John C. Bancroft and grad students AVO, migration apertures Fresnel zones, stacking, Q, anisotropy, and fast inversions John C. Bancroft and grad students University of Calgary CREWES 2003 NSERC
Grad students Shuang Sun Pavan Elapavuluri Kun Liu (and Hugh Geiger) John Millar Xiang Du Zhihong (Nancy) Cao Chunyan (Mary) Xiao
Outline Differentiators Fresnel zones AVO migration apertures Q Wavefield extrapolators Multiples and radon transforms Inversion
Digital differentiator Accuracy Speed Beauty (1, -1) 1, 0, -1) better Theoretical shape?
Inverse FFT of (j)
Derivative assumes: Maximum bandwidth Ideal interpolator Sinx/x operator Derivative of sinx/x SincFn.m
What is the basic unit of laryngitis? Shuang Sun Stacking specula energy Fresnel zone for offset and dipping reflectors Balanced AVO in areas of poor geometry Caution: model based migration One hoarsepower What is the basic unit of laryngitis?
What size hole can we see with seismic?
What happens at the edge of a reflector?
Fresnel zone Fresnel Size: frequency depth velocity Claerbout
Depends on the size of the Fresnel radius Only for stacked data!!!
Analytic function for zero-offset case (no wavelet) Solution to the acoustic wave equation for zero-offset, the diffraction response can be written as: Diffraction operator D0: te : minimum two-way travel time to the edge of the reflector; t’=t-te: time measured after onset time te ; e : angle between the normal to the reflector and raypath of the minimum travel time to the edge of the reflector. Berryhill, 1977
Diffraction amplitude Time Angle
Diffraction amplitude
Scaled amplitude (evaluate phase)
Fresnel zone with Ricker wavelet source Source signature with dominant frequency 50hz Reflected signal amplitude
Fresnel zone with fixed bandwidth Frequency from 0-150 hz Frequency from 0-120 hz
Reflector smaller than Fresnel zone: width 50m
Reflector width: 1000m
Amplitude (no wavelet)
Diffraction modelling
Valid area for AVO ?
Comments Locations for AVO analysis before migration should be at least a Fresnel radius away from the target edge. The amplitudes will be in error if the target size is smaller than or equal to the Fresnel zone. Wavelet will contribute to the size of contamination area. Only zero offset was considered Should consider a prestack migration to perform AVO analysis.
Offset traveltime 2h DSR eqn.
Cheop’s pyramid - DSR eqn. Hyperbola CMP at T0 Zero off. DSR eqn. Const. Off CMP gath.
Prestack data, flat… 7 Cheop’s . Specula energy
Reflections from dipping event Reflecting element
Prestack data, dipping... Cheop’s p. Defines exact shape of surface, specula energy
Cheop’s summation: specula energy Fresnel zones
Prestack Fresnel zone horizontal layers
P. S. Fresnel zone for dipping event
Contours : equal angle of incidence Offset increases with x Angle 1
Constant offset migration
What about dipping specula energy? Smear of dips Still OK ???
CMP gather
Limited aperture EO gather
Comments … Prestack specula energy can be identified, Limited aperture horizontal and dipping Limited aperture faster, better SNR, may preserve amplitudes Optimum size of a super-CMP gather can be defined using Fresnel zones Should use prestack migration gathers for AVO Use model based migration with caution
Comments Constant offset migration EOM smears AVO energy (OK) requires velocity information EOM also smears AVO energy (OK) Stacking within the Fresnel zone may reduce acquisition geometry artifacts
What is one millionth of a mouth wash? Pavan Elapavuluri Q What is one millionth of a mouth wash? One microscope
Wavelets with varying Q and Time Build Q wavelet table: [Q, time] Q: 30, 50, 70, 90, 110, … T: 0 to 6.0 sec Cross-correlate (coherence) trace with all wavelets At a given time, peak corresponds to Q
9 wavelets T=1000 T=3000 T=5000 Q=90 Q=70 Q=50
3 Spikes on trace Q = 70 T=1000 T=3000 T=5000 Q=90 Q=70 Q=50
Noise on trace Q = 70 T=1000 T=3000 T=5000 Q=90 Q=70 Q=50
Kun Liu (and Hugh Geiger) Wavefield extrapolation What is 2000 pounds of Chinese soup? Won ton
Input
200m depth extrapolation …with adaptive taper
What is half a large intestine? Xiang Du Finite element method integrated with Finite difference migration A semicolon What is half a large intestine?
Comparison: FE-FDM with FX-FDM Result of FE-FDM Result of FX-FDM (SU)
Trace comparisons
Steep oblique model Velocity model Seismic section
Comparison of methods Result of FE-FDM Result of FX-FDM (SU)
Runtime and memory requirements FE-FD 26 secs 2.2 meg FK-FD 31 secs 1.3 meg.
What is one thousand aches? Zhihong (Nancy) Cao Chunyan (Mary) Xiao Tutorials Review of Multiple suppression techniques Multiple attenuation using the Radon transform What is one thousand aches? A kilohurtz
after Kabir and Marfurt, 1999
Semblance plot is a Radon transform
Lorraine Bloom David Henley Larry Lines The model and inverse
Modelling process Inversion expert
The Model (Not the inverse)
The Inverse
The artifacts (of inversion) Before modelling After inversion
Of course, … some things are easier to model and invert.
Inversion Mathematical process Geophysical terminology Used in many geophysical applications: decon., statics, tomography, tau-p, … Geophysical terminology Estimation of rock properties Modelling <-> Inversion
What is the ratio of an igloo’s circumference to its diameter? Millar time Multigrid inversion Borrowed from fluid dynamics Recursive Variable grid size Solutions are frequency dependent Fast What is the ratio of an igloo’s circumference to its diameter? Eskimo pi
Iterative solution
Laplace’s equation
Variable grid size
Analogous to FFT ???
Sorting a 2-D array into a vector
Extractions, smoothings and interpolations S = Solution with one iterative step 33x33 17x17 3x3 9x9 5x5 3x3 9x9 5x5 Pass defects down and errors up
Performance
Conclusions: Migration apertures speed, SNR, balanced amplitudes CMP AVO: consider Fresnel zones Finite element / Finite difference migration Multiples and Radon transforms Q and differential operators Multi grid inversions
The end