Presentation is loading. Please wait.

Presentation is loading. Please wait.

Zero-Offset Data d = L o ò r ) ( g = d dr r ) ( g = d

Similar presentations


Presentation on theme: "Zero-Offset Data d = L o ò r ) ( g = d dr r ) ( g = d"— Presentation transcript:

1 Zero-Offset Data d = L o ò r ) ( g = d dr r ) ( g = d
(aka poststack data) d = L o o r ) ( ò g = d dr o r ) ( g = d V/2 Depth Time

2 d =  L m Born Forward Modeling ~  e  ij j i d(x) =  m(x’) i
A(x,x’) xx’ i e ~ d =  L m ij j i d(x) = x’ g(x|x’) m(x’) reflectivity

3 Seismic Inverse Problem
Given: d = Lm Find: m(x,y,z) Soln: min || Lm-d || 2 Waveform inversion (non-linear) Least squares migration (linear) m = [L L] L d T -1 Migration L d T

4 Migration Intuitive: Modeling Least Squares Poststack Mig Prestack
Green’s Theorem

5 Smear Reflections along Fat Circles
ZO Migration Smear Reflections along Fat Circles xx + T o Exploding Reflector assumption: Let c = v/2, or ½ actual velocity so Exploding reflector time=2-way time 2-way time (x-x ) + y 2 c xx = x Where did reflections come from? Thickness = c*T /2 o x d(x , ) xx

6 Smear Reflections along Fat Circles
ZO Migration Smear Reflections along Fat Circles x & Sum 2-way time d(x , ) xx Hey, that’s our ZO migration formula

7 Smear Reflections along Circles
ZO Migration Smear Reflections along Circles x & Sum 2-way time Out-of--Phase In-Phase d(x , ) xx m(x)=

8 ZO Data Migration ZO Data 0 km 3 km 0 km km

9  m(x) =  d (g, ) ZO Migration: Smear Trace Sample over Circle g xg
Loop over data for ixtrace=1:ntrace; for ixs=istart:iend; for izs=1:nz; r = sqrt((ixtrace*dx-ixs*dx)^2+(izs*dx)^2); time = 1 + round( r/c/dt ); mig(ixs,izs) = mig(ixs,izs)/r + data(ixtrace,time); end; Loop over x in model Loop over z in model Smear over circle Traveltime

10  m(x) =  d (g, ) ZO Migration: Smear Trace Sample over Circle g xg
Loop over data for ixtrace=1:ntrace; for ixs=istart:iend; for izs=1:nz; r = sqrt((ixtrace*dx-ixs*dx)^2+(izs*dx)^2); time = 1 + round( r/c/dt ); mig(ixs,izs) = mig(ixs,izs)/r + data(ixtrace,time); end; Loop over x in model Loop over z in model Traveltime

11 ZO Migration: Sum Trace Samples along migration hyperbola into m(x)
(x,z) (x’,z’) Loop over x in model Loop over z in model for ixtrace=1:ntrace; for ixs=istart:iend; for izs=1:nz; r = sqrt((ixtrace*dx-ixs*dx)^2+(izs*dx)^2); time = 1 + round( r/c/dt ); mig(ixs,izs) = mig(ixs,izs)/r + data(ixtrace,time); end; Loop over data Sum samples along hyperbola

12 ZO Diffraction Stack Migration
d (g, ) xg m(x) = g Trial image pt x traces x g It is often thought that RTM does not enjoy filtering tricks of KM such as U+D separation, obliquity factor, angle gather separation, anti-aliasing filter, etc. This is not true as shown above. The RTM formula is shown above in traditional form: apply adjoint Green’s function to data and backpropagate data, then zero-lag correlation with source field. Rearranging brackets gives different interpretation: RTM is just like KM in the sense that you apply a dot product of the hyperbolas to the data to get migration image. In this case the hyperbolas conatin all the scattering events and the Green’s functions are computed by FD solves rather than ray tracing.

13 2D dot product of migration
ZO Diffraction Stack Migration d (g, ) xg m(x) = g Trial image pt x traces 2D dot product of migration Operator and d(g,t) x g It is often thought that RTM does not enjoy filtering tricks of KM such as U+D separation, obliquity factor, angle gather separation, anti-aliasing filter, etc. This is not true as shown above. The RTM formula is shown above in traditional form: apply adjoint Green’s function to data and backpropagate data, then zero-lag correlation with source field. Rearranging brackets gives different interpretation: RTM is just like KM in the sense that you apply a dot product of the hyperbolas to the data to get migration image. In this case the hyperbolas conatin all the scattering events and the Green’s functions are computed by FD solves rather than ray tracing. Migration Image

14 ZO Diffraction Stack Migration: C(x,z)
d (g, ) xg m(x) = g Trial image pt x Ray tracing It is often thought that RTM does not enjoy filtering tricks of KM such as U+D separation, obliquity factor, angle gather separation, anti-aliasing filter, etc. This is not true as shown above. The RTM formula is shown above in traditional form: apply adjoint Green’s function to data and backpropagate data, then zero-lag correlation with source field. Rearranging brackets gives different interpretation: RTM is just like KM in the sense that you apply a dot product of the hyperbolas to the data to get migration image. In this case the hyperbolas conatin all the scattering events and the Green’s functions are computed by FD solves rather than ray tracing.

15 3D ZO Diffraction Stack Migration
d (g, ) xg m(x) = g Trial image pt x Impulse Response of Mig. Op. It is often thought that RTM does not enjoy filtering tricks of KM such as U+D separation, obliquity factor, angle gather separation, anti-aliasing filter, etc. This is not true as shown above. The RTM formula is shown above in traditional form: apply adjoint Green’s function to data and backpropagate data, then zero-lag correlation with source field. Rearranging brackets gives different interpretation: RTM is just like KM in the sense that you apply a dot product of the hyperbolas to the data to get migration image. In this case the hyperbolas conatin all the scattering events and the Green’s functions are computed by FD solves rather than ray tracing.

16 Migration Intuitive: Modeling Least Squares Poststack Prestack
Green’s Theorem

17 3D Prestack Diffraction Stack Migration
Motivation: ZO only good if no lateral vel change It is often thought that RTM does not enjoy filtering tricks of KM such as U+D separation, obliquity factor, angle gather separation, anti-aliasing filter, etc. This is not true as shown above. The RTM formula is shown above in traditional form: apply adjoint Green’s function to data and backpropagate data, then zero-lag correlation with source field. Rearranging brackets gives different interpretation: RTM is just like KM in the sense that you apply a dot product of the hyperbolas to the data to get migration image. In this case the hyperbolas conatin all the scattering events and the Green’s functions are computed by FD solves rather than ray tracing. s g x

18  m(x) = 3D Prestack Diffraction Stack Migration = d(x’,  +  ) s g x
sx s,g xg m(x) = Trial image pt x It is often thought that RTM does not enjoy filtering tricks of KM such as U+D separation, obliquity factor, angle gather separation, anti-aliasing filter, etc. This is not true as shown above. The RTM formula is shown above in traditional form: apply adjoint Green’s function to data and backpropagate data, then zero-lag correlation with source field. Rearranging brackets gives different interpretation: RTM is just like KM in the sense that you apply a dot product of the hyperbolas to the data to get migration image. In this case the hyperbolas conatin all the scattering events and the Green’s functions are computed by FD solves rather than ray tracing. s g x

19 Prestack Migration Question: Why Prestack when poststack
migration seems good enough? Answer: Stacking to get stacked section assumes layered medium assumption. Solution: Migrate shot gathers so no layer assumption needed. This is prestack migration.

20 Diffraction Stack Migration: Prestack
Down time Up time T(s,g) = sx xg + Where is scatterer? s x g sx xg s,g d(s,g, ) sx xg + Narrow band case: direct wave correlated with data

21 Diffraction Stack Modeling: Prestack
115. Diffraction Stack Modeling: Prestack m = L d T d = L m i i ~ W( ) ~ e m(x) ~ sx x xg e d(s,g) = w 2 A(s,x) A(g,x)

22 Diffraction Stack Migration: Prestack
115. Diffraction Stack Migration: Prestack m = L d T - - d(s,g,  +  ) sx xg d ò i i ~ W( ) ~ * e sx s,g xg e w 2 m(x) = ~ d(s,g) A(s,x) A(g,x) Broadband case W( )=1 ~ .. A(s,x) s,g A(x,g) = m(x) Narrow band case: direct wave correlated with data

23 MATLAB Inefficient Prestack Migration
Data Loops Model Loops for isx=1:nx % Loop over shot for igx=1:nx % Loop over receivers for ix=1:nx % Loop over model x for iz=1:nx % Loop over model z t=timer(ix,iz,isx)+timer(ix,iz,igx) sample=gather(isx,igx,t) % Shot gather has 2 time derivatives mig(ix,iz)=mig(ix,iz)+sample end

24 MATLAB Prestack Migration

25 Poststack vs Prestack Migration

26 Poststack vs Prestack Migration

27 Prestack Migration 1. No 1D assumption about velocity model
2. More sensitive to velocity model errors compared to poststack migration 3. More than 100 times slower than poststack migration 4. More sensitive to velocity model than time migration

28 Summary m(x) =    m(x) = 3D ZO Diffraction Stack Migration d (g, )
Migration Motivation: diffractions, dipping layers, conflicting dips, out-of-plane reflections 3D ZO Diffraction Stack Migration d (g, ) xg m(x) = g Trial image pt x 3D Prestack Diffraction Stack Migration = d(x’,  +  ) sx s,g xg m(x) = Trial image pt x

29 Migration Intuitive: Modeling Least Squares Poststack Prestack
Green’s Theorem

30 Iterative Least Squares Migration
Step 1: Step 2: Step 3: Step 4:

31 MATLAB SD Least Squares Migration
p=p % Data without direct wave m=adjoint(p,c) % Initial reflectivity model c % Velocity model for i=1:niter p=forward(m,c) % predicted data alpha=step(p,p0,c,m) % step length dP=p-p % data residual dm =adjoint(dP,c) % migrate residual m = m –alpha*dm % Update model end

32 Dot Products and Adjoint Operators
Recall: (u,u) = u* u i Recall: (v,Lu) = v* ( L u ) j i ij [ L v* ]u j i ij = [ L* v ]* u j i ij = So adjoint of L is L  i ij L*

33 Dot Product Test with CG code
Actual model Predicted model Actual data Predicted data d Lm =(d,Lm) = (Lm,d) = m L d T T T d=forward(m,c) m=adjoint(d,c) d d = T m T m All migration codes should pass the dot product test

34 Migration r = L L L Least Squares: Intuitive: Modeling Examples
Footprint Intuitive: Modeling Poststack Prestack Migration r = L L L T . . . 1 . Green’s Theorem . . Migration butterfly

35 Migration = Blurred r m = L d d = L r T but Migrated Section Modeling
Data

36 True Reflectivity Model r
Migration = Blurred r T m = L but d = L r L r Migrated Section Migration Image m = True Reflectivity Model r

37 r L = m Migration Deconvolution = Reflectivity Migration image T
Migration Green’s function

38 L r L = m = Migration Deconvolution ] [ Migration Deconvolution -1 1 T

39 Migration Deconvolution
-1 ] [ r = m = 1

40 L r = m = Migration Deconvolution ] [ -1 1 T
Assume Local v(z) Approximation

41 Migration Noise Problems
Note: Artifacts stronger near surface. Why? Footprint Migration noise and artifacts Depth (km) Weak illumination Irregular acquisition geometry Limited recording aperture Footprint and aliasing Some footprint caused by Irregular acquisition geometry Limited recording aperture aliasing and illumination loss in migration imaging 3.5

42 Wave Equation Migration Before LSM
Depth (km) 10 X (km) 20

43 Wave Equation Migration after MD X (km)
Depth (km) 10 X (km) 20

44 Acquisition Footprint (Geophone Aliasing)
Coarse Kirchhoff Migration Image (15 Hz) 2 Depth (km) X (km) LSM Image (15 Hz) Actual Model 2 Depth (km) X (km) Note: Artifacts stronger near surface. Why?

45 Standard Kirchhoff Image vs LSM Image
Kirchhoff Migration Image (15 Hz) 2 Depth (km) X (km) LSM Image (15 Hz) Actual Model 2 Depth (km) X (km)

46 Migration Least Squares: Intuitive: Modeling Examples Poststack
Footprint Intuitive: Modeling Poststack Prestack Migration Green’s Theorem

47 2D Poststack Data from Japan Sea
JAPEX 2D SSP marine data description: Acquired in 1974, Dominant frequency of 15 Hz. 5 TWT (s) 20 X (km) 48

48 LSM vs. Kirchhoff Migration
LSM Image 0.7 1.9 Depth (km) 2.4 4.9 X (km) 0.7 1.9 Depth (km) 2.4 4.9 X (km) Kirchhoff Migration Image

49 Kirchhoff MD

50 Kirchhoff LSM

51 Migration Summary m(x) =    m(x) = d (g, )
Migration Motivation: diffractions, dipping layers, conflicting dips, out-of-plane reflections Full Waveform Inversion m = m - aL dm (k+1) (k) T Least Squares Migration m = m - aL dm (k+1) (k) T 3D ZO Diffraction Stack Migration d (g, ) xg m(x) = g Trial image pt x 3D Prestack Diffraction Stack Migration = d(x’,  +  ) sx s,g xg m(x) = Trial image pt x

52 ZO Summary  d(x, ) .. m(x’)  = 1. ZO migration: cos xx’ x A(x,x’)
q obliquity .. xx’ m(x’) x d(x, ) = 1. ZO migration: cos q Approx. reflectivity A(x,x’) 2. ZO migration assumptions: Single scattering data 3. ZO migration matrix-vec: m=L d T ~ Compensates for Illumination footprint and poor illumination 4. LSM ZO migration matrix-vec: m=[L L] L d T -1 5. ZO migration smears an event along appropriate doughnut

53 Summary  d(x, ) .. m(x’)  = 1. ZO migration: cos xx’ x A(x,x’)
q obliquity .. xx’ m(x’) x d(x, ) = 1. ZO migration: cos q Approx. reflectivity A(x,x’) 2. ZO migration assumptions: Single scattering data 3. ZO migration matrix-vec: m=L d T ~ Compensates for Illumination footprint and poor illumination 4. LSM ZO migration matrix-vec: m=[L L] L d T -1 5. ZO migration smears an event along appropriate doughnut

54 Kirchhoff MD

55 Kirchhoff MD

56 Kirchhoff MD

57 Least Squares  Recall: Lm=d Find: m that minimizes sum of squared
j ij Find: m that minimizes sum of squared residuals r = L m - d i (r ,r) = ([Lm-d],[Lm-d]) = m L Lm -2m Ld-d d (r ,r) d dm i = 2 m L Lm -2 m Ld = 0 For all i L Lm = Ld Normal equations


Download ppt "Zero-Offset Data d = L o ò r ) ( g = d dr r ) ( g = d"

Similar presentations


Ads by Google