Kirchhoff (diffraction stack) migration

an integral (or summation) solution to the wave equation

a point scatterer/diffractor creates a hyperbola in the zero-offset section
a reflector is equivalent to a line of closely spaced point diffractors

Yilmaz 2001

scheme one: spread a single sample on the unmigrated

trace to all possible reflection points
this creates the semi-circular impulse response
the circle is stretched vertically if velocity increases
with depth – analytical solution exists
the circle is deformed if there is lateral velocity
variation – requires ray-tracing to find it
do this for all samples on unmigrated traces, and add
where neighbouring samples agree with each other,
they will add together– constructive interference
where they disagree, they will cancel – destructive
interference
Yilmaz 2001

Hole: GEOS 4174 4.3-8 Reflection Data Processing: Poststack Signal Processing
Kirchhoff (diffraction stack) migration

scheme two: compute a single sample on a migrated trace

by adding together all unmigrated samples along the
diffraction hyperbola (diffraction stack)
if diffractor (or piece of reflector) exists, a signal will
add up at the migrated sample point
the overlying rms velocity defines the hyperbola for
1D velocity structure
a non-hyperbolic shape is created by lateral
velocity variation – must be ray-traced
repeat for all samples in the migrated image
Yilmaz 2001

the two schemes are equivalent; the latter is more common

during the summation, the amplitude and phase must be corrected for spreading and obliquity of the
ray paths
the corrected summation along the hyperbola is called the “Kirchhoff summation”

migration aperture is the distance along the line that the summation follows the hyperbola
deeper reflectors require longer aperture
steeper dips require larger aperture

performance of Kirchhoff migration:

all dips; complex velocity structure; slow (especially with lateral velocity variation)

Hole: GEOS 4174 4.3-9 Reflection Data Processing: Poststack Signal Processing
 finite difference migration
a finite-difference solution to the wave equation

in the exploding reflector model: at time t = 0 , the upgoing wavefield has the reflector’s shape
this is known as the imaging principle
the acoustic (P-wave only) wave equation for pressure P( x, z,t ) or P( x, z, ω ) is:
∂2P ∂2P 1 ∂2P ∂2P ∂2P ω2 ∂2P
2
+ 2 − 2€ 2 =0 or 2
+ 2 + 2 2
=0
∂x ∂z v ∂t ∂x ∂z v ∂ω
the observed zero-offset stacked section is the wavefield at the surface P( x, z = 0,t )
€ €
downward continuation
€ using a finite-difference approximation of the€wave equation, backwards extrapolate the
wavefield from its known value at depth z = 0 down to a new depth z = Δz
€
this is equivalent to computing the wavefield that would have been observed with the
receivers at this greater depth
at this depth, the wavefield at zero time has the shape of reflectors at this depth
repeat extrapolation to greater depths
€ €
the migrated reflection section is the value of the wavefield at zero time P( x, z,t = 0)

€ Yilmaz 2001

Hole: GEOS 4174 4.3-10 Reflection Data Processing: Poststack Signal Processing
finite difference migration

the finite-difference wave propagation scheme can be implemented in the time or frequency domain
two common finite-difference approximations exist:
the “15˚ approximation” (parabolic equation) can image dips to about 35˚ and is usually
implemented in the time (t-x) domain
the “45˚ approximation” can image dips to about 65˚ and is usually implemented in the
frequency (f-x) domain
higher-dip approximations exist, but are computationally slow
steeper dips are under-migrated and dispersed (spread out in time and frequency) => artifacts
both vertical and lateral velocity variation can be easily included in the finite difference scheme

the step size in depth (or in time for time migration) for the downward continuation is critical
steps too large will produce under-migration
steps too large or too small produce dispersion – spreading of the arrival in time & frequency
steeper dips require smaller steps
optimum step size is about half of the dominant wavelength (or period for time migration)

performance of finite-difference migration:

moderately fast; complex velocity structure; dip-limited (unless slow high-dip approx. used)

Hole: GEOS 4174 4.3-11 Reflection Data Processing: Poststack Signal Processing
frequency-wavenumber (f-k) migration
a Fourier transform solution to the wave equation

Yilmaz 2001
Stolt migration
for constant velocity, migration operator in the f-k domain (actually f-kx-kz domain) shifts the
ik z
reflector by an amount e z
for variable velocity, first locally stretch the time axis to simulate constant velocity
introduces inaccuracy, so only works for smooth velocity
phase-shift (= Gazdag) migration
for depth-varying€velocity, perform the operation in the f-kx-z domain by downward continuation
in small steps Δz
for laterally variable velocity, first locally stretch the time axis to simulate laterally constant vel.
introduces inaccuracy, so only works for smooth velocity
spatial aliasing of steep dips produces artifacts in f-k migration
performance of frequency-wavenumber migration:
€
fast; all dips; limited to laterally smooth (almost 1D) velocity structure

Hole: GEOS 4174 4.3-12 Reflection Data Processing: Poststack Signal Processing
prestack migration

events with different dips can overlap in the CMP gather => a single stacking velocity cannot work
dipping events are not centered at zero offset => NMO correction cannot work
=> requires prestack time migration
the semi-circular impulse response of stacked (zero-
offset) trace becomes an ellipse with the source
and receiver at the foci; the ellipse is stretched by
vertical velocity variation
generally, this can be replaced by DMO + post-stack
time migration

Yilmaz 2001

strong lateral velocity variation along ray paths (i.e., at shallower depth than reflection target) creates
non-hyperbolic moveout => violates assumptions of NMO/DMO; stacking cannot work
=> requires prestack depth migration
the impulse response ellipse is distorted by vertical and lateral velocity variation
requires a better velocity model than time migration

Hole: GEOS 4174 4.3-13 Reflection Data Processing: Poststack Signal Processing
another migration example: salt flanks
stack prestack time migration

Yilmaz 2001
prestack time migration has imaged the steep dips on either side of the salt

Hole: GEOS 4174 4.3-14 Reflection Data Processing: Poststack Signal Processing
another migration example: complex salt structure

Ikelle & Amundsen 2005

prestack depth migration has imaged the bottom of salt and underlying strata

Hole: GEOS 4174 4.3-15 Reflection Data Processing: Poststack Signal Processing