-
Identification of lithology in the Gulfof Mexico
FRED HILTERMAN and JOHN W. C. SHERWOOD, Geophysical Development
Corporation, Houston, TexasROBERT SCHELLHORN, Chieftain
International, Dallas, TexasBRAD BANKHEAD, ORYX, Dallas, TexasBRIAN
DEVAULT, Colorado School of Mines, Golden, Colorado
n a small town outside of Houston,a local rancher was overh e a
rd saying,Do you have any 3-D seismic acro s syour place? You ought
to get some,because it tells you exactly whatsdown there and where
to drill.
Yes, the transfer of technologyhas been accelerated by new
elec-t ronic media such as the Internet, butis it possible that it
bypassed the geo-physicists?
Maybe the rancher was looking atbright spots plotted in red.
Suchresults are possible but not guaran-teed. Throughout the
Tertiary basinsin the Gulf of Mexico (GOM) area reas where acoustic
impedance val-ues of shales and gas sands area p p roximately
equal. This meansh y d rocarbon zones do not appear asbright spots
and are difficult to detectwith conventional 3-D seismic data.F u r
t h e r m o re, in some areas, geo-physicists have had no success
usingAVO for predicting exactly where todrill. This normally occurs
when therock properties are not calibrated tothe various AVO
attributes.
To resolve this dilemma, a 3-DAVO study was conducted utilizingn
u m e rous well-log suites, core analy-ses, and field production
histories.With the inclusion of anisotro p i ce ffects, a robust
AVO analysis basedon a lithologic model was possible.Results from
this study are illustrat-ed in Figure 1: A conventional 3-Dsection
with the AVO analysis over-plotted in red and yellow. Corre l a t i
o nto the well-log curves and the fieldp roduction histories
indicates that allred and yellow events are associatedwith proven
hydrocarbon zones.Could this be the 3-D seismic that therancher was
talking about?
It is obvious that the re f l e c t i o namplitudes on the
conventional 3-Dseismic do not identify lithology ifthe red and
yellow events are trulyh y d rocarbon events. However, as
theremainder of this article will show, insome environments, the
petrophysi-cal AVO model can be constrained sothat reflections from
very clean wet
EDGENET HTTP://WWW.EDGE-ONLINE.ORG FEBRUARY 1997
I
Figure 1. Conventional migrated section with AVO anomalies
super-imposed as red and yellow events that represent known
hydrocarbon reservoirs.
Figure 2. Suite ofsonic logs across Ter-tiary basin (upper)and
synthetic strati-graphic section gen-erated by the normal-incidence
reflectioncoefficient equation(lower). Chronostrati-graphic
surfaces arenumbered. (AfterVail, 1977).
Figure 3. Poissonsratio curves for shaleand sand as a functionof
P-wave velocity. IfVp,sand > .84Vp,shale , thensand
< shale , a conditionthat occurs in most ofGOM.
February 1998 THE LEADING EDGE
-
sands and gas sands overshadow allother reflections. These
dominantlithostratigraphic reflections arerelated to Poissons
ratio.
Chronostratigraphic and lithostrati-graphic reflections. Vail
and col-leagues at Exxon presented the basicprinciples underlying
seismic strati-graphy 20 years ago in AAPG Memoir2 6 . F i g u re
2, one of the most re m e m-b e red illustrations from that
work,displays several sonic logs across aTertiary basin in South
America alongwith three chronostratigraphic orequal-time surfaces
numbered 15, 10,and 8. Of importance is that chro n o
s-tratigraphic surface 8 cuts rightt h rough a major sand deposit
whichoverlies an unconformity. The lowerportion of Figure 2 shows a
series ofnormal-incidence synthetic seismo-grams generated from the
sonics.What astounded geophysicists whenthis synthetic was first
presented wasthat the reflection events follow thec h
ronostratigraphic surfaces and notthe upper and lower surfaces of
thesand package which would be thelithostratigraphic surfaces. From
thisexample and the many that have beenp roduced since 1977, it was
conclud-ed that the normal-incidence sectionand conventional
seismic data basi-cally consisted of chro n o s t r a t i g r a p
h-ic reflections. It should be noted thatthis conclusion results
from empiricalobservations and not a rigid theore t-ical model.
Shortly afterward (1982), anotherExxon geoscientist (Paul Tu c k
e r )warned (in Pitfalls Revisited) thatstacking enhances
continuity andparallelism of the reflection, ... butover-stacking
can destroy the geolo-gy. ... Stacking can also distort
thestratigraphy ... thus playing havocwith successful stratigraphic
map-ping. These astute observationsw e re made before the advent of
AV Oand before the information contentof the reflection stack was
fullyunderstood. Is it possible that, becauseof the long offsets
employed in todaysseismic, that a mixing of two petrophys -ical
properties acoustic impedanceand Poissons ratio leads to strati
-graphic distortion? Are chro n o s t r a t i -graphic and
lithostratigraphic re f l e c t i o n sbeing mixed together?
Petrophysical model. The relation-ship of chronostratigraphic
andlithostratigraphic events to thereflection amplitude is
difficult toenvision if the exact Zoeppritz equa-tion is
examined.
H o w e v e r, a decade ago Shuey(GEOPHYSICS, 1985) presented a
lin-ear approximation of the reflection-coefficient equation that
was recent-ly modified by Verm and Hilterman(TLE, August, 1995)
to:
RC() NI cos2() + PR sin2()
where
NI = Normal-incidence reflectivity =
(2V2- 1V1)/(2V2+ 1V1), PR = Pois-son reflectivity = (2- 1 )/(l-
avg)2 ,and , V and are respectively thed e n s i t y, P-wave
velocity and Pois-
EDGENET HTTP://WWW.EDGE-ONLINE.ORG FEBRUARY 1997
Figure 4. Suite of well-log curves in study area. While the
Poissons ratiocurve correlates with the SP curve, the
acoustic-impedance and sand-per-centage curves do not
correlate.
Figure 5. Crossplot of ln [acoustic impedance] versus Poissons
ratio.Quick-look prediction tool: vertical distance between two
lithologic pointsrepresents NI while horizontal distance represents
PR.
0.14 Poissons Ratio 0.44
February 1998 THE LEADING EDGE
-
sons ratio for the lower medium (2)and the upper medium (1), and
avgis (2+1)/2.
This equation provides usefulinsight into the AVO response, and
itis often used as the model for esti-mating NI and PR from seismic
CDPgathers. The main benefit is that theresulting PR can be thought
of as asignal that reflects from the earthsPoissons ratio profile.
Since Vail re l a t-ed NI to chro n o s t r a t i g r a p h y, we
ask,Can it be that PR is related to lithos-tratigraphy; i.e., sand
versus shale?Verm and Hilterman, in fact, notedhow the Poissons
ratio curve in asand-shale sequence closely re s e m-bles the SP
curve, a primary lithos-tratigraphic tool of the log analyst.The
correlation between SPand Pois-sons ratio curves can be explained
byexamining Castagnas empirical Vp-to-Vs relationships (Figure
3).
An interesting observation fromFigure 3 is that sand appears to
havea Poissons ratio that is consistentlysmaller than shales
Poissons ratio.This is similar to an SP curve wherethe sand value
falls beneath the shalebase line. From the diagram, for aPoissons
ratio curve to resemble anS P curve, the P-wave velocity ofsand and
shale should be approxi-mately the same. If shale and sandhave the
P-wave velocities as depict-ed by points A and B in Figure 3,then
the Poissons ratio for sand isless than shale. This is similar to
anS P relationship. However, as thesand P-wave velocity decreases
fro mpoint B to C, the Poissons ratio dif-ference between sand and
shale dis-appears. Thus, for the correlation ofthe SP and Poissons
ratio to be sim-ilar, the sand P-wave velocity mustbe greater than
.84 of shales P-wavev e l o c i t y. This condition exists formost
of the GOM. In short, the Pois-sons ratio curve indicates
lithologysimilar to an SP curve, and thusreflection amplitude
associated withthe Poisson reflectivity will be
lithos-tratigraphic. Also, as the Ve r m -Hilterman equation
indicates, theconventional seismic stack doesindeed mix the
lithostratigraphic PR,with the chronostratigraphic NIespecially in
a Class 2 environmentwhere NI is small.
The PR contribution to the re f l e c -tion amplitude is not
significant untillarger incidence angles are reached.When the sourc
e - receiver offset isthe same as the depth of investiga-tion, the
incidence angle is approxi-mately 30. At this angle, 75% of the
EDGENET HTTP://WWW.EDGE-ONLINE.ORG FEBRUARY 1997
Figure 6. Migrated CDP gather and isotropic AVO synthetic from
well atsame location. Both are NMO-corrected, based on isotropic
media. Offsetsrange from 1000 to 20 000 ft.
Figure 7. Migrated CDP gathers, NMO-corrected with isotropic
equation(upper) and anisotropic equation (lower). Offsets range
from 1000 to 20 000 ft.
February 1998 THE LEADING EDGE
-
reflection amplitude (according tothe Verm-Hilterman equation)
isf rom chronostratigraphy (NI) and25% from lithostratigraphy
(PR).H o w e v e r, at 60, the situation isopposite because 25% is
NI and 75%is PR. Thus, if lithology estimation isthe goal, it is
desirable to re c o rd seis-mic data at angles approaching 60.
If reflections at angles of 60 aredesired and assuming that the
criti-cal angle has not been reached, thep e t rophysical model
described bythe Verm-Hilterman equation mustbe reviewed. A h i g h
e r- o rder termwhich can be approximated by 1/2 [(V2 - V1)/(V2 +
V1)] (tan2- sin2)was dropped from the right side oftheir
approximation. This term cantbe ignored unless the change in P-wave
velocity across the interface issmall. Thus, the direct evaluation
oflithology from the large offset reflec-tion amplitude is modeled
for smallvelocity variations which occurfor Class 2 AVO
anomalies.
Well-log data. The study are ainvolved Tertiary rocks that are
com-monly found in the transition zone ofo ff s h o re Texas and
Louisiana. Figure4 illustrates a typical suite of well-logcurves
from the area. The SP c u r v eresembles the Poissons ratio
curveshown next to it. Pay zones aro u n d8000 ft are indicated by
red re c t a n g l e s .The acoustic impedance curve that
isoverplotted on the sand-perc e n t a g ecurve has little
character re s e m b l a n c eto it, suggesting that acoustic
imped-ance is not a good indicator of lithol-ogy in this enviro n m
e n t .
In order to calibrate this well tothe AVO response, a crossplot
of thenatural log of acoustic impedanceversus Poissons ratio for
the depthinterval 6100-9000 ft was generated(Figure 5).
With the normal-incidence re f l e c-tion coefficient expressed
as NI = . 5 [ 1 n (V )2 - 1n(V )1], the vertical dis-tance on the
graph linearly relates toNI. The Poissons ratio axis can thenbe
scaled so that a relative re l a t i o n-ship of NI to PR can
easily beobtained. For instance, the verticaldistance (NI) between
the center ofthe shale points and the center of thegas-sand points
is small compared tothe horizontal distance (PR) betweenthese
points. This distance re l a t i o n-ship suggests that the NI for
ashale/gas sand interface will be smallc o m p a red to its
corresponding PR.Also, the vertical distance (NI) fro mthe shale
center point to the gas-sand
EDGENET HTTP://WWW.EDGE-ONLINE.ORG FEBRUARY 1997
Figure 8. Anisotropic ray-theory AVO synthetic and migrated CDP
gathersat a well site with both anisotropic and isotropic NMO
corrections.
Figure 9. Conventional migrated section (upper), estimated PR
section byinversion (middle), and estimated PR section by
large-angle stack (lower).
February 1998 THE LEADING EDGE
-
center point is about the same as thevertical distance (NI) from
the shaleto the wet sand. This means brightspots will not be
evident. However,the horizontal distance (PR) for ashale/gas sand
interface is appro x i-mately four times larger than the
hor-izontal distance (PR) for a shale/wetsand interface. This
indicates that PR,and not NI, will diff e rentiate litholo-g y. The
ln[V] versus c rossplot pro-vides a quick-look method of
cali-brating NI and PR to variouslithologic interfaces.
Anisotropic eff e c t . In an effort to sta-bilize the
extraction of PR from aCDP gather, offset distances whichwere
greater than depth (incidenceangles greater than 30) weredesired.
In Figure 6, a 3-D migratedC D P gather at a well location isshown
beside the wells AVO syn-thetic response. The model wasi s o t
ropic and NMO-corrected withthe wells RMS velocity. At 2.25 s (
8000 ft), the models 14 000-ft off-set trace exhibits an NMO
overcor-rection of 45 ms (caused by thei s o t ropic ray bending).
The field datahad an overc o r rection at the sameo ffset of 125
ms. This additionalovercorrection (125 ms - 45 ms) wasidentified as
an anisotropic effect. Inessence, the horizontal velocity isfaster
than the vertical velocity.
It is believed that the media aredominantly transversely isotro
p i c(TI). 3-D fields of both velocity andanisotropy are obtained
from NMOanalyses and can then be applied inthe NMO, DMO and
migration pro-cessing steps.
F i g u re 7 presents three CDPgathers with conventional (isotro
p i c )NMO and anisotropic NMO correc-tions. An interesting effect
in the fartraces is the reduction of NMOstretch when anisotropy is
includedin the NMO correction. Often, theNMO overc o r rection on
the fartraces is not observed because theCDP gathers are muted at
the linewhich is equivalent to offset = depth.H o w e v e r, as the
Ve r m - H i l t e r m a nequation and the crossplot indicate,the
reflections on the right side of themute line contain information
aboutlithology that needs to be preserved.This lithologic content
can be veri-fied by AVO modeling.
Well-log curves along with thea n i s o t ropic properties
measured fro mthe 3-D seismic were used to generatethe AVO
anisotropic synthetic shownin Figure 8. To the right of the
syn-thetic is the anisotro p i c - p ro c e s s e d
C D P gather at the well location. Tw ogas zones are evident on
the SP a n dresistivity logs. The thickness of theupper gas sand
(at 2.25 s) is 40 ft whilethat of the lower sand (2.39 s) is 35
ft.O ffset traces up to 16 000 ft wereusable in the CDP g a t h e
r. Because thea n i s o t ropy in this area causes thew a v e f
ront to flatten from its isotro p-ic spherical shape, the incidence
angleat 2.39 s on the maximum offset traceis 50 for anisotropic
modeling, whilei s o t ropic modeling predicts 60. Thismeans
critical angle reflections occuron traces with larger offsets than
con-ventionally assumed.
A n i s o t ropic AVO synthetics, inplace of 1-D synthetics,
offer severali n t e r p retational benefits. First, theAVO
synthetics correlate to the CDPgathers, especially if the thick
cleansands and gas sands are matched atthe far traces. This is
evident by com-paring the AVO synthetic to the fieldCDP gather in
Figure 8. Also, whenthe acoustic impedances of the sandand shale
are almost equal, smallerrors for the sonic and density val-ues in
the uninvaded zones have sig-nificant impact on the 1-D
synthetic.Not only can the magnitude of thereflection be several
times off, butalso the polarity of the reflection canbe reversed,
thus making it difficultto match the 1-D synthetic to thefield
data. However, the Poissonsratio assigned to a depth interval iss t
rongly dependent on the more
robust estimate of lithology from thegamma and SP logs. Thus, on
theAVO synthetic, the far-offset tracesare more likely to show the
correctcorrelation with the field data thanthe 1-D normal-incidence
synthetic.As an experiment, try tying an SP l o gdisplayed in time
to the far- o ff s e ttraces in a CDP gather (far- o ff s e ttraces
offset > 1.5 depth).
Field data. The study area was con-fined to one off s h o re
block (ninemi2). A typical three-mile line acrossthe block is
displayed in the upperportion of Figure 9. This conven-tional 3-D
migrated section includesreflections with incidence angles of0-26.
Two control wells are on thisline. Because of the small NI
com-pared to the PR (as predicted fromthe well-log crossplot), this
sectionexhibits a mixture of both NI and PRevents, accounting for
some of thewormy events. Several eventshave amplitudes that are
larger thanthe background reflectors. However,none of these bright
events are asso-ciated with the two known gas zonesin well B.
The middle section of Figure 9 isan AVO inversion for PR based
onthe Verm-Hilterman equation. Theoverplotted red and yellow
eventsa re associated with those amplitudescalibrated to be
gas-sand re f l e c t i o n s .Near well B, this PR section has al
a rge reflection for the upper gas
EDGENET HTTP://WWW.EDGE-ONLINE.ORG FEBRUARY 1997
Figure 10. Conventional migrated section (upper) and estimated
PR sectionby large-angle stack. Red-yellow events are well-log
calibrated as gas-sandreflections.
February 1998 THE LEADING EDGE
-
zone around 2.55 s, but evidence ofthe deeper gas zone near 2.98
s ismissing. This is labeled as MI forMissing Indicator. Another MI
occursnear well D. What is also discourag-ing about this PR section
is that abright reflection occurs at 2.66 s atwell B. This
reflection appears in asand-poor interval just below theonset of
geopre s s u re. This is a FalseIndicator (FI) of hydro c a r b o n
s .
The existence of missing andfalse hydrocarbon indicators fro
mthe AVO inversion of PR based onthe Verm-Hilterman equation
sug-gested that a more robust estimatorof PR was needed. This is
shown inthe lower portion of Figure 9. It is alarge-angle stack
that included inci-dence angles (26-55). From the cor-relation to
the many wells in thearea, the red and yellow events canbe tied to
known production. Nofalse indicators nor missing indica-tors
occurred at well locations. Thei m p rovement of the larg e - a n g
l estack section over the PR inversionsection can be attributed to
the factthat the inversion uses differences ofamplitudes to
estimate PR. Howev-er, the large-angle stack would nothave been a
good estimator of PR ifthe magnitude of NI and PR were thesame and
also if the large off s e tangles were not used. Rememberthat the
contribution of PR becomeslarger than NI after 45. In short, forh y
d rocarbon identification, the PRestimate from large-angle stacks
isnot always better than a PR inver-sion based on a mathematical
model.It depends on the magnitude rela-tionship of NI to PR in the
local area,and this should be evaluated beforeany interpretation of
seismic AV Oattributes is conducted.
The seismic line in Figure 10 isnear the one in Figure 9. The
upperportion contains the conventionalstack, while the lower
portion is thelarge-angle stack. Two control wellsare on this line,
and two are project-ed onto the line. The top of geopres-sure once
again occurs around 2.6 s.
No false indicators or missingindicators of gas zones are on
thelower section. What is interesting isthe high that appears
between wellsF and D at approximately 2.8 s. Orig-i n a l l y, this
was a prospect to bedrilled. However, subsequent inter-p retation
of the PR 3-D volumechanged this decision. The twoevents at 2.8 s
near wells F and D arenow interpreted to be two separatechannels
that cut into the flanks of anexisting topographic feature. The
crest of this feature does not appearto be sand prone.
The lower portion of Figure 10 isa good re p resentation of
lithologywhile the upper portion containsboth lithostratigraphic
and chronos-tratigraphic reflections. In fact, Fig-ure 1 is a
combination of the upperportion of Figure 10 with the red andyellow
events from the lower portionof Figure 10 superimposed.
F i g u re 11 shows time slices at 2.54s across both the
conventional 3-Dvolume and the large-angle stack 3-Dvolume. The
upper portion is fro mthe conventional 3-D volume whilethe lower is
the large-angle stack.Both time slices have similar pat-
terns, but only the large-angle stackp roperly depicts the known
limits ofthe major reservoir in the northeast-ern portion of the
study are a .
D i s c u s s i o n . When a wet sand is veryclean, its Poissons
ratio lies betweenthat of a gas sand and a slightly shalysand.
Also, clean sands tend to beblocky in appearance on the SP logand
thus do not suffer from a reduc-tion in PR amplitude caused by
atransitional shaliness. These obser-vations suggest that caution
shouldbe exercised when evaluating highamplitude PRs on the
lithologic sec-tion as they can be clean-sand reflec-tions.
Normally, however, clean wet
EDGENET HTTP://WWW.EDGE-ONLINE.ORG FEBRUARY 1997
Figure 11. Time slice at 2.54 s from conventional 3-D volume
(upper) andequivalent time slice from large-angle stack 3-D volume
(lower). Knownfield is depicted by red in northeastern part of
block (lower time slice).
February 1998 THE LEADING EDGE
-
sands can be differentiated from gassands by crossplotting the
NI versusthe PR volumes as described byVerm and Hilterman.
It was fortuitous that this are ahad numerous gas reservoirs
andalso had several clean blanket sands.At the same time, without
the avail-ability of the 3-D long-offset seis-mic data in a good
reflection area, the application of the anisotropic p rocessing
would not have beenattempted. The repeated high-ampli-tude
reflections throughout the blockfrom the very clean sands and
gassands allowed the anisotropic factorto be analyzed with NMO
corre c-tions. Other reflections were buriedin the noise on the
far-offset traces.
One obvious omission in thisarticle is the contribution ofa n i
s o t ropy to the reflection coeff i-cient equation. Using the
estimate ofthe function from the NMO analy-sis and using the
available sonic log,low frequency estimates of theThomsens
anisotropic parameters and are obtainable. These two timefunctions
are then adjusted using theshale volume curve to force thea n i s o
t ropy to reside in the shalezones and to allow the sand zones tobe
basically isotropic. This pro c e d u re
provides the additional parametersneeded to generate TI AVO
synthet-ics. However, no compensation hasbeen assigned to the
inversion algo-rithms. This is a project for
futureinvestigation.
C o n c l u s i o n s . When the acousticimpedance of sands and
shales arenearly equal, the Poissons ratiocurve is similar to the
SP curve. Inthis situation, the amplitude on thelarge-angle traces
is essentially thePoisson re f l e c t i v i t y, which is an
indi-cator of lithology. The conventionalstack, especially in Class
2 environ-ments, will be composed of chronos-tratigraphic and
lithostratigraphicreflections. Assuming shale to be thebounding
media, PR reflections fro mgas sands have the highest ampli-tudes,
very clean wet sands have thenext highest amplitude, and shalywet
sands are weaker. By calibratingthe color scale to highlight the
PRamplitudes, a section results that def-initely fits the tale of
the local ranch-er it tells you exactly whatsdown there and where
to drill.
H o w e v e r, a geophysicist willstill be needed to make sure
therancher is in the right class of AV Oa n o m a l i e s .
Suggestions for further reading.There have been numerous
publica-tions in GEOPHYSICS and TLE fromthe Center for Wave
Phenomena,Colorado School of Mines coveringa n i s o t ro p y. Ilya
Tsvankins article, P -wave signatures and notation for trans
-versely isotropic media: An overviewcontains the basic principles
and ref-e rences necessary for anisotro p i cseismic processing.
The connectionof seismic to the rock properties canbe found in the
excellent review byCastagna et al. Rock physics The linkbetween
rock properties and AV Oresponse in Offset-Dependent Reflectiv -ity
Theory and practice of AVO analy -sis, SEG IG No. 8.
Acknowledgments: We thank Fairfield Indus -tries for its
cooperation and permission toshow the seismic from its
shallow-water 3-Dsurvey. At GDC, Mark Wilson provided thep e t
rophysical analysis; Connie Van Schuyver,the seismic data
processing; Martin Wood,the software code, and, Jim DiSiena, sugges
-tions to the manuscript. Finally, GDC appre -ciates the
opportunity to apply the excellentanisotropic work published by
Thomsen andTsvankin.
C o r responding author: Fred Hilterman,[email protected]
EDGENET HTTP://WWW.EDGE-ONLINE.ORG FEBRUARY 1997
LE
February 1998 THE LEADING EDGE