SEISMIC DATA PROCESSING for Oil for Oil & Gas Exploration HAGI Guest Lecturing Program
Universitas Brawijaya Malang, 15 March 2014
Teguh Suroso HAGI – Pertamina UTC
Outline
Introduction
Fundamentals
Concepts
Seismic data processing in practice
Advanced processing
1
INTRODUCTION
Seismic Products Advanced Processing Processing
Basic Processing
Final CMP Gathers
Field records
Time migrated section
Acquisition Time migrated gather
Velocity model building
Depth migrated section
Well seismic tie
Acoustic Impedance section
AVO analysis analysis
Intercept-Gradient section
2
Seismic DP Purposes Diephuis , 2008 2008 : To transform the raw field records into an interpretable volume/line depicting reflection coefficient in the subsurface
Gluyas & Sw arbrick , 200 2004 4: To enhance the interpretable (useful) seismic information relative to the noise in the signal and place the reflectors reflectors in their correct x,y,z x,y,z space
IPIMS, 2010 : The main goal of seismic processing processing is to obtain the best image of the subsurface.
Reservoir characterization: - AVO and Inversion
Sample of Field Record Shot Point Ch-1
Ch-n
3
Geology Model
Field Record – along the line Displayed in every 10 SP
4
Overlay field records with geology model NOT interpretable data
Overlay field records with geology model Seismic imaging, final product of processing
5
Interpretable data
Seismic imaging (stacked trace)
Field record
6
Seismic imaging (stacked trace)
Field record
FUNDAMENTALS
7
Seismic Wave
Seismic wave is a sound wave
Wave propagation is three dimensional phenomenon Type of seismic wave P‐wave Body wave S‐wave Seismic wave Love wave Surface wave Rayleigh wave
Seismic Wave Ilustration Body waves Propagate through the Earth’s interior
a. P‐wave > Compressional wave = longitudinal wave > Propagates in solids, liquids and gasses b. S‐wave > Shear wave = transversal > Propagates in solids only Surface waves Propagates along the Earth’s surface
c. Love wave > low velocity layer overlaying high velocity layer d. Rayleigh wave > ground roll
8
Body Wave Velocity Comparison
Propagation of P‐wave Propagation of S‐wave S‐waves propagate more slowly than P‐waves Vs < Vp
Wave Propagation source
surface
Isotropic media Wavefront -Surface of equal time
Ray path -Line everywhere perpendicular to wavefront
9
P‐wave If P-wave strikes a boundary between two media with different velocities of propagation and/or different densities, the P-wave will be : reflected, transmitted , and converted into reflected and transmitted S-wave The sum of the reflected and transmitted amplitudes is equal to the incident amplitude.
Reflection & Refraction If amplitude of incident wave = A0 amplitude of reflected wave = A1, and amplitude of transmitted wave = A2 A0 = A1 + A2 Relative size of the reflected and the transmitted amplitudes depend on The contrast in acoustic impedance Aco us ti c Im pedan ce (AI ) AI = ρ . V ρ = density V = P-wave velocity
Reflection Coeffici ent (RC) R = A1/A0
Transmis ion Coeffici ent (TC) T = A2/A0 T =1 -R
10
Snell’s Law the reflected angle is equal to the incident angle
Head wave
critical angle
11
Huygens’ Principl e Every point on an advancing wavefront is a new source of spherical wave. Huygens’ Principle provides a mechanism by which a propagating seismic pulse loses energy with depth. Seismic waves propagate away from the source : - the wavefront become larger - the surface become larger - energy per unit area become smaller Spherical (geometrical) spreading Seismic amplitudes are proportional to the square root of energy per unit area.
Fermat’s Principle A light ray traveling from one point to another will follow a path such that, compared with nearby paths, the time required is either a minimum or a maximum or will remain unchanged (Danbom, 2007) Minimum time path (Diephuis, 2008)
12
CONCEPTS
Consider a single sine wave of 30Hz In a medium of 2500m/s
83m
13
Resolusi
Menurut Rayleigh, agar dapat resolved: ketebalan lapisan harus setidak-tidaknya ¼
(Sherriff, 1997)
Resolusi Data Seismik Harris dan Langan (1991) Dalam kaitannya dengan resolusi vertikal:
Data seismik antar ‐ sumur mengisi gap antara VSP dan log sonik Resolusi maksimum : ~1 m Fraksi reservoir: 10‐2 – 10‐5
14
Harris dan Langan (1991): Perbandingan resolusi seismik‐ permukaan, seismik antar‐sumur dan log sonik
15
Convolutional Model M O D EL I N G
AI(t)
RC(t)
W(t)
S(t)
*
Geologic model
INVERSION
Convolutional Model
AI(t)
RC(t)
W(t)
S(t)
*
Geologic model
16
Convolutional Model
AI(t)
RC(t)
W(t)
S(t)
* Geologic model
Convolutional Model
AI(t)
RC(t)
S(t)
Geologic model
17
Convolutional Model
AI(t)
RC(t)
S(t)
Geologic model
Convolutional Model
AI(t)
RC(t)
S(t)
Geologic model
18
Convolutional Model
AI(t)
RC(t)
S(t)
Geologic model
Convolutional Model
AI(t)
RC(t)
S(t)
Geologic model
19
Convolutional Model
Seismic trace S(t) = RC(t) * W(t) + n(t) n(t) = noise
FORWARD MODELING
AI(t)
RC(t)
W(t)
S(t)
*
Geologic model
INVERSION
Signal Domain Time domain - Amplitude vs Time
Signal
Frequency domain - Amplitude vs Frequency - Phase vs Frequency
20
Relationship Time‐Frequency Inverse Fourier Transform sum
decompose Fourier Transform
Single frequency sinusoids
sum
-
A TIME domain wavelet can be synthesized by summing a set of single FREQUENCY sinusoids
-
A TIME domain wavelet can be decomposed into a set of single FREQUENCY sinusoids
21
22
23
24
25
Semakin banyak frequency contain-nya, gelombang seismik akan semakin spike, Sehingga daya-pisahnya semakin besar. Simple quiz: Gambarkan bagaimana kira-kira sketsa spektrum amplitude-nya!
26
Relations hip betw een Time and Frequency domain
Inverse Fourier Transform sum
Single frequency sinusoids
sum
-
A TIME domain wavelet can be synthesized by summing a set of single FREQUENCY sinusoids
-
A TIME domain wavelet can be decomposed into a set of single FREQUENCY sinusoids
Relations hip betw een Time and Frequency domain
decompose Fourier Transform
Single frequency sinusoids
sum
-
A TIME domain wavelet can be synthesized by summing a set of single FREQUENCY sinusoids
-
A TIME domain wavelet can be decomposed into a set of single FREQUENCY sinusoids
27
Relations hip betw een Time and Frequency domain
Inverse Fourier Transform sum
decompose Fourier Transform
Single frequency sinusoids
sum
-
A TIME domain wavelet can be synthesized by summing a set of single FREQUENCY sinusoids
-
A TIME domain wavelet can be decomposed into a set of single FREQUENCY sinusoids
Time Domain Change in amplitude with TIME at a particular LOCATION
T-domain (time)
Seismic signal Change in amplitude with DISTANCE at a particular TIME
X-domain (space)
Time domain -
Period (T)
= time required to complete one cycle
-
Frequency (F)
= number of cycle/second
Space domain -
Wavelength ( λ ) = distance required to complete one cycle
-
Wavenumber (k) = number of cycle/unit distance
28
Time Domain Xmin
Offset
Xmax
T i m e
Transformation T‐X to F‐K aliased
T-X domain
F-K domain
Signal is crossed by noise in T-X plane but separated in F-K plane
29
Source strength
absorption scattering
Curved reflector Amplitude variation with angle (AVA)
Dynamic range
Receiver response
Receiver strength Geophone arrays
30
SEISMIC DATA PROCESSING IN PRACTICE
Processing Stages Data Preparation PREProcessing Pre-Migration Migration Post-Migration Archieving
31
Processing Stages Data Preparation PREProcessing Pre-Migration Migration Post-Migration
No.
Items
Remark
1
Survey type
3-D Land
2
Seismic data available
Raw record from field tape
3
Data format
SEG-D
4
Observer report available
Softcopy & hardcopy
5
Geometry/navigation data available
SPS
6
Field data available
Elevation, Uphole time
7
Signature available for marine survey
softcopy
8
Acquisition report available
hardcopy
9
Field/On Board processing report available
softcopy
10
Data legacy (from old process) available
Poststack time migration volume (SEG-Y) from Elnusa. Powerpoint slides with interpretated lines
11
Other supporting data available
-Well -Horizon interpretation
Archieving
Standard Seismic Data Processing (Pre-Migration) Reformating
Reformating
Geometry Assignment
Seismic-Navigation Merge
Trace Editing/Denoise Geometric Spreading (Amp)Corr.
Trace Editing/Denoise
PREProcessing
Geometric Spreading (Amp)Corr.
Statics Correction
Tidal Correction
Denoise
Swell Noise Attenuation, Linear Noise Attenuation
Deconvolution
Designature
Velocity Analysis-1
Tau-p Deconvolution
Residual Statics Correction-1
SRME (if necessary)
Velocity Analysis-2 Residual Statics Correction-2 Surface Consistent Amplitude Corr.
CMP Gathers
Pre-Migration
Velocity Analysis Demultiple (Hi-res Radon) Surface Consistent Amplitude Corr. CMP Gathers
32
Standard Seismic Data Processing for Land2 CMP Gathers
NMO, muting & Stacking Migration Precondition Post Stack Time Migration
Offset Regularization Migration Precondition Pre-Stack Time Migration
NMO & muting & Stacking Migration Precondition
Pre-Stack Depth Migration
Post Stack Depth Migration
Depth to Time Conversion
Depth to Time Conversion
Velocity Analysis
Velocity Analysis
NMO, muting & Stacking
NMO, muting & Stacking
Post Stack Processing : Noise Attn, Filter, Scaling
Post Stack Processing : Noise Attn, Filter, Scaling
Post Stack Processing : Noise Attn, Filter, Scaling
Post Stack Processing : Noise Attn, Filter, Scaling
Datum Correction
Datum Correction
Datum Correction
Datum Correction
Time to Depth Conversion
Time to Depth Conversion
Final Volume/Line
Final Volume/Line Volume/Line
Final Volume/Line
Final Volume/Line
Standard Seismic Data Processing for Marine CMP Gathers
NMO, muting & Stacking Migration Precondition Post Stack Time Migration
Post Stack Processing : Noise Attn, Filter, Scaling
Gun & Cable Correction
Final Volume/Line
Offset Regularization Migration Precondition Pre-Stack Time Migration
NMO & muting & Stacking Migration Precondition
Pre-Stack Depth Migration
Post Stack Depth Migration
Depth to Time Conversion
Depth to Time Conversion
Velocity Analysis
Velocity Analysis
Residual (Radon) Demultiple
Residual (Radon) Demultiple
NMO, muting & Stacking
NMO, muting & Stacking
Post Stack Processing : Noise Attn, Filter, Scaling
Post Stack Processing : Noise Attn, Filter, Scaling
Post Stack Processing : Noise Attn, Filter, Scaling
Time to Depth Conversion
Time to Depth Conversion
G Gun un & C Cable ab le Correction Co rr ec ti on
Gun G un & & Cable C ab le Correction Co rr ec ti on
Gun & Cable Correction Correction Final Volume/Line
Final Volume/Line
Final Volume/Line Volume/Line
33
Seismic Data Processing In Practice
PREPROCESSING
Reformat
34
Refor mat: review lo aded data
Data Sampling Continuous Analog Signal
Digitized Signal
Reconstructed Signal
35
Frequency Aliasing Input Output : 1ms sampling Output : 2ms sampling Output : 4ms sampling Output : 8ms sampling
aliased Memastikan sebelum resampling data di-Hi-Cut filter sekitar Frekuensi Nyquist
Geometry assignment Geometry assignment -Geometry update. -Trace labelling. -Assign unique numbers. -Specify coordinate for all source & receiver position.
Data must be updated with the correct geometry. The wrong geometry assigned will be very fatal. The processing can not be continued to the next step if the geometry is not correctly updated.
36
Incorrect geometry Stack section with GEOMETRY ERROR
Survey coverage
37
SPS – Header file
Geometry/Navigation file
SPS – Geometry file (XPS)
SPS – Source file (SPS)
Geometry/Navigation file
SPS – Receiver file (RPS)
38
Geometry QC Shot Point Gather with LMO applied showing GEOMETRY ERROR
Geometry QC Shot Point Gather with LMO applied showing CORRECTED GEOMETRY
39
Incorrect binning
Distribusi fold kurang merata
Correct binning
Distribusi fold lebih merata
40
Trace editing Trace editing is the process of removing or correcting any traces or records which, in their originally recorded form, may cause a deterioration of the stack. Individual traces may be affected by polarity reversals or by noise, (IPIMS, 2010). Polarity reversal
Noisy trace
Spike
Raw record
41
Denoising: low cut filter applied
Denoising Before Noise Attenuation in Shot Domain
42
Denoising After Noise Attenuation in Shot Domain
Denoising Difference Noise Attenuation
43
Denoising: noise modeling Input Processing: Transformation Denoise Filtering
subtraction Noise modeling subtraction Output
Denoising: low frequency target
88
44
Denoising: low frequency target
89
Denoising: low frequency target
90
45
Denoising: QC on stack Before Noise Attenuation
Denoising: QC on stack After Noise Attenuation
46
Denoising: QC on stack Difference Noise Attenuation
before after Differences
“Smile” effect
47
“Smile” effect removed
STATICS
48
Shot record
Before Statics Correction
Shot record
After Statics Correction
49
Koreksi statik, statik: koreksi yang diterapkan pada data seismik untuk mengkompensasi efek dari variasi elevasi, low velocity layer (LVL) near surface, ketebalan lapisan lapuk dengan referensi sebuah datum.
Surface A
Respon seismik T0
B C
D
Reflektor
Travel time A ke B > Travel time C ke D
50
Surface
Respon seismik
Reflektor
Surface A
B C
D
Datum
Reflektor Travel time A ke B > Travel time C ke D. Perlu referensi yang sama
51
A
B
A’
Surface
Bagaimana mengkompensasi bagian ini?
B’
C
D
C’
D’
Datum
Reflektor Travel time A ke B > Travel time C ke D. Perlu referensi yang sama
• Elevation Correction • Delay-Time • GLI (bagus untuk model layer-based) • Traveltime Tomography (model grid-based bagus untuk complex geology) • Waveform Tomography (lebih detail)
52
HOW ABOUT MARINE DATA?
Statics on marine data
Water Column Static s Water column statics are a manifestation of physical c hanges in the water column caused by salinity, temperature, etc., over the period of acquisition. (Geotrace, 2010)
53
To remove the dynamic temporal changes in seismic data due to velocity change in the water. Before water velocity correction
* WesternGeco, 2008
After water velocity correction
54
Statics on marine data
Streamer
Gun
Statics = (Streamer depth + Gun depth)/water velocity
STATICS QC ON STACK
55
Stack without statics correction
Stack with statics correction
56
Stack with residual statics correction
DECONVOLUTION
57
Deconvolution : -
Improve the temporal (vertical) resolution
-
Remove coherent noise of multiple
-
Inversion process based on convolutional model of the seismic trace
S(t) = W(t) * R(t)
-1
R(t) = S(t) * W(t)
Seismic source from dynamite
Seismic source from vibroseis
Seismic source from airgun
Deconvolution : 1.
Sp ik in g Dec on : the desire wavelet is a spike or impulse.
2.
Predictive/Gap Decon : use early part of the trace to predict and deconvolve the later part.
3.
Wi en er Fi lt er : designing a filter which when convolved with an input signal minimises the difference between actual output and the desired output.
4.
Si gn at ur e Dec on : the output is desired wavelet.
Deconvolution Parameters : 1.
Length of input data window (gate).
2.
Length of decon operator.
3.
Whitenoise stability factor.
58
Shot Point Gather without deconvolution
Shot Point Gather with deconvolution
59
VELOCITY ANALYSIS
Why we need velocity? -Amplitude compensation -NMO correction (for stack) -Defining angle mute -Migration -Conversion to Depth -Identifying rock type
How to get the “correct” velocity?
60
Velocity analysis Raw
2264 m/s
2000 m/s
Overcorrected
uncorrected
velocity correct
vel oci ty t oo sl ow (need to be speeded up)
2500 m/s
Undercorrected
vel oci ty to o f ast (need to be slowed down)
Velocity analysis Survey 3-D
Survey 2-D Tools in velocity analysis : -Semblance -CMP gather -Multi velocity function stacks -Control stack -Isovelocity overlay dengan control stack -Basemap
61
Velocity analysis Normally velocity increase with depth, this is becaused of overburden pressure effect Velocity Time
Velocity analysis containing multiple
62
Velocity analysis containing multiple
Velocity analysis Multiples were removed
63
Velocity analysis QC: overlay velocity with the stack
Stack with single velocity function
64
Stack with multi (analized) velocity function
AMPL A MPLITUDE ITUDE CORRECTION CORRECTION
65
- Also known as geometric spreading amplitude correction and true amplitude recovery (TAR). - The Decrease in wave strength (energy per unit area of wavefront) with distance as a result of geometrical spreading.
Amplitude (A) at time T ~ 1/r ~ 1/(V.T), (r, is the radius of spherical wave front) For a constant velocity medium, V=const., A(T) ~ 1/T But when the velocity increases between layers, and in practice it increases with depth within layers, A(T) ~ 1/TV
2
QC amplitude correction
Raw Shot Gather
Less Compensation
Good Compensation
Too much Compensation
66
Stack without Surface Consistent Amplitude Correction (SCAC)
SURFACE PROBLEM
Stack with Surface Consistent Amplitude Correction (SCAC)
AFTER SURFACE CONSISTENT AMPL ITUDE CORRECTION
67
Shot gather without SCAC
135
Shot gather with SCAC
136
68
REGULARIZATION
Offset regularization Common Offset
Fold of coverage before 3-D Offset regularization
Common Offset
Fold of coverage After 3-D Offset regularization
RMS amplitude After 3-D regularization
With offset regularization the data distribution in every single bin became “balanced”. And the QC on the RMS amplitude over the offset cube is very usefull to look at the amplitude distribution before proceed the migration.
69
MIGRATION
Migration Concept
Reaching the reflector, the wavefront will be reflected, and some energy will propagate back to the source. The reflected wavefront will have the same circular form as the incident. Point reflector : the point at which the wavefront is reflected off the interface. Each source-receiver pair has a uniqe point reflector that yield the shortest traveltime .
70
Migration Concept
Migration Concept
71
Migration Concept
Migration Concept
72
Migration Concept
Migration Concept
73
Migration is a tool to get an accurate image of underground layer and structures. Migration : -Geometric reposition of recorded events to their true position -Move dip events to their true position -Collapse the diffraction
Type of Migration : 1. Kirchhoff - Most popular in recent years - Trace by trace - Not the best for imaging complex structures or area with strong lateral velocities variation 2. Finite-Difference (FD) - Much more accurate than Kirchhoff - Time consumming 3. Frequency-wave number or Fourier transform - More efficient than FD migration - More accurate than Kirchhoff - Not accurate for strong lateral velocity variation
74
Survey
Input Data
2-D
Post Stack Migration
Migration
3-D
Pre-Stack
Domain Time Migration Depth
Migration Strategies (from Yilmaz, 2001) Case
Migration Strategies
Dipping events
Time migration
Conflicting dips with different stacking velocities, complex non-hyperbolic moveout
Pre-stack migration
3-D behavior of fault planes and/or salt flanks
3-D migration
Strong lateral velocity variations associated with complex overburden structures
Depth migration
75
A
MIGRATION
B
MIGRATION
76
Migration Comparison
Post Stack Kirchhoff Time Migration result.
Pre-Stack Kirchhoff Time Migration gives benefit on the steep dip structure. The good data will help the interpreter to produce more accurate interpretation. The accurate interpretation of course will reduce the risk.
Special Processing
ANISOTROPY
77
B
A www.treehugger.com
Thoms en (2002) : Anisotropy is the variation of a physical property depending on the direction in which it is measured. Seismic anisotropy is defined to be the dependence of seismic velocity upon angle.
78
X
Z
X
Z Isotropic
Anisotropic
P-wave propagation
Anisotropy y r t e m m y s f o s i x A
y r t e m y s f o s i x A
t0 y r t e m y s f o s i x A
t0 + t
Slower velocity
V V for all azimuth
79
Anisotropy
Anisotropy • Well misties • ‘Hockey stick’ effects • Velocity variations correlating with structure • Problems with imaging different dips
80
Anisotropy Well Mis-tie
Anisotropy Hockey stick effects
81
Anisotropy Hockey stick effects’ corrected
Special Processing
SURFACE-RELATED MULTIPLE
82
Multiples
Surface-related multiple
Interbed multiple Sea surface
Sea bottom
Surface-Related Multiple Elimination Surface Related Multiple
Marine data with strong surface-related multiple
Stack section after surface-related multiple elimination
Removing the surface-related multiple has increased the S/N ratio and m ade the primaries came up. It will very much help on the interpretation.
83
Data contains surface-related multiple
Surface Related Multiple
Data after removing surface-related multiple
Surface Related Multiple free
84
Special Processing
COMMON REFLECTION SURFACE
85
(Baykulov et al., 2011)
Azimuthal processing processing
86
Azimuthal processing processing
Perbandingan data di-stack dengan velocity orisinil (V-0) vs velocity baru (V-1) : Time Slice 1800ms
Stack dengan V-0 Stack dengan V-1 V-1 di-analisis setelah data di-rotate pada azimuth tertentu
87
Advanced Processing Processing
DEPTH IMAGING
Time Migr Migr ation Image
88
Depth Migr ation Image
When we need Depth Migr ation? S
R
S
R
Time Migration : retrieves the velocity profile at the CMP location and ray traces through this local 1D model, i.e. no lateral velocity variations are comprehended, the ray path is always symmetric. Depth Migration: Velocity Model used as provided in it’s full complexity and Ray tracing comprehends velocity changes vertically and laterally. The ray path is non symmetric and summation surfaces shape becomes complex.
We need Depth Migration in subsurface that has strong lateral velocity variation. Lateral variation may be caused by faults, carbonate build-up, anticline/syncline, salt diapir, facies changing, gas pocket etc.
89
Flat reflector, con stant velocit y m odel CMP point
Surface
Reflector P NMO and Stack adequate to correctly im age and position point P.
Dipping reflector, constant velocity mod el CMP point
Surface
Reflector P Time migration will correctly image data.
90
Flat reflector, laterally varyin g velocit y mod el CMP Point
Surface
- VE
+VE
Reflector P Requires depth migration to correctly image data.
Image position comparison
0 offset stack trace
Time migrated trace
Depth migrated trace
Surface
Normal incidence ray
Image ray Full ray tracing
90
Reflector
P
Apparent position of P on stack trace
P Apparent position of P on stack trace
P Apparent position of P on stack trace
91
Benefit of Depth Migration
Correct vertical positioning - if the velocity model is good enough, the image will be free of structural distortions related to lateral velocity variations that cause pull-ups and sags.
Correct l ateral positioni ng - if the velocity model is good enough, the events will be placed in their correct lateral position.
Improved resolutio n - the image will have higher resolution than that obtained by time imaging because it doesn’t rely on the hyperbolic moveout assumption.
Al lows veloc it y and depth esti matio n - it provides it’s own diagnostics for deriving the accurate velocity model. If the depth model is correct, imaging with that model yields an identical image at all offsets/angles.
Depth Imaging
SAMPLES
92
Case : Fault i mage Challenge : Fault image on the flower structure.
PSTM section
Case : Fault i mage The antitetics fault now appears clearly on the flower structure by running DEPTH migration with accurate velocity model.
PSDM section
93
