some techniques of enhancing seismic reflection...

JOURNAL OF THE CANADIAN SOCIETY OF EXPWRATTION GEOP”“SICISTS VOL. 21. NO. 1 ,DEC. 19851. P 15-29

SOME TECHNIQUES OF ENHANCING SEISMIC REFLECTION DATA FOR STRATIGRAPHIC INTERPRETATION’

SUDHIR JAIN?

ABSTRACT

The computation of seismic velocities from amplitude vatia- tions on the seismic trace is influenced by the limitedfrequency- band of Ihe data. static correction errors and limited energy penetration. These problems are being attacked in the data acquisitilJnphase. Inaddition. thereareproblemsoforganized misc trains. intrabed multiples. random noise and wavelet distortion fram trace 10 trace. This paper discusses the techniques o,f solving these problems by using the data itself. Random and organired noise is attenuated by means ofa short two-dimensional operator, intrabed multiples by an adaptive deconvolution technique, and wavelet distortion by determining a wavelet for each trace and applying aperators to reduce it to a spike.

All these techniques increase the reliability of computed velocity variations but do not provide absolute velocities. Thus, il is possible lo predict lithologic variations but “01 absolute lithologies.

In the last twenty-five years acquisition and processing techniques for seismic reflection data have made incredible strides, even enabling the prediction of lithologic variations with accuracy and detail. The field records used as late as in 1960 for mapping structures are not even known to many seismologists of today. Immense advances in computer technology, coupled with the application of communications theory to harness this technology for the seismologist, have made it possible to explore small reservoirs at great depths with success ratios as high as fifty percent.

CDPrecordinganddigitalprocessingremain themost significant advances in seismic reflection technology since the introduction of the reflection seismograph. Another advance was the computation of the seismic velocityfromastaked trace(Delaseta/., 1970;Lavergne, 1974: Lindseth and Street, 1974; Lindseth, 1979: Cooke and Schneider, 1983; Berteussen and Ursin, 1983; Oldenburg,etal, 1983;WalkerandUlrych. 1983),partic-

ularly after the computation of the propagating wavelet and the reduction of the stacked trace to the reflectivity index series (Jainand Wren, 1977). The experience with computed velocities over the years has pointed to certain deficiencies in the stacked data that prevent the accurate computation of velocity sections, and various techniques have been developed to overcome these deficiencies. Some of those techniques are the subject of this paper. They have been developed in the industry over the last ten years and are in wide use now.

The problems in acquisition and prestack processing -accurate static and dynamic corrections, true amplitude recovery, recording and preservation of wider frequency bands-are significant and need to be carefully investigated. However, they are not the subject of this paper.

SOME IMPERFECTIONS OF STACKED SEEMIC DATA

In spite of the best efforts in data acquisition and basic processing, seismic data contain, to a variable degree, the following important distortions:

I. Complex structures are misrepresented on stacked section. In addition, diffraction patterns caused by sharp structures confuse the interpretation of underlying interfaces.

2. The attenuation of reflection energy as it travels downward is oftencompensated by applying some variation of automatic gain control (AGC). This procedure is based on the assumption that the average of reflectivity indices over an interval is approximately constant. This leads to sections that are convenient for structural mapping but are notaccuraterepresentationsoflithologicvariations.

3. In addition to primary events, the stacked sections contain non-reflection energy such as shot- generated noise, converted waves, etc.

4. The energy contained at higher frequency levels, generally above SOHz, is considerably less than at median frequencies of 25 to 40 Hr. When the

‘Presented at the Joint CSEG-AGU Meeting held in Calgary. May 1985. ‘Commonwealth Geophysical Development Company, Ltd.. #1X,1. 505 3 Street S.W.. Calgary. Alberta T2P 3E6

The author has worked closely for sev.val years with A.E. Wren and J.D.T. Crane of Petrel Consultants. Lorne Kclsch of PanCanadian Petroleum and W.W. Soukoreff of P&o-Canada I.A.C. and had benefitted greally from these associ&ms. Colleagues at Commonwealth Geophysical were very cooperative during the preparation of this paper.

The kindness of several companies in permitting publication of these data is gratefully acknowledged.

15

SUIIHIK JAlN

signalinoiseratioat these frequencies isconsidered, this reduction in high-frcqucncy energy hccomcs even “lore Critical.

S. The deconvolution normally applied in prestack processingetfectively attenuates short-period rcver- herations and also stahilires the wavelet at various depth levels. However, the section still contains i&abed multiples of periods ranging up to several hundred milliseconds.

6. As indicated earlier, some of the changes in the seismic sections caused by small stratigraphic variations are too weak to be observed on normal sections. It is desirable to have some means of presentationwhereonlythechangesareohserved.

SOME SOLUTIONS FOR IMPFKFECTION~

IN SEISMIC DATA

STRUCTURAL DIST~U~-I(~N/DII’~KA~~I~NS

The stacking process normally assumes that the retlec- tion point is located exactly in the middle of the source and receiver points, and that the reflection point is located vertically below the trace location. This is true when the reflectors (and all the interfaces above) arc horizontal. In the presence of any structure or lateral velocity variations the reflection point is displaced from themidpoint. Thcmagnitudcofthisdisplacement depends on the dip of the reflector, the magnitude of lateral velocity changes and the distance between the source and the receiver. The displaccmcnt is always such that the apparent location of the reflection point (on the stacked section) is downdip ofthe true location and is in theoppositcdircctionsfortwolimbsofafold.Therefore, the anticlines appear gentler and the synclines sharper than they actually are. In sharp synclines, the two limbs may he displaced enough to cause them to appear as anticlines.

Another complication caused by complex structures is the appearance of diffractions that are reflections from point sources such as the edges of faults. On occasion, the diffractions help to identify faults. On the other hand, there are situations where they cause amplitude anomalies by interfering with genuine reflections. In the Western Canadian Basin, diffractions may or may not be significant in exploration of Cretaceous sand reservoirs. In detailed investigation of Devonian reefs in ccntl-al Alhcr-ta or of Mississippian gas rcsc~-- voirs in the Alberta Foothills, diffractions are a source of confusion and need to he collapsed. Both - the collapse of in-lint diffractions and correction of strut- tural distortion - can be achieved in one round of migration of reflection points to their true positions (Rockwell. 1971 :Clat-eboutandDoherty, 1972:Sattlegar and Stiller, 1973: Stolt, 1978). Generally the migration is applied to stacked traces. When dips or source-receiver intervals are large, the displacement changes within the

CDP groups can be detrimental to the signal quality and migration may he desirable before stack. Jain and Wren (1980)described aneconomic processforprestack migration based on Kirchhoff s summation, and illustrated it by an example shown in Figures I, 2 and 3. Figure I is a stacked section from the Alberta Foothills and is dominated by n gentle anticline and numerous diffractions. Migration of these data by means of finite-differcncc solution of the wave equation is shown in Figure 2. Incidentally. the Kirchhoff summation produced a very similar section. Prestack migration is shown in Figure 3. Both migration processes succeeded in collapsing diffractions and in correcting the structural distortion in marker A, but the prestack migration improved marker B more then the poststack migration. In our experience thih example is an exception, and in most cases the improvementmightnotjustifyconsiderablyhighercosts.

Migration proccssing causes some attenuation in the high-frequency component of the data. Therefore. the accu~-ate representation of structure must he balanced against possible loss of resolution of thin beds. This factor becomes important when data are being analyzcd for delineation of thin reservoirs (Jai” (‘I al., IYXI).

12 FOLD STACK

Fig. 1. Stacked section lrom Albella Foothills

FINITE DIFFERENCE MIGRATION VERTICAL AMPLITUDE VARIATIONS

210 200 790 180 170 160 Various processing systems have been developed to compensatefordifferential attenuationofseismicenergy with tl-avel time (Jain, 1975; Taner and Koehler. 1981; O’Brien er ul., 1982). Most commonly. the gain applied during recording is subtracted and replaced by some linear function. The spherical divergence compensa- tion is made based on a generalized average velocity curve. This is followed by a detailed analysis of amplitudes for individual source and receiver locations to computenecessaryamplitudecompensationsforsource and receiver environments. Often this procedure is bypassed in favour of window averaging. which can be called slow digital AGC. While this AGC products sections of uniform amplitudes and helps in mapping weak retlections, it suppresses amplitude variations laterally and vertically. To illustrate this difference. a synthetic trace was computed from an offshore well without amplitude adjustments in the reflectivity index (RI) series and with a 300 msec AGC. Figure 4 shows the synthetic traces. computed RI series, and original computed velocities from both synthetics. RI and velocities were computed by using the technique described by Jain and Wren (1977). The match hctween computed and well velocities is very close for the length of the sonic when AGC was not applied. On the AGC synthetic. the characters of the sonic and inversion v&cities are

Fig. 2. Poststack migration of section in Figure 1 very close, but the magnitudes of velocity changes are distorted in various places, particularly above I. I sec.

MIGRATION BEFORE STACK

2.5-

ENHANCING; SElSMlC REFLKTION DATA I7

and below 2.2 sec. Apparently, AGC can distort velocity variations laterally as well as across the section, ;md the swxess of true amplitude recovery in the stacking process is crucial to the recovery of the magnitude of velocity val-iations.

ATTENUATION OF SPUINNJS EVENTS

Theuseoffrequency-limitingfil~ersroattenuate’noise’ is as old as seismic recording. Spatial filtering in the form of weighted or straight mix is also an established practice. The first application of two-dimensional filter using both - time and spatial - elements was by Embree et ul. (1963). They introduced the f-k domain filters with pass- and reject-bands passing through the origin. The operators are quite wccessful in attenuating random and organized noise and enhancing particular dips. However, these operators have to he quite long in the space domain. Consequently, the signal character is invariably distorted, and small features that may be diagnostically imporrant are attenuated. Worse still, the reflection energy in any particular dip direction may be enhanced and the rest attenuated at the processor’s will. Thus, any structure may hc created by a determined processor.

Fig. 3. Prestack migration corresponding to Figure 2,

The purpose of high-resolution seismic reflection sw veys is to record wider spectral band than is possible withmultiplegeophoneand/orshot arrays. Unfortunately,

18 S”DHlR~JAlN

INVERSION COMPUTED

2.5

Fig. 4. inversion of synthetic with and without AGC.

SYNTHETIC SYNTHETIC asc

+ 0

0.5

1.0

1.5

2.0

2.5

ENHANCtNG SElSMlCKEFL.ECTlON DATA 19

randomandorganizednoise-trainslikewindnoise, hole- blast, and distant traffic movement are also recorded most faithfully. Some of this extraneous noise may be attenuated by using frequency band limiting filters, which defeats the purpose of high-resolution surveys. Five- or seven-trace mix with constant or variable weights has had some success in attenuating wind noise, hut the discrimination between signal and noise is very crude.

Jain (1979) discussed straight two-dimensional filtering using short-operators, which overcome the ohjec- tions outlined above. The operator is designed by inverse Fourier tl-ansform of the desired two-dimensional frequency response using a hybrid truncator, which pro- videdoptimumresponseinalargevarietyofexperiments. The operator is applied to the whole data set, which minimirer the probability of enhancing anomalies beyond their appropriate significance level as frequently hap- pens in filtering based on coherency estimation.

Figure 5 shows a seismic section from central Alberta where the! prospect is Cretaceous sand at approximately 0.9 sec. Figure 6 shows the noise component extracted from this section by using the two-dimensional reject filter designed from the desired frequency response. The absolute gain has been adjusted for display purposes. The display consists mainly of very low velocity events but some flat events are present around I sec. These apparent events between 0.9 and I .2 sec. were analyzed

STACK

for time, wavelength and amplitudes, and one such analysis is shown in Figure 7. Amplitude of the event changes abruptly from frace 10 wace by about 4 db, wavelengths by about 20 msec, and time of peaks by more than IO and up to 20 msec. Considering large trace fo trace variations, it is fair to regard these apparent events as mis-stacked component of signal - in this case probably due to static correction errors.

Figure 8 shows the section obtained by subtracting the noise trace from the stack data. Trace-to-trace variations in time and amplitudes have been reduced. It is now possible to pick shallow events that are hidden by noise in Figure 5, and small changes in the zone of interest are more reliable than on the stacked section.

HIGH-FREQUENCYENHANCEMENT

After the noise component has been attenuated by two-dimensional filtering, it is possible to apply zero- phase operators to enhance the high-frequency component of the data. Figure 9 shows a portion of a stacked marine section. The data have been filtered by a two- dimensional operator. After applying a zero-phase operator, which is determined from the power spectrum of the trace and is designed to enhance the high- frequency component ofthe data, we obtained the section in Figure IO. The improvement in resolution is obvious

Fig. 5. Stacked section from central Alberta.

20 SUIX,IR lA,N

2 0 RLJFCT

Fig. 6. Noise coomponent of data in Figure 5.

REJECT I

F ? ;- 11 4l 1.n

Fig. 7. Character study of event at 1 .O set on Figure 6.

Fig. 8. Figure 6 after subtracting noise component (Fig. 7).

2-D FILTER

I 27 54

2-D FlLTE’(

HIGH FREQUENCY ENHANCED

I 27 54

Fig. 9. A portion of stacked offshore data after two-dimensional filter. Fig. 10. Data in Figure 9 after high-frequency enhancement.

22 S”DH,K JAlN

at all levels and particularly in the boxed area. Note that amplitude relationships of the events arc preserved throughout.

Figure I I compares the amplitude spectrum of the data shown in Figures 9 and IO. The two-dimensational filter has preserved the spectrum in all details. High- frequency enhancement has increased the amplitude levels above 40 Hz. The peak frequency is 50 Hz instead of 25 Hz, and 25% down points are 34 Hz apart instead of 20 Hz apart, indicating a broader spectrum.

Figure I2 compares the computed wavelets corresponding to the spectra in Figure I I. The wavelength of the wavelet after high-frequency enhancement has been reduced to 18 msec from 28 msec on the stacked data. Following the quarter-wavelength rule (Wide\% lY73),

I *-

FRtOUtWCloHl

“I “I

- STICI

~--~ 10 SILISF

- iD l,lTrr _,r~ I”rO r*I^YLr-rYT

Fig. 11. Spectraofatracein Figure9 beforeandanertwo-dimensional filter and after high-frequency enhancement.

- j. s,LTrP wlr^ .msm EN*L*CLs”INI Fig. I 2. Computed propagating waveI& corresponding to spectra in Figure 11.

the thickness of a thin bed that can be detected is reduced to 4.5 msec from 7.0 mscc of two-way time.

IwtutxL) MULTIPIIS

Seismicdatainmanyareascontaincontinuousevents that do not match the sonic logs and cannot be explained from known geology. Often, the major markers in an area show local structures that do not stand the test of the drill. Detailed analysis shows that some of these eventsareintrabedmultiples. whichinterferewithgenu- ine rctlectors tocausepseudo-structures. Identification andattcnuationoftheseeventsisnecessaryforsuccess- ful exploration in such areas.

Normal-moveout differential has been used toattenu- ate long-period multiples. either by filters based on coherency Clain. 1976) or by f-k filtering (Ryu, 1980). However, in the case of intrabed multiples the moveout differential is very small. In these situations, an adaptive deconvolution technique has been used successfully on stacked data (Griffiths rrn/., 1977). The top part of Figure I3 showsaportionofastacked section. There is evidently no significant multiple problem on the section. However, by using time-continuous forward and reverse adaptive deconvolution operators. intrabed multiples of various periodicities were predicted from the data. These multiples are shown at the bottom of Figure 13. Subtracting them from the stacked data produced the section in the middle of Figure 13, which is very similar to the original stack except for the feature within the box. On the original section of the dip reversal has been masked by the multiples. Unfortunately, many of the stratigraphicfeaturesofexplorationinterest arecompa- noble to the event shown in this example, so that it is crucial to attenuate intrabed multiples even if these are not significant for structural mapping.

IDENTIFICATION OF SMALL ANOMALIES

In many cases, stratigraphic variations do not cause dramatic amplitude or structural anomalies on the section. A carbonate section becoming porous or the presence ofgranite wash may cause achange in amplitude of 10% or less, which is hard to see on a section where the maximum trace amplitude is about five millimetres. Detailed examination of these anomalies is helped by the application of techniques conventionally used to isolate small anomalies on potential field maps (Jain, 1979). A regional seismic section is computed by applying alarge(71 traces, 36 msec) two-dimensional operator, which is designed to pass only the anomalies of large nrcnl extent. The difference between the original and regional sections - the residual section - shows the anomalous component of the stacked data.

Figure I4 shows a twelvefold section from central Alberta. The boxed area is the zone of a known Nisku reef. The regional section (Fig. IS) looks very similar to the stxked section. Figures I6 and 17 show both polarities

ENHANclNGSElSMlCREFLECTlONDATA 23

FINAL STACK

PRIMARIES

INTERBED MULTIPLES

Fig. 13. A poinion of IEfold section (top), predicted intrabed multiples (boflom). and after subtracting multiples (middle).

of the residual section obtained after subtracting the regional from the stacked section. The extent of the reef is very clearly identified on these sections, and the calculations; about vertical and lateral dimensions can be made more accurately on the residual than on the stacked section. Evidently, the residual section has successfully isolated the anomalous reef. The same process is applicable to the inverted seismic sections, to detine the nragnitude of velocity change over an anom- aly more clearly than a normal velocity display.

DEDUCTIONOFSTKATIGRAPHYFROMSEISMICSECTION

MODELLLNG

Assumingthat theseismicsectionisatruerepresenta- tion of subsurface, there are two ways of deducing stratigraphic details: modelling and inversion. Model- ling is generally cheaper and quicker. The structural or lithologic variations are assumed; reflectivity indices are computed for a set of traces and filtered by a prespecified wavelet. The model synthetic (Fig. 18) is

24 SUDHlK JPJN

IZ- FOLD STACK I7 Ii l3Oll 9 ,la ,_ 7 5 3 (

L.... _ :‘y.p”.. -.,,.,,,,,, ,*.* ,,,,....-...... *... .,...

*tt:,,i(i:“‘s..“.“,mmmFiimril "i~li~I1li),iiiiiiiil~!!! ,,,,,, L:l~r”l,ll.,,,,,,..,,,,,~,,,, t

L6- ,, ,,,,,,,,,, ,A~Y~!~,m .y- -r,N 1,t1b>1,P., .,.,,,,),,,,,,,I. _ ,)))))~wumr~~b,~r~~,)~~~~~)~~,D~~~~LlliLbbbr~~,,.~,l,,,,.,,rr,,,~,~ ~~.~

2,0 '-c~~-'-..;.;..~.u"~ --.i,r)rriri),r,,,,...,...,., -..,., ,..G. -*rru,,r,,,,r!:~rrr-,,r,,..*,,,,i,,,,rr.rrrrrrr,, * ,,,.,,.,,...>.....v I,.,-),,, ~~l,~~~~.,......-.,,. ,,,.,...,,,., ,,,t

Fig. 14. Twelvef0ld stacked section from central Aiberia.

REGIONAL GECTDN

15 Boll 9 7 5 r...’ :..‘.:

3 ~ ..:...’ 1 i

nnHHH,,,,,,,,,,,,,,,,,,,,

,,,,,,,,,,,,,,,,,,,,,,,... Pp.““.“““” . . . . . . . “” . . . . . . . . . . . . . . . . . ..I -,‘I’,‘,‘~‘,,))),,~ ,,,,,n ~,,~,,llr,l,r,llrl,l~L)rlr)lrl,))),),),,,~,~,~~~~~~~~~~~~~~~~

3, ,,,,,, ..,,,,,,,,,,,~~~~~..,~~,,~,.,.,,..,,,,,,.,.,.,..,..,,,~,,,,~.~~~~ 1.6,i111ir~r),~,,i,~,~,,,,it,,,,,l,,,,i,,,,,~~~~~~~~~~~~;~~~,~~~~,~,~~,~~~~~~

-“,,~,~.~.~.~.~*~.~.~‘.““.~...‘...... . . . . . . . . . . . . ..I......... ...,..,,.,,~,~,lll~l ~~,I~I~~I~IrlI,I,,lii~l,~,,~~~,lr~~r~,I~,I,l,i ,111 !,j /,/,, !,l,,i;!,!l,!l~! ,)),,),),,,,,,,,1)1/1,,,,,,,,~,,,,~,,,,,,,,,,,,,,,,~,,,,,~,,,,,,,,,,,,,,,

,,6*““““~rrr..,rr.....r..~.....~......*.~.~~.,,,,,,,,,,,,,,,,,,,,,,,,,, ,,,,, ~~~~I~,~,1/,1,~,,1,1,11,,,,1,,

,,,,,,,,,,,,,,,),))))))))))))))))))))))1,,,~,~~~~,,~,~,,,~,~~~,,,,,,,~,,, -;ij;;i;,iiii,iiu,i~b~~~bb~~~~~~~~~~~~~~~;~i~i~i~~~~~b~~bbb~~b~~~b~bb~

Fig. 15. ‘Regional’ section corresponding to Figure 14

Fig. 15. ‘Residual’ section computed tram Figures 14 and 15

RESIDUAL SECTION, REVERSE POLARITY

I3 ,,2’:>,,,I’,....;~;‘: :~~ A,

01, 9 7 5

,,~,~,,,I~~~~r,~~~~r~~,!~lllli(i~i~I~~~~i~~;’i~‘,‘~,~~il:~~i~~, !:;;;I,, 7, .~ ;

‘~‘iill.ll:I;;;jI,‘!‘!j,!i’!lliji,rii,’,’,!,!!i”’ \’ I,i,l)l~~;‘:,“l~lil~~,,, -‘cl ,:I L;~i),i,,i,,Lill;:’ ‘l-,;,:,;I!::I;:::!ii,;,i~,iil;l;;:;,;;tjiji:,:

,,.....~>.:1u,,. ))),” Y...)bb, ,.4~~~~““““““. ,,,,,,, ,y;,L,:!:!!!ll”

~““L~‘)‘,.,,), Ili!i,,,,,,,,,,,,,rrrI;,,, ‘t(;: ~~!!~~\‘!,!1,!,,.I!!!~‘~”

.,1,,111111,),),),,1Ilr)) _i:!::,::.,‘,,~~,,,,,,,‘i;;;:;:I;,;,,,,,

,:,:‘,‘il,l,,l:iliii~l, ,,I, ,!“‘l~::;,,,,II::;:;;l~,rllli;;;

,,,,,r, II!i,,,,)~,(““‘~ ,,,. b’.ibll*‘,i,,,,,.~:~

-,: ,: ,,, ,,“‘,),!,,1;.1!~1,!:,~,

‘Illll!::,,,),l,li,,IJ/)))li:ll,” ‘“~,~~r~~~~;i~~~~,;.;.;.~:.,~.~~~~~:i -J,,..be..s~;ji b ,,,,,!,,,~,...,,,I,,,*~ ~)i,ii),ll”““““““““”

,‘il”l’l,,,,:,,;,.,~ ,t>;,.,,,t111.>v .,,‘t”r:!!! ,,,,,, :,>, “,,,,..m,,,,..::;;;ir:“::rii,,,,,),,,,,,,,’, ‘,I), ” ,,,,1,, ~.L,~h~~illi!“! ,.6,!,“:~:,i,,,+‘:::.“: I,,) Il,.‘!)‘“...~.-‘~‘:‘::I.: I,,, 1,,,...),, ;.:Qi),ill’ ‘~1,!!!!flllll~.::::;;;,., ,,I,,

,)) / ,,,,), ‘,i,iii;iii )),, ,,

:,,,,,,,.,.,,, 1’)’ ,_,, :;,,; l;,l:! -,1I;~,il:j:l~iiiji)l~~!li,,!,!,f)r:lI:..-!.:.::l;,:;,i,:!~lr,i,jl~,,,,ji

,,,,,,,; ,: !l,r~~“,‘,!““‘f: .,,,,,,i,i)‘,),))l,),i),,,i, ~, ,,,),, ,,:l~‘,‘,.‘,,,i.s;.,,i;,,, z.o!r,!:,l:.;,.~!L f’,

I ,r,..““‘“~~~~~~.~~::ii;i;,lr,,‘l,,,,,-l~ ,f%%w )))). ‘I:!!‘,.‘~l-;~~,:i~,~:,,~“‘”

,i,iiil:II)) ‘t>,>,;> ,,),,,,. : ,),I ,,,,, ;..,,‘:‘?r,,’ ,,l,!l’,~?,i, _:::~I::::::~..,r~,,ii,i....,“.~i~!:,::::,,, ,,,... ..I,,,,,,.,. ,-...*.,. .,,,..,_. ~~~..~. >

Fig. 17. ‘Residual’ section with opposite polarity with respect to Fig ure 16.

now compared with the stacked section. The model is altered until it generates B synthetic that matches the section within allowable limits. Instead of starting from an assumed model. one can start from a sonic log and make changes in the zone of interest. Again, reflectivity indices and synthetic traces are computed and compared with the stacked traces to select the model that best matches the data (Pig. IY).

The modelling works vuy well with thick anomalous zones. However, when very small features are being explored. the following problems are encountered:

a) The wavelet is a critical factor in determining model response. On a seismic section the wavelet may change from trace to tract.

b) Thcwavcletintegratesthercsponseoverthedepth range represented by its length. Thus. even if a model inr~w& to show the change in reflection character due to porous sand, for example, the aclual match may be due to a coal band in the vicinity.

c) The comparison of the model response and the seismic sction is made visually, and is obviously restrictcd bythedynamicrangeoftheplot. Changes smaller than 102, in amplitude or 4 mscc in wavc- length may not be noticed by many interpreters.

INVERSIOU: COMI’UI’.UION OF VFI.OCI’IY FROM StlSMlC ‘I‘RACF

This process is generally a two-stage one. In the first stage. the stacked tract is used to compute a wavelet. and an opcratol- is designed to reduce the wavelet to a spike. This operator is applied to the trace. Thus, a series of spikes approximately proportional to the reflectivity index (RI) coefficients is obtained for each trace. The RI series is, in turn. used to compute acoustic impedance for each sample. Commonly. some empiri-

ENHANCtNG SEISMIC

MODEL

Fig. 18. Synthetic seismicsection generated from assumed geological model (shown in the top part of the section).

KEFLECTlON DATA 25

cal density-velocity relationship (for example. Gardner era/., 1974) is assumed to present final results as velocities (Jain and Wren, 1977). These velocities can be directly related to lithology in the zones of interest.

The problems with this approach, even with excel- lent data, are:

4

b)

Evenassumingthatempiricalvelocity-densityrela- tionshipsapply. RItracecontainsinformationabout velocity changes, not the absolute velocity itself. To compute actual velocities, a velocity value must be known at any one point on the trace. The RI section does not contain valid information below approximately IO Hz. The significance of this limitation was demonstrated by Lindseth and Street (1974) and Wren and Jain (1978). The information must be obtained from some other means. Sonic logs from nearby wells and detailed normal- moveout analysis are two common methods.

The recorded frequency band is also restricted on the high side. Good data contain frequencies up to 80 Hz in ideal circumstances. The frequency rangeofthedatadetermines the wavelengthofthe propagating wavelet which, in turn, determines the limit of resolution of the data. Widess (1973) showed that the limit of resolution in the noise-

e MODIFIED LOG RESPONSE It

Fig. 19. Synthetic seismic traces generated from a sonic log lor assumed changes in zones of interest.

26

VELOCITIES I w ii

0.5

I.0

I- 0

-0 370

,0.5 - 41 I

-1.0 -1714

RESPOr

WAVEI

SUDHIR JAN

R.I.

USE

LET

0

STACK 2D-FILTER

-0

-0.5

-1.0

Fig. 20. A set of stacked traces from southern Alberta. RI s?ction and computed velocities.

ENHANCING SEISMIC REFLECTION DATA 27

free (iata is one quarter of the signal wavelength. For ,thinner beds, the only information on the sections is in the form of acombined reflection for the bed’s top and bottom interfaces. Beds thinner than the resolution limit may be identified from changes in reflection time and amplitude. Gener- ally thin beds are seen in the form of small velocity changes on the inversion section. The magnitude of change depends on both thickness and velocity contrast. The recovered velocity contains this blurr,ed information, and deduction of the absolute value ofthe changes in thickness and/or velocity is often not possible.

c) RI values recovered from seismic traces are proportional to actual RI ~values, and the propor- tionality constant has to be determined. Without this i,nformation, the magnitude of velocity variations cannot be predicted.

d) The wavelet can be determined fairly accurately from the data, and its polarity established in most cases,. However, to be absolutely certain, some external confirmation is desirable.

e) Seismic traces are independent of each other. Therefore, oneneedstheproportionalityconstant, low-frequencyvelocityfunction,and knownveloc- ity point for each rrucr. It is usually assumed that constants determined at a known point can be extrapolated by taking structureintoaccount. This assumption may often not be valid.

Experience with data from all over the world shows that, if a sonic log is available in the vicinity of the seismic dal!a, all but the last of the above problems can be largely resolved, to the extent that one can predict the variations in velocity and therefore in lithology. The prediction of absolute velocity, however, is not possible. The imperfections in computed velocities make the preparation of depth sections from inversion velocities very hazardous because of their cumulative effect down the section.

Figure 20 shows a normal stacked section from southern Alberta on the right and the reflectivity index section derived from the stacked data in the middle. The wavelet computed for each trace, and the response of the operator used to reduce the wavelet to a spike, are shown at the end of each RI trace. On the left of Figure 20 is the velocity section computed from the reflectivity indices. The sonic log from the well located close to the station is shown after integration at 2-msec intervals. A very close match between the computed velocities and actual soni,: velocities is possible because a reasonably accurate wwelet could be computed and an accurate RI trace derived. This allows identification of relatively thin zones with confidence, and makes it possible to predict the behaviour of these zones away from the well.

In favourable circumstances, given enough control, it is possible to define a reservoir in great detail, as

shown in the study of the Angel gas field by Jai” et al. (1981). Figure 21, taken from this paper, shows the net pay map of the Mount Head based on the geologic interpretationandinversionofsevenseismiclines. Zero line on the map represents a gas/water interface at the depth of 4337 ft below sea level, and the net pay thickness is contoured.

If properly done, inversion of the reflectivity index section provides a reasonably accurate replica of sonic logs with maximum possible resolution. It matches the actual integrated sonic log withoutanysmoothing m-filtering of the sonic log. Unfortunately. the velocities computed from seismic traces are not sufficiently accurate in ahsolure terms to be able to convert inverted sections intodepth-velocity logs without introducing unwarranted structural distortions. However, in most cases, the velocities are accurate in relative terms, and reasonable conclusions can be made regarding the presence of sand beds, carbonate stringers, porosity changes in known beds, the presence of porosity in carbonate formations, etc. While it is true that the reliability coefficient of these conclusions is directly proportional to the amount of prior knowledge about the area, careful application of inversion techniques can be useful even in virgin areas.

CONCLUSION

The attempts to locate stratigraphic traps on reflection seismic data brought to focus many shortcomings of the data - the presence of organized and random noise, the occurrence of intrabed multiples, misrepre- sentation of true structure, difficulty in isolating small structures, and insufficient resolution. Processing techniques have been developed to minimize these factors, to isolate very small features and to predict lithologic variations. Even if absolute lithologies cannot be predicted without well control, a skilled interpreter can obtain some quite useful information on lithologic variations.

REFERENCES

Berteusren, K.A.. ,983 Approximate computaliun uf the ac~wtic impedance from seismic data: Geophysics, Y. 8. p, 1351-1358.

Claerbout, J.F. and Doherty. S.M.. 1972. Dawnward continuation of moveoul-corrected seismograms: Geophysics. v. 37. p. 741-768.

Cooke. 0.A. and Schneider. W.A.. 1983. Generalised linear inwrsion uf reflection seismic data: Geophysics. v. 48. p. 6654776.

Delas, 0.. Beauchomp. J.B., delombares. G.. Fourmann. J.M. and Postic, A., 1970, An example of practical velocity detcrminati”ns from seismic traces: Paper presented at EAEC. 32nd Meeting in Edinburgh, Scotland.

Embree,Peter,Burg,J.P.andBackus,M.M., 1!263,Wide-bandvetoc- ity filtering. Ihe Pie-Slice Process: Geophysics. v. 2X, p. 948-974.

Gardner. G.H.F.. Gardner. L.W. and Gregory, R.A.R.. ,974, Forma- tion velocity and dewily -the diagnostic basics for Stratigraphic traps: Geophysics. Y. 19, p. 770-78”.

Griffiths, L.J.. Smulka. P.R. and Trembtay. L.D.. 1977. Adaptive deconvotution ~ a new technique for processing time-varying seismic data: Geophysics, v. 42. p. 742.759.

SIIIIHIK JAlN

: J .* 1%: ‘; k % ,

i LEGEND

ANGEL FIELD

NET PAY MAP MISSISSIPPIAN MT. HEAD

CUT-OFF 3% 4

KALE: 1:50.000 CONTOUR INTERVAL: IO F I t. Fig. 21, Net pay map for Angel Gas Field computed from seismic inversion and geological interpretation.

ENHANUNG SEISMIC REFLECTION DATA 29

Jain. S.; lW!i, A sysrem of real amplitude recovery in land seismic recordings: CanadianSocieryofExploralionCleophysiciilsJournal, Y. I I, p. x47.

~. I976. The application ofcoherent slack to data dominated by multiples 01 by incoherent noise: Czanadian Society of Exploratmn Geophysicists Journal. v. 12. p. 23-37.

~, 1979, Two-dimensional filtering of seismic data: Canadian Society of Exploration Geophysicists Journal. Y. IS. p. 40~49.

~, Kelrch. L., Hamilton. D. and Andrews, M.. 1981. Reservoir delineation of the Angel Gas Field in the Alberta Foothills using seismic inwrsion: Slst SEG Meeting. Los Angeles.

-and Wren. A.E.. 1977, Theapplicationoftheoptimalwavelets io seismic inversion: Canadian Society of Exploration Geophysi- cists Jaurnid, Y. 13, p. 3-14.

__ and Wren, A.E.. ,980. Migrakm before stack-procedure and significance: Geophysics, v, 45. p. X4-212.

Lavergne. M., 1974. Pseudo-diagraphics de Vilesses en offshore profande: Geophysical Prospecting. v. 23. p. 695-71 I.

Lindseth, R.,O. and Street, A.V.. ,974, Velocity information and seismic traces: Paper presented at the 44th Annual Meeting of the SEG in Dallas. Texas.

-. 1979, Synthetic sonic Iogs, a ~)rocess for stratigraphic mter- prcmtiun: Geophysics, Y. 42, p. 3-26.

O‘Brien. J.T.. Kamp. W.P. and Hoover. G.M.. 1982, Sign bit amplitude recovery with applications to seismic data: Geophysics. Y. 47, 0. 1527~1539

Oldenburg. D.W.. Scheur,T.andLevy.S.. 1983.Rec0veryofacc~urtic impedance from reflection seismograms: Geophysics. v. 48. P. 1318.1337.

Rockwell. D.W., 1971. Migration stack aids interpretation: Oil and Gas hurna,, April 19.

Rvu. J.V.. ,980. Decomposition (DECOM) of seismic gathers into klocity components by a space lime filter: Presented at the 50th Annual Meeting of the SEG. Houston. Texas.

Sattlegar. J.W. and Stiller. P.K.. 1973. Section migration before stack. after stack or in between: Geophysical Prospecting, Y. 22. p. 297.314.

Stult, R.H., ,978. Migration by Fourier transform: Geuphysics, v. 43. p. 23~48.

Taner. M.T. and Koehler. F., 1981. Surface consistent corrections: Geophysics. v. 46, p. 17-22.

Walker. C. and ~lrych. T.J.. 1983, Autoregressive recovery of the RCOUSI~C impedance: Geophysics, Y. 48, p. 1338-1350.

Widess.M.B..Howlhinisathinbed?~eophysicr.v.38.p. 1176-1180.

Wren, A.E. andJain. S.. ,978.Theapplicationofinversion~aselsmic inlerpretatiunproblemsinthe Fuuthi,,s:CanadianSocietyofExplo- ration Geophysicists Journal. Y. 14, p. 66-7,.

some techniques of enhancing seismic reflection...

Documents