Introduction'to Seismic Inversion Methods Brian Russell

PART 4 - SEISMIC PROCESSINGCONSIDERATIONS

Part 4 - Seismic Processing Considerations Page 4 - 1

•ntroduction to Seismic •nvers•on Methods [Link] Russell

4.1 Introduction

Havinglookedat a simple model'of the seismic trace, anu at the

recursire inversion alogorithmin theory, we will nowlook at the problem of
processingreal seismiceata in order to get the best results fromseismic
inversion. We may group the key processingproblemsinto the following
categories:

( i ) Amp
1i tu de recovery.
(i i) Vertical resolution improvement.
(i i i ) Horizontal resol uti on improvement.
(iv) Noise elimination.

Amplitudeproblemsare a majorconsiderationat the early processing

stagesandwewill look at both deterministicamplitude
recoveryandsurface
consistent residual static time corrections. Vertical resolution improvement
will involve a discussion of aleconvolution and wavelet processing techniques.
In our discussion of horizontal resolution we will look at the resolution
improvement
obtained in migration, using a 3-D [Link], wewill
consider several approaches
to noiseelimination, especially the elimination
of multi pl es.

Simply stateu, to invert our seismic data we usually assume the

one-dimensional model given in the previous section. And to arrive at an
approximationof this model (that is, that each trace is a vertical,
band-limited reflectivity function) we must carefully process our data with
these considerations in minU. Figure 4.1 showsa processing flow which could
be useUto do preinversion processing.

Part 4 - Seismic Processing Considerations Page 4 - 2

Introduction to Seismic Inversion Methods Brian Russell

INPUT RAW DATA

DETERMINISTIC
AMPLITUDE
CORRECTIONS

,. _•m SURFACE-CONSI STENT

AMPt:ITUDE ANAL'YSIS

mlm

SURFACE-CONS ISTENT

DECONVOLUTIO,
NFOLLOWED
BY HI GH [Link] DECON
i

SURFACE-CONSISTENT
STATI CS ANAIJYSIS

VELOCITY ANAUYSIS

APPbY STATICS AND VEUOCITY

MULTIPLE ATTENUATION

STACK
ß •

MI GRATI ON
,

Fig. 4.1. Simpl

i fiedi nversi
onprocessing
flow.

ll , ß ' ß I , _ i 11 , m - -- m _ • • ,11

Part 4 - Seismic Processing Considerations Page 4 - 3

Inl;roducl:ion 1:oSeJsmlc Invers1on Nethods BrJan Russell

4.2 Am.p'l
i [Link]..P,.ecovery

The most dJffJcult job in the p•ocessing of any seismic line

ß
is
•econst•uctingthe amplJtudes
of the selsmJc
t•aces as they wouldhavebeenJf
the•e were no dJs[urbJnginf'luences present. We normally make the
simplJficationthat the distortionof the seJsmic
amplJtudes
maybe put into
three main categories'sphe•Jcaldivergence, absorptJon,and t•ansmJssion
loss. Based on a consideration of these three factors, we maywrJte aownan
approximatefunctJonfor the total earth attenuation-

(b
At: AO* / t) * exp(-at),

where t = time,
At = recorded
amplitude,
anU
A0 = true ampl
i tude,
a,b = constants.

Thus, if we estimate the constants in the above equation from the seismic
data, the true amplitudesof the data coulUbe recoveredby using the inverse
formula. The deterministic amplitude correction and trace to trace mean
scaling will accountfor the overall gross changesin amplitude. However,
there may still be subtle (or even not-so-subtle) amplitude problems
associatedwith poorsurface conditions or other factors. To compensate
for
these effects, it is often advisable to computeand apply surface-consistent
gain corrections. This correction involves computing
a total gain value for
each trace and then decomposing
this single value in the four components

Aij=
SixRjxG
kxMkX
•j,
where A = Total amplitude factor, M = Offset component,
S = Shot component, X = Offset distance,
R: Receiver component, i,j = shot,receiver pos.,
G = CDP component, and k = CDP position.

Part 4 - Seismic Processing Considerations Page 4 -

Introduction to Seismic .Inversion Methods Brian Russell

SURFACE
SUEF'A•
AND
CONSIb'TEh[O{

T |tV•E :

,RiL-rER ß

Fig. 4.2. Surface and sub-surface geometryand

surface-consistent decomposition. (Mike Graul).

, ,

Part 4 - SeismicProcessing
Considerations Page 4 - 5
Introduction to Seismic Inversion Methods Brian Russell

Figure 4.g (from Mike Graul's unpublished course notes) shows the
geometryusedfor this analysis. Notice that the surface-consistent
statics
anti aleconvolutionproblemare similar. For the statics problem,the averaging
can be •1oneby straight [Link] the amplitude problemwemust
transform the above equation into additive form using the logarithm:

InAij=
InSi +InRj+InGk+lnkMijX•.
The problem can then be treated exactly the same way as in the statics
case. Figure 4.3, fromTaner anti Koehler (1981), showsthe effect of doing
surface consistent amplitude and statics corrections.

4.3 I•mp.
rov.
ement_
o.[_Ver.
[Link].1..Resoluti
on

Deconvolution is a process by which an attempt is made to remove the

seismic wavelet from the seismic trace, leaving an estimate of reflectivity.
Let us first discussthe "convolution"part of "deconvolution" starting with
the equation for the convolutional model

st--wt* r t where st= the seismictrace,

wt= the seismicwavelet,
rt= reflection coefficient series,
* = convolution operation.
In the frequency domain

S(f) • W(f) x R(f) .

The deconvol ution process is simply the reverse of the convolution

procedure and consists of "removing" the wavelet shape to reveal the
reflection coefficients. We must design an operator to do this, as in the
fol 1owlng equati on-

rt: st* o whereOr--operator-- inverseof wt .

Part 4 - Seismic Processing Considerations

,
Page 4 - 6
Introduction to Seismic Inversion Methods Brian Russell

ii 11

ß 1'

ii

'..,•' •, ," " " ß d.

Preliminary
stack
bet'ore
surface
consistent
static
andompli- ß Stockwith surfaceconsistent
staticand amplitudecor-
rections.
lude corrections.

Fig. 4.3. Stacks with and without surface-consi stent

corrections. (TaneranuKoehler,1981).
ß ,

Part 4 - Seismic Processing Considerations Page4 - 7

Introduction to Seismic Inversion Methods Brian Russell

In the frequency domain,this becomes

R(f) = W(f) x 1/W(f) .

After this extremely simple introduction, it may appear that the

deconvolutionproblemshouldbe easy to solve. This is not the case, and the
continuing research into the problem testifies to this. There are two main
problems. Is our convolutionalmodelcorrect, and, if the modelis correct,
can we derive the true wavelet from the data? The answer to the first
questionis that the convolutionalmodelappearsto be the best modelwe have
comeup with so far. The main problem is in assuming that the wavelet does
not vary with time. In our discussionwe will assumethat the time varying
problemis negligible within the zoneof interest.

The secondproblemis much more severe, since it requires solving the

ambiguousproblemof separatinga wavelet and reflectivity sequencewhenonly
the seismic trace is known. To get around this problem, all deconvolution or
wavelet estimation programsmakecertain restrictive assumptions,either about
the wavelet or the reflectivity. There are two classes of deconvolution
methods: those which make restrictive phase assumptions and can be considered
,

true wavelet processingtechniquesonly whenthese phaseassumptionsare met,

and those which do not make restrictive phase assumptions and can be
consideredas true wavelet processingmethods. In the first category are

(1) Spiking deconvolution,

(2) Predictive deconvolution,
(3) Zero phase deconvolution, and
(4) Surface-consi stent deconvoluti on.

Part 4 - Seismic Processing Considerations Page 4 -

Introduction to Seismic Inversion Methods Brian Russell

(a)

Fig. 4.4 A comparison of non surface-consistent and surface-consistent

decon on pre-stack data. {a) Zero-phase deconvolution.
{b) Surface-consistent soikinB d•convolution.

(b),
Fig. 4.5 Surface-consistent decon comparisonafter stack.
(a) Zero-phase aleconvolution. (b) Surface-consistent
deconvol ution.
'--'- , ß , ,• ,t ß ß _ , , _ _ ,, , ,_ , ,

Part 4 - .Seismic
Processing
Consioerations Page4 -
Introduction to Seismic [Link] Methods Brian Russell

In the second category are found

(1) Wavelet estimation using a well 1og (Strat Decon).

(Hampson
and Galbraith 1981)

(2) Maximum-1ikel ihood aleconvolution.

(Chi et al, lg84)

Let us illustrate the effectiveness of one of. the methods,

surface-consi stent aleconvolution. Referring to Figure 4.•, notice that a
surface-consi stent scheme involves the convolutional proauct of four
components. We must therefore average over four different geometry
di recti ons- common source, commonreceiver, common depth point (CDP), and
con,non offset (COS). The averaging must be performed iteratively and there
are several different ways to perform it. The example in Figures 4.4 ana 4.5
shows an actual surface-consi stent case study which was aone in the following
way'

(a) Computethe autocorrelations of each trace,

(b) average the autocorrelations in each geometry eirection to get four
average autocorrel ati OhS,
(c) derive and apply the minimum-phase inverse of each waveform, and
(•) iterate through this procedure to get an optimumresult.

Two points to note when you are looking at the case study are the
consistent definition of the waveform
in the surface-consistentapproachan•
the subsequent improvementof the stratigraphic interpretability of the stack.

We can compareall of the above techniques using Table 4-1 on the next
page. The two major facets of the techniques which will be comparedare the
wavelet estimation procedure and the wavelet shaping procedure.

Part 4 - Seismic Processing Considerations Page 4 - 10

Introduction to Seismic Inversion Methods Brian Russell

Table 4-1 Comparison of Deconvolution MethoUs

m m ß ß m

METHOD WAVELET ESTIMATION WAVELETSHAPING

Spiking [Link]
phase assumption Ideally shapedto spike.
Deconvol ution Randomrefl ecti vi ty In practice, shapedto minimum
assumptions. phase,higherfrequencyoutput.

Predi cti ve No assumptions about Does not whiten data well.

Deconvol uti on wavelet•
Removes
short andlong period
multiples. Does not affect
phaseof wayelet for long lags.
..1_, m

Zero Phase Zero phaseassumption. Phase is not altered.

Deconvol utton Randomrefl ectt vi ty Amplitude spectrumi$
assumption. whi tened.

Canshapeto desired output.

Surface-cons. Minimumor zero phase. Phasecharacteri s improved.

Deconvolution Randomreflecti vi ty Ampli rude spectrum i s
assumption. whitenedless than in single
trace methods.

Stratigraphic No phase assumption. Phase of wavelet is zeroed.

Deconvol ution However, well must match Amplitude
spectrum
not
sei smi c. whi tened.

Maximum- No phase assumption. Phase of wavelet is zeroed•

L ik el i hood Sparse-spike assumption. Amp
1i rude spectrumi s
deconvol ution whi tened.

Part 4 - Seismic Processing Considerations Page 4 11'

Introduction to Seismic Inversion Methods Brian Russell

4.4 Lateral Resol uti on

The complete three-dimensional (3-D) diffraction problemis shownin

Figure 4.6 for a modelstudy taken fromHerman,et al (1982). Wewill look'at
line 108, which cuts obliquely across a fault and also cuts across a reef-like
structure. Note that it misses the second reef structure.

Figure 4.7 shows the result of processing the line. In the stacked
section we maydistinguish two types of diffractions, or lateral events which
do not represent true geology. The first type are due to point reflectors in
the plane of the section, and include the sides of the fault and the sharp
corners at the base of the reef structure which was crossed by the line. The
secondtype are out-of-t•e-plane diffractions, often called "side-swipe". This
is most noticeable by the appearance of energy from the second reef booy which
was not crossed. In the two-dimensional (2-D) migration, we have correctly
removed the 2-D diffraction patterns, but are still bothere• by the
out-of-the-plane diffractions. The full 3-D migration corrects for these
problems. The final migrated section has also accounted for incorrectly
positioned evehts such as the obliquely dipping fault. This brief summaryhas
not been intended as a complete summaryof the migration procedure, but rather
as a warning that migration {preferably 3-D) mustbe performedon complex
structural lines for the fol 1 owing reasons:

(a) To correctly position dipping events on the seismic section, and

(b) To remove diffracted events.

Although migration can compensatefor someof the lateral resolution

problems, we must rememberthat this is analogous to the aleconvolution problem
in that not all of the interfering effects may be removed. Therefore, we must
be aware that the true one-dimensional seismic trace, free of any lateral
interference, is impossible to achieve.

Part 4 - Seismic Processing Considerations Page 4 - 12

 Introduction to Seismic Inversion Methods Brian Russell

lol 71

131

101

131 108

(a] 3- D MODEL

ß
ß ß
ß ß
ß

..................................

LINE
.............................

.........................................

....................................

{hi 8•8•0 LAYOU•

Fig. 4.6. 3-D model experiment. (Herman

et al, 1982).
i mm _ ml j mm

Part 4 • Seismic Processing Considerations Page 4 - 13

Introduction to Seismic Inversion Methods Brian Russell

4.5 Notse Attenuation

As we' discussed in an earlier section, seismic noise can be classified as

either •andom 'or coherent. Random noise is reduced by the stacking process
quite well unlessthe signal-to-noise ratio dropsclose to one. In this case,
a coherencyenhancement programcan be used, which usually involves sometype
of trace mixing or FK filtering. However,the interpreter mustbe aware that
any mixing of the data will "smear" trace amplitudes, makingthe inversion
result on a particular trace less reliable.

Coherent noise is much more difficult to eliminate. One of the major

sources of coherent noise is multiple interference, explained in section 2.4.
Two of the major methodsused in the elimination of multiples are the FK
filtering method,and the newerInverse Velocity Stacking method. The Inverse
VeiocityStackingmethod
involvesthe followingsteps:

(1) Correct the data using the proper NMOvelocity,

(2) Model the data as a linear sumof parabolic shapes,
(This involves transforming to the Velocity domain),
(3) Filter outthe parabolic
components
with a moveout
greaterthansome
pre-determined
limit (in the order of 30 msec),and
(4) Perform the inverse transform.

Figure 4.8, taken from Hampson(1986), showsa comparisonbetweenthe two

methodsfor a typical multiple problemin northernAlberta. The displays are
all' co•on offset stacks. Notice that although both methods have performed
well on the outside traces, the Inverse Velocity Stacking methodworks best on
the inside traces. Figure4.9, also fromHampson
(1986), showsa comparison
of
final stacks with and without multiple attenuation. It is obvious 'from this
comparison
that the result of inverting the section whichhas not had multiple
attenuation would be to introduce spurious velocities into the solution. The
importanceof multiple elimination to the preprocessingflow cannot therefore
be overemphasized.

m i i m , i . i m _ i i _ L ,=•m__ _ i m ß •

Part 4 - Seismic Processing Consideration• Page ½ - 14

Introduction to Seismic Inversion Methods Brian Russell.

!lilt
tiiti
ll!1111iitt
i)tt
iltli
ii/lit
tttl•
ill

(b] LINEld8 - 2-D MIGRATION

IIIIIIll!!1111111111111it
I!1111111
I!11111111111illl
illIiIIIIIIIIIil!111111tllilil!illlllll!111illlll
[1111111111111111111111111
III!!1111
I!111111111111111
IIIIIIIilllllllll!111111111111111111111
?•111[•i••
IIIIIIIII
!1111111111111111
IIII!IIIiill•illlllillllllllllliillllllllllllh

•., }!l!iilll
•lllllilllllll
i!iiJ
:illllllllllllilitiilillit!illllll
,o

111lllllllllllllllllllll1111llllll
Iilllllll!ll!llll
I111
illllllllilllllllllllllllllllllllllii{lilllllllllllll
"•fillllllllll!1111illi!111
IIIIIIIII
IIIIIII1111111111
IIIIIlillilllllll!1111!1!111111111
'•
Col LINE 108 - 3-D MIGR•ATION
F•g. 4.7. Migrationof modeldata shown
in F•g. 4.6.
- - -- (Hermanet al, 1982).
Part 4 - Seismic ProcessingConsiderations
ß
Page 4 - 15
 Introduction to Seismic Inversion Methods Brian Russell

AFTER AFTER
INVERSE VELOCITY STACK F-K MULTIPLE
INPUT MULTIPLE ATTENUATION ATTENUATION
J. ' ' ')'%':!•!t!'!11!1'1 ';.•m,:'!:',./-•-•l- •r'm-- all

" "';;:.m;: .... ,;lliml;..• .

m#l

Fig, 4.8. Commonoffset stacks calculated from data before multiple

attenuation, after inverse velocity stack multiple attenuation,
and after F-K multiple attenuation. (Hampson, 1986)

888
Zone d
Interest

1698
-4

Secondreal-dataset conventional
stackwithout
multiple
attenuation.

'•",• ......
•,,t.•/:,.•t.,. ). ;•,•<,:u(•:'J,.•J
L,.•.,!-
I',,', ,'; •,
• •,
, , I• •,,,,,ß..... '1"' ',''.•..;•t(•' •,)"•,'.m,,•""•.
•,,,•• '•;••
• ,ii%' .t .% '.
, ,,,, • ..•'•t,..'•"•'i•' •-
lee ---';•-•'
"" • "" • • ' "' "•'"t" 1•%';J•
ß' ß' • •t•, ß.... -.... ; -'".' ,•..'....
'. 2•>.': '..'•, •;,%"'•1
'" "' Zone of

,,, .t•iill••)•.•);•l',"P,'•)'•"•'".•r'"mm"•""•P"•
,,..,.,,..,,,_. "•)r'"
gt••'
•..,,,. ' '" •- ..... ,•
•,,.,....•.,...,..,...,..,.•..,....,,,.,.,.. Interest
••,•,•p}•h•?.•,.•,,
r•.•,•. •.}•.U.•,..,
,'......•.•_ ), •.•,
,,,,.,.•,,,,,..,
.l•,• • ,•,•m,l,•,
' ,,{..'•' ',•'•...r'•
.. ,,,
........ ,nm,
....
,,,, '.""::•"'•'•""""="'""•"
../•.• ,•.•'•, .'•-•%....
";'
Fig. 4.9. Second real data stack after inverse velocity stack
multiple attenuation. (Hampson, 1986)

Part 4 - Seismic Processing Considerations Page 4 - 16