You are on page 1of 6

Downloaded 11/04/16 to 129.11.253.115. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.

org/

Whatever
happmed
ground roll?

to

By NIGEL A. ANSTEY
Geophysical Consultant
Boston, Massachusetts
(Editors Note: In August1985,the SEC;Workshopon Seismic
Field Techniques
washeld in Monterey,CalQornia,underthe
chairmanshipof Dr. Elmer Eisner.This talk wasgivenas the
keynoteaddress;
it is reproduced
substantiallyverbatim.Part 2,
Fieldtechniques
for highresolution,
follows in the next issue
of THE LEADING EDGE.)

adiesand gentlemen,first let me acknowledge,quite


cheerfully,that I haveabsolutelyno business
kickingoff
a workshop
on fieldtechniques.
For althoughI havealways
been drawn to field techniques,almost all my present
everydaywork is asan interpreter.So, if whenI talkabout
field techniques,I soundout of touch,out of date,out of
the picture. . . well, I am. I even live in Boston.
Needingas I do an updateon recentadvancesin field
techniques,what am I hopingto hear?
I realize,of course,thatthe expertswill be announcing
solutionsto problemsthat I didnt evenknow we had and I promiseto be suitablyimpressedby all that. But
thereare a few problemswhich I do know we have,and
I would be overjoyedto hear that they are solvedalso.
Youknow,Im a simplechap,andI lovesimplethings.
Thingsthatareclever,butsimple.Like thewheel,andthe
weedwhacker,and hanggliders.For that reason,I like
electricmotors,but I abominatethe internalcombustion
engine.Too damn complicated.When I standand look
at a Vibroseisvibrator, I get the same feeling - too
damncomplicated.So I am hopingto hear someoneannouncea newtype of vibrator,simpleenoughthatsimple
chapscan understandit, and simple enoughthat simple chapscan keep it working.
For thatmatter,I get the samefeelingwith marineairgun arrays. All those problemsthat come with using
different-sizeguns - incomprehensible
directionalresponses,heaven-knows-what
spectrumfor the sourcegeneratednoise . . Wouldsomeonepleaseannouncethe
deathof the air-gunarray,soI dont haveto makethe effort any more?
what elseam I hopingto hear?That someonehasfinally
clearedup all the confusionandcontradictions
aboutwhat
happensin the near field of the Vibroseis vibrator (or,
rather,in the near field of an array of three or four vibrators).You rememberthe problem:what is the shape
of a Vibroseisreflection?In the 1960swe usedto tell interpretersthat the reflectionpulseon a Vibroseisrecord
was basically
a zero-phase
pulse,modifiedonlybytheearth
andthegeophone.This wassobecausewe lockedthepar-

40 GEOPHYSICS:

THE LEADING

EDGE OF EXPLORATION

MARCH

1986

title velocity of the baseplateto the control signal,and


becauseour geophonesalso measuredparticlevelocity,
in the far field. Then we hesitated,rememberingthat(in
a fluid, at least)theparticlevelocitychanges
phaseby 90
degreesasthe pulsepropagates
throughthenearfield. So
perhapsour reflectionswere actually skew-symmetric
pulses,90 degreepulses.What to do?First we got out
all the old classicalpapersagain.Then, lookingat all that
mathematics,we decidedit mightbe betterjust to put a
geophonedowna hole, andmeasureit. But researcherA
disagreedwith researcherB, andB with C, andthe Russiansdisagreedwith everybody,and- unlessIve missed
somerecentannouncement
- we still dont know what
shapeour Vibroseisreflectionsare. And thats awful; we
really shouldknow.So I am delightedto seethat, according to the program,all thisconfusionwill be completely
clearedup. This very afternoon.
AnythingelseIm hopingto hear?Well, youknowwhat
the hi-fi salesmensay:the stronglink in the hi-fi chain
is theamplifier,andtheweaklink is thespeakers.
Sospend
your moneyon the speakers.In seismicinstrumentation,
I supposethe stronglink is againthe amplifier, and the
weak link is the geophone.So if I havea concernwith
our instruments,its in the geophone.Oh, of courseits
muchbetter than it usedto be, but doesthat meanthat
its harmonicdistortionis so smallthat we can forgetit?
Underreal conditionsof plant,andreal conditionsof tilt?
I hopethatsomeonewill setmy mindat rest, on . . . lets
see .
tomorrow,or perhapsFriday?Good.

So

thoseare someof the solutionswhichI hopeto take


awaywith me from thisworkshop- in addition,asI said,
to the solutions
of all theotherproblemsthatI didnt know
I had.
Buttheprogramalsotalksof giveandtake.The operative
word, Elmer tells us, is sharing. I can take awaysomething, but I mustalso bring something.
So, I thought,what can I bring - what do I haveto
saythatsnew?Well,I do havesomething
to say,but,oddly
enough,I haveno idea whetherits new or not. In part,
of course,this is becauseI know that if I live in Boston
I mustbe out of touch.In theotherpart, it is becausewhat
I haveto sayis too simpleandtoo obviousto be new.
Let me tell you aboutit.
In additionto beinga workinginterpreter,I do some
teaching.Specifically,I freelancefor IHRDC, in guiding
the geophysicsportionof their Video Library. And you
know that there is nothinglike teachingto make you
organize whatyou know,andto makeyou realize(oh my

Downloaded 11/04/16 to 129.11.253.115. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/

goodness!)what you dont know. time and again it happens. You understandthe basicconcepts,you apply the
conclusionsevery day, but when you try to explain it to
somebodyelse- somethingsticks,somethingwont come
out right. So I havelearned much from havingto teach.
Often, I find myself going back, filling in gaps in my
knowledgewhich I shouldhavefilled (and meant to fill)
many years ago.
One suchthing I alwaysmeant to do was to work out
whatactuallyhappensto the groundroll in the stack.Well,
of course,it getsattenuated.But is this a purely random,
statistical,JN- type of attenuation. . . or is there some
systematicorder to it?
The answer- which is my piece of sharingtoday - is
that there is indeed order to it.
You know what happenswheneveryou make what you
think is an announcement.Somebodyjumps up and says,
I thoughtof thatyearsago.No you didnt. YesI did.
No you didnt. Yes I did. Well, there will be none
of that here. What I haveto sayis so obviousthat I cannot
expectthat I am the first personto notice it. Indeed, my
immediatereaction, on noticing it, was one of shamethat it shouldhavetakenthisold ploddersolong to realize
what everybodyelsemusthaveknownsincethe discovery
of fire. Yet, I dont knowof anypublicationof it. Of course,
that couldbe my fault again- for missingit in the literature. But then, I think to myself, how little of what we
actually do is in the literature! And conversely. . . oh
researchers!
howlittle of whatis in the literaturewe actually
do! Then again,this view of how ground roll getsattenuated in the stack leads to a set of criteria for multiple
coveragein the field, and many modern sectionswhich
passacrossmy deskviolate thosecriteria. So maybethe
criteria are not widely known. Anyway, here goes.
Whether its ancienthistory,or hot off the press,here it is.

The receivedwisdom, astaughtin textbooksand papers


beingpublishedtoday,is thatground roll - or, in a wider
sense,source-generated
noise- shouldbe attenuatedby
arrays. . . by geophoneand sourcearrays.We are taught
to makenoisetestsin the field, to establishthe troublesome
rangeof wavelengthsin the ground roll, andto designarraysfor which this rangeof wavelengthsis down among
the sidelobes.Thus traditionalarray designis basedon,
and directed at, the source-generatednoise.
Whereverthe source-generated
noiseincludeslongwavelengths(in particular, in ground-roll country) this traditional approachleadsto long arrays.However, we do not
like long arrays.They attenuatethe first breaks,lose the
high frequenciesat early times and long offsets, lose
thehighfrequencies
in roughterrain,and- mostseriouslose the high frequenciesfrom dipping reflectors. Long
arraysare weighty and tanglesomein the field. And they
are conceptuallyout of tune with modernpractice,in that
they makea final decision- an irrevocableexclusivedecision - in the field.
Where the longestnoisewavelengths
are associated
with
the lowestfrequencies,we canshortenthe necessaryarrays
by excludingtheselow frequencieswith a low-cut filter.
Then, with shorterarrays,we are more likely to preserve
our high frequencies- on shallow events, on dipping
events,and on the first breaks. In other words, we can
maintainthe high frequenciesif we sacrificethe low fre-

quencies.But in the modern world we do not wish to do


either;the low frequenciesare critically importantto inversion, and to the recognitionof transitionalreflectors,and
to polarity estimation.Therefore we seekto be rid of this
needto compromise.We seeka method of obtainingextremely long arraysfor the suppression
of eventhe lowest
ground-roll frequencies,while presentingshort arraysto
the first breaks, and the far groups, and the steepdips.
Further, of course, we seek arrays which are easy to lay.
All of these objectivesare satisfiedby what I call the
stack-arrayapproachto array design.The stack-arrayapproachjunks the traditionalapproachto array design. It
has no concernwith the wavelengthsof the ground roll,
and thereforeit needsno noisetests.It says:To hell with
longarrays.To hell with tangledstrings.To hell with array
responses,
to hell with complexity,to hell with Chebyshev.
(Most of all, to hell with Chebyshev.)It says:just select
a sensiblegroup interval on the usual considerationsof
havingenoughtracesto seethe geologicfeaturewere looking for, and of avoidingspatialaliasing- and then just
set the effective array length equal to the group interval.
The group length equal to the group interval. Always.
The effect of this is to provide a uniform successionof
geophonesalong the spread- equally spaced,equally
weighted, even, regular, continuous.This is our fundamentalstarting-point.All right, nothingunfamiliarin that.
Now we considera field record, with first breaks,and
reflections, and (in this context) ground roll (Figure 1).
Normally we think of the ground roll as a wave train on
a trace- a time function at one instantof space(Figure
2a). Equally well, we could think of it as a wave train in
space- a snapshotof the surfaceat an instantof time
(Figure 2b). Then from that instantof time to somelater
instantof time the ground roll moves in the fashion of
Figure 3.
Now lets do a terrible thing - lets just add all the
tracesof the field record. At each instantof time we are
adding all the geophonesalong a line thefull length of
the spread. In effect, we are forming one long array,
perhapsa thousandor more equally-spacedgeophonesin
one array whose length is the length of the spread. The

Figure I. A rudimentary field record, with one trace shown


in full. Normally we think of the wavesas traces, that is, as
functions of time

GEOPHYSICS: THE LEADING

EDGE OF EXPLORATION

MARCH

1986 41

Downloaded 11/04/16 to 129.11.253.115. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/

array is many times longer than the longestwavelength


in the ground roll, and so the ground roll is zapped.
Of course,the reflectionsare zappedalso, so that wont
do. We have the problem of dipping reflections, and we
havethe problemof normal moveout.To removethe problem of dip first, we turn (as usual) from the field record
to the gather.

ets start with the gather derived from an end-on


spread - group length equal to the group interval and lets set the sourceinterval to huff the group interval. Then, of course,from the stackingdiagram- a fold
of stackequal to the numberof channels(Figure 4a). Six
channels,six fold - six traces on the field record, six
traceson the gather.Indeed if the field record looks like
Figure 1, the gather looks the same. Now lets draw the
arrays, the groups,on the stackingdiagram(Figure 4b).
We see that if the groupsare even, continuous,end-toend on the field record, then they are even, continuous,
end-to-endon the gather (Figure 4~). So adding all the
tracesof the gather(thatis, stacking)yieldsonecontinuous
uniformarray,havingthe samelengthasthe spread(Figure
4d). Again, the ground roll is zapped.
Then the problem of normal moveout.Well, when you
think about it, it really isnt much of a problem. First,
becauseat the offsetsand timeswhere the groundroll arrives, the NM0 is small anyway. And second, because

Figure 3. Ground-roll wavetrain asa functionof spaceand

time
the physicalarray acrossthe gather is still uniform and
continuous
- its just that, whenwe sumalongthe moveout
hyperbola, the sample of ground roll put into the sum
jumpsforwarda few milliseconds,in time betweengroups.
At the frequenciesof the groundroll, and for the smooth
natureof the NM0 curve,the groundroll remainssubstantially zapped.
Now, all that wasfor an end-on spreadwith sourceinterval equal to half the group interval. Lets consider a
moreusualcasefor landwork - a splitspread,againwith
the grouplengthequalto the groupinterval,but now with
the sourceintervalalso equalto the groupinterval. Then,
in thissituation,it is desirableto shootbetweenthe groups
- not on the flags. Now the fold is only half the number
of channels(Figure 5a). The groups- althoughstill endto-end and continuousin the field (Figure 5b) - are no
longerend-to-endandcontinuousacrossthe gather(Figure
5~). Indeed they are regularly on-off-on-off acrossthe
gather.But then if we look at the other sideof the gather,
we have the sameeffect, but now off-on-off-on. We can
see the effect of stackingthe gather if we flip the lower
side to the upper side (Figure 5d); the offsetsinterleave
to give, onceagain,a continuousevenarray stretchingthe
whole length of one side of the split spread.Again, the
ground roll is rapped.
So both of the spreadgeometriesweve just considered
- the end-on spreadwith sourceinterval half the group
interval,and the split spreadshotbetweenthe groupswith
sourceinterval equalto the group interval - both satisfy
the basicstack-arraycriterion: the multiplecoveragemust
be suchthat there is an even, continuous,uniform successionof geophonesacrossthe gather. Not acrossthe
spread- acrossthe gather. When this is so, the operation of stackingforms a long uniform array the lengthof
the spread, and the ground roll is zapped.
Lets see what happensif we violate this criterion.

Figure 2. Ground-roll wave train as a function of time at


an instant of space (a), and as a function of space, at an

instantof time (b).

42

GEOPHYSICS: THE LEADING

EDGE OF EXPLORATION

MARCH 1986

Suppose, for example,that we halve the fold. In the endon case,insteadof usinga sourceintervalof half the group
interval, we use a sourceinterval equal to the group interval (Figure6a). (You seelotsof crewsdoingthat.) Then
the stack-array is on-off-on-off all across the gather,
and there is no suppression- no suppressionat all of ground-roll wavelengthsequal to two group intervals

Downloaded 11/04/16 to 129.11.253.115. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/

Figure4. (a) Stackingdiagram for an end-onspread.(h) Endto-end groups shown in the common-sourcedirection. (c)
End-to-endgroupsshownin the common-midpointdirection.
(d) The addition of the gatheredtracesforms the stackarray.

Figure 5. Counterparts of Figure 4 for a split spread shot


betweenthe groups. The groups interleave to form a continuous stack array the length of one side of the spread.

GEOPHYSICS:

THE LEADING

EDGE OF EXPLORATION

MARCH

1986 43

Downloaded 11/04/16 to 129.11.253.115. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/
44

(Figure 6b). Clearly, this is a bad thing to do.


If we draw the stackingdiagrams,we can quickly see
that severalothermuch-practiced
techniquesalsoviolate
the stack-array
criterion,in particular,end-onspreads
with
a sourceintervalequalto or greaterthanthe groupinterval, splitspreads
witha sourceintervalof doublethegroup
interval,split spreadsof any type shoton the flags, and
- checkit for yourself- anydynamiteshootingin which
the arraysare not equalin lengthto a groupinterval(or
an integralmultiple of the group interval).
In fact, for thefield-manspreferenceof a grouplength
equalto one groupinterval,thereare only two multiplecoveragetechniqueswhich satisfythe basicstack-array
criterion. This allowsus to restatethe criterion as two
criteria: for end-onspreads,the sourceintervalmustbe
half the groupinterval;for split spreads,the sourceinterval mustbe equalto the group interval,and shotbetween the flags. Those are the two techniques.
(Incidentally,if we are preparedto rni~ traces,either
in the processingor on the ground,thenthere are other
arrangements
whichsatisfythebasicstack-array
criterion.
However,in all casestheseother arrangements
involve
somesacrifice- eitherof steepdipsor of resolution.
After
this sacrifice,the resultsare probablyno betterthan we
wouldhaveobtainedwith one of thesetwo preferredarrangements
andfewerchannels;weve paidgoodmoney

for nothing.)
So,if lotsof crewsareoutthererightnowviolatingthese
criteria,whathappens?
Well. perhapstheresnottoomuch
source-generated
noise,and the resultsare fine anyway.
Perhapsthe noisewavelengths
are short,andshortgroups
are sufficientto attenuate
it - fine again.Butmaybesome
wavelengths
of the noise come right throughthe stack
unattenuated.
and give the appearanceof low-frequency
multipleson thefinal section. . weve all seenthat.Then
theprocessors
do sometestsandfind theresultsare much
better if they apply f-k filtering on the shotrecords,or
if they apply a low-cutfilter. Nobody recognizes
those
spuriouseventsas an error in the designof the multiple
coverage.
Or the processors
seesomeherringboneeffects
on the section(weve all seenthat too). which they find
can be suppressed
by f-k filtering after stack. Again,
nobodyrecognizes
it asan error in thedesignof themultiple coverage.
OK - a few otherpointsaboutthestack-array
approach.

Figure 6. End-on spread with source interval equal to the


group interval (a) yields no suppression of ground-roll
wavelengths equal to two group intervals (b).

Figure 7. On a field record (a), the traces display the


physically-propagating wave train across the spread. On a
gather (b), differences in the ground roll from shot to shot
disturb the picture of the wave train.

GEOPHYSICS:

THE LEADING

EDGE OF EXPLORATION

MARCH

F.wt. it reallyisquitedistinctanddifferentfromthetraditionalapproach
to arraydesign.In thetraditionalapproach,
thelengthof thearrays(thegroups)is basedon thelongest
apparentwavelength
of thesource-generated
noise.In the
stack-arrayapproach,the lengthof the groupsis just the
groupinterval,whichis half theshortestapparentwave-

1986

Downloaded 11/04/16 to 129.11.253.115. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/

length of the Signal. Thusthe stack-array


approachreally
doesrequireno ground-rolltests,no noisespreads,nolost
time and no compromises
betweenlow frequenciesand
high frequencies.
Second,it is true that the stackarray doesnecessitate
a highfold of stack.SO if thisis unthinkable(for example,
in portabledynamitework in thejungle), we abandonthe
stack-arrayapproach,and go back to traditionalarrays.
But the stack-arrayapproachis ideal for work with a
mobile source,andin particular,of course, it is ideal for
Vibroseis.
Third, with Vibroseiswe are likely to be usingboth
sourceandgeophonearrays.Interestingly,the stack-array
criterion(if otherwisesatisfied)remainssatisfiedif either
the source array or fhe geophone array is equal to the
groupinterval.So, if yourbossis a sticklerfor traditional
arrays- no problem,just takethe convenience
of a geophonegrouplengthequalto the groupinterval,andthen
spaceout the vibratorsto give him his traditionalsource
array. And that sourcearray can also be an area1array,
if there is noisescatteredback from the side.
Fourth,in the stack-arrayapproach,the suppression
of
the groundroll doesnot occuruntil the stack.Doesthis
meanthatwe takerecordsjust swamped
with groundroll,
and somehowmusterthe courageto say, Dont worry,
itll all be systematically
killed in the stack?No. it does
not. With the stack-arrayapproach,we are totally free to
raisethe naturalfrequencyof the geophones,and to use
a ground-rollfilter in the field, and thusto removethe
groundroll from the field records.But doesntthatnegate
one of the advantages
of the stack-arrayapproach- that
it doesnot requireany sacrificeof the low frequencies?
No again- becausethe responseof the geophone,and
of the low-cutfilter, doesnot changethe ratio - the ratio
- of the low-frequency
signalto the low-frequencynoise.
We canapplythe low-cutfilteringin thefield, to getgood
recordingand a usablemonitorrecord . . then zap the
groundroll in the stack.andfinally usea determinedinversefilter to recoverthe low frequencies.Thus we can
userugged14 Hz geophones,and 12 Hz or even 18 Hz
ground-rollfilters, and still maintaina final processed
bandwidthdown to 5 Hz.
Next, the stackarrayis aslongasthe spread,andmany
timeslongerthanthelongestwavelength
in thegroundroll.
It containsmany,many traces,but it doeshavean end.
What happenswhenthe groundroll getsto the endof the
gather? Well, we get a few minor side-lobes.If we feel
stronglyaboutthese,we can give a little lessweightto
the last trace or two - taper the stackat the ends.The
same.too, if we havea gap in the gatheron accountof
rivers and houses- just taper the edges.
Finally, in the contextof the stackarray, we haveto
recognizethat there is a differencebetweena very long
arrayformedacrossthegatheranda verylongarrayformed
acrossthe spread.Thuson a field recordthetracesdisplay
the actualpassageof a physicallypropagating
wavetrain
acrossthe spread(Figure7a), andwe canbe totallyconfidentthattheadditionof thesetraceswouldkill theground
roll dead. But on a gather,eachtrace comesfrom a different source, in a different place; perhapsthe amplitude
of the groundroll is different, or its velocity,or its frequency.In the worst case, the samplesof ground roll
recordedby the tracesat any one time could be quite

randomFigure 7b). Thenwhatwouldthe stackarray give?


It wouldno longerkill the groundroll dead,it would just
give a statisticalJN-type attenuation.Here, then, we see
the range of suppression
providedby the stack array.
If the groundroll is substantiallyconstant from shotto
shot, the ground roll is zapped. If the ground roll
is extremely erratic from shot to shot, the ground
roll is reducedonly by thesquarerootof the fold of stack.
With in-betweensin between.

Youknow,whenI cameintothisgame,themajorenemy
of the geophysicist- the devil himself - was called
groundroll. Everywherethe questionwasthe same:what
canwe do aboutgroundroll?Then, with surfacesources,
the groundroll actuallygot worse!But nowadays
- whoevertalksaboutgroundroll?Abouthalfthecrewsoutthere
take no accountof groundroll in their arrays;geophone
frequencies
go lowerandlower,andground-rollfiltersare
very unfashionable.
So whateverhappened
to groundroll?
In the heydayof fancygeophonearrays- Chebyshev
andall that - a wise old man whisperedto me that the
bestthat arrayscould do (howeverfancy)wasto reduce
the groundroll by 15 dB. Say to 20 percent.We agreed
just now that the worstthe stackarraycoulddo - if the
groundroll wasentirelyrandomfrom shotto shot,or indeedif the stack-arraycriterionwasviolatedin any way
- wasto reducethe groundroll by the squareroot of the
fold of stack.At 25fold, therefore,the worst-casestack
array is aboutas goodas traditionalarrays.And that, I
think, is whateverhappenedto groundroll. When we got
to 24-fold, we found- probablyquiteempirically- that
we didnt needlongarraysanymore.But whatwe did not
notice(or at leastwhatI did notnotice)wasthatby walking the extra inch - designingthe multiplecoverageto
satisfythestack-array
criterion- we couldhavedoneeven
better.
Sothereis the stack-arrayapproachto arraydesign.As
I saidearlier, I doubtvery muchwhetherI am the first
to formalizethe stack-arraycriterion. So if you got there
first. 15yearsago, be my guest- feel free to claim it.
But then, be preparedto be outdoneby the guy who got
there 25 yearsago.And, of course,theresalwaysHarry
Mayne,whosethinkingmayhavegotthisfar 35 yearsago.
As always,Harry is hard to beat.

Justone other point aboutthe stack-arrayapproach-

if yourcompanyis a sponsorof IHRDCs Video Library,


you alreadyhavethe detailsof all this in-house,in the
manualfor moduleGP305, on array design.IE

NigelA. Anstey hasparticipated in all facets of the Seismic exploration method. Following his graduationfrom the University
of Bristolin England in 1948, he worked for SeismographService Ltd. until 1968 when hejoined SeiscomDelta, Inc. in 1975
he became an independent consultant. Anstey iS recipient Of
many awards from the SEG, EAEG and AAPG (see TM
January 1986, p. 26). In addition to his mktier as a Seismic
interpreter, Anstey is a geophysical instructor and writer of
international renown. He is also the guiding force behind the
geophysicsportion of the Video Library of the International
Human ResourcesDevelopment Corp. to which he refersin his
article.

GEOPHYSICS:

THE LEADING

EDGE OF EXPLORATION

MARCH

1986

45

You might also like