You are on page 1of 35

AVO NMO

Jonathan Downton
and
Larry Lines
Theme

• Traditionally AVO inversion performed on NMO


corrected CMP gathers
– NMO is a kinematic correction which distorts the
amplitude and phase in an offset dependent fashion
– These distortions become more problematic as we
include larger offsets which are important for 3 term
AVO
• This presentation demonstrates a methodology of
performing NMO and AVO inversion
simultaneously
Literature Review
• Ursin and Ekren (1995)
– suggested flattening the CDP gather on a particular
event rather that performing NMO
– Addresses NMO stretch but ignores offset dependent
tuning issues
• Bakke and Ursin (1998)
– suggested a way to correct for offset dependent tuning
– too many restrictive assumptions in derivation to apply
in general fashion
• Swan (1997)
– Suggested wavelet processing technique to correct for
gradient
Message
• By doing simultaneous AVO NMO more accurate
estimates of reflectivity attributes are obtained than
by performing NMO and then AVO inversion
– The offset dependent tuning is explicitly built into the
forward model and treated as part of the inversion
– NMO stretch is not an issue since the input data is never
NMO corrected. The AVO analysis is performed along
travel time events described by some travel time equation
such as a higher term NMO equation
– General approach which can be applied to any linear AVO
inversion formulation
Outline

• Theory
– NMO distortions
• NMO stretch
• Offset dependent tuning
– AVO NMO
• Example
• Conclusions
NMO
• SEG Dictionary of Exploration Geophysics
– NMO: “variation of reflection arrival time because of
shotpoint to geophone distance”
– NMO correction: “the time correction applied to
reflection time because of normal moveout.”
• Key points
– NMO is a kinematic correction
– The NMO correction is a conjugate operation, not an
inverse operation (Claerbout, 1992)
– The NMO correction introduces amplitude and
character distortions to the offset gather
NMO stretch
• Frequency spectrum
gets shifted and
amplified due to
NMO stretch (Dunkin
and Levin, 1973)
1  f 
Sx  f  
~
S x  
x x 
~
where S x NMO corrected data
S x data before NMO
 x compressio n factor
NMO Stretch
Offset (m)

Time (sec.)
Logs in time reflectivity input x-plot

Filtered reflectivity input x-plot Distortions introduced by NMO


Offset dependent tuning
• Why don’t we just back out NMO stretch?
– Offset dependent tuning introduces zeros into
the amplitude spectrum in an offset dependent
fashion
• hyperbolic moveout tunes events differently as a
function of offset
– tuning is more severe at far offsets since events are closer
together

• Because of differential tuning it is difficult


to remove the effects of NMO stretch
Velocity Model Elastic Shot Model

BQ

Vp r Ip
ID Synthetic NMO Corrected Shot Synthetic Near Far
Stack Stack
Near Offset Analysis Far Offset Analysis
AVO theory
• Use convolutional model as basis of AVO NMO
inversion scheme
– Assumes locally, earth is composed of a series of flat,
homogenous, isotropic layers
– Approximation of Zoeppritz equation used to model how
reflectivity changes as a function of offset
• Plane waves assumed
– Ray tracing is done to map relationship between angle of incidence
and offset
– Transmission losses, converted waves and multiples are not
incorporated in this model and so must be addressed through pre-
processing
– Gain corrections such as spherical divergence, absorption,
directivity and array corrections can be incorporated in this model
but are not considered for brevity and simplicity
AVO theory: Linear approximation
to Zoepprtiz equation
• To linearize problem two term Fatti approximation
(Fatti et al, 1994) used to approximate Zoepprit
equation
  V 2 
 
x   1  tan 2  RP  8 s  sin 2   RS ,
  V p  
 
• where  is the average angle of incidence,
• x() is the offset dependent reflectivity,
• VP and VS are the P- and S-wave velocity,
• RP and RS are the P- and S-impedance
reflectivity
AVO theory
• Fatti equation may be written in matrix notation
• For example, consider the case of two offsets
– a near offset x1 and a far offset x2 where subscript
indicates the offset
•  x1   f1 g1   RP 
x    f g 2   RS 
•  2  2

• where

f n  1  tan 2  n 
VS2
gn  8 2 sin 2

VP
AVO theory
• Typically solve AVO inverse problem one
interface at a time
– Ignores band limited nature of seismic data
– Can solve problem for many interfaces at once
• To illustrate this, solve 2 interface simultaneously
 x11   f11 g11 0 0   RP1 
 1   1  
 x2    f 2 g 21 0 0   RS1 
 x12    0  2     2  ,
0 f12  g1 RP
 2    2    2  
 x2   0 0 f 22  g 2   RS 

Where superscript has been introduced to indicate interface


AVO theory
• Rearrange matrix so column vector is organized along
common offsets rather than common time samples
 x11   f11 0 g11 0   RP1 
 2     
 x1    0 f12  0 g12    RP2  
 x21   f 21 1 .
0 g 21 0 RS  
 2    2    2  
 x2   0 f 22  0 g 2   RS 

x1   F1 G1  rP 
x   F G  r ,
 2  2 2  S 

Where elements of linear operator matrix are diagonal


matrices composed of their respective weights for each offset
further simplify by writing elements composing common offsets as
vectors
AVO theory: NMO
• NMO can be written as a linear operator
(Claerbout 1992).
• A reflectivity sequence referenced to zero offset
time xn can be transformed to offset dependent
travel time dn by the linear operator Nn so that

• dn=Nnxn.
• matrix Nn can be constructed using whatever
offset/ travel time relationship one desires
– We use shifted hyperbola formula (Castle 1994)
AVO theory: AVO NMO
• Combine NMO operator dn=Nnxn with multi-layer AVO
model
x1   F1 G1  rP 
x   F    ,
 2  2 G 2  rS 
• and wavelet W to get simultaneous AVO NMO model
– This model may be generalized for more offsets and interfaces
d1   WN 1F1 WN 1G1  rP 
d   WN F WN G  r .
 2  2 2 2 2  S 
• Can estimate P and S impedance reflectivity by solving
this as an inverse problem
– Problem is underdetermined and ill-conditioned
– Use Bayes theorem to constrain problem (similar to Downton and
Lines, 2001)
d  Gr
Outline

• Theory
– NMO distortions
• NMO stretch
• Offset dependent tuning
– AVO NMO
• Example
• Conclusions
Synthetic example
• Model constructed
from Swan (1997)
Table 1
– All events are tuned
• Generated synthetic
gather using
– ray tracing
– Zoeppritz equation
– Wavelet 10/14-60/70
Hz
– Data sampled at 4
msec.
– S/N ratio = 10
Traditional AVO
Input Reconstruction 2X Difference

Offset (m)
Traditional NMO Zero-offset reflectivity
(10/14-60/70 Hz) Cross-plot
P-impedance S-impedance
reflectivity reflectivity

actual
estimate

Traditional AVO
Reflectivity Cross-plot
Traditional NMO Zero-offset reflectivity
Cross-plot
(10/14-50/60
P-impedance
Hz)
S-impedance
reflectivity reflectivity

actual
estimate

Traditional AVO
Reflectivity Cross-plot
AVO NMO
Input Reconstruction 2X Difference

Offset (m)
AVO NMO Zero-offset reflectivity
Cross-plot
P-impedance S-impedance
reflectivity reflectivity

actual
estimate

AVO NMO reflectivity


Cross-plot
Data Example
• Line part of project shot to explore for Halfway
sand potential (Downton and Tonn, 1998)
– 2 Bright spots
• Producing field
– Expect both velocity and density anomaly
• Uneconomic field (porous sand with low gas saturations)
– Expect velocity anomaly to be bigger than density anomaly
– Data has good S/N
– 10/14-110/130 Hz zero phase wavelet used
A C E F
A C E F
Conclusions
• NMO introduces distortions into the AVO analysis particularly when using
long offset data
• Simultaneous AVO NMO inversion gives superior results compared to the
traditional approach of first applying NMO and then doing AVO inversion.
– Better estimates of reflectivity lead to more correct trends and tighter
clusters in the cross-plot domain
• Improvement is true even for events showing significant offset dependent tuning.
– Greater bandwidth than the traditional method
• Able to invert up to input frequency of model
– Better signal-to-noise ratio than traditional method
• Model data both temporally and spatially leading to less bias introduced
by random noise bursts
• Can apply NMO AVO methodology to any linearized approximation of the
Zoeppritz equations including 3 term parameterizations
• Improvements come at a significant computational cost
Acknowledgements

• The authors would like to thank Core Lab,


N.S.E.R.C. and the C.R.E.W.E.S. sponsors
for funding this research.
Outline

• Theory
– NMO distortions
• NMO stretch
• Offset dependent tuning
• RNMO and Higher order NMO corrections
– AVO NMO
• Example
• Conclusions
RNMO

• In addition to the distortions introduced by


the NMO correction, if there is RNMO this
will further distort the AVO inversion.
• Since we are concerned with using data
with large angles need to be concerned
with ignoring corrections for large offsets
• Will use 3 term correction following Castle
(1994)
reflectivity input x-plot Filtered reflectivity input x-plot

Distortions introduced by NMO Distortions introduced by RNMO, NMO


Stretch and offset dependent tuning Stretch and offset dependent tuning

You might also like