essential for analyzing seismic data. It applies to

almost all processing phases. A seismic trace is a seismic wavefield recorded at the
place of the receiver. The digital form of a seismic trace is a time series that can be
described completely as a discrete sum of several sinusoids (Yilmaz 2001).

Fourier transform is one of the most commonly used techniques in signal

processing and control theory. It allows a frequency content (spectral) analysis of a
signal and is suitable for periodic signals (Hlavác 2018). The principle of Fourier
says that each signal can be described by a sum of Sin and Cos functions. The Fourier
transformation transforms a time queue from time domain (amplitude as function of
time) to the frequency domain (amplitude as function of frequency) as shown in
figure (2.1), through the following equations: (Yilmaz 2001).

17
 CHAPTER TWO

+∞
Fourier transformation: G(f) = ∫−∞ g(t)e−i2πft . 𝑑𝑡 … . … … … … … . (2 − 1)
+∞
Back transformation: g(t) = ∫−∞ G(f)e−i2πft . 𝑑𝑡 … … . … … … … . (2 − 2)

Where G(f) is the function in the frequency domain and g(t) is the function in time
domain.

Figure 2.1: An example of Fourier transformation after (Dobrin and savit 1988).

2.4 Sampling Rate

The continuous (analog) seismic signal in digital recording is sampled in time at

a fixed rate, called the sampling interval (or sampling rate). Typical sampling
interval values range from 1 to 4 ms for most seismic reflections. Sampling intervals
as small as 0.25 Ms is required for High-resolution studies, the reconstructed signal
may lack the details present in the original analog signal. These details correspond
to parts of high frequency that were lost through sampling. If a smaller sampling
interval was selected, the reconstructed signal would describe the initial signal more
accurately (Yilmaz 2001).

18
 CHAPTER TWO

2.5 Seismic Data Processing Steps

Seismic data must pass through several processing steps before the interpretation
stage figure (2.2).

Figure 2.2: General seismic data processing sequence (Seismic discussion).

Below is a definition of some of the processing steps that were used in this thesis.

2.5.1 Reformatting
This process takes the data coming from the receivers and puts them into trace
order. Normally the data are written to tape at this stage in one of the designated
industry formats so that the raw records are retained to form the basis of possible
later reprocessing. If starting from field tapes, reformatting includes converting the
data from standard industry format into whatever format the processing system uses
(Bacon et all 2003).

19
 CHAPTER TWO

2.5.2 Geometry Definition

Applying the field geometry to the seismic data is an important step in the
preprocessing. Application of the geometry assigns all the positional information to
the data, and relationships between shots and receivers. This therefore defines the
fold and offsets etc. The geometry is acquired from the observer’s log, and The field
geometry should be incorporated with the seismic traces. Before this was done
manually, but nowadays it is done by the processing software’s. The procedure
called Labeling or Merging (Lee, et al., 1999).

2.5.3 Spherical Diversion Compensation

A field record is a wavefield produced by a single shot. Conceptually, a single

shot is thought of as a point source that generates a spherical wavefield. The earth
affects a propagating wavefield in which the energy density decreases in a
homogeneous medium in proportion to 1/r2, where r is the wavefront radius. Wave
amplitude is proportional to the energy density square root; it decays as 1/r (Yilmaz
2001).

2.5.4 Noise Attenuation

Noise attenuation is an important step in seismic data processing, and must be held
in a way so that at the end of the acquisition and processing sequence, the desired
signal can be reliably interpreted and the noise is suppressed as much as possible.
There are main types of noise such as multiples, ground roll and scattered energy.
How much low-velocity noise can be suppressed depends on the choice of field
arrays, the stack response, filtering and on various processing steps, (Vermeer 2002)

2.5.5 CMP Sorting

The acquisition of Seismic data with multifold coverage is usually done in shot-
receiver (s, g) coordinates. Figure (2.3 a) describes the recording geometry and ray

20
 CHAPTER TWO

paths that is associated with a flat reflector. On the base of the field geometry
information, every single trace is located to the midpoint of the distance between the
shot and receiver locations. Those traces that have the same midpoint location are
summed together, making up a CMP gather. Figure (2.3 b) shows the CMP gather
and raypaths that is associated with a flat reflector (Yilmaz, 2001).

Figure 2.3: (a) Seismic data in shot-receiver (s, g) coordinates. (b) Seismic data
processing in midpoint-offset coordinates after (Yilmaz, 2001).

2.5.6 Velocity Analysis

Velocity analysis is the calculation of stacking or NMO velocity from
measurements of normal moveout. It generally deals with finding the best velocity
associated with the best-fit hyperbola to common-midpoint data. However, even in
the absence of noise and errors, time-offset data is not hyperbolic except in the case
of constant velocity, and the stacking velocity value is often somewhat dependent
on the amount of data included in the analysis. The stacking velocity is roughly the
rms velocity when all reflectors are horizontal and when velocity varies only with
depth. Velocity analyses are usually more than 3–10 CMP on average and may
require selecting about 50 t-x pairs (Sheriff 2002).

21
 CHAPTER TWO

Velocity spectrum analysis is generally the most used technique; the velocity
spectrum is acquired when the stacking results for a range of velocities are plotted
side by side on a velocity plane versus two-way travel time in a panel for each
velocity. (Taner and Koehler 1969). Using the semblance is one of the most used
methods to determine a velocity spectrum. The semblance measure is an essential
tool for seismic record velocity analysis. Conventional seismic velocity analysis
scans various values of effective moveout velocity, Calculates the semblance of
flattened gathers and produces velocity spectrum for subsequent velocity picking
(Yilmaz, 2001). Figure (2.4) (A) shows a synthetic CMP gathers and (B) shows
velocity analysis panels using conventional semblance scans and the corresponding
automatic velocity picks.

A B

Figure 2.4: (A) synthetic CMP gathers (B) semblance scans for CMP gathers. Black
curves indicate velocity picks after (Fomel 2009).

22
 CHAPTER TWO

2.5.7 NMO-Correction
In reflection seismology, the normal moveout (NMO) is describing the distance
between the source and the receiver and how it affect the arrival time of the signal
from the reflector in the form of an increasing the time with offset figure (2.5). The
relationship between the offset and the arrival time is hyperbolic which is an
important principal criterion that the geophysicist uses to decide whether a specific
event is a reflection or not (Sheriff and Geldart 1995). The normal moveout (NMO)
is usually depending on a complex combination of factors which include the velocity
that is above the reflector, the offset, the dip of the reflector and the azimuth of the
source receiver in relation to dip of the reflector (Yilmaz 2001). For a flat and a
horizontal reflector, the travel time equation is:

𝑥2
𝑡 2 = 𝑡02 + … …. … … … (2 – 3)
𝑣2

Where x = offset.
v = velocity of the medium above the interference.
t0 = time travel at the zero offset.

Figure 2.5: Seismic data that is sorted by CMP then corrected by NMO (Seg.Wiki).

23
 CHAPTER TWO

The velocity that flattens the best of the reflection hyperbola is the best velocity
that corrects for NMO before traces is stacked figure (2.6). Furthermore, for a case
of a single horizontal reflector, the best velocity is equal to the velocity of the
medium above the reflector (Yilmaz 2001).

Figure 2.6: (a) CMP gather of single event with a 2264 m/s velocity (b) NMO
corrected gather by using the best velocity (c) over corrected gather by using low
velocity 2000 m/s (d) under corrected gather using high velocity 2500 m/s, after
(Yilmaz 2001).

2.5.8 Stacking
The traces in a CMP can be corrected for NMO Once the velocity is known, in
order to correct each trace to the equivalent of a zero-offset trace. These traces will
have the same reflection pulses at the same times, but with different random and
coherent noise. Combining all these traces in a CMP together will decrease the noise,
and will increase the signal-to-noise ratio (SNR). This process is termed stacking
figure (2.7) (Kearry et.all 2002).

24
 CHAPTER TWO

A conventional stacking, which is performed by averaging a migrated or NMO-

corrected data set, is used only when the noise components in all traces are
uncorrelated, normally distributed, and of equal magnitude (Liu et al. 2009).
Therefore, a different stacking methods have been proposed, along with an
improvement in optimizing the weights of seismic traces.

Figure 2.7: The stacking process after sorting to CMP and applying NMO.
(Schlumberger Oilfield Glossary).

2.6 Noise in Seismic Data

Generally, before attempt to solve any problem it must be defined what is signal
(S) and what is noise (N) before improving S/N.
Noise is defined by Russell et al. (1990) as “anything on the seismic data that
does not fit our conceptual model of the data, that is, as clean seismic reflections”.
The term signal is used by Sheriff and Geldart (1995) to refer to any event on the
seismic record from which an information about geological structure is obtained,
everything else is noise.

25
 CHAPTER TWO

2.6.1 Signal to Noise Ratio (S/N)

The signal-to-noise ratio, abbreviated S/N, is the ratio of the signal energy to a
specified portion of the record to the total noise energy in the same portion.
Whenever the signal-to-noise ratio is small, poor records are resulted. (Telford et al,
1990).

Different factors can affect S/N ratio, such as the fold of coverage which is: the
multiplicity of common-midpoint data or the number of midpoints per bin, therefore
the fold varies from bin to bin and for different offsets (figure 2.8) (Sheriff, 2002).
Fold Taper: is the area around a 3-D survey in which the fold increases from zero to
full-fold needs to build up full fold (Vermeer, 2002). Fold controls the signal-to-
noise ratio (S/N), if the fold is doubled, a 41% increase in S/N is accomplished.
Doubling the S/N ratio requires quadrupling the fold, assuming that the noise is
distributed in a random fashion (incoherent noise) (Cordsen et al., 2000).

Fold = Constant × (S/N) 2 …. …. …. …. … (2 – 4)

Figure 2.8: Fold of coverage in seismic survey after (Bacon et al., 2003).

26
 CHAPTER TWO

Another factor affecting the S/N ratio is the bin, which is: a small rectangular or
square area that usually has the dimensions (SPI/2) x (RPI / 2) (figure (2.9). All
midpoints that lie inside this area, or bin, are assumed to belong to the same common
midpoint (Vermeer, 2002). Bin size is controlled by the spatial resolution required
for reservoir definition, source and receiver points spacing. Smaller bin size
increases the resolution and the cost. The S/N is directly proportional to the length
of one side of the bin (Stone, 1998).

S/N= Constant x bin size … … … … … (2 – 5)

Figure 2.9: Midpoint in seismic survey after (Chaouch and Mari 2006).

2.6.2 Types of Seismic Noise

The noise can be classified according to nature in two main groups: Coherent
and incoherent (random) noises, (Elboth et al., 2008).

These two types of noise can often be discriminated in the frequency domain.
Coherent noise like ground roll is low frequency whereas random noise is can be
distributed in random frequencies (1-120), as shown in figure (2.10) (Li and Tang,
2005).

27
 CHAPTER TWO

Figure 2.10: A typical distribution of signal and noise after Al-Sadi 2017.

Random noise does not correlate either with the neighboring channels or along
the same channel. Coherent noise, however, is part of the data that correlates
spatially and/or temporally. The possible sources of these seismic noises can be
placed under four categories: (1) ambient sources, (2) wave propagation related
noise, (3) data acquisition related noise, and (4) data processing artifacts. The
severity and types of noise can differ between marine and land acquisition
environment. (Arora et.al 2011).

2.6.2.1 Coherent Noise

According to (Telford et al, 1990), coherent noise can be divided into:

a. Noise that travels essentially horizontally and is repeatable.

b. Noise that reaches the spread more or less vertically.

28
 CHAPTER TWO

There are three properties that characterize coherent noise which is: coherency,
travel direction, and repeatability.

Different type of noise can be within the coherent noise category such as:

(A) Ground-roll
Surface waves are kind of seismic waves that exist at the boundary of the
propagating medium. One kind of the surface waves called the (Rayleigh Wave). It
exhibits a retrograde elliptical particle motion. The Rayleigh wave is usually
recorded on seismic records taken on land. It is then usually called (ground-roll), an
example of the ground roll in the seismic data is shown in figure (2.11). The ground-
roll is characterized by relatively low frequency, low velocity and high amplitude,
that usually tends to cover part of the desired reflection signals. Different methods
such as stacking, f-k filtering and frequency filtering are used to reduce it. (Gadallah
and Fisher 2008).

Figure 2.11: Ground roll in 3D shot gather (Seg.wiki).

29
 CHAPTER TWO

(B) Direct and Refracted Waves

A direct wave travels directly along surface of the earth’s figure (2.12). It is
important to keep in mind that the direct wave is not a surface wave but a body wave
that travels parallel to and just below the surface of the top layer (Lowrie, 2007).

Refractions or head waves travel horizontally in the subsurface sediments and reach
the receivers after they are refracted figure (2.12). They travel with the velocity of
subsurface sediment units and hence they always have higher velocity and lower dip
than direct waves on the shot records (Dondurur 2018). A shot gather with direct and
refracted waves is shown figure (2.13).

Figure 2.12: Head wave (refracted) and direct wave in seismic data acquisition after
(Dondurur 2018).

30
 CHAPTER TWO

Figure 2.13: Direct (D) and refracted (R) waves in successive shot gathers, after
(Dondurur 2018).

(C) Multiple Reflections

A wave spreading through an arbitrary inhomogeneous medium will be scattered
and thus a scattered wavefield will be produced. If the wavefield is only scattered
once between source and receiver or multiply on a small scale, the arrived wave is
the primary part of the wavefield. Multiple reflections are formed once this primary
wave field is scattered from a second scatterer, which is at a distance larger than a
wavelength. Figure (2.14) shows the multiple reflection in seismic data acquisition,
(Shapiro and Hubral 1999).

In seismic data multiple reflections are usually regarded as unwanted noise that
often seriously disturbs the accurate mapping of the underground geology in search
of oil and gas reservoirs. In seismic exploration, the problem of multiple reflections
is contaminating the seismograms and thus disguising important information about
subsurface reflectors, (Essenreiter et al 1999). Figure (2.15) shows both primary and
multiple arrivals in seismic data.

31
 CHAPTER TWO

Figure 2.14: Seismic data acquisition. Travel paths of the primary reflections (solid
lines) and several multiple reflections (dashed lines) after (Essenreiter 1999).

Primaries

Multiples

Figure 2.15: Primary and multiple reflection in a shot gather. The multiples normally
have a higher curvature than the primaries after (Fomel 2009).

32
 CHAPTER TWO

(D) Air Waves

Air blasts, ground-coupled air waves, and air-coupled ground are all potential
problems in recording seismic signals. The reflection signals of the subsurface can
be certainly overwhelmed by these types of noise, especially when the noise source
is near the receivers. In addition, unacceptably high noise levels can be created if the
winds are above certain speed, (Stewart 1998). The air wave at a speed of 300 m / s
can be a severe issue when shooting with surface charges such as Geoflex, Poulter
or ground air gun, (Yilmaz 2001). Figure (2.16) shows an example of an air wave
on the seismic data.

Air wave

Figure 2.16: Seismic data with air wave noise, after (Dragoset 2005).

(E) Backscattering noise

Source-generated coherent noise can contaminate Seismic reflection surveys and
limits the interpretation and identification of subbasement reflections. Coherent
noise arrivals are produced by scattering from the ground around the seismic line.
Back-scattered surface waves are horizontally travelling surface waves that scattered
and travelling backward along the receiver line, what makes a distinctive arrival
33
 CHAPTER TWO

pattern of opposing slope compared to those of forward propagation that travels

directly from the source. Those types of arrivals are considered as a noise and
conventional common midpoint (CMP) stacking can enhanced it, (Calvert 1997).
Figure (2.17) shows an example of back of backscattered wave in a seismic data.

Figure 2.17: An example of back scattered wave in seismic data, (parkSEIS).

2.6.2.2 Random Noise

Random noise is any noise spikes or bursts which are not coherent in nature. This
type of noise is having a different characteristic from the surrounding data in
frequency and amplitude, (Dingus 2010). In oppose to coherent noise Random noise
cannot be correlated from trace to trace and can often be suppressed by repeated
measurement, averaging and by stacking, (Kearry et.al 2002). Random noise can be
generated in the area where seismic acquisition work is being carried out, by the
activities in the environment. In a land acquisition, random noise can be created by
the people working in the survey area, acquisition truck, wind, vehicles, electrical

34
 CHAPTER TWO

power lines and animal movement. This noise appears in a seismic record as spikes
figure (2.18) (Onajite 2014).

Figure 2.18: Random noise (spikes) in seismic record after (Onajite 2014).
2.7 3D Data Filtering
2.7.1 Anomalous Amplitude Attenuation (AAA) Filter
The anomalous amplitude attenuation filter can be used to attenuate different
type of noise modes, including marine swell noise, ground roll, wind noise,
biological noise, near-surface scatterers, and cultural noise. It is applicable to all
types of seismic data, whether land or marine, legacy or new acquisition. The filter
is applied in both CMP/CDP and shot gather in frequency domain, (Muchlis 2015).

Noise characterized by anomalous amplitudes is attenuated from prestack seismic

data by transforming the seismic data into the frequency domain and then applying

35
 CHAPTER TWO

a spatial median filter, in which frequency bands with amplitudes that deviate from
the amplitude of the median filter threshold are either scaled (multiplied by a
specified scale factor) or replaced with an interpolated band using neighboring
traces. A frequency band is considered anomalous when its mean amplitude exceeds
or falls below a computed threshold, the threshold is computed by multiplying the
median amplitude of frequency bands from specified number of surrounding traces
by a specified factor, (Xie et.al 2015).

The mean amplitude of a window is defined by: (Western Geco 2015a).

1 𝑛𝑓𝑡𝑘
𝐸𝑓𝑡𝑘 = ∑𝑖=1 𝐴𝑖𝑓𝑡𝑘 … … … … (2 – 6)
𝑛𝑓𝑡𝑘

Where:
𝐸𝑓𝑡𝑘 = Amplitude (energy) of frequency band f of window t of trace k
𝐴𝑖𝑓𝑡𝑘 = Amplitude of sample i in frequency band f of window t of trace k
𝑛𝑓𝑡𝑘 = number of samples in frequency bands f window t of trace k

Anomalous frequency bands can be either attenuated or replaced with

interpolated data. The noisy frequency bands are replaced by a weighted average of
the real and imaginary complex frequencies. The average is computed using the
closet traces that are not anomalous, one on each side of the anomalous trace. The
default value of the traces that is computed is 3 traces, if two non-anomalous traces
cannot be found within this number of traces, one on each side of the noise trace, the
trace cannot be interpolated and is attenuated, therefore any anomalous frequency
bands in the first or last trace of the gather is always attenuated rather than
interpolated, (Western Geco 2015a).

36
 CHAPTER TWO

After transforming the data into the frequency domain, the median filter will
divide the shots into several windows and every window will be divided into several
frequency bands, then applying the specified threshold on every frequency band
separately in which any amplitude above the specified threshold will be attenuated.
There are two different types of threshold can be defined, either a Percentile in which
the value of threshold will be applied to all windows, or can be defined by median
deviation in different values of threshold will be applied to different windows or
different time in which usually the threshold factor will decrease with time
increasing. Figure (2.19) shows the work flow of AAA filter.

Figure 2.19: Work flow of AAA filter.

37
 CHAPTER TWO

A median absolute deviation threshold is computed through: (Western Geco 2015a).

𝑀𝐴𝐷𝑓𝑡𝑘 = 𝑚𝑒𝑑𝑘 [| 𝐸𝑓𝑡𝑘 − 𝑚𝑒𝑑𝑘 ( 𝐸𝑓𝑡𝑘 )|] …. …. …. (2 – 7)

Where:

𝑀𝐴𝐷𝑓𝑡𝑘 = Median absolute deviation for frequency band f of window t of trace k

𝑚𝑒𝑑𝑘 = median operation where median is taken of amplitudes from all traces k
within the median width for frequency band f of window t

2.7.2 Tau-p Linear Noise Attenuation (TPLNA) Filter

The tau-p transform is a powerful processing tool to preserve uniqueness of the
seismic data as it affords an increased separation between different seismic waves
(Basak et.al 2012)

The filter can be used for noise suppression on 3D data, by transforming the input
wave field of poststack seismic trace into the tau-p domain. Only the energy from
events within the defined range of dips is transformed to the τ-p domain. Events
outside the range of user specified moveouts are eliminated by inverse transforming
the data from the tau-p domain to the t-x domain, (Western Geco 2015b).

Tau-p transform is an attempt to preserve the seismic data characteristics in the

wavefield. A seismic section in the τ -p domain can offer a different view in which
all subsurface reflectors are illuminated by incident energy of a fixed ray parameter.
Working in the τ-p domain gives advantage so that we can study the different wave
modes as function of their corresponding slowness values (p=1/v), where v is the
propagation velocity. The τ-p transform is a useful processing tool because it

38
 CHAPTER TWO

provides an increased separation between different seismic waves (i.e., ground-roll,

multiples, P and S waves among others), (Donati and Martin 1995).

The radon transform is performed for any of the parabolic, hyperbolic or linear
events, and in the case of (TPLNA) the radon transform will be used to transform
only the linear events. So, the τ-p transform is a special case of radon transform,
where the data are decomposed as a series of straight lines which will be mapped
into a points in the τ -p domain, (Basak et.al 2012). The radon transform makes a set
of p-traces in the τ-p domain by applying parabolic or linear moveout through the
following equations:

Linear moveout: (Western Geco 2015c).

−𝑚
∆𝑇𝑥,𝜌 = 𝑝𝑋 ; 𝑝 = … …. …. …. (2 – 8)
𝑋𝑚𝑎𝑥

Parabolic moveout: (Western Geco 2015c).

-m
∆Tx,ρ = pX 2 ;p= …. … …. …. (2 – 9)
Xmax2

Where:
∆𝑇𝑥,𝜌 = Moveout at offset x for a given p trace
X = Offset distance
Xmax = Max offset parameter
M = moveout at offset Xmax

The coordinate in the tau-p domain are zero-offset time (tau) and moveout
applied to the far offset trace (p). A p-trace is created by summing across constant
time paths all the traces in a t-x domain panel of traces. The traces in the panel are

39
 CHAPTER TWO

statically time shafted by an amount dependent on the shift trajectory (TI), the first
p-trace has a p-value of (TI). The next p-trace is obtained by summing across traces
that have different shift trajectory (TI). (Western Geco 2015b). The shift trajectory
(TI) is computed from: (Western Geco 2015b).
𝑀𝐴𝑋𝑝 −𝑀𝐼𝑁𝑝
TI = … …. …. …. (2 – 10)
𝑁𝑃−1

Where;
MAXp is the maximum p-trace
MINp is the minimum p-trace
NP is the number of p-traces.
The number of P-traces is calculated using the following equation:
𝑛𝑢𝑚𝑝 = (𝑀𝐴𝑋𝑝 − 𝑀𝐼𝑁𝑝 ) 𝑋 2 𝑋 𝑓𝑚𝑎𝑥 𝑋 (𝑛𝑢𝑚𝑥 − 1)/ 1000 + 1 …. … (2 – 11)
Where;
nump is the number of p-traces.
pmax is the maximum p-trace.
pmin is the minimum p-trace.
fmax is the maximum frequency.
numx is the number of x-traces.

As shown in figure (2.20), any linear event in time domain will be transformed
into point in the tau-p domain and eventually forming p-traces. The points in tau-p
domain will be distributed depending on the dip of the linear event in the t-x domain,
in which the white line in the figure will transform into a point in the center in tau-p
domain, the yellow line will be positioned in the positive side while red line is in
negative side because of the inverse dipping.

40
 CHAPTER TWO

Figure 2.20: The Tau-p transform after (Western Geco 2015b).

2.7.3 F-X Coherent Noise Suppression (FXCNS) Filter

The F-X coherent noise suppression is used to attenuate coherent noise that often
corrupts seismic data. Much of this noise originated at near surface in the form of
dispersive surface waves, and strong trapped modes that have the same apparent
velocity as refraction in which will interfere with the primary reflection data. The
removal of coherent noise is essential because such energy can adversely affect the
prestack algorithms and it often survives the stacking process, (Western Geco
2015d).

Although many acquisition and processing techniques are successful in

suppressing coherent noise for 2D data, such methods are often ineffective for the
3D case. 2D data spatially samples the seismic wave field in a regular manner along
a single direction. Thus frequency- wavenumber (F-K) and some other filtering

41
 CHAPTER TWO

methods are effective in reducing coherent noise, for 3D data signal and noise arrive
to the receivers from a wide range of directions, therefore in any given propagation
direction the wavefield is not uniformly sampled, so the straightforward of using (F-
K) are excluded, (Gaiser 1995).

The FXCNS uses a frequency-space domain fan filter and a least-squares

optimization scheme. The noise is locally estimated at each receiver for a specified
range of apparent velocities and frequencies then subtracted from the data. The
estimation procedure uses a trace’s nearest neighbors, to determine the content of
coherent noise for a specified range of apparent velocities. The least-squares
estimate is performed independently for each frequency over a specified portion of
the bandwidth (Barbero and Suarez 2008).

A shot record is Fourier transformed to frequency domain, is modeled by equation:

(Western Geco 2015d).

d (ꞷ, x) = s (ꞷ, x) + c (ꞷ, x) + r (ꞷ, x) … … …. …. (2 – 12)

Where:
s (ꞷ, x) = Desired signal
c (ꞷ, x) = Coherent noise
r (ꞷ, 𝑥) = Random noise
ꞷ = Angular frequency
x = Source-receiver offset

The least-square error for estimating coherent noise for each ꞷ from the data is
minimized by equation: (Western Geco 2015d).

Φ (ꞷ) = ∑n [ d (ꞷ, xn ) − 𝑓 (ꞷ, xn ) a (ꞷ, xn )]2 … … …. …. (2 – 13)

42
 CHAPTER TWO

Where:
Φ (ꞷ) Objective function
𝑓 (ꞷ, 𝑥𝑛 ) 𝑎 (ꞷ, 𝑥𝑛 ) = Coherent noise
With
𝑓 (ꞷ, 𝑥) = Time delay and time advance operators
𝑎 (ꞷ, 𝑥) = Weighting function

2.7.4 Frequency-Wavenumber (F-K) Filter

The use of frequency-wavenumber filtering (also known as velocity filtering) is
to extract coherent noise occurrences from seismic records based on the specific
angles at which the events dip. The dip angle of an event is determined by the
apparent velocity with which it propagates across a spread of detectors, (March &
Bailey 1983).

A seismic pulse travelling with velocity v at an angle α to the vertical will

propagate across the spread with an apparent velocity (Va= v/sina) figure (2.21).
Along the spread direction, each individual sinusoidal component of the pulse will
have an apparent wavenumber ka related to its individual frequency f, where (f = Va
ka). Hence, a straight-line curve with a gradient of Va will be formed from the plot
of frequency against apparent wavenumber for the pulse. Any seismic event
propagating across a surface spread will be characterized by an F-K curve radiating
from the origin at a particular gradient determined by the apparent velocity with
which the event passes across the spread. The overall set of curves for a typical shot
gather containing reflected and surface propagating seismic events is shown in figure
(2.22), (Kearry et.al 2002).

43
 CHAPTER TWO

Figure 2.21: A wave travelling at an angle α to the vertical will pass across an in-
line spread of surface detectors at a velocity of v/sina. after (Kearry et.al 2002).

Figure 2.22: An f–k plot for a typical shot gather containing reflection events and
different types of noise, after (Kearry et.al 2002).

It is apparent that different types of seismic event fall within different zones of
the f–k plot and this fact provides a means of filtering to suppress unwanted events
on the basis of their apparent velocity. The normal means by which this is achieved,
known as f–k filtering, is to enact a two-dimensional Fourier transformation of the
seismic data from the t–x domain to the f–k domain, then to filter the f–k plot by
removing a wedge-shaped zone or zones containing the unwanted noise events and
finally to transform back into the t–x domain, (March & Bailey 1983).

44
 CHAPTER TWO

Figure (2.23) shows how events with different dips appears in t-x and in f-k
domain. Events will be plotted in the f-k domain in negative or positive wave number
depending on their dip.

Figure 2.23: Comparing Events in time (t-x) domain and in frequency wavenumber
(f-k) domain, (XGECO).

2.8 Synthetic Seismic Data Processing

Synthetic seismic data was generated based on the inline stacked section of Al-
Samawa area and processed using open source program package seismic Unix. In
order to understand the orders and the scripts of Seismic Unix, an advanced concept
of Seismic Unix is shown below.

2.8.1 Advanced Seismic Unix (SU) Concepts

SU is easy to use because it does not require extensive programming knowledge,
only basic Unix commands for file management, pipes, and shell redirection.
Although SU can be run from the command line, shell scripts are the most effective
way to use SU. Shell scripts allow us to combine several processing programs into
one job comparable to commercial seismic processing packages, (Forel et.al 2005)

45
 CHAPTER TWO

The following Seismic Unix commands are not complete. These are a selection
of commands that considered helpful to understanding the processing in SU.

(A). Invoke the Shell: #! /bin/sh

To start the Bourne shell interpreter, the first line of any script we make must be:

#! /bin/sh

(B). Comment line: #

A comment line is one that is not seen by the shell. It is for user-comments within a
shell script. A comment line begins with the # character. For example:

# Name input and output files

(C). Pipe: |

The “Pipe” connects two processes and allows data to flow from one process to
another. For example:

suplane | suxwigb

suplane creates data, then the pipe sends the data to the imaging program, suxwigb
to be seen on the screen. In other words, data flow is through the pipe, left to right.

(D). Shell redirection < and > and >>

The redirect < sends data from right to left. For example:

suwind < seis1.su

data file seis1.su is sent as input to program suwind, a windowing program that will
be used later.

The redirect combination below sends data file (a) to process (b); the output of (b)
is stored in file (c):
46
 CHAPTER TWO

suwind < seis1.su > seis2

b < a > c

The first command below types the contents of file README to the screen. The
second command below redirects the result of the cat command to file info2.txt; the
contents of file README do not get to the screen. The third command below
appends the contents of file README to the bottom of file info2.txt; that is, the
original contents of file info2.txt are not overwritten.

cat README

cat README > info2.txt

cat README >> info2.txt

(E). Shell internal variables ( = ) and ( $ )

Inside a shell script, variables are defined with an equal (=) sign and referenced (used
in the script) with the $.

The line offset=2000 sets the variable offset to 2000. The value of the variable is
later used as $offset.

(F). Continuation mark \

Sometimes we place instructions on more than one line for the sake of visual clarity.
In the following instruction, the title of the output plot is placed on the second line.
When the continuation mark (backslash) is used, no other character or space can be
to its right.

suximage < shot$1.su perc=95 \

title=”SP # $1”

47
 CHAPTER TWO

(G). Exit the shell

At the end of a script the command exit can be used. If exit is used, the shell
terminates at this point, even if there are commands after exit. The command is:

Exit

2.8.2 Trace Headers

When seismic traces are in SEG-Y format, the SU trace format, and many other
formats, the beginning of every trace, the trace header, has information about the
trace. You can think of these as slots of information above the data part of the trace.
The data part is the time amplitude series that we see in a seismic display.

Trace header information might include the trace number and the offset of the
trace (for shot or CMP gathers). In an SU seismic data set, the number of trace header
slots is the same to ensure that every trace in a data set has the same length (in terms
of bytes of storage) (Forel et.al 2005). SU doesn't call them headers; it calls them
keys. The following table (2.1) lists some SU keys

Table 2.1: Seismic Unix keys and their definitions.

Key Definition
dt Sample interval in micro second
ns Number of samples in this trace
ntr Number of traces
offset offset
tracf Trace number within field record
tracl Trace sequence number within line
tracr Trace sequence number within reel (entire data set)

48
 CHAPTER TWO

2.8.3 Synthetic Seismic Data Filtering

Three different techniques were used for noise attenuation of the synthetic
seismic data generated by SU:

2.8.3.1 SUFXDECON
Sufxdecon is used to attenuate random noise by fx-deconvolution. The method
of noise reduction known as FX-Deconvolution or FX-Prediction was first
introduced by Canales (1984). The appearance of linear dipping events in the F-X
(Frequency vs. space) domain led to the realization that a complex Wiener Filter,
one for each frequency, would successfully predict the complex series formed by
taking samples at the same frequency from the Fourier Transforms of each trace,
(Galbraith 1991). In which Each trace is transformed to the frequency domain, for
each frequency, Wiener filtering, with unity prediction in space, is used to predict
the next sample. At the end of the process, data is mapped back to t-x domain.

The Algorithm of the filter is (Seismic Unix help 2005):

- read data
- loop over time windows
- select data
- FFT (t -> f)
- loop over space windows
- select data
- loop over frequencies
- autocorelation
- matrix problem
- construct filter
- filter data

49
 CHAPTER TWO

- loop along space window

- FFT (f -> t)
- reconstruct data
- output data
Sufxdecon is applied in SU through the following order:
Sufxdecon < input [parameters] > output
The parameters are:
Taper = length of taper
Fmin = minimum frequency to process in Hz
Fmax = maximum frequency to process in Hz
Twlen = entire trace time window length (minimum 300ms for lower freqs)
Ntrw = number of traces in window
Ntrf = number of traces for filter (smaller than ntrw)

2.8.3.2 SUFILTER
Sufilter is used to attenuate noise based on the frequency content of the data.
Frequency filtering is the process of directly modifying the amplitude spectrum of
the seismic data. The primary goal for frequency filtering is that the amplitudes of
signal and noise appear in the seismic data at various frequency components. If noise
frequencies are close to the signal frequencies, the frequency filtering may also
attenuate the reflection signals along with the noise ; (Dondurur 2018).

Several different filter designs exist to properly modify the amplitude spectrum
of the input data (figure 2.24) in which:

(A). Low-pass filters, remove the frequencies higher than a specified cut-off
frequency value. Low-frequency cut-off (f1) is zero.

50
 CHAPTER TWO

(B) High-pass filters, remove the frequency band lower than a specified cut-off
frequency value (f1). High-frequency cut-off (f2) is the Nyquist frequency.

(C). Band-pass filters, keep the specified frequency band between a low- and a high-
frequency cut-off value (f1 and f2, respectively).

(D). Notch filters remove only one specified frequency value or a very narrow
frequency band typically used to filter out specific monofrequency noise
components, such as 50- or 60-Hz.

Sufilter is applied using the following order:

Sufilter < input [parameters] > output
The parameters are:
F = f1,f2… array of filter frequencies (HZ)
Amps = a1,a2… array of filter amplitudes
dt = (from header) time sampling interval (sec)

Examples of filters:
Bandpass: sufilter < data f =10,20,40,50 amps = 0.,1.,1.,0 |
Bandreject: sufilter < data f =10,20,30,40 amps =1.,0.,0.,1. |
Lowpass: sufilter < data f =10,20,40,50 amps =1.,1.,0.,0. |
Highpass: sufilter < data f =10,20,40,50 amps =0.,0.,1.,1. |
Notch: sufilter < data f =10,12.5,35,50,60 amps =1.,.5,0.,.5,1. |

51
 CHAPTER TWO

Figure 2.24: Schematic illustration of pass-bands of different filter designs in the

frequency domain. (A) Low-pass, (B) high-pass, (C) band-pass, and (D) notch
filters. f1 and f2 represent cut-off frequencies. After (Dondurur 2018).

2.8.3.3 SUCONV
Suconv is a convolution with user-supplied filter. The convolution is Change in
waveshape as a result of passing through a linear filter (Sheriff 2002). The filter is
applied using the order:

Suconv < input filter= [parameters] > output

The required parameter is one of two, either sufile which is a convolution with a file
containing SU trace to use as filter. Or convolution user-supplied filter.

Examples:

suconv< data sufile=trace

Here, the su data file is convolved trace by trace with the single su trace,

52
 CHAPTER TWO

suconv< data filter=1,2,1

Here, the su data file is convolved trace by trace with the filter shown.

53
 Modeling and
Processing Al-
Samawah Strata
Using Seismic Unix
Software
 CHAPTER THREE

3.1 Preface
Seismic Unix is an open source software environment (softwares package) used
by exploration geophysicists, seismologists, environmental engineers, software
developers and others for seismic studies and data processing. It is used by scientific
staff in both small companies and major oil and gas companies, and by academics
and government researchers. The Seismic Unix project is partially funded by the
Society of Exploration. Geophysicists (SEG), and by the Center for Wave
Phenomena (CWP), Department of Geophysical Engineering, Colorado School of
Mines. Past support for SU has included these groups, as well as the Gas Research
Institute (GRI), (Cohen and Stcokwell 1994). The package's philosophy is to provide
a free processing and development environment in a proven structure that can be
retained and extended to suit a range of users requirements. The package is not
necessarily limited to seismic assignments. However, a broad suite of wave-related
processing can be done with SU, making it a somewhat more general package than
the word “seismic” implies. SU is intended as an extension of the Unix operating
system, and therefore shares many characteristics of the Unix, including Unix
flexibility and expandability (Al-Shuhail, 2007).

Seismic Unix contains a group of seismic processing applications complemented

with a variety of applications for manipulating and displaying seismic data. It is also
providing a robust set of subroutine libraries that greatly facilitate the development
of new applications. The applications provided with seismic Unix are classified
according to their purpose (Murillo, 1996).

The aim of this chapter is to generate a noisy synthetic seismic data based on the
stacked section of Al-Samawa area, then applying different techniques to reduce the
noise, the major processing steps of this chapter are introduced in table (3.1).

54
 CHAPTER THREE

Table 3.1: The processing steps.

Build a model
Acquire seismic data
Amplitude correction
Added random noise
Noise reduction
(Filtering)

Sort from shot gathers to CMP gathers

Velocity analysis
Normal moveout correction
Stack

3.2 Seismic Unix Programs

Seismic Unix has programs for performing many of the operations of geophysical
signal processing. It is possible to use it to manipulate and create your own
seismograms, and also to convert them between the SU standard file and the industry
standard SEG Y, (Stockwell and Cohen 2008).

Below a list of the programs that have been used in this chapter:

3.2.1 Editing, Sorting and Manipulation Programs

Suchw : Change Header Word using one or two header word fields.

Sukeycount : COUNT the number of unique values for a given key word.

Surange : get max and min values for non-zero header entries.

Suwind : window traces by key word.

55
 CHAPTER THREE

Susort : sort on any segy header keywords.

Suaddnoise : add noise to traces.

3.2.2 Filtering, Transforms and Attributes

Suconv : convolution with user-supplied filter.

Sufilter : applies a zero-phase, sine-squared tapered filter.

Sufxdecon : random noise attenuation by FX-Decovolution.

Suspecfx : Fourier spectrum (T -> F) of traces.

3.2.3 Gain, NMO, Stack and Standard Processes

Sugain : apply various types of gain to display traces.

sunmo : NMO for an arbitrary velocity function of time and CDP.

Suatck : stack adjacent traces having the same key header word.

Suvelan : compute stacking velocity semblance for cdp gathers.

3.2.4 Model Building and Data Acquisition

Trimodel : make a triangulated sloth (1/velocity^2) model.

Triseis : make a synthetic seismograms for a sloth model.

Triray : dynamic ray tracing for a triangulated sloth model.

56
 CHAPTER THREE

3.2.5 Graphics

Suxwigb X-windows wiggle plot of a segy data set.

Suximage : X-windows image plot of a segy data set.

Suxgraph : X-windows GRAPH plot of a segy data set.

3.2.6 Import / Export

Suoldtonew : convert existing su data to xdr based system-independent format.

3.3 Model Building

An important aspect of seismic exploration and research are programs for

creating synthetic data. Such programs find their use, both in the practical problem
of modeling real data, as well as in the testing of new processing programs. A
processing program that will not work on idealized model data will likely not work
on real seismic data (Stockwell and Cohen 2008).

A 9 layers triangulated (sloth) model were generated based on a stacked section

to simulate Al-Samwah area layers figure (3.1), using the TRIMODDEL program.
All the layers in this chapter are homogeneous and isotropic, each layer has a single
acoustic (P-wave) velocity. Sloth model is a slowness model, where velocities in
different layers are entered by (1/velocity^2). The model is done by interpolating
points using the Delaunay Triangulation (Forel et.al 2005).

The model figure (3.2) goes from zero to six kilometers in both x direction
(distance) and z direction (depth). The script used to build the model was found by
(Forel etal 2005) and modified to the wanted parameters as shown:

57
 CHAPTER THREE

1 #! /bin/sh
2 # File: model1.sh
3
4 # Set messages on
5 set -x
6
7 # Experiment Number
8 num=1
9
10 # Name output binary model file
11 modfile=model${num}.dat
12
13 # Name output encapsulated Postscript image file
14 psfile=model${num}.eps
15
16 # Remove previous .eps file
17 rm -f $psfile
18
19 trimodel xmin=0 xmax=6 zmin=0 zmax=6 \
20 1 xedge=0,6 \
21 zedge=0,0 \
22 sedge=0,0 \
23 2 xedge=0,6 \
24 zedge=0.4,0.4 \
25 sedge=0,0 \
26 3 xedge=0,6 \
27 zedge=0.6,0.6 \
28 sedge=0,0 \
29 4 xedge=0,6 \
30 zedge=1,1 \

58
 CHAPTER THREE

31 sedge=0,0 \
32 5 xedge=0,6 \
33 zedge=1.4,1.4 \
34 sedge=0,0 \
35 6 xedge=0,6 \
36 zedge=2,2 \
37 sedge=0,0 \
38 7 xedge=0,6 \
39 zedge=2.25,2.25 \
40 sedge=0,0 \
41 8 xedge=0,6 \
42 zedge=2.5,2.5 \
43 sedge=0,0 \
44 9 xedge=0,6 \
45 zedge=3,3 \
46 sedge=0,0 \
47 10 xedge=0,6 \
48 zedge=4.5,4.5 \
49 sedge=0,0 \
50 11 xedge=0,6 \
51 zedge=6,6 \
52 sedge=0,0 \
53 kedge=1,2,3,4,5,6,7,8,9,10,11 \
54 sfill=0.1,0.1,0,0,0.50,0,0 \
55 sfill=0.1,0.5,0,0,0.47,0,0 \
56 sfill=0.1,0.8,0,0,0.44,0,0 \
57 sfill=0.1,1.2,0,0,0.41,0,0 \
58 sfill=0.1,1.6,0,0,0.38,0,0 \
59 sfill=0.1,2.15,0,0,0.35,0,0 \
60 sfill=0.1,2.35,0,0,0.32,0,0 \

59
 CHAPTER THREE

61 sfill=0.1,2.75,0,0,0.29,0,0 \
62 sfill=0.1,3.5,0,0,0.26,0,0 \
63 sfill=0.1,4.75,0,0,0.23,0,0 > $modfile
64 ## x,z
65
66 # Create a Postscipt file of the model
67 # Set gtri=1.0 to see sloth triangle edges
68 spsplot < $modfile > $psfile \
69 gedge=0.5 gtri=2.0 gmin=0 gmax=1 \
70 title="Earth Model - 9 layers [M${num}]" \
71 labelz="Depth (km)" labelx="Distance (km)" \
72 dxnum=1.0 dznum=0.5 wbox=6 hbox=6
73
74 # Exit politely from shell
75 exit

Figure 3.1: Stacked section 45485 of Al-Samawa area used for model building.

60
 CHAPTER THREE

Figure 3.2: Nine homogeneous isotropic layers model that simulate the stacked
section of Al-Samawa area.

The script can be divided into several sets:

System:

Line 1 invokes the shell, line 5 turns on messages, and line 55 exits the shell.

Variables:

Line 11 assigns a name to the output binary model file that is used on line 63.

Line 14 assigns a name to the output .eps image file that is used on line 68.

Program TRIMODEL: Lines 19-63 create the model. The 11 sets of xedge, zedge,
and sedge triplets (lines 20-52) define layer boundaries and velocity gradients. There
is a requirement that each set of triplets have the same number of values. A triplet

61
 CHAPTER THREE

consists of xedge, zedge pairs that is an interference control point and sedge represent
velocity gradient at the control point, Line sedge has only zeros because all layers
are isotropic and homogeneous.

Program trimodel fills the model with triangles of (1/velocity)2. While

(1/velocity) is called “slowness,” (1/velocity)2 is called “sloth.”. The sloth values
were added in the sfill lines (lines 53-63). These lines are written for isotropic,
homogenous layers, which is why most of the values are zero. The sfill parameters
are shown in Table (3.2).

Table 3.2: sfill variables.

x Z x0 z0 S00 (sloth) ds/dx ds/dz

Each x-z pair of sfill is point in the layer and each sfill line describes the sloth
value that fills the layer, the model consists of 11 boundaries, therefore there are 10
sfill lines. Table (3.3) shows the sloth values used in the model. These P-wave
velocity values are not realistic, just take in consideration it is increasing with depth.

Table 3.3: Values of sfill in the model.

X (m) Z (m) s00 (sloth) Velocity (1/√𝑠𝑙𝑜𝑡ℎ (m/s)
0.1 0.1 0.5 1414
0.1 0.5 0.47 1458
0.1 0.8 0.44 1507
0.1 1.2 0.41 1561
0.1 1.6 0.38 1622
0.1 2.15 0.35 1690
0.1 2.35 0.32 1767
0.1 2.75 0.29 1856
0.1 3.5 0.26 1961
0.1 4.75 0.23 2085

62
 CHAPTER THREE

Line (68-72) program spsplot plot the triangulated sloth function as a postscript
(image) file. One thing to note that the dimension of the image must be the same as
the model, such as in line 72 wbox and hbox values are 6 which represent the model
dimension in x (distance) and z (depth). (The rest of the scripts used in this chapter
are too long and will be added in Appendix A).

3.4 Synthetic Seismic Data Acquisition

A 2D line of seismic data has been generated using the 9 layers model that was
previously built. 120 shot was generated using script found by (Forel et al 2005) and
modified to generate a synthetic 2D seismic line (Appendix A) with the survey
parameters shown in table (3.4).

Table 3.4: Acquisition parameters for the synthetic seismic data survey.

Shots number 120

Shots interval 50 m
Geophones number 60
Geophones interval 50
Geophone offset range -1475 to +1475
Shot locations (0 – 6) km
Geophone locations (-1475 – 7475)
Type of spread Split spread
Number of samples (ns) 1500
Samples interval (dt) 0.004 ms

63
 CHAPTER THREE

The final data set contain 7200 traces (120 shots x 60 geophones per shot) with
a 6 second length (dt x ns) and nine hyperbolas referring to the nine layers form
earth model. Figure (3.3) shows three shot gathers in time domain from x=0 , x=3
and x=6 positions using suwind program.

Suwind < input.su key=fldr min=() max=() | suxwigb perc=99

Below is the surange program output for the seismic data

tracl = (1 - 7200), tracr = (1 - 7200), fldr = (1 - 120), tracf = (1 - 60),

offset = (-1475 - 1475), sx = (0 - 5950), gx = (-1475 - 7425) ns = (1501)
dt = (4000)

Figure 3.3: Shots number (1,60,120) acquired from the nine layers earth model.

64
 CHAPTER THREE

3.5 Amplitude Correction

Although the data are synthetic, but it’s affected by spherical diverging in which
amplitude of the signal reduced with the depth or time. So, in order to recover the
signal, a compensation based on time was applied using sugain program using Tpow
(time power) compensation that will compensate the amplitude based on time.
Sugain program was tested and applied three different time using tpow of (1,2,3)
respectively, through the following order: Sugain < input.su tpow=() > output

The results are shown in figure (3.4), in which:

(A) a raw shot gather that shows the difference between amplitude in upper and
lower part of the shot.

(B) the shot after applying gain with (Tpow=1), note the signal in all parts of the
shot are almost has an equal amplitude.

(C) the shot after applying gain with (Tpow=2), the signal in the lower part now has
a higher amplitude than the upper part since the compensation is increasing with
time resulting in decreased energy in the upper part

(D) the shot after applying gain with (Tpow=3), the amplitude in the lower part has
much higher amplitude comparing to upper part.

After testing different parameters, a gain with tpow=1 has been choosed.

65
 CHAPTER THREE

Figure 3.4: Applying gain to a shot gather (A) shot without gain, (B) shot after
applying gain (tpow=1), (C) shot after applying gain (tpow=2), (D) shot after
applying gain (tpow=3).

66
 CHAPTER THREE

3.6 Create Random Noise

Since the data are synthetic, its free of noise. But it’s possible to add random noise
using suaddnoise program through the following order:

Suaddnoise < input sn=() > output

Where sn refer to signal/noise ratio.

Signal / noise of 10% (10 % signal – 90% noise) of random noise with random
frequencies and amplitudes has been added to the previously generated synthetic
data as shown in figure (3.5). It’s easier to attenuate random noise of higher S/N
ratio, but S/N = 10 was chosen in order to make the process a bit more challenging.
The suaddnoise program can be specified to add random noise with specified
frequency and amplitude.

Figure 3.5: Added random noise of signal / noise = 10% to a shot gather.
67
 CHAPTER THREE

3.7 Random Noise Attenuation

Three techniques have been applied in order to attenuate random noise that was
previously added. The aim is to attenuate as much as possible of the random noise
without affecting the signal. It’s might be a hard task to attenuate all the random
noise without affecting the signal, since the data consist of 90% noise of different
amplitudes and frequencies.

The first filter used was Sufxdecon using the following order:

Sufxdecon < input [parameters] > output

The filter was tested and applied two times using the following parameters:

(A) ntrw=60 ntrf=59 fmin=1 fmax=120 twlen=0.5 taper=0.1

(B) ntrw=60 ntrf=59 fmin=1 fmax=120 twlen=2 taper=1

Since the data contain noise of random frequencies, the filter was set to process
all the range of frequencies (1-120 or Nyquist). Ntrw and ntrf in both are the same
since they depend on the trace number of the shot gather and the change was based
on time-window length (twlen) and taper. The results are shown in figure (3.6) which
is (A) is the result of the first set of parameters and (B) is the second.

Comparing between the two results in figure (3.6), the first set of parameters (A)
is a medium filtering that gives an excepted result, although a small part of the signal
is removed with the noise comparing to (B) which is a sever filtering resulted from
increasing twlen and the taper. Clearly can be seen that is although the second set of
parameters removed a much higher portion of noise, but in the same time removed
a bigger part of signal comparing to first set specially in the lower part it’s clearly
seen that the deeper hyperbola (in the box) is almost completely removed in (B)
comparing to (A).Many other sets of parameters were tested and applied, but no
better result was found.
68
 CHAPTER THREE

Figure (3.7) shows the frequency-amplitude spectrum before and after applying
sufxdecon, note that the range of frequencies between (1-5) and (80-120) all were
removed and considered as a noise based on the filter estimation. Also note the
higher amplitude in first spectrum (A) is 70, but after filtering the highest amplitude
shown in the second spectrum is reduced to 50. Indicating that the filter removed all
the amplitude above 50 as it may considered as a high amplitude spikes that was
covering the signal.

1 2

Figure 3.6: Applying Sufxdecon on a shot gather: (1) shot before filtering (2) after
filtering (A) medium filtering (B) severe filtering.

69
 CHAPTER THREE

Figure 3.7: shot gather spectrum (A) before filtering (B) after filtering.

In order to attenuate the remaining noise, other techniques must be used, since
any further filtering using sufxdecon will severely affect the signal. So, the data has
been checked using frequency-offset spectrum (fx-spectrum), using sufxspec
program figure (3.8) using the following order:

suspecfx < input | suop op=norm | suximage perc=99 cmap=hsv2

70
 CHAPTER THREE

Figure 3.8: Frequency-Offset spectrum for shot gather after applying Sufxdecon.

Based on figure (3.8), it can be said that the data within the range of (15-70) Hz
are signal plus some random noise which cannot be rejected. But the frequencies
below 15 Hz and above 70 Hz are considered noise with no signal overlapping
between it, so in order to remove the mentioned frequencies, sufilter program was
used to remove the unwanted frequencies using the following order:

Sufilter < input f=14,15,70,71 amps=0,1,1,0

The rejected data are shown using the following order:

Sufilter < input f=14,15,70,71 amps=1,0,0,1

The result of the passed and rejected data are shown in figure (3.9).

71
 CHAPTER THREE

Figure 3.9: applying Sufilter on a shot gather (A) before filtering (B) after filtering
(the passed data) (C) after filtering (the rejected data).

Although it’s difficult to notice the difference before and after filtering (A) and
(B), but the rejected data (C) shows there is a huge amount of noise that was rejected.

The remaining data is between (15-70) Hz, which is a mix of signal + noise and
it’s not possible to use sufilter to reject the noise alone without affecting the signal.
So, in order to further enhance the signal, a convolution has been applied using
suconv program which is a convolution with the user-supplied filter. After several
testings, suconv was applied using (4,2,4,2) filter through the following order:

Suconv < input filter=4,2,4,2 > output

The results are shown in figure (3.10), clearly can be seen the effect of
convolution on the data specially in removing the high amplitude spikes which is
more obvious in the magnified part from 4 to 6 second of the data. Figure (3.11)
shows a comparision of the shot gather before and after noise attenuation.

72
 CHAPTER THREE

High amplitude spikes

Figure 3.10: Applying Suconv with (4,2,4,2) filter on a shot gather (A) before
convolution, (B) after convolution.

73
 CHAPTER THREE

Attenuated Signal

Figure 3.11: Noise attenuation on a shot gather (A) before noise attenuation (B)
after noise attenuation.

Comparing between A and B in figure (3.11), it can be seen that a very large
amount of noise has been attenuated and the hyperbolic shape of the signal which
was completely coverd with noise in (A), can be clearly seen now in (B). But in the
same time, removing noise using previous techniques led to remove some part of the
signal, especially those indicated by the boxes in figure (3.11). And it can be tell
judging from the previous results that sufxdecon filter is responsible for removing
part of the signal while denoising; since the rejected data from sufilter shows no
indication of a rejected signal, and the same goes for suconv.

74
 CHAPTER THREE

At this point, the results are not accepted specially in case of the real data, since
removing the signal will affect the interpretation processes. Different parameters and
expeiriment has been made using sufxdecon filter, but no better than this result was
obtained. Since sufxdecon is a filter made on estimation of the noise, it can be tell
that the filter accidentaly estimated and considered the signal as a noise and removed
it. Since the signal/noise ratios is very low.

So, one last experiment was made to solve the problem, by applying normal
move-out correction on the shot gather before attenuating noise. Since the filter
made on estimation, giving a bigger seperation between the noise and the signal may
help the filter to precisely estimated the noise. And that is done by applying a normal
move-out correction to the shot gather and flattening the shape of the signal.

Todorov and Margrave (2010), applied a CDP noise attenuation and found that
noise attenuation on NMO-corrected gathers in offset direction will show a very
strong performance, and in the same time preserving the relative amplitude changes.
And concluded that noise attenuation in NMO-corrected gathers is an effective tool
for removal of high-amplitude spiky noise and uncorrelated random noise.

On the other hand, Chen et all (2014) applied median filter in order to remove
spiky random noise and proposed to apply median fltering to remove the noise after
normal moveout correction, otherwise it will harm much of the useful energy.

75
 CHAPTER THREE

So, in order to apply NMO correction to the shot, time-velocity picking must be
held first. Using sevelan program time-velocity picking on the shot has been made
as shown in Table (3.5) bellow:

Table 3.5 Time-Velocity picking values.

Time 0.59141 0.8585 1.3831 1.8887 2.6422 2.9284 ,3.2146 3.7583 5.2845
Vel 1393.05 1452.45 1482.15 1511.8 1586. 1600.9 1600.9 1675 1838.5

After applying NMO using sunmo program on the shot, the same techniques were
applied on the NMO-corrrected gather. The results of noise attenuation before and
after applying NMO are shown in figure (3.12), in which (A) is the noise
attenuation before NMO, (B) noise attenuation after NMO and (C) is the result
after inversing the NMO for the sake of better comparesion. It can be clearly seen
that the result is very good and the effect of NMO on noise attenuation is obvious,
not only it’s solved the problem of the lost signal, as shown and indicated by boxes
from 3 to 6 second in (B) and (C), but also much higher ratio of random noise has
been attenuated and resulting much higher signal/noise ratio than before.

76
 CHAPTER THREE

Figure 3.12: Applying noise attenuation on a shot gather (A) before NMO (B) after
NMO (C) after inverse NMO.

77
 CHAPTER THREE

3.8 Sort From Shot Gather to CMP Gather

In order to sort the data to the cmp gather, we need to update the geometry with
the cdp values since the header of the data does not contain cdp headers.

Suchw program was used to to create cdp header and assign values to it using
the following equation:

Key1 = ( a + b * key2 + c * key3 ) / d

Where key1 is cdp, key2 is geophone position (gx) and key3 is source position
(sx). The scalars a,b,c and d must be computed from the geometry, since the
reciever and source positions increment are 50, (a) value is determined to be 50
more than largest positive offset which is 1475 + 50 =1525, (b) and (c) are both set
to be 1 because the cdp gathers output must increment by 1, and finaly d is set to
be 50 which is the cdp increment values. so the equation become:

Cdp = ( 1525 + 1 * gx + 1 * sx ) / 50

To compute cdp header, suchw program was used through the folllowing order:

Suchw < input key1=cdp key2=gx key3=sx a=1525 b=1 c=1 d=50 > output

After computing cdp header susort program was used to sort the data from shot to
cdp gathers: susort < input cdp offset > output

Below is the output of surange after cdp header update:

tracl = (1 - 7200), tracr = (1 - 7200), fldr = (1 - 120), tracf = (1 - 60),

offset = (-1475 - 1475), sx = (0 - 5950), gx = (-1475 - 7425) ns = (1501)
dt = (4000), cdp (1 – 298).

So, as a result, 298 cdp has been updated to the header of the data.

78
 CHAPTER THREE

3.9 Velocity Analysis and Normal Move-out

After sorting the data from shot gathers to cmp gathers, the total was 298 cmp
gathers. But not all of them is full fold. So, in order to compute the number of full
folds cdp gathers, sukeycount was used:

Sukeycount < input key=cdp > output,txt

Then the output txt file was plotted in excel as shown in figure (3.13).

The results shows that the first full fold is cdp 60 and the last one is 240, note
that full fold contain 30 traces were the shot gather full fold was containing 60
traces which is due to the acquisition geometry.

Figure 3.13: Fold values for the CMP gathers data.

A full-folds cdp from (60 to 230) were chosen to make a reasonable detailed
velocity analysis at regular intervals along the line.

79
 CHAPTER THREE

Suvelan program was used through script found by (Forel et al 2005) and
modified for the velocity analysis of the cdp from (60 – 230) (Appendix A) at
interval of 10. Figure (3.14) shows time-velocity picking on cdp 60 using
semblance, where (1) is the semblence plot for cdp 60 and (2) is the picked
velocities. The same procedure was applied on the rest of the full-fold cdp and the
results of time-velocity picks were saved in a text file that used to apply normal
move-out on the data using sunmo program.

Sunmo < input.su t-vpicks.txt > output.su

After applying NMO correction, sustack program used for stacking the cdp figure
(3.15)

Sustack < input.su > output.su

1 2

Figure 3.14: Velocity analysis of cdp 60 where (1) the semblance plot for cdp 60
and (2) the picked velocities.

80
 CHAPTER THREE

Figure 3.15: Stacked section after random noise attenuation.

Figure (3.16) shows a stacked section before and after noise attenuation where
(A) before noise attenuation and (B) after noise attenuation. The difference
between A and B can be clearly seen especially the deep reflectors in which the
reflector above 4 second is barely can be noticed where as the last reflector bellow
5 second is completely vanished due to their low amplitude although gain process
has been applied and although the stacking process enhased the signal / noise ratio
for both sections.

81
 CHAPTER THREE

Figure 3.16: Stacked section (A) before noise attenuation and (B) after noise
attenuation.

A comparision between the 9 layers model and the stacked section is shown in
figure (3.17). The stacked section was transformed from time to depth using suttoz
program, in order to accurately compare between the depth of the reflectors in the
model and the stacked section which is almost identical as it seen in the figure.

82
 CHAPTER THREE

Figure 3.17: A comparison between the 9 layers earth model and the stacked section acquired from it.

83
Noise Attenuation for
3D Al-Samawah Cube
 CHAPTER FOUR

4.1 Preface
The seismic trace is a combination of both signal and noise, the signal which is
the wanted data is the representation of the geological feature but the presence of the
noise makes it look different. So, in order to enhance the signal to noise ratio,
different technique and filters are used in this chapter to remove or reduce the noise
such as:

1- Amplitude Filtering: at which noise can be removed based on its amplitude.

Usually used to remove high amplitude random noise such as spikes.
2- Frequency Filtering: in this case the noise is removed based on its frequency
content by cutting the frequency range that noise lies in, such as ground-roll
3- Moveout Filtering: cutting noise by its moveout, usually used for linear or
parabolic noise due to their high moveout
4- Velocity Filtering: used to remove noise from seismic data on the base of
velocity or the angle at which the noise dipping. Used also for linear noise and
can be used to remove ground-roll.

4.2 Acquisition Parameters

The acquisition parameters that are used in the survey are shown below in Table
(4.1).
Table 4.1: Acquisition parameters.
Specifications Parameters
Type of spread Theoretical split spread (60+60)
Method Conventional 3D acquisition
Number of receiver lines 14
Number of active channels per receiver 120
line
Receiver station interval 50 m

84
 CHAPTER FOUR

Source station interval 100 m

Source line spacing 300 m
Receiver line spacing 400 m
Bin size 25 x 50 m
Nominal CDP fold 70
Minimum offset 35 m
Maximum offset 4375 m

4.3 Work Flow

Table 4.2 shows the processing steps applied in this chapter starting from
reformatting and updating the geometry then finishing it with the radon filter.

Table 4.2: Work flow for noise attenuation of Al-Samawah 3D seismic data.

3D Raw Seismic Data Input

Reformat
Geometry Update
Geometrical Spreading compensation
Anomalous Amplitude Attenuation (AAA)
Tau-p Linear Noise Attenuation (TPLNA)
F-X Coherent Noise Suppression (FXCNS)
Anomalous Amplitude Attenuation (AAA) 2nd pass
Multi-Channel Dip Filter (F-K Filter)
Radon (Kris-Cross) Removal

The processing steps and filters in table (4.2) are done using Schlumberger
Omega seismic data processing software.

85
 CHAPTER FOUR

4.4 Reformat
The seismic data which were supplied in SEG-D format, converted to internal
seismic file format (OMEGA FORMAT), then the data were checked using the
transcription program print out file. The input and output record number matched
with the observer information. The data were extracted with 6000 msec time depth.

4.5 Geometry Update

Geometry information was received in SPS file format. A geometry data base
was built, and the geometry data base was updated into filed data information of
shots using the geometry up data SFM (seismic function model), so old headers were
updated, new headers were added.

4.6 Geometrical Spreading (Spherical Diversion) Compensation

The Geometrical Spreading Compensation was applied in shot gather to recover
decay in amplitude caused by geometric divergence. An RMS constant velocity
function was used to compute V2t amplitude compensated function. The velocity
applied should be tested because we do not want the deep data be over compensated.
The input parameters that the program needs to compensate the lost amplitudes is a
time-velocity pairs which have been taken from an old 2D lines of the study area.

Figure (4.1) shows a raw shot before and after applying geometrical spread
compensation, and as clearly shown both the signal and the noise are now can be
seen clearly especially in the lower part of the figure in which the effect of the
spherical diversion is larger (the deeper the signal penetrates the earth the larger
amplitude loss will occur). Figure (4.2) shows a raw stack line before and after
applying GSC.

86
 CHAPTER FOUR

Ground-roll Signal

high
amplitude
spikes

Before gain After gain

Figure 4.1: The result of applying GSC on a raw shot gather, (A) before
compensation (B) after compensation.

Before gain After gain

Figure 4.2: The result of applying GSC on a stacked inline section, (A) before
compensation (B) after compensation.

87
 CHAPTER FOUR

4.7 Noise Attenuation

4.7.1 Anomalous Amplitude Attenuation (AAA)

There are many Processing techniques to reduce or attenuate noise characterized

by anomalous amplitude. One of them is Anomalous Amplitude Attenuation. It can
be used before and after deconvolution and may be run several times within a
project. This technique had very good effect in removing Abnormal Amplitued
noise, as well as some type of coherent noise such as air waves.

In order to use the filter several required parameters must be inserted which are
shown below:
 Type of window: moveout or fixed number window
 Window length
 Width of median filter (window width)
 Overlap percentage between windows
 Width of frequency bands
 Type of threshold: either Percentile or defined by using median deviation
 Threshold value
 Minimum frequency: (0 – 125) or Nyquist
 Maximum frequency: (0 – 125) or Nyquist

This technique was applied three times with different parameters as shown below:

AAA Pass 1 parameters

Type of window: moveout window.
Min frequency: 0 Hz
Max frequency: NYQUIST
Width of spatial median filter:31.

88
 CHAPTER FOUR

AAA Pass 2 parameter

Type of window: Fixed Number Overlap Window.
Gate length in time: simple window 1000 Ms.
Overlap percentage between adjacent windows =50 Ms.
Min frequency: 0 Hz
Max frequency: 80 Hz
Width of spatial median filter :31

AAA Pass 3 parameters

Type of window: Fixed Number Overlap Window.
Gate length in time: simple window: 1000 Ms.
Overlap percentage between adjacent windows: 50 Ms.
Min frequency: 0 Hz
Max frequency: 50 Hz
Width of spatial median filter: 31.

Below are the threshold parameters used for pass 2 and pass 3 Table (4.3):

Table 4.3: AAA threshold factor parameters for pass 2 and 3.

AAA pass 2 threshold parameters AAA pass 3 threshold parameters

Window number Threshold
Threshold Factor Time Factor
1000 4 1 10000
100 1000 2 1000
3 100
90 1500
4 99
80 2000
5 98
70 4000
6 97
60 6000 7 96
8 95
9 94
10 93
11 92

89
 CHAPTER FOUR

The filter was applied 3 different times, the first pass was applied with moveout
window and fixed threshold that applied to the whole window. The second and third
passes of the filter were applied with a fixed number of windows, each window have
a different threshold factor to be applied which is shown in Table (4.3). Another
thing to mention that a normal move out was applied before the third pass to
flattening the shot, and inversed after the noise removal, as done and tested in the
synthetic data processing in ch3. This step can help the filter and the user to
recognize the noise since only the signal will be affected by the velocity file which
is used to apply NMO.

The results of the 3 passes of the filter on a shot are shown in figures (4.3) and
(4.4). Note the lower part of the shot in figure (4.3) (A) contains an anomalous or
abnormal amplitude spikes (which is mostly the vehicle noise) that is covering the
signal behind it due to its high amplitude. And figure (4.4) which contains air waves
that also have been significantly removed. The results are shown in (B) and the
removed noise in (C) of the same figures. Clearly, it can be seen how the abnormal
spikes and the air waves were removed although not all of it were removed and there
was residual noise, because severe filtering may probably remove a part of the signal
and that is the reason to apply the same filter more than once, to slowly remove the
noise without affecting the signal. The results of applying the filter on a stacked
inline section are shown in figure (4.5)

90
 CHAPTER FOUR

The difference
(A-B)

Removed spikes

Abnormal
Before AAA amplitude
spikes After AAA
Residual
B
noise

Figure 4.3: The final result of applying AAA on a shot: (A) before applying the filter (B) after applying
the filter (C) the difference (the removed noise).

91
 CHAPTER FOUR

The difference
(A-B)

Air wave
Before AAA After AAA

Air wave Residual noise

Figure 4.4: The final result of applying AAA on a shot: (A) before applying the filter (B) after applying
the filter (C) the difference (the removed noise).

92
 CHAPTER FOUR

High amp spikes

Before AAA

After AAA

The difference
(A-B)

Figure 4.5: The final result of applying AAA on a stacked inline section 45660:
(A) before applying the filter (B) after applying the filter (C) the difference (the
removed noise).

93
 CHAPTER FOUR

4.7.2 Tau-p Linear Noise Attenuation (TPLNA)

To eliminate linear noise within the data, first a normal move out was applied to
the shot gathers, then transformed into the Tau-p domain. The velocities applied to
the NMO will only affect the signal, hence the desired signal will be transformed
from a dipping event into a horizonal that is almost zero dipping event, whereas the
remaining data which are in this case (the linear noise) will remain linear.

So, applying the forward radon transform to the NMO corrected shot gathers will
transform all the signal in the center (small P value), whereas the linear noise which
has a larger dipping will be mapped in the far side away from the signal (greater P
value). After applying the transformation, Tau-p filtering or muting can be used to
attenuate linear noise by dividing the Tau-p transform into pass and reject zones, by
allowing to the centered events to pass and rejecting the rest, the linear noise can be
removed leaving only the desired signal. The pass and rejected zones can be
specified either manually or using velocity parameters. Then the filter will transform
the data back to the t-x domain and finally an inverse NMO will be applied.

The procedure steps for TPLNA filter are summarized in figure (4.6) where;

(A) Is a shot contain linear noise after applying the AAA filter.
(B) Applying NMO to the shot (Note the stretching caused by the NMO is most
severe near the first arrival, and diminishes at later times).
(C) Transforming the data from t-x domain to the tau-p domain using forward radon
transform, in which at this case the signal should be in the middle and the linear
noise distributed at the sides. The pass and rejected zones are specified manually as
show by the two red and blue lines.
(D) Muting the noise and allowing only the signal to pass.
94
 CHAPTER FOUR

(E) Transforming the data back to the t-x domain.

(F) Applying inverse NMO with the same velocity file.
(G) The removed linear noise (A-F).

Figure 4.6: Procedure steps for TPLNA were (A) a noisy shot; (B) applying NMO
(C) transforming to the tau-p domain; (D) muting or rejecting the noise manually;
(E) transforming back to the time domain; (F) applying inverse (NMO) and (G) The
removed linear noise (A-F).

95
 CHAPTER FOUR

The required parameters for Radon Forward transform are shown below:
 Min moveout: smallest moveout at max offset (negative number).
 Max moveout: largest moveout at max offset (positive number).
 Offset.
(The parameters above use to define the range of the filter)
 Moveout display start time: the start time of the previously defined range.
 Maximum frequency to process: (0-120).
 Number of P-traces: either computed internally or manually added.
 Velocity parameters: in case of using auto pass and rejection zones.

The Forward Radon Transform was applied one time using the following parameters
Table (4.4):

Table 4.4: Tau-P (Radon) transformation parameters.

AGC/applied & removed 500

Transformation Type Linear
Minimum Moveout in T-X -3500
Maximum Moveout in T-X 3500
Reference Offset Distance 5000
Number of P-traces Computed
Muting Manually

Automatic gain control was applied in order to clearly observe the data, then
NMO applied in order to flatten the hyperbolic form of the signal. The result of the
Tau-p transform using the above parameters on a shot gather are shown in figure
(4.7). In which (A) a shot gathers before applying TPLNA. (B) after applying
TPLNA, and (C) the removed linear noise. The input data for the radon forward
transform are the pre-stack of any type (shot or CMP). The result of applying the
filter on a stacked inline section are show in figure (4.8).

96
 CHAPTER FOUR

Residual random
Residual Linear noise
Linear noise The difference
noise
(A-B)

Before TPLNA After TPLNA

Figure 4.7: The result of applying Tau-P linear noise attenuation filter (TPLNA) on a shot (A) before filtering (B)
after filtering (C) the difference.

97
 CHAPTER FOUR

Before TPLNA
After TPLNA

The difference
(A-B)

Figure 4.8: The result of applying Tau-P linear noise attenuation filter (TPLNA) on
an inline stacked section 41090 (A) before filtering (B) after filtering (C) the
difference.

98
 CHAPTER FOUR

4.7.3 F-X Coherent Noise Suppression (FXCNS)

FXCNS filter was used to remove linear coherent especially ground-roll using
frequency and velocity parameters.

Using frequency-space (f-x) domain, the noise is locally estimated at each

receiver for a specified range of frequencies and apparent velocities and then
subtracted from the data. In order to determine the range of frequencies that the
ground-roll lies in, the shot was divided into multiple panels with different frequency
bands figure (4.9). The shot was divided into (7) panels from (3 to 70) Hz at (10)
intervals, in which (A) is the first panel which is the original shot with frequency (1-
120) Hz, and the rest contain the data of specified frequencies. Note that the ground-
roll is start appearing at the second panel (B) of (3-10) Hz and strongly appear at the
next panel (C) of (10-20) Hz, but slightly disappear at the next panels (D-E) of (20-
40), and completely vanish at higher frequencies except of (F) at (40-50) Hz which
appear to contain a residual of the ground-roll. So, from these panels it can be clearly
seen that the ground-roll is located at the range of (3-40 or 42) Hz.

Removing the ground-roll by cutting of the range of frequencies that is located

on is possible, but it’s there is a risk of losing some of the signal that lies in the lower
frequencies and overlapping with the noise. So, in order to completely removing the
ground-roll without affecting the signal, another parameter which is velocity can be
calculated for the ground-roll. Since the signal and the noise have a different
velocity, it’s possible to remove the ground-roll using (frequency-velocity)
parameters. Figure (4.10) shows how the velocities were picked. The green and the
red lines resemble the high-stop and the low-stop velocities respectively, the high-
pass and low-pass velocities should be within the range of picked velocities. The
velocities picked in the figure are not accurate and used for test purposes.

99
 CHAPTER FOUR

Figure 4.9: Frequency panels for the Ground-roll where the range of frequencies in
(A) is (0-120), (B) is (3-10), (C) is (10-20), (D) is (20-30), (E) is (30-40), (F) is (40-
50), (G) is (50-60) and (H) is (60-70).

100
 CHAPTER FOUR

Figure 4.10: Linear velocity picking in which green line represent the high pass
velocity and blue line represent the low pass velocity.
The required parameters to apply the filter is the values for the (high-stop, high-
pass, low-stop, and low-pass) for both the frequencies and velocities which will
defines the taper. After measuring the required frequencies and velocities, the
FXCNS filter was applied two times with the parameters shown in Table (4.5) and
Table (4.6).

Table 4.5: FXCNS 1st pass parameters.

Velocities Values Frequencies Values
High - Stop 2400 High - Cut 38
High - Pass 2300 High - Pass 33
Low - Pass 1800 Low - Pass 3
Low - Stop 1700 Low - Cut 1

101
 CHAPTER FOUR

Table 4.6: FXCNS 2nd pass parameters.

Velocities Values Frequencies Values

High - Stop 1500 High - Cut 35
High - Pass 1000 High - Pass 30
Low - Pass 500 Low - Pass 5
Low - Stop 300 Low - Cut 3

The FXCNS filter was applied two times using the parameters in Table (4.5) and
(4.6). In most cases only one pass is required to eliminate the ground-roll, since the
ground-roll usually appears as linear events, with low frequencies, high amplitudes
and low velocities. But in this case a higher velocity ground roll appears due to the
study area containing relatively high velocity layers, and thus a second pass of
FXCNS with a higher velocity values to (cut in) is needed.

The final result of applying FXCNS on a shot gather is shown in figure (4.11).
The ground-roll might be hard to notice since its overlapping with the signal, but it
can be clearly seen in the difference. Figure (4.12) shows a stacked inline section
after applying FXCNS.

For quality control purposes, an F-K transform was applied on a shot after
removing the ground-roll for a chosen shot gather as shown in figure (4.13). The F-
K spectrum (D) is done for the removed ground roll (C). The lower frequency data
in the lower part of the F-K spectrum is the ground-roll, and no other than the
ground-roll appears on the spectrum in which indicating that there is no overlapped
signal was removed during filtering.

102
 CHAPTER FOUR

Ground-roll Signal

After FXCNS
Before FXCNS Residual The difference
random noise (A-B)

Figure 4.11: The result of applying FXCNS on a shot gather (A) before filtering (B) after filtering (C) the difference.

103
 CHAPTER FOUR

Before FXCNS After FXCNS

The difference
(A-B)

Figure 4.12: The result of applying FXCNS on Stacked inline section 45330 (A)
before filtering (B) after filtering (C) the difference.

104
 CHAPTER FOUR

Ground-roll
Ground-roll Ground-roll
+ Signal
Signal

Before FXCNS After FXCNS

Figure 4.13: Quality control of FXCNS filter on a shot (A) before filtering (B) after filtering (C) the difference
(D) F-K spectrum for the difference.

105
 CHAPTER FOUR

4.7.4 Anomalous Amplitude Attenuation (AAA) 2nd pass

After removing the random noise using AAA and linear coherent noise by
TPLNA and FXCNS, it was seen that a different portion of the data has amplitude
values that were lower than the rest. And hence surface constant amplitude
compensation (SCAC) was applied in order to bring the low amplitude portion of
the data to the same amplitude as the rest of the data. (SCAC) is a tool that
compensate for shot, detector and offset amplitude variation that are caused by
acquisition effects and are not a result of the subsurface geology.

As a result, the amplitude for both the signal and the noise were compensated.
And it was noticed the appearing of random high amplitude spikes in different
portions of the data that were not removed in the first pass of AAA due to their low
amplitude which was below the threshold factor. So, a 2nd pass of AAA is needed in
order to remove the rest of the random noise.

The parameters used for AAA 2nd pass is shown below:

AAA Pass 1 Parameters
Type of window: Fixed Number Overlap Window.
Gate length in time: simple window 2000 Ms.
Overlap percentage between adjacent windows =50 Ms.
Min frequency: 0 Hz
Max frequency: 45 Hz
Width of spatial median filter :11
Width of the frequency band: 5

AAA Pass 2 Parameters

Type of window: Fixed Number Overlap Window.
Gate length in time: simple window 2000 Ms.
Overlap percentage between adjacent windows =50 Ms.
Min frequency: 20 Hz
Max frequency: Nyquist
Width of spatial median filter :11
Width of the frequency band: 5

106
 CHAPTER FOUR

AAA Pass 3 Parameters

Type of window: Fixed Number Overlap Window.
Gate length in time: simple window 2000 Ms.
Overlap percentage between adjacent windows =50 Ms.
Min frequency: 0 Hz
Max frequency: Nyquist
Width of spatial median filter :41.
Width of the frequency band: 5.

AAA Pass 4 Parameters

NMO velocity file was applied to flattening the shots, and inversed after noise
removal.
Type of window: Fixed Number Overlap Window.
Gate length in time: simple window: 1000 Ms.
Overlap percentage between adjacent windows: 50 Ms.
Min frequency: 20 Hz
Max frequency: 80 Hz
Width of spatial median filter: 201.
Width of the frequency band: 5.

AAA Pass 5 Parameters

NMO velocity file was applied to flattening the shots,
and inversed after noise removal.
Type of window: Fixed Number Overlap Window.
Gate length in time: simple window: 1000 Ms.
Overlap percentage between adjacent windows: 50 Ms.
Min frequency: 20 Hz
Max frequency: 80 Hz
Width of spatial median filter: 201.
Width of the frequency band: 10.

AAA Pass 6 Parameters

NMO velocity file was applied to flattening the shots, and inversed after noise
removal.
Type of window: Fixed Number Overlap Window.
Gate length in time: simple window: 500 Ms.
Overlap percentage between adjacent windows: 50 Ms.
Min frequency: 0 Hz

107
 CHAPTER FOUR

Max frequency: 20 Hz
Width of spatial median filter: 101.
Width of the frequency band: 5.
Threshold defined by Median Deviation Percentile Method = 7.0.

Table (4.7) below shows the AAA2nd threshold parameters for pass (1-5):

Table 4.7: Threshold parameters for AAA2nd filter pass (1-5).

AAA pass 1 parameters AAA pass 2 parameters AAA pass 3 parameters

Threshold Time Threshold Time Threshold Time

Factor Factor Factor

15 4 40 4 500 4

15 500 30 500 200 2000

15 3700 15 1000 21 2600

7 5000 10 3700 15 3300

5 6000 7 5000 9 4000

5 6000 7 6000

AAA pass 4 parameters AAA pass 5 parameters

Window Threshold Window Threshold

number Factor number Factor
1 100 1 100
2 50
2 50
3 30
3 30 4 20
4 20 5 15
5 15 6 5

6 5

108
 CHAPTER FOUR

As mentioned previously it’s better to apply AAA more than once with smooth
filtering, because severe filtering may strongly affect the signal. Hence AAA2nd pass
was applied to the data 6 times with different parameters. The threshold factor was
defined using median deviation which is being increasing with time or window
number, except the last pass which was defined by percentile method in which the
same value of the threshold factor is applied to the whole data in the range of the
defined frequencies.

Figure (4.14) shows the result of applying AAA 2nd pass on a shot. Note there is
a residual of the ground-roll from FXCNS filter that has been removed using AAA,
due to their high amplitude after applying SCAC, as well as removing the high
amplitude random spikes.

4.7.5 Multi-Channel Dip Filter (F-K Filter)

Dipping events that overlap in t-x domain can be separated in the f-k domain.
This separation enables events with unwanted dips to be attenuated while the
remaining dips are unaffected. F-K filter can be used to reject linear noise from either
non-stacked or stacked data. The filter is used to remove direct wave and the residual
coherent linear noise which is usually has a steep dip in both directions, by
selectively passing and rejecting dipping event on the 3d seismic data, which is
determined by their apparent velocity

The filter is designed in the frequency-wavenumber (f-k) domain by specifying

either a fan-shaped region, seismic dips or polygonal regions that is defined
manually. It’s recommended to apply normal move-out before transforming and
rejecting the noise, then removing it after filtering. So that that the flattened signal
will be plotted in the center of the f-k transform away from linear noise, to ensure
no signal will be lost while rejecting.
109
 CHAPTER FOUR

Ground-roll
Random noise

Residual

Ground-roll High Amp Spikes

The difference

Before AAA After AAA (A-B)

Figure 4.14: The final result of applying AAA 2nd pass on a shot (A) before filtering (B) after filtering (C) the
difference.

110
 CHAPTER FOUR

The f-k filter was applied along the following procedure figure (4.15):

(A). Input Common shot gathers.

(B). Apply NMO correction using the Raw stacking velocity.
(C). Transform the data from t-x domain to f-k domain then apply fan filter to
identify the pass and reject zones.
(D). Transform the data back to t-x domain.
(E). Apply the inverse NMO correction, using the same velocity used in the step 2.

The result of applying multichannel dip (f-k) filter on a shot gather and an inline
stacked section is shown in figure (4.16) and figure (4.17) respectively. Note there
is a small amount of linear noise removed due TPLNA and FXCNS filters who both
removed large amounts of coherent linear noise and the remaining noise (residual
and direct wave) was removed by F-K filter.

111
 CHAPTER FOUR

Transform the data

to the f-k domain

Fan filter

Transform to Apply
Shot in t-x domain After apply NMO
t-x domain inverse NMO

Figure 4.15 Procedure steps for F-K filter. Where (A) shot in time domain (B) After applying NMO (C) After
transform the data to f-k domain and rejecting linear noise (D) After transform the data to time domain (E) After
apply inverse NMO.

112
 CHAPTER FOUR

The removed
direct wave

Before F-K After F-K The difference

(A-B)

Figure 4.16 The result of applying F-K filter to a shot (A) Before filtering (B) After filtering (C) The difference.

113
 CHAPTER FOUR

Before F-K filter After F-K filter

The difference
(A-B)

Figure 4.17: The result of applying F-K filter to a stacked inline section 44145.
Where (A) before filtering (B) after filtering (C) the difference.

114
 CHAPTER FOUR

4.7.6 Radon (Kris-Cross) Removal

(Kris-Cross) is a residual coherent noise within the data, formed from multiple of
parabolic events with different and opposite directions and dips, forming a shape of
(Kris-Cross). To eliminate these noises, the Weighted Least Square Radon is
applied. And as said earlier (Ch2). The radon transform is performed for any of the
parabolic, hyperbolic or linear events, previously we used linear transformation to
remove linear noise using (TPLNA), and this time a parabolic transformation will
be held in order to remove the parabolic Kris-cross noise within the data.

The procedure steps of this filter are not that different from TPLNA steps, except
that this time after applying Normal Move Out, the radon transform will target the
parabolic shaped events and placed them in the both sides of the radon transform
domain because of their high moveout, whereas the signal will be positioned in the
center. After that the data in the radon domain will be divided manually into pass
and reject zones then subtract the Kris-Cross noise from the data. Finally, the data
will be transformed back into T-X domain after removing the Normal Move Out
effect.

The weighted Least Square Radon (RWD) Transform was applied one time by
using the following parameters table (4.8):

Table 4.8 Radon (RWD) parameters.

Transformation Type parabolic

Minimum Moveout in T-X -5000
Maximum Moveout in T-X 5000
Minimum Moveout for signal -400
Minimum Moveout for signal 400
Reference Offset Distance 4075
Muting Manually

115
 CHAPTER FOUR

The minimum and maximum moveout defining the range of the work, in which
the RWD will transform any parabolic event in the range of moveout of (-5000 –
5000) excluding (-400 – 400) which is referring for the signal. Figure (4.18) shows
the result of applying RWD on a shot after rejecting the noise and removing the
NMO effect. The result of applying RWD on an inline section is show on figure
(4.19).

4.8 The Final Results

The final result of applying all the previous filters on a shot is shown in figure
(4.20). Note how the hyperbolic form of the signal in the box shown in (B) is clearly
seen, where else it’s almost impossible to note and observe the signal on the same
spot before noise attenuation (A), due to overlapping noise. Figure (4.21) shows a
stacked inline section before and after noise attenuation.

A spectrum of signal to noise ratio before and after noise attenuation is shown in
figure (4.22). The spectrum is computed internally using auto-correlations and cross-
correlations from either a 2-d rectangular group of traces or for each gather in the
data set. The spectrum can estimate the signal and noise components and may be
used as quality indicators for the data volume. Note the massive dropping of the
blue line (noise) after noise attenuation (B) figure (4.22), resulting in rising the red
signal line.
Finally, a time slice at 3500 MS is shown in figure (4.23) before and after noise
attenuation. The time slice (B) shows a massive enhancement comparing to (A) in
which it’s almost completely covered with noisy data.

116
 CHAPTER FOUR

Kris-cross

The difference
Before filtering After filtering
(A-B)

Figure 4.18: The result of applying RWD on a shot (A) before filtering (B) after filtering (C) the difference.

117
 CHAPTER FOUR

Before filtering After filtering

The difference
(A-B)

Figure 4.19: The result of applying RWD on stacked inline section 45565. Where:
(A) before filtering (B) after filtering (C) the difference.

118
 CHAPTER FOUR

Signal Before noise

attenuation Signal After noise
attenuation

Before noise attenuation After noise attenuation

Figure 4.20: Full noise attenuation on a shot gather (A) before noise attenuation (B) after noise attenuation.

119
 CHAPTER FOUR

Before noise attenuation After noise attenuation

Figure 4.21: Full noise attenuation on staked inline section 45830. Where: (A) before noise attenuation (B) after
noise attenuation.

120
 CHAPTER FOUR

Signal
Noise

Signal
Noise

Before noise attenuation After noise attenuation

Figure 4.22: Signal to noise ratio spectrum (A) before noise attenuation (B) after noise attenuation.

121
 CHAPTER FOUR

Time slice
3500 MS
before
noise
attenuation

Time slice
3500 MS
after noise
attenuation

Figure 4.23: Time slice at 3500 MS in which: (A) before noise attenuation and (B)
after noise attenuation.

122
 Conclusions and
Recommendations
 CHAPTER FIVE

5.1 Conclusions

Based on the result of processing of Synthetic seismic data using Seismic Unix
(SU), the following conclusions can obtained:

1- Although the data are synthetic, but it is affected by spherical diverging in

which the amplitude of the signal reduced with the depth or time. So,
amplitude compensation has been applied.
2- Sufilter was used to remove random noise bellow 15 Hz and above 70 Hz that
is not interpolated with signal.
3- Sufxdecon filter considered the best filter of all three and was used to process
all the frequencies of the data (1-120) Hz, that is interpolated with the data
within the range of frequencies of (15-70) Hz, resulting in removing some part
of the signal.
4- NMO was applied prior to sufxdecon filter in order to affectively attenuate
random noise without affecting the signal. since the filter is based on
prediction, applying NMO gives a better separation between the signal and
noise resulting in a very good result.
5- Suconv filter was used to remove some of the residual random spikes that
couldn’t be removed using previous filters.

Based on the result of processing for 3-D seismic data and quality control in this
thesis, the following conclusions can be obtained:

1- The dominant noise in the seismic data is the coherent noise especially linear
coherent noise.
2- Anomalous Amplitude Attenuation (AAA) overall is used 9 different times in
order to remove random noise, because severe filtering may strongly affect the
signal resulting removing random noise as well as part of the signal.

123
 CHAPTER FIVE

3- Although Anomalous Amplitude Attenuation (AAA) is meant to remove high

amplitude random noise, but the filter was capable of removing air wave noise
which considered coherent noise.
4- Removing linear noise in 3D seismic data is much harder than 2D data, since
linear noise will reach the receivers from different directions and azimuth.
5- Three different filters were used to remove the linear noise effectively without
affecting the signal which is: Tau-p, FXCNS and F-K filter
6- Filtering in the tau-p domain resulted in an effective attenuation of coherent
linear noise and produced a better-quality result and thus considered the best
for linear noise attenuation. So, Tau-p Linear Noise Attenuation filter
(TPLNA) was the main filter to remove linear noise.
7- F-X Coherent Noise Suppression (FXCNS) filter is used to remove Ground-
roll instead of F-K filter. Since 3D seismic data has irregular-offset sampling,
Ground-roll loses its linearity and changes apparent velocity with the direction
of wavefront propagation. So, the straightforward of using F-K filter is
excluded.
8- F-K filter was used to remove only the residual linear noise that couldn’t be
removed using TPLNA and FXCNS filters, since the data has a large amount
of linear noise.
9- The weighted Least Square Radon is used to effectively remove parabolic
noise that form the shape of (kris-cross).
10- Applying NMO before filtering gives a very good results in 3D data
processing as tested and applied in synthetic seismic data.

124
 CHAPTER FIVE

5.2 Recommendations

According to the study results, the following recommendations can be

suggested:
1- Test and apply Seismic Unix filters that were used in this thesis on more
complicated models and real seismic data.
2- Apply post-stack random noise attenuator (RNA) filter of Omega software
to remove residual random noise if founds.
3- It is recommended to apply all filters used in 3D data processing at the
same order they were applied, since applying TPLNA filter with the
presence of high amount of high amplitude random noise will result in
spreading the spikes over a wide range of traces within gather after
applying inverse Tau-p transform.
4- Different area of the same 3D survey might have different velocities. So,
it’s recommended to apply different tests regarding the attenuation of
ground-roll using FXCNS filter in 3D seismic data.
5- It is highly recommended to apply NMO prior to filtering, as applied and
tested in both synthetic and real seismic data, since it gave a verygood
results.

125
References
 REFERENCES

References
Al-Adhami, A. F., 1998. Analytical study of some seismic noise tests at selected
area in Iraq, Thesis of Master of Science in Geophysics (unpublished), Geology
Department, Baghdad University, 117p. (In Arabic).

Al-Dossary, S. and Wang, Y. E., 2009. Estimating incoherent noise in post-

stack seismic data. In SEG Technical Program Expanded Abstracts, Society of
Exploration Geophysicists, pp 1044-1047.

Ibrahime, R. K., 2019. Noise reduction along two seismic sections East
Diwaniya city, Southeastern Iraq, Thesis of Master of Science in Geophysics
(unpublished), Geology Department, Baghdad University.

Treky, L. F., 2018. Applications seismic UN*X (SU) package in seismic and
ground penetration radar (GPR) data processing, Thesis of Master of Science in
Geophysics (unpublished), Geology Department, Baghdad University.

Al-Sadi, H. N., 2017. Seismic Hydrocarbon Exploration: 2D and 3D

Techniques, Springer, pp, 39-40

Al-Sadi, H. N., 1992. 3D Seismic Survey, the first national experiment, research
held in the 10th Iraqi geological convention, 19 p. (In Arabic).

Al-Shuhail, A., 2007. Laboratory Manual for Seismic Data Processing Courses
at KFUPM using the Seismic UN*X Software under the Linux Operating System,
King Fahd University of Petroleum & Minerals (KFUPM), 138 p.

126
 REFERENCES

Anvari, R., Kahoo, A. R., Mohammadi, M. and Pouyan, A. A., 2017. Random
noise attenuation in 3D seismic data by iterative block tensor singular value
thresholding, In 2017 3rd Iranian Conference on Intelligent Systems and Signal
Processing (ICSPIS), pp164-168.

Arora, K., Cazenave, A., Engdahl, E. R., Kind, R., Manglik, A. and Uyeda,
S., 2011. Encyclopedia of Solid Earth Geophysics, Springer Science & Business
Media., pp (1085-1509).

Aswad, N. N., 2016. Enhancement Of Seismic Data West Luhias Area Using
Static Correction Program In Governorate Of Al- Muthana-South Of Iraq. Thesis
of Master of Science in Geophysics (unpublished), Geology Department,
Baghdad University.

Awaina, F. H., 1996. Three-Dimension Seismic of Kifl Area – Middle Iraq, PhD.
thesis of Science in Geophysics (unpublished), Geology Department, Baghdad
University, 142 p. (In Arabic).

Bacon, M., Simm, R. and Redshaw, T., 2003. 3-D Seismic Interpretation,
Cambridge University Press, 53 p.

Barbero, G. M., & Suarez, G. M., 2008. Full-wave Seismic Analysis: Source
Comparisons, Land Streamer Tests, and Converted-wave Processing (PhD thesis,
University of Calgary). p 24.

Basak, R. L., Rana, K. S., Rao, A. K., Gangaiah, A. and Chandrasekaran, C.

R., 2012. Removal of noises using Tau-P transformation-an indigenous tool for
noise attenuation in shallow seismic data. hyderbad. Pp 1-6.

Bording, P., 1995. Seismic Wave Propagation Modeling and Inversion,

Computational Science Education Project, 50 p.

127
 REFERENCES

Buday,T. and Jassim, S. Z., 1987. The Regional Geology of Iraq, Vol.2,
Tectonism, Magmatism and Metamorphism Publication of Geosurv, Baghdad,
352 p.

Calvert, A. J., 1997. Backscattered coherent noise and seismic reflection

imaging of the oceanic crust, Journal of Geophysical Research: Solid Earth. Pp
1-12.

Canales, L. L., 1984. Random noise reduction, In SEG Technical Program

Expanded Abstracts, Society of Exploration Geophysicists, pp 525-527.

Chaouch, A. and Mari, J. L., 2006. 3-D Land Seismic Surveys: Definition of
Geophysical Parameter. Oil & Gas Science and Technology Pp 613-617.

Chen, Y., Yuan, J., Jin, Z., Chen, K. and Zhang, L., (2014). Deblending using
normal moveout and median filtering in common-midpoint gathers. Journal of
Geophysics and Engineering, Vol 11, pp 1-22.

Chiu, S. K, 2013. Coherent and random noise attenuation via multichannel

singular spectrum analysis in the randomized domain, Geophysical Prospecting,
Vol 61. pp 1-9.

Claerbout, J. F, 1985. Fundamentals of Geophysical Data Processing,

Blackwell Scientific Publications, 15 p.

Cohen, J. K. and Stockwell, J. W., 1994. The SU User’s Manual. Center for
Wave Phenomena Colorado School of Mines, Gas Research Institute, Chicago,
Illinois, pp 9-18.

Cordsen, A., Gabraith, M. and Peirce, J., 2000. Planning Land 3-D Seismic
Surveys, Edited by Bob A. Hardage Series Editor: Stephen J. Hill, SEG, 199p.

128
 REFERENCES

Corsmit, J., Versteeg, W. H., Brouwer, J. H. and Helbig, K, 1988. High-

resolution 3- D reflection seismics on a tidal flat: acquisition, processing and
interpretation: First Break, Vol 6, pp 9–23.

Dingus, C., 2010. Seismic Processing-Noise Attenuation Techniques for relative

Amplitude Processing: Petroleum Africa, Technology and Solutions, 47 p.

Dobrin, M. B., and Savit, C. H., 1988. Introduction to Geophysical Prospecting,

(Vol. 4), New York, McGraw-hill, pp 21-30.

Donati, M. S. and Martin, N. W., 1995. Seismic reconstruction using a 3D tau-

p transform. CREWES Research Report. 2 p.

Dondurur, D., 2018. Acquisition and Processing of Marine Seismic Data,

Elsevier, pp 240-255.

Dragoset, B., 2005. A historical reflection on reflections, The Leading Edge, 49

p.

Elboth, T., Qaisrani, H. and Hertweck, T., 2008. De-noising seismic data in
the time-frequency domain, SEG Technical Program Expanded Abstracts, Vol
27(1), pp 2622-2626.

Essenreiter, R., Karrenbach, M., and Treitel, S., 1999. Multiple reflection
attenuation in seismic data using backpropagation, IEEE transactions on signal
processing, P 2.

Fadhil, M. S., 2010. Seismic Reflection Study of Lower Cretaceous Rocks in

Kifl Oil Field, Iraq, thesis of Master of Science in Geophysics (unpublished),
Geology Department, Baghdad University, 142 p.

Fomel, S., 2009. Velocity analysis using AB semblance. Geophysical

Prospecting. University of Texas. pp 311-321.
129
 REFERENCES

Forel, D., Benz, T. and Pennington, W. D., 2005. Seismic data processing with
Seismic Un*x, Society of Exploration Geophysicists, pp 26-274.

Gadallah, M. R., & Fisher, R. (2008). Exploration Geophysics, Springer

Science & Business Media., pp 18-81.

Gaiser, J. E., 1995. 3-D prestack fx coherent noise suppression, In SEG

Technical Program Expanded Abstracts, Society of Exploration Geophysicists,
pp 1354-1357.

Galbraith, M., 1991. Random noise attenuation by fx prediction: A tutorial. In

SEG Technical Program Expanded Abstracts, Society of Exploration
Geophysicists, pp 1428-1431.

Galway, R. S., 2000. The Integration of Multibeam Sonar Data with Huntec Sub-
bottom Profile Data into a Marine GIS, M.sc., published, Thesis, University of
British Columbia.

Hlavác, V., 2018. Fourier transform, in 1D and in 2D. Czech Institute of

Informatics, Robotics and Cybernetics. Czech Technical University.

Huang, W., Wang, R., Chen, Y., Li, H. and Gan, S. 2016. Damped
multichannel singular spectrum analysis for 3D random noise attenuation.
Geophysics, Vol 81, pp 261-270.

Henley, D.C., 2007. Radial filtering 3D data, CREWES Research Report,

Volume 19, 1p.

Kearey, P., Brooks, M., and Hill, I., 2002. An Introduction to Geophysical
Exploration. Blackwell science Ltd. Pp (45-51).

Kareem, R. M., 2005. Using filters in field processing of 3D seismic data to

improve signal to noise ratio (s/n) at Amara oil field in southeastern Iraq, Thesis
130
 REFERENCES

of Master of Science in Geophysics (unpublished), Geology Department,

Baghdad University.

Larner, K., Chambers. R., Yang. M., Lynn. W. and Wai. W., 1983. Coherent
noise in marine seismic data, Geophysics, Vol 48, pp 854-886.

Lee, M. W., Agena, V. F., Grow, J. A. and Miller J. J., 1999. Seismic
Processing and Velocity Assessments, Arctic National Wildlife Refuge, Alaska,
U. S. Geological Survey Open File Report, Vol. 8, No. 36, PP 98-34.

Li, Y. and Tang, D., 2005. Background noise identification and attenuation
using point receiver seismic data, SEG/Houston, Annual Meeting, 5 p.

Liu, G., Fomel, S., Jin, L. and Chen, X., (2009). Stacking seismic data using
local correlation. Geophysics, Vol 74, pp 43-48.

Liu, G. and Chen, X., 2013. Noncausal f–x–y regularized nonstationary

prediction filtering for random noise attenuation on 3D seismic data. Journal of
Applied Geophysics, Vol 93, pp 60-66.

Lorenzo, J.M., Hicks, J. and Vera, E., 2014. Integrated seismic and cone
penetration test observations at a distressed earthen levee: Marrero, Louisiana,
U.S.A., Elsevier Engineering Geology, Vol. 8, No. 168, PP 59-68.

Lowrie, W., 2007. Fundamentals of Geophysics, 2nd Edition, Cambridge

university press, New York, 374 p.

March, D.W. and Bailey, A.D., (1983). Two-dimensional transform and seismic
processing, First Break, pp 9-21.

Martinez, R. D., Kamps, W. H. and Takh, V. A. 2000. A robust method for

3D noise attenuation. In SEG Technical Program Expanded Abstracts, Society of
Exploration Geophysicists, pp 2077-2080.
131
 REFERENCES

Majeed, R. F., 2016. 3D seismic processing of Al-Samawa area, Oil exploration

company, Internal report.

Muhsin, S. U., 2012. 3D Seismic Structural Study of Subsurface in Balad Oil

Field – Center of Iraq, thesis of Master of Science in Geophysics (unpublished),
Geology Department, Baghdad University, 121 p.

Muchlis, M., 2015. Anomalous Amplitude Attenuation Method to Enhance

Seismic Resolution, Journal of Aceh Physics Society, 2 p.

Mukherjee, S. and Banerjee, D., 2006. A comparative study of the various wave
equation migration techniques applied to synthetic data, International Conference
& Exposition on Petroleum Geophysics Kolkata.

Murillo, A. E., 1996. Distributed Seismic Unix, M.Sc, Published Thesis, center
for wave phoneme, 12 p.

Öhman, I., Heikkinen, E. and Lehtimäki, T., 2006. Seismic 2D Reflection

Processing and Interpretation of Shallow Refraction Data, working report, Posiva
Ltd.

Oil Exploration Company (O.E.C), 2009, Final report of Gharraf project

(unpublished), seismic survey dept, 140 p. (In Arabic).

Oil Exploration Company (O.E.C), 2010. Final report of Nasiriya project

(unpublished), seismic survey dept, 132 p. (In Arabic).

Oil Exploration Company (O.E.C), 2011b. Final report of Badra project

(unpublished), seismic survey dept, 108 p. (In Arabic).

Oil Exploration Company (O.E.C), 2011c. Final report of Akkas project

(unpublished), seismic survey dept, 136 p. (In Arabic).

132
 REFERENCES

Oil Exploration Company (O.E.C), 2012a. Final report of Margan-West Kifl

project (unpublished), seismic survey dept, 206 p. (In Arabic).

Oil Exploration Company (O.E.C), 2012b. Final report of Mansuria project

(unpublished), seismic survey dept, 130 p. (In Arabic).

Oil Exploration Company (O.E.C), 2013. Final report of Anah-Graben project

(unpublished), seismic survey dept, 168 p. (In Arabic).

Oil Exploration Company (O.E.C), 2015. Final report of Al-Samawa project

(unpublished), seismic survey dept. 8 p (In Arabic).

Onajite, E., 2014. Seismic Data Analysis Techniques in Hydrocarbon

Exploration. Elsevier, pp 37-64.

Parkseis.

Website: http://www.parkseismic.com/PSBSACO.html

Percival, D., (2000). Applied Physics Lab, University of Washington, 16 p.

Rajput, S. and Rajput, S., 2006. Signal preserving seismic interference noise
attenuation on 3D marine seismic data. In SEG Technical Program Expanded
Abstracts, Society of Exploration Geophysicists, pp 2747-2751.

Roizman, M. D., 2013. Pre-Stack noise reduction in 3d-3c seismic data using
Common Offset Vectors Gathers. In 13th International Congress of the Brazilian
Geophysical Society & EXPOGEF, Rio de Janeiro, Brazil, Society of
Exploration Geophysicists and Brazilian Geophysical Society, pp 1453-1457.

Russell, B., Hampson, D. and Chun J., 1990. Noise elimination and the Radon
transform, part 1: The Leading Edge, Vol 9, No. 10, pp 18–23.

133
 REFERENCES

Schleicher, K., 2012. Open Data/Open Source: Seismic Unix scripts to process
a 2D land line, The University of Texas.

Schlumberger Oilfield Glossary. Stack

Website: https://www.glossary.oilfield.slb.com/en/Terms/s/stack.aspx

Seismic discussion.

Website:http://debu-interpretation.blogspot.com/2011/03/discussion-seismic-
data-processing.html

Seg.wiki (NMO).

Website: https://wiki.seg.org/wiki/Step-by-step_NMO_correction

Seg.wiki (Ground-roll).

Website: https://wiki.seg.org/wiki/Ground_roll

Seismic Unix Help 2005.

Website:
http://sepwww.stanford.edu/oldsep/cliner/files/suhelp/sufxdecon.txt

Shapiro, S. A., and Hubral P., (1999). Elastic waves in random media -
fundamentals of seismic stratigraphic filtering. Springer Lecture Notes in Earth
Sciences, 67 p.

Sheriff, R. E., and Geldart, L. P., 1995. Exploration Seismology. Cambridge

university press. 86 p.

Sheriff, R. E., 2002. Encyclopedic Dictionary of Applied Geophysics. Society of

exploration geophysicists. pp 80-378.

134
 REFERENCES

Sissakain, V.K., Mahdi, A.I., Amin, R.M. and Selman, B., 1997. The Nfayil
Formation, a new lithostratigraphic unit in the Western Desert of Iraq. Iraqi Geol.
Jour., Vol. 30, no.1, pp 61-65.

Soupios, P, M. and Toth, T., 2005. Seismic Reflection Tomography: A Case

Study of a Shallow Lake Survey in Lake Balaton, journal of Balkan geophysical
society, Vol. 8, No.1, pp 20-27.

Stewart, R. R., (1998). Air-noise reduction on geophone data using microphone

records, CREWES Res, Vol 10, pp 1-2.

Stockwell, J. W., and Cohen, J. K., 2008. The new SU user’s manual, Version
4.0, Center for Wave Phenomena, Colorado School of Mines, Golden, USA, pp
12-48.

Stone, D. G., 1998. Designing seismic surveys in two and three dimensions,
Society Exploration Geophysicists, pp 12-17.

Stork. C., Kapoor. J., Zhao. W., Dragoset. B., Dingwall. K, 2006. Predicting
and removing complex 3D surface multiples with WEM modelling-an alternative
to 3D SRME for wide azimuth surveys, Society of Exploration Geophysicist, pp
2006-2079.

Taner, M. T. and Koehler, F., 1969. Velocity spectra – digital computer

derivation and applications of velocity functions. Geophysics. P 34.

Telford, W. M., Geldart, L. P. and Sheriff R. E., 1990. Applied Geophysics,

2nd ed., Cambridge Univ. Press. Pp 34-50.

Todorov, T. I and Margrave, G. F. (2010). CDP noise attenuation using local

linear models. CREWES Research Report, Vol 80, 22 p.

135
 REFERENCES

Torres, J.A. and Goyes, Y.P., 2013. Seismic Processing a Synthetic Structural
Section Perijanero Flank Central, Central Basin of Venezuela, Researchgate.

Vermeer, G. J., 2002. 3-D seismic survey design. Society of Exploration

Geophysicists. p 18.

Western Geco Manual, 2015a. Anomalous Amplitude Attenuation 27.2

(ANOMALOUS-AMP-ATTEN).

Western Geco Manual, 2015b. Simple Tau-P Domain Filter 2.12

(TAU_P_FILT).

Western Geco Manual, 2015c. Forward Radon Transform 12.13 (Radon-FWD-

Transform).

Western Geco Manual, 2015d. F-X Coherent Noise Suppression 3.16

(FXCNS).

XGECO.

Website: http://www.xsgeo.com/course/basic.htm

Xie, H., Wu, C., Luo, C., Duan, W., Zhao, Y. and Liu, Q., 2015. Imaging
complex geology through challenging surface terrain–a case study from West
China. First Break, 61 p.

Yagci, F, G., 2016. Reprocessing and structural interpretation of multichannel

seismic reflection data from Penobscot, Nova Scotia, M.sc., Thesis, published,
Missouri university of science and technology.

Yilmaz, Ö. 2001. Seismic Data Analysis: Processing, Inversion, and

Interpretation of Seismic Data, Society of exploration geophysicists, pp 81-274.

136
 REFERENCES

Yu, M. C., 2012. Elimination of seismic interference noise using a 3D filtering

approach. In Istanbul 2012-International Geophysical Conference and Oil & Gas
Exhibition Society of Exploration Geophysicists and The Chamber of
Geophysical Engineers of Turkey, pp. 1-4.

137
Appendix A.
Bellow the acquisition acq.sh script used to generate synthetic seismic data in ch3.
The survey layout is:
1- The angles of the rays are (201 to -65), number of rays = 65, sampling rate =
0.004 and number of time samples = 1501 (line 7).
2- 120 shots (line 35).
3- Shots are equally spread at 50 m interval (line 39 and 42).
4- 60 split spread traces are recorded from each shot location (line 49).
5- Geophone spacing is 50 m (line 53 and 56).
6- Geophone offset range from -1475 to +1475 (line 59).
7- Shots locations range from 0 to 5.95 km (line 77).
8- Geophones locations range from -1.475 to 7.425 km (line 77).

1 #! /bin/sh
2 # File: acq4.sh
3 # Set messages on
4 ##set -x
5
6 # Assign values to variables
7 nangle=201 fangle=-65 langle=65 nt=1501 dt=0.004
8
9 # Model
10 num=1
11 echo " --Model number = $num"
12
13 # Name input model file
14 inmodel=model$num.dat
15
16 # Name output seismic file
17 outseis=seis${num}.su

138
18
19 # Remove survey file
20 rm -f survey${num}.txt
21 # Name survey file
22 survey=survey${num}.txt
23
24 #=================================================
25 # Create the seismic traces with "triseis"
26 # i-loop = 120 source positions
27 # j-loop = 60 geophone positions (split-spread)
28 # per shot position
29 # k-loop = layers 2 through 8
30 # (do not shoot layers 1 and 9)
31
32 echo " --Begin looping over triseis."
33
34 i=0
35 while [ "$i" -ne "120" ]
36 do
37
38 fs=`bc -l <<-END
39 $i * 0.05
40 END`
41 sx=`bc -l <<-END
42 $i * 50
43 END`
44 fldr=`bc -l <<-END
45 $i + 1

139
46 END`
47
48 j=0
49 while [ "$j" -ne "60" ]
50 do
51
52 fg=`bc -l <<-END
53 $i * 0.05 + $j *0.05
54 END`
55 gx=`bc -l <<-END
56 $i * 50 + $j * 50 - 1475
57 END`
58 offset=`bc -l <<-END
59 $j * 50 - 1475
60 END`
61 tracl=`bc -l <<-END
62 $i * 60 + $j + 1
63 END`
64 tracf=`bc -l <<-END
65 $j + 1
66 END`
67
68 echo " Sx=$sx Gx=$gx fldr=$fldr Offset=$offset tracl=$tracl\
69 fs=$fs fg=$fg"
70 echo " Sx=$sx Gx=$gx fldr=$fldr Offset=$offset tracl=$tracl\
71 fs=$fs fg=$fg" >> $survey
72
73 k=2

140
74 while [ "$k" -ne "11" ]
75 do
76
77 triseis < $inmodel xs=0,5.95 xg=-1.475,7.425 zs=0,0 zg=0,0 \
78 nangle=$nangle fangle=$fangle langle=$langle \
79 kreflect=$k krecord=1 fpeak=40 lscale=0.5 \
80 ns=1 fs=$fs ng=1 fg=$fg nt=$nt dt=$dt |
81 suaddhead nt=$nt |
82 sushw key=dt,tracl,tracr,fldr,tracf,trid,offset,sx,gx \
83 a=4000,$tracl,$tracl,$fldr,$tracf,1,$offset,$sx,$gx >> temp$k
84
85 k=`expr $k + 1`
86
87 done
88 j=`expr $j + 1`
89
90 done
91 i=`expr $i + 1`
92
93 done
94
95 echo " --End looping over triseis."
96
97 #=================================================
98
99 # Sum contents of the temp files
100 echo " --Sum files."
101 susum temp2 temp3 > tempa

141
102 susum tempa temp4 > tempb
103 susum tempb temp5 > tempc
104 susum tempc temp6 > tempd
105 susum tempd temp7 > tempe
106 susum tempe temp8 > tempf
107 susum tempf temp9 > tempg
108 susum tempg temp10 > $outseis
109
110 # Remove temp files
111 echo " --Remove temp files."
112 rm -f temp*
113
114 # Report output file
115 echo " --Output file ** $outseis **"
116
117 # Exit politely from shell script
118 echo " --Finished!"
119 exit

142
Below is the velocity analysis iva.sh script used in velocity analysis of the
synthetic data in ch3.
• The CMP from (60 to 240) used to velocity analysis is plotted from line 13
to line 18 with an increment of 10.
• The number of velocities, velocity intervals and first velocity are from line
38 to line 40.
• The inputs of suvelan program are from line 304 to line 311.

1 #! /bin/sh
2 # File: iva.sh
3 # Run script iva.scr to start this script
4
5 # Set messages on
6 ##set -x
7
8 #================================================
9 # USER AREA -- SUPPLY VALUES
10 #------------------------------------------------
11 # CMPs for analysis
12
13 cmp1=60 cmp2=70 cmp3=80
14 cmp4=90 cmp5=100 cmp6=110
15 cmp7=120 cmp8=130 cmp9=140
16 cmp10=150 cmp11=160 cmp12=170
17 cmp13=180 cmp14=190 cmp15=200
18 cmp16=210 cmp17=220 cmp18=230 cmp19=240
19
20 numCMPs=18
21

143
22 #------------------------------------------------
23 # File names
24
25 indata=cmp.su # SU format
26 outpicks=vpick.txt # ASCII file
27
28 #------------------------------------------------
29 # display choices
30
31 myperc=98 # perc value for plot
32 plottype=0 # 0 = wiggle plot, 1 = image plot
33
34 #------------------------------------------------
35 # Processing variables
36
37 # Semblance variables
38 nvs=100 # number of velocities
39 dvs=27 # velocity intervals
40 fvs=1200 # first velocity
41
42 # CVS variables
43 fc=1200 # first CVS velocity
44 lc=3900 # last CVS velocity
45 nc=10 # number of CVS velocities (panels)
46 XX=11 # ODD number of CMPs to stack into central CVS
47
48 #================================================
49

144
50 # HOW SEMBLANCE (VELAN) VELOCITIES ARE COMPUTED
51
52 # Last Vel = fvs + (( nvs-1 ) * dvs ) = lvs
53 # 5000 = 500 + (( 99-1 ) * 45 )
54 # 3900 = 1200 + (( 100-1 ) * 27 )
55
56 # Compute last semblance (velan) velocity
57 lvs=`bc -l << -END
58 $fvs + (( $nvs - 1 ) * $dvs )
59 END`
60
61 #------------------------------------------------
62
63 # HOW CVS VELOCITIES ARE COMPUTED
64
65 # dc = CVS velocity increment
66 # dc = ( last CVS vel - first CVS vel ) / ( # CVS - 1 )
67 # m = CVS plot trace spacing (m = d2, vel units)
68 # m = ( last CVS vel - first CVS vel ) / ( ( # CVS - 1 ) * XX )
69
70 # j=1
71 # while [ j le nc ]
72 # do
73 # vel = fc + { [( lc - fc ) / ( nc-1 )] * ( j-1) }
74 # j = j + 1
75 # done
76 # EXAMPLE:
77 # vel = 1200 + ( (( 3900 - 1200 ) / ( 10-1 )) * ( 1-1) )

145
78 # vel = 1200 + ( (( 3900 - 1200 ) / ( 10-1 )) * ( 2-1) )
79 # .
80 # .
81 # .
82 # vel = 1200 + ( (( 3900 - 1200 ) / ( 10-1 )) * (11-1) )
83
84 #================================================
85
86 # FILE DESCRIPTIONS
87
88 # tmp0 = binary temp file for input CVS gathers
89 # tmp1 = binary temp file for output CVS traces
90 # tmp2 = ASCII temp file for managing picks
91 # tmp3 = binary temp file for stacked traces
92 # tmp4 = ASCII temp file for "wc" result (velan)
93 # tmp5 = ASCII temp file for stripping file name from tmp4 (velan)
94 # tmp6 = ASCII temp file to avoid screen display of "zap"
95 # tmp7 = ASCII temp file for picks
96 # tmp8 = binary temp file for NMO (flattened) section
97 # panel.$picknow = current CMP windowed from line of CMPs
98 # picks.$picknow = current CMP picks arranged as "t1 v1"
99 # "t2 v2"
100 # etc.
101 # par.# (# is a sequential index number; 1, 2, etc.)
102 # = current CMP picks arranged as
103 # "tnmo=t1,t2,t3,...
104 # "vnmo=v1,v2,v3,...
105 # par.uni.# (# is a sequential index number; 1, 2, etc.)

146
106 # = current CMP picks arranged as
107 # "xin=t1,t2,t3,...
108 # "yin=v1,v2,v3,...
109 # for input to xgraph to display velocity profile
110 # par.cmp = file of CMP number and sequential index number;
111 # for example: "40 1"
112 # "60 2"
113 # etc.
114 # par.0 = file "par.cmp" re-arranged as
115 # "cdp=#,#,#,etc." NOTE: # in this line is picked CMP
116 # "#=1,2,3,etc." NOTE: # in this line is "#"
117 # outpicks = concatenation of par.0 and all par.# files.
118
119 #================================================
120
121 echo " "
122 echo " *** INTERACTIVE VELOCITY ANALYSIS ***"
123 echo " "
124
125 #------------------------------------------------
126 # Remove old files. Open new files
127 rm -f panel.* picks.* par.* tmp*
128
129 > $outpicks # Write empty file for final picks
130 > par.cmp # Write empty file for recording CMP values
131
132 #------------------------------------------------
133 # Get ns, dt, first time from seismic file

147
134 nt=`sugethw ns < $indata | sed 1q | sed 's/.*ns=//'`
135 dt=`sugethw dt < $indata | sed 1q | sed 's/.*dt=//'`
136 ft=`sugethw delrt < $indata | sed 1q | sed 's/.*delrt=//'`
137
138 # Convert dt from header value in microseconds
139 # to seconds for velocity profile plot
140 dt=`bc -l << -END
141 scale=6
142 $dt / 1000000
143 END`
144
145 # If "delrt", use it; else use zero
146 if [ $ft -ne 0 ] ; then
147 tstart=`bc -l << -END
148 scale=6
149 $ft / 1000
150 END`
151 else
152 tstart=0.0
153 fi
154
155 #------------------------------------------------
156
157 # Initialize "repick" -- for plotting previous picks on velan
158 repick=1 # 1=false, 0=true
159
160 #------------------------------------------------
161 # BEGIN IVA LOOP

148
162 #------------------------------------------------
163
164 i=1
165 while [ $i -le $numCMPs ]
166 do
167
168 # set variable $picknow to current CMP
169 eval picknow=\$cmp$i
170
171 if [ $repick -eq 1 ] ; then
172 echo " "
173 echo "Preparing CMP $i of $numCMPs for Picking "
174 echo "Location is CMP $picknow "
175 fi
176
177 #------------------------------------------------
178 # Plot CMP (right)
179 #------------------------------------------------
180
181 suwind < $indata \
182 key=cdp min=$picknow max=$picknow > panel.$picknow
183 if [ $repick -eq 1 ] ; then
184 if [ $plottype -eq 0 ] ; then
185 suxwigb < panel.$picknow xbox=634 ybox=10 wbox=300 hbox=450 \
186 title="CMP gather $picknow" \
187 label1=" Time (s)" label2="Offset (m)" key=offset \
188 perc=$myperc verbose=0 &
189 else

149
190 suximage < panel.$picknow xbox=634 ybox=10 wbox=300 hbox=450 \
191 title="CMP gather $picknow" \
192 label1=" Time (s)" \
193 perc=$myperc verbose=0 &
194 fi
195 else
196 if [ $plottype -eq 0 ] ; then
197 suxwigb < panel.$picknow xbox=946 ybox=10 wbox=300 hbox=450 \
198 title="CMP gather $picknow" \
199 label1=" Time (s)" label2="Offset (m)" key=offset \
200 perc=$myperc verbose=0 &
201 else
202 suximage < panel.$picknow xbox=946 ybox=10 wbox=300 hbox=450 \
203 title="CMP gather $picknow" \
204 label1=" Time (s)" \
205 perc=$myperc verbose=0 &
206 fi
207 fi
208
209 #------------------------------------------------
210 # Constant Velocity Stacks (CVS) (middle-left)
211 # Make CVS plot for first pick effort.
212 # If re-picking t-v values, do not make this plot.
213 #------------------------------------------------
214
215 # repick: 1=false, 0=true
216 if [ $repick -eq 1 ] ; then
217

150
218 # number of CMPs - 1; for windowing
219 X=`expr $XX - 1`
220
221 # Window CMPs around central CMP (+/- X/2). Write to tmp0
222 k1=`expr $picknow - $X / 2` # Window from CMP to CMP - X/2
223 k2=`expr $picknow + $X / 2` # Window from CMP to CMP + X/2
224 suwind < $indata key=cdp min=$k1 max=$k2 > tmp0
225
226 # Calculate CVS velocity increment
227 # dc = ( last CVS vel - first CVS vel ) / ( # CVS - 1 )
228 dc=`bc -l << -END
229 ( $lc - $fc ) / ( $nc - 1 )
230 END`
231
232 # Calculate trace spacing for CVS plot (m = d2, vel units)
233 # m = ( last CVS vel - first CVS vel ) / ( ( # CVS - 1 ) * XX )
234 m=`bc -l << -END
235 ( $lc - $fc ) / ( ( $nc - 1 ) * $XX )
236 END`
237
238 # CVS velocity loop
239 j=1
240 while [ $j -le $nc ]
241 do
242
243 vel=`bc -l << -END
244 $fc + $dc * ( $j - 1 )
245 END`

151
246
247 # uncomment to print CVS velocities to screen
248 ## echo " vel = $vel"
249
250 sunmo < tmp0 vnmo=$vel verbose=0 |
251 sustack >> tmp1
252
253 j=`expr $j + 1`
254 done
255
256 # Compute lowest velocity for annotating CVS plot
257 # loV = first CVS velocity - ( ( CMP range - 1 ) / 2 ) * vel inc
258 loV=`bc -l << -END
259 $fc - ( $X / 2) * $m
260 END`
261
262 suximage < tmp1 xbox=322 ybox=10 wbox=300 hbox=450 \
263 title="CMP $picknow Constant Velocity Stacks" \
264 label1=" Time (s)" label2="Velocity (m/s)" \
265 f2=$loV d2=$m verbose=0 \
266 perc=$myperc n2tic=5 cmap=rgb0 &
267
268 fi
269
270 #------------------------------------------------
271 # Picking instructions
272 #------------------------------------------------
273

152
274 echo " "
275 echo "Preparing CMP $i of $numCMPs for Picking "
276 echo "Location is CMP $picknow "
277 echo " Start CVS CMP = $k1 End CVS CMP = $k2"
278 echo " "
279 echo " Use the semblance plot to pick (t,v) pairs."
280 echo " Type \"s\" when the mouse pointer is where you want a pick."
281 echo " Be sure your picks increase in time."
282 echo " To control velocity interpolation, pick a first value"
283 echo " near zero time and a last value near the last time."
284 echo " Type \"q\" in the semblance plot when you finish picking."
285
286 #------------------------------------------------
287 # Plot semblance (velan) (left)
288 #------------------------------------------------
289
290 # repick: 1=false, 0=true
291 if [ $repick -eq 0 ] ; then
292
293 #--- --- --- --- --- --- --- --- --- ---
294 # Get the number of picks (number of lines) in tmp7 |
295 # Remove blank spaces preceding the line count.
296 # Remove file name that was returned from "wc".
297 # Store line count in "npair" to guide line on velan.
298
299 wc -l tmp7 | sed 's/^ *\(.*\)/\1/' > tmp4
300 sed 's/tmp7//' < tmp4 > tmp5
301 npair=`sort < tmp5`

153
302 #--- --- --- --- --- --- --- --- --- ---
303
304 suvelan < panel.$picknow nv=$nvs dv=$dvs fv=$fvs |
305 suximage xbox=10 ybox=10 wbox=300 hbox=450 perc=99 \
306 units="semblance" f2=$fvs d2=$dvs n2tic=5 \
307 title="Semblance Plot CMP $picknow" cmap=hsv2 \
308 label1=" Time (s)" label2="Velocity (m/s)" \
309 legend=1 units=Semblance verbose=0 gridcolor=black \
310 grid1=solid grid2=solid mpicks=picks.$picknow \
311 curve=tmp7 npair=$npair curvecolor=white
312
313 else
314
315 suvelan < panel.$picknow nv=$nvs dv=$dvs fv=$fvs |
316 suximage xbox=10 ybox=10 wbox=300 hbox=450 perc=99 \
317 units="semblance" f2=$fvs d2=$dvs n2tic=5 \
318 title="Semblance Plot CMP $picknow" cmap=hsv2 \
319 label1=" Time (s)" label2="Velocity (m/s)" \
320 legend=1 units=Semblance verbose=0 gridcolor=black \
321 grid1=solid grid2=solid mpicks=picks.$picknow
322
323 fi
324
325 #------------------------------------------------
326 # End first set of plots
327 #================================================
328
329 #------------------------------------------------

154
330 # Manage picks (1): Prepare picks for sunmo
331 #------------------------------------------------
332
333 sort < picks.$picknow -n |
334 mkparfile string1=tnmo string2=vnmo > par.$i
335 echo "cdp=$picknow" >> tmp2
336 cat par.$i >> tmp2
337
338 #================================================
339 # Begin second set of plots
340 #------------------------------------------------
341
342 #------------------------------------------------
343 # Flattened seismic data (NMO) plot (middle-right)
344 #------------------------------------------------
345
346 sunmo < panel.$picknow par=tmp2 verbose=0 > tmp8
347 if [ $plottype -eq 0 ] ; then
348 suxwigb < tmp8 xbox=634 ybox=10 wbox=300 hbox=450 \
349 title="CMP $picknow after NMO" \
350 label1=" Time (s)" label2="Offset (m)" \
351 verbose=0 perc=$myperc key=offset &
352 else
353 suximage < tmp8 xbox=634 ybox=10 wbox=300 hbox=450 \
354 title="CMP $picknow after NMO" \
355 label1=" Time (s)" \
356 verbose=0 perc=$myperc &
357 fi

155
358
359 #------------------------------------------------
360 # Stack window (right)
361 #------------------------------------------------
362
363 j=1
364 while [ $j -le 8 ]
365 do
366
367 # Append stack trace into tmp3 multiple times
368 sustack < tmp8 >> tmp3
369
370 j=`expr $j + 1`
371 done
372
373 suxwigb < tmp3 xbox=946 ybox=10 wbox=200 hbox=450 \
374 title="CMP $picknow repeat stack trace" \
375 label1=" Time (s)" d2num=50 key=cdp \
376 verbose=0 perc=$myperc &
377
378 #------------------------------------------------
379 # Manage picks (2): Prepare picks for vel profile
380 #------------------------------------------------
381
382 sed < par.$i '
383 s/tnmo/xin/
384 s/vnmo/yin/
385 ' > par.uni.$i

156
386
387 #------------------------------------------------
388 # Velocity profile (left)
389 #------------------------------------------------
390
391 unisam nout=$nt fxout=$tstart dxout=$dt \
392 par=par.uni.$i method=mono |
393 xgraph n=$nt nplot=1 d1=$dt f1=$tstart x2beg=$fvs x2end=$lvs \
394 label1=" Time (s)" label2="Velocity (m/s)" \
395 title="CMP $picknow Stacking Velocity Function" \
396 -geometry 300x450+10+10 -bg white style=seismic \
397 grid1=solid grid2=solid linecolor=2 marksize=1 mark=0 \
398 titleColor=black axesColor=blue &
399
400 #------------------------------------------------
401 # Dialogue with user: re-pick ?
402 #------------------------------------------------
403
404 echo " "
405 echo " t-v PICKS CMP $picknow"
406 echo "----------------------"
407 cat picks.$picknow
408 echo " "
409 echo " Use the velocity profile (left),"
410 echo " the NMO-corrected gather (middle-right),"
411 echo " and the repeated stack trace (right)"
412 echo " to decide whether to re-pick the CMP."
413 echo " "

157
414 echo "Picks OK? (y/n) " > /dev/tty
415 read response
416
417 rm tmp*
418
419 # "n" means re-loop. Otherwise, continue to next CMP.
420 case $response in
421 n*)
422 i=$i
423 echo " "
424 echo "Repick CMP $picknow. Overlay previous picks."
425 repick=0
426 cp picks.$picknow tmp7
427 ;;
428 *)
429 echo "$picknow $i" >> par.cmp
430 i=`expr $i + 1`
431 repick=1
432 echo "-- CLOSING CMP $picknow WINDOWS --"
433 zap xwigb > tmp6
434 zap ximage > tmp6
435 zap xgraph > tmp6
436 ;;
437 esac
438
439 done
440
441 #------------------------------------------------

158
442 # Create velocity output file
443 #------------------------------------------------
444
445 mkparfile < par.cmp string1=cdp string2=# > par.0
446
447 i=0
448 while [ $i -le $numCMPs ]
449 do
450 sed < par.$i 's/$/ \\/g' >> $outpicks
451 i=`expr $i + 1`
452 done
453
454 #------------------------------------------------
455 # Remove files and exit
456 #------------------------------------------------
457 echo " "
458 echo " The output file of t-v pairs is "$outpicks
459 pause
460 rm -f panel.* picks.* par.* tmp*
461 exit
462

159
 ‫الخالصة‬
‫الهدف من هذا البحث هو تحسين نسبة األشارة الى الضوضاء للبيانات الزلزالية بأستخدام طرق و مرشحات‬
‫مختلفة‪ ,‬لغرض ازالة اكثر ما يمكن ازالته من الضوضاء العشوائية والمتجانسة من دون التأثير على األشارة‪.‬‬
‫نوعان من البيانات الزلزالية تم استخدامها ومعالجتها في هذه الرسالة‪ ,‬بيانات زلزالية صناعية تم توليدها‬
‫بأستخدام منضومة اليونكس الزلزالية‪ ,‬وبيانات زلزالية ثالثية األبعاد لمنطقة السماوة والتي تبعد ‪ 082‬كم جنوب‬
‫شرق بغداد‪ .‬ان البيانات الزلزالية الثالثية األبعاد تم تسجيلها من قبل الفرقة الزلزالية الثالثة التابعة لوزارة النفط‬
‫العراقية في سنة ‪.0202‬‬
‫تم توليد البيانات الزلزالية الصناعية بأستخدام منضمومة اليونكس الزلزالية باألعتماد على مقطع زلزالي من‬
‫البيانات الزلزالية الثالثية األبعاد التابعة لمنطقة السماوة‪ .‬في البداية تم بناء موديل ارضي‪ ,‬ثم تم توليد ‪002‬‬
‫شوتة من البيانات الزلزالية من خالل الموديل األرضي‪ ,‬ثم تم معالجة البيانات الحقا ً من خالل طرق ومرشحات‬
‫عديدة بعد عملية تصحيح السعات المفقودة واضافة الضوضاء عشوائية بمقدار نسبة اشارة الى ضوضاء ‪%02‬‬
‫اي (‪ %02‬اشارة – ‪ %02‬ضوضاء)‪ ,‬بحيث ان البيانات الصناعية ال تحتوي على اي ضوضاء‪ .‬ثالثة مرشحات‬
‫مختلفة تم استخدامها ألزالة الضوضاء العشوائية التي تمت اضافتها‪ ,‬وهي مرشح األحتواء العكسي في نطاق‬
‫المسافة والتردد‪ ,‬المرشح الترددي و المرشح األحتوائي‪ ,‬وان النتائج التي تم الحصول عليها من خالل تطبيق‬
‫هذه الفالتر اعطى نتائج سيئة في البداية على الرغم من ان كمية كبيرة من الضوضاء العشوائية تمت ازالتها‪,‬‬
‫لكن جزء من األشارة تمت ازالتها ايضاً‪ .‬وبعد تجارب عديدة تم اكتشاف ان تطبيق عملية األنتقال العادي قبل‬
‫تطبيق ال مرشحات الثالثة سيعطي انفصال اكبر بين األشارة والضوضاء والتي نتج عنها ترشيح افضل وحل‬
‫لمشكلة األشارة المفقودة‪.‬‬
‫الجزء الثاني لهذه الرسالة سوف يتعامل مع معالجة وتوهين الضوضاء للبيانات الزلزالية الثالثية األبعاد لمنطقة‬
‫السماوة‪ .‬لتحسين نسبة األشارة الى الضوضاء‪ ,‬تم تطبيق مرشحات عديدة بأستخدام منضومة اوميغا‪ .‬في البداية‬
‫تم تحويل صيغة البيانات الزلزالية لصيغة منضومة اوميغا‪ ,‬و من ثم تم تحديث البيانات الهندسية لرؤوس‬
‫البيانات الزلزالية‪ ,‬وبعدها تم تعويض السعات المفقودة الناتجة من ظاهرة األنتشار الكروي‪.‬‬
‫تم البدء بمرحلة توهين الضوضاء بأستخدام مرشح توهين السعات الشاذة ألزالة الضوضاء الزلزالية الشاذة‬
‫والغير طبيعية‪ .‬وبسبب ان البيانات تحتوي على كمية كبيرة من الضوضاء العشوائية‪ ,‬تم تطبيق المرشح عدة‬
‫مرات بأستخدام مدخالت مختلفة‪ ,‬ثالث مرات في التطبيق األول و ‪ 6‬مرات في التطبيق الثاني بعد ان تم تصحيح‬
‫السعات السطحية‪ .‬وتم ساتعمال مرشح (تاو ب) ألزالة الضوضاء الخطية المتجانسة من خالل تحويل البيانات‬
‫الى نطاق ال (تاو ب) بعد تطبيق عملية األنتقال الطبيعي و عكسه‪ ,‬و من ثم ازالة الضوضاء يدوياً‪.‬‬
‫تم تطبيق مرشح ازالة الضوضاء المتجانسة في نطاق التردد والمسافة ألزالة الموجات األرضية السطحية‬
‫بأستخدام معامالت السرعة والتردد‪ .‬بعد تقسيم البيانات الى لوحات ترددية‪ ,‬تم اكتشاف ان الموجات السطحية‬
‫تقع ضمن نطاق ترددي بين (‪ 02-3‬او ‪ )00‬وان المرشح اعطى نتائج ممتازة بما يخص ازالة الموجات اسطحية‪.‬‬
‫اما الضوضاء الخطية المتبقية فتم ازالتها بأستخدام مرشح التردد والطول الموجي من خالل تحويل البيانات‬
‫الى نطاق التردد والطول الموجي ومن ثم ازالة الضوضاء بأستخدام مرشح المروحة‪.‬‬
‫واخيرا ً تم استخدام مرشح الرادون ألزالة الضوضاء المتقاطعة من خالل استهداف المرشح لهذا النوع من‬
‫الضوضاء وازالتها يدوياً‪ .‬الجدير بالذكر ان عملية األنتقال الطبيعي تم تطبيقها قبل مرشح التردد والطول‬
‫الموجي ومرشح الرادون ومن ثم عكسه بعد عملية ازالة الضوضاء‪ .‬ان النتائج النهائية كانت ممتازة واظهرت‬
‫تحسين واضح للبيانات الزلزالية والتي يمكن التأكد منها من خالل طيف األشارة والضوضاء و الشريحة‬
‫الزمنية‪.‬‬
 ‫جمهورية العراق‬
‫وزارة التعليم العالي والبحث العلمي‬
‫جامعة بغداد‬
‫كلية العلوم‬
‫قسم علم االرض‬

‫أزالة الضوضاء للتسجيل الثالثي األبعاد في منطقة السماوة‬

‫جنوب غرب العراق‬
‫رسالة مقدمة الى‬
‫كلية العلوم ‪ -‬جامعة بغداد‬
‫كجزء من متطلبات درجة الماجستير‬
‫في علم األرض – الجيوفيزياء‬

‫من قبل‬
‫أحمد حسين علي‬
‫بكلوريوس علم األرض ‪6102‬‬

‫أشراف‬

‫أ‪.‬د‪ .‬علي مكي الرحيم‬

‫‪ 2020‬م‬ ‫‪ 1440‬هـ‬